Iter #50: [tensor([-0.0989,  0.0341,  0.0069, -0.0431, -0.0550, -0.0378, -0.0135, -0.0614,
        -0.0294, -0.0286, -0.0189,  0.0371,  0.0293,  0.0330, -0.0317, -0.0017,
        -0.0073,  0.0287, -0.0024, -0.0249,  0.0253, -0.0304, -0.0035,  0.0596,
        -0.1168,  0.1408,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0159, -0.0357,  0.0381,  0.0048, -0.0281, -0.0411, -0.0451,  0.0094,
         0.1977, -0.0251,  0.0072, -0.0471,  0.0148,  0.0096, -0.0525,  0.0379,
        -0.0400, -0.0646, -0.0221,  0.0363,  0.2268,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1410, -0.0370,  0.0394,  0.0167, -0.0005,  0.0161,  0.0068, -0.0282,
         0.0169,  0.0137, -0.0048,  0.0137,  0.0080,  0.0043,  0.0003,  0.0203,
        -0.0358, -0.0113,  0.0567,  0.0706,  0.0404,  0.0031,  0.0231,  0.0080,
        -0.1253,  0.0509,  0.0493, -0.0064,  0.0128, -0.0038,  0.0404,  0.0185,
         0.0012, -0.0489, -0.0149,  0.0110,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0306, -0.3064, -0.0853, -0.0151, -0.0339, -0.0873,  0.0478, -0.0492,
        -0.0157, -0.0057,  0.0078, -0.0372, -0.0497, -0.0521,  0.0032,  0.0192,
        -0.0422,  0.0482,  0.0247,  0.0387,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0649, -0.0866,  0.0169, -0.0133,  0.0492,  0.0361, -0.0401, -0.0103,
        -0.0165, -0.0185, -0.0369,  0.0599, -0.0243, -0.0393, -0.0263,  0.0219,
        -0.0461, -0.0742, -0.0945,  0.0708,  0.1534,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0042, -0.2127,  0.1197, -0.1051, -0.0727,  0.0897,  0.0497,  0.0025,
         0.0448, -0.0085,  0.0584, -0.0422, -0.1900,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0905, -0.0681, -0.0508,  0.0690, -0.0446, -0.0534, -0.0166,  0.0084,
        -0.0002,  0.0310,  0.0186, -0.0078,  0.0211, -0.0162,  0.0018, -0.0234,
        -0.0007, -0.0212,  0.0055,  0.0232, -0.0275, -0.0252, -0.0609, -0.0473,
         0.0027, -0.0044,  0.0412, -0.0056,  0.0020, -0.0522, -0.0515, -0.0068,
         0.0071,  0.0108, -0.0118, -0.0199, -0.0508], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1038, -0.0181,  0.0625,  0.0528,  0.1179, -0.0320,  0.0247,  0.1442,
         0.0412,  0.0683, -0.0112,  0.0207, -0.1087,  0.0070, -0.0122,  0.0116,
        -0.0067,  0.0163,  0.0364, -0.0097, -0.0342, -0.0145,  0.0215,  0.0238,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1074,  0.0019,  0.0205,  0.0555, -0.0597, -0.0573,  0.0121,  0.0243,
        -0.0098, -0.0613,  0.0116,  0.0134,  0.0227,  0.0083, -0.0121,  0.0224,
         0.0338, -0.0366, -0.0051,  0.0234, -0.0563, -0.0441, -0.0055,  0.0092,
        -0.0066,  0.0073, -0.0079,  0.0124, -0.0144,  0.0205, -0.0314,  0.1855,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0337,  0.0118, -0.0349,  0.0057,  0.0227,  0.0223,  0.0380, -0.0254,
        -0.0149,  0.0809, -0.0010,  0.0131, -0.0179,  0.0180,  0.0219,  0.0205,
        -0.0075,  0.0087, -0.0590, -0.0444,  0.0721, -0.0963,  0.0206,  0.0436,
        -0.0340, -0.0240, -0.0113,  0.0531, -0.0057,  0.1182, -0.0186,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1010,  0.0036,  0.1119,  0.0603,  0.1540,  0.0400,  0.0214, -0.0529,
         0.0109,  0.0003, -0.0673,  0.0242,  0.0194,  0.0299,  0.0061, -0.0064,
         0.0010,  0.0357,  0.0180,  0.0234,  0.0157, -0.1144,  0.0823,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0361, -0.0081, -0.0369,  0.0138, -0.0056,  0.0114, -0.0139, -0.0108,
        -0.0103,  0.0348, -0.0229,  0.0663,  0.0156, -0.0908, -0.0731, -0.0060,
        -0.0487, -0.0070, -0.0073,  0.0076,  0.0117, -0.0535,  0.0242,  0.0382,
         0.0433,  0.0512, -0.1688, -0.0820,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 0.1088, -0.0449, -0.0619, -0.0160,  0.0165, -0.0181,  0.0155, -0.0081,
        -0.0385, -0.0274,  0.0247, -0.0199, -0.0305, -0.0118, -0.0026,  0.0269,
         0.0172, -0.0096,  0.0181, -0.0427, -0.0456, -0.0192,  0.0395,  0.0141,
        -0.0147, -0.0204,  0.0555, -0.1679,  0.0635,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0652, -0.0205, -0.0084,  0.0115,  0.0567,  0.0035,  0.0136,  0.0110,
         0.0625, -0.0709,  0.0082, -0.0078, -0.0186,  0.0288,  0.0885,  0.0286,
         0.0059, -0.0159, -0.0301, -0.0757, -0.1044,  0.0551,  0.0129, -0.1958,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.7202e-02,  8.9773e-02,  3.1152e-02,  5.1548e-02, -3.5998e-05,
         3.4362e-02,  1.9750e-03,  3.1022e-02,  3.3119e-03,  5.2381e-02,
         1.7956e-02, -2.7877e-02,  3.4300e-02,  1.3181e-02,  3.7969e-02,
         1.3804e-04, -1.1840e-02, -7.1941e-03,  3.8793e-02, -5.3625e-02,
        -6.2955e-02, -1.3674e-02, -7.9053e-03, -7.0822e-03, -3.2841e-03,
         1.3989e-03,  2.0483e-03, -2.6840e-02,  2.3764e-02, -5.8154e-02,
         2.1735e-02, -2.7649e-02, -5.2395e-02, -6.5481e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0448,  0.0275, -0.1992, -0.0347, -0.0380, -0.0692, -0.0779, -0.0387,
        -0.0509, -0.0124, -0.0237, -0.0061, -0.0143, -0.0285, -0.0264, -0.0653,
        -0.0722, -0.0129,  0.0321, -0.0341,  0.0295, -0.0613,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2161, -0.0699,  0.0634,  0.0486,  0.0173,  0.0351, -0.0047, -0.1104,
         0.0171, -0.0102, -0.0097, -0.0295, -0.0187,  0.0462,  0.0335,  0.0160,
        -0.1245,  0.1289,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1371,  0.0136,  0.0437,  0.0308, -0.0331,  0.0108, -0.1389,  0.0250,
        -0.0320, -0.0272, -0.0162,  0.0031,  0.0021, -0.0280, -0.0235, -0.1167,
         0.0404,  0.0061, -0.0660, -0.0083, -0.0105,  0.1277, -0.0593,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0470,  0.0376, -0.1780, -0.0874, -0.0158,  0.0914,  0.1133,  0.0636,
         0.0568,  0.0816,  0.0220,  0.0210, -0.0108, -0.1091,  0.0309, -0.0082,
         0.0257,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0263,  0.0859, -0.0317,  0.0040, -0.0356,  0.1450,  0.1385,  0.1253,
        -0.0614,  0.0571,  0.0252,  0.0100,  0.0587,  0.0324, -0.0454, -0.0226,
         0.0350, -0.0600,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1779, -0.0140,  0.0674,  0.0078, -0.0440, -0.0078, -0.0206, -0.0628,
        -0.0288,  0.0003, -0.0171, -0.0054,  0.0101, -0.1088, -0.0092,  0.0097,
        -0.0059,  0.0473,  0.0498, -0.0021,  0.0056, -0.0197,  0.0034, -0.0191,
        -0.1396,  0.1159,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1052,  0.0467,  0.0252,  0.0072,  0.0723,  0.2314,  0.0037,  0.0079,
        -0.0292,  0.0111,  0.0170,  0.0109,  0.0116, -0.0191,  0.0159, -0.0070,
        -0.0612, -0.0332,  0.0044, -0.0067, -0.0098,  0.0457,  0.0081,  0.0077,
        -0.0162, -0.0509,  0.0166, -0.0136,  0.0704,  0.0134, -0.0204,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0004, -0.0859, -0.0191,  0.0273, -0.0277, -0.0400, -0.0225, -0.0347,
        -0.0683, -0.1175, -0.0156, -0.0291, -0.0499,  0.0244,  0.1008,  0.0951,
        -0.1458, -0.0959,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1801, -0.0831,  0.2620, -0.1169, -0.1474, -0.0121,  0.1747,  0.0237,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 0.0003,  0.0354,  0.0420, -0.0481, -0.0656, -0.0005, -0.0381,  0.0175,
         0.0340, -0.0094, -0.0030, -0.0780, -0.0038,  0.0251, -0.0002, -0.0182,
        -0.0060,  0.0438, -0.0014, -0.0379, -0.0017, -0.0153,  0.0355, -0.0056,
         0.0013, -0.0185,  0.0342,  0.0297,  0.0625, -0.0248,  0.0345, -0.0347,
        -0.0196, -0.0223,  0.0121, -0.0605, -0.0144,  0.0381,  0.0266,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0752,  0.0179,  0.0442,  0.0242, -0.0242, -0.0110, -0.0232, -0.0022,
        -0.0014, -0.0089, -0.0020, -0.1127, -0.0187, -0.0103,  0.0229,  0.0039,
         0.0575,  0.0039,  0.0167, -0.0781,  0.1264,  0.0075, -0.0192,  0.0093,
        -0.0573, -0.0401,  0.1507,  0.0306,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1197,  0.1623,  0.0955,  0.0343, -0.0108,  0.0482,  0.0857, -0.0692,
         0.0102,  0.1146, -0.1395, -0.1097,  0.0004,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0580, -0.1667,  0.0096, -0.0947, -0.0312, -0.0352, -0.0801,  0.0754,
        -0.0078, -0.0523, -0.0104,  0.0572,  0.0387,  0.1387, -0.1441,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1564e-02,  1.4234e-02,  2.3524e-02,  5.1035e-02, -6.0982e-03,
         2.8563e-02,  1.0973e-02, -1.7008e-02,  4.7869e-03, -1.9677e-02,
        -9.9109e-03,  4.1112e-02,  5.1284e-03, -6.1796e-03,  3.1650e-02,
         8.9848e-03, -2.7209e-02,  4.7222e-02,  7.0818e-03,  3.2911e-02,
         3.9912e-03,  1.7790e-02, -1.4911e-03,  3.3963e-02,  4.4437e-02,
        -4.3057e-02,  1.4760e-02,  1.3769e-02,  2.1222e-02,  2.5007e-02,
         2.2802e-02, -1.8876e-02,  2.4563e-02, -2.8646e-03,  2.4988e-03,
         1.2872e-02, -4.8132e-02, -5.6258e-06,  6.2342e-03, -2.2049e-02,
         1.3738e-02,  2.7968e-02, -1.3065e-02, -2.5935e-03, -3.6223e-02,
        -6.7115e-02, -1.4060e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0819, -0.0829, -0.0080,  0.0578, -0.0133, -0.0450, -0.0101,  0.0030,
         0.0537, -0.0223,  0.0777,  0.0686,  0.0071, -0.0100,  0.0008, -0.0108,
        -0.0113,  0.0173, -0.0360,  0.0031,  0.0100,  0.0039, -0.0045,  0.0265,
        -0.0387, -0.0132, -0.0331, -0.0033, -0.0253,  0.0108, -0.0185,  0.0028,
         0.0032, -0.0095,  0.0147,  0.0065, -0.0143, -0.0233, -0.0015,  0.0112,
        -0.0128, -0.0914,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0273, -0.0416, -0.0723,  0.0075, -0.0198,  0.0695,  0.1336,  0.0444,
        -0.0229,  0.0439,  0.1161, -0.0317, -0.0017, -0.0067,  0.0348, -0.0425,
        -0.0380, -0.0289, -0.0199, -0.0391, -0.0230,  0.0035,  0.0271,  0.0711,
         0.0330,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0093,  0.0006,  0.0122,  0.0296,  0.0430,  0.0417,  0.0677,  0.0203,
        -0.0006,  0.0040,  0.0091, -0.0357,  0.0158, -0.0408,  0.0149,  0.0136,
         0.0128,  0.0313,  0.0838,  0.1368, -0.0983, -0.1221,  0.1560,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1664,  0.0880,  0.0411, -0.0059,  0.0164,  0.0548,  0.0006, -0.0133,
        -0.0120, -0.0245, -0.1339, -0.0447, -0.0137, -0.0103,  0.0155, -0.0073,
        -0.0841, -0.0090,  0.0090, -0.0257, -0.0634, -0.1310,  0.0296,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0783,  0.1654, -0.0390, -0.0241,  0.0637,  0.0205,  0.0652,  0.1549,
        -0.0400,  0.1050,  0.0155,  0.1912,  0.0372,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0104e-01,  1.2245e-01,  2.0754e-02, -1.4800e-02,  2.7283e-04,
         2.2343e-02,  3.7470e-02, -8.2863e-02, -4.2531e-02, -9.2955e-02,
         1.6285e-02,  1.7534e-01, -2.6041e-02, -4.4858e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0722,  0.0623, -0.0457,  0.0833,  0.0736,  0.0625,  0.0216,  0.0708,
        -0.0302,  0.0387,  0.0060,  0.0485,  0.0147,  0.0299,  0.0230, -0.0980,
         0.1532, -0.0113, -0.0546,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 0.1221,  0.0023, -0.0187, -0.0402, -0.0228, -0.0409,  0.0403,  0.0342,
         0.0108,  0.0548,  0.0407,  0.0481,  0.0124,  0.0128, -0.0037,  0.0614,
         0.0579,  0.0215, -0.0360,  0.0064, -0.0286,  0.0034, -0.0294, -0.0297,
        -0.1562,  0.0647,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1506,  0.0886, -0.0107, -0.0325, -0.0105, -0.0422, -0.0638,  0.0577,
         0.0027,  0.0526,  0.0304, -0.0400,  0.0864,  0.0568, -0.1546, -0.1199,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1641, -0.0265,  0.0124, -0.0841, -0.0447, -0.1075, -0.0771,  0.0274,
        -0.0455,  0.0356,  0.0038,  0.1175, -0.0186,  0.0239,  0.0296, -0.0315,
        -0.1502,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.0112e-02, -1.2331e-02, -3.0920e-01,  2.6740e-04,  2.7595e-02,
         1.6359e-02,  1.1185e-02,  3.0020e-02, -5.0293e-01,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0641, -0.0081,  0.0069,  0.0196,  0.0254, -0.0644,  0.0235, -0.0127,
        -0.0195,  0.0099,  0.0079, -0.0111,  0.0317,  0.0598,  0.0146, -0.0098,
         0.0037, -0.0006, -0.0111, -0.0335, -0.0097,  0.0144,  0.0222,  0.0028,
         0.0041,  0.0046, -0.0713,  0.0524, -0.0058,  0.0072,  0.0066,  0.0259,
         0.0189,  0.0168, -0.0060,  0.0241, -0.0130, -0.0266,  0.0271,  0.0026,
         0.0020,  0.0361, -0.0152,  0.0466,  0.0305, -0.0697], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1231, -0.0433, -0.0441, -0.0495,  0.0026, -0.0097,  0.0100,  0.0340,
         0.0153, -0.0258, -0.0442, -0.0509, -0.1041, -0.0250,  0.0095, -0.0306,
        -0.0120,  0.0074,  0.0181,  0.0193,  0.0419,  0.0134, -0.0182,  0.0008,
        -0.0054,  0.0252,  0.0176, -0.0166, -0.0869, -0.0951,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.2457,  0.0817,  0.1613,  0.3186,  0.1223,  0.0130, -0.0574,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1511,  0.0081, -0.0949,  0.0848,  0.0817, -0.2455,  0.0073, -0.0060,
         0.1053, -0.0934, -0.1219,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0963, -0.0188, -0.0640, -0.0747, -0.0153,  0.0419,  0.1333, -0.0151,
         0.2078,  0.0508,  0.0028,  0.0433,  0.1575,  0.0784,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0172,  0.0068,  0.0353,  0.0716,  0.0080, -0.0172, -0.0007, -0.0396,
         0.0184, -0.0314, -0.0227,  0.0617,  0.0179, -0.0185, -0.0003, -0.0334,
         0.0039, -0.0622,  0.0217, -0.0023, -0.0031, -0.0215,  0.0507, -0.0188,
         0.0295,  0.0441, -0.0514, -0.0432, -0.0867, -0.0496,  0.0046, -0.0099,
        -0.0625,  0.0336,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0827,  0.0841,  0.0541,  0.0435, -0.0059, -0.0171,  0.0557, -0.0558,
        -0.0025, -0.0002,  0.0143,  0.0045,  0.0001, -0.0029,  0.0113,  0.0185,
        -0.0931,  0.0143, -0.0148, -0.0040,  0.1246,  0.0091,  0.0451,  0.0048,
        -0.0596, -0.0152,  0.1268, -0.0354,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0633,  0.0070, -0.0208, -0.0309, -0.0107,  0.0356,  0.0146, -0.0149,
        -0.0041, -0.0429, -0.1008, -0.0337, -0.0280, -0.0190,  0.1238, -0.0490,
         0.0470, -0.0089, -0.0956, -0.0355, -0.0633, -0.0235,  0.1000, -0.0271,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([-0.1050,  0.1625,  0.0791,  0.0138,  0.0104, -0.0120,  0.0141, -0.0261,
        -0.0279, -0.0047,  0.0031, -0.0272,  0.0220, -0.0415,  0.0016, -0.0005,
         0.0307,  0.0039,  0.0262, -0.0300, -0.0011,  0.0218, -0.0026,  0.0085,
        -0.0033,  0.0251,  0.0044, -0.0058,  0.0419, -0.2009, -0.0423,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0136,  0.0496,  0.0632,  0.0433,  0.0262,  0.0776, -0.0392, -0.0281,
        -0.3141, -0.0790,  0.1051,  0.1610,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0086,  0.1357, -0.0884,  0.0061,  0.0280,  0.0342,  0.0553,  0.0619,
         0.0198, -0.0996, -0.0362,  0.0041, -0.0258,  0.3150, -0.0814,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0350, -0.0590,  0.0021, -0.0392, -0.1074,  0.0852, -0.0698,  0.0045,
        -0.0085,  0.0362,  0.0274,  0.0167,  0.0041,  0.0201, -0.0249,  0.0146,
        -0.0130,  0.0142, -0.0258, -0.0478,  0.0176,  0.0051,  0.0230,  0.0057,
        -0.0215,  0.0055, -0.0088, -0.0038, -0.0021,  0.0002, -0.0222,  0.0290,
         0.2000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2857, -0.0003, -0.1324,  0.0195, -0.0269,  0.0187,  0.0178,  0.0145,
         0.0651,  0.0339, -0.0334,  0.0219,  0.0302,  0.0112,  0.1760,  0.1125,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2180, -0.0466, -0.2400, -0.2387, -0.0597,  0.0314, -0.0447, -0.1191,
        -0.0018,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.3093,  0.1887, -0.0377,  0.3268, -0.1375,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0752,  0.0112, -0.0285, -0.0284, -0.0558, -0.0188, -0.0248,  0.0088,
         0.0414, -0.0805, -0.0910, -0.0348,  0.0803, -0.0707, -0.0019, -0.0317,
         0.0057, -0.0510, -0.0117, -0.0168,  0.0132, -0.0276, -0.0114, -0.0228,
         0.0133,  0.1066,  0.0359,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2675, -0.1567, -0.1138, -0.1865, -0.1788, -0.0780,  0.0187,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0330, -0.2139,  0.0866, -0.0052, -0.0234,  0.0119, -0.0532, -0.0354,
        -0.0307,  0.0006, -0.0185,  0.0082, -0.0432,  0.0601,  0.0304, -0.0281,
        -0.0091, -0.0044, -0.0547, -0.1191, -0.0670, -0.0634,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0501,  0.0442,  0.0404,  0.0413, -0.1369,  0.0223,  0.0485, -0.0809,
         0.0275, -0.0598,  0.0549, -0.0344, -0.0171, -0.0738, -0.0059,  0.1265,
         0.0487,  0.0057,  0.0810,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1059, -0.1065, -0.0248, -0.0407, -0.0259, -0.0876,  0.0462, -0.0335,
        -0.0265,  0.0124, -0.0117,  0.0055, -0.0341, -0.0298, -0.0012, -0.0274,
        -0.0003, -0.0066, -0.0042,  0.0647,  0.0071,  0.0354,  0.0054, -0.0080,
        -0.0152,  0.0245, -0.0025, -0.0226, -0.0720,  0.1113,  0.0005,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 0.0840, -0.0100,  0.0143,  0.0291,  0.0176,  0.0147,  0.0015,  0.0286,
        -0.0338,  0.0145,  0.0150, -0.0014,  0.0201, -0.0005, -0.0426, -0.0275,
         0.1528,  0.0041, -0.0635, -0.0046, -0.0145,  0.0136, -0.0295, -0.0150,
        -0.0834, -0.0050, -0.0637, -0.0355, -0.0006,  0.0131,  0.1202, -0.0256,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1661, -0.0936,  0.0243, -0.0168,  0.0199,  0.0322, -0.0227, -0.0165,
         0.0856, -0.0167, -0.0580,  0.0344,  0.0016, -0.1628, -0.0360,  0.0086,
        -0.0020, -0.0076, -0.0218, -0.0401,  0.1177, -0.0151,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0414,  0.0072, -0.0137,  0.0321, -0.0280, -0.0241, -0.0409, -0.0255,
        -0.0371,  0.0430,  0.0143,  0.0078,  0.0089, -0.0234,  0.0403, -0.0012,
        -0.0347, -0.0046,  0.0105, -0.1530, -0.0110,  0.0749,  0.0132, -0.0049,
         0.0160, -0.1120, -0.0382, -0.0432,  0.0204,  0.0743,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1696, -0.0301, -0.0703,  0.0550, -0.0149, -0.0167, -0.0219,  0.0035,
        -0.0350,  0.0082, -0.0207, -0.0329, -0.0322,  0.0482,  0.0375, -0.1118,
        -0.1965, -0.0950,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0299,  0.0070, -0.0125, -0.0092, -0.0208, -0.0181, -0.0269, -0.0120,
        -0.0513,  0.0143, -0.0468,  0.0360, -0.0346, -0.0054, -0.0533, -0.0153,
        -0.0157, -0.0458, -0.0098, -0.0623,  0.0221,  0.0346, -0.0359, -0.0035,
        -0.0184,  0.0097, -0.0148, -0.1084,  0.0315,  0.0116, -0.0200, -0.0316,
        -0.0078, -0.0121, -0.0135, -0.0242, -0.0732,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0021,  0.0126,  0.0022,  0.0501, -0.0314, -0.0340,  0.0744,  0.0122,
         0.0006,  0.0007, -0.0130, -0.0414,  0.0683, -0.0089,  0.0560,  0.0092,
         0.2436, -0.0472, -0.0528, -0.0074,  0.1212,  0.0584, -0.0407, -0.0116,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0690,  0.0119, -0.0124,  0.0077, -0.0111,  0.0092, -0.0456, -0.0031,
        -0.1622, -0.0088, -0.0145,  0.0181, -0.0417,  0.0007,  0.0199,  0.0082,
         0.0016,  0.0235,  0.0112,  0.0074, -0.0011,  0.0029, -0.0061, -0.0231,
        -0.0572, -0.0127, -0.0127,  0.0016, -0.0143,  0.0367, -0.0012, -0.0484,
         0.0085, -0.0118, -0.0179, -0.0103, -0.0164, -0.0192,  0.0123, -0.0142,
        -0.0137,  0.0067,  0.0097,  0.0035,  0.0338, -0.1164,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0532,  0.0483,  0.0595,  0.0029, -0.0542, -0.0426,  0.0347, -0.0077,
         0.0048, -0.1055,  0.0650,  0.0097, -0.1432, -0.0154, -0.0208,  0.0031,
        -0.0783, -0.0384,  0.0019,  0.0119,  0.0174,  0.1109,  0.0705,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0359,  0.1103, -0.1140, -0.0534,  0.0155,  0.0868, -0.1233, -0.1075,
        -0.0390,  0.2596,  0.0547,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1365, -0.0242, -0.0280,  0.0482,  0.0776,  0.0994,  0.1213, -0.0069,
         0.0097, -0.0655, -0.1504,  0.0334,  0.1833,  0.0157,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1495,  0.2263,  0.1158,  0.0290, -0.0167,  0.0298,  0.0725,  0.0377,
         0.0502,  0.0017, -0.1070,  0.1637,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0822, -0.0038,  0.0257,  0.0433,  0.0375, -0.0044,  0.0091, -0.0214,
        -0.0024, -0.0374, -0.0033, -0.0650,  0.0066, -0.0052, -0.0133,  0.0099,
         0.0023,  0.0358,  0.0023,  0.0189,  0.0053,  0.0107, -0.0143,  0.0167,
        -0.0382,  0.0202,  0.0139, -0.0142, -0.0354,  0.0301, -0.0347,  0.0174,
         0.0120, -0.0112,  0.0069, -0.0171,  0.0067, -0.0119,  0.0090,  0.0103,
        -0.0015, -0.0010,  0.0020,  0.0029,  0.0046, -0.0175,  0.0450,  0.0127,
         0.0072,  0.0017,  0.0321,  0.0247,  0.0156,  0.0067, -0.0585],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 0.2707, -0.0089,  0.0464, -0.0011,  0.0018,  0.0499, -0.0359,  0.0269,
         0.0578, -0.0076, -0.0613,  0.0052, -0.0357,  0.0434, -0.0825, -0.0510,
         0.0158,  0.0153,  0.0298, -0.0069,  0.0461,  0.0883, -0.0116,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1000,  0.0447,  0.0247, -0.1689, -0.0429,  0.0205, -0.0333, -0.0433,
        -0.0448, -0.0021, -0.0303, -0.1209, -0.0040, -0.0497,  0.0436, -0.0582,
        -0.0705, -0.0977,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.2473, -0.0115,  0.0350, -0.0199,  0.0461,  0.0138, -0.1252,  0.0313,
         0.0191,  0.0033,  0.0300,  0.0031, -0.1297,  0.0185,  0.0449,  0.0620,
        -0.0437, -0.1158,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0865, -0.0295, -0.0512, -0.0784,  0.0621,  0.0877,  0.0799, -0.0572,
        -0.0420,  0.0761,  0.0185, -0.1561,  0.0346, -0.0397, -0.1004,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6733e-01, -2.3474e-02,  5.1809e-02, -2.7581e-02,  1.2007e-02,
         1.0134e-02, -1.9693e-02,  5.8825e-03,  5.7121e-02, -8.1636e-03,
         7.5467e-03,  4.0208e-02,  1.4423e-03,  1.6252e-02, -3.3467e-03,
        -4.8374e-02, -2.1421e-02,  2.5132e-02, -5.1454e-05,  2.1415e-02,
        -3.4032e-03, -2.0709e-02,  1.8805e-02,  1.4463e-02,  1.8338e-02,
         6.0239e-03, -7.5138e-02,  2.1180e-02, -2.5644e-02,  7.6742e-02,
         5.5484e-02, -9.5684e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1294,  0.0128,  0.0071, -0.0133, -0.0295,  0.0074, -0.0197, -0.0103,
        -0.0266,  0.0274,  0.0413, -0.0163, -0.0225, -0.0047,  0.0153,  0.0077,
        -0.0148,  0.0607,  0.0183, -0.0009, -0.0240,  0.0812,  0.0494, -0.0009,
        -0.0463, -0.0275, -0.0177, -0.0424, -0.0339, -0.0244, -0.0169, -0.0627,
         0.0871,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0688,  0.0091, -0.0034, -0.0125, -0.0057, -0.0170,  0.0097, -0.0016,
         0.0236,  0.0366,  0.0075,  0.0057,  0.0336, -0.0043,  0.0702, -0.0087,
        -0.0019,  0.0207,  0.0137,  0.0096,  0.0109,  0.0029,  0.0276,  0.0109,
        -0.0258,  0.0048,  0.0169, -0.0038,  0.0161,  0.0239,  0.0191,  0.0142,
         0.0120, -0.0051, -0.0185,  0.0352, -0.0264, -0.0124, -0.0075, -0.0004,
        -0.0029, -0.0106, -0.0196,  0.0082, -0.0002,  0.0095,  0.0050, -0.0687,
         0.0053, -0.0195, -0.0697, -0.0066,  0.0077, -0.0040,  0.0259,  0.0062,
         0.0559, -0.0159], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0170,  0.0488,  0.0147, -0.0237,  0.0010,  0.0301,  0.0363, -0.0212,
        -0.0066,  0.0062, -0.0055, -0.0540,  0.1139,  0.0068,  0.0167, -0.0681,
         0.0174, -0.0120, -0.0139, -0.0109, -0.0914, -0.0989, -0.0227,  0.0510,
        -0.0503, -0.1610,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0498,  0.1646, -0.0105,  0.0749,  0.0151,  0.0376, -0.0037, -0.0308,
        -0.0354,  0.0887, -0.0058, -0.0095,  0.0118,  0.0154, -0.0259, -0.0238,
        -0.0603,  0.0021,  0.0153, -0.0089,  0.0170, -0.0472, -0.0150, -0.1973,
        -0.0335,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3704e-02, -1.1294e-01, -3.8809e-02,  5.7992e-03,  9.2718e-03,
         3.3642e-02, -7.3334e-02,  4.5667e-02,  9.5977e-03, -4.7179e-02,
        -4.1042e-02,  6.1466e-03, -3.3680e-02,  2.2570e-03,  3.6951e-03,
        -7.4515e-02, -8.8198e-03,  7.5428e-03,  4.9978e-03,  6.2023e-03,
         4.9567e-03, -1.1223e-03,  1.5376e-02,  2.8003e-03, -3.4838e-02,
        -1.6422e-03,  1.6026e-02,  9.7636e-03,  1.0586e-02, -2.3909e-03,
         1.5252e-02,  1.9348e-02, -2.4871e-02,  6.4683e-05, -2.7032e-02,
         1.2548e-02, -4.8372e-03, -3.3089e-03,  7.2814e-03,  2.0853e-02,
        -1.0850e-02,  2.7669e-02,  3.2224e-02, -3.0312e-02,  8.5203e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0043e-01, -8.9186e-03,  1.7326e-02,  7.1899e-02, -3.3484e-03,
         4.0972e-02, -3.5131e-02, -8.1978e-02,  2.4797e-02,  8.8989e-03,
        -1.6370e-02,  1.4095e-02, -9.1229e-02,  8.8891e-05, -2.1792e-03,
         5.2793e-02, -2.1590e-02, -9.6557e-03,  1.5456e-02, -2.3186e-02,
         3.6946e-03,  1.6754e-02,  1.3920e-01,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1677, -0.3852,  0.0318, -0.0972,  0.0019, -0.0530, -0.0007,  0.0114,
         0.0309, -0.0755, -0.0543,  0.0905,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 0.1111, -0.0154, -0.0287,  0.0023, -0.0237, -0.0073, -0.0058, -0.0477,
         0.0690, -0.0014,  0.1282,  0.0611,  0.0288,  0.0368,  0.0772,  0.0540,
        -0.0321, -0.0733, -0.0809,  0.0221, -0.0240, -0.0131,  0.0559,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0277,  0.1045,  0.0401,  0.0091,  0.0659, -0.0204, -0.0128,  0.0060,
        -0.0130, -0.0180, -0.0120,  0.0061,  0.0114, -0.0121,  0.0005, -0.0031,
         0.0004, -0.0509, -0.0238, -0.0312,  0.0394, -0.0335, -0.0092,  0.0151,
         0.0177,  0.0394, -0.0359,  0.0364, -0.0118,  0.0598, -0.0311, -0.1146,
        -0.0871], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1386,  0.0751, -0.0968, -0.0140, -0.1687, -0.0460, -0.0387, -0.0481,
        -0.0246, -0.0032,  0.0129,  0.0529,  0.0268, -0.0192,  0.0112, -0.0392,
        -0.0313, -0.0982, -0.0542,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0344,  0.2201,  0.0581,  0.0423,  0.0649, -0.0057,  0.0683, -0.0846,
         0.0827, -0.0937, -0.0199, -0.0073,  0.0384, -0.1794,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0348, -0.2669, -0.0089,  0.0085,  0.0325, -0.0405, -0.0174,  0.0365,
        -0.0779,  0.0367, -0.0018, -0.0565, -0.0017, -0.0882, -0.0148, -0.0210,
        -0.0519, -0.0342,  0.0124,  0.0505, -0.0711,  0.0353,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0208, -0.0611,  0.0221,  0.0070,  0.0008, -0.0125,  0.0030,  0.0021,
        -0.0058,  0.0192, -0.0707,  0.1056,  0.0315, -0.0190,  0.0041, -0.0147,
         0.1234,  0.0066, -0.0378, -0.0068,  0.0208, -0.0246,  0.0833, -0.0408,
         0.0385,  0.0133, -0.0291, -0.0622,  0.0884,  0.0245,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1842, -0.1528,  0.0336, -0.0558,  0.0205, -0.0119, -0.0052,  0.0390,
         0.0497, -0.0723, -0.0783, -0.0039, -0.0643,  0.0387, -0.0449,  0.0195,
         0.1254,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0904, -0.0013,  0.0674,  0.0524, -0.0299,  0.0843,  0.0688,  0.0807,
         0.3106,  0.1366, -0.0775,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.3405,  0.1696,  0.0439, -0.0398,  0.0459, -0.0469,  0.0105, -0.0280,
        -0.0150,  0.0293,  0.0062, -0.0290, -0.0672,  0.0055, -0.0706,  0.0477,
         0.0045,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1574,  0.0741, -0.0395,  0.0248,  0.1241,  0.0255,  0.0063, -0.0323,
         0.0192,  0.0062,  0.0053,  0.1119, -0.0092,  0.0129,  0.0069,  0.0265,
         0.0359, -0.0517,  0.0590,  0.0154,  0.0208,  0.0649,  0.0216, -0.0055,
         0.0209,  0.0221,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1211,  0.0203,  0.0264, -0.0379,  0.0436,  0.0644, -0.0833, -0.0397,
         0.0798, -0.0274,  0.0131, -0.0273,  0.0270, -0.0292, -0.0178, -0.0345,
        -0.0010, -0.0150,  0.0204, -0.0697,  0.0274,  0.0137,  0.0148, -0.0013,
         0.0085, -0.0050, -0.0271,  0.0187, -0.0004,  0.0840,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.6264e-02, -1.4933e-01, -2.3448e-02, -3.5439e-02, -2.2524e-02,
        -2.5665e-02,  2.5147e-02, -4.3537e-02,  4.5937e-02,  1.9470e-03,
         5.7755e-03,  1.7963e-02,  5.8640e-03, -2.1782e-02, -4.3893e-02,
         3.9619e-02,  2.4846e-04, -8.4070e-02,  3.1155e-01,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 0.2433, -0.0615, -0.0551, -0.0230, -0.0305, -0.0137, -0.0220, -0.0187,
         0.0041,  0.0107,  0.0095,  0.0169, -0.0098,  0.0131, -0.0174, -0.0276,
         0.0144,  0.0214, -0.0078, -0.0332,  0.0199, -0.0229,  0.0015, -0.0340,
         0.0346, -0.0512,  0.1822,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0790, -0.0037, -0.0316, -0.2444, -0.0176, -0.0676, -0.0633,  0.0405,
         0.1471,  0.0285, -0.0913, -0.0740, -0.0552, -0.0561,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0276,  0.0989,  0.1161, -0.0059, -0.0645,  0.0132, -0.0015,  0.0582,
        -0.0327, -0.0153,  0.0704, -0.0403, -0.0109, -0.0325,  0.0340,  0.0010,
        -0.0123, -0.0225, -0.0738, -0.0131, -0.0004,  0.0119, -0.0110, -0.0682,
        -0.0990,  0.0650,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1541,  0.0713, -0.0479, -0.0374, -0.0631,  0.0137,  0.0828, -0.0893,
         0.0232,  0.1061,  0.0477, -0.1253,  0.0146, -0.1235,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0296,  0.0295,  0.0074, -0.0245,  0.0249,  0.0153, -0.0175, -0.0436,
        -0.0051, -0.0225, -0.0102,  0.0005,  0.0168, -0.0525,  0.0280,  0.0082,
        -0.0225, -0.0465,  0.0183, -0.0037, -0.0460,  0.0558, -0.0290, -0.0520,
        -0.1031, -0.0216,  0.0405, -0.0822, -0.0426, -0.0132, -0.0667,  0.0205,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.3691,  0.0835, -0.0412,  0.0089,  0.0091, -0.0832,  0.0063, -0.0197,
        -0.0027,  0.0466, -0.0062,  0.0023,  0.0306, -0.0110,  0.0195,  0.0006,
        -0.0317, -0.1040, -0.0137,  0.0026, -0.0067,  0.0186, -0.0164,  0.0129,
        -0.0530,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.3119,  0.1045,  0.1075,  0.0624, -0.1834,  0.2303,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0885,  0.1145, -0.0355, -0.0096,  0.0545,  0.0055, -0.0231, -0.0392,
        -0.0241, -0.0305, -0.0329,  0.0051, -0.0093,  0.0116,  0.0126,  0.0063,
         0.0626, -0.0433, -0.0023,  0.0074,  0.0013, -0.0250, -0.0377,  0.0101,
         0.0018,  0.0245, -0.0086, -0.0017, -0.0302,  0.0197,  0.0751,  0.0308,
        -0.0312,  0.0014, -0.0043, -0.0146, -0.0638], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0277,  0.0892,  0.0288,  0.0129,  0.0041,  0.0135, -0.0430, -0.0200,
         0.0005, -0.0017,  0.0211,  0.0139,  0.0134,  0.0542,  0.0135, -0.0308,
        -0.0549, -0.0237,  0.0177, -0.0144,  0.0059, -0.0273,  0.0082, -0.0122,
         0.0556, -0.0893, -0.0231,  0.0678, -0.0098, -0.0057, -0.0059,  0.0162,
         0.0152, -0.0904, -0.0684,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.2214,  0.1049, -0.0113, -0.0371, -0.1073, -0.0206,  0.0164, -0.0062,
        -0.0197,  0.0771,  0.0452,  0.0617, -0.0622, -0.1573, -0.0517,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0052, -0.1004, -0.0138, -0.0054,  0.0450, -0.0276, -0.0415,  0.1096,
         0.0532,  0.0050, -0.0301,  0.0138, -0.0215,  0.0033,  0.0259,  0.0043,
         0.0012, -0.0265, -0.0877, -0.0306, -0.0406, -0.0149, -0.0408,  0.0109,
        -0.0283, -0.0495, -0.0313, -0.0115, -0.0009, -0.0123, -0.0354,  0.0618,
        -0.0102,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1054,  0.1092,  0.0351,  0.0577,  0.0307, -0.0574, -0.0221,  0.0099,
        -0.0074,  0.0037,  0.0089,  0.0121,  0.0020, -0.0484, -0.0284,  0.0310,
         0.0164, -0.0720,  0.1002, -0.0229,  0.0202, -0.0014, -0.0182, -0.0011,
         0.0108, -0.0034, -0.0389, -0.1243,  0.0010,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #500: [tensor([-0.1799, -0.1854, -0.0623,  0.0053, -0.0117, -0.0393, -0.0870,  0.0315,
        -0.0505, -0.1538,  0.0393,  0.0744, -0.0074,  0.0472, -0.0251,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1257,  0.0324,  0.0180,  0.0120,  0.0110,  0.0364,  0.0296, -0.0148,
         0.0799,  0.0549,  0.0330, -0.0295,  0.0031, -0.0200,  0.0055, -0.0104,
        -0.0131,  0.0026, -0.0002, -0.0199, -0.0026, -0.0231, -0.0148,  0.0123,
        -0.0231,  0.0053,  0.0004, -0.0182, -0.0191, -0.0011,  0.0052,  0.0178,
        -0.0013, -0.0063,  0.0101,  0.0287,  0.0158, -0.0110, -0.0053, -0.0128,
         0.0078, -0.0123, -0.0325, -0.0186, -0.0049,  0.0124,  0.0218, -0.0634,
         0.0400], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0132, -0.0315, -0.0414,  0.0261,  0.0076, -0.0535, -0.0780,  0.0200,
         0.0351, -0.0585,  0.0092, -0.0333, -0.0544, -0.0132,  0.0108, -0.0381,
         0.0734,  0.0046, -0.0070,  0.0418,  0.0470,  0.0067,  0.0190, -0.0140,
        -0.0122, -0.0078,  0.0031, -0.0209, -0.0091,  0.0027, -0.0026,  0.0344,
         0.0068, -0.0842,  0.0788,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0234, -0.0137, -0.0509,  0.0370,  0.0061,  0.0238,  0.0521,  0.0334,
         0.0203,  0.0597,  0.0941,  0.0684,  0.0425,  0.0070, -0.1125,  0.0467,
        -0.0422, -0.2289,  0.0372,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0989,  0.2121,  0.0757,  0.0024, -0.0508, -0.0362,  0.1131, -0.0010,
         0.0795, -0.0199, -0.0254,  0.0454,  0.0544, -0.1684, -0.0168,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1004, -0.0623,  0.0989,  0.1041,  0.0244,  0.0148,  0.0219,  0.0862,
        -0.0241,  0.0214, -0.0197,  0.0756, -0.0545, -0.0143,  0.0100, -0.0987,
        -0.0168, -0.0068,  0.0341, -0.0401,  0.0420,  0.0290,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0386,  0.0894,  0.1032,  0.0336, -0.0108,  0.0264,  0.0934, -0.0305,
         0.0358,  0.0262, -0.0098, -0.0181, -0.0088,  0.0188,  0.0062, -0.0010,
        -0.0087, -0.0215, -0.0514,  0.0069,  0.0557,  0.0369, -0.0625,  0.0197,
        -0.0102,  0.0256, -0.0440,  0.0205,  0.0526, -0.0332,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1709, -0.0986, -0.0319, -0.0877, -0.2031,  0.0380,  0.0205,  0.1007,
        -0.0373, -0.1085,  0.1028,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0476,  0.0688, -0.0705,  0.0167, -0.0342, -0.1533, -0.0299, -0.0280,
        -0.0020,  0.0236, -0.0807, -0.0411, -0.0007, -0.0188, -0.0094,  0.0418,
         0.0093, -0.0024, -0.0194, -0.0438, -0.0437, -0.2145,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0117,  0.1257, -0.1055,  0.0032,  0.0094, -0.0999, -0.0206,  0.0443,
        -0.0302, -0.0036, -0.0004,  0.0991, -0.0010, -0.0084,  0.0174,  0.0553,
         0.0114,  0.0127, -0.0118,  0.0221,  0.0077, -0.0136,  0.0694,  0.0427,
        -0.0175, -0.0158, -0.0068, -0.0156,  0.0208,  0.0496,  0.0086,  0.0255,
        -0.0128,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1078, -0.0713, -0.0302, -0.0174, -0.0338, -0.0015,  0.0227,  0.0136,
        -0.0195, -0.0116, -0.0083, -0.0138,  0.0514, -0.0242, -0.0328, -0.0679,
         0.0029,  0.0979,  0.0154,  0.0988,  0.0671, -0.0773, -0.0094,  0.0043,
        -0.0079, -0.0283,  0.0365,  0.0260,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0142, -0.2947, -0.0524, -0.0057,  0.0383, -0.0154, -0.0205, -0.0495,
        -0.0898,  0.0055, -0.0850,  0.0259, -0.0114,  0.0401, -0.0158,  0.0678,
        -0.0019,  0.0117,  0.0123, -0.0175, -0.0070, -0.0093,  0.0712,  0.0257,
        -0.0116,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 0.0038,  0.1279, -0.0101,  0.0150,  0.1060,  0.0338, -0.0046,  0.0092,
         0.0299, -0.0460, -0.0052, -0.0071, -0.0230, -0.0277,  0.0073, -0.0105,
         0.0146,  0.0241,  0.0182, -0.0035, -0.0134,  0.0043,  0.0276,  0.0094,
         0.0206,  0.0079,  0.0144,  0.0227,  0.0403, -0.0939, -0.0067, -0.0181,
        -0.0092,  0.0226,  0.0243, -0.0226, -0.0007, -0.0095,  0.0423, -0.0048,
        -0.0574,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.2560, -0.1049, -0.0244,  0.0373, -0.0015,  0.0543, -0.0478, -0.0193,
        -0.0291, -0.0022,  0.0200, -0.0044,  0.0500,  0.0237,  0.0156,  0.0032,
        -0.0352, -0.0405,  0.0189, -0.0251,  0.0131, -0.0039,  0.0145,  0.1040,
         0.0511,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0814,  0.0873, -0.0372,  0.0598,  0.0040,  0.0791,  0.0769,  0.0196,
        -0.0631, -0.0429, -0.1334, -0.0272, -0.0041,  0.0129,  0.0448,  0.0113,
         0.0009, -0.0340, -0.1063, -0.0421, -0.0319,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0019, -0.2427, -0.0324,  0.0589, -0.0259, -0.0596, -0.0170, -0.0656,
         0.0219, -0.2125, -0.0376,  0.0249, -0.0013, -0.0778,  0.0033, -0.0280,
         0.0450, -0.0437,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0441, -0.0213,  0.0353,  0.0131,  0.0083, -0.0045, -0.0255, -0.0291,
         0.0105, -0.0022,  0.0312, -0.0182, -0.0357, -0.0649, -0.0043,  0.0270,
        -0.0391, -0.0529,  0.0259,  0.0714,  0.0289, -0.0091,  0.0064, -0.0056,
         0.0587,  0.0032,  0.0228,  0.0217,  0.0333,  0.0013, -0.0057,  0.0162,
        -0.0030, -0.0350,  0.0376,  0.0256, -0.1089,  0.0121,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0976,  0.0797,  0.0806, -0.0264,  0.0579,  0.0003,  0.0584, -0.0690,
         0.0109, -0.0157, -0.0259,  0.0271,  0.0378,  0.0232,  0.0390,  0.0541,
         0.0048, -0.1001, -0.0778, -0.1140,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2226e-01,  1.1800e-02, -6.1464e-02, -1.3507e-02, -3.0070e-02,
         1.1208e-01,  8.7046e-03, -7.9636e-02,  7.6886e-02,  3.6531e-02,
        -4.3610e-02,  6.5688e-05,  9.8679e-02, -1.0471e-01,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0290,  0.0449,  0.0355,  0.0344, -0.0430, -0.0388, -0.0383, -0.0099,
        -0.0426,  0.0403, -0.0101,  0.0012,  0.0068, -0.0257,  0.0041, -0.0075,
         0.0665, -0.0271, -0.0587, -0.0294,  0.0479, -0.0564, -0.0273, -0.0383,
        -0.0132, -0.0155,  0.0028, -0.0221, -0.0021,  0.0507,  0.0278, -0.1022,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1332, -0.0176, -0.0553, -0.0945,  0.0757,  0.0463, -0.0380, -0.0014,
        -0.0005,  0.1196, -0.0253, -0.0420, -0.0267, -0.0842, -0.0553, -0.0063,
         0.0229,  0.0085, -0.0587, -0.0878,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.1216e-02, -1.4139e-01, -5.4027e-02,  2.6644e-02, -6.0012e-02,
        -2.2164e-02,  1.1214e-02,  6.3436e-03,  1.4114e-02,  5.6464e-03,
        -1.2626e-02, -4.1894e-03,  2.9181e-03,  4.3881e-03,  1.2309e-02,
         1.1423e-02,  2.0872e-03,  1.5087e-02,  2.1752e-02,  1.1556e-02,
         2.1109e-02,  9.1105e-04, -1.6137e-03, -1.6420e-02,  6.3295e-03,
         2.1362e-02, -3.7570e-03, -1.9339e-03, -2.3037e-02,  2.4409e-02,
        -7.7266e-02,  2.1799e-03, -1.3386e-02, -3.0713e-02,  8.8038e-03,
        -1.2504e-02, -2.0975e-02, -8.4397e-06,  9.8298e-03, -2.1115e-02,
         1.9355e-02, -1.5357e-02,  6.9852e-03,  3.3065e-02, -3.7821e-03,
        -1.9310e-02, -7.3380e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1860, -0.0287,  0.0993,  0.0817, -0.0488,  0.0448,  0.0100, -0.0436,
         0.0727, -0.0408,  0.1021,  0.0075,  0.0307,  0.1517, -0.0516,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0649e-01, -5.2824e-02,  2.9054e-03, -3.8456e-02, -8.3720e-03,
        -1.2352e-02, -4.8769e-03, -8.2371e-03,  2.0950e-02,  6.0116e-03,
         5.7356e-03,  3.0657e-03, -2.3571e-02, -6.1054e-03,  1.0648e-02,
        -7.8831e-03, -9.0812e-03,  1.0697e-02,  3.3031e-02,  1.5204e-02,
         2.5799e-02,  4.8162e-02,  3.0498e-02, -1.1216e-02, -6.7948e-03,
        -6.3191e-03, -7.5902e-03,  3.6475e-02, -2.9957e-03, -7.3421e-03,
        -3.0355e-02, -1.3150e-02, -5.9407e-02, -1.0459e-02,  2.4195e-02,
         2.8844e-04,  9.4197e-04,  9.8884e-03, -6.2752e-03,  1.6050e-02,
        -2.7684e-05,  9.2292e-03, -2.7672e-02, -6.5116e-03, -1.1294e-02,
        -3.1109e-02,  6.4453e-02, -1.0900e-01], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-0.1266,  0.1505,  0.0354,  0.0291, -0.0129,  0.0301, -0.0192, -0.1099,
        -0.1575, -0.0227, -0.0085,  0.0077, -0.0120,  0.0025,  0.0069, -0.0033,
         0.0016,  0.0609,  0.0169,  0.0479,  0.0026, -0.0128,  0.0104,  0.0340,
        -0.0012,  0.0771,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1195, -0.2303, -0.0324, -0.0701, -0.0639,  0.0076,  0.1031,  0.0471,
         0.0606,  0.0420,  0.0022, -0.0343,  0.0169,  0.0136,  0.0117, -0.0342,
        -0.0178,  0.0193,  0.0323,  0.0194,  0.0216,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1687, -0.0573, -0.0127,  0.0680,  0.0376,  0.0455,  0.0362, -0.0456,
         0.0020, -0.0111, -0.0034,  0.0112,  0.0243,  0.0131,  0.0082, -0.0056,
        -0.0351,  0.0279, -0.0268, -0.0482, -0.0069, -0.0065,  0.0079,  0.0173,
        -0.0007,  0.0079, -0.0013, -0.0180, -0.0182, -0.0090,  0.0139,  0.0268,
         0.0071,  0.0071, -0.1566, -0.0062,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.2493,  0.1404, -0.0089,  0.0627, -0.0021, -0.0583,  0.1927, -0.0463,
        -0.0147,  0.0151,  0.0247, -0.0008, -0.0043, -0.0016, -0.0021,  0.0435,
        -0.0183,  0.0086,  0.0761, -0.0292,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1668, -0.3130, -0.0146, -0.0100,  0.0100,  0.0361,  0.0120,  0.0563,
         0.0544,  0.0055, -0.0313, -0.0245,  0.0250,  0.0201,  0.0086,  0.0439,
        -0.0286,  0.0054,  0.0135,  0.0556, -0.0646,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1293, -0.3065, -0.1481,  0.1118,  0.0192, -0.0003, -0.0714,  0.0414,
        -0.0186, -0.0524,  0.0280,  0.0598, -0.0133,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1358e-01, -1.0481e-01, -4.8490e-02, -2.5968e-02,  4.1374e-02,
         1.8213e-02,  3.1202e-02,  5.5505e-03, -3.2771e-03, -5.3080e-03,
         5.4970e-03,  1.2985e-02,  2.6455e-02,  1.0902e-02,  1.4293e-02,
        -8.7459e-03,  1.1990e-03,  4.5557e-02,  1.6582e-02, -5.0651e-03,
         3.7908e-02, -2.3686e-02, -2.6539e-02, -6.9834e-02,  3.4986e-02,
        -3.8836e-06,  4.4990e-02, -2.1778e-02,  9.6271e-03, -8.6194e-03,
         1.8638e-02, -2.5469e-02, -1.6393e-02, -1.8832e-02,  2.4468e-02,
        -2.1768e-02, -5.1411e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.2101, -0.1578, -0.1031, -0.0616,  0.0373, -0.0241, -0.0317, -0.0005,
        -0.0446,  0.0147,  0.0266, -0.0336, -0.0067, -0.0205, -0.0098, -0.0132,
        -0.0064, -0.0034, -0.0063,  0.0038,  0.0117, -0.0115,  0.0570, -0.1038,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4164e-01,  6.8453e-03, -4.3522e-03, -1.4896e-02, -7.3075e-03,
        -2.0260e-02, -2.9253e-02,  8.9035e-03,  5.0775e-03, -6.0755e-03,
         8.2503e-03,  2.3843e-02,  1.9951e-02, -4.2602e-03, -2.0913e-02,
        -1.2857e-02,  1.6026e-02,  8.1236e-03,  7.3400e-03,  3.9564e-02,
        -1.0113e-01, -1.8259e-01, -6.3381e-03, -2.9686e-02,  6.3137e-02,
         1.4544e-04, -3.2794e-02,  2.4260e-02,  3.9646e-03,  2.2754e-02,
        -5.0649e-02,  7.6815e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0016,  0.1286, -0.0015, -0.0705,  0.0205, -0.0202,  0.0061,  0.0702,
        -0.0402, -0.0100,  0.0300,  0.0099,  0.0106,  0.0134,  0.0017, -0.0127,
         0.0048,  0.0298,  0.0121,  0.0111,  0.0148, -0.0243, -0.0861,  0.0053,
         0.0479,  0.1221, -0.0271,  0.0297,  0.0743,  0.0404,  0.0222,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7914e-01, -6.6179e-02,  4.7912e-02,  7.0399e-02,  3.4394e-03,
         1.6769e-02, -7.5106e-03, -5.8748e-02, -8.4164e-02,  6.8334e-03,
         2.5435e-02, -3.5737e-02, -4.5607e-02, -6.4490e-02, -1.4907e-02,
        -3.7337e-02, -7.9234e-03, -7.7503e-05, -4.0087e-02,  4.4663e-02,
        -2.5070e-02,  5.5038e-02,  6.2530e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0365,  0.3327,  0.0720, -0.0039, -0.0119, -0.0567,  0.0260,  0.0174,
        -0.0067,  0.0071, -0.0402,  0.0778, -0.0012, -0.0843,  0.0234, -0.0034,
        -0.0268, -0.0046, -0.0176,  0.0076,  0.0563,  0.0006, -0.0046, -0.0121,
         0.0089, -0.0108,  0.0188, -0.0300,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 2.4981e-02,  3.8304e-01, -2.8755e-02,  2.0875e-02, -2.8468e-02,
        -7.3376e-03,  1.4867e-02, -2.7834e-02, -3.8865e-02, -1.3121e-02,
         1.3908e-02, -5.6871e-03, -1.5585e-04, -2.4182e-02, -6.4278e-03,
        -1.2769e-02,  2.0623e-02, -1.3916e-02,  4.1333e-02, -1.8718e-02,
         6.0498e-04, -6.7661e-03, -8.3508e-03,  1.5542e-02, -3.5255e-02,
         3.3676e-02,  7.4815e-02,  3.1263e-02,  4.7867e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0932,  0.3085,  0.0623,  0.0058, -0.0371, -0.0558, -0.0366,  0.0343,
        -0.0431, -0.0699, -0.0069, -0.0213,  0.0154,  0.0044,  0.0126, -0.0075,
        -0.0025,  0.0272,  0.0169,  0.0046, -0.0275, -0.0369, -0.0493, -0.0202,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1129,  0.1820,  0.0395, -0.0418,  0.0058, -0.0038, -0.0292, -0.0103,
        -0.0087,  0.0226,  0.0307,  0.0115,  0.0480,  0.0269,  0.0151,  0.0034,
         0.0127, -0.0002, -0.0325,  0.0102,  0.0078, -0.0225,  0.0049, -0.0143,
         0.0034,  0.0124,  0.0123,  0.0075,  0.0189, -0.0125,  0.0727,  0.0632,
         0.0955,  0.0044], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0236, -0.5239, -0.1080,  0.0125,  0.0324, -0.0177, -0.0781,  0.0276,
         0.0210,  0.0514,  0.0027, -0.0081,  0.0213,  0.0094, -0.0093,  0.0033,
        -0.0091, -0.0196,  0.0082, -0.0041, -0.0013, -0.0077,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6472e-01, -3.4178e-01,  5.4992e-02, -3.6383e-02,  3.8497e-02,
        -2.8619e-04, -2.4732e-02, -8.4659e-03,  3.7689e-03,  4.3469e-02,
        -3.8860e-03, -1.1136e-02,  1.0848e-02, -2.2485e-03,  4.4276e-03,
        -2.3912e-02,  1.1073e-01, -1.1571e-01,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0626, -0.2789, -0.0831,  0.0516,  0.0997, -0.0574,  0.0116, -0.0127,
        -0.0189, -0.0522,  0.0044,  0.0085, -0.0115, -0.0006,  0.0048,  0.0420,
         0.0027,  0.0093, -0.0347, -0.0137, -0.0195,  0.0633,  0.0563,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1021, -0.3773, -0.0023,  0.0037,  0.0730,  0.0271,  0.0503, -0.0081,
         0.0203,  0.0266, -0.0677,  0.0007,  0.0032, -0.0543,  0.1086,  0.0158,
        -0.0590,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2667, -0.1032, -0.0746, -0.0312,  0.0234,  0.0038, -0.0910, -0.0119,
         0.0434, -0.0987, -0.0304,  0.0318,  0.0005,  0.0104, -0.0102, -0.0043,
        -0.0969, -0.0678,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0035e-01, -1.6561e-01, -1.8155e-01,  3.4097e-02,  1.1568e-02,
        -1.7885e-02,  1.8829e-02,  7.5562e-03,  1.1548e-05, -6.7507e-03,
         3.2289e-04,  1.5870e-03,  5.3616e-02,  3.3181e-02,  2.4804e-02,
        -2.0434e-02,  3.8795e-02, -1.1668e-02, -3.9421e-03,  3.9758e-02,
        -4.7634e-02, -4.2406e-04, -5.8370e-02, -6.6545e-02,  3.8358e-02,
        -1.6351e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1913, -0.0987,  0.2037, -0.0027, -0.0318, -0.0392,  0.0029,  0.0092,
        -0.0170, -0.0074,  0.0115,  0.0008,  0.0031,  0.0254,  0.0155, -0.0154,
         0.0246,  0.0010,  0.0114,  0.0448, -0.0009, -0.0220,  0.0019, -0.0083,
         0.0220, -0.0246, -0.0390, -0.0093, -0.0932,  0.0164, -0.0053,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0327, -0.3092, -0.0519,  0.0407,  0.0384, -0.0949,  0.0508,  0.0322,
         0.0200,  0.0699, -0.0512,  0.0120, -0.0426, -0.0203,  0.0554,  0.0047,
        -0.0676,  0.0054,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1145, -0.5577,  0.0006, -0.0290, -0.1343, -0.0108,  0.0246, -0.1285,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 2.4761e-02,  3.0714e-01,  5.6311e-02, -3.2081e-02,  6.4542e-03,
        -1.1499e-03, -1.0241e-02,  1.3014e-02,  9.4032e-03,  2.2092e-02,
         1.0335e-02,  2.9656e-02,  6.1825e-03,  1.2574e-03, -7.1796e-04,
         6.2322e-03, -1.0335e-02, -1.7736e-02, -2.6114e-02,  4.5987e-02,
        -6.3238e-02, -2.7011e-02, -1.1892e-02,  9.7922e-03,  8.6787e-03,
         1.0452e-02, -1.3309e-02,  6.5045e-03, -2.1032e-02,  8.5390e-03,
         3.1098e-02,  5.1984e-03,  1.0654e-02,  8.9448e-03, -1.7499e-04,
         3.3349e-02, -7.8892e-03, -4.1840e-02, -4.3208e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0457,  0.2975, -0.0869,  0.0161,  0.0274, -0.0124,  0.0181, -0.0214,
        -0.0287,  0.0613,  0.0730, -0.0074,  0.0424,  0.0123, -0.0067, -0.0009,
        -0.0247, -0.0161, -0.0160, -0.0362,  0.0134,  0.0130,  0.0207, -0.0081,
         0.0098, -0.0076, -0.0680,  0.0086,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0361,  0.3858,  0.1970,  0.0078, -0.0565,  0.0369,  0.0569, -0.0575,
        -0.0092, -0.0126, -0.0848,  0.0046, -0.0542,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.3238, -0.3139, -0.0121, -0.0365,  0.0127, -0.0140, -0.0394,  0.0207,
        -0.0034, -0.0230, -0.0004, -0.0477, -0.0294,  0.0972, -0.0256,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.7928e-02, -1.8523e-01, -5.4234e-02, -6.8455e-02, -5.8660e-02,
        -2.0715e-02,  5.1534e-03, -2.0111e-02, -3.1108e-02, -3.1355e-02,
         1.1458e-02,  1.5302e-02, -1.0122e-02, -4.9961e-03, -5.8855e-03,
         1.6419e-03,  5.4132e-04, -8.4697e-03,  6.8779e-03,  3.2741e-03,
        -6.1129e-03, -1.0988e-02, -1.2933e-02,  1.1161e-03, -7.0830e-02,
        -3.2241e-02, -4.6747e-03, -3.2124e-03, -5.3827e-03,  3.1668e-04,
         1.5561e-02, -1.5679e-02, -9.5282e-06, -7.9514e-03, -5.2560e-02,
        -2.1416e-02,  1.8575e-02,  1.3879e-02, -1.4606e-02, -2.5437e-05,
         2.3064e-03, -4.1853e-03, -1.0828e-02,  7.3340e-03, -1.2354e-02,
         1.1816e-02, -2.1587e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5835e-01, -3.9483e-01,  6.4031e-03,  4.8012e-02, -2.1695e-03,
         1.6626e-02,  2.5254e-02,  6.3146e-04,  4.3290e-02, -6.7762e-03,
         7.7986e-03, -2.9118e-02,  2.0726e-02, -5.1007e-03,  6.4336e-03,
         5.9100e-03,  1.3088e-02,  2.9496e-02,  7.7281e-03, -1.7081e-04,
        -4.8735e-04, -4.3746e-04, -8.9381e-03, -1.3580e-02,  1.1929e-03,
         6.7971e-03, -4.6379e-03, -2.8271e-03,  1.0667e-02,  6.4718e-03,
         8.0178e-03,  8.7222e-03, -5.2426e-03,  1.6181e-03,  5.7146e-03,
         5.8024e-03,  1.1990e-02,  8.4725e-03,  3.1175e-03,  3.5120e-03,
         1.9657e-02, -3.4183e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0448, -0.4733, -0.1265, -0.0086, -0.0138, -0.0173,  0.0197, -0.0094,
         0.0236, -0.0290, -0.0277, -0.0215,  0.0052, -0.0091,  0.0162, -0.0132,
         0.0015, -0.0207, -0.0158, -0.0044, -0.0034,  0.0066, -0.0068, -0.0193,
        -0.0625,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0253, -0.1202, -0.0641,  0.0866,  0.0227,  0.0092, -0.0455,  0.0127,
        -0.0419,  0.0455,  0.0061, -0.0215, -0.0234,  0.0051, -0.0347,  0.0742,
         0.0026, -0.0245, -0.0195,  0.2035,  0.0166, -0.0227,  0.0719,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0613,  0.3526,  0.0306,  0.0139,  0.0222,  0.0581,  0.0388, -0.0255,
        -0.0092,  0.0674,  0.0158, -0.0140,  0.0624,  0.0151, -0.0086,  0.0008,
        -0.0083,  0.0478, -0.0256, -0.0186, -0.0008, -0.0626, -0.0399,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0756, -0.5100, -0.0682,  0.0014, -0.0070, -0.0143, -0.0514, -0.0360,
         0.0254, -0.0505,  0.0762, -0.0250, -0.0590,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1553,  0.4190, -0.0567, -0.0709,  0.0084, -0.0308, -0.0183, -0.0345,
        -0.0362, -0.0464, -0.0605, -0.0147, -0.0203, -0.0278,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2765,  0.3287,  0.0566, -0.0036,  0.0443,  0.0069, -0.0004,  0.0069,
        -0.0024, -0.0216,  0.0041,  0.0072, -0.0029, -0.0470,  0.0300, -0.0276,
         0.0121, -0.0678,  0.0535,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-0.0154,  0.2242, -0.0270, -0.0025, -0.0418, -0.0297, -0.0124,  0.0185,
         0.0388,  0.0065,  0.0111,  0.0368, -0.0121, -0.0202,  0.0297, -0.0359,
         0.0090, -0.0107, -0.0333, -0.0048,  0.0151,  0.0120, -0.0358, -0.0300,
         0.1190,  0.1679,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.1408e-02, -5.8211e-01, -5.9623e-03,  8.2773e-02,  3.6263e-02,
        -2.9542e-02,  5.2341e-02,  5.1493e-03,  1.8302e-02,  6.3300e-03,
         8.0500e-05, -1.9751e-04,  6.6261e-02,  3.4733e-02, -9.0339e-03,
         9.5110e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1124, -0.3819,  0.0308, -0.0891,  0.0708, -0.0482, -0.0394, -0.0266,
         0.0124,  0.0032, -0.0186,  0.0423, -0.0201,  0.0075, -0.0217, -0.0217,
        -0.0534,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1484, -0.5874, -0.0115,  0.0853, -0.0212,  0.0065,  0.0161, -0.0740,
         0.0495,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1056, -0.1533, -0.0976,  0.0018, -0.0010,  0.0160,  0.0460,  0.0028,
         0.0246, -0.0039,  0.0480,  0.0036,  0.0003, -0.0091, -0.0092,  0.0209,
         0.0070, -0.0284, -0.0054, -0.0050,  0.0115,  0.0033,  0.0162, -0.0233,
        -0.0101, -0.0018,  0.0173, -0.0137, -0.0054,  0.0205,  0.0024,  0.0065,
        -0.0051, -0.0086,  0.0077,  0.0162, -0.0040, -0.0183, -0.0193,  0.0005,
        -0.0020,  0.0214,  0.0209, -0.0438,  0.0678, -0.0430], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0896,  0.2001, -0.2072, -0.0022, -0.0343,  0.0174, -0.0818, -0.0125,
        -0.0148, -0.0666, -0.0110, -0.0049, -0.0274, -0.0022,  0.0016, -0.0009,
        -0.0160,  0.0056, -0.0049, -0.0076, -0.0015,  0.0057,  0.0097, -0.0187,
        -0.0133,  0.0038, -0.0012, -0.0209, -0.0117, -0.1051,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0695, -0.6801, -0.0155,  0.0811, -0.0775, -0.0358, -0.0404,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0775, -0.5263,  0.0388, -0.0472,  0.0158,  0.0265, -0.0628,  0.0111,
        -0.0569,  0.0774, -0.0599,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0228, -0.4297,  0.0750,  0.0869,  0.0372,  0.0520, -0.0081, -0.0100,
        -0.0894,  0.0242,  0.0012, -0.0146, -0.1358,  0.0132,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0369,  0.3441,  0.0142,  0.0141,  0.0017, -0.0102,  0.0119,  0.0288,
        -0.0133, -0.0112, -0.0116, -0.0011, -0.0107, -0.0047,  0.0118, -0.0490,
        -0.0076, -0.0145,  0.0055,  0.0058,  0.0201, -0.0146,  0.0008, -0.0032,
        -0.0306,  0.0150, -0.0077,  0.0283, -0.0950, -0.0032,  0.0419, -0.0073,
        -0.0637,  0.0599,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1938,  0.1840,  0.0372,  0.0284, -0.0275,  0.0641, -0.0413, -0.0168,
         0.0171,  0.0623, -0.0036,  0.0030,  0.0272, -0.0070,  0.0114, -0.0030,
         0.0287,  0.0013,  0.0043, -0.0105, -0.0501, -0.0135, -0.0060, -0.0093,
        -0.0107, -0.0071, -0.1213, -0.0098,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1940,  0.2613, -0.0172, -0.0129, -0.0427,  0.0138,  0.0236, -0.0128,
         0.0250,  0.0035,  0.0132, -0.0085,  0.0387, -0.0602,  0.0364,  0.0502,
         0.0260, -0.0018, -0.0376,  0.0170, -0.0131, -0.0201,  0.0612, -0.0089,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 0.0580, -0.3557,  0.0624, -0.0809, -0.0248, -0.0198, -0.0275,  0.0051,
        -0.0058,  0.0031, -0.0058, -0.0118,  0.0110,  0.0225,  0.0226,  0.0168,
         0.0146,  0.0440,  0.0166, -0.0232, -0.0225,  0.0080,  0.0313, -0.0005,
         0.0067, -0.0154,  0.0080,  0.0090, -0.0085, -0.0047, -0.0536,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0624, -0.3711,  0.1456,  0.0284,  0.0758, -0.0367, -0.0407,  0.0450,
        -0.0113,  0.0657,  0.0337,  0.0836,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0229, -0.4359, -0.0544,  0.0592,  0.0542,  0.0366, -0.1246, -0.0047,
        -0.0100,  0.0056, -0.0137, -0.0071,  0.0476, -0.0548,  0.0687,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1441e-01, -3.5166e-01, -9.8545e-02, -1.8697e-04, -9.8186e-02,
        -6.3860e-03, -2.3869e-02,  5.6460e-04,  5.1875e-04, -8.0912e-03,
         5.2212e-03,  8.2564e-03,  4.2183e-03,  9.4170e-03,  8.1390e-03,
         1.3939e-02, -9.1834e-03,  1.2175e-02,  4.2899e-02,  1.2548e-02,
        -3.1198e-03,  6.6450e-03, -2.5690e-03,  5.0819e-03,  6.1228e-03,
        -1.7968e-03,  3.4834e-03,  3.8721e-03,  8.6779e-04,  4.1216e-03,
        -1.6545e-02,  9.1327e-03, -1.0823e-01], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0940, -0.2823,  0.0165,  0.0935, -0.0004,  0.0763,  0.0308, -0.0115,
         0.0474,  0.0307,  0.0696,  0.0015, -0.0086,  0.1019,  0.0896,  0.0452,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1750,  0.1571,  0.1394, -0.0378, -0.0568,  0.0969, -0.1135, -0.0782,
         0.1452,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.2425, -0.2851,  0.4413, -0.0264,  0.0047,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0426, -0.4522,  0.0063, -0.0200,  0.0191, -0.0169,  0.0022, -0.0170,
        -0.0392,  0.0257, -0.0162,  0.0130,  0.0423,  0.0689, -0.0145,  0.0015,
         0.0041, -0.0137, -0.0207, -0.0154,  0.0216, -0.0147, -0.0281,  0.0257,
        -0.0058, -0.0310, -0.0216,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0022, -0.5917, -0.0649, -0.1580,  0.1033, -0.0254,  0.0545,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0472, -0.5605,  0.0256,  0.0317,  0.0206,  0.0135,  0.0551, -0.0090,
         0.0373, -0.0040,  0.0121, -0.0034,  0.0022,  0.0091,  0.0202,  0.0255,
        -0.0088,  0.0088,  0.0040, -0.0345,  0.0339,  0.0331,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3999e-01,  3.0704e-01,  2.3873e-04, -8.6871e-03,  3.2747e-02,
        -3.4584e-02, -3.9073e-02,  6.1282e-02, -3.7736e-02, -3.9786e-03,
        -4.7043e-04, -2.3652e-02,  2.7512e-02,  1.7042e-02, -3.4103e-02,
        -8.8054e-03,  1.8488e-02, -7.1007e-02,  3.3568e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4696e-02,  5.5510e-01, -1.3530e-02, -3.6562e-02, -3.4885e-02,
        -1.0504e-02,  1.9439e-02, -2.6498e-03,  1.2143e-02,  9.6058e-03,
        -2.3993e-02,  3.4606e-03,  2.5673e-03,  2.2807e-02, -5.5007e-03,
         1.3378e-02, -2.7706e-03, -6.1483e-03,  3.4269e-03, -2.4395e-02,
         9.5966e-03,  2.8797e-02,  1.3943e-03, -3.6298e-03,  2.8647e-03,
         1.6332e-02,  4.1445e-05,  1.8092e-02,  2.8711e-02, -5.3026e-03,
        -5.7672e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 0.1019, -0.4218,  0.0801,  0.0012,  0.0297, -0.0103, -0.0094,  0.0137,
        -0.0297,  0.0114,  0.0035, -0.0053,  0.0059,  0.0024, -0.0084,  0.0063,
         0.0222,  0.0088,  0.0397,  0.0034,  0.0043,  0.0140,  0.0198,  0.0184,
        -0.0191, -0.0063, -0.0026, -0.0023,  0.0126,  0.0369,  0.0040,  0.0449,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1233,  0.3656,  0.1034, -0.0329,  0.0171, -0.0161, -0.0304, -0.0073,
        -0.0254,  0.0087,  0.0096, -0.0133, -0.0012,  0.0366, -0.0674,  0.0322,
         0.0067,  0.0301, -0.0114,  0.0125, -0.0476, -0.0013,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0163, -0.4393,  0.0497, -0.0110,  0.0196,  0.0011,  0.0666, -0.0554,
        -0.0654,  0.0083, -0.0084, -0.0121,  0.0035,  0.0081,  0.0098, -0.0074,
         0.0050, -0.0008, -0.0346,  0.0424, -0.0166,  0.0177,  0.0067,  0.0066,
        -0.0263, -0.0144, -0.0028,  0.0043, -0.0125,  0.0271,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0641,  0.4217, -0.1387,  0.0063,  0.0043,  0.0268, -0.0259, -0.0108,
        -0.0169, -0.0069,  0.0483, -0.0448, -0.0161, -0.0056, -0.0184,  0.0083,
         0.0951, -0.0410,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0200, -0.3428,  0.0121, -0.0582, -0.0512, -0.0444, -0.0139,  0.0103,
        -0.0036,  0.0181,  0.0046, -0.0039,  0.0030, -0.0053, -0.0135,  0.0373,
         0.0211,  0.0063, -0.0093,  0.0111,  0.0078,  0.0020,  0.0382,  0.0015,
         0.0031, -0.0014, -0.0201,  0.0091, -0.0063,  0.0113, -0.0017, -0.0038,
        -0.0025, -0.0173, -0.0038,  0.1431, -0.0371,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.7019e-03, -5.8048e-01, -5.2380e-02, -1.2788e-02,  2.3728e-02,
        -1.5701e-03, -6.1949e-03, -1.9228e-02,  1.7203e-02, -4.7362e-03,
         3.4617e-03, -2.0472e-02, -2.2101e-02, -2.0739e-02,  2.1760e-02,
         1.3549e-02,  1.7703e-02, -1.3184e-02, -1.3648e-03, -4.3890e-04,
        -3.2759e-02,  5.5666e-03,  8.4219e-02, -1.5671e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.0978e-03, -3.5970e-01, -2.2251e-02,  1.4867e-02, -6.5264e-02,
         7.4901e-02, -1.5139e-02, -5.8252e-03, -5.9104e-02,  2.3751e-02,
         7.2553e-03,  1.2508e-02, -3.6247e-03,  3.4595e-03, -8.8316e-06,
         9.0083e-03,  5.4202e-03,  1.1879e-02, -5.2555e-03,  3.1612e-03,
         2.4166e-03,  1.0503e-02,  7.1894e-03,  1.6192e-02, -2.5774e-02,
        -7.1928e-03,  6.8425e-03,  6.7250e-04,  5.5667e-03, -1.5734e-02,
        -3.1307e-02,  1.4660e-02, -3.2155e-03,  6.5440e-03, -2.1142e-03,
         2.7987e-02,  6.3719e-03,  5.7253e-03,  3.6354e-03,  2.5438e-03,
         8.8650e-03,  7.2064e-03,  1.0290e-02, -5.2630e-05,  1.1470e-02,
        -4.9451e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1261, -0.2903,  0.0040, -0.0556,  0.0262, -0.0196,  0.0303,  0.0127,
         0.0245,  0.0158, -0.0210, -0.0017, -0.0340,  0.0184, -0.0320, -0.0088,
         0.0675, -0.0111, -0.0181, -0.0061, -0.0268, -0.1083, -0.0410,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0229, -0.6595, -0.0750,  0.0320, -0.0066,  0.0782, -0.0203, -0.0082,
         0.0368,  0.0484, -0.0122,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1461,  0.2835, -0.0824,  0.0806, -0.0511,  0.0046, -0.0947, -0.0123,
         0.0021, -0.0124, -0.0807, -0.0084,  0.1120,  0.0292,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0701,  0.4201, -0.0540, -0.0840,  0.0253, -0.0144, -0.0338,  0.0062,
        -0.0259, -0.2389, -0.0018, -0.0255,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0988e-02, -3.2478e-01,  5.3858e-02,  1.1287e-01, -4.2126e-03,
         2.6693e-02,  1.2433e-02,  5.8726e-03,  1.9310e-02,  6.5101e-03,
        -3.1468e-03,  1.3307e-02,  7.3569e-03,  7.4052e-03,  2.2550e-02,
         6.4320e-03,  2.0157e-02,  1.8220e-03,  1.6825e-02, -1.4802e-03,
         3.9181e-03,  1.7599e-02,  4.3225e-03, -2.2582e-02,  3.6055e-04,
        -8.1993e-05, -7.0793e-03,  3.2607e-02,  4.2532e-03,  1.8331e-02,
         1.2676e-02,  4.9693e-03, -1.7856e-02, -1.6085e-02,  9.6156e-03,
         6.4742e-04,  1.5022e-03,  4.7707e-03, -1.9155e-03, -1.9576e-03,
        -9.6851e-04, -6.0722e-04, -1.3734e-02,  4.5989e-03, -8.6878e-03,
        -2.3293e-02,  1.2503e-03, -1.0554e-03,  4.6465e-03,  6.3799e-03,
        -9.4783e-03,  7.4304e-03,  5.8413e-03,  2.8641e-03,  1.8026e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-0.2450, -0.3141, -0.0583, -0.0116,  0.0497,  0.0262, -0.0040, -0.0020,
        -0.0113,  0.0138,  0.0136,  0.0149,  0.0116, -0.0091,  0.0310,  0.0064,
         0.0314, -0.0054,  0.0026,  0.0085, -0.0123,  0.0665, -0.0506,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1158, -0.3877, -0.0970,  0.0451,  0.0051, -0.0064, -0.0128,  0.0734,
         0.0741, -0.0077,  0.0403,  0.0287,  0.0193,  0.0086,  0.0035,  0.0044,
         0.0360, -0.0342,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.2568, -0.3038,  0.0874,  0.0510,  0.0486,  0.0363, -0.0611,  0.0053,
         0.0161,  0.0234,  0.0050,  0.0305,  0.0181,  0.0083, -0.0019, -0.0061,
         0.0180, -0.0224,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1542,  0.5701,  0.0140, -0.0476, -0.0026,  0.0248,  0.0281, -0.0307,
        -0.0139, -0.0076,  0.0283, -0.0327, -0.0015,  0.0114, -0.0324,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0848,  0.2548,  0.0341,  0.0571,  0.0289,  0.0074,  0.0028,  0.0089,
         0.0029,  0.0018, -0.0289, -0.0276,  0.0176, -0.0232, -0.0487, -0.0757,
         0.0145,  0.0309,  0.0193,  0.0098, -0.0008,  0.0009, -0.0005,  0.0055,
        -0.0092, -0.0127, -0.0055, -0.0032, -0.0318,  0.0214,  0.0397, -0.0891,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0783, -0.3334, -0.0067,  0.0317,  0.0098, -0.0022, -0.0091,  0.0076,
         0.0232,  0.0088, -0.0191,  0.0088,  0.0207,  0.0085, -0.0017,  0.0192,
         0.0005, -0.0152,  0.0238,  0.0235, -0.0513,  0.0094,  0.0106,  0.0309,
         0.0198,  0.0329,  0.0077, -0.0082,  0.0224,  0.0244,  0.0199,  0.0570,
        -0.0537,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.5389e-02,  2.6743e-01, -3.1585e-03,  2.5906e-02, -1.3831e-02,
        -5.7898e-02,  4.3730e-04, -3.2917e-03,  3.9174e-03,  1.6770e-02,
        -6.8596e-03, -2.4774e-02,  9.7333e-03, -4.3647e-02, -3.9350e-02,
         5.0621e-02,  7.3735e-03,  2.0666e-03, -1.2475e-03,  5.9287e-03,
        -3.8694e-03,  3.5180e-03, -1.7855e-03,  4.5143e-03, -1.0635e-03,
        -6.3304e-03,  1.7171e-02,  2.5653e-02, -2.0249e-03, -1.8763e-02,
         1.3825e-02,  8.5841e-03, -2.5531e-03,  1.3244e-02,  1.6798e-03,
        -1.9968e-02, -4.1254e-03,  8.6472e-03, -3.1547e-05,  7.0652e-03,
         4.9668e-03, -1.4767e-03,  2.7782e-03, -1.7728e-02, -5.0368e-03,
         1.3768e-02, -1.4299e-02, -6.7773e-03, -1.5419e-02, -3.1497e-03,
        -2.8093e-02, -9.9632e-03, -6.1218e-03, -2.5447e-03, -1.9031e-02,
         1.5120e-03, -2.2310e-02, -1.0981e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0498, -0.4291, -0.1566, -0.0195, -0.0226,  0.0065,  0.0215, -0.0059,
        -0.0226,  0.0101, -0.0132, -0.0045, -0.0281, -0.0133, -0.0023, -0.0053,
         0.0201,  0.0087,  0.0102,  0.0054,  0.0156, -0.0020,  0.0125,  0.0380,
         0.0278,  0.0489,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.6746e-02,  4.7184e-01, -2.9627e-02, -1.0097e-03,  2.6587e-02,
        -1.3831e-02, -9.6376e-03,  1.6524e-02, -6.1307e-03, -2.3036e-03,
         3.1742e-04, -1.1842e-02,  1.9846e-02,  1.4549e-03, -6.1726e-03,
         5.9132e-03, -1.7639e-02,  4.6495e-03,  9.8680e-03,  5.2561e-02,
         1.4456e-02,  3.4470e-02, -2.6897e-02,  9.1435e-02, -5.8239e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8216e-02, -2.8814e-01,  3.6542e-02,  1.3902e-01, -2.3950e-02,
        -2.5884e-02, -3.3793e-03,  1.9017e-02,  2.3307e-02, -1.5945e-02,
         3.3532e-02, -2.3929e-02, -9.5125e-04,  1.2602e-02, -7.8633e-03,
        -1.0296e-02, -3.0277e-02, -6.5412e-03,  6.0693e-03, -3.0141e-03,
        -1.2434e-02, -1.5809e-02, -1.0068e-03, -2.6393e-02, -6.8893e-03,
         1.9938e-02,  1.2364e-02,  1.1394e-03, -1.1591e-02,  3.5155e-03,
        -5.5636e-03, -4.4665e-03, -1.4473e-02,  8.9027e-03, -2.1516e-02,
        -2.8512e-03, -9.9794e-03,  8.9369e-04, -2.2175e-03, -4.0267e-05,
        -2.4979e-03,  1.3494e-02, -6.1583e-03,  1.0861e-02,  5.6532e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0556,  0.4593, -0.0417,  0.0624,  0.0041,  0.0417,  0.0428, -0.0007,
        -0.0289,  0.0052, -0.0204,  0.0234,  0.0094,  0.0089,  0.0227, -0.0646,
         0.0348, -0.0078,  0.0024,  0.0043,  0.0268,  0.0232, -0.0089,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0276, -0.3809,  0.0957, -0.1357, -0.0357, -0.0361, -0.0288, -0.0186,
         0.0293,  0.0331,  0.0186,  0.1600,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-0.1562,  0.4100,  0.0443,  0.0163, -0.0366, -0.0016, -0.0220,  0.0081,
         0.0093, -0.0093,  0.0079,  0.0082, -0.0033,  0.0305, -0.0244, -0.0061,
        -0.0046,  0.0482,  0.0021, -0.0067,  0.0303, -0.0144, -0.0996,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8215e-02,  3.1322e-01,  1.9707e-01,  3.9566e-02, -7.2133e-03,
        -6.9102e-03,  4.0238e-03, -2.9283e-03,  1.4052e-02,  1.4398e-02,
         1.2741e-02,  7.2029e-03,  1.2052e-02, -8.0352e-03, -3.9237e-02,
         1.5977e-03, -7.3397e-03,  2.3817e-02,  1.2180e-03,  3.5539e-03,
        -1.6780e-02, -8.7967e-03, -7.6891e-03, -6.2745e-03,  2.7671e-04,
        -2.1085e-02,  1.4230e-02, -1.5646e-03,  1.5740e-02,  5.4837e-03,
         1.8257e-02, -2.6668e-02,  9.2759e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0130, -0.1378, -0.0318, -0.1280, -0.0916,  0.0083, -0.0378,  0.0243,
         0.0760,  0.0066,  0.0580,  0.0396, -0.0587, -0.0555,  0.0153, -0.0148,
         0.0244, -0.0411,  0.1374,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1284, -0.4804, -0.0465, -0.0684, -0.0745, -0.0007,  0.0144, -0.0007,
         0.0201,  0.0179, -0.0038, -0.0388,  0.0343, -0.0711,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0465, -0.4761, -0.0521,  0.0012, -0.0058,  0.0158,  0.0231,  0.0036,
        -0.0007,  0.0063, -0.0243,  0.0078,  0.0057,  0.0971, -0.0219,  0.0079,
         0.0377, -0.0034,  0.0053,  0.0047, -0.0429, -0.1100,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0635, -0.3687,  0.0150, -0.0452, -0.0068, -0.0228, -0.0152,  0.0528,
         0.0006,  0.0427, -0.0216, -0.0139,  0.0343,  0.0063,  0.0104,  0.0265,
        -0.0119, -0.0309, -0.0112,  0.0297, -0.0216, -0.0093,  0.0206,  0.0185,
        -0.0148,  0.0150, -0.0063, -0.0012, -0.0102, -0.0523,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1046,  0.1983, -0.0382, -0.0449,  0.0402,  0.0818, -0.0562, -0.0407,
        -0.0098,  0.0055, -0.0188, -0.0165, -0.0482,  0.0231,  0.0316, -0.0734,
        -0.1682,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0403, -0.5671, -0.0570, -0.0744,  0.0521,  0.0494,  0.0203,  0.0651,
         0.0024,  0.0224,  0.0495,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1767, -0.3702, -0.0421,  0.0129, -0.0236,  0.0830, -0.0196,  0.0050,
         0.0145,  0.0095, -0.0348,  0.0016, -0.0212, -0.0101,  0.0580, -0.0024,
        -0.1147,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0744,  0.4476,  0.0664,  0.0233,  0.0021, -0.0081,  0.0144, -0.0023,
         0.0092, -0.0130,  0.0293, -0.0184,  0.0120,  0.0340, -0.0544,  0.0044,
        -0.0052, -0.0044, -0.0575,  0.0188, -0.0182,  0.0016, -0.0098, -0.0059,
        -0.0532,  0.0120,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5519e-02,  4.8969e-01,  5.8056e-02, -1.5752e-02, -2.4804e-02,
        -2.0005e-02, -3.0859e-02,  3.5952e-02, -4.0891e-02,  3.8957e-03,
         2.2105e-04,  2.0129e-02,  4.4697e-04, -5.8034e-04, -4.9696e-03,
        -4.0818e-02, -1.8513e-02,  3.7689e-03, -1.0627e-02, -5.5730e-03,
        -3.6990e-02, -1.2335e-02, -1.1098e-02, -3.6773e-03,  8.3809e-03,
        -1.0387e-04, -5.7266e-03, -8.6600e-03, -4.4813e-02,  1.7140e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1556, -0.3317,  0.0143,  0.0286, -0.0971, -0.0227, -0.0076, -0.0515,
        -0.0087,  0.0848,  0.0075, -0.0308, -0.0296,  0.0019, -0.0317, -0.0053,
        -0.0005,  0.0628,  0.0273,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 0.0104,  0.5065,  0.0589,  0.0019, -0.0153, -0.0396,  0.0072, -0.0060,
        -0.0133, -0.0062, -0.0052, -0.0154, -0.0118, -0.0039, -0.0296, -0.0180,
         0.0112, -0.0329,  0.0028,  0.0022,  0.0060, -0.0049,  0.0137, -0.0056,
         0.0117, -0.1420,  0.0179,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1112,  0.3066,  0.0115, -0.1396,  0.0495, -0.0055, -0.0753, -0.0238,
        -0.0047,  0.0318, -0.0635, -0.0167,  0.0955,  0.0648,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0232,  0.4280,  0.0378,  0.0272, -0.0139,  0.0523,  0.0114, -0.0398,
        -0.0478,  0.0195,  0.0070,  0.0046,  0.0111,  0.0247,  0.0514,  0.0110,
        -0.0159, -0.0034,  0.0399,  0.0171,  0.0031,  0.0118, -0.0198,  0.0283,
         0.0464,  0.0039,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0764,  0.3233,  0.0479, -0.0251, -0.0299, -0.0336, -0.0139,  0.0070,
        -0.0774,  0.0510,  0.0054, -0.1817,  0.1073,  0.0202,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0178, -0.2929,  0.1060,  0.0102,  0.0175,  0.0357,  0.0137,  0.0223,
         0.0121, -0.0077, -0.0125, -0.0066,  0.0082,  0.1377,  0.0282,  0.0167,
         0.0502,  0.0062, -0.0044,  0.0018,  0.0148,  0.0141,  0.0176,  0.0053,
        -0.0109,  0.0138, -0.0005, -0.0046,  0.0526, -0.0255,  0.0165, -0.0154,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9626e-02,  5.1260e-01, -3.5648e-02,  1.3667e-02,  3.0960e-02,
        -2.3486e-02, -8.5221e-03,  5.0665e-02,  2.3434e-02, -9.8309e-05,
        -2.4812e-02,  1.3829e-02,  8.0966e-03, -7.2504e-03,  9.5839e-03,
         9.7042e-03, -4.0043e-02,  1.5378e-02,  1.5565e-02,  7.3611e-03,
        -1.0797e-02,  6.7158e-03,  6.2406e-03,  7.1563e-02, -2.4350e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0355,  0.5336, -0.1332, -0.0086, -0.1894, -0.0995,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1025,  0.4152,  0.0517, -0.0031, -0.0524,  0.0061, -0.0156,  0.0056,
        -0.0087, -0.0130, -0.0165,  0.0089,  0.0125,  0.0126, -0.0100, -0.0186,
        -0.0051,  0.0066, -0.0119, -0.0183, -0.0184, -0.0078,  0.0177, -0.0087,
         0.0007,  0.0159, -0.0005, -0.0259, -0.0005,  0.0015,  0.0014,  0.0276,
         0.0054,  0.0058, -0.0146,  0.0455,  0.0072], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0595, -0.5812, -0.0198, -0.0381,  0.0124, -0.0252, -0.0070, -0.0145,
        -0.0025,  0.0061,  0.0036,  0.0065,  0.0064,  0.0045,  0.0040, -0.0023,
        -0.0168,  0.0057,  0.0073,  0.0098,  0.0408,  0.0036,  0.0052, -0.0139,
         0.0166,  0.0156, -0.0071, -0.0104,  0.0123,  0.0078,  0.0027,  0.0036,
         0.0145,  0.0104,  0.0020,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0849, -0.4152, -0.0811,  0.1012, -0.0238,  0.0314, -0.0157,  0.0064,
         0.0042,  0.0157,  0.0054,  0.0310,  0.0316, -0.1220,  0.0305,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.4782e-02, -4.7654e-01, -6.7021e-02,  8.0948e-02,  1.4243e-02,
         1.4548e-02, -2.9642e-02, -1.0549e-03, -3.2167e-02,  3.7433e-04,
         1.1777e-02,  1.8828e-02,  1.4065e-02,  9.3361e-03, -7.0868e-03,
         3.4393e-02,  7.3489e-03, -9.7931e-03,  2.2120e-02, -5.5269e-03,
        -8.6870e-03, -2.0942e-03,  8.0331e-03,  4.4153e-03,  6.1446e-03,
        -7.8881e-03,  4.0532e-03, -4.2681e-03, -4.2146e-03,  1.2663e-02,
        -2.3926e-03,  1.0538e-03,  1.2493e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0504, -0.5323,  0.0091,  0.0112,  0.0219, -0.0192,  0.0126,  0.0038,
         0.0062,  0.0037, -0.0071,  0.0301,  0.0277,  0.0569,  0.0383, -0.0028,
        -0.0027, -0.0109, -0.0056, -0.0068,  0.0180, -0.0007,  0.0124, -0.0023,
         0.0059, -0.0062,  0.0542,  0.0349,  0.0063,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 0.0226,  0.5180, -0.1057, -0.0690, -0.0749, -0.0242,  0.0200,  0.0042,
        -0.0233, -0.0321,  0.0293, -0.0049, -0.0201, -0.0386, -0.0132,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2393e-02, -3.1740e-01, -2.8516e-02,  8.8208e-02,  2.7634e-02,
         6.9369e-03,  9.1086e-03, -1.1704e-03, -5.2972e-02,  4.6681e-02,
         1.1096e-02,  5.9655e-03,  1.1085e-02,  3.4315e-02, -7.8132e-03,
        -3.2271e-02,  9.7944e-03,  7.0571e-04,  2.8214e-03, -8.0988e-04,
        -1.7401e-03, -6.5974e-03,  2.7865e-03,  3.3201e-03,  4.1081e-03,
        -1.2793e-03,  4.9143e-05,  5.6094e-03,  5.0959e-02, -5.5822e-03,
         3.8709e-03, -2.5263e-03,  5.1049e-03,  3.7065e-03, -9.6525e-03,
         6.6462e-04,  3.4782e-03, -4.4481e-04, -6.1067e-03, -7.2099e-03,
        -3.4886e-03,  6.6031e-03,  1.6268e-03,  1.0869e-02, -1.2853e-02,
         1.7899e-02, -8.8328e-04,  5.5779e-02, -3.7500e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.2288,  0.2407,  0.1551,  0.0200, -0.0003,  0.0129,  0.0151,  0.0134,
         0.0186,  0.0376,  0.0030, -0.0027, -0.0010, -0.0042,  0.0036, -0.0045,
         0.0180,  0.0026,  0.0099,  0.0004,  0.0099,  0.0116, -0.0004,  0.0010,
        -0.0073, -0.0136, -0.0046,  0.0079,  0.0102, -0.0088, -0.0022,  0.0024,
        -0.0060,  0.0917,  0.0301,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1670e-01,  4.4330e-01,  1.4701e-02,  1.0649e-02, -2.0480e-02,
        -2.1635e-03,  1.5732e-02, -8.3410e-03, -7.6554e-03, -2.0027e-02,
        -1.2798e-02, -5.5662e-03, -1.6316e-04, -8.8793e-03,  4.6142e-03,
        -9.1019e-03, -4.9735e-04, -9.5946e-02, -2.6849e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.2725,  0.3740,  0.0033, -0.0112,  0.0078,  0.0008,  0.0192,  0.0053,
         0.0478,  0.0037, -0.0339,  0.0180,  0.0177, -0.0797, -0.1051,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1081, -0.2312, -0.1095, -0.0130, -0.0167, -0.1058, -0.0122,  0.0061,
        -0.0086, -0.0386,  0.0210, -0.0313, -0.0628, -0.0018, -0.0437, -0.0216,
        -0.0083,  0.0275, -0.0605, -0.0481,  0.0209, -0.0026,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7418e-02, -5.0006e-01, -1.1300e-02,  3.2627e-02, -1.1385e-02,
         3.0367e-04,  3.0970e-02, -1.5424e-02, -3.8744e-02, -1.6744e-02,
        -4.5103e-03,  2.0401e-02,  1.1342e-02,  3.2811e-02, -3.8287e-03,
        -5.8354e-03, -1.5379e-02, -2.2338e-02, -2.4008e-02, -6.4340e-03,
        -5.5421e-04, -3.3136e-03,  1.8825e-02, -2.0664e-02, -5.1446e-03,
         1.5367e-02,  2.4674e-02, -1.2048e-02,  4.2775e-02,  2.4774e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0719, -0.4440,  0.1035,  0.0347,  0.0135,  0.0602,  0.0302, -0.0546,
         0.0084,  0.1763, -0.0026,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.3377e-02, -4.7622e-01,  7.7157e-02, -2.8199e-02, -3.3653e-02,
        -4.2947e-05, -4.7289e-02, -1.1909e-02, -2.8750e-03,  1.5943e-04,
         1.5515e-02, -1.7378e-02, -2.8695e-02, -1.3113e-02,  5.0902e-03,
        -2.0835e-02,  2.2001e-02,  5.8986e-03,  1.5279e-02, -7.1465e-03,
         8.2365e-03,  7.9927e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2186e-01, -4.0429e-01,  8.0977e-02, -5.9334e-03, -2.3767e-02,
         6.7464e-02, -2.5108e-03, -1.0689e-02,  2.1101e-02, -1.5134e-04,
         3.8988e-02,  1.0687e-02, -5.6048e-03,  4.2082e-03,  2.7292e-03,
         1.1607e-02,  2.4589e-03,  1.4821e-02, -4.1792e-04, -7.7653e-03,
         8.8363e-04, -9.5747e-03, -2.5835e-02, -4.0583e-03, -5.3936e-03,
         6.0184e-03, -1.7484e-02,  1.9437e-02,  6.3405e-03,  1.0507e-02,
         9.9359e-03, -2.4277e-02, -2.2221e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0366, -0.3497, -0.0976,  0.0236, -0.0044, -0.0061, -0.0134, -0.0049,
        -0.0050,  0.0166,  0.0016, -0.0142,  0.0069,  0.0109,  0.0132, -0.1142,
         0.0018, -0.0153,  0.0013,  0.0122, -0.0185,  0.0054, -0.0041,  0.0083,
        -0.0035, -0.0178,  0.1580,  0.0348,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0700, -0.3132, -0.0806, -0.0391, -0.0087, -0.0225, -0.0157, -0.0638,
        -0.1176, -0.0295,  0.0033,  0.0065,  0.0050, -0.0027, -0.0085, -0.0229,
        -0.0410, -0.0083, -0.0317, -0.0125,  0.0026, -0.0265,  0.0023, -0.0345,
         0.0309,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 0.0168, -0.4894, -0.0706, -0.0047,  0.0226, -0.0020,  0.0036,  0.0029,
         0.0101,  0.0196,  0.0105,  0.0113, -0.0008, -0.0073,  0.0097,  0.0152,
         0.0223, -0.0104,  0.0030,  0.0066, -0.0052,  0.0046,  0.0103,  0.0201,
        -0.0213, -0.0039,  0.0434, -0.0128,  0.0181,  0.0045, -0.0041,  0.0071,
        -0.0080,  0.0079,  0.0014,  0.0231, -0.0022, -0.0092,  0.0048, -0.0429,
         0.0056,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1379,  0.4018,  0.0403, -0.0470,  0.0490, -0.0333, -0.0005, -0.0202,
         0.0287, -0.0284, -0.0167,  0.0415,  0.0275, -0.0083,  0.0071, -0.0065,
        -0.0119, -0.0009,  0.0006,  0.0160,  0.0040, -0.0340,  0.0041, -0.0249,
        -0.0089,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1742,  0.3983, -0.0405, -0.0675,  0.0155,  0.0268,  0.0492, -0.0176,
         0.0098,  0.0037, -0.0066, -0.0058, -0.0302, -0.0040,  0.0097,  0.0290,
         0.0411, -0.0158,  0.0201,  0.0331, -0.0013,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1535,  0.3485,  0.0908, -0.0040, -0.0123,  0.0170, -0.0078, -0.0124,
        -0.0294, -0.0811,  0.0693, -0.0250, -0.0090, -0.0311,  0.0044, -0.0258,
        -0.0173, -0.0614,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0723,  0.3724, -0.0468,  0.1028,  0.0038,  0.0320,  0.0182, -0.0412,
        -0.0194,  0.0227, -0.0027, -0.0028, -0.0017,  0.0044, -0.0161,  0.0091,
        -0.0159, -0.0045,  0.0170, -0.0194,  0.0017, -0.0254, -0.0132, -0.0101,
         0.0219,  0.0163,  0.0048,  0.0043,  0.0005,  0.0036,  0.0048,  0.0050,
        -0.0083, -0.0059,  0.0138,  0.0106,  0.0082,  0.0163,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0695,  0.4587, -0.0097, -0.0205, -0.0410, -0.0010, -0.0070,  0.0192,
        -0.0118,  0.0070, -0.0424, -0.0152, -0.0122,  0.0148, -0.0176,  0.0332,
        -0.0131, -0.1233,  0.0336, -0.0492,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3937e-01, -3.0912e-01, -1.3336e-01, -1.2192e-02,  1.2435e-02,
         1.9205e-03,  3.0667e-04, -3.3942e-02,  2.8551e-03,  4.2610e-03,
         1.2694e-02,  1.3950e-02,  7.9929e-02, -1.4367e-01,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1361, -0.4061,  0.0249,  0.0271,  0.0338, -0.0160, -0.0192,  0.0121,
         0.0090, -0.0172, -0.0010,  0.0213, -0.0073, -0.0036, -0.0074,  0.0294,
        -0.0236,  0.0018,  0.0116, -0.0021,  0.0049, -0.0451, -0.0153, -0.0412,
        -0.0190, -0.0103, -0.0089, -0.0157, -0.0088, -0.0174, -0.0021,  0.0004,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4806e-01,  4.9973e-01,  1.2858e-02,  5.2477e-02, -3.1750e-02,
         2.1723e-02,  2.6335e-02, -2.4822e-02,  4.5947e-03, -5.7085e-03,
        -9.2717e-03, -1.3046e-02,  2.9711e-03,  1.3408e-02,  8.5447e-03,
         4.6129e-03, -4.7717e-03,  2.7288e-04,  3.2093e-02, -8.2953e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1680e-02,  3.5170e-01,  8.1284e-02,  1.3746e-02,  2.5429e-02,
        -2.3044e-02, -8.7529e-03, -8.7077e-03, -2.5500e-02, -2.0889e-02,
         9.1790e-03, -1.1424e-02, -4.2006e-03,  6.3008e-03, -7.0699e-05,
         4.9216e-04, -1.5275e-02, -9.4028e-03, -1.4636e-02, -1.8830e-02,
        -2.2678e-02,  8.0761e-03,  3.5934e-02,  1.7095e-02, -3.7262e-02,
         7.4262e-03, -1.3012e-02,  3.3276e-03, -3.5871e-02, -1.1744e-02,
         1.0418e-02,  3.6549e-03, -1.0968e-02,  5.9822e-04, -9.9918e-03,
        -1.1637e-02, -1.0984e-02, -3.5408e-03, -4.8345e-03, -7.4517e-03,
         7.7198e-03, -2.9530e-02, -1.0564e-02,  1.3518e-03,  1.1444e-03,
        -2.3102e-03, -2.0334e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0137,  0.4196,  0.0741, -0.0186, -0.0401,  0.0395,  0.0047, -0.0118,
         0.0754, -0.0009,  0.0212,  0.0487, -0.0338,  0.1916,  0.0064,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0124e-03, -3.9548e-01,  1.6468e-02,  1.1863e-02, -7.8475e-04,
         5.9559e-02, -1.8525e-02,  2.3718e-02, -4.3848e-02, -5.0507e-03,
        -2.7574e-02, -6.1998e-03, -2.9608e-03, -3.5970e-03, -9.1966e-04,
         2.0605e-02,  1.1012e-02,  1.2634e-02,  6.4645e-02, -1.0682e-02,
        -8.0829e-03,  4.3670e-03, -1.2446e-03, -2.2303e-03, -8.8106e-03,
         2.1490e-02, -1.8429e-02, -2.8748e-02, -8.3377e-03,  2.4169e-03,
        -6.5748e-03,  4.6403e-04, -4.3991e-03, -3.5614e-04, -2.4844e-03,
        -8.9905e-03, -3.4514e-03, -6.3447e-03,  9.8211e-03, -7.6015e-03,
        -2.7934e-03, -3.9287e-03,  1.0593e-02, -4.4812e-03,  5.8056e-03,
         4.7374e-03, -2.7723e-02, -4.5153e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 1.5278e-02,  4.4577e-01,  7.8051e-03,  3.7741e-02, -1.7077e-02,
        -1.8990e-02,  1.3425e-02,  4.4864e-02, -4.4835e-02, -2.8368e-02,
         2.8648e-02,  2.5035e-02,  9.8524e-03, -3.7918e-03, -1.0793e-03,
        -1.7330e-04,  6.7072e-03,  1.9249e-02, -6.9856e-03, -1.7475e-03,
         1.2352e-03, -3.2333e-02, -1.6173e-02, -4.6199e-02,  5.2595e-02,
        -7.4048e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1074,  0.5300,  0.0423, -0.0234,  0.0182,  0.0143,  0.0264,  0.0300,
        -0.0104,  0.0343, -0.0011, -0.0144, -0.0042,  0.0113, -0.0083, -0.0090,
         0.0061,  0.0065,  0.0223,  0.0769, -0.0034,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2843e-01,  2.8203e-01, -4.5894e-02, -1.0614e-01, -1.8011e-02,
        -7.0563e-03,  8.4503e-03, -3.6515e-03, -1.1937e-02, -1.5173e-02,
        -1.8812e-03,  1.3649e-03, -1.2731e-02, -1.3119e-02, -9.4085e-03,
        -2.4634e-03, -3.7688e-03, -3.4988e-02,  5.0109e-02,  6.3429e-03,
         1.3326e-02,  2.1229e-03,  5.7419e-02,  1.5580e-02,  3.2593e-02,
         2.0182e-03, -1.9309e-02, -9.7572e-03,  3.4529e-04,  1.0820e-02,
        -1.0607e-03,  6.0504e-05,  1.5169e-02, -1.1342e-02, -1.7693e-02,
        -2.8441e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0924,  0.3528,  0.1007, -0.0282, -0.0086, -0.0246,  0.0904, -0.0266,
         0.1046,  0.0061, -0.0068,  0.0170,  0.0070, -0.0053, -0.0176,  0.0189,
         0.0124, -0.0013, -0.0022, -0.0765,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9943e-02,  5.4520e-01,  5.5066e-03, -9.1997e-03, -1.8633e-03,
        -1.9012e-04,  6.9756e-03,  2.1383e-02, -4.3513e-02, -1.8847e-02,
         7.6585e-03, -2.2137e-02, -1.0796e-02, -1.9158e-02, -2.1145e-02,
        -2.8771e-02, -1.6338e-02, -1.4902e-02, -2.4323e-02, -5.9205e-02,
        -6.2941e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0291,  0.6182,  0.0639, -0.0399,  0.0108,  0.0365, -0.0104,  0.0568,
        -0.0134, -0.0216, -0.0100, -0.0453,  0.0441,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0536,  0.3717, -0.0694,  0.0148,  0.0059, -0.0065, -0.0421,  0.0075,
        -0.0041,  0.0213,  0.0258, -0.0088, -0.0154, -0.0013,  0.0390, -0.0005,
        -0.0250, -0.0176,  0.0173, -0.0244,  0.0146,  0.0192,  0.0180,  0.0379,
        -0.0073, -0.0150, -0.0094,  0.0190,  0.0218,  0.0152,  0.0006,  0.0077,
         0.0211, -0.0049, -0.0075, -0.0077,  0.0012], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.8923e-02, -4.3362e-01,  7.4231e-02, -2.6813e-02,  2.1920e-02,
        -1.3399e-02,  1.0137e-02, -4.4543e-02, -4.2202e-04, -4.1429e-03,
        -1.2048e-02, -2.4219e-03,  2.4008e-02,  1.3323e-02, -1.8747e-02,
         1.5544e-02, -2.2248e-03,  2.1078e-02,  6.9569e-03,  1.5429e-02,
         3.4139e-02,  3.0762e-02, -2.3228e-02,  6.1942e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0803,  0.3198, -0.0227, -0.0054, -0.0257, -0.0091, -0.0290,  0.0104,
        -0.0293,  0.0329, -0.0052, -0.0722, -0.0431, -0.0095, -0.0173, -0.0228,
        -0.0210,  0.0135, -0.0068,  0.0143, -0.0288, -0.0514, -0.0116,  0.0120,
        -0.0193, -0.0109, -0.0084, -0.0108,  0.0065,  0.0256, -0.0096, -0.0147,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0408,  0.4002,  0.0281,  0.0346,  0.0146,  0.0054, -0.0356, -0.0798,
         0.0248,  0.0032,  0.0116,  0.0022, -0.0206, -0.0052, -0.0063,  0.0213,
        -0.0144,  0.0222,  0.0146,  0.0114,  0.0122, -0.0040,  0.0162, -0.0027,
        -0.0152, -0.0144, -0.0239,  0.0273,  0.0075, -0.0451, -0.0344,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0119,  0.4680,  0.0584,  0.0045,  0.0557,  0.0306, -0.0073, -0.0188,
         0.0286,  0.0052, -0.0189,  0.0118,  0.0393, -0.0314,  0.0228,  0.0344,
         0.0040,  0.0188,  0.0149,  0.0100,  0.0333,  0.0476, -0.0239,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.2739e-04,  5.0542e-01,  9.5830e-02,  1.0264e-01, -1.9298e-02,
         2.6330e-02, -2.2863e-02,  2.4500e-02, -1.0317e-02, -2.4395e-02,
        -1.9498e-03, -8.9001e-03, -7.4676e-03,  1.8664e-02,  8.0398e-03,
        -2.8134e-03, -2.3471e-02, -1.1233e-02, -1.8930e-02, -4.8487e-03,
        -5.5471e-03, -5.7772e-03,  6.1313e-03, -5.7471e-03,  1.3698e-02,
        -3.0141e-05, -1.6211e-02,  8.4252e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-0.0316,  0.5070, -0.0031,  0.0304,  0.0025,  0.0068,  0.0098,  0.0075,
        -0.0284, -0.0221, -0.0319, -0.0162, -0.0112, -0.0130,  0.0049, -0.0053,
         0.0077, -0.0379, -0.0230,  0.0040, -0.0163, -0.0011, -0.0138,  0.0036,
        -0.0072,  0.0099,  0.0160, -0.0910, -0.0370,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1196, -0.2991, -0.0126,  0.0399, -0.0422,  0.0157,  0.0130,  0.0112,
         0.0385,  0.0452,  0.0320, -0.0016,  0.0029,  0.0291,  0.0254, -0.0041,
        -0.0047, -0.0132,  0.0006, -0.0270,  0.0132,  0.0384,  0.1405,  0.0304,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0350,  0.3269,  0.0739, -0.0093,  0.0467, -0.0306,  0.0275,  0.0304,
         0.0185, -0.0051,  0.0185, -0.0123, -0.0136,  0.0028, -0.0131, -0.0143,
        -0.0330, -0.0116, -0.0176, -0.0018,  0.0424, -0.0065,  0.0023, -0.0129,
         0.0088, -0.0155, -0.0043, -0.0046, -0.0049,  0.0026, -0.0281, -0.0489,
        -0.0719,  0.0034], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.7481e-02, -5.0968e-01, -4.7397e-02,  1.1375e-02, -7.2362e-03,
        -1.5754e-02, -3.1953e-02, -3.3895e-02, -8.3880e-03, -1.4565e-02,
        -3.3976e-02, -3.9560e-03,  4.4085e-03, -2.3095e-02, -3.4029e-02,
         3.5396e-02, -7.0973e-03, -2.3858e-02,  1.8446e-02, -1.6080e-04,
         5.1616e-02, -6.2398e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1390, -0.3070, -0.0377,  0.0757,  0.0519,  0.0361,  0.0165, -0.0401,
        -0.0123,  0.0079, -0.0983,  0.0052,  0.0119, -0.0093,  0.0169,  0.0605,
         0.0588, -0.0150,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9566e-01,  7.4778e-02, -5.0709e-02, -2.9260e-02, -4.6890e-02,
        -2.1195e-03,  4.9910e-03, -8.8486e-04, -1.3996e-02, -2.2723e-04,
         7.3500e-03, -1.3456e-02,  1.6662e-03,  7.5255e-03, -6.8080e-03,
         3.7230e-03,  1.1266e-02,  4.6231e-03,  1.0930e-02,  5.6741e-03,
         8.3452e-03, -1.7839e-01, -2.0733e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1511, -0.3312, -0.0081, -0.0189,  0.0731, -0.0309, -0.0633, -0.0454,
        -0.0407, -0.0643,  0.0089, -0.0067, -0.0297, -0.0034,  0.0019,  0.0765,
        -0.0460,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0933,  0.3623,  0.0800, -0.0311, -0.0416,  0.0318, -0.0099, -0.0193,
        -0.0245, -0.0158, -0.0027, -0.0283, -0.0398, -0.0182,  0.0244, -0.1258,
        -0.0275,  0.0237,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1187, -0.2959,  0.0834,  0.0053,  0.0607, -0.0087,  0.0105, -0.0254,
        -0.0132, -0.0651,  0.0079, -0.0081,  0.0151,  0.0218, -0.0183,  0.0082,
         0.0048, -0.0210, -0.0181,  0.0294,  0.0536, -0.0009, -0.0169,  0.0348,
        -0.0223, -0.0322,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0735,  0.2432,  0.0538,  0.0005, -0.0056,  0.0426, -0.0188, -0.0969,
         0.0020, -0.0070, -0.0360, -0.0513, -0.0108, -0.0260, -0.0205,  0.0160,
        -0.0107,  0.0204, -0.0104,  0.0064, -0.0356,  0.0118, -0.0318,  0.0147,
        -0.0207,  0.0141,  0.0078, -0.0133,  0.0573, -0.0055, -0.0350,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0682,  0.6109, -0.0364, -0.0626,  0.0037,  0.0082, -0.0215, -0.0095,
         0.0418,  0.0501,  0.0035, -0.0232,  0.0189, -0.0045, -0.0062, -0.0094,
        -0.0186, -0.0028,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0714,  0.5699,  0.0049, -0.0815,  0.0247,  0.0321, -0.0865,  0.1289,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 8.4018e-02,  4.3300e-01,  2.1748e-02,  1.1660e-02, -9.3381e-03,
         1.4830e-02,  1.6865e-02, -3.7783e-04, -7.9746e-02,  3.0421e-02,
        -7.6318e-03, -4.2089e-02, -4.0233e-03,  1.0765e-02, -1.7989e-03,
         2.3741e-02,  2.2556e-03, -1.5202e-02, -1.1502e-02, -2.9982e-03,
         2.0477e-03, -1.9613e-03, -2.0321e-02, -1.0495e-02, -1.4306e-02,
        -1.0598e-02,  6.3424e-03,  4.8834e-03, -2.7880e-03,  9.5631e-03,
        -2.2322e-02, -4.0122e-03, -1.2872e-02, -9.6359e-03, -8.3072e-03,
         3.0484e-03,  1.2066e-02, -1.0704e-02, -9.7206e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2634e-02,  4.5640e-01, -4.4588e-04, -5.3771e-03, -2.5240e-02,
        -1.0413e-02, -7.6446e-03, -1.6016e-02,  1.3452e-02, -3.6666e-02,
        -1.0419e-02, -9.6438e-03, -3.9753e-02, -1.9263e-02, -4.2834e-03,
         1.4813e-03,  1.1143e-02, -4.7691e-03,  9.4881e-04, -2.1263e-02,
        -6.0839e-02, -3.4543e-02, -1.5196e-02, -2.3594e-03, -1.2757e-02,
        -3.9070e-03, -7.1712e-02, -5.1429e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0184,  0.6191, -0.0016,  0.1593,  0.0306, -0.0273,  0.0155, -0.0178,
         0.0027,  0.0036,  0.0104,  0.0007, -0.0931,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0494,  0.5883,  0.0256, -0.0057, -0.0292,  0.0055, -0.0443, -0.0170,
         0.0213,  0.0158, -0.0211,  0.0114, -0.0284, -0.0521, -0.0850,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2718e-02,  3.9519e-01,  1.4434e-01,  3.6204e-02,  1.4533e-02,
         7.0563e-03, -5.6749e-03, -1.8950e-02, -7.9201e-03,  8.6894e-04,
         2.1719e-02,  1.3915e-02,  1.7329e-02, -1.3329e-04, -2.5269e-03,
         1.4087e-02,  3.9375e-03,  9.1055e-03, -4.2135e-03, -1.4931e-02,
         1.8604e-03,  7.1070e-04, -6.3502e-03,  1.1295e-02,  1.4825e-02,
        -2.1664e-02, -7.0282e-03,  1.5432e-03,  1.9020e-02,  1.1275e-02,
         2.8793e-03, -1.1324e-03,  7.9253e-03,  2.0618e-02,  1.3564e-02,
         2.5227e-02,  7.4553e-03,  1.5465e-03, -1.4741e-03,  1.4746e-03,
         2.7559e-03,  6.5308e-03, -6.7025e-03, -7.6317e-04,  9.1999e-03,
        -2.8792e-03,  4.6945e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0332,  0.3850, -0.0703, -0.0062, -0.0528, -0.0500,  0.0123,  0.0193,
         0.0053, -0.0132,  0.0175, -0.0042,  0.0163,  0.0065,  0.0053, -0.0025,
         0.0093, -0.0370,  0.0081,  0.0011,  0.0118, -0.0058,  0.0074, -0.0054,
        -0.0065,  0.0062, -0.0131,  0.0031, -0.0156,  0.0107, -0.0020, -0.0006,
        -0.0142,  0.0031, -0.0014, -0.0038, -0.0112, -0.0047, -0.0081,  0.0163,
        -0.0214,  0.0722,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1205, -0.4302, -0.0319, -0.0125,  0.0911, -0.0098, -0.0099, -0.0368,
         0.0297,  0.0237,  0.0276,  0.0269, -0.0178,  0.0029, -0.0095, -0.0025,
         0.0328, -0.0167,  0.0132,  0.0185,  0.0152,  0.0082,  0.0036,  0.0049,
         0.0037,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1124, -0.5006, -0.0015, -0.0511,  0.0052, -0.0068, -0.1196, -0.0041,
        -0.0008, -0.0048, -0.0080,  0.0055, -0.0091, -0.0137, -0.0099, -0.0184,
        -0.0127, -0.0430, -0.0044, -0.0165,  0.0207,  0.0049, -0.0265,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0862, -0.5069, -0.0364, -0.0158, -0.0545,  0.0312, -0.0134,  0.0452,
        -0.0186, -0.0311,  0.0360,  0.0068, -0.0180, -0.0185,  0.0066, -0.0093,
         0.0089, -0.0008,  0.0190,  0.0010,  0.0072,  0.0177,  0.0110,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0084, -0.4638,  0.1154,  0.1414,  0.0748,  0.0027, -0.0121, -0.0108,
         0.0243,  0.0281, -0.0319,  0.0283,  0.0580,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1840, -0.4105, -0.0318,  0.0075,  0.0422,  0.0053,  0.0092,  0.0008,
         0.0230,  0.0736,  0.0422,  0.0499,  0.1176, -0.0023,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1511, -0.3978,  0.0493,  0.0785, -0.0299, -0.0196, -0.0339, -0.0347,
        -0.0108,  0.0692, -0.0040, -0.0267, -0.0024,  0.0005, -0.0259,  0.0081,
         0.0052,  0.0252,  0.0271,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-0.2136,  0.2923, -0.0478,  0.0452,  0.0628, -0.0010, -0.0111,  0.0087,
        -0.0232,  0.0033, -0.0053, -0.0097, -0.0034, -0.0227,  0.0219, -0.0124,
         0.0089,  0.0010, -0.0217, -0.0018,  0.0126,  0.0139, -0.0041, -0.0097,
         0.0902,  0.0516,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0133,  0.5765,  0.0538,  0.0010, -0.0560,  0.0323, -0.0235, -0.0436,
        -0.0204, -0.0293, -0.0387,  0.0036, -0.0220,  0.0029, -0.0744,  0.0086,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0797,  0.6082,  0.0150,  0.0760, -0.0097, -0.0300,  0.0184,  0.0253,
        -0.0504,  0.0317, -0.0019, -0.0031,  0.0015, -0.0007, -0.0070,  0.0337,
         0.0075,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.2013,  0.5251,  0.0072, -0.0483, -0.0105,  0.0164, -0.0296, -0.0195,
         0.1421,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0831, -0.4096, -0.1218, -0.0207,  0.0265,  0.0052,  0.0122,  0.0237,
         0.0099,  0.0081,  0.0071, -0.0070, -0.0013, -0.0083, -0.0120,  0.0022,
        -0.0013,  0.0013, -0.0006, -0.0116, -0.0062, -0.0082, -0.0153, -0.0054,
         0.0009,  0.0054, -0.0147, -0.0060, -0.0112,  0.0040, -0.0013, -0.0036,
         0.0020,  0.0089, -0.0144,  0.0017,  0.0097, -0.0073,  0.0073, -0.0069,
         0.0076,  0.0032, -0.0096, -0.0232, -0.0196,  0.0231], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0537,  0.3329, -0.1167,  0.0018,  0.0961,  0.0439,  0.0265, -0.0096,
         0.0059,  0.0122,  0.0023, -0.0043, -0.0096,  0.0164,  0.0121,  0.0554,
         0.0060,  0.0311,  0.0057,  0.0064,  0.0062,  0.0041,  0.0373, -0.0058,
        -0.0122, -0.0026,  0.0397,  0.0107,  0.0120, -0.0208,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0598, -0.6703, -0.0296, -0.0498,  0.0314,  0.0765, -0.0827,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.2568, -0.4636, -0.0905,  0.0457,  0.0140,  0.0172, -0.0215,  0.0137,
         0.0196,  0.0543,  0.0032,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0164,  0.2714,  0.0903, -0.0720, -0.0734,  0.0850,  0.0088, -0.0710,
         0.0503, -0.0053, -0.0067,  0.0068,  0.1014, -0.1413,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5293e-02, -5.3797e-01, -8.9201e-02, -9.1697e-03,  6.3226e-03,
         2.6957e-02,  6.1211e-03, -7.0945e-03,  2.5036e-02,  1.1379e-03,
         4.0316e-03,  1.9896e-02,  8.7439e-03,  1.5441e-02, -8.7065e-04,
         1.4852e-02,  2.3080e-02,  6.2060e-03, -4.3338e-03, -4.8465e-03,
        -1.4442e-02,  5.3708e-03,  2.2892e-02, -2.4633e-03,  2.6282e-03,
         2.2344e-02, -1.6229e-03,  2.7499e-04,  8.7897e-03, -1.5421e-02,
        -1.0968e-02,  2.0896e-02, -1.0710e-02, -2.4574e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3174e-01, -3.1858e-01, -2.5664e-02, -8.5379e-03,  1.5270e-02,
        -8.7732e-03, -1.5021e-02,  7.6743e-03, -1.9912e-02,  5.1063e-02,
        -1.3226e-02, -7.3239e-03, -1.1826e-03, -1.0973e-02, -5.4773e-02,
         1.2446e-02, -1.7078e-02,  2.5921e-02,  3.0360e-02,  2.7270e-04,
        -1.0641e-02, -1.5047e-03,  1.3231e-03,  2.0036e-03, -1.1612e-02,
         3.2156e-03,  7.4044e-02,  1.9867e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0083, -0.5587,  0.0144,  0.0524,  0.0119,  0.0058,  0.0076,  0.0169,
         0.0292,  0.0035,  0.0177,  0.0187,  0.0473, -0.0238,  0.0125,  0.0096,
         0.0114,  0.0018,  0.0356,  0.0091,  0.0188, -0.0052, -0.0306, -0.0490,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 9.7899e-02, -3.9503e-01, -1.8111e-02, -2.1353e-02,  5.7842e-03,
         1.6812e-02,  1.9084e-02, -1.1200e-02, -2.8901e-02, -3.4920e-03,
         5.5572e-03,  1.4171e-03,  2.7384e-02, -1.2766e-04, -7.7638e-04,
         2.9235e-02, -6.1944e-02,  2.5337e-02,  4.6068e-02,  2.8568e-02,
        -1.2357e-02,  5.8396e-03, -1.7513e-02, -1.4188e-02, -3.6163e-03,
        -3.0883e-02,  1.6526e-03, -1.0242e-02, -5.4062e-03, -4.8642e-03,
         4.9356e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.2420, -0.3724,  0.0750, -0.0057, -0.0088, -0.0681, -0.0368, -0.0226,
        -0.0325, -0.0098, -0.0572,  0.0690,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.2974, -0.3143, -0.0456,  0.1238,  0.0047, -0.0081, -0.0161,  0.0035,
        -0.0049,  0.0306, -0.0096,  0.0106, -0.0093,  0.1038,  0.0177,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0141, -0.4346,  0.0323, -0.0317, -0.0454, -0.0098, -0.0445, -0.0044,
         0.0047, -0.0292, -0.0192,  0.0240, -0.0139,  0.0037, -0.0055, -0.0096,
        -0.0382,  0.0165,  0.0029,  0.0178, -0.0022,  0.0180, -0.0008,  0.0298,
        -0.0212, -0.0014,  0.0083,  0.0119,  0.0039,  0.0206, -0.0031, -0.0314,
        -0.0457], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1546,  0.3596, -0.1679, -0.0038, -0.0008,  0.0163,  0.0039, -0.0038,
         0.0097,  0.0196,  0.0109,  0.0173,  0.0027, -0.0112, -0.0235, -0.1942,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0637,  0.6089, -0.0429,  0.1160, -0.0454, -0.0421,  0.0169, -0.0177,
         0.0463,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1018,  0.6934, -0.0770,  0.0735, -0.0543,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.7270e-02, -2.9382e-01, -3.4462e-02,  5.4691e-02,  3.6842e-02,
         4.2787e-02,  2.3592e-02,  1.1458e-02,  1.1463e-02, -6.7265e-03,
        -6.9877e-02, -3.6867e-02,  1.2211e-04, -4.6853e-03, -1.1854e-02,
        -2.1409e-03, -5.8318e-03, -3.0240e-02, -2.6309e-02, -3.7854e-02,
        -1.2120e-02, -7.4452e-03,  6.4830e-03,  2.1349e-02, -1.5748e-02,
        -8.1484e-03, -8.9811e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.2449, -0.2512, -0.0769, -0.0823,  0.0098, -0.2335,  0.1014,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0741, -0.5306, -0.0118,  0.0047,  0.0063, -0.0075,  0.0029, -0.0077,
         0.0157,  0.0131,  0.0012, -0.0102, -0.0086,  0.0023, -0.0224, -0.0038,
        -0.0269,  0.0162,  0.0250, -0.0288,  0.1414,  0.0390,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0798,  0.5055, -0.0132,  0.0162, -0.0513,  0.0300, -0.0077, -0.0864,
         0.0085, -0.0131,  0.0350,  0.0216, -0.0219, -0.0271, -0.0035, -0.0137,
        -0.0154,  0.0220, -0.0281,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8429e-02, -5.3491e-01,  4.2138e-02,  1.2281e-02,  3.2658e-04,
        -5.8139e-03, -6.5935e-03, -1.0890e-02, -2.6740e-02,  1.5347e-02,
         1.0267e-02, -2.0741e-03, -7.1004e-04, -2.8937e-02, -5.2019e-03,
         2.7516e-03, -1.4126e-02,  3.3039e-02, -5.8546e-03,  4.9732e-03,
         1.7933e-02, -2.5230e-02,  3.8729e-03,  6.5019e-03, -6.8825e-03,
         5.5907e-03,  3.3858e-02, -3.0904e-02,  2.2097e-02,  3.1542e-02,
         3.4177e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-4.1919e-02,  4.8570e-01, -1.6342e-02, -5.4954e-02, -1.3041e-02,
         1.9770e-02, -9.2101e-03, -3.4312e-03, -3.0916e-02, -1.5025e-02,
         9.1347e-03,  1.1059e-02, -2.4815e-02, -1.9957e-02,  3.9333e-02,
         2.2574e-02, -6.9267e-02,  1.0917e-02, -9.2099e-04,  2.2732e-02,
         3.5302e-04, -3.3627e-03,  1.6476e-03, -1.0886e-02, -6.4218e-03,
         9.3394e-03,  2.2213e-04, -8.5731e-03, -4.5539e-03,  1.8425e-02,
        -9.9897e-03, -5.2094e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1360, -0.2678, -0.1075, -0.0493, -0.0138, -0.0301, -0.0587, -0.0249,
        -0.0224, -0.0095, -0.0055,  0.0049,  0.0274, -0.0879,  0.0505, -0.0264,
         0.0126, -0.0150,  0.0005, -0.0205, -0.0006,  0.0282,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0797,  0.2765,  0.0438, -0.0326,  0.0047,  0.0431, -0.0646,  0.0189,
         0.0039, -0.0140, -0.0020,  0.0040,  0.1281,  0.0099,  0.0169, -0.0127,
        -0.0219,  0.0120,  0.0158, -0.0236,  0.0154, -0.0048,  0.0099,  0.0034,
        -0.0058, -0.0131, -0.0023,  0.0316,  0.0227, -0.0623,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0795, -0.5422, -0.1183, -0.0403,  0.0229,  0.0095, -0.0009, -0.0051,
        -0.0079,  0.0161, -0.0029, -0.0287,  0.0157, -0.0182, -0.0007,  0.0170,
        -0.0393, -0.0347,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0205, -0.2674,  0.0981,  0.0646,  0.1251, -0.0273, -0.0496, -0.0497,
         0.0218,  0.0080, -0.0104,  0.0128, -0.0083, -0.0109, -0.0032,  0.0011,
        -0.0329,  0.0041, -0.0065,  0.0116,  0.0122,  0.0013,  0.0044, -0.0025,
        -0.0103,  0.0019, -0.0084, -0.0023, -0.0093,  0.0043, -0.0063, -0.0193,
        -0.0071, -0.0205, -0.0075, -0.0418,  0.0067,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.9470e-02,  4.8356e-01, -2.9201e-02,  1.2677e-01, -2.0531e-03,
        -1.0474e-02,  8.0466e-03, -2.8249e-02, -8.2772e-03,  2.9817e-03,
         8.9689e-03,  1.2959e-03, -5.6788e-03, -1.9073e-02, -5.5849e-03,
         6.1349e-05,  2.0665e-02, -5.3729e-03,  1.5052e-02, -1.2355e-02,
        -3.3870e-02,  1.2534e-02, -6.1989e-02,  1.8415e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.6730e-02, -3.5147e-01,  4.2307e-02,  4.4420e-02, -2.3254e-02,
         2.4808e-02, -3.7076e-02, -5.7437e-02,  1.4832e-02, -4.9287e-04,
         3.3580e-04, -9.3103e-03,  3.6710e-04, -1.0699e-02, -2.6443e-03,
        -6.6985e-03, -2.5560e-03,  3.8028e-03,  1.0753e-02,  8.9378e-03,
         3.6808e-03,  4.9809e-03,  2.3641e-02, -1.5890e-02, -2.3384e-02,
        -4.3406e-02, -2.9235e-04, -1.9647e-03,  5.6456e-03, -5.1033e-03,
        -1.4207e-02, -2.6131e-03,  1.3348e-04, -5.4209e-03,  2.0589e-03,
         2.2614e-02,  4.7542e-03,  5.2269e-03, -3.6128e-03, -6.5552e-04,
         2.5929e-02,  2.8494e-03,  4.3543e-03, -1.6435e-02,  4.7673e-02,
        -4.5448e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0867, -0.4884,  0.0074, -0.0195,  0.0698, -0.0195, -0.0025, -0.0083,
        -0.0114,  0.0058,  0.0088, -0.0113, -0.0353,  0.0043, -0.0089, -0.0121,
         0.0218, -0.0230, -0.0179,  0.0057,  0.0100,  0.1000,  0.0215,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1424, -0.6680, -0.0491, -0.0060,  0.0401,  0.0103, -0.0362,  0.0159,
         0.0055,  0.0097, -0.0169,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1278,  0.4706,  0.1271,  0.0076, -0.0181,  0.0811,  0.0497, -0.0008,
         0.0086, -0.0297,  0.0220,  0.0153, -0.0214,  0.0201,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1576,  0.3112,  0.0362, -0.1400,  0.0523, -0.0173,  0.0365, -0.0370,
        -0.0047, -0.0650,  0.0727,  0.0692,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1860e-02, -3.9395e-01,  3.1024e-02,  1.1400e-01,  1.6744e-02,
        -3.2383e-02,  5.4892e-03, -2.6377e-03, -9.1400e-03,  1.0215e-02,
        -2.3006e-03,  2.1473e-02, -4.1998e-03, -4.1085e-03,  7.4170e-03,
         8.0887e-04,  1.5667e-02, -1.1677e-02,  1.4998e-03,  1.4184e-02,
        -2.5564e-03,  1.7502e-03,  5.0646e-03, -2.1386e-02, -2.2268e-03,
        -2.1718e-02, -2.5228e-03,  7.4651e-03, -2.4281e-03, -7.5079e-03,
        -2.2445e-03,  2.1886e-02, -8.7829e-03,  9.4373e-04, -5.3528e-03,
         5.3949e-03,  3.3601e-04, -7.4294e-03,  2.5372e-03, -6.5679e-03,
        -3.9713e-03,  2.4599e-03, -3.8138e-03,  4.0877e-03,  8.1658e-03,
         2.8606e-03, -3.6218e-04,  5.0127e-03, -1.7278e-03, -7.7793e-03,
        -3.2422e-03,  1.0829e-02, -5.2382e-03, -2.1568e-02, -2.2006e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-2.4896e-01,  3.1028e-01,  4.5443e-02, -3.0453e-02, -1.3199e-01,
        -1.8751e-02,  5.5622e-02,  1.0157e-02, -5.9057e-03,  8.9198e-05,
         7.3371e-03, -3.6192e-03,  4.0677e-03, -1.9351e-02,  1.1854e-02,
        -7.4269e-03, -1.5517e-02, -8.2448e-03, -2.3867e-03,  6.6484e-05,
        -2.0679e-03,  2.3470e-02,  3.6939e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.1270e-02,  4.9152e-01,  9.5267e-02,  1.7146e-02, -1.7211e-02,
         1.3509e-02,  2.6473e-05,  1.5265e-02, -4.4458e-02, -2.8983e-03,
        -3.1571e-02,  2.2140e-02, -2.3666e-03,  1.7874e-02, -1.4141e-02,
        -1.4598e-02, -8.5488e-02, -6.3250e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.2275,  0.3306, -0.0321, -0.0791,  0.0063,  0.0237,  0.0632,  0.0361,
        -0.0090, -0.0280,  0.0068, -0.0010,  0.0667, -0.0180, -0.0110,  0.0027,
        -0.0487,  0.0097,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1179,  0.4142,  0.0238, -0.0284,  0.0143,  0.0433, -0.0399, -0.0742,
        -0.0219, -0.0091,  0.0087,  0.0533,  0.0040,  0.0767,  0.0704,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0230,  0.4785,  0.0453,  0.0217,  0.0370, -0.0137,  0.0012,  0.0151,
         0.0089, -0.0071,  0.0359,  0.0040,  0.0180, -0.0044,  0.0385, -0.0422,
        -0.0048,  0.0243,  0.0137,  0.0056,  0.0006, -0.0086, -0.0048, -0.0127,
        -0.0104, -0.0173,  0.0415,  0.0122,  0.0116, -0.0058,  0.0083,  0.0233,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.7897e-02, -3.8952e-01,  1.0603e-02, -2.8042e-02,  5.3655e-02,
        -2.5815e-02, -4.2641e-03,  1.2319e-02,  2.3785e-02,  3.7245e-03,
         1.2989e-02, -1.1147e-02,  1.5030e-02, -1.4638e-02, -1.1244e-02,
         1.9353e-02, -1.5737e-02,  1.1749e-02, -3.2247e-02,  9.4179e-03,
         9.9735e-03, -2.6938e-02,  3.2800e-02, -1.3377e-02,  5.7538e-03,
         2.9888e-03, -8.8205e-05, -9.7452e-03,  1.8912e-02,  3.8735e-02,
         8.0477e-03,  2.3247e-03, -4.7136e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1860e-02,  5.2779e-01, -3.6010e-02,  5.8707e-02,  2.1353e-03,
        -1.2394e-02,  2.6356e-03, -1.6065e-02, -9.5638e-03, -1.2878e-02,
        -5.5407e-03, -1.0039e-02,  3.9342e-04, -9.3130e-03, -3.1639e-02,
         1.9497e-02,  1.2451e-02, -2.6905e-03, -3.4313e-04,  8.1422e-03,
        -3.9866e-03,  6.7329e-03, -3.5937e-03,  8.3479e-03, -1.1605e-02,
        -2.3646e-03,  7.5276e-03, -9.5978e-03,  3.4303e-03, -9.9946e-04,
         1.4848e-03,  3.9356e-03, -5.7790e-03, -6.3059e-03,  2.0038e-03,
        -3.2573e-03,  4.1760e-03, -1.0441e-03,  3.1146e-03, -1.2699e-03,
        -7.2548e-04, -3.2322e-03, -2.1886e-03, -6.0063e-03, -1.6290e-03,
         3.5042e-03,  1.3133e-03, -5.0848e-03, -5.8779e-03, -2.4331e-03,
         8.5358e-04,  3.6415e-03, -1.1892e-02,  2.2998e-03, -4.6167e-03,
        -2.4192e-03, -1.8564e-02,  2.3074e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0126, -0.5118,  0.0128,  0.0664,  0.0138,  0.0194,  0.0063, -0.0116,
        -0.0078, -0.0043,  0.0051, -0.0047,  0.0201,  0.0182,  0.0191, -0.0051,
         0.0060,  0.0186, -0.0057,  0.0062,  0.0017, -0.0119,  0.0270,  0.0041,
         0.0461,  0.1339,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1685, -0.4911, -0.0446, -0.0115,  0.0096, -0.0064,  0.0217, -0.0074,
         0.0051,  0.0219,  0.0667,  0.0071,  0.0020, -0.0099,  0.0114, -0.0030,
         0.0115,  0.0065, -0.0034, -0.0010,  0.0045,  0.0108,  0.0118, -0.0296,
         0.0331,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6614e-02, -3.4339e-01,  1.9631e-04, -2.9040e-03,  7.8000e-02,
         1.5723e-02,  2.7246e-03,  3.8135e-02,  2.7992e-02, -5.2249e-03,
         2.3059e-02,  2.7640e-02, -1.0419e-03, -4.4199e-02, -2.5677e-02,
        -2.7734e-02, -2.5651e-02, -2.0315e-02, -5.1767e-03, -1.5136e-02,
        -2.0318e-03, -1.5620e-02, -6.3411e-04, -1.1191e-02, -4.0283e-02,
         7.5933e-03, -4.6934e-03, -5.7988e-03, -4.1049e-04,  4.1615e-03,
        -1.3165e-02,  2.9238e-03,  1.1616e-02, -2.4402e-03,  3.9599e-03,
        -7.4076e-03, -2.4965e-02, -6.0059e-03, -2.6241e-03, -9.9774e-03,
        -5.6639e-03,  5.1785e-04, -2.8578e-03,  3.7041e-02,  1.9889e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0333, -0.4017, -0.0695, -0.0797, -0.0058, -0.0103, -0.0466, -0.0311,
         0.0169,  0.0110, -0.0134, -0.0094, -0.0118, -0.0122, -0.0170,  0.0179,
        -0.0208, -0.0527, -0.0040,  0.0111,  0.0057,  0.0340,  0.0843,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1128,  0.5472, -0.0959,  0.0075,  0.0061,  0.0537, -0.0634, -0.0080,
         0.0150,  0.0376,  0.0291,  0.0238,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 0.2366, -0.2919, -0.0992, -0.0120, -0.0222,  0.0210, -0.0845,  0.0046,
         0.0266, -0.0032, -0.0134, -0.0047,  0.0084,  0.0025, -0.0047, -0.0210,
         0.0105, -0.0403,  0.0161, -0.0022,  0.0089, -0.0339, -0.0316,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0082, -0.4374, -0.0729, -0.0443, -0.0847,  0.0042, -0.0262,  0.0243,
        -0.0083, -0.0038,  0.0179, -0.0088,  0.0055, -0.0031,  0.0506, -0.0014,
        -0.0034, -0.0136,  0.0077,  0.0045,  0.0026,  0.0013, -0.0225,  0.0154,
         0.0119,  0.0090,  0.0046, -0.0025,  0.0063,  0.0064,  0.0021, -0.0489,
        -0.0354], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1337,  0.4735,  0.0140,  0.0249,  0.0175, -0.0325,  0.0370,  0.0347,
         0.0072, -0.0104,  0.0225,  0.0088, -0.0234,  0.0564,  0.0009, -0.0084,
        -0.0458,  0.0225,  0.0261,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1065, -0.5723,  0.0388,  0.0142, -0.0276, -0.0518,  0.0338,  0.0097,
         0.0192, -0.0446, -0.0129, -0.0159,  0.0121, -0.0408,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0472,  0.6438,  0.0349, -0.0056,  0.0263, -0.0216,  0.0074,  0.0207,
         0.0067, -0.0046,  0.0427,  0.0096, -0.0093,  0.0315, -0.0035, -0.0124,
         0.0081,  0.0075,  0.0062, -0.0072,  0.0169, -0.0263,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0581,  0.4637,  0.0659, -0.0438,  0.0367,  0.0220,  0.0131, -0.0359,
        -0.0141, -0.0040,  0.0192, -0.0140, -0.0321, -0.0057, -0.0076, -0.0128,
        -0.0173, -0.0170,  0.0168, -0.0078, -0.0173, -0.0013, -0.0055,  0.0103,
        -0.0063,  0.0049,  0.0005,  0.0127, -0.0293, -0.0041,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0571, -0.1600, -0.1573, -0.1037, -0.1301, -0.0478,  0.1026,  0.0011,
         0.0192,  0.0069, -0.0179, -0.0111, -0.0186,  0.0025,  0.0337, -0.0811,
         0.0492,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0421,  0.5753, -0.0286,  0.1181, -0.0056, -0.0019,  0.0347,  0.0685,
        -0.0868, -0.0108, -0.0275,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.2804,  0.3661,  0.0561, -0.0082, -0.0265, -0.0380,  0.0243, -0.0032,
        -0.0012,  0.0192,  0.0330,  0.0098,  0.0035,  0.0611, -0.0248, -0.0114,
         0.0331,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3634e-02,  4.8405e-01,  6.8646e-02,  1.2418e-03, -1.1374e-02,
         1.7460e-02,  1.6153e-02, -3.6423e-03,  1.9326e-02,  2.4837e-02,
        -5.2975e-03, -3.0703e-03, -2.1241e-04,  4.4133e-02, -3.6195e-03,
         7.0362e-03,  2.2665e-02, -3.8479e-04,  4.5250e-02,  5.2900e-03,
        -3.5244e-03, -2.2837e-02, -2.4187e-02, -1.3283e-02, -2.3566e-02,
        -1.1528e-01,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.6886e-02,  5.5296e-01,  1.4924e-02, -3.3730e-03, -1.5499e-03,
        -1.4587e-03, -1.3080e-02,  3.9448e-02,  3.3415e-02,  2.4997e-04,
        -3.1274e-02, -6.0346e-03,  6.0357e-03,  9.7165e-03, -3.2843e-03,
        -1.0085e-02, -8.9391e-03, -1.4250e-02, -2.3229e-03, -7.2592e-03,
        -3.4321e-02, -4.5135e-03,  2.8469e-03,  1.5481e-02, -6.9581e-03,
        -1.1134e-03, -3.3267e-03, -8.0317e-03, -5.8238e-02, -2.8621e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0173, -0.4783, -0.0542,  0.0269,  0.0555,  0.0065, -0.0230, -0.0221,
        -0.0290,  0.0022, -0.0047,  0.0249,  0.0115, -0.0216, -0.0562, -0.0275,
         0.0297, -0.0266, -0.0822,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 1.0773e-01,  4.9729e-01,  9.9458e-03,  2.7004e-02,  4.8509e-02,
         1.7525e-04,  4.4208e-02, -7.3845e-04, -3.4148e-03,  9.2539e-03,
         3.4035e-03, -1.5845e-03,  5.7426e-03,  1.0761e-02, -1.7254e-02,
        -1.9361e-03,  9.6298e-03,  2.9463e-03, -5.9623e-03,  1.5645e-03,
        -1.1302e-02, -3.9931e-03, -1.2581e-02,  1.0267e-02, -2.4059e-03,
        -7.3972e-02, -7.6423e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.3107, -0.2224, -0.0077, -0.0082, -0.0007, -0.0167, -0.0400,  0.0054,
        -0.0964,  0.0467,  0.0850, -0.0374,  0.0710,  0.0517,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0306,  0.4707,  0.0483,  0.0170,  0.0086,  0.0047, -0.0184,  0.0161,
        -0.0482,  0.0177,  0.0045, -0.0438,  0.0150, -0.0022,  0.0135, -0.0281,
         0.0145,  0.0034,  0.0322,  0.0134, -0.0060,  0.0171, -0.0146, -0.0128,
         0.0295, -0.0689,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1549, -0.4755,  0.0739,  0.0135,  0.0340, -0.0478, -0.0012,  0.0052,
         0.0058, -0.0723,  0.0232,  0.0674,  0.0013,  0.0239,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0503e-01,  4.7558e-01,  6.6774e-02,  7.0509e-03,  2.3904e-02,
        -1.0496e-02,  2.0389e-02, -2.9972e-03, -7.0806e-03, -5.8548e-04,
         2.3798e-03, -2.6907e-02,  1.8828e-03, -5.6452e-02,  1.4753e-03,
        -3.8101e-04, -3.7430e-02,  1.1393e-03, -6.9108e-03, -3.2874e-03,
        -1.5789e-02, -2.9140e-02, -2.1743e-02, -2.6971e-03, -5.3880e-03,
        -8.9277e-03, -3.6638e-03, -7.4087e-03,  2.8125e-03, -4.6585e-03,
        -3.0290e-02,  9.3466e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.1640e-03,  5.2547e-01,  8.0880e-02,  1.6308e-03, -2.5134e-02,
         9.4480e-04, -2.9637e-02, -5.1972e-03,  2.6372e-02, -1.1841e-02,
         8.8590e-04, -4.1217e-03,  4.6821e-03, -2.2692e-03,  1.6244e-02,
         1.4883e-04, -2.1203e-02, -8.2694e-03, -8.6503e-03, -6.3671e-03,
         3.1633e-03, -4.3019e-03,  7.5379e-03,  1.3117e-01, -6.5711e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0180, -0.6013, -0.1719,  0.0651,  0.0165,  0.1273,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.2773e-02,  4.9538e-01, -1.2613e-03, -3.1434e-02, -4.4864e-02,
         8.7587e-03, -2.0555e-03,  3.0082e-02,  2.3718e-03, -2.3826e-02,
         1.0209e-02,  2.2347e-02,  7.6590e-03,  1.0109e-02, -1.3119e-02,
         1.7769e-02,  3.6024e-03,  1.4055e-02, -3.5187e-03, -7.4363e-03,
        -6.6377e-03,  2.0523e-03, -5.5145e-03, -1.3648e-02, -1.4435e-02,
        -1.0985e-02,  1.7280e-03, -1.4435e-02,  9.4031e-03,  4.3078e-03,
         1.5819e-02,  1.3174e-02,  1.4216e-02, -1.1468e-03, -6.3111e-03,
        -2.2661e-04,  3.3326e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7753e-04, -5.1503e-01,  4.0591e-02, -7.2308e-02,  1.6941e-02,
        -9.6540e-03, -4.5421e-02,  3.2044e-02, -1.4014e-02, -8.4637e-03,
         4.1182e-03,  2.6424e-03,  7.9093e-03,  2.0291e-02,  1.0859e-02,
         4.9909e-03,  1.8393e-02,  5.6157e-03,  1.3079e-02,  6.2991e-03,
         2.7424e-02,  6.0682e-03,  1.0058e-02, -6.2086e-03,  8.9885e-03,
         1.4293e-02, -1.3440e-02,  5.0026e-03,  1.8783e-03, -6.9312e-03,
         6.7229e-03,  5.1790e-04, -1.5256e-06,  3.8053e-02,  5.5661e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0382, -0.3915,  0.2058,  0.0048, -0.0259, -0.0166, -0.0266, -0.0094,
         0.0223, -0.0439, -0.0121, -0.0216,  0.0467, -0.0844,  0.0500,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7158e-02, -4.4247e-01, -3.0309e-02,  1.0314e-01, -1.3961e-02,
         1.1924e-02, -9.2411e-03, -1.1630e-03, -3.2500e-02,  6.2245e-03,
        -1.3286e-02,  1.1885e-03,  1.1021e-02,  4.7047e-03,  2.4801e-03,
        -1.2748e-02,  1.1329e-02, -2.2908e-02, -1.9432e-03, -1.5209e-02,
         2.2352e-03, -8.9942e-04, -2.0465e-02, -3.2483e-03, -1.1093e-02,
        -2.3801e-02, -4.2629e-03,  1.9196e-03, -9.3135e-03,  5.0789e-03,
        -1.9859e-02,  5.2707e-02, -2.0876e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9163e-02,  6.4153e-01,  4.6609e-02,  7.6937e-04, -9.8074e-03,
        -5.9019e-02, -8.1357e-04,  6.9442e-04,  3.0984e-04, -8.5072e-03,
         3.4005e-03, -7.5978e-04, -1.2146e-02, -2.7611e-02, -2.9396e-02,
        -1.9646e-03,  3.1009e-03, -9.0119e-04, -1.1152e-02, -9.1920e-03,
        -5.7894e-03,  9.0995e-04, -2.6126e-03, -5.7499e-03, -4.5241e-03,
         5.2744e-03, -2.5314e-02, -3.2445e-02, -2.0531e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([-0.1050, -0.4317,  0.0576,  0.0652,  0.0708,  0.0213, -0.0096, -0.0255,
        -0.0374,  0.0460, -0.0094, -0.0265,  0.0300, -0.0049,  0.0591,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8839e-02,  4.6174e-01,  3.6931e-02, -2.7794e-02, -1.2475e-02,
        -1.9263e-02,  1.9536e-02, -7.9550e-03,  5.2149e-02, -3.1367e-02,
        -1.5096e-02, -7.4454e-03, -1.6515e-02, -1.7901e-02, -1.8213e-03,
        -4.4465e-03, -1.9351e-02, -1.8413e-03, -8.9709e-04,  2.1514e-03,
        -1.2554e-03, -4.9525e-03,  2.1248e-03, -1.2431e-03, -3.2369e-03,
         4.4757e-03, -7.9178e-03, -1.2555e-02,  2.9735e-03,  2.6601e-03,
         1.0796e-03,  6.2714e-03, -2.8973e-04, -2.5299e-03,  2.6016e-03,
         6.8786e-03,  3.2074e-03, -1.7311e-03,  6.9347e-03, -2.9487e-03,
         3.0746e-03,  5.3660e-03,  2.0951e-02,  3.2314e-03, -6.3295e-03,
         9.4668e-03,  2.3005e-04, -1.8155e-02,  4.9815e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1250, -0.5011,  0.0553, -0.0080, -0.0012, -0.0253, -0.0155,  0.0120,
         0.0144, -0.0238,  0.0142, -0.0026,  0.0023,  0.0044,  0.0068,  0.0060,
        -0.0153, -0.0016, -0.0045, -0.0154, -0.0166, -0.0073,  0.0050,  0.0074,
         0.0047,  0.0044,  0.0204, -0.0245,  0.0010,  0.0112,  0.0067, -0.0006,
        -0.0045, -0.0265,  0.0043,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.2156e-03, -5.1845e-01,  4.2502e-02,  1.1700e-01,  2.6290e-02,
        -3.0579e-02,  9.5853e-03, -3.4319e-03,  8.9600e-05, -9.6690e-03,
        -7.4905e-03,  9.2978e-04,  3.8486e-03,  1.5502e-03,  1.8225e-03,
        -3.3007e-03, -5.1422e-03, -1.2071e-01,  9.1392e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0262,  0.4839,  0.0998,  0.0138,  0.0380,  0.0148, -0.0082,  0.0041,
        -0.0323, -0.0364, -0.0349,  0.0296, -0.0197, -0.1539, -0.0043,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0740,  0.5513,  0.0258,  0.0436,  0.0081, -0.0106, -0.0331, -0.0185,
         0.0063, -0.0201, -0.0355,  0.0022, -0.0136, -0.0156, -0.0046, -0.0318,
         0.0006, -0.0165,  0.0274,  0.0032, -0.0339, -0.0236,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0716,  0.4274,  0.0574,  0.0278, -0.0101,  0.0040,  0.0249,  0.0237,
        -0.0169, -0.0271, -0.0141,  0.0415,  0.0198,  0.0098, -0.0188, -0.0016,
         0.0060,  0.0104,  0.0090, -0.0018, -0.0037,  0.0061,  0.0092,  0.0058,
        -0.0124,  0.0167, -0.0130,  0.0049, -0.0673, -0.0372,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6068e-02,  5.0193e-01, -1.1575e-01, -1.7189e-02, -7.5991e-02,
        -4.9288e-04, -6.8793e-02, -3.9720e-02,  3.1408e-02, -7.6401e-02,
         3.6260e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0601,  0.7285, -0.0037, -0.0041, -0.0190, -0.0216,  0.0142, -0.0008,
         0.0139,  0.0106, -0.0098,  0.0181,  0.0033,  0.0020,  0.0081,  0.0055,
         0.0077,  0.0041,  0.0031, -0.0085,  0.0057,  0.0477,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0696e-02, -5.4633e-01, -3.0677e-02, -1.7180e-02, -4.4871e-02,
        -1.7990e-02, -7.6240e-03,  2.2631e-02, -8.2273e-03,  6.7765e-04,
         4.6095e-03,  2.6329e-02,  3.3399e-03,  4.4800e-03,  1.6793e-02,
         2.2304e-02, -2.3515e-03,  1.2132e-03,  1.0727e-02,  2.1553e-04,
         6.6760e-03, -4.5421e-03, -8.2888e-03,  8.6505e-03,  6.9704e-03,
         1.7863e-02,  2.6105e-03,  1.2652e-02,  1.0500e-02,  1.0730e-02,
         3.7078e-02, -1.3807e-02, -2.0366e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0105, -0.4289, -0.1023,  0.0573,  0.0451, -0.0212, -0.0240,  0.0118,
        -0.0139,  0.0069, -0.0091, -0.0085,  0.0225,  0.0158,  0.0315, -0.0572,
         0.0052,  0.0020,  0.0037, -0.0057,  0.0067,  0.0100, -0.0006,  0.0047,
        -0.0021, -0.0098,  0.0383,  0.0449,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1038,  0.3629,  0.0591,  0.0042,  0.0349, -0.0595, -0.0126,  0.0306,
         0.0618, -0.0126, -0.0094, -0.0034, -0.0173,  0.0198,  0.0173,  0.0157,
        -0.0032, -0.0413,  0.0136, -0.0150, -0.0220,  0.0057,  0.0320,  0.0333,
         0.0089,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 1.7815e-02, -5.1888e-01,  6.1068e-02,  1.8205e-02,  4.6979e-02,
        -1.8451e-03,  1.2038e-02,  1.8167e-02,  1.4240e-02,  2.3125e-03,
        -1.5380e-02,  2.3632e-02,  1.5803e-03,  3.4250e-03, -1.5753e-03,
         2.4582e-02, -1.2646e-02, -9.3029e-03, -8.5762e-03, -1.8330e-02,
         1.4695e-02,  4.7545e-03, -2.5632e-02, -4.0015e-03,  1.0451e-02,
        -4.5841e-03,  7.2396e-03, -1.5510e-02,  8.2006e-03,  7.5345e-03,
         8.7994e-03,  1.9795e-03,  4.4870e-04, -1.2920e-03, -6.1471e-03,
         9.6327e-03,  6.6611e-03,  7.4809e-03, -7.2690e-03, -1.0153e-02,
        -6.9541e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0258,  0.4924, -0.0316, -0.0556,  0.0565,  0.0259,  0.0111, -0.0323,
        -0.0421, -0.0014,  0.0142, -0.0045, -0.0302,  0.0062,  0.0254, -0.0036,
        -0.0026, -0.0027, -0.0121, -0.0383, -0.0147, -0.0065, -0.0074, -0.0385,
        -0.0186,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4198e-02,  5.3830e-01,  4.8295e-02, -1.2218e-04, -2.5867e-02,
         1.0199e-02,  4.2466e-02,  2.8538e-02,  1.9269e-02, -6.7422e-02,
        -3.8687e-02, -2.1926e-03, -1.8143e-03, -7.0875e-05, -1.0397e-02,
         7.9472e-03,  3.8668e-03, -2.5574e-02, -4.7423e-02,  4.8171e-03,
        -5.2531e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.6300e-03,  5.2159e-01,  2.0990e-04,  1.9608e-02, -5.9346e-02,
        -1.2057e-03,  3.7425e-02, -2.1322e-02,  1.4403e-02,  8.1177e-04,
        -3.6802e-02,  8.2279e-03, -6.2725e-03, -1.4085e-03, -1.6031e-02,
        -1.5837e-02, -1.6298e-01, -6.9888e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1137, -0.4261, -0.0071, -0.0335,  0.0135,  0.0320,  0.0051,  0.0209,
        -0.0015,  0.0089, -0.0174,  0.0089, -0.0046, -0.0165, -0.0109, -0.0097,
         0.0086,  0.0312, -0.0131, -0.0387, -0.0110,  0.0107,  0.0036,  0.0040,
         0.0051, -0.0065, -0.0062,  0.0075, -0.0058,  0.0076, -0.0058, -0.0064,
        -0.0039, -0.0113,  0.0146, -0.0182, -0.0249, -0.0251,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0961,  0.4771,  0.0939, -0.0631,  0.0179, -0.0015, -0.0107, -0.0098,
         0.0127, -0.0380,  0.0212,  0.0011,  0.0150,  0.0186, -0.0130,  0.0132,
         0.0261, -0.0408,  0.0230, -0.0071,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0060, -0.6375,  0.0878, -0.0198, -0.0039, -0.0153, -0.0103, -0.0353,
        -0.0177,  0.0111,  0.0433,  0.0343, -0.0048, -0.0731,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0878, -0.0787, -0.0731, -0.0200, -0.0564,  0.0282, -0.0434,  0.0359,
         0.0557, -0.0286,  0.0065, -0.0017, -0.0115, -0.0073,  0.0210,  0.0103,
        -0.0071, -0.1243,  0.0454, -0.0349,  0.0280, -0.0018,  0.0138, -0.0296,
         0.0077, -0.0125, -0.0057, -0.0154,  0.0093,  0.0099,  0.0521,  0.0366,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1147, -0.5544,  0.0474, -0.0308, -0.0006,  0.0197, -0.0549, -0.0151,
        -0.0015,  0.0037,  0.0220,  0.0013, -0.0051,  0.0042, -0.0347,  0.0146,
        -0.0076, -0.0105, -0.0466,  0.0105,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0306, -0.4653, -0.0620,  0.0130,  0.0508, -0.0051,  0.0011,  0.0060,
         0.0176,  0.0077,  0.0196,  0.0055, -0.0062,  0.0014, -0.0008,  0.0142,
         0.0031, -0.0030, -0.0047, -0.0031,  0.0052, -0.0237, -0.0100,  0.0049,
         0.0054, -0.0146, -0.0022,  0.0225, -0.0215,  0.0022, -0.0111,  0.0153,
        -0.0083, -0.0025, -0.0023, -0.0053, -0.0020, -0.0029, -0.0031, -0.0017,
        -0.0097,  0.0158, -0.0093, -0.0109, -0.0153,  0.0371,  0.0144,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0099,  0.7066, -0.0556,  0.0049, -0.0038,  0.0282,  0.0024, -0.0195,
         0.0385,  0.0116, -0.0193, -0.0033, -0.0013, -0.0395, -0.0556,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.8889e-02,  3.7900e-01,  1.0665e-01,  6.5320e-04,  2.4367e-03,
        -5.6771e-03,  1.8311e-02, -6.3191e-03,  3.5526e-02, -2.4980e-03,
         1.2723e-02, -9.9491e-04, -5.8981e-03,  1.5650e-02,  3.5141e-03,
         9.1854e-03, -6.6717e-03, -5.9762e-03, -2.1156e-02,  1.1311e-02,
         1.4486e-02,  1.2129e-02, -1.3304e-02, -1.6756e-02,  6.9657e-03,
        -2.5125e-03,  2.5716e-02,  1.0829e-02, -2.4788e-03,  2.7902e-03,
        -1.0504e-02,  4.0748e-03,  1.8118e-02,  1.8352e-03,  3.1391e-03,
        -6.4992e-03,  1.7473e-03,  2.5918e-03, -5.5858e-03,  9.2044e-04,
         1.2168e-03,  1.0899e-03, -6.7070e-03, -7.0125e-03,  2.9040e-03,
         2.5937e-04, -1.3707e-02,  5.5076e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 0.0378,  0.5715,  0.0231,  0.0169,  0.0061,  0.0186, -0.0066,  0.0185,
         0.0262, -0.0036,  0.0127,  0.0435,  0.0136,  0.0053, -0.0032,  0.0042,
         0.0377, -0.0105, -0.0083,  0.0141,  0.0142, -0.0024,  0.0123,  0.0053,
         0.0204,  0.0634,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0192,  0.5771,  0.0139, -0.0562, -0.0147, -0.0109,  0.0228, -0.0158,
        -0.0120, -0.0022,  0.0058, -0.0287, -0.0237, -0.0097, -0.0040, -0.0081,
        -0.0284,  0.0080, -0.0858, -0.0271, -0.0258,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.6626e-02,  4.6472e-01,  2.5906e-02, -5.7057e-02, -7.2733e-03,
         3.4220e-02,  3.2921e-02, -1.0679e-02,  3.3943e-04,  4.9315e-03,
        -1.3189e-02, -6.1661e-03, -8.4673e-03,  5.4031e-03,  2.8489e-03,
        -2.9178e-03, -1.3008e-02, -1.3006e-02,  9.8009e-03, -1.2451e-03,
         2.0433e-03, -1.1021e-02, -5.7363e-03, -8.0629e-03, -1.6180e-02,
         7.2815e-03, -9.2173e-03, -1.4647e-02, -1.3412e-02, -8.4329e-04,
        -5.9851e-03, -1.2005e-02, -4.1693e-03, -2.0425e-02, -5.8837e-02,
        -9.4125e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1603,  0.3355, -0.0241,  0.0053,  0.0533,  0.0108, -0.0063, -0.0258,
         0.0261,  0.0224, -0.0308,  0.0141,  0.0434,  0.0133, -0.0003,  0.0098,
        -0.0018, -0.0158,  0.1020, -0.0987,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7172e-02, -5.2358e-01, -5.1325e-02, -3.2487e-02, -4.5042e-02,
         9.3576e-03, -1.6611e-02,  3.3052e-02,  7.1228e-02,  8.5503e-05,
        -3.4379e-02, -6.5778e-03,  1.0570e-03, -1.5606e-02,  4.0298e-03,
        -1.1686e-03,  1.4380e-02, -1.1250e-02,  1.9366e-02,  5.7657e-02,
        -1.4583e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0255,  0.5830,  0.1458, -0.0174,  0.0594, -0.0158,  0.0135, -0.0551,
         0.0220,  0.0023,  0.0037,  0.0417, -0.0148,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7501e-02,  4.9651e-01,  8.5422e-02,  1.9531e-02, -5.3393e-04,
        -2.5286e-02, -5.2679e-02, -3.0507e-03,  5.1013e-03,  1.1054e-02,
         1.8530e-02, -7.0713e-03, -1.9371e-03, -2.0830e-03,  1.3761e-02,
         2.8383e-02, -1.5475e-02, -2.1848e-02, -1.0080e-02,  1.0166e-02,
         3.0522e-04,  1.2889e-02,  1.0711e-02,  1.4947e-02,  7.4441e-03,
        -4.0407e-03, -7.5112e-03, -6.3539e-04,  6.7688e-03, -2.0441e-02,
         3.5794e-03,  4.3995e-03, -4.6639e-03, -1.0123e-03,  4.4838e-03,
        -3.9737e-02,  1.0427e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0396,  0.4780,  0.0304, -0.0586, -0.1114, -0.0170, -0.0063, -0.0208,
         0.0044, -0.0139, -0.0024, -0.0109,  0.0403, -0.0098, -0.0112, -0.0132,
         0.0007,  0.0179,  0.0091, -0.0148, -0.0250,  0.0067, -0.0170,  0.0407,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0148, -0.5031, -0.0245, -0.0325,  0.0319,  0.0027,  0.0489, -0.0377,
         0.0029,  0.0111, -0.0188, -0.0074, -0.0374, -0.0014,  0.0173,  0.0076,
        -0.0034,  0.0034,  0.0020, -0.0245,  0.0134,  0.0357,  0.0101, -0.0029,
         0.0098,  0.0048,  0.0085, -0.0101, -0.0038, -0.0071,  0.0257, -0.0347,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0473, -0.5526, -0.0886,  0.0030,  0.0094,  0.0634, -0.0188,  0.0065,
        -0.0125, -0.0184,  0.0118,  0.0043,  0.0100,  0.0009,  0.0096,  0.0125,
         0.0174,  0.0066,  0.0038,  0.0101, -0.0072,  0.0065,  0.0105, -0.0126,
        -0.0087,  0.0165,  0.0069,  0.0014,  0.0041,  0.0173,  0.0008,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0573,  0.4096,  0.1208,  0.0041, -0.0097, -0.0542, -0.0344,  0.0027,
         0.0012, -0.0187,  0.0273,  0.0065,  0.0471, -0.0172, -0.0060, -0.0112,
        -0.0025,  0.0013, -0.0247, -0.0225, -0.0588, -0.0253, -0.0369,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7544e-02, -4.9873e-01, -2.6718e-02, -7.9740e-02, -2.2502e-02,
        -6.9098e-03, -2.9923e-04, -3.7670e-02, -8.2708e-03, -3.1597e-03,
         2.1606e-02, -8.3587e-03,  1.1427e-03,  3.1952e-02,  6.2740e-03,
        -2.1943e-02, -5.4957e-03,  6.1297e-03, -1.4766e-02,  1.2568e-02,
         1.7366e-02,  4.8612e-03,  9.4594e-03,  8.7029e-03,  7.2019e-03,
        -6.8672e-03,  2.0289e-02, -1.3475e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 1.6205e-05,  6.1078e-01,  7.3503e-02,  2.5865e-03,  1.3222e-02,
         4.8015e-03, -2.5985e-03, -1.5376e-02, -1.5760e-02,  7.5624e-03,
        -2.4950e-02, -6.0709e-03, -7.9764e-03, -8.6852e-03, -1.3878e-02,
         6.4150e-03, -1.7238e-03, -1.7612e-02,  7.5613e-04, -2.2092e-02,
        -8.2232e-03,  9.3197e-03,  1.1535e-02, -3.1981e-02, -7.7322e-03,
         5.3830e-03, -1.7671e-02,  2.3575e-02,  2.8212e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.8205e-02,  5.4302e-01,  7.8092e-03, -5.6847e-03, -4.4722e-02,
         1.3786e-02, -3.2557e-04, -2.7889e-02, -6.1979e-02, -5.1150e-03,
        -1.7263e-02, -6.5770e-02, -5.1864e-03, -1.0035e-02,  3.0694e-02,
         3.6109e-03,  1.3728e-02,  2.4081e-02, -1.4523e-03,  1.9509e-02,
        -6.4393e-03,  2.8034e-02,  3.7157e-03,  1.9420e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5270e-02, -5.0404e-01, -4.8674e-03,  1.0067e-02, -3.0874e-02,
         4.7557e-02,  9.3136e-04, -5.7878e-04,  2.1870e-03,  1.4827e-02,
         3.1331e-02, -3.4979e-04,  7.4903e-03, -2.5782e-03,  6.9113e-04,
         6.5147e-03,  9.8180e-03,  1.0650e-02, -2.0479e-02,  4.2087e-03,
        -2.1100e-02, -8.8344e-03, -3.2726e-03, -4.8036e-03,  7.4307e-04,
        -4.8047e-03, -4.1600e-03, -1.8750e-03, -7.9845e-03, -1.8278e-02,
         7.6575e-03, -4.8822e-02,  3.2219e-02,  1.0013e-01], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0663,  0.5922,  0.0016, -0.0257, -0.0196, -0.0229,  0.0288,  0.0224,
         0.0212,  0.0411, -0.0118,  0.0176,  0.0044, -0.0131,  0.0036,  0.0031,
         0.0064,  0.0065, -0.0206, -0.0070,  0.0079, -0.0562,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1274, -0.3988, -0.0130,  0.0296,  0.0246, -0.0816, -0.0132,  0.0063,
        -0.0108,  0.0067, -0.0358, -0.0147, -0.0147, -0.0062,  0.0138,  0.0535,
        -0.0644,  0.0848,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0767, -0.4078, -0.0421, -0.0314, -0.0223, -0.0585, -0.0191,  0.0161,
        -0.0328, -0.0052, -0.0408, -0.0102, -0.0011, -0.0248, -0.0028, -0.0351,
        -0.0108,  0.0034,  0.0055,  0.0126, -0.0474,  0.0404, -0.0533,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0023,  0.5599, -0.0546,  0.0881, -0.0431, -0.0017,  0.0146,  0.0011,
        -0.0020,  0.0039,  0.0332,  0.0198,  0.0109,  0.0211, -0.0530, -0.0878,
        -0.0029,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0194,  0.5118,  0.1523,  0.0958,  0.0086,  0.0216,  0.0148, -0.0184,
         0.0099, -0.0072,  0.0031,  0.0084, -0.0024,  0.0110,  0.0161, -0.0013,
        -0.0933,  0.0046,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3009e-02, -5.6310e-01, -6.5608e-02,  1.7869e-02, -4.0976e-04,
         7.8433e-03,  1.8296e-02, -2.3311e-03,  3.1873e-03, -1.8137e-02,
         2.0343e-02,  8.6348e-03,  1.9128e-04,  2.2806e-02,  2.8100e-02,
         8.0791e-04,  1.7744e-02,  8.8184e-03, -1.9393e-02,  2.3845e-02,
         7.5288e-03, -8.7696e-03,  4.8397e-03,  4.6728e-03,  5.6641e-02,
        -3.7069e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4365e-02, -3.6959e-01, -1.0024e-02,  1.8289e-02, -2.5980e-03,
         2.8174e-02, -4.2245e-02,  6.6547e-03,  3.7686e-02,  2.1714e-02,
        -7.1511e-03,  3.3720e-03,  1.2524e-02,  2.3384e-02,  1.1927e-02,
         3.8964e-03, -3.2526e-02, -1.4742e-03, -2.9054e-02, -1.8766e-02,
        -1.5178e-02, -1.8066e-02, -5.6632e-03,  1.3633e-02,  2.4682e-04,
        -1.7730e-02, -8.0334e-03,  1.5684e-02,  2.2192e-02, -3.3104e-02,
         1.1506e-01,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0441,  0.5136,  0.0738, -0.0203,  0.0409,  0.0041, -0.0070,  0.0334,
         0.0091, -0.0394, -0.0569, -0.0143, -0.0109, -0.0156, -0.0247,  0.0072,
        -0.0545, -0.0303,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0824,  0.6218,  0.0287, -0.0341, -0.0361, -0.0411,  0.0832,  0.0725,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 0.0817,  0.5680,  0.0414, -0.0191,  0.0107,  0.0020,  0.0176,  0.0076,
         0.0020, -0.0071, -0.0080, -0.0125, -0.0012, -0.0079, -0.0068, -0.0071,
         0.0018, -0.0103, -0.0016, -0.0039, -0.0125, -0.0189, -0.0155, -0.0102,
        -0.0031,  0.0015, -0.0129,  0.0033, -0.0046,  0.0056,  0.0187,  0.0007,
        -0.0024,  0.0027,  0.0177,  0.0061, -0.0156, -0.0159,  0.0135,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.5653e-03,  5.6036e-01,  1.3507e-02,  1.4374e-02,  7.8308e-03,
         1.9042e-02,  7.0826e-03, -3.7635e-04,  2.4289e-02,  3.9477e-02,
         2.9150e-02, -8.0074e-03, -5.1327e-02, -8.5526e-03, -2.5647e-02,
         2.4365e-02,  2.3876e-03,  3.0068e-04, -4.5367e-03,  8.6253e-03,
         2.0941e-02, -3.4734e-02,  1.5111e-02,  4.3108e-03, -9.9263e-03,
         8.7340e-03, -3.3239e-02,  1.5202e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0857,  0.5279,  0.0891, -0.0184,  0.0119, -0.0206,  0.0479,  0.0290,
         0.0030,  0.0267, -0.0197, -0.0871, -0.0329,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0834,  0.6406,  0.0387, -0.0589, -0.0157, -0.0014, -0.0232, -0.0103,
         0.0108, -0.0139, -0.0077, -0.0125, -0.0093, -0.0507, -0.0228,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1946,  0.3196,  0.0422, -0.0080,  0.0374, -0.0124, -0.0285, -0.0052,
        -0.0076,  0.0051, -0.0048, -0.0034,  0.0049,  0.0221,  0.0158, -0.0010,
        -0.0085, -0.0038,  0.0005, -0.0045,  0.0039,  0.0024,  0.0078,  0.0046,
        -0.0129, -0.0062,  0.0018,  0.0019,  0.0183, -0.0012, -0.0026,  0.0058,
        -0.0228,  0.0086,  0.0103,  0.0004,  0.0154,  0.0099,  0.0037,  0.0048,
         0.0004,  0.0050, -0.0077,  0.0025, -0.0099,  0.0152, -0.0843],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2655e-01,  4.9455e-01, -8.3584e-05, -1.6385e-02, -7.3361e-02,
        -2.1924e-02, -7.4382e-03,  2.1488e-02,  7.9561e-03, -7.9886e-03,
        -4.2963e-03,  1.2882e-03, -6.1158e-04, -1.0615e-02,  5.8614e-03,
        -4.5452e-03,  8.3980e-03, -9.9794e-03, -4.2435e-03, -7.0535e-04,
        -3.2061e-03,  2.5066e-04, -1.3397e-02, -3.6611e-03, -2.6695e-03,
        -4.8107e-03,  9.8294e-03,  9.3111e-03, -2.8669e-03, -2.8309e-02,
        -3.6535e-04,  3.3276e-03, -1.4212e-02, -1.2033e-03, -5.4729e-03,
        -7.9764e-03,  4.3883e-03, -1.9717e-03, -3.2350e-03, -2.1236e-02,
        -1.7199e-02,  1.2834e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8188e-02,  6.0800e-01,  5.6264e-02,  7.1556e-03,  2.4794e-03,
         3.9066e-02,  2.3061e-02,  3.0837e-02, -2.6839e-03,  2.0644e-03,
         1.8728e-02, -2.4777e-02,  2.1237e-02,  1.9205e-02,  5.9111e-03,
         2.1850e-02,  6.0859e-03,  5.7507e-03,  3.5315e-03, -2.0664e-04,
        -8.9682e-03, -2.1842e-02, -7.9245e-03, -1.4693e-02,  9.4862e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1826,  0.2026, -0.0046,  0.0192,  0.0025, -0.0010,  0.0294, -0.0043,
        -0.0090,  0.0103,  0.0027,  0.0005, -0.0032,  0.0116, -0.0239,  0.0100,
         0.0079, -0.0207, -0.1513,  0.0495,  0.1408, -0.0796, -0.0329,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1185,  0.4864,  0.0120,  0.0079, -0.0640,  0.0192,  0.0051, -0.0137,
        -0.0052,  0.0770,  0.0363, -0.0170, -0.0060,  0.0167, -0.0068, -0.0010,
        -0.0147,  0.0168, -0.0057, -0.0181,  0.0017, -0.0399, -0.0104,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1079, -0.5403, -0.0451, -0.0381, -0.0030,  0.0078, -0.0474, -0.0671,
        -0.0298, -0.0038, -0.0122,  0.0518,  0.0458,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1925,  0.5462,  0.0197, -0.0260, -0.0182,  0.0029, -0.0298, -0.0206,
        -0.0233,  0.0185, -0.0096, -0.0089, -0.0731, -0.0107,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0446, -0.6733,  0.0314,  0.0320, -0.0146, -0.0194,  0.0033, -0.0231,
        -0.0150, -0.0043,  0.0126, -0.0094, -0.0013, -0.0038, -0.0026, -0.0187,
         0.0432,  0.0310,  0.0163,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 2.1106e-01,  5.3780e-01, -1.8527e-02,  1.7984e-02,  1.2765e-02,
        -1.4502e-03,  8.6422e-03,  6.4194e-03, -6.8263e-03,  1.1951e-02,
        -1.3590e-02, -1.0611e-02, -5.7699e-03, -5.6663e-03,  1.4565e-02,
         2.3989e-03,  2.3195e-03, -1.1993e-02, -9.1112e-03, -4.4327e-03,
        -2.8457e-03, -5.1168e-03,  4.2919e-04, -5.7895e-03,  4.6468e-02,
        -2.5465e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2260e-01, -5.3001e-01, -8.7069e-02,  5.0315e-02, -5.4410e-02,
        -2.0844e-02, -5.7129e-03,  4.4565e-03,  8.8324e-03,  2.0238e-02,
         1.4403e-02,  1.2170e-03,  2.4162e-04, -3.7129e-02, -1.6250e-02,
        -2.6268e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0411, -0.5119,  0.0357, -0.0870, -0.0219, -0.0232, -0.0373, -0.0301,
        -0.0232, -0.0310, -0.0171, -0.0024, -0.0091, -0.0444, -0.0424, -0.0115,
        -0.0306,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1656e-01,  7.7053e-01,  5.6825e-02, -1.7984e-03, -4.2324e-02,
         3.0672e-03, -1.1308e-04,  5.6646e-03, -3.1183e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2203e-01,  3.9741e-01,  6.9254e-02,  2.0873e-02, -4.8178e-02,
        -1.3369e-02, -7.8152e-03,  7.8275e-04,  2.9066e-03, -1.8589e-03,
         1.1267e-02,  2.4150e-03,  4.7698e-03, -8.7975e-03,  2.4384e-03,
        -8.7113e-03,  2.2439e-02,  3.7832e-03, -3.3439e-03,  1.6275e-02,
         3.0878e-02, -1.7276e-04,  7.6112e-03,  9.2014e-03, -8.0832e-03,
         6.3706e-03,  1.8022e-02, -3.9402e-03,  2.6179e-03, -1.1629e-03,
         1.5848e-03, -1.1722e-02, -1.5825e-03,  4.4435e-03, -2.5552e-03,
         1.7501e-03,  1.3729e-02,  3.8435e-02, -4.9544e-03,  1.6291e-03,
         2.5759e-03,  2.5981e-03, -3.0884e-02, -5.6628e-03, -3.0099e-03,
         1.6103e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2902e-03,  5.7090e-01, -4.9581e-02, -1.2013e-02, -2.7536e-03,
        -8.8601e-03,  4.0441e-02, -3.6053e-03,  3.0477e-02, -7.4084e-02,
        -3.7729e-03, -3.5485e-03,  7.9863e-03,  9.8164e-03, -1.4392e-03,
         5.4708e-03,  2.3236e-03,  2.8528e-05, -3.9326e-03,  2.2601e-03,
         1.7126e-02,  1.8719e-02,  2.2528e-02, -2.8285e-03, -1.9201e-02,
         2.2036e-02,  2.9816e-02, -3.0126e-03,  4.7279e-03, -2.3419e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1321, -0.7167, -0.0332, -0.0215, -0.0361,  0.0221, -0.0382,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0460,  0.6951,  0.0137,  0.0633,  0.0070, -0.0115,  0.0109, -0.0194,
        -0.0136, -0.0469,  0.0725,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0548,  0.6315,  0.0130,  0.0097, -0.0040, -0.0194,  0.0093, -0.0442,
        -0.0125, -0.0209, -0.0229, -0.0284, -0.0498, -0.0798,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1076, -0.5142, -0.0745, -0.0126,  0.0068,  0.0147,  0.0120, -0.0008,
         0.0108,  0.0048, -0.0048, -0.0015,  0.0153,  0.0210,  0.0047,  0.0142,
         0.0029,  0.0029,  0.0031,  0.0157, -0.0061, -0.0036,  0.0130,  0.0080,
         0.0124, -0.0027, -0.0048,  0.0052,  0.0081,  0.0213,  0.0016,  0.0090,
        -0.0228, -0.0364,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0480,  0.3988,  0.0266, -0.0295, -0.0184,  0.0228,  0.0246, -0.0117,
         0.0236, -0.0409,  0.0091, -0.0030,  0.0115,  0.0208,  0.0210, -0.0263,
         0.0044,  0.0094,  0.0058,  0.0078,  0.0098, -0.0092,  0.0013,  0.0105,
        -0.0023, -0.0144, -0.1223,  0.0663,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0367,  0.4459,  0.1184, -0.0756, -0.0252,  0.0239,  0.0129, -0.0044,
         0.0253, -0.0052, -0.0061, -0.0194, -0.0081, -0.0091, -0.0078,  0.0025,
        -0.0722, -0.0234,  0.0191,  0.0043, -0.0023, -0.0075,  0.0328,  0.0118,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([-0.2461, -0.3584,  0.0328,  0.0662,  0.0378,  0.0076, -0.0015, -0.0006,
         0.0050, -0.0017,  0.0026,  0.0368, -0.0081, -0.0030,  0.0070, -0.0004,
         0.0012,  0.0302,  0.0021,  0.0039, -0.0092,  0.0077,  0.0115, -0.0008,
        -0.0101, -0.0015,  0.0270,  0.0059,  0.0039, -0.0655, -0.0041,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0175, -0.6997,  0.0513,  0.0495,  0.0698,  0.0065, -0.0162, -0.0051,
         0.0194,  0.0020, -0.0058, -0.0573,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1562, -0.5067,  0.0121,  0.0693, -0.0085, -0.0144, -0.0372, -0.0005,
        -0.0068,  0.0499, -0.0358, -0.0300, -0.0447,  0.0143, -0.0137,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4939e-02, -5.3746e-01, -1.2764e-02,  2.4991e-02, -6.8084e-02,
        -1.6186e-02, -3.2225e-02, -4.3407e-03, -3.4187e-02, -2.9072e-03,
         2.9791e-02,  2.4343e-03,  1.4007e-02,  1.7938e-02,  6.9475e-03,
        -1.2163e-03,  2.6003e-03, -2.0833e-02, -1.3146e-02, -2.0949e-02,
        -4.5057e-03, -7.1867e-03, -8.7418e-03, -2.7662e-03, -4.4362e-04,
        -1.3096e-02,  1.0638e-02, -1.0948e-02,  8.0441e-03, -4.3620e-04,
        -1.8353e-02,  1.9600e-02, -1.7294e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0836,  0.4158, -0.0800, -0.0614,  0.0275,  0.0263, -0.0139,  0.0044,
         0.0148,  0.0069, -0.0058,  0.0235, -0.0064, -0.0618, -0.0891, -0.0789,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.2621,  0.3654,  0.0335,  0.0838,  0.0218,  0.0674, -0.0641, -0.0751,
         0.0268,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2335,  0.6006,  0.0439, -0.0652, -0.0567,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0641,  0.6069,  0.0337,  0.0190, -0.0066, -0.0221,  0.0086, -0.0078,
        -0.0115,  0.0143,  0.0227, -0.0086, -0.0102,  0.0055,  0.0066,  0.0168,
         0.0089,  0.0159,  0.0083, -0.0034, -0.0011,  0.0008,  0.0066, -0.0254,
         0.0137, -0.0029,  0.0479,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1642,  0.6277, -0.0461,  0.0332, -0.0065,  0.0458, -0.0764,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1586, -0.5650, -0.0172, -0.0151, -0.0228, -0.0178,  0.0088, -0.0028,
        -0.0052, -0.0059, -0.0057, -0.0117,  0.0037,  0.0061,  0.0173,  0.0050,
        -0.0219,  0.0112, -0.0083, -0.0349,  0.0534,  0.0015,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0719, -0.5428, -0.0446,  0.0204,  0.0211, -0.0105,  0.0095,  0.0105,
        -0.0156,  0.0196, -0.0349, -0.0065, -0.0186, -0.0324,  0.0010, -0.0358,
        -0.0305,  0.0158,  0.0580,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0119,  0.5530,  0.0454,  0.0127,  0.0390, -0.0449, -0.0170,  0.0136,
         0.0258, -0.0064, -0.0208,  0.0041, -0.0085, -0.0071,  0.0038,  0.0072,
        -0.0091, -0.0033, -0.0121, -0.0322,  0.0108,  0.0048,  0.0007,  0.0109,
         0.0107, -0.0026, -0.0041,  0.0046, -0.0289,  0.0073, -0.0366,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #300: [tensor([-0.0184, -0.4240, -0.0629,  0.0013, -0.0432, -0.0101, -0.0294, -0.0135,
        -0.0078,  0.0023, -0.0158, -0.0074,  0.0424, -0.0115,  0.0005, -0.0336,
         0.0135, -0.0417,  0.0456, -0.0128, -0.0048,  0.0062,  0.0024,  0.0444,
        -0.0548,  0.0008,  0.0012,  0.0065,  0.0148,  0.0070, -0.0046, -0.0151,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0394, -0.4837, -0.0946, -0.0162,  0.0305, -0.0093, -0.0438,  0.0114,
        -0.0616,  0.0018, -0.0126, -0.0126, -0.0082, -0.0505,  0.0018,  0.0270,
         0.0098, -0.0199, -0.0126, -0.0170,  0.0075,  0.0280,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0230,  0.4791,  0.0034,  0.0133,  0.0155,  0.0267, -0.0826,  0.0701,
         0.0446,  0.0056,  0.0109,  0.0160,  0.0023,  0.0076, -0.0151, -0.0026,
        -0.0140,  0.0040,  0.0054,  0.0198,  0.0185, -0.0039, -0.0087, -0.0168,
        -0.0010, -0.0068, -0.0005, -0.0442,  0.0323, -0.0059,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0259, -0.6203,  0.0007, -0.0258, -0.0024,  0.0268, -0.0208, -0.0153,
         0.0245,  0.0500, -0.0070,  0.0317, -0.0054, -0.0062, -0.0044, -0.0027,
        -0.0649, -0.0651,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2940e-01,  5.1046e-01,  4.2268e-02,  3.1743e-02,  2.9787e-02,
         1.9575e-02, -1.1100e-02,  8.8783e-03, -3.4746e-03, -5.0692e-03,
         9.1228e-03, -5.0084e-06, -6.3051e-03,  1.1139e-03, -1.8400e-03,
        -1.0376e-02,  1.4700e-02, -2.0346e-04,  2.2735e-02,  1.5854e-02,
        -8.2424e-03, -9.6076e-05, -1.2973e-02,  1.8679e-03, -1.1516e-02,
        -5.1099e-03, -2.1281e-02, -1.3597e-03,  3.4016e-05,  3.7202e-03,
         7.9561e-03, -2.2653e-03, -7.8518e-05, -2.3598e-02, -1.0886e-02,
         1.3982e-02, -1.0222e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3448e-02, -4.9613e-01, -6.1624e-02, -1.1844e-01, -7.4425e-03,
        -5.1470e-03, -1.4660e-02,  3.2582e-03, -3.7102e-03, -1.3406e-02,
         9.8539e-04, -7.8848e-03, -1.8538e-02, -1.3300e-02,  3.3895e-03,
         9.7091e-03, -2.2131e-02,  2.1993e-04, -1.8408e-02,  6.3552e-03,
        -3.3123e-02, -5.1963e-03, -3.0051e-02, -1.3443e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7504e-02,  4.9060e-01,  1.5364e-02, -2.0688e-02,  2.7606e-02,
        -4.5858e-03,  2.0830e-02,  1.1411e-02,  2.0149e-02,  2.2758e-02,
         1.0502e-02, -2.3463e-03, -1.9437e-02,  5.5653e-03,  1.0866e-02,
         2.2173e-03, -2.2529e-02,  6.8532e-03,  3.3696e-04, -2.0362e-03,
        -3.5320e-03, -9.7464e-03, -1.1899e-02, -8.2007e-03,  1.0454e-02,
         2.8620e-03,  2.3005e-03,  1.3444e-02, -4.9378e-03,  8.4763e-04,
         4.0068e-02, -6.2354e-04, -1.0972e-02,  9.2593e-03,  1.5013e-02,
        -1.5093e-02, -8.9156e-04, -7.2944e-03,  2.7409e-03,  3.5149e-03,
        -9.1692e-03,  5.1733e-03, -9.2169e-04, -9.9509e-03, -4.3256e-02,
        -1.3648e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1150, -0.5108, -0.0542, -0.0040, -0.0209, -0.0287, -0.0423, -0.0137,
         0.0084,  0.0020,  0.0107,  0.0078, -0.0399,  0.0060,  0.0056,  0.0345,
         0.0083,  0.0077,  0.0319, -0.0068, -0.0114,  0.0142, -0.0152,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0382, -0.7401,  0.0125,  0.0289,  0.0043,  0.0040, -0.0029, -0.0350,
         0.0282,  0.0720,  0.0339,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0821,  0.4946,  0.0574, -0.0732, -0.0142,  0.0693,  0.0511, -0.0076,
        -0.0367, -0.0112,  0.0203, -0.0068,  0.0423,  0.0330,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1316, -0.5618, -0.0422,  0.0861, -0.0527, -0.0185,  0.0039,  0.0158,
         0.0365, -0.0091, -0.0097, -0.0321,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1735e-02,  5.0539e-01,  9.2070e-02, -4.7093e-02,  1.2918e-02,
         2.1442e-02, -1.2815e-02,  2.6213e-02,  1.9305e-03, -2.4383e-03,
        -5.9086e-03, -1.5019e-02, -1.5198e-03, -9.6648e-04, -2.0549e-04,
        -9.8080e-03, -1.5407e-02, -3.6947e-03, -3.7687e-03, -2.0893e-02,
        -9.4796e-03,  3.1147e-03, -6.1261e-04,  3.8700e-03,  2.0192e-03,
         6.7527e-03,  2.4274e-02, -3.4500e-03, -3.4293e-03,  3.4264e-03,
        -7.2272e-04, -1.4407e-02,  9.7686e-03,  5.9159e-03,  7.5908e-03,
        -4.3657e-03, -2.7968e-03, -7.3557e-03,  2.7714e-03, -4.9083e-03,
         2.0102e-03,  6.0052e-03,  5.7990e-03,  1.3440e-03,  2.6696e-03,
         1.1418e-03, -1.2203e-03, -5.8846e-04,  1.4653e-03, -6.8732e-03,
         4.6017e-04, -2.7533e-03,  2.5113e-03, -2.1957e-03,  1.0702e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 1.0404e-01, -5.6504e-01, -3.6483e-02, -1.3178e-03,  1.5366e-02,
         4.0032e-02, -5.9882e-03,  9.7132e-03,  1.8674e-02,  3.2012e-03,
         2.9163e-02,  5.2473e-03, -4.6519e-04,  2.2514e-02,  3.1039e-02,
         8.0687e-03,  3.6902e-03,  3.7132e-04, -2.5854e-04,  1.2224e-02,
         1.6637e-02, -1.4812e-03, -6.8989e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1365, -0.4950, -0.0351, -0.0234, -0.0374, -0.0035, -0.0018, -0.0340,
        -0.0132, -0.0117,  0.0564,  0.0009,  0.0148,  0.0381,  0.0238,  0.0192,
         0.0203,  0.0349,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1288,  0.4830, -0.0214, -0.0473,  0.0020,  0.0302,  0.1112,  0.0021,
         0.0046, -0.0046,  0.0197,  0.0084,  0.0383, -0.0098,  0.0212,  0.0180,
        -0.0177, -0.0317,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1477, -0.4501,  0.0238,  0.0161,  0.0272,  0.0830,  0.0063,  0.0652,
         0.0648, -0.0022,  0.0070, -0.0161, -0.0038,  0.0734,  0.0133,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0430, -0.5044, -0.0319, -0.0189,  0.0020, -0.0314,  0.0288,  0.0104,
        -0.0210, -0.0078,  0.0095,  0.0128,  0.0006, -0.0177,  0.0239,  0.0407,
        -0.0296, -0.0113, -0.0069, -0.0311, -0.0007,  0.0238, -0.0022, -0.0060,
        -0.0077,  0.0069, -0.0117, -0.0036,  0.0054, -0.0012,  0.0256, -0.0217,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7329e-02, -6.1135e-01, -1.2117e-02, -2.6511e-02,  1.9949e-02,
         6.8868e-03, -2.6392e-03, -1.2645e-03, -1.5485e-02, -1.2499e-03,
        -2.1687e-02,  7.3503e-03, -3.4948e-03, -1.4435e-02,  1.6767e-03,
         2.1349e-02, -1.6337e-03, -1.5648e-03, -2.0284e-04,  3.3394e-02,
         1.7311e-02, -1.7497e-02,  8.8380e-03,  2.1410e-02,  7.3159e-03,
         2.1879e-02,  1.1380e-02, -1.1616e-02,  8.8977e-03,  2.0022e-02,
        -6.7895e-03,  7.0651e-03, -8.4107e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.8181e-02,  4.7004e-01,  2.2992e-02,  2.0722e-02,  2.7007e-02,
        -2.9396e-02, -1.0592e-03,  1.3436e-02,  6.3836e-03,  6.8648e-03,
        -1.5172e-02, -1.0420e-02, -1.9564e-02,  2.5106e-02, -1.2903e-02,
        -7.3333e-03,  5.5875e-03,  3.2504e-03,  1.2858e-03, -5.0603e-03,
         1.5180e-04,  9.5115e-05, -1.5811e-03, -1.0550e-02,  1.9517e-03,
        -7.6933e-03,  1.0529e-02, -5.5594e-03,  8.8504e-04,  8.3607e-03,
         7.8911e-03,  2.9469e-03, -6.6873e-03, -1.3068e-03,  9.2479e-04,
        -9.2376e-03,  9.6637e-05, -3.3078e-03, -9.0460e-04, -4.8945e-03,
         1.9762e-03,  9.3696e-05, -5.5117e-03, -8.3376e-03, -5.8797e-03,
         6.8817e-03, -4.9788e-03,  1.0118e-02, -8.1601e-03,  8.3995e-03,
        -1.3887e-02,  1.5096e-03, -1.1706e-02, -1.0368e-02, -2.7304e-03,
        -1.1933e-02, -2.3936e-02, -1.6274e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.2181e-02,  5.7096e-01,  6.9310e-02, -4.4277e-04,  2.2462e-03,
        -1.1341e-02, -1.2796e-02,  1.4223e-02, -9.0581e-03,  3.6392e-03,
        -1.5470e-02, -1.6194e-02, -2.5142e-02, -3.3883e-03, -2.6878e-03,
        -1.5002e-03, -3.4717e-03, -6.2111e-03,  5.7886e-03,  1.3285e-02,
         4.3667e-03, -4.2414e-03, -2.2213e-02, -1.4920e-02, -1.3556e-02,
        -7.1371e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5198e-01,  6.1108e-01, -2.8735e-05,  2.9532e-03,  1.7150e-04,
        -1.5509e-02, -2.6182e-02, -6.2495e-03, -3.6198e-02,  8.9135e-03,
         5.8970e-03, -3.5269e-03, -8.6152e-03, -2.1033e-03, -2.2516e-03,
         2.9404e-03,  1.4458e-02, -8.6302e-03, -1.9713e-03, -1.3172e-02,
        -1.0184e-02, -4.8405e-03, -1.8152e-02,  3.9495e-02, -4.4957e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0272,  0.3306,  0.0864,  0.0160, -0.0520, -0.0192,  0.0485, -0.0134,
         0.0054,  0.0397, -0.0193,  0.0096,  0.0359,  0.0017,  0.0019,  0.0238,
         0.0062,  0.0048,  0.0013,  0.0091, -0.0098,  0.0529, -0.0029,  0.0013,
         0.0172, -0.0004, -0.0024,  0.0080,  0.0032,  0.0036, -0.0084, -0.0168,
        -0.0247, -0.0016,  0.0097,  0.0030,  0.0016,  0.0049, -0.0094, -0.0066,
        -0.0017, -0.0007, -0.0100, -0.0121, -0.0352,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0640,  0.5269, -0.0216,  0.0122,  0.0051, -0.0275,  0.0250,  0.0214,
        -0.0460,  0.0135, -0.0174,  0.0092, -0.0119, -0.0085,  0.0236,  0.0253,
         0.0411, -0.0064, -0.0201, -0.0013,  0.0171, -0.0314,  0.0235,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1490,  0.6632, -0.0656, -0.0529,  0.0034,  0.0132,  0.0043,  0.0071,
        -0.0145,  0.0068, -0.0132, -0.0068,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-6.6918e-02, -5.9956e-01, -6.4652e-02, -1.3416e-02, -8.8960e-04,
        -2.6297e-02, -4.2424e-02, -2.1350e-02, -4.9731e-03,  1.4615e-02,
         5.2857e-03, -1.2719e-02, -6.8834e-03, -1.0262e-02,  3.5087e-04,
        -1.7368e-02,  1.3627e-02,  9.2862e-03,  5.8820e-03,  2.8615e-03,
         3.3396e-02,  1.1496e-02, -1.5491e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0074, -0.6263, -0.0029,  0.0104,  0.0205,  0.0349, -0.0033, -0.0074,
        -0.0043, -0.0122, -0.0053, -0.0102, -0.0103, -0.0223, -0.0147,  0.0085,
         0.0009,  0.0027,  0.0135,  0.0030, -0.0130, -0.0023,  0.0049,  0.0025,
         0.0044, -0.0092, -0.0525, -0.0057,  0.0019,  0.0016,  0.0062,  0.0386,
        -0.0361], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0708, -0.5205,  0.0532, -0.0390,  0.0583,  0.0124,  0.0813,  0.0228,
         0.0139,  0.0045,  0.0144, -0.0064, -0.0010, -0.0464, -0.0044,  0.0174,
         0.0027,  0.0163,  0.0143,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.2908, -0.2571, -0.0082,  0.0118,  0.1837, -0.0096,  0.0011, -0.0048,
         0.0250,  0.0132,  0.0327, -0.0087, -0.1445,  0.0089,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.7248e-03, -5.1474e-01, -1.6068e-02, -2.5067e-02, -1.4515e-01,
        -3.6941e-02,  1.0767e-02,  4.0391e-02,  6.5892e-03,  6.7698e-03,
         3.8574e-04, -7.4403e-03, -2.1311e-02,  3.2345e-02, -5.8876e-03,
         2.2478e-03,  1.9061e-02, -3.7604e-03, -2.6990e-03, -9.8971e-03,
         7.1095e-02,  1.4665e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0425,  0.5391,  0.0541,  0.0444,  0.0300, -0.0174,  0.0013, -0.0319,
        -0.0019, -0.0251,  0.0086,  0.0015, -0.0309,  0.0050, -0.0024,  0.0015,
        -0.0104, -0.0029,  0.0100, -0.0167, -0.0035,  0.0035,  0.0111,  0.0076,
        -0.0219, -0.0084, -0.0097,  0.0083, -0.0432, -0.0054,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2479,  0.2884,  0.0359, -0.0487,  0.0119, -0.0368,  0.0054,  0.0423,
        -0.0204, -0.0237, -0.0137, -0.0352, -0.0073,  0.0017,  0.0342, -0.1349,
        -0.0115,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0636, -0.6075, -0.0578, -0.0808, -0.0062,  0.0096, -0.0137,  0.0441,
         0.0054,  0.0560,  0.0551,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6725e-03, -6.9278e-01, -4.3862e-02,  1.4132e-02, -4.0172e-03,
         8.7180e-02, -1.9171e-02, -5.8274e-03,  8.8987e-03, -9.7145e-03,
         5.7454e-03, -9.9902e-04,  1.5205e-02, -4.0047e-02, -3.3593e-04,
         2.9661e-02,  1.6748e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1034,  0.4334,  0.0132,  0.0307, -0.0440, -0.0046, -0.0175,  0.0086,
         0.0069, -0.0149, -0.0133,  0.0074, -0.0257,  0.0041, -0.0646, -0.0107,
        -0.0013, -0.0089,  0.0030, -0.0154, -0.0141, -0.0135,  0.0114,  0.0014,
        -0.0641, -0.0639,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0239,  0.5799,  0.0902,  0.0038, -0.0333,  0.0113, -0.0047,  0.0174,
        -0.0229, -0.0172, -0.0091, -0.0160, -0.0036,  0.0130, -0.0102,  0.0020,
         0.0053,  0.0164, -0.0079, -0.0054,  0.0034,  0.0007,  0.0092,  0.0042,
         0.0097,  0.0054,  0.0058,  0.0103, -0.0383,  0.0197,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.8674e-02, -5.7113e-01, -1.6925e-02, -8.5787e-02,  9.4355e-03,
        -1.4833e-02,  1.5043e-02,  1.6133e-02,  2.3973e-02,  2.4106e-02,
        -4.1779e-04,  8.5789e-03,  1.9228e-02, -1.5126e-02, -4.1874e-02,
        -1.3792e-02,  1.5260e-02,  4.6676e-02,  1.3013e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-1.9400e-02, -5.6674e-01, -1.0712e-01, -6.2277e-02, -2.2843e-02,
        -1.2254e-04,  2.3754e-02,  1.1474e-03, -1.4415e-02,  3.6295e-03,
         2.1884e-03,  8.5614e-03, -1.5615e-02, -1.2665e-02,  6.5050e-03,
         7.1249e-03, -2.5363e-02,  8.7963e-03, -1.5388e-02, -1.4376e-02,
        -3.4595e-03,  3.5326e-03, -2.7872e-03, -8.9366e-03, -5.9497e-03,
         3.4382e-02,  2.9237e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.0713e-02,  4.4817e-01, -5.9699e-02, -7.7165e-02,  8.1994e-04,
        -5.0512e-02, -2.2003e-02,  3.7865e-04,  2.9137e-02,  1.9360e-02,
        -9.4393e-02, -1.6958e-02,  3.8529e-02, -8.2166e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0811e-02,  6.1075e-01,  4.6888e-03, -1.3702e-02, -3.3384e-04,
         1.4144e-02,  1.7757e-02, -1.2071e-03, -1.4115e-02, -1.3997e-02,
        -9.0475e-03, -2.0030e-02,  1.5398e-02, -8.6285e-03, -1.1499e-02,
         2.1716e-02,  5.1096e-03,  6.0850e-03, -2.2744e-02,  3.0642e-03,
        -9.2233e-03, -5.0397e-03, -1.0703e-02, -1.5135e-02, -2.8242e-02,
        -7.6832e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0310,  0.6450,  0.0337, -0.0161, -0.0161, -0.0477,  0.0039, -0.0338,
        -0.0529,  0.0594,  0.0226, -0.0240, -0.0094,  0.0043,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2061e-02,  5.3647e-01,  1.5010e-02,  5.8668e-03,  2.0268e-02,
        -4.4971e-02,  6.3255e-03, -1.1978e-03,  9.0345e-03,  1.1639e-02,
        -1.1334e-02, -1.0548e-02, -1.0713e-02, -2.4035e-02,  1.5312e-02,
        -3.3957e-04, -4.0518e-03, -5.5226e-03, -1.2969e-02,  2.0106e-02,
        -8.9714e-03,  3.8600e-03, -1.3325e-02,  6.1279e-03,  6.7231e-03,
        -6.3633e-03, -1.0489e-02, -3.4045e-02,  7.8037e-03, -7.9291e-03,
        -6.4952e-03,  9.0098e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2616e-01,  4.0911e-01,  3.1102e-02, -1.7507e-02, -1.7246e-01,
        -6.4732e-04, -2.2192e-02,  5.2854e-03, -1.7984e-02, -1.6430e-02,
        -1.4803e-02,  1.2727e-02,  6.3309e-03, -7.8548e-03,  1.5787e-02,
        -2.4083e-03, -2.3910e-04,  3.5117e-03,  1.5236e-04,  6.3681e-03,
         1.7350e-03,  2.4522e-03, -5.6834e-03,  8.2031e-02, -1.9033e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1124,  0.6633,  0.0687, -0.0756, -0.0380,  0.0420,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.4083e-02,  6.1591e-01,  8.0121e-02, -1.9000e-02,  1.6615e-02,
         1.1829e-02,  9.7577e-03, -6.2588e-03,  3.2989e-03, -2.0081e-04,
         7.6961e-03,  4.2024e-03, -2.4601e-03, -5.3916e-04,  1.2214e-03,
        -1.4282e-03,  7.2434e-03,  1.4073e-02,  7.5720e-03,  5.8118e-03,
        -1.0687e-02, -1.0096e-02, -9.9104e-05, -9.5556e-04,  5.2836e-03,
         5.4014e-03, -2.4045e-03, -3.5389e-03, -6.4360e-03,  1.3654e-03,
         1.2814e-02,  7.7753e-03, -2.8748e-03,  4.2019e-03, -2.5962e-04,
         2.1839e-02,  1.4651e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2938e-02, -5.2503e-01,  6.4375e-02, -3.4197e-02,  4.4716e-02,
        -2.6138e-02,  4.4242e-03,  4.3351e-03,  2.4762e-05, -6.1155e-03,
         1.2142e-03, -1.3339e-02, -1.3580e-02, -1.9796e-03,  1.4598e-02,
         1.3479e-02,  1.1995e-02, -1.2843e-02,  6.6436e-03,  1.3122e-02,
         2.7422e-02,  3.2734e-04,  1.5392e-03, -8.7674e-04, -2.5713e-05,
         7.2717e-03, -1.3526e-02, -1.6380e-03,  1.1575e-02,  1.0976e-02,
         2.9315e-03, -2.8593e-05, -1.0540e-04,  7.6425e-03, -6.9024e-02,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0909,  0.3113, -0.2374, -0.0526, -0.0116,  0.0109,  0.0589, -0.0051,
        -0.0498, -0.0366,  0.0073,  0.0011, -0.0077,  0.0399,  0.0791,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4091e-02,  5.9665e-01,  4.3666e-02, -5.0023e-02,  1.0951e-02,
         2.0549e-03,  1.5913e-02, -2.4709e-02,  8.7724e-03,  1.0624e-02,
         3.5912e-03, -2.0979e-03, -1.6892e-03, -1.0413e-02,  8.1333e-03,
        -2.0352e-02, -1.3107e-03,  1.5465e-02, -5.9747e-03, -7.4635e-03,
        -2.4736e-03, -1.7716e-02, -7.0116e-03, -2.0920e-03, -4.5456e-03,
        -3.1553e-03, -5.5382e-03, -3.4110e-03, -7.4366e-03, -4.2451e-04,
        -1.7065e-02,  1.5071e-02, -1.0116e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6562e-03,  6.8731e-01,  1.2701e-02, -3.0663e-03, -1.2574e-02,
        -2.8381e-02,  7.6708e-03, -4.7817e-03, -2.6021e-02, -8.8567e-03,
        -1.8683e-02, -1.9968e-02, -1.4127e-02, -4.7946e-03, -1.4460e-02,
        -6.7354e-04, -8.2474e-03, -9.0619e-03,  7.5301e-03, -8.8780e-03,
        -2.8558e-03, -2.8026e-03, -8.8908e-03,  5.2157e-04,  2.7498e-03,
        -1.0883e-02, -2.4397e-02,  6.3689e-03,  3.8088e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([-0.0711, -0.4799,  0.0805,  0.0630,  0.0326, -0.0104,  0.0373, -0.0017,
        -0.0045,  0.0498,  0.0005,  0.0197,  0.0160,  0.0361,  0.0970,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9132e-02,  4.9453e-01, -4.4879e-02, -3.9662e-02,  2.0031e-02,
        -1.7799e-02,  1.5296e-02,  2.6277e-02, -2.1661e-02, -2.3577e-02,
        -5.7058e-03, -4.0818e-03,  6.7026e-03, -5.1908e-03,  7.7705e-03,
        -1.9172e-02, -1.2472e-02, -4.6800e-03,  5.3823e-04, -3.4993e-03,
         2.7129e-04, -1.5244e-02, -9.1117e-03, -3.9485e-03, -6.8708e-03,
        -9.7375e-03, -3.0366e-03, -1.4264e-03, -1.6906e-04,  3.1441e-03,
         3.1504e-03,  5.9070e-03,  7.7538e-03, -7.7152e-03, -1.0549e-02,
        -7.9198e-03, -6.1150e-03, -3.6981e-03, -1.0319e-02, -6.8497e-03,
        -3.6427e-03, -7.6219e-03,  6.2223e-03, -5.6402e-03,  1.3669e-04,
        -2.9706e-03,  1.7562e-03, -2.5713e-02,  2.0705e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1208,  0.4641,  0.0224, -0.0574,  0.0030,  0.0362,  0.0085, -0.0162,
        -0.0400, -0.0018, -0.0095, -0.0066,  0.0100,  0.0066,  0.0091, -0.0146,
         0.0158,  0.0024,  0.0019,  0.0167, -0.0034,  0.0034, -0.0029, -0.0056,
         0.0064,  0.0057, -0.0152,  0.0167,  0.0080, -0.0114, -0.0019, -0.0057,
         0.0144, -0.0215, -0.0141,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0175, -0.6129, -0.0039,  0.0034,  0.0299,  0.0286,  0.0261,  0.0205,
        -0.0051, -0.0105, -0.0288,  0.0120,  0.0043, -0.0030, -0.0041,  0.0028,
         0.0409, -0.0141,  0.1316,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1400, -0.4798, -0.0200, -0.0012,  0.0717, -0.0240,  0.0256, -0.0448,
         0.0680,  0.0227,  0.0161, -0.0108, -0.0039, -0.0460,  0.0255,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.7067e-02, -5.8015e-01, -8.0851e-02,  6.7253e-02, -1.4102e-03,
        -2.4378e-02,  1.8398e-02,  4.4509e-04,  2.9988e-03, -1.8060e-03,
         1.5218e-02,  2.3792e-02,  2.5475e-02, -2.7600e-03,  4.3383e-03,
         2.5551e-02,  2.1633e-04,  8.3456e-03,  1.9450e-03, -3.1907e-02,
        -1.7367e-06,  5.6914e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3757e-03,  4.9764e-01,  1.6898e-02,  8.7094e-02,  4.0450e-02,
         3.0757e-02, -3.3895e-02, -1.5333e-02, -3.4256e-03,  1.3523e-02,
        -1.1956e-02, -4.0344e-02, -8.7383e-03, -2.4799e-02, -2.7637e-03,
        -4.1386e-03, -2.0797e-02, -1.2380e-02,  1.7982e-02,  2.1727e-03,
        -1.5342e-02,  2.0515e-02, -8.7703e-03, -2.7003e-03,  2.3163e-02,
        -1.6345e-03, -1.1546e-02,  2.8330e-03, -2.3802e-02, -2.3233e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0393, -0.4992,  0.1192, -0.0123,  0.0366, -0.0437,  0.0683, -0.0084,
         0.0569, -0.0659,  0.0503,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1924e-02,  6.6590e-01, -7.5507e-02, -1.2143e-02, -2.0748e-02,
         7.3172e-04, -3.1398e-03,  1.7608e-02,  1.8621e-02,  1.5766e-02,
        -5.4805e-04, -2.5853e-03, -5.0833e-03, -4.2894e-03,  1.0719e-03,
         6.8367e-03, -8.1198e-03,  1.3760e-04, -2.0254e-03, -7.2213e-03,
         2.6352e-02, -5.3645e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.7225e-02, -6.3607e-01, -1.4063e-02, -2.5217e-02, -1.7554e-03,
         2.9820e-03, -4.1789e-03,  1.2720e-02,  7.2412e-03,  1.9057e-02,
         2.3074e-02,  2.1728e-02,  1.2320e-02,  7.7793e-03,  6.1625e-03,
         8.5652e-04,  1.0113e-04, -3.6447e-03, -1.1133e-02, -1.1146e-03,
         4.3539e-03,  2.7183e-03,  1.3224e-02,  2.4407e-03,  3.0378e-04,
         7.7594e-03,  1.2012e-02,  7.9749e-03,  1.1764e-02,  1.7688e-02,
         2.7242e-03,  1.6010e-02,  1.2609e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0109,  0.5465,  0.0702, -0.0739, -0.0347, -0.0042,  0.0035,  0.0043,
        -0.0025, -0.0242, -0.0060,  0.0036, -0.0058, -0.0235, -0.0214,  0.0482,
         0.0021,  0.0034,  0.0012, -0.0148, -0.0069, -0.0248,  0.0040, -0.0109,
        -0.0020, -0.0011, -0.0304,  0.0151,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.9653e-03,  6.4285e-01, -9.3889e-03,  1.3786e-03,  1.0105e-02,
        -1.1410e-02, -3.2480e-02, -5.4289e-02, -2.3765e-02, -2.0050e-03,
        -1.8077e-02,  1.0957e-02, -2.0382e-03, -4.9853e-03,  2.8602e-03,
        -2.8256e-02, -1.2121e-04, -9.5864e-03, -6.1211e-03,  2.8007e-02,
         3.2434e-03, -2.2676e-04, -4.0092e-02,  2.5215e-02,  2.2577e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-0.0151, -0.4826,  0.0125, -0.0057,  0.0235,  0.0096,  0.0336,  0.0217,
         0.0089,  0.0099,  0.0014,  0.0285, -0.0019, -0.0095,  0.0053,  0.0180,
        -0.0040,  0.0067,  0.0039, -0.0008,  0.0057,  0.0148,  0.0121, -0.0126,
         0.0341, -0.0072,  0.0554,  0.0119,  0.0097, -0.0254, -0.0339,  0.0008,
        -0.0137, -0.0022, -0.0088, -0.0037, -0.0030, -0.0149, -0.0105, -0.0067,
        -0.0101,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1025,  0.5994, -0.0363, -0.0358, -0.0307, -0.0186,  0.0176, -0.0245,
        -0.0068,  0.0015,  0.0049, -0.0222,  0.0068,  0.0034,  0.0098,  0.0079,
         0.0121,  0.0102, -0.0078, -0.0045,  0.0019,  0.0054,  0.0039, -0.0132,
         0.0122,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.4390e-03, -6.4207e-01, -2.5970e-02, -1.3425e-02, -4.1101e-03,
        -2.5938e-02,  3.3492e-04, -4.5747e-02, -2.5426e-02,  4.5068e-02,
        -2.0587e-03, -8.5477e-03, -5.5702e-03, -3.6027e-04,  1.3742e-02,
         2.2789e-02,  1.3363e-02,  4.1441e-02,  1.9363e-02,  3.1951e-02,
        -5.2830e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0592,  0.6521,  0.0391,  0.0203,  0.0362,  0.0091,  0.0082, -0.0279,
         0.0039,  0.0030, -0.0090,  0.0036, -0.0062, -0.0138, -0.0307, -0.0117,
        -0.0116, -0.0543,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0438,  0.4466,  0.1009,  0.0289,  0.0043, -0.0472,  0.0239, -0.0019,
        -0.0076,  0.0013, -0.0012, -0.0102,  0.0040,  0.0121,  0.0122, -0.0047,
        -0.0190, -0.0052, -0.0006,  0.0148, -0.0037, -0.0028,  0.0102, -0.0007,
        -0.0127,  0.0133,  0.0043, -0.0345, -0.0040,  0.0253,  0.0026,  0.0052,
        -0.0037, -0.0028,  0.0199,  0.0096, -0.0385, -0.0157,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0839, -0.5170,  0.0859,  0.0628, -0.0128,  0.0101,  0.0288, -0.0083,
         0.0186, -0.0093,  0.0068,  0.0224, -0.0038,  0.0163, -0.0106, -0.0168,
        -0.0247,  0.0153,  0.0380, -0.0079,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1464, -0.5975, -0.0178, -0.0130, -0.0100, -0.0209, -0.0229, -0.0051,
        -0.0033,  0.0188, -0.0128,  0.0127,  0.0800,  0.0386,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6806e-02,  3.9130e-01, -6.7957e-02, -5.5818e-02,  1.4462e-02,
         4.1296e-02,  8.5360e-03,  3.2432e-02,  2.8190e-03,  4.9746e-02,
         2.9488e-02,  8.8052e-04, -1.8102e-02,  6.2660e-06,  1.9455e-02,
         3.0137e-03,  1.4946e-02,  1.2275e-02, -3.4016e-03, -6.5793e-03,
         1.7124e-02,  3.0738e-03,  8.1099e-03, -1.6296e-02,  6.5514e-04,
        -1.0074e-02,  8.9050e-03,  1.3706e-02,  1.4780e-02,  8.3464e-04,
         6.6496e-02,  5.0624e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6453e-02,  6.2821e-01,  1.0744e-01,  5.2448e-02, -5.6790e-04,
        -1.9265e-03,  1.0377e-02,  1.9211e-03, -6.6261e-03,  2.6751e-02,
        -1.1140e-02, -1.3149e-02,  2.5972e-03,  6.6071e-03,  1.5103e-02,
         1.0425e-02,  2.2666e-04, -1.0923e-02,  1.2410e-02, -6.4693e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8196e-02,  4.0970e-01,  3.6829e-02, -8.6951e-02, -2.9829e-02,
         1.3980e-02, -5.4458e-03, -5.8278e-03,  3.3304e-03, -2.0609e-02,
        -1.5010e-02, -1.0541e-02, -2.0162e-03, -1.4169e-03,  2.1899e-03,
        -5.8268e-03,  1.4774e-03, -2.8815e-03, -1.6227e-02,  4.0817e-04,
        -1.7913e-02, -1.4641e-02,  3.1464e-03,  9.6071e-03, -6.3522e-03,
        -4.1834e-03,  1.2295e-03,  1.6157e-03,  2.8352e-03, -5.4282e-03,
        -1.6920e-02, -1.6398e-02,  7.0079e-04, -7.1809e-03, -1.5456e-02,
         7.5285e-05,  3.4542e-03, -5.5941e-03, -4.5266e-03,  3.6150e-03,
        -1.4567e-02, -2.3549e-03,  2.7437e-02,  4.7784e-03, -3.8773e-03,
        -3.0848e-02,  7.6578e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0599, -0.6625, -0.0034, -0.0086, -0.0702, -0.0069, -0.0033,  0.0171,
        -0.0166, -0.0041, -0.0097, -0.0126,  0.0035,  0.0301,  0.0915,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.3593e-03, -5.9240e-01, -6.7110e-02,  1.1903e-02,  3.2995e-03,
         1.9168e-02, -4.0500e-03, -7.4848e-03, -2.2513e-02,  1.8874e-02,
        -4.1954e-03,  2.3853e-03, -8.4619e-03, -5.1053e-03,  1.0734e-02,
        -3.8741e-03, -3.0670e-04,  1.0266e-02,  8.5909e-03, -9.0957e-03,
        -5.5960e-03,  5.8008e-03,  3.6683e-03,  7.8335e-03, -8.4513e-04,
        -6.7830e-03, -1.1943e-02, -4.3415e-02, -1.6249e-03, -3.7533e-04,
         3.0027e-03,  8.4428e-03, -2.5841e-03, -5.5477e-04, -5.0747e-04,
         6.3106e-03,  1.8174e-03, -1.5331e-03,  1.2883e-02, -1.9546e-03,
        -3.4086e-03,  2.5958e-03,  3.4358e-03,  1.4973e-03,  2.2807e-03,
        -2.9459e-03,  2.3254e-02, -1.8926e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 3.8081e-02, -5.4689e-01, -6.3477e-02, -4.1494e-02,  1.4431e-04,
         2.8956e-03, -3.0404e-02, -2.9513e-02, -3.9930e-02, -1.5816e-02,
        -2.2670e-02, -2.7632e-02,  7.8753e-03, -1.6740e-02, -6.4862e-03,
        -1.4394e-03,  1.5416e-02, -5.2915e-03,  5.9037e-03, -1.7724e-02,
        -1.5588e-02,  2.0843e-03,  2.0971e-02,  5.9522e-03,  1.2330e-02,
        -7.2512e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0113, -0.5886, -0.0134,  0.0131,  0.0181, -0.0364,  0.0564,  0.0128,
         0.0145, -0.0196, -0.0391,  0.0191, -0.0031,  0.0145,  0.0077, -0.0058,
        -0.0102,  0.0542, -0.0079,  0.0381,  0.0163,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8039e-02,  4.4908e-01,  3.7000e-02, -2.9985e-02, -1.4545e-02,
        -1.3410e-02,  7.2095e-03, -1.1308e-02, -1.7145e-02, -2.1317e-03,
        -1.2171e-02, -5.5677e-03, -3.2228e-02, -3.0060e-02, -1.3082e-02,
        -6.1448e-03, -2.2694e-02, -3.1025e-03,  3.6702e-02,  2.8377e-02,
         1.4006e-02,  2.5770e-04,  4.1926e-02,  6.9029e-03,  1.8969e-03,
        -8.5011e-03, -3.1549e-03, -1.9416e-03, -4.4237e-03,  1.4339e-03,
        -1.2808e-02,  4.5953e-03,  8.2707e-03, -5.0990e-03, -2.6793e-02,
         5.8011e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0289,  0.6812, -0.0037,  0.0055,  0.0296, -0.0172, -0.0403,  0.0131,
         0.0096,  0.0209, -0.0036,  0.0212,  0.0092,  0.0067, -0.0165,  0.0051,
         0.0063,  0.0020, -0.0552, -0.0243,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3789e-02, -5.9012e-01, -3.9783e-02, -3.8096e-02, -8.8940e-03,
         1.6216e-03, -8.8075e-03, -4.7198e-02,  2.1258e-02,  8.0887e-05,
        -2.8274e-02,  1.2987e-02, -1.5424e-02, -1.6974e-02, -1.2716e-03,
        -2.8139e-02, -1.1811e-03, -3.5551e-03,  2.8891e-02,  2.3137e-02,
         3.0519e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0561, -0.6432, -0.0683, -0.0063, -0.0154,  0.0028,  0.0452,  0.0506,
        -0.0280,  0.0028,  0.0177,  0.0370, -0.0266,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3485e-02, -5.3520e-01, -5.1856e-02,  3.2773e-04, -9.9963e-03,
         1.2576e-02,  2.2243e-02, -1.7835e-02, -7.1471e-03, -1.8768e-02,
         1.5082e-03,  1.8869e-02,  6.4077e-03, -1.1832e-03, -1.5253e-02,
         1.2094e-02,  2.4646e-02,  3.0538e-03,  8.9079e-04,  5.7563e-04,
        -2.8978e-02,  2.8095e-02, -6.6118e-04,  1.4694e-02, -3.1892e-03,
         8.1848e-03,  5.6244e-03,  1.4738e-03, -2.7241e-03,  1.7335e-02,
         5.6911e-02, -6.8273e-03, -1.6027e-02, -3.7638e-03,  4.3944e-03,
        -1.3446e-02, -1.3756e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0344, -0.6137, -0.1196, -0.0147,  0.0385, -0.0013, -0.0017,  0.0071,
        -0.0138, -0.0050, -0.0047,  0.0170, -0.0141, -0.0100, -0.0050, -0.0082,
        -0.0007, -0.0011,  0.0014,  0.0077, -0.0193,  0.0117,  0.0052, -0.0441,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2456e-02,  3.4833e-01, -8.5240e-03,  6.1525e-02, -2.6410e-02,
        -6.9098e-03, -5.5189e-02,  4.6958e-02,  6.7925e-03, -3.0317e-02,
        -2.2872e-02, -4.6419e-02, -1.1445e-02,  5.2013e-03, -2.8743e-03,
        -1.7379e-02, -7.0596e-02, -4.1341e-02, -3.2502e-02, -2.2804e-02,
        -9.2849e-03, -3.8526e-02, -1.9284e-03,  9.4757e-03, -1.9926e-02,
         4.1963e-04,  2.2067e-03,  3.8018e-03,  1.2845e-04,  2.1561e-02,
         1.7110e-03, -4.1860e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0571, -0.6069,  0.0022, -0.0153,  0.0008,  0.0301, -0.0034,  0.0113,
        -0.0190, -0.0222,  0.0090, -0.0075,  0.0050,  0.0142,  0.0061,  0.0069,
         0.0110, -0.0070,  0.0143,  0.0046,  0.0070, -0.0012,  0.0045,  0.0049,
         0.0101,  0.0301,  0.0261, -0.0091,  0.0012,  0.0337, -0.0181,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.8831e-02,  6.2673e-01,  1.0431e-02,  1.8984e-04,  5.4960e-02,
         1.7621e-02, -1.4205e-02, -2.8383e-02,  1.2993e-02, -1.4927e-03,
         1.6334e-02,  4.8985e-03,  6.0724e-04, -7.0869e-03,  9.8991e-03,
         9.4966e-03,  2.8396e-05, -1.7650e-03,  8.8918e-03, -2.6608e-02,
         3.7502e-02,  5.2080e-02, -1.8968e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5156e-02,  5.6806e-01,  1.3358e-02,  3.2865e-02,  1.0229e-02,
         5.0445e-02, -7.8812e-04,  2.6870e-02, -1.4597e-02, -8.9586e-03,
        -1.5948e-02,  2.1083e-02,  8.8283e-03, -1.1325e-02,  2.2957e-02,
        -4.6017e-03, -7.8749e-03, -2.8314e-03, -1.5392e-03,  1.1106e-02,
        -4.1834e-02,  2.2956e-02,  2.2246e-02, -4.1068e-04, -9.2973e-04,
        -5.1449e-03, -1.0399e-02,  6.6563e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-6.1886e-02,  5.7057e-01, -5.4046e-02,  2.6849e-02,  3.7788e-03,
         1.2914e-02,  7.3032e-03, -2.2762e-02, -1.3733e-02, -1.2132e-02,
        -2.3563e-02, -1.2020e-02,  1.2778e-02, -2.6802e-04, -1.1424e-03,
        -2.6184e-03, -1.3276e-02,  4.1340e-03, -1.4548e-02,  9.4276e-03,
         2.9239e-02,  4.1639e-04, -2.4444e-02, -6.2373e-03, -1.1649e-02,
        -1.5139e-02,  9.6397e-03, -8.9671e-03,  1.4519e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6166e-02, -6.0715e-01,  6.4461e-04, -4.9716e-03,  4.2171e-02,
        -8.9051e-03, -8.1281e-03,  4.3650e-03,  3.0989e-02, -6.9211e-03,
         7.3455e-03,  1.2956e-02,  3.3394e-02,  1.1317e-02,  1.1389e-02,
         5.9573e-04,  1.5522e-02,  2.0369e-02,  2.7568e-03,  2.2859e-02,
         2.4671e-02, -2.6410e-02,  4.5696e-02,  2.4309e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3484e-03,  4.7922e-01,  1.5326e-02, -4.5296e-02,  3.8483e-02,
         6.3912e-03,  2.5656e-02, -2.9169e-02,  2.4853e-02,  3.1743e-03,
        -3.6657e-02,  2.6712e-03,  2.4216e-02,  1.8727e-02, -1.0348e-02,
        -4.2389e-03,  5.3517e-03, -4.7570e-03,  9.4153e-03,  2.9573e-03,
        -2.3687e-02, -1.4426e-02, -2.7570e-04,  1.6481e-02, -2.0575e-03,
         9.9099e-03,  4.4734e-03, -5.7372e-03,  9.6693e-03,  7.7168e-03,
         1.3267e-02, -7.4000e-03,  1.8858e-02, -7.6788e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0925, -0.5451, -0.0656,  0.0012,  0.0076, -0.0013, -0.0438, -0.0304,
         0.0011, -0.0179,  0.0061, -0.0050,  0.0011,  0.0083,  0.0013, -0.0136,
        -0.0215, -0.0272,  0.0211, -0.0140,  0.0135, -0.0607,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0045, -0.5911,  0.0347, -0.0185,  0.0507, -0.0639, -0.0141,  0.0170,
        -0.0220,  0.0189,  0.0309,  0.0097, -0.0093, -0.0099,  0.0201,  0.0153,
        -0.0583, -0.0111,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1235, -0.4846, -0.0326, -0.0587, -0.0412, -0.0011, -0.0233, -0.0047,
         0.0093, -0.0253, -0.0068, -0.0019,  0.0278,  0.0040,  0.0097,  0.0159,
         0.0206,  0.0043,  0.0102,  0.0056,  0.0071, -0.0238,  0.0580,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3753e-02, -6.2119e-01, -3.8153e-02, -1.3683e-01, -1.6935e-02,
         1.8699e-02, -1.5006e-02, -2.1621e-02, -6.2840e-03, -1.6347e-03,
         8.3770e-04, -2.3197e-03, -1.0978e-02,  4.6141e-03,  2.2022e-05,
         1.2888e-02,  5.8235e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1130,  0.5095,  0.0951,  0.0480, -0.0045,  0.0180,  0.0180, -0.0110,
        -0.0115, -0.0134,  0.0009, -0.0051,  0.0049, -0.0147,  0.0099,  0.0163,
         0.0337,  0.0727,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0566, -0.5439, -0.0110,  0.0273, -0.0104, -0.0452, -0.0112, -0.0290,
         0.0043, -0.0138, -0.0127, -0.0182, -0.0137,  0.0014,  0.0145, -0.0152,
        -0.0243, -0.0283, -0.0118,  0.0103,  0.0287, -0.0035, -0.0319,  0.0042,
        -0.0110,  0.0177,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0373, -0.4908,  0.0730,  0.0338,  0.0459,  0.0221,  0.0198,  0.0041,
         0.0226,  0.0070, -0.0052, -0.0085, -0.0155,  0.0155,  0.0040, -0.0117,
        -0.0134, -0.0042,  0.0116,  0.0181,  0.0013, -0.0045, -0.0008,  0.0076,
        -0.0132, -0.0125,  0.0013, -0.0012, -0.0237,  0.0114,  0.0580,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.4262e-02,  6.7639e-01,  5.3671e-02,  4.1941e-03, -1.1215e-02,
         2.1817e-02, -1.9959e-02,  6.3459e-02, -2.6724e-02,  1.4008e-04,
        -9.4215e-03,  2.3230e-03, -2.5692e-04,  1.3455e-03, -1.7744e-02,
         3.1373e-03, -2.4164e-02, -9.7772e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0183, -0.5585, -0.0214, -0.1369, -0.1168,  0.0423, -0.0709, -0.0349,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([-3.1138e-02, -5.5540e-01, -4.4096e-02,  2.5726e-03, -4.8984e-03,
         2.0037e-02,  1.9564e-02,  1.6794e-03,  1.5447e-03,  6.3741e-03,
         9.6221e-03, -1.4223e-02, -4.3812e-03, -8.7548e-03, -1.9860e-02,
        -7.3124e-03,  2.8443e-03,  2.1404e-02,  5.8535e-03,  3.7489e-03,
        -3.6446e-02,  6.4187e-02, -1.7226e-02, -4.1220e-03,  4.5601e-03,
         9.7775e-03, -5.1159e-04, -7.0388e-03,  5.0158e-03, -3.0157e-03,
        -3.3849e-04, -1.4197e-03,  6.1506e-03,  7.8772e-03,  5.8919e-03,
         9.5929e-04,  9.0000e-03, -3.9921e-03,  2.7157e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0057, -0.4849,  0.0680,  0.0616, -0.0017, -0.0095, -0.0152, -0.0211,
        -0.0159,  0.0012, -0.0144, -0.0304,  0.0028, -0.0152,  0.0340, -0.0018,
         0.0211,  0.0301,  0.0204,  0.0319,  0.0266,  0.0034, -0.0104,  0.0044,
        -0.0073,  0.0072,  0.0147,  0.0391,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0169, -0.5965, -0.0549, -0.1371,  0.0238,  0.0026, -0.0549,  0.0262,
        -0.0167, -0.0217, -0.0072,  0.0355,  0.0062,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0742,  0.6134, -0.0626, -0.0272,  0.0061,  0.0359,  0.0221,  0.0032,
         0.0048, -0.0159, -0.0221, -0.0404,  0.0017, -0.0225, -0.0480,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0834e-02,  4.2905e-01, -1.1261e-03, -8.6467e-02,  4.9619e-04,
         3.5289e-02, -3.5910e-02,  7.2588e-03, -8.6695e-03, -9.5674e-03,
        -1.9236e-04,  1.1827e-02, -2.2588e-02, -1.7679e-02,  3.0480e-03,
        -2.4439e-03,  5.5907e-04, -1.3951e-02,  5.9502e-06,  6.3181e-03,
         1.1301e-02,  3.0859e-03, -1.0834e-05,  8.6862e-03, -2.8718e-02,
         7.4818e-03, -7.1007e-03, -4.2976e-03, -1.2765e-02,  3.6228e-03,
         9.0360e-03,  3.1420e-04,  1.6468e-02,  1.0733e-02,  2.1942e-02,
         4.8124e-03,  1.0607e-03,  4.4318e-03, -2.6909e-04,  3.5581e-03,
        -1.7679e-02, -5.3428e-04, -3.4294e-03,  1.0825e-02,  2.0775e-02,
         4.5982e-02, -2.7796e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8373e-02, -5.4785e-01, -1.3707e-03,  2.8541e-03, -3.0557e-02,
         5.6250e-02,  5.2889e-03, -2.9159e-03,  2.8062e-02,  9.2726e-03,
         1.0271e-02,  8.4697e-03, -6.9294e-04,  2.5922e-03,  1.5019e-03,
        -1.6548e-02, -2.5649e-03,  3.9298e-02, -1.7248e-03,  4.5119e-04,
        -3.0249e-03, -4.5748e-03,  3.0941e-03,  1.3056e-03,  1.7773e-02,
         1.5463e-02,  1.2878e-02,  2.2419e-03, -9.1377e-04, -4.0655e-03,
        -3.1099e-05,  2.5809e-03,  8.3421e-03, -1.9042e-03, -9.1568e-03,
        -8.9252e-03, -2.0429e-03,  5.1694e-04,  2.4237e-03,  2.6243e-03,
        -9.1323e-04, -8.8292e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0299,  0.5421,  0.0515, -0.0103, -0.0258, -0.0342, -0.0069, -0.0343,
        -0.0037, -0.0185,  0.0379, -0.0109, -0.0106,  0.0135,  0.0037,  0.0090,
        -0.0260,  0.0106,  0.0068, -0.0030, -0.0238, -0.0117, -0.0014, -0.0260,
        -0.0480,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7634e-02, -6.5661e-01, -7.1601e-02, -2.5775e-03,  7.7250e-03,
        -2.9045e-02,  3.7161e-02,  2.6061e-02,  1.1334e-02, -2.4880e-03,
         8.1103e-03, -2.0900e-03, -1.5641e-03,  9.5730e-03, -4.3035e-04,
        -1.4711e-02,  3.1436e-02, -3.9570e-03, -2.3117e-03,  4.0895e-03,
         5.4157e-03, -2.5694e-02, -1.8385e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.3037, -0.3882,  0.0089, -0.0040, -0.0444,  0.0274,  0.0151, -0.0051,
         0.0254, -0.0438,  0.0116,  0.0235,  0.0242, -0.0188,  0.0093, -0.0082,
         0.0058,  0.0021,  0.0121, -0.0058, -0.0034,  0.0052,  0.0042,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0613,  0.5308, -0.0379,  0.0279, -0.0551, -0.0079,  0.0279,  0.0449,
         0.0224, -0.0055,  0.0694,  0.0352,  0.0738,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1191, -0.4785,  0.0474,  0.0231,  0.0094,  0.0199,  0.0149,  0.0150,
         0.0262,  0.0209,  0.0444,  0.1085,  0.0581,  0.0144,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4707e-02,  6.1481e-01,  5.1074e-02,  1.8906e-02,  2.6293e-02,
         1.8337e-02,  2.3433e-04, -7.9605e-04,  9.4524e-03,  1.4770e-02,
        -7.3850e-03,  5.2333e-04,  3.3823e-02,  1.7704e-02,  2.2654e-02,
         1.6261e-02, -3.3156e-02, -5.1429e-03,  6.3976e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-1.7587e-01, -3.8534e-01, -3.1614e-02,  1.5150e-01, -1.6173e-02,
         1.0892e-02, -1.2518e-02, -6.9963e-03,  9.5255e-03, -2.5527e-03,
         1.3518e-02, -7.5052e-03, -1.9575e-03,  5.4776e-03,  2.1865e-02,
        -5.0837e-03,  4.6585e-03,  2.9923e-02,  1.5075e-04, -5.4481e-03,
         5.4023e-03,  1.9182e-02, -2.5848e-03,  7.4731e-04, -2.5101e-02,
         4.8415e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0293, -0.6598, -0.0405,  0.0264,  0.0295, -0.0069,  0.0350,  0.0328,
        -0.0009, -0.0083,  0.0188,  0.0102,  0.0388, -0.0140,  0.0462,  0.0026,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1605,  0.5548,  0.0314,  0.0648,  0.0221,  0.0350,  0.0168, -0.0052,
        -0.0008,  0.0147,  0.0127, -0.0182,  0.0082, -0.0011,  0.0180,  0.0166,
         0.0192,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0071, -0.6321,  0.1261, -0.0045, -0.0513,  0.0437,  0.0189,  0.0578,
        -0.0586,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0357,  0.4251, -0.0355, -0.0539, -0.0106, -0.0063,  0.0146,  0.0324,
         0.0312,  0.0164,  0.0265,  0.0190, -0.0113, -0.0107,  0.0076, -0.0104,
         0.0027, -0.0096, -0.0102, -0.0144,  0.0056, -0.0035,  0.0005, -0.0137,
        -0.0030, -0.0058, -0.0118,  0.0044,  0.0087,  0.0049,  0.0059,  0.0005,
         0.0254, -0.0070, -0.0015,  0.0013,  0.0085,  0.0062, -0.0037,  0.0111,
         0.0066,  0.0286, -0.0041, -0.0019, -0.0134,  0.0285], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0750, -0.5808,  0.0416, -0.0013, -0.0075, -0.0053, -0.0084,  0.0375,
         0.0152, -0.0463,  0.0133,  0.0084, -0.0021, -0.0032,  0.0011,  0.0048,
         0.0029,  0.0228,  0.0101, -0.0081, -0.0028,  0.0028, -0.0027,  0.0090,
         0.0093,  0.0007, -0.0383, -0.0162,  0.0059,  0.0164,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0638, -0.7866, -0.0341,  0.0543, -0.0168, -0.0071, -0.0373,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.2481e-02, -8.5338e-01, -3.8195e-03,  9.0043e-03,  1.6356e-03,
        -5.3802e-05,  6.6203e-05,  1.4675e-02, -2.0501e-02,  4.6314e-04,
         1.3915e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0418,  0.5392,  0.0289,  0.0802, -0.0249,  0.0360, -0.0044, -0.0293,
        -0.0008, -0.0148, -0.0021,  0.0012, -0.0543, -0.1422,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0552, -0.6193, -0.0512,  0.0025, -0.0159, -0.0131,  0.0008,  0.0214,
         0.0027,  0.0020, -0.0012,  0.0064,  0.0015,  0.0102,  0.0045,  0.0174,
        -0.0081,  0.0021, -0.0010, -0.0186, -0.0051, -0.0008,  0.0060,  0.0015,
         0.0075,  0.0139, -0.0107,  0.0051,  0.0053,  0.0234,  0.0271,  0.0066,
        -0.0099,  0.0220,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2411e-02, -6.7096e-01, -6.3766e-02, -1.3153e-03,  5.9607e-03,
        -1.5090e-02, -9.5040e-03,  1.1300e-02,  1.2809e-02, -1.5567e-02,
         3.9930e-03,  5.5768e-03, -6.8291e-04,  5.5703e-04,  2.9118e-02,
        -5.5986e-03,  7.8371e-03,  2.5739e-02, -3.4383e-03, -3.1907e-03,
         1.5140e-02, -1.6441e-02, -6.0506e-03, -6.1898e-03,  2.7915e-03,
         3.9399e-03,  9.3380e-03,  5.6913e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1213e-01,  3.1166e-01,  1.6562e-02, -1.3029e-01, -3.0046e-02,
         7.1847e-03,  9.5826e-03,  2.3097e-03,  2.2723e-03, -1.8856e-02,
         1.5116e-03, -1.8688e-02,  1.4718e-02,  2.0220e-02, -2.2758e-02,
         5.3392e-03, -4.1861e-02,  6.4992e-03, -5.5935e-03,  1.1931e-02,
         3.0044e-04, -4.5288e-03,  8.9140e-02, -1.6007e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 4.0062e-02, -5.1124e-01, -4.5478e-02, -4.8179e-02, -2.6738e-02,
         5.5105e-03,  1.7284e-02, -2.9642e-03, -1.7715e-05, -1.7080e-02,
        -1.1053e-02,  3.0194e-02, -1.6317e-02, -1.5963e-03, -2.1304e-03,
        -1.5604e-02, -5.1675e-03,  1.0625e-02,  3.1164e-02,  2.5954e-04,
         9.9191e-03,  4.6614e-03,  1.1668e-02,  1.3039e-02,  4.9544e-03,
        -4.9741e-03,  1.7378e-02, -2.8079e-03,  1.9901e-02,  6.3449e-03,
         6.5683e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0058, -0.5620,  0.0594,  0.0565,  0.0474,  0.0202, -0.0143,  0.0449,
         0.0524,  0.0217,  0.0375,  0.0779,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0079, -0.7024, -0.0271,  0.0327, -0.0177, -0.0068,  0.0218,  0.0062,
        -0.0138,  0.0050, -0.0127,  0.0067, -0.0397,  0.0379,  0.0616,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0147, -0.5777, -0.0156,  0.0251, -0.0352, -0.0258,  0.0661,  0.0010,
        -0.0226, -0.0069, -0.0078,  0.0039, -0.0063,  0.0163,  0.0146, -0.0116,
        -0.0105,  0.0037,  0.0037,  0.0062, -0.0027,  0.0031, -0.0218, -0.0057,
        -0.0269,  0.0119,  0.0101,  0.0047,  0.0051,  0.0103,  0.0099, -0.0007,
        -0.0118], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0515, -0.6767, -0.0144,  0.0301, -0.0088,  0.0072,  0.0020, -0.0076,
         0.0061, -0.0166, -0.0168, -0.0012, -0.0169,  0.0860,  0.0231,  0.0352,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0324,  0.5278,  0.0294, -0.1853, -0.0417, -0.0105,  0.0348,  0.0489,
         0.0891,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3146e-01,  6.6314e-01, -2.7644e-02, -1.4013e-04, -7.7623e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3472e-02,  4.3731e-01, -3.9984e-02, -1.3755e-02,  9.3831e-03,
         7.3370e-03,  1.0791e-02, -1.3823e-02, -3.0392e-02, -3.0146e-03,
         1.4304e-02,  2.0479e-02, -2.3720e-02,  7.6805e-02,  1.0437e-02,
         1.6863e-02,  1.8493e-02,  7.3342e-03,  2.9232e-02,  1.6671e-02,
         3.3449e-03,  1.6531e-02,  1.5700e-02, -1.7190e-02,  4.2921e-04,
         2.3896e-02,  3.9310e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0406,  0.8194, -0.0155,  0.0540, -0.0365, -0.0209,  0.0131,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3504e-03,  6.2027e-01, -4.1456e-02,  1.7909e-03,  9.9966e-03,
         3.3956e-02,  3.6346e-03,  1.0149e-02,  4.0538e-03, -1.5262e-02,
         1.3848e-03,  3.1261e-02, -2.9913e-03,  4.6138e-04, -1.4372e-02,
        -1.3500e-02,  2.0747e-02, -2.9057e-02, -8.8445e-03,  2.2377e-02,
        -6.7983e-02, -4.3098e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0371,  0.5728,  0.0247,  0.0041, -0.1271, -0.0039, -0.0284, -0.0039,
         0.0008, -0.0559, -0.0185, -0.0207, -0.0195,  0.0064, -0.0156,  0.0094,
        -0.0068, -0.0366,  0.0078,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0318,  0.6625,  0.0089, -0.0060, -0.0042, -0.0484,  0.0119,  0.0042,
         0.0068, -0.0065, -0.0086,  0.0050, -0.0045, -0.0128, -0.0015, -0.0068,
        -0.0367, -0.0050, -0.0041,  0.0105,  0.0156,  0.0123,  0.0016,  0.0088,
        -0.0028,  0.0041,  0.0007, -0.0062, -0.0021,  0.0065, -0.0525,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #300: [tensor([-1.4492e-01,  4.4301e-01,  3.9700e-02, -1.5197e-02,  6.6556e-03,
        -5.7837e-03,  9.2140e-04, -1.0016e-03, -7.0677e-03, -7.3964e-03,
        -8.5114e-05,  1.5481e-02, -6.8306e-04,  1.6182e-02,  1.4409e-04,
         3.1564e-02,  6.3177e-03,  3.9360e-03, -2.3030e-02, -3.1243e-03,
        -7.2529e-03, -7.5042e-04, -1.4492e-02, -9.6446e-03,  2.7237e-02,
         1.0267e-02,  1.4423e-02, -7.1112e-02, -5.8774e-03,  2.7739e-02,
        -1.4794e-02,  2.4210e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0150,  0.5917,  0.1172,  0.0030, -0.0092, -0.0106,  0.0027, -0.0095,
         0.0376,  0.0102,  0.0278,  0.0073,  0.0077, -0.0179,  0.0067, -0.0179,
        -0.0283,  0.0110, -0.0066,  0.0075,  0.0357,  0.0190,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2385e-03,  6.7743e-01, -1.4177e-02, -4.9673e-02, -6.9523e-03,
         3.7242e-03, -3.2629e-02,  2.5876e-02,  1.6668e-02, -1.0136e-02,
         3.9201e-03,  1.3989e-02,  9.9958e-03,  1.4053e-03, -8.9446e-05,
        -1.1482e-02,  1.0434e-02,  5.9126e-03, -1.0872e-02,  2.1583e-03,
         6.4679e-03, -1.7557e-02, -1.1851e-02, -2.6934e-03, -1.5184e-03,
        -6.7258e-03,  8.9645e-03,  1.1183e-02,  6.4341e-03, -1.6846e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0605,  0.6830,  0.0311, -0.0032,  0.0195,  0.0076,  0.0185,  0.0047,
        -0.0084, -0.0280,  0.0076,  0.0081,  0.0245, -0.0121,  0.0062, -0.0258,
        -0.0117,  0.0395,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3731e-02,  5.2889e-01,  1.1010e-01,  3.6747e-02, -2.0917e-02,
         4.2281e-02,  2.5899e-02, -1.0253e-02,  6.2205e-03, -6.1182e-03,
        -1.4752e-02, -2.0803e-03, -3.7012e-03, -1.2698e-02,  2.9909e-03,
        -5.1374e-04, -9.1850e-03, -2.8645e-02,  8.2804e-03, -1.0251e-02,
        -1.2130e-02, -1.3742e-02,  8.4979e-03,  7.0556e-03,  5.3835e-03,
         8.3661e-04, -6.3223e-03, -3.4173e-04,  6.5056e-03,  4.0099e-03,
         1.6236e-03, -6.0158e-04, -1.7712e-03, -1.0363e-02,  6.1377e-03,
        -1.6083e-02,  4.3379e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6971e-02, -6.1752e-01, -2.0236e-02, -3.9769e-02,  8.8722e-03,
         6.0308e-03,  2.3239e-03,  2.5575e-02,  1.5375e-03,  6.5421e-05,
        -1.4348e-02,  1.9374e-02,  3.4468e-02, -1.1804e-02,  8.9349e-03,
         6.6673e-04,  1.0003e-02,  6.3032e-03,  5.7099e-03,  1.9048e-02,
         1.2067e-02, -3.3861e-02, -3.0585e-02, -2.3927e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.5956e-03,  4.7785e-01,  5.3372e-02,  3.0284e-02,  6.5994e-02,
         1.1746e-02, -4.1707e-04,  1.1860e-02, -1.7872e-03,  5.5469e-03,
        -1.2497e-02, -1.9241e-02, -8.4837e-03, -5.5189e-03,  1.5296e-02,
        -4.4043e-03, -5.6688e-03, -2.4767e-03,  5.7891e-04,  9.9608e-04,
        -4.3523e-03, -1.3033e-02, -7.9182e-03,  1.0092e-02, -5.5713e-03,
         1.2785e-02, -3.3892e-03,  6.0841e-03, -8.0944e-03,  3.8827e-04,
         2.4202e-02, -5.4862e-03, -9.1481e-03,  1.6420e-02,  1.1804e-02,
        -2.0664e-02, -1.9249e-02, -7.5079e-03, -7.7950e-03, -2.7752e-03,
        -8.8970e-03, -8.0333e-03, -3.7876e-03,  6.5706e-04, -3.3627e-02,
        -7.6239e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1166,  0.4386,  0.1151,  0.0197, -0.0346,  0.0030,  0.0035,  0.0034,
        -0.0081,  0.0156, -0.0177,  0.0057,  0.0386, -0.0137, -0.0156,  0.0018,
        -0.0135, -0.0100, -0.0204,  0.0012,  0.0209, -0.0377, -0.0450,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1120, -0.4804, -0.0412, -0.0553, -0.1176,  0.0072, -0.0152,  0.0757,
         0.0425,  0.0493,  0.0036,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1711, -0.5538, -0.1003,  0.0186,  0.0164, -0.0041, -0.0199, -0.0074,
        -0.0093, -0.0083, -0.0170,  0.0075,  0.0344,  0.0320,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0679,  0.6602,  0.0586, -0.0009,  0.0046, -0.0061, -0.0262, -0.0083,
        -0.0143, -0.1048,  0.0336,  0.0146,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0205e-02,  5.0366e-01,  4.0291e-02, -3.9258e-02,  5.5656e-03,
        -2.1999e-03, -5.9756e-03,  2.0642e-02,  2.3961e-02, -2.4045e-03,
        -4.4948e-03, -2.1691e-02, -3.6197e-03, -6.9406e-03, -6.0486e-03,
        -1.0018e-02, -8.4552e-03, -9.8060e-03, -1.1483e-02, -3.3321e-02,
        -7.7440e-03, -5.4951e-03, -8.2547e-03,  1.1867e-02,  1.1383e-02,
         3.9676e-04, -9.3403e-04, -5.7892e-03,  5.1751e-03, -7.4803e-03,
         4.5209e-03, -6.8150e-03, -1.4758e-03,  3.0046e-03, -1.0508e-02,
        -7.2009e-03, -2.6467e-03,  1.6477e-04, -6.8292e-03, -1.0722e-02,
        -2.1109e-03, -1.2728e-03,  1.4666e-02, -1.3143e-03, -7.2483e-03,
         1.2054e-02, -1.0323e-03, -8.5680e-03, -4.9955e-03, -1.1084e-02,
         1.4694e-03,  6.5101e-03, -4.6599e-03,  2.1660e-03,  2.3967e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-0.0037, -0.6822, -0.0793,  0.0150,  0.0279,  0.0194, -0.0023,  0.0105,
        -0.0125, -0.0160, -0.0065,  0.0026, -0.0071, -0.0112, -0.0198,  0.0074,
         0.0189,  0.0091, -0.0008, -0.0104,  0.0041, -0.0049,  0.0283,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0264, -0.6028, -0.0579,  0.0143,  0.0151, -0.0172,  0.0116, -0.0412,
        -0.0070,  0.0103,  0.0226, -0.0025,  0.0292,  0.0454,  0.0234,  0.0233,
         0.0338, -0.0161,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1417,  0.5863,  0.0248, -0.0285,  0.0058,  0.0346,  0.0078, -0.0179,
        -0.0268, -0.0247, -0.0188, -0.0130,  0.0178, -0.0063, -0.0204, -0.0130,
        -0.0036,  0.0083,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0818, -0.5119,  0.0063, -0.0014,  0.0024,  0.0160,  0.0379,  0.0987,
         0.0388,  0.0108,  0.0234,  0.0405, -0.0238,  0.0428,  0.0637,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.4613e-03,  5.5266e-01,  4.6036e-02,  2.1407e-02,  1.3355e-02,
         3.3011e-03,  2.5476e-02, -1.9676e-02,  6.4708e-03,  9.2507e-03,
         1.4362e-02,  5.1156e-04,  1.0450e-02,  1.5021e-02, -8.4721e-03,
         1.0243e-01, -1.3190e-02, -1.0977e-02,  5.8556e-04, -3.4509e-03,
         1.7874e-03, -1.5909e-02,  5.7381e-03, -1.5490e-03, -7.1474e-03,
         6.2093e-03,  3.9020e-03, -1.0748e-02,  1.6715e-04,  2.4668e-02,
         1.0769e-02,  2.5865e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1322e-02,  4.1804e-01,  7.8129e-02, -3.9687e-02, -1.7763e-02,
         3.1241e-03, -6.8920e-03, -1.6660e-02, -1.8334e-02,  4.5744e-03,
        -1.2136e-02, -1.9772e-02,  2.4702e-02,  2.5545e-02,  1.7604e-03,
         1.9774e-03,  2.7135e-03, -3.0900e-02, -9.1637e-03,  9.5246e-03,
        -4.1388e-02, -7.5234e-03, -3.5429e-02, -8.7248e-03, -8.9636e-03,
        -4.2676e-03, -1.4770e-02,  2.1230e-02,  2.7370e-04, -2.1027e-02,
        -1.7930e-03,  4.9376e-03,  4.6948e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.1573e-02,  5.1265e-01,  5.2060e-02,  5.6401e-02,  1.0039e-02,
         2.2202e-03, -3.5911e-03,  6.8153e-03, -4.2470e-02, -5.9916e-02,
        -3.9531e-03,  1.9236e-02, -7.1021e-03, -1.3424e-02, -3.1266e-03,
         1.6906e-02,  6.3759e-04, -2.9620e-04, -6.2886e-03, -4.6424e-03,
        -2.3222e-03,  2.7165e-04, -2.8052e-03, -1.7684e-02,  1.9356e-03,
        -2.0432e-03,  5.8087e-03,  1.0164e-02,  3.4891e-03,  8.2848e-05,
         1.5432e-03, -2.7280e-03, -7.3929e-04, -3.0540e-03,  1.8446e-03,
        -1.0796e-02, -2.9667e-03,  9.6981e-04,  7.2503e-04,  1.6704e-03,
        -6.4957e-04, -1.5205e-03,  6.2747e-04, -6.2088e-03,  2.1640e-03,
         8.4816e-03, -1.3180e-03, -1.0433e-02,  4.9318e-03, -3.6730e-03,
        -5.5925e-03,  1.4325e-03,  6.1268e-03,  1.5299e-03, -1.4250e-03,
        -2.4490e-04,  5.6515e-03,  9.9244e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0470, -0.5490, -0.0208,  0.0283,  0.0139,  0.0081,  0.0173,  0.0150,
         0.0094, -0.0007,  0.0406,  0.0071,  0.0022,  0.0289, -0.0093, -0.0078,
        -0.0149,  0.0104,  0.0023,  0.0100, -0.0016, -0.0216,  0.0198, -0.0107,
         0.0575,  0.0457,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.5558e-03, -6.4200e-01, -2.5923e-02, -5.4370e-02, -8.0084e-03,
        -1.2561e-02,  2.4912e-02,  5.6075e-03,  3.1288e-02,  3.5546e-02,
        -1.4069e-02,  7.5584e-05, -1.6655e-02, -3.0606e-03, -7.4716e-03,
         1.2202e-03,  1.9880e-02, -2.0882e-03,  8.3546e-04, -6.3720e-03,
        -4.7896e-03,  1.7033e-02,  4.8111e-02,  1.1304e-03,  9.4376e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0598e-01, -4.7116e-01, -4.5224e-02, -7.6483e-03,  4.4001e-02,
        -2.5565e-03,  4.0435e-03,  4.0025e-02,  9.1009e-03,  8.4462e-03,
         1.1418e-02,  3.3410e-03, -2.2008e-03, -4.0595e-02, -2.3360e-02,
         8.4135e-03, -3.0469e-03, -8.1426e-03, -1.2098e-03, -2.4691e-03,
         3.2207e-03,  4.6805e-03,  6.6998e-03,  9.3397e-04,  7.1161e-03,
         1.3478e-02,  6.4080e-04, -6.0307e-03,  3.5736e-03,  4.0736e-03,
         1.9160e-04,  1.8254e-03,  1.4088e-02,  3.9787e-03,  4.3703e-03,
         1.1007e-04,  5.5795e-03, -1.1199e-02,  4.7361e-03,  2.6565e-03,
         5.1358e-03, -2.0941e-03, -6.2950e-03,  2.7901e-02, -1.7008e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0294,  0.5597,  0.0247,  0.0455,  0.0536, -0.0247,  0.0064, -0.0047,
        -0.0369, -0.0029, -0.0143, -0.0163, -0.0060, -0.0124,  0.0100,  0.0167,
        -0.0228, -0.0056, -0.0143,  0.0181,  0.0328, -0.0347, -0.0075,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0122,  0.7148, -0.0938, -0.0148,  0.0029,  0.0214,  0.0603, -0.0016,
        -0.0376,  0.0008, -0.0195,  0.0203,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 1.3548e-02, -5.6514e-01, -1.2451e-01, -3.2254e-02, -4.9086e-03,
        -5.6219e-02, -2.8067e-02,  3.2406e-04, -2.9501e-02,  1.8489e-02,
         1.1486e-03, -1.2117e-02, -3.1140e-03, -3.3171e-03,  5.0220e-03,
        -2.5550e-02, -9.4333e-04,  1.0269e-02,  5.9534e-03,  6.5260e-03,
         8.2510e-03,  7.9568e-03,  3.6875e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.2233e-02,  4.7371e-01,  2.9975e-02, -7.7503e-03, -6.1333e-02,
         5.9321e-03,  2.8151e-02, -7.5510e-03,  2.0394e-02,  3.8620e-03,
        -3.0589e-03,  7.1459e-04, -3.6202e-03,  1.8689e-02, -1.5145e-02,
        -5.1155e-03,  1.3851e-02, -3.4750e-02,  1.0248e-02,  2.5444e-04,
         1.4221e-02,  3.6239e-03,  1.1232e-02, -1.0243e-02,  3.9959e-03,
         6.5572e-03, -1.4912e-03,  8.5725e-03,  3.7816e-03,  1.0743e-02,
         3.0099e-03,  2.8025e-02,  6.8160e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0668e-01, -5.6031e-01, -9.8231e-03,  5.2641e-04, -3.7440e-02,
        -2.1741e-02,  1.9320e-03, -1.3524e-02, -3.9882e-02, -5.9155e-03,
        -2.8766e-02,  1.1678e-02,  2.0000e-02, -1.0982e-02, -3.2968e-02,
         5.0355e-02,  1.0022e-02,  2.4228e-02,  1.3228e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1141, -0.6852, -0.0021,  0.0012, -0.0076, -0.0360,  0.0307,  0.0183,
         0.0051,  0.0007,  0.0249, -0.0518,  0.0152,  0.0071,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.9932e-03,  6.8320e-01, -1.8269e-02, -1.8223e-03,  2.0271e-02,
         6.0406e-03, -2.0730e-02,  5.6631e-03, -1.1081e-02, -4.1326e-03,
         2.1693e-02, -3.8349e-04,  1.9385e-02, -4.2405e-02, -8.5775e-03,
         5.6604e-03,  2.3765e-02,  1.6917e-02,  2.5402e-03,  1.4431e-02,
         3.7038e-02,  2.9007e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0815,  0.6091,  0.0107,  0.0034, -0.0135,  0.0045,  0.0209, -0.0182,
         0.0275,  0.0094, -0.0121, -0.0262, -0.0183, -0.0057, -0.0009, -0.0022,
        -0.0239, -0.0019,  0.0055, -0.0166, -0.0014,  0.0213, -0.0093, -0.0055,
        -0.0084, -0.0097, -0.0105,  0.0042,  0.0117,  0.0060,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.7154e-02,  4.8124e-01,  5.5660e-02,  6.6284e-02,  2.3851e-02,
         9.7391e-02,  1.7613e-04, -1.9512e-02,  3.6383e-02,  1.7178e-02,
         3.4229e-03, -3.1964e-02, -1.0893e-02, -2.4839e-02,  7.5586e-03,
        -2.3658e-02, -1.2838e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5295e-01, -6.3271e-01, -4.8329e-02, -2.0780e-02, -4.4540e-04,
         1.5343e-02,  4.5859e-02, -1.2608e-03,  1.3235e-02,  3.1635e-02,
        -3.7455e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6169e-02,  6.4451e-01,  6.5528e-02, -9.7101e-03,  4.5388e-05,
        -7.2203e-02,  1.8058e-02,  3.4289e-03,  5.3956e-03,  1.5705e-02,
        -8.5827e-03, -9.4941e-03, -7.0764e-03,  2.4351e-02,  2.4606e-02,
        -1.0476e-02,  5.4659e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0335,  0.6254,  0.0164,  0.0142,  0.0009,  0.0054, -0.0068,  0.0036,
         0.0059, -0.0134, -0.0128,  0.0063, -0.0182,  0.0029, -0.0604, -0.0038,
         0.0055, -0.0134, -0.0299, -0.0009, -0.0086, -0.0236,  0.0049, -0.0124,
        -0.0033, -0.0675,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0902,  0.5149,  0.0091, -0.0238,  0.0094, -0.0282, -0.0428,  0.0130,
         0.0250, -0.0044,  0.0179, -0.0017, -0.0115, -0.0010, -0.0008,  0.0020,
        -0.0085,  0.0156, -0.0134, -0.0027, -0.0150, -0.0161, -0.0007,  0.0093,
        -0.0277, -0.0127, -0.0038, -0.0049, -0.0312, -0.0426,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9514e-02, -6.2958e-01, -3.1800e-02,  1.5812e-02, -5.8202e-02,
         5.0629e-03, -2.5347e-04,  2.3956e-02,  5.2424e-03,  8.6864e-03,
         1.5741e-02, -3.5565e-03,  2.1689e-02, -1.5571e-02,  2.0968e-02,
        -1.1780e-02,  3.2502e-02, -1.4370e-02, -4.5719e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 2.8895e-03,  5.7822e-01,  6.7201e-02,  6.4154e-02, -7.1620e-03,
        -1.1972e-03,  9.0841e-02,  7.5999e-03,  2.6568e-02,  4.6085e-03,
        -1.9432e-03, -2.6757e-03,  1.0160e-02,  1.2748e-02,  4.2023e-04,
        -1.8362e-02,  8.5003e-03,  1.7651e-02,  4.9813e-04,  6.8244e-03,
        -3.3189e-03, -3.9850e-03,  3.8299e-03,  2.6006e-03, -1.7024e-02,
         1.4804e-02,  2.4217e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0766,  0.5541,  0.0228, -0.0566, -0.0509, -0.0032,  0.0061,  0.0058,
        -0.0060, -0.0133, -0.0066, -0.0312,  0.0899, -0.0769,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0621,  0.5464,  0.0217,  0.0236,  0.0087, -0.0080, -0.0099, -0.0075,
         0.0083,  0.0035, -0.0410,  0.0097, -0.0195,  0.0071, -0.0245,  0.0019,
        -0.0127, -0.0099, -0.0100,  0.0098, -0.0090, -0.0048, -0.0195, -0.0264,
        -0.0239, -0.0706,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1374, -0.5541, -0.0548,  0.0615, -0.0065,  0.0023,  0.0116,  0.0072,
         0.0458, -0.0090,  0.0197,  0.0031, -0.0282, -0.0588,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6394e-02,  5.1158e-01,  5.2324e-03,  1.5246e-03,  8.9267e-03,
        -2.9435e-02,  1.0489e-03, -1.4130e-02,  3.4481e-03, -9.2690e-04,
         5.3847e-03, -2.1853e-02, -1.0741e-02, -2.9405e-02,  2.1040e-02,
        -1.3979e-02, -2.3837e-02,  1.2500e-02, -2.0009e-03, -2.0327e-03,
        -1.0170e-02, -1.0722e-02, -1.3513e-02,  5.2447e-02, -2.6945e-03,
        -4.7434e-04, -1.0761e-02, -1.3769e-02,  4.9860e-03, -2.3312e-02,
         4.2147e-02,  5.9585e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0947,  0.6173,  0.0526, -0.0144,  0.0042,  0.0100, -0.0138, -0.0053,
         0.0108, -0.0083,  0.0152, -0.0021, -0.0161,  0.0021,  0.0039,  0.0027,
        -0.0211, -0.0111, -0.0173, -0.0073, -0.0086, -0.0129,  0.0032,  0.0092,
        -0.0357,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0901,  0.7597, -0.0315, -0.0507, -0.0226, -0.0454,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.2154e-02,  5.5600e-01, -3.3242e-03,  1.8746e-03, -1.9495e-02,
         2.5116e-02,  1.5637e-02, -2.4552e-03,  7.2745e-03, -1.0849e-02,
         8.5786e-03,  3.2452e-02,  2.5414e-02, -5.8165e-03,  6.7970e-04,
        -8.7672e-03, -3.4837e-02,  8.4542e-03,  1.3821e-03, -4.8776e-03,
        -7.6097e-03, -4.7622e-04,  1.7961e-02, -5.5530e-03, -1.0820e-03,
        -6.8370e-04,  5.8121e-03, -1.1891e-04, -1.1707e-03, -9.8820e-03,
         1.7567e-02,  1.0322e-02,  4.7645e-03,  8.9877e-03,  1.0291e-02,
         2.5166e-02,  1.7110e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.8195e-03, -4.2382e-01,  2.1521e-02, -8.3804e-02,  2.7214e-02,
        -2.5940e-02, -1.7088e-02,  1.7457e-02, -2.0462e-03, -3.5643e-03,
        -6.5725e-03, -1.7719e-02, -4.7432e-03,  1.7987e-03, -2.8446e-02,
        -2.7411e-02,  9.8578e-03,  1.0944e-02,  6.4141e-03,  2.3278e-04,
         2.4776e-02,  9.4661e-03,  5.5822e-03,  5.5386e-02, -3.2409e-03,
         9.9858e-03, -1.7716e-02,  1.3323e-02,  1.7169e-03, -4.0290e-04,
        -7.3631e-03,  1.1159e-02, -1.7231e-02,  1.3341e-02, -6.5891e-02,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0626, -0.5866,  0.2331, -0.0047, -0.0036, -0.0116,  0.0063, -0.0028,
         0.0012, -0.0090, -0.0027,  0.0094,  0.0153, -0.0278, -0.0234,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0236,  0.5275,  0.1025, -0.0373, -0.0196,  0.0135,  0.0071, -0.0195,
         0.0245,  0.0283, -0.0054, -0.0047,  0.0006,  0.0046,  0.0115,  0.0182,
        -0.0007,  0.0339,  0.0023,  0.0055,  0.0008, -0.0073, -0.0083, -0.0024,
        -0.0015, -0.0078,  0.0047,  0.0017, -0.0058, -0.0119,  0.0049,  0.0259,
         0.0262,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0244,  0.6350,  0.0498,  0.0086,  0.0068,  0.0059, -0.0017,  0.0029,
        -0.0086, -0.0042, -0.0112, -0.0124, -0.0058, -0.0032, -0.0282, -0.0110,
        -0.0007,  0.0008,  0.0026, -0.0189, -0.0255, -0.0042,  0.0016, -0.0057,
        -0.0072, -0.0039, -0.0188, -0.0230, -0.0675,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 0.1013,  0.6168, -0.0349, -0.0358,  0.0213,  0.0078, -0.0161, -0.0183,
        -0.0050, -0.0296, -0.0113, -0.0108, -0.0069,  0.0020, -0.0823,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.3691e-02,  5.8253e-01, -7.3857e-03, -3.6510e-02,  3.1370e-02,
        -3.0784e-03,  1.7267e-02,  6.7212e-03, -1.5870e-02, -1.9815e-02,
        -4.4412e-03, -8.6213e-03, -9.6195e-03,  4.9075e-03,  3.0266e-03,
         5.9385e-03,  2.3211e-03, -1.2138e-03,  2.7110e-03, -6.5285e-03,
         8.9505e-03, -7.3632e-03, -4.6920e-03, -6.3336e-03, -3.6633e-03,
        -1.5617e-03,  4.1261e-03, -1.4122e-02, -3.4502e-03,  2.6469e-03,
         2.3946e-03, -1.3267e-03,  1.2141e-03, -4.4485e-03, -1.8973e-03,
        -2.5025e-03,  5.2664e-03,  3.0330e-03,  6.3479e-03, -2.8033e-03,
         2.4980e-04,  2.2350e-03,  4.8983e-03, -5.8053e-03,  2.2966e-03,
         1.0330e-02,  1.7249e-03, -8.0014e-03,  3.2752e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6187e-02, -4.1076e-01, -1.4824e-02,  4.9137e-02,  2.4073e-02,
        -2.7038e-02,  1.0302e-01,  4.2763e-02,  6.3302e-02,  5.0802e-02,
         5.8740e-03, -7.0790e-03, -9.2417e-03,  2.5221e-03,  1.1600e-02,
         8.2804e-03,  2.3241e-04,  2.3286e-03,  2.2881e-03, -2.4649e-03,
         6.2417e-03,  1.5422e-02,  4.2711e-03,  9.1875e-03,  4.3524e-03,
        -1.6673e-03,  1.1229e-03, -1.7768e-02, -3.1230e-03,  9.1302e-03,
        -1.2163e-02, -5.1146e-03,  8.0984e-04, -5.2995e-02, -2.8195e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0849,  0.6599,  0.0375, -0.0009,  0.0208,  0.0070, -0.0228,  0.0155,
        -0.0178,  0.0117, -0.0446, -0.0076,  0.0025,  0.0193, -0.0047,  0.0207,
         0.0039, -0.0031,  0.0148,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0524e-02,  5.8568e-01, -2.4675e-02,  3.3420e-02, -3.1354e-04,
         1.8184e-02, -2.1612e-03, -1.2379e-02,  3.7092e-02, -1.8627e-02,
         4.4564e-03, -1.3677e-03,  1.2783e-02, -8.4538e-02, -8.3798e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0607, -0.6084,  0.0301,  0.0040,  0.0251,  0.0231,  0.0040,  0.0182,
        -0.0082,  0.0168,  0.0225,  0.0285,  0.0074,  0.0102,  0.0240,  0.0215,
         0.0020,  0.0238,  0.0009,  0.0007,  0.0506, -0.0094,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.7426e-04,  5.7034e-01,  4.6890e-02,  3.1811e-02,  2.7504e-02,
        -1.0047e-03, -4.7050e-02,  3.9028e-03, -6.3947e-03,  9.5490e-03,
        -1.0724e-02,  1.0666e-02, -5.4042e-04, -4.1501e-02, -2.6280e-02,
        -7.5106e-03, -4.1234e-03, -1.3483e-02,  1.1467e-02,  4.3210e-03,
         1.3905e-02,  1.0699e-02, -1.4186e-02,  4.2398e-03, -4.9847e-02,
        -7.4031e-03, -7.2846e-03,  1.8798e-03,  6.9179e-03,  7.7964e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.3105,  0.2532, -0.0768, -0.0157,  0.0287,  0.0138, -0.0812, -0.1029,
        -0.0022,  0.0424, -0.0727,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.2408e-02, -6.2853e-01,  2.2749e-02,  4.9093e-02,  1.7336e-02,
         2.5467e-02, -3.2050e-02, -2.9622e-03, -5.5617e-03,  9.9309e-04,
        -6.1303e-03, -1.3418e-02,  2.5524e-02, -7.2022e-03,  2.4432e-03,
        -2.9292e-04,  9.2932e-03,  1.1487e-02,  1.9209e-02, -7.8026e-03,
         2.0130e-02,  1.9921e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.6264e-02, -5.7209e-01, -9.9382e-02,  6.4576e-03, -4.6398e-03,
        -4.1542e-02, -1.3180e-02,  1.5417e-04,  1.8977e-04,  3.6106e-03,
         1.2607e-02,  3.0678e-02,  2.1577e-03,  1.1621e-02, -4.3878e-04,
         7.3570e-03,  5.2982e-04, -6.9032e-03, -2.7392e-03,  7.1103e-04,
        -3.0949e-03,  4.9049e-03,  1.5526e-02,  1.2886e-02,  4.5613e-03,
         4.1157e-03,  4.1861e-04,  8.6006e-03,  3.1877e-02,  1.1807e-02,
         1.1800e-03, -1.4960e-02, -2.2815e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.3525e-02, -5.8425e-01, -5.3754e-02,  5.1093e-02,  2.3222e-02,
        -1.0701e-02, -1.1685e-02,  1.3225e-02,  7.2684e-05,  3.7575e-03,
         8.5825e-03,  4.8716e-03,  2.1997e-02,  2.7077e-02,  1.7312e-03,
        -2.8530e-02,  8.5912e-03,  8.8103e-03,  2.7993e-03,  5.7756e-03,
         1.3510e-02,  8.7303e-03, -2.3774e-03,  1.5546e-03,  3.2999e-03,
        -2.3728e-03,  8.7732e-03,  2.5334e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6957e-02, -6.0917e-01, -1.1996e-01, -2.8318e-03,  4.7918e-03,
         1.1850e-02,  5.6064e-03,  2.4570e-02, -2.9338e-02, -2.3283e-03,
         9.8045e-03,  1.5657e-04,  1.7616e-02,  9.2842e-03,  1.0743e-02,
        -2.5889e-03, -4.9220e-03,  5.3971e-03,  1.0329e-02,  1.5900e-03,
         2.2944e-02, -5.1448e-03, -1.9601e-02,  4.0203e-03,  4.8452e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 3.6278e-02, -5.8314e-01, -2.3707e-02,  2.0632e-02,  2.9279e-02,
         1.3706e-02,  1.3686e-02,  9.4238e-03,  1.2709e-02,  9.6521e-03,
         1.1573e-02,  1.2661e-02,  4.1786e-03,  3.2296e-02, -4.8921e-04,
         1.1659e-02,  9.7830e-03, -8.0672e-03, -6.3254e-03, -7.9294e-04,
        -1.9664e-03,  1.1346e-02,  6.7089e-03,  1.7361e-03, -1.5656e-02,
         2.1308e-03,  8.6956e-03,  4.6859e-03,  5.1288e-03,  9.6720e-03,
         9.7230e-03,  7.0893e-03, -2.7531e-03,  4.2811e-03,  4.5057e-03,
         8.2701e-03,  8.1718e-03,  2.7215e-03,  8.6600e-03, -1.0720e-02,
        -1.5341e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0296e-02, -6.7861e-01, -1.0311e-01,  1.7143e-02, -2.3677e-02,
         7.3076e-03, -4.7102e-03, -5.6335e-04,  9.8012e-03, -9.6317e-03,
         5.3295e-03, -3.7752e-03,  3.6785e-03,  6.3734e-03, -1.0879e-02,
         3.8685e-03,  9.7507e-03, -2.5662e-03, -2.2809e-03, -1.4547e-02,
         7.2705e-03,  2.0546e-02,  1.2096e-04, -1.2817e-02,  2.1347e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0353,  0.5942,  0.0212,  0.0974, -0.0209,  0.0202,  0.0047, -0.0037,
        -0.0173, -0.0455,  0.0131, -0.0092,  0.0032, -0.0231, -0.0026, -0.0121,
         0.0011, -0.0273,  0.0108,  0.0179,  0.0191,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0502, -0.6841, -0.0330, -0.0324, -0.0095, -0.0080, -0.0262,  0.0051,
         0.0048,  0.0117, -0.0210,  0.0099,  0.0071, -0.0014,  0.0103,  0.0292,
         0.0227,  0.0333,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0342,  0.5534,  0.0409, -0.0280,  0.0115,  0.0019,  0.0099, -0.0485,
         0.0036, -0.0041, -0.0012, -0.0063, -0.0076,  0.0070,  0.0074, -0.0228,
        -0.0186,  0.0284, -0.0088,  0.0274, -0.0068, -0.0039, -0.0062, -0.0053,
         0.0036,  0.0145, -0.0055, -0.0032, -0.0136,  0.0125, -0.0039, -0.0032,
        -0.0111, -0.0063,  0.0052, -0.0185, -0.0009,  0.0046,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0085, -0.5880, -0.0503,  0.0255, -0.0259, -0.0063,  0.0128, -0.0111,
        -0.0210, -0.0070,  0.0099,  0.0015, -0.0101,  0.0310, -0.0265, -0.0195,
        -0.0152,  0.0479,  0.0373,  0.0447,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0353, -0.6563,  0.0645,  0.0409,  0.0102, -0.0022,  0.0076, -0.0024,
        -0.0130, -0.0226,  0.0308,  0.0367, -0.0582,  0.0194,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.4024e-02,  4.5495e-01,  2.7231e-02, -3.0952e-02, -2.1246e-02,
         4.8239e-03,  4.3966e-02, -3.9610e-02,  1.8365e-02, -1.1404e-02,
         2.5775e-02,  5.6546e-03, -8.7629e-03, -7.1960e-03,  1.6168e-02,
         1.4665e-02,  1.8797e-02, -8.7021e-03, -1.9154e-02, -2.2008e-02,
         9.1111e-03,  1.5611e-02, -7.0260e-03, -8.6233e-03, -1.0790e-02,
        -2.2173e-02, -2.5307e-03,  1.7959e-04,  1.8925e-03, -1.4371e-02,
        -1.5575e-02,  1.8659e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6653e-01, -5.6716e-01,  1.4835e-02, -1.9081e-02, -1.5715e-02,
         2.8128e-03, -1.2804e-02,  2.9274e-02,  8.2898e-03, -1.3052e-02,
         6.3824e-03,  9.4177e-03, -1.3764e-03, -1.1975e-02, -6.8658e-05,
        -6.0575e-03, -1.7195e-04,  3.4250e-02, -2.9990e-02,  5.0755e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1551e-02,  5.3073e-01,  4.4479e-02, -1.0263e-02, -2.0493e-02,
        -3.9627e-03, -2.0180e-02, -1.3538e-02,  6.9684e-03, -2.9706e-03,
        -1.2635e-02, -1.7918e-02, -8.1434e-03, -4.1865e-03, -2.4213e-03,
        -8.4011e-03, -2.2538e-03, -7.5367e-04, -1.6185e-02, -8.3874e-03,
        -1.2011e-02,  1.5109e-03,  1.5823e-03, -1.0978e-02, -6.8384e-03,
        -9.5761e-03, -1.3178e-02, -2.2116e-02, -2.3875e-02, -2.0000e-03,
        -6.1065e-03,  8.3323e-03, -5.1624e-03,  5.1145e-03, -2.2202e-02,
        -8.0166e-03, -3.3832e-03,  1.5903e-03, -7.2763e-03, -3.8909e-03,
         9.4025e-03,  1.0377e-02, -4.2288e-03,  1.0379e-02,  7.9531e-03,
         4.1401e-04,  2.6080e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0544, -0.6178,  0.0075, -0.0231, -0.0029,  0.0272,  0.0012, -0.0028,
         0.0594, -0.0212, -0.0105,  0.0533,  0.0379,  0.0482, -0.0326,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5304e-03,  5.7634e-01,  6.1543e-02, -4.4287e-03,  5.6185e-03,
         6.3963e-03,  6.1049e-03, -1.0641e-02, -6.2353e-03,  3.7200e-04,
         5.9810e-03, -3.9945e-04,  3.4618e-03,  2.3124e-02,  1.5154e-02,
        -5.1859e-03, -3.2637e-03, -1.5395e-02,  1.8031e-02, -5.9414e-03,
        -4.5712e-03, -1.4836e-02, -1.3707e-02,  7.1516e-04, -4.5377e-04,
        -1.7920e-02, -1.7250e-03, -2.1978e-02, -1.0561e-02,  1.4845e-02,
        -7.8263e-04,  3.0088e-03, -3.4219e-03, -6.3998e-03,  1.8697e-03,
         1.6156e-03, -3.5224e-03,  8.9417e-03, -3.6506e-03,  2.0133e-04,
         5.6132e-03,  4.7374e-03, -1.4366e-03,  8.7788e-03, -5.2514e-03,
         3.4524e-03,  2.7052e-02,  3.2800e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-0.1096, -0.5490, -0.0439,  0.0075, -0.0279, -0.0073,  0.0094, -0.0079,
         0.0259, -0.0017, -0.0072, -0.0195, -0.0161,  0.0015,  0.0103, -0.0037,
        -0.0023,  0.0130, -0.0034, -0.0041, -0.0061,  0.0078, -0.0017,  0.0192,
        -0.0784, -0.0157,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0304,  0.5710,  0.1221, -0.0120,  0.0088, -0.0362, -0.0344, -0.0015,
         0.0063,  0.0176,  0.0133,  0.0012,  0.0023,  0.0011, -0.0071,  0.0123,
        -0.0052, -0.0117, -0.0327, -0.0156,  0.0571,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5870e-02, -5.5717e-01, -9.8482e-02, -1.8566e-02,  5.5148e-02,
         1.0729e-02,  9.2804e-03, -1.6513e-02,  8.9997e-03,  1.7173e-03,
         6.1938e-03,  7.5210e-03,  1.0212e-02,  2.4869e-02, -1.7898e-03,
         2.4793e-03, -1.0769e-04,  2.8953e-02,  3.0308e-03,  2.1724e-02,
         1.5090e-02,  9.5718e-03,  4.4369e-03, -1.9389e-03, -2.5460e-03,
         6.9693e-03, -9.5179e-05,  8.6979e-03,  6.8443e-03, -3.3214e-03,
         1.0559e-02, -4.3144e-03, -4.9432e-03,  1.5168e-02, -2.5133e-03,
         3.6394e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0956,  0.4426,  0.0787,  0.0060,  0.0264,  0.0248, -0.0349,  0.0341,
        -0.0198,  0.0308,  0.0439,  0.0066,  0.0179, -0.0133, -0.0036,  0.0066,
         0.0019,  0.0385,  0.0098, -0.0644,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0025, -0.6663, -0.0784,  0.0016, -0.0142,  0.0171, -0.0217,  0.0030,
         0.0297,  0.0182, -0.0338,  0.0228,  0.0132,  0.0023,  0.0055, -0.0011,
         0.0038,  0.0200,  0.0042, -0.0274,  0.0133,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0682,  0.6522,  0.0684, -0.0218, -0.0145, -0.0120, -0.0179, -0.0166,
        -0.0258,  0.0310, -0.0414, -0.0222, -0.0080,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0268,  0.5633,  0.0523, -0.0136,  0.0214, -0.0308, -0.0417, -0.0027,
         0.0028,  0.0026,  0.0222, -0.0103, -0.0115, -0.0136,  0.0137,  0.0039,
        -0.0053, -0.0055,  0.0006,  0.0110,  0.0118, -0.0060, -0.0020, -0.0013,
         0.0050, -0.0200, -0.0061,  0.0090,  0.0012, -0.0030, -0.0024, -0.0073,
         0.0021, -0.0076, -0.0038, -0.0150,  0.0407], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0309, -0.6702, -0.0384, -0.0052,  0.0281, -0.0190, -0.0171, -0.0028,
        -0.0083,  0.0052, -0.0034, -0.0097, -0.0395,  0.0208,  0.0053, -0.0215,
         0.0009, -0.0061, -0.0065,  0.0032,  0.0085,  0.0071, -0.0171, -0.0249,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9206e-02,  5.3476e-01, -7.0617e-03, -4.2532e-04,  2.5219e-02,
        -3.5418e-03, -2.3139e-02, -3.1034e-03, -1.5508e-02,  2.3619e-02,
         9.4370e-03, -1.3382e-02,  3.6374e-02,  1.7351e-02,  2.0813e-02,
         2.9521e-02, -1.1715e-02,  6.0722e-02,  9.5149e-03,  1.1304e-04,
         4.3416e-03,  2.2715e-02,  3.2779e-03,  3.7128e-03, -2.1210e-02,
         2.9950e-03,  9.5380e-04, -1.1741e-02, -4.0597e-03,  9.5908e-03,
         4.7308e-02, -3.5680e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8644e-02,  6.0843e-01,  4.6339e-02,  7.9445e-03, -1.7594e-02,
         3.5137e-03, -1.8085e-02, -2.0327e-02,  2.5950e-02, -1.0416e-02,
         2.1886e-03, -1.9151e-03, -8.7620e-03, -1.2849e-02, -8.4640e-03,
        -3.7564e-03, -1.1814e-02, -1.7920e-03, -8.7806e-03, -1.6707e-02,
         2.3754e-03,  4.8548e-03, -1.4077e-03, -6.9263e-03, -4.5771e-02,
        -2.3370e-02, -2.8026e-03, -2.6111e-03, -8.5813e-05, -2.3371e-03,
         3.3191e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0565,  0.5900, -0.0873, -0.0137, -0.0066, -0.0080,  0.0082, -0.0050,
         0.0101, -0.0042,  0.0197,  0.0266,  0.0221, -0.0036,  0.0185,  0.0273,
         0.0050,  0.0181,  0.0219,  0.0012,  0.0259,  0.0078, -0.0125,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0706e-02,  6.8113e-01, -5.1184e-03,  3.7901e-02,  1.7224e-02,
         8.5213e-03,  7.4764e-03,  2.3430e-02, -1.9147e-02,  4.1837e-03,
         7.0799e-03, -1.5409e-02, -4.7559e-03, -1.1032e-02, -2.5793e-03,
        -4.9044e-03,  6.9847e-03, -4.4423e-04,  4.1339e-03, -3.0743e-02,
        -3.6724e-02, -1.2438e-03, -8.7271e-03,  4.6425e-03, -1.0554e-03,
        -5.7978e-03, -1.7763e-02, -1.1145e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-1.1710e-02,  6.4529e-01, -2.7092e-02, -9.7377e-03,  1.0872e-02,
         3.1399e-04,  1.3340e-02, -6.8882e-03, -1.0871e-02, -1.7949e-02,
        -1.6357e-02, -2.1008e-02,  1.2396e-02, -8.2363e-03, -7.0988e-04,
        -1.6933e-02, -8.5733e-03,  5.8016e-03, -1.6684e-02,  2.0974e-02,
         1.6414e-02, -2.3324e-04, -1.6375e-02, -4.5566e-03, -9.6675e-03,
        -1.2012e-02, -1.3337e-02,  3.0729e-02, -1.4938e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.0361e-03, -6.3746e-01,  5.9040e-03, -2.0804e-02, -1.7017e-02,
        -1.2264e-02,  3.8600e-03, -4.0881e-03,  7.2606e-02,  1.8512e-02,
         3.7426e-02,  3.1258e-02,  1.3702e-02, -3.9025e-04,  1.3694e-02,
         2.2473e-03, -2.9378e-06, -1.8536e-03, -1.3756e-02,  1.4139e-04,
        -3.1121e-03,  1.6602e-02,  6.7921e-02,  3.3911e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0047,  0.6018,  0.0290, -0.0292,  0.0268, -0.0114,  0.0083, -0.0035,
        -0.0236,  0.0034, -0.0252,  0.0031,  0.0054, -0.0017, -0.0268, -0.0021,
        -0.0095, -0.0064,  0.0179, -0.0041, -0.0032, -0.0182, -0.0014,  0.0069,
         0.0009, -0.0027, -0.0140, -0.0098, -0.0127, -0.0073,  0.0082,  0.0144,
        -0.0472, -0.0091], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.0818e-02, -5.9729e-01,  1.1311e-01,  2.7986e-02, -1.0993e-02,
         1.2343e-02, -6.9890e-02, -4.9911e-03,  1.0635e-02,  9.7996e-03,
        -6.2322e-03,  3.6040e-04, -5.4590e-03,  5.3474e-04, -7.7671e-03,
        -1.0508e-02, -2.1993e-03, -6.4431e-03,  7.2553e-03, -2.0135e-02,
         9.7872e-03,  5.4595e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.3623e-03, -4.9428e-01, -6.1397e-02, -2.4084e-03,  7.0443e-02,
         5.9760e-02, -5.4115e-02,  3.0645e-02, -4.3308e-04,  1.7994e-02,
        -1.0275e-01, -1.2777e-02, -9.6648e-03, -1.2307e-02, -1.7239e-02,
        -7.1980e-03, -1.7956e-02, -1.9271e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0495,  0.5870,  0.0369,  0.0225, -0.0289,  0.0116, -0.0206,  0.0341,
        -0.0118, -0.0235,  0.0113,  0.0039, -0.0150, -0.0015, -0.0086, -0.0128,
         0.0124, -0.0051, -0.0020, -0.0061,  0.0086, -0.0743,  0.0118,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.7115e-04, -5.9795e-01,  1.2556e-01, -1.1251e-01, -1.9913e-02,
        -2.4934e-02, -1.0628e-02,  1.8123e-02,  3.0282e-02,  1.1611e-02,
        -9.9041e-03, -1.8296e-03, -5.3972e-06, -1.6119e-02, -7.0135e-03,
        -3.5341e-03,  9.2150e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1325, -0.6026, -0.1211,  0.0186, -0.0060, -0.0013,  0.0123,  0.0310,
         0.0023,  0.0056, -0.0126,  0.0081, -0.0063,  0.0140,  0.0009,  0.0007,
         0.0032,  0.0208,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7244e-02,  6.0880e-01,  6.9070e-02, -2.6826e-03, -1.1230e-02,
         2.0656e-03, -8.9894e-03, -1.2467e-02, -1.7590e-04, -3.4431e-02,
         2.2395e-02, -4.0730e-03, -2.9587e-03, -5.3315e-04,  9.8569e-03,
         3.3138e-03, -1.4607e-02,  2.1883e-02, -3.0663e-02, -3.3185e-03,
        -9.0110e-03,  1.6745e-02,  1.5560e-02, -4.6111e-03, -2.5553e-02,
        -2.7760e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1017,  0.3620,  0.0802, -0.0258, -0.0285, -0.0087, -0.0918,  0.0109,
         0.0128, -0.0218, -0.0096, -0.0086, -0.0065, -0.0346, -0.0116,  0.0032,
        -0.0082,  0.0053, -0.0077, -0.0049, -0.0110, -0.0031, -0.0099, -0.0027,
        -0.0317, -0.0008,  0.0008, -0.0019,  0.0226,  0.0606, -0.0103,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0096,  0.5516,  0.0394,  0.0483,  0.0389,  0.0089,  0.0014,  0.0229,
        -0.0091,  0.0851,  0.0108, -0.0471,  0.0032, -0.0007,  0.0152, -0.0080,
         0.0283, -0.0713,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0916,  0.6592,  0.0338,  0.0361,  0.0259, -0.0144,  0.0786,  0.0604,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([-7.4788e-02, -6.1728e-01, -2.0566e-02,  2.9388e-02,  1.4543e-03,
        -1.7556e-02, -2.7970e-02, -6.3527e-03,  2.2494e-02,  2.1111e-03,
         1.2950e-03,  2.2283e-02, -8.6489e-03, -6.2752e-03, -6.2039e-03,
        -3.5566e-03,  3.3211e-03,  1.0249e-02,  5.9453e-03,  3.3021e-03,
        -6.8860e-03, -3.9557e-03, -8.4289e-03, -3.8400e-03,  4.3240e-03,
        -5.8541e-03, -3.2941e-04, -6.6809e-03,  1.2076e-02, -1.3282e-03,
        -1.3048e-02,  3.2986e-03, -9.3750e-04, -4.8295e-03,  4.2677e-03,
        -4.3353e-03,  7.5199e-03, -1.0595e-02,  6.4239e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0023,  0.5094, -0.0350,  0.0337,  0.0251,  0.0161,  0.0071,  0.0059,
         0.0206,  0.0145,  0.0276,  0.0194,  0.0198,  0.0206,  0.0006, -0.0042,
        -0.0183,  0.0211,  0.0134,  0.0259,  0.0587, -0.0320, -0.0079,  0.0081,
        -0.0040,  0.0020,  0.0243, -0.0229,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0261, -0.7984, -0.0321, -0.0092, -0.0120,  0.0191,  0.0161,  0.0079,
         0.0140,  0.0356, -0.0029,  0.0080,  0.0186,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7225e-02, -6.8532e-01,  1.7888e-02,  5.1747e-02,  1.0251e-02,
        -1.7915e-02,  4.8415e-02,  9.3989e-03,  1.1903e-02, -3.2077e-03,
         2.7626e-02,  3.1616e-02,  1.6112e-03, -4.5337e-04, -2.5418e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0821,  0.4057, -0.0296, -0.0752, -0.0037, -0.0252, -0.0184,  0.0047,
         0.0263,  0.0063, -0.0005,  0.0121,  0.0355,  0.0232,  0.0034,  0.0016,
        -0.0128,  0.0053, -0.0081,  0.0007,  0.0042, -0.0090, -0.0045,  0.0088,
        -0.0029,  0.0107, -0.0067,  0.0067,  0.0025,  0.0032, -0.0084, -0.0037,
        -0.0098,  0.0067,  0.0093,  0.0117,  0.0116,  0.0005, -0.0052, -0.0005,
        -0.0018,  0.0007, -0.0023, -0.0059, -0.0064, -0.0215, -0.0542],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1480e-02,  5.9321e-01,  5.5910e-02,  1.1107e-02, -1.0734e-02,
        -5.3078e-02, -1.6426e-02,  1.9745e-03, -7.5299e-03, -8.4210e-03,
        -9.4697e-03,  1.1491e-02, -1.0472e-02, -7.4449e-03, -3.2035e-03,
        -1.4862e-02, -7.0031e-03, -2.1788e-02, -7.0973e-03,  4.6901e-03,
         1.7043e-03, -4.8114e-03,  7.0624e-03,  4.1962e-03, -8.8500e-03,
        -6.3017e-03, -6.7502e-03,  5.1843e-04, -9.8531e-04, -1.6941e-02,
         1.1257e-03, -3.1558e-03, -2.2218e-03, -7.4977e-03,  5.4363e-04,
        -3.4001e-03, -9.0190e-03, -2.6050e-03, -5.2202e-03, -5.0608e-03,
        -2.2602e-03, -2.3745e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2281e-03,  6.5876e-01,  5.8509e-02, -5.1157e-03,  8.3733e-03,
         1.0767e-02,  3.4994e-02,  3.5336e-02, -1.0202e-02,  1.8010e-02,
         6.4700e-03, -3.5930e-02,  4.4965e-03, -2.1707e-04,  3.2145e-03,
         2.6093e-03,  9.1018e-03,  5.7154e-03, -6.3142e-03,  1.2656e-02,
        -1.3260e-02,  4.9686e-03, -1.4182e-02,  2.5317e-02,  1.4254e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2791e-02, -6.5210e-01, -6.4538e-02, -3.7113e-02, -1.8960e-02,
         4.8454e-03, -3.4589e-02, -9.1726e-03, -1.0582e-02,  1.2996e-02,
        -6.0254e-03,  2.4357e-02,  2.1932e-04, -4.6541e-03,  1.0700e-02,
         1.6626e-02,  2.1342e-02, -1.0622e-02,  2.8911e-02, -4.4328e-03,
        -6.0736e-03,  1.8100e-03,  6.5356e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1723e-02, -6.6292e-01,  7.9076e-03,  1.9693e-02,  3.2674e-02,
         2.0738e-02,  1.3935e-02,  2.3579e-02,  2.4972e-02, -5.8660e-03,
        -1.5823e-02, -1.6037e-04,  1.5002e-03, -1.1572e-02,  1.4635e-02,
        -6.3544e-03,  4.3299e-02,  1.3236e-03,  1.9639e-02, -1.9245e-03,
         9.6838e-03, -9.2054e-03, -3.0877e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7370e-01, -6.2435e-01, -2.2898e-02,  2.3275e-02,  7.2790e-03,
         7.3698e-03, -1.7011e-02,  1.7499e-04,  1.8566e-02, -3.8192e-02,
        -2.1131e-02, -3.1355e-02, -1.4698e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5989e-01,  5.4410e-01, -1.1614e-02,  3.4024e-02, -2.7307e-02,
        -3.9248e-02, -6.6427e-03,  6.2537e-03, -1.2453e-02, -2.7351e-02,
         5.4206e-04,  5.5498e-02, -2.0743e-02,  5.4327e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.2826e-02,  6.8140e-01,  3.5215e-02,  2.8777e-02, -1.9801e-03,
        -8.0088e-03,  8.9007e-03,  1.0061e-02,  1.8917e-04, -2.1220e-02,
        -3.1687e-02, -4.4419e-03,  1.9405e-02,  9.5348e-05,  3.9483e-03,
         1.5395e-03, -1.8611e-02,  1.4337e-02,  3.7354e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-1.1082e-01, -7.4380e-01,  3.3411e-03,  1.6237e-02,  8.6940e-03,
         1.0910e-02,  1.7412e-03, -2.5422e-03,  7.7193e-03,  8.9647e-03,
         1.0759e-03,  3.0225e-04,  3.6043e-04,  1.7449e-02, -3.4202e-03,
         2.0823e-03, -7.3537e-03, -2.2142e-03,  1.9668e-03,  4.3063e-03,
         1.0988e-02, -9.8705e-04,  2.5130e-03,  7.3371e-03,  1.0261e-02,
         1.2606e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0409, -0.7092, -0.0550,  0.0078,  0.0141,  0.0227,  0.0324, -0.0133,
         0.0182,  0.0236,  0.0098,  0.0126,  0.0032, -0.0087,  0.0255,  0.0029,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1164, -0.5136,  0.0508, -0.0384,  0.0131, -0.0290, -0.0391,  0.0234,
         0.0207,  0.0084,  0.0247,  0.0335,  0.0150, -0.0026,  0.0165,  0.0481,
        -0.0067,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0999,  0.4836,  0.1372,  0.0628,  0.0406,  0.0389, -0.0289,  0.0630,
         0.0450,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.6746e-02,  4.3162e-01,  7.1204e-02,  3.1311e-03, -1.5881e-02,
        -1.8848e-02, -1.1619e-02,  4.9224e-03, -1.7019e-02, -1.2108e-02,
        -3.2458e-05,  1.1103e-02, -1.1733e-03,  8.7567e-03, -2.4554e-03,
        -2.0544e-02, -5.5755e-03, -4.1724e-03,  1.0384e-03,  1.0635e-02,
        -2.8178e-03,  6.7046e-03, -1.2276e-03,  1.8705e-03,  8.9555e-04,
        -1.4981e-02,  1.2548e-02,  4.6073e-03,  3.1982e-03,  1.7992e-03,
        -5.2982e-03,  7.3926e-04,  1.0615e-03,  1.3369e-02, -9.2218e-04,
        -3.6217e-02, -1.7815e-03,  2.3160e-02, -5.9431e-03,  3.7609e-03,
        -1.7041e-02,  1.4451e-02, -1.1980e-02, -2.2741e-02,  6.5915e-03,
         5.5707e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1133,  0.5561, -0.0031, -0.0220, -0.0089,  0.0252,  0.0294, -0.0116,
         0.0057,  0.0056, -0.0076,  0.0019, -0.0049,  0.0056,  0.0023,  0.0191,
         0.0139,  0.0336, -0.0054,  0.0112, -0.0216,  0.0069, -0.0135, -0.0070,
         0.0079, -0.0022,  0.0240, -0.0017, -0.0133, -0.0151,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0224, -0.7624, -0.0015,  0.0141,  0.0721,  0.0907, -0.0367,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0254, -0.8093,  0.0242, -0.0212, -0.0181, -0.0262,  0.0186, -0.0183,
         0.0092,  0.0102,  0.0194,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0282, -0.6641, -0.0735, -0.0631, -0.0125, -0.0219, -0.0281, -0.0015,
        -0.0477, -0.0052, -0.0028, -0.0047,  0.0142, -0.0325,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1755e-02,  6.3522e-01, -1.9565e-02,  3.8154e-03, -1.1010e-02,
         6.4114e-03, -7.8486e-03,  9.5367e-03,  5.5790e-03, -3.4435e-02,
         1.3818e-02,  6.3947e-03, -5.3899e-04, -9.5187e-03,  1.0986e-02,
        -1.9134e-04, -2.8439e-04,  1.2108e-03, -9.7862e-03,  1.8487e-03,
        -7.4613e-03, -6.2775e-03, -1.5016e-02,  3.4843e-03,  1.7513e-03,
         1.9547e-02, -1.2961e-02, -6.6185e-03, -1.4333e-02, -3.1088e-03,
        -5.0693e-03,  3.9128e-03,  4.0765e-02,  2.9944e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.4539e-02,  5.2846e-01,  1.2984e-03,  2.9783e-02, -9.7487e-03,
        -8.7914e-03,  1.8209e-02, -5.9185e-05,  8.7120e-03,  1.9211e-02,
        -2.7904e-02, -7.0151e-03, -9.9217e-03,  3.9671e-03, -1.5762e-02,
        -1.3894e-02, -2.2236e-02, -2.7821e-02, -1.1091e-02,  3.5511e-03,
        -4.1303e-02, -5.0764e-03, -3.0277e-02, -2.4391e-03,  2.9263e-02,
        -2.5907e-02,  8.1356e-03,  3.5627e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0077, -0.5566, -0.0550, -0.0183, -0.0286, -0.0029,  0.0039,  0.0067,
        -0.0120, -0.0104,  0.0078,  0.0159, -0.0587,  0.0181,  0.0631,  0.0070,
        -0.0051,  0.0328,  0.0014,  0.0087,  0.0062,  0.0148, -0.0453,  0.0130,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([-0.1268, -0.5048, -0.0395, -0.0440,  0.0036, -0.0066, -0.0029, -0.0018,
         0.0005, -0.0036,  0.0376, -0.0203,  0.0218, -0.0063, -0.0073,  0.0014,
         0.0171,  0.0020,  0.0012, -0.0171, -0.0053,  0.0139,  0.0071, -0.0046,
        -0.0067, -0.0113, -0.0102, -0.0140,  0.0065,  0.0358,  0.0185,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0043, -0.6018,  0.1455, -0.0128,  0.1006, -0.0187,  0.0054, -0.0329,
        -0.0145,  0.0240, -0.0201, -0.0193,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0954, -0.6940, -0.0155, -0.0237, -0.0239, -0.0041,  0.0053,  0.0037,
        -0.0325,  0.0232, -0.0262, -0.0045, -0.0159, -0.0263,  0.0058,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3685e-02,  6.4167e-01, -2.9411e-02, -1.5888e-02, -2.0874e-02,
        -1.6704e-02, -2.7380e-02,  4.0558e-03,  1.9145e-04, -1.3472e-02,
         1.2834e-02,  5.5064e-03, -5.5439e-03, -9.5188e-03, -8.8600e-03,
         1.6646e-02, -2.3941e-03, -8.1600e-03, -1.1978e-02, -1.8897e-02,
        -2.3599e-03, -2.0643e-04, -6.8230e-03, -6.7228e-03,  6.8362e-03,
        -1.5962e-02, -3.6486e-03, -4.5171e-03, -6.1989e-03,  1.2661e-03,
        -9.9840e-03, -2.2437e-04,  4.1580e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3958e-02, -6.9314e-01,  2.4872e-02,  6.4998e-02, -2.4934e-02,
        -6.1297e-03,  1.8340e-03, -6.3384e-03,  1.2548e-02,  1.0824e-04,
        -1.9321e-02,  7.7308e-03,  8.2192e-03,  4.0843e-02,  3.5492e-02,
         2.9536e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0615, -0.6735, -0.0342,  0.1022,  0.0089,  0.0198, -0.0483,  0.0384,
        -0.0132,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0472,  0.7701,  0.0031,  0.1171,  0.0625,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0552, -0.5488,  0.0478, -0.0193, -0.0022,  0.0124,  0.0132,  0.0013,
         0.0236, -0.0106, -0.0253,  0.0031, -0.0481,  0.0268, -0.0040, -0.0176,
        -0.0104, -0.0196, -0.0091, -0.0016, -0.0171, -0.0095, -0.0118, -0.0058,
        -0.0220,  0.0203,  0.0135,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0236,  0.7002,  0.0645,  0.0317, -0.0625,  0.0236, -0.0938,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.7668e-03, -7.0985e-01, -2.6795e-02, -3.6525e-03,  1.0857e-03,
         3.7659e-03,  1.1929e-02, -4.5338e-04,  1.1222e-03,  2.3888e-02,
         1.1548e-02,  1.2502e-02, -6.7442e-03,  1.5363e-02, -1.5406e-03,
         8.3408e-03, -1.3381e-02,  1.2493e-03,  1.2733e-03,  2.6145e-03,
         9.8014e-02, -3.8119e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4151e-02, -7.0993e-01, -3.1603e-02,  2.6252e-03, -2.1143e-02,
        -5.9170e-03, -3.0753e-04,  1.9545e-02,  3.0810e-02, -1.3182e-02,
        -1.9214e-02,  3.3396e-03,  1.2852e-02, -1.3085e-02, -4.4767e-04,
         4.8862e-03,  2.2606e-02, -7.1521e-03,  4.7201e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0685, -0.6528, -0.0226,  0.0133,  0.0139,  0.0089, -0.0137, -0.0201,
         0.0148, -0.0018,  0.0042, -0.0046,  0.0032,  0.0216, -0.0044,  0.0039,
        -0.0013, -0.0014,  0.0007, -0.0143, -0.0015,  0.0082,  0.0153,  0.0025,
         0.0065, -0.0033,  0.0014, -0.0144,  0.0160,  0.0326,  0.0086,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 1.5302e-02,  5.5754e-01,  9.9122e-02, -3.5312e-02, -2.4757e-03,
         5.6344e-03,  1.4774e-03, -6.9049e-03, -1.8842e-03,  8.7404e-04,
         2.9518e-03, -5.5272e-03, -2.2887e-02, -1.2158e-02, -2.2942e-03,
         1.3862e-02, -2.4711e-02,  1.8135e-02, -4.0355e-02,  9.3193e-03,
        -7.2520e-03, -1.9587e-02, -4.4874e-03, -8.0750e-03,  3.2099e-03,
        -4.1492e-03,  4.4606e-03, -1.1921e-02, -3.0247e-02, -1.6109e-02,
        -1.1493e-02,  2.8038e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1814e-01,  5.7770e-01, -2.9931e-02,  2.2397e-03,  5.8938e-03,
        -2.0616e-02,  1.5118e-02,  5.3540e-03,  6.0340e-02,  2.0042e-03,
         2.4858e-03,  2.1165e-02, -9.0726e-03, -5.3613e-03,  3.2362e-02,
        -1.1801e-02, -2.3317e-03,  9.1647e-03, -3.0918e-03, -2.4135e-04,
         2.1928e-03, -6.3396e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0084, -0.5351,  0.0267, -0.0229,  0.0251, -0.0259,  0.0483, -0.0404,
        -0.0282, -0.0147,  0.0019,  0.0117, -0.0175, -0.0006, -0.0091, -0.0135,
         0.0120, -0.0048, -0.0294,  0.0041, -0.0142,  0.0093, -0.0102, -0.0011,
         0.0096, -0.0074,  0.0034, -0.0200,  0.0179,  0.0263,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0265,  0.6598,  0.0157, -0.0236,  0.0263, -0.0298, -0.0489,  0.0030,
         0.0059, -0.0009, -0.0089, -0.0289,  0.0039, -0.0123, -0.0070,  0.0160,
         0.0517,  0.0312,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5232e-02, -6.3070e-01, -3.9388e-02, -3.0864e-02, -4.1633e-04,
         4.2013e-02, -1.6371e-03,  3.9644e-03,  5.9525e-03, -3.6438e-03,
         7.3790e-03,  3.0858e-03, -8.3017e-03, -3.6612e-03, -1.9271e-02,
        -1.8825e-02, -2.3241e-02,  2.5198e-04,  1.1447e-02, -9.5001e-03,
        -2.6823e-05,  1.3033e-02,  1.4814e-02, -9.0300e-04, -5.6448e-03,
        -2.1959e-03,  6.7256e-03, -3.8778e-03, -1.2616e-02, -4.4157e-03,
        -1.0939e-02,  1.2215e-03,  2.2066e-03,  2.4488e-03,  2.4853e-03,
         1.1030e-02,  2.6642e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0205e-02, -6.3897e-01,  1.1807e-01,  1.6375e-02,  1.4293e-02,
        -2.2402e-04,  3.1284e-03, -2.8109e-02,  8.1923e-03, -6.1166e-03,
        -2.7705e-03,  4.4146e-03,  6.6656e-03, -1.5407e-02,  1.9786e-03,
        -1.0097e-02,  5.1461e-03, -1.6299e-02,  3.3423e-03, -1.0656e-02,
         1.3836e-02, -5.0391e-03, -1.3358e-02, -7.2998e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4795e-02, -4.4268e-01, -3.5971e-02, -3.7206e-02, -3.1861e-02,
         1.3140e-02, -1.9741e-02,  5.0628e-03, -7.1970e-02,  1.0848e-02,
        -5.9550e-03,  1.0022e-02,  1.3962e-02, -5.0304e-03, -4.3162e-03,
        -1.3008e-03,  7.6588e-03,  2.9385e-03,  1.6000e-03,  3.7053e-03,
        -1.1885e-03,  7.8614e-03,  1.0864e-02,  4.0072e-03, -1.3970e-02,
        -8.4441e-03, -7.9497e-03, -7.4703e-03,  4.8437e-03,  1.7705e-03,
        -3.2141e-02, -4.2184e-03, -3.0106e-03, -2.9960e-03, -9.7923e-03,
         1.3744e-02,  1.1728e-02,  2.9037e-03, -2.7564e-03,  4.1529e-03,
         5.3360e-02, -9.5969e-06,  2.3164e-04, -1.0933e-02,  1.3850e-02,
         1.6036e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2408e-01, -6.2705e-01,  1.2037e-04, -6.0109e-03,  6.9231e-02,
        -7.1168e-03,  1.0292e-02, -8.0323e-03, -1.5230e-03, -8.3160e-03,
        -9.3155e-04,  2.0452e-02, -7.3618e-03,  4.2397e-04, -5.9912e-04,
         2.7882e-03,  3.2726e-04, -4.6234e-03,  1.2183e-02, -7.4223e-04,
        -4.8981e-03,  5.5457e-02,  2.7438e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.4036e-03,  7.8481e-01, -4.1851e-05, -6.8988e-04,  5.1668e-02,
         3.4866e-03, -1.9148e-02, -8.6374e-03, -4.5457e-02, -2.6439e-03,
         7.5014e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0746e-02,  7.0026e-01,  5.2925e-02, -2.4248e-02, -4.0829e-02,
         2.5209e-02,  1.3581e-02,  1.2151e-04,  2.0843e-02, -1.2688e-02,
        -2.9882e-02, -1.3275e-02,  1.5392e-02, -2.0000e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0025, -0.6557, -0.0317,  0.1354,  0.0510,  0.0282,  0.0109, -0.0109,
         0.0014,  0.0564, -0.0030, -0.0129,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.3849e-02,  6.0358e-01,  4.4661e-02, -3.4181e-03, -4.1681e-03,
        -5.8218e-03,  7.0945e-04,  1.3737e-02,  7.4824e-03, -9.2427e-03,
        -3.4294e-03, -1.8267e-02,  1.0457e-02, -1.5847e-02,  2.6053e-03,
         6.1237e-03, -4.0555e-03,  7.0241e-03, -4.5432e-03, -2.7294e-02,
        -2.3011e-03,  2.8736e-03,  5.8974e-03,  9.6005e-03,  1.1634e-03,
         2.6890e-02, -4.3486e-04, -5.5651e-03, -6.4551e-03, -3.9007e-03,
        -4.2604e-03,  1.8480e-03,  1.2653e-03,  2.8766e-03,  5.7687e-04,
         4.4849e-03,  1.1946e-03,  1.6198e-03,  1.8098e-03,  1.3345e-03,
         4.0245e-03, -1.6935e-03,  3.5962e-03,  5.8498e-03,  5.7550e-06,
         3.8121e-03,  2.7976e-03,  3.0672e-03,  1.5137e-03, -5.2576e-03,
         7.2835e-03, -3.0608e-03, -2.9031e-03,  2.3508e-03,  1.0116e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-4.2111e-02, -6.3238e-01, -3.2056e-02,  8.2973e-03, -1.8110e-02,
        -6.0906e-03,  1.2894e-02,  7.6550e-03,  7.6670e-03,  4.2832e-03,
         2.2870e-02,  1.6096e-03, -2.3000e-02, -7.8732e-03, -3.0001e-03,
        -1.8145e-03,  1.1562e-02, -2.9917e-02,  2.3841e-03,  2.8514e-03,
        -9.0366e-05,  6.5062e-02,  5.6418e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1143, -0.5515, -0.0740, -0.0228,  0.0022, -0.0156, -0.0070, -0.0131,
        -0.0189, -0.0062,  0.0354, -0.0022,  0.0124,  0.0537,  0.0276, -0.0016,
        -0.0262,  0.0153,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3008e-01,  3.2621e-01,  6.9773e-02, -3.6056e-02, -1.0424e-04,
         3.0942e-02,  1.1057e-02, -2.1147e-02, -2.9638e-02, -3.4948e-02,
         4.3280e-05, -2.0604e-02,  2.7564e-02, -2.5607e-02, -1.4454e-02,
         5.6892e-02,  3.4072e-02, -3.0806e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5717e-01, -4.5160e-01,  4.8467e-02,  6.3653e-02,  1.6856e-02,
         2.7869e-02, -8.9661e-03,  1.0351e-01,  1.3532e-02,  1.6122e-06,
        -3.4035e-03,  8.6882e-03,  2.5847e-03,  7.5684e-02,  1.8018e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0774, -0.5716,  0.0107, -0.0127,  0.0034, -0.0159,  0.0153,  0.0029,
        -0.0350, -0.0096,  0.0132, -0.0035,  0.0074, -0.0065, -0.0054, -0.0022,
        -0.0070, -0.0088,  0.0079,  0.0205, -0.0024,  0.0185,  0.0020,  0.0033,
         0.0231,  0.0073,  0.0211,  0.0019,  0.0548, -0.0165, -0.0103,  0.0019,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0621e-01, -4.9287e-01,  2.2236e-02, -3.2205e-03, -1.3878e-02,
         1.1027e-02,  6.8443e-03,  1.5043e-02,  4.7998e-03, -1.9131e-03,
        -1.6312e-03, -1.4319e-02, -7.1952e-03, -5.1546e-03, -5.0025e-03,
         6.1737e-03,  6.6769e-05,  3.0366e-02, -1.9898e-02,  5.1269e-02,
        -2.8851e-02,  2.4186e-02,  2.3331e-02,  5.4640e-03, -9.4860e-04,
        -8.4055e-03,  1.2633e-02, -1.9636e-02,  2.4846e-03,  8.5275e-03,
        -1.6106e-02, -6.7793e-03, -2.3529e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4671e-02, -5.3348e-01,  2.1931e-02, -1.7704e-02, -1.9611e-02,
        -4.2521e-02, -1.5558e-02, -1.8519e-02, -6.8944e-03,  4.5259e-03,
         1.6618e-03,  3.6531e-02, -7.4148e-03, -4.4421e-03,  1.2599e-02,
         1.7326e-02,  6.1454e-03,  4.8670e-03,  1.2087e-02,  7.7596e-03,
         8.2644e-03, -1.7436e-03,  5.5066e-03,  1.2610e-02,  2.2261e-03,
        -1.5639e-03, -4.0074e-03, -1.3604e-02, -6.6286e-03, -2.3316e-03,
        -5.1124e-03, -1.0781e-03,  2.4876e-04, -4.9801e-03, -4.3944e-03,
        -3.2986e-04,  3.0001e-03,  4.7152e-04,  8.7181e-04, -1.0822e-02,
        -4.7200e-04,  5.0621e-04,  2.9242e-03,  5.1543e-03,  1.5497e-03,
        -5.7784e-03,  8.3207e-04,  1.9994e-03, -4.4496e-03, -1.4811e-03,
         2.7462e-02, -1.4822e-03, -3.6017e-03, -8.0466e-03, -9.0276e-03,
        -7.7308e-03, -4.4025e-03, -7.0554e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3669e-03, -6.1590e-01, -1.0151e-01, -4.7492e-02, -8.2230e-03,
        -5.5639e-03, -3.4546e-03,  1.8710e-03,  9.3826e-03, -1.2738e-03,
        -1.3639e-02,  7.3603e-03,  3.8990e-02,  2.5194e-02,  1.4011e-02,
         6.5728e-04, -8.8415e-03,  3.4931e-03,  4.6786e-05, -1.1334e-03,
         3.2004e-03, -1.0481e-02,  8.9497e-03, -4.9310e-03,  3.3478e-02,
         2.4560e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7515e-01, -6.1539e-01, -2.6185e-03,  1.5048e-02,  2.0584e-02,
         5.8970e-05,  1.8452e-02, -2.3329e-03,  4.9961e-03,  1.3519e-02,
        -9.9644e-03,  6.3566e-04,  1.3807e-02,  4.2678e-03,  4.0793e-03,
         5.0479e-04, -2.3199e-03, -2.5866e-03, -8.6373e-03,  9.2466e-03,
         3.3721e-03,  1.5394e-02,  2.2266e-02,  2.5116e-02,  9.6468e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1217e-02, -5.6180e-01,  9.3994e-03,  1.9938e-02, -3.8018e-02,
         3.4883e-02, -9.0624e-03, -8.5908e-03,  1.7826e-02, -1.8189e-02,
         1.6739e-02,  2.5003e-03,  6.5437e-04,  2.6120e-03, -7.7499e-03,
        -2.2636e-02,  1.4677e-03, -7.4842e-03,  1.8945e-03,  1.7731e-03,
         9.2755e-03,  1.2062e-02,  2.4737e-03,  1.4778e-02, -4.8018e-03,
         8.6244e-03,  1.8055e-02,  7.9890e-03,  3.0102e-04, -9.8761e-03,
         1.2122e-02, -7.3196e-04, -1.1744e-02,  8.6449e-04, -1.5653e-02,
        -2.4342e-03,  1.9252e-03, -3.2084e-02, -5.8174e-03,  1.9007e-03,
        -2.4909e-04,  2.4646e-03, -1.0687e-02, -1.2580e-02, -6.0693e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5749e-03, -7.0323e-01, -8.8192e-02, -2.8800e-02,  2.3143e-02,
         1.3560e-03, -3.7850e-03,  9.6348e-04,  1.1449e-02, -2.9715e-03,
         1.1271e-02,  2.5183e-03,  9.4611e-03,  3.1293e-03, -1.1919e-03,
        -1.2618e-02, -2.1260e-02,  3.9288e-03,  5.4753e-04, -1.3204e-02,
        -6.7240e-03,  3.3465e-02,  1.3211e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1853,  0.6377,  0.0706,  0.0063,  0.0171, -0.0135,  0.0034,  0.0072,
         0.0071,  0.0115, -0.0070, -0.0332,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 0.0335, -0.6501, -0.0985,  0.0293, -0.0067, -0.0045, -0.0142, -0.0113,
        -0.0168,  0.0167, -0.0035, -0.0074, -0.0034, -0.0115,  0.0056,  0.0066,
         0.0073, -0.0009,  0.0052,  0.0308,  0.0106,  0.0213,  0.0044,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.1137e-02, -6.3383e-01, -2.8462e-02,  2.2076e-02,  7.9119e-03,
         6.1164e-03,  9.8914e-03,  9.8739e-04, -2.9052e-04, -1.1097e-02,
        -7.4321e-03, -7.9489e-03, -6.3813e-03, -1.5818e-02,  3.2789e-02,
         6.9462e-03,  8.2001e-03, -3.5394e-03,  6.7130e-03,  2.7798e-02,
         1.4256e-02, -5.8159e-03,  2.6055e-03,  4.2395e-03,  8.0294e-03,
        -4.4681e-03, -1.3053e-02, -1.6437e-03, -7.9433e-03, -1.2385e-03,
         7.3591e-04, -1.7615e-02,  2.9887e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0138, -0.7283,  0.0242,  0.0450,  0.0297, -0.0014,  0.0056,  0.0084,
         0.0056,  0.0038, -0.0079, -0.0253,  0.0047, -0.0179, -0.0263,  0.0179,
         0.0136, -0.0045,  0.0159,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0078, -0.7636, -0.0137, -0.0636, -0.0183, -0.0152,  0.0149,  0.0077,
        -0.0218,  0.0162, -0.0121, -0.0119, -0.0028, -0.0303,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1070e-01,  7.1352e-01,  1.9505e-02, -5.3921e-03, -5.0647e-03,
         1.1548e-02, -1.0285e-02,  1.2808e-02,  5.4851e-04,  6.3462e-04,
        -4.5048e-03,  7.6789e-04, -7.0578e-03, -1.7371e-02, -5.7242e-03,
        -6.8964e-03,  8.2770e-03,  4.9182e-04, -1.6406e-02, -6.3659e-03,
         5.2989e-03,  3.0829e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5916e-02,  6.4256e-01,  3.3158e-02,  4.7709e-02,  2.5178e-02,
        -5.7268e-03, -3.8677e-03, -3.3253e-02, -1.9644e-02, -6.3869e-03,
        -3.1823e-02, -4.6113e-03, -2.0693e-02,  1.2801e-02,  1.9035e-03,
         5.8349e-04, -5.3581e-03, -2.5552e-03, -3.9664e-03,  4.9008e-03,
        -9.9000e-03,  5.3309e-03, -9.9643e-04, -1.2462e-02, -9.5810e-03,
        -1.4922e-03,  6.6976e-04,  4.3531e-03, -1.2090e-02, -5.2756e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0942,  0.4193,  0.0248, -0.0590,  0.0089,  0.0797,  0.0232, -0.0755,
         0.0095, -0.0174, -0.0496,  0.0788,  0.0342,  0.0010, -0.0057, -0.0178,
         0.0012,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1061,  0.6516,  0.0505, -0.0472, -0.0095, -0.0078, -0.0062,  0.0164,
        -0.0716, -0.0171, -0.0159,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4569e-02,  5.5624e-01,  4.3999e-02,  8.0582e-03, -2.1681e-02,
        -1.1455e-01, -2.0139e-02, -2.0217e-02, -5.4835e-03,  1.1957e-02,
         2.1162e-02,  6.8452e-03, -1.8622e-03,  5.6687e-02,  2.1200e-04,
        -4.8942e-02,  1.7402e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0378,  0.5963,  0.0506,  0.0032,  0.0105,  0.0055,  0.0238,  0.0021,
        -0.0235, -0.0113, -0.0191, -0.0296, -0.0141,  0.0025, -0.0645, -0.0066,
        -0.0012, -0.0030, -0.0190, -0.0041, -0.0022, -0.0205, -0.0107, -0.0061,
        -0.0033, -0.0288,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0779e-02,  5.9453e-01, -4.4569e-02,  1.0678e-02,  7.8037e-03,
        -2.7661e-02, -1.7127e-02,  1.5179e-02,  3.7924e-02, -2.5326e-03,
         1.4401e-02,  3.9922e-03,  1.5181e-02,  2.4796e-02,  8.1904e-03,
        -9.8000e-03, -1.4334e-02,  2.7609e-02, -9.8700e-03, -6.6774e-03,
        -1.3065e-02, -4.5480e-03,  1.1280e-04,  1.3135e-02, -2.3908e-03,
        -7.7694e-03,  9.3709e-03,  6.9114e-03, -2.4297e-02,  1.4766e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8234e-03,  7.3040e-01,  1.2456e-02, -2.4698e-02,  1.6434e-02,
        -9.4378e-03,  2.1421e-02, -2.8994e-04,  1.2583e-02, -3.2866e-02,
        -4.6090e-03,  3.6516e-02,  1.2192e-02,  3.8223e-03, -7.9818e-03,
         6.9325e-03,  2.0894e-03, -4.1546e-03,  5.8290e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-3.6006e-02,  6.6934e-01,  6.0652e-02, -2.3872e-02, -7.9015e-03,
        -1.2124e-03, -1.3152e-02, -1.2389e-02, -1.1437e-03, -3.8694e-03,
         4.2736e-03, -1.9561e-04,  4.7666e-03,  4.7916e-03,  1.2179e-02,
        -1.3705e-02,  1.4509e-02,  1.5822e-02,  1.0377e-02,  9.0618e-03,
        -7.7479e-03, -1.3939e-02,  7.0206e-03,  9.9735e-03, -1.0035e-02,
        -2.1848e-02,  1.0220e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1168, -0.5518, -0.0586,  0.0202, -0.0285, -0.0147, -0.0225, -0.0076,
        -0.0342, -0.0314, -0.0089,  0.0613,  0.0069,  0.0366,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.8320e-02,  7.0124e-01,  3.5184e-02, -4.5802e-03,  4.6946e-04,
         4.8046e-03, -1.9880e-02,  1.3224e-02,  8.5968e-03, -2.4739e-02,
        -2.3436e-02,  1.1658e-02, -8.7468e-03,  3.5605e-03,  4.1146e-03,
        -7.0452e-04, -7.6194e-03,  8.0820e-03,  8.2041e-03,  3.3969e-03,
        -5.5284e-03,  4.9909e-03,  1.3291e-03,  2.2509e-03, -9.2341e-04,
        -2.4414e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.7212e-03,  6.6288e-01,  7.5974e-02,  2.6393e-02,  1.3047e-02,
        -4.6215e-03,  1.8495e-04,  2.0125e-02, -3.4869e-03,  2.5345e-02,
        -6.0063e-03,  6.0408e-02,  5.0718e-02, -4.3088e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6730e-02,  5.9409e-01, -1.9327e-02,  1.3890e-02, -7.6348e-04,
        -1.2600e-02,  2.4937e-02, -2.6470e-03, -3.1568e-03, -9.0696e-03,
        -1.0403e-02, -1.1624e-02, -1.1478e-02,  2.7558e-02, -2.0230e-02,
        -6.8769e-04, -3.1438e-02,  3.0429e-04, -6.4051e-03, -1.0170e-02,
        -9.0881e-03, -2.5110e-02,  1.8071e-02,  1.6975e-02,  1.9056e-03,
        -1.0778e-02, -9.8964e-03, -1.9906e-02,  3.3494e-04, -7.1710e-03,
        -2.8948e-02, -4.3060e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4236e-01, -5.9286e-01, -2.3751e-02, -3.1289e-02, -2.4456e-02,
         8.9583e-03,  2.0528e-02,  7.7000e-03,  8.0471e-03,  1.1564e-02,
         9.7661e-03,  1.6718e-02, -2.1562e-03,  6.7783e-03,  9.1361e-03,
        -3.9799e-04,  8.0895e-03,  8.7047e-03,  5.1481e-03, -6.2228e-03,
         1.7941e-03, -2.8118e-03, -2.5480e-03, -2.1481e-02,  2.6732e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0254,  0.7129,  0.1738, -0.0415, -0.0145,  0.0319,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1670,  0.5232,  0.0445, -0.0040,  0.0306,  0.0011,  0.0059, -0.0017,
         0.0015,  0.0068, -0.0027,  0.0067,  0.0043, -0.0114, -0.0129, -0.0080,
        -0.0046, -0.0007, -0.0016,  0.0055, -0.0055, -0.0062,  0.0229, -0.0014,
        -0.0104,  0.0138, -0.0008,  0.0010, -0.0121, -0.0034,  0.0206,  0.0125,
         0.0099,  0.0031, -0.0022,  0.0266,  0.0031], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.8061e-02, -5.4477e-01, -5.5259e-02, -4.3304e-02, -1.0510e-02,
        -1.6437e-02, -2.0776e-03,  1.0830e-02,  3.9243e-03,  5.5084e-03,
         4.3065e-03, -3.9703e-03, -8.6237e-03,  1.0807e-02, -2.0611e-02,
         3.9978e-03,  2.1244e-02, -2.6512e-02,  1.1339e-02, -8.8759e-03,
         2.6505e-02,  5.4691e-03,  1.1853e-02,  1.0556e-02,  6.3303e-03,
         4.0164e-03, -4.2743e-04, -1.5161e-02,  4.0381e-03,  3.4804e-03,
         5.4690e-03, -1.4546e-03,  3.3376e-04,  8.1189e-03,  1.5821e-02,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0442, -0.6840,  0.0452,  0.0161,  0.0119, -0.0159,  0.0015,  0.0059,
         0.0332,  0.0053,  0.0011, -0.0180,  0.0077, -0.0603,  0.0496,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0214, -0.6800, -0.0058,  0.0160,  0.0042,  0.0037, -0.0043,  0.0581,
        -0.0187,  0.0172, -0.0058,  0.0009,  0.0030,  0.0029, -0.0018, -0.0136,
        -0.0007, -0.0326, -0.0094, -0.0045,  0.0011,  0.0038,  0.0086,  0.0127,
         0.0178,  0.0025,  0.0019,  0.0032,  0.0054,  0.0088,  0.0036,  0.0023,
        -0.0236,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0381, -0.6976,  0.0046,  0.0096, -0.0128,  0.0331, -0.0362, -0.0056,
         0.0239,  0.0059, -0.0030, -0.0010,  0.0092, -0.0151, -0.0136,  0.0013,
         0.0078, -0.0047,  0.0107,  0.0008, -0.0013, -0.0010, -0.0033,  0.0014,
        -0.0036,  0.0092, -0.0165, -0.0271,  0.0018,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 1.5884e-01,  6.5575e-01,  1.3504e-02, -9.3814e-03,  5.8264e-03,
        -3.5395e-02, -3.2627e-03, -1.7536e-02,  3.5192e-03, -2.7804e-02,
         5.7578e-03, -1.0218e-02,  4.5249e-04,  1.1624e-02, -4.1126e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.2178e-02,  6.2555e-01,  2.8119e-02, -2.7421e-02,  8.2587e-03,
        -3.8274e-03,  1.4423e-02,  1.2645e-02,  2.0123e-02,  7.0175e-04,
        -2.3985e-03, -7.4483e-03, -6.0972e-03, -5.9574e-03, -1.3280e-03,
        -9.8147e-03, -6.6919e-04,  2.2653e-03, -4.4406e-03,  4.0233e-03,
         2.9389e-03, -2.5587e-04,  3.4628e-03,  1.6737e-03,  2.8635e-03,
         8.3062e-03, -1.7517e-03,  1.1603e-02,  8.8465e-03,  5.5459e-03,
         4.9845e-03,  2.5749e-03,  1.2242e-03, -9.6622e-04, -5.1670e-04,
         4.4996e-03,  9.2699e-04, -5.8303e-04,  5.7992e-03,  4.2391e-03,
         3.7595e-03,  3.3067e-03,  4.8449e-03,  1.3518e-03,  3.3150e-04,
         4.9640e-03,  1.5907e-02,  2.7469e-02,  2.6812e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0246,  0.6329,  0.0372, -0.0391,  0.0086, -0.0158, -0.0085, -0.0055,
        -0.0140,  0.0114,  0.0048,  0.0030,  0.0023,  0.0057,  0.0023, -0.0099,
        -0.0254, -0.0030,  0.0122, -0.0101,  0.0031,  0.0018, -0.0048,  0.0042,
         0.0024, -0.0124, -0.0113,  0.0130,  0.0045,  0.0017,  0.0029,  0.0141,
        -0.0056,  0.0394, -0.0022,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.8661e-02, -6.5727e-01, -3.5674e-02, -1.1722e-02, -2.4380e-03,
         1.9869e-02, -1.1561e-02, -1.4741e-02, -6.6901e-06,  2.0531e-02,
         1.6319e-02,  2.3008e-02,  1.8079e-02,  3.7697e-03, -5.1430e-03,
         4.0478e-03,  7.7081e-03,  6.0391e-02,  2.9060e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0208, -0.7429,  0.0177, -0.0242, -0.0140, -0.0312, -0.0142, -0.0184,
        -0.0319, -0.0059,  0.0159, -0.0025, -0.0091,  0.0460, -0.0052,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0453,  0.6074,  0.0637, -0.0066, -0.0045,  0.0019, -0.0411, -0.0087,
        -0.0069, -0.0280, -0.0278, -0.0283, -0.0258, -0.0159,  0.0191, -0.0205,
        -0.0133, -0.0120,  0.0053,  0.0014, -0.0048, -0.0117,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.4336e-02,  6.7969e-01,  1.8921e-02, -5.0165e-03,  4.5055e-03,
         4.7578e-03, -8.3093e-03, -7.0654e-04,  8.1286e-05, -1.0715e-02,
         8.9148e-03, -1.3411e-02, -1.5041e-02, -1.3012e-02,  5.5043e-03,
        -1.1822e-02, -9.3982e-03, -5.2950e-03,  7.8682e-03, -3.1036e-03,
         1.2909e-02, -7.9282e-03,  2.2292e-03,  2.0675e-02, -2.5253e-03,
         4.9715e-03, -6.5985e-03, -1.0177e-02,  1.2284e-02,  2.9292e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0367, -0.7536,  0.0248, -0.0329,  0.0605,  0.0013,  0.0280, -0.0071,
        -0.0231,  0.0224, -0.0096,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0129,  0.7568, -0.0014, -0.0163,  0.0357, -0.0127,  0.0215, -0.0016,
        -0.0021,  0.0131, -0.0026,  0.0163,  0.0008, -0.0096,  0.0053,  0.0170,
         0.0068,  0.0111,  0.0090, -0.0135,  0.0220, -0.0117,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0940e-04, -6.5317e-01, -7.1080e-02, -4.3481e-03, -1.5360e-02,
         4.3394e-03, -7.0696e-03,  2.6134e-03, -2.8953e-03,  6.3962e-03,
         3.2013e-02,  1.7467e-02,  2.3976e-02, -5.7919e-03,  8.6370e-03,
         1.3480e-02, -4.6546e-03,  1.6550e-02,  6.6704e-04,  5.5866e-03,
         7.0597e-03, -8.3942e-04, -2.1274e-02,  9.1491e-03, -2.2697e-03,
         6.9702e-03,  8.7771e-03,  2.1086e-03, -2.2629e-03,  5.6950e-03,
         1.3782e-02,  1.6788e-02,  6.3163e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.0448e-02, -5.8086e-01, -9.9624e-02,  2.7910e-02, -1.0427e-02,
         1.2695e-02, -8.4242e-03,  1.9961e-02, -7.4818e-03, -4.8045e-03,
         4.6239e-03, -1.5942e-02, -8.1909e-03, -8.5398e-03,  1.7812e-02,
        -2.6809e-02,  6.9710e-03,  3.7268e-04,  6.7636e-03,  6.1428e-04,
        -2.1903e-04,  7.1265e-03, -1.2454e-02,  1.7076e-02,  4.6305e-03,
         7.8926e-04, -2.1759e-02,  1.6676e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5337e-02, -6.3042e-01, -1.0700e-01, -2.2483e-02, -1.4801e-02,
         7.0119e-03, -8.6465e-03,  8.3516e-03,  3.3335e-02,  1.8738e-03,
         1.1660e-02, -1.7060e-02, -5.0314e-03, -1.8757e-03, -8.5706e-04,
        -1.7609e-02,  1.0159e-02,  3.8802e-04,  2.9621e-03, -2.2243e-03,
        -3.5663e-03,  1.9204e-03, -1.0582e-02,  8.6608e-03,  1.6179e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 2.8934e-02,  6.1674e-01,  3.3795e-02, -2.4545e-03, -6.8617e-03,
         1.0279e-02, -6.6807e-03, -3.3787e-03, -8.7886e-03, -4.4773e-03,
         6.3050e-03, -1.2319e-02,  2.2514e-03,  1.2299e-04,  4.4021e-03,
        -2.3370e-02,  2.2730e-03, -6.1611e-03, -1.8815e-03, -1.3428e-02,
         1.5182e-02, -1.0777e-03, -9.7989e-03, -2.2336e-03,  2.6330e-02,
         1.0838e-02, -6.3384e-03,  4.8487e-04, -4.0811e-03, -9.6721e-03,
         1.5887e-04, -5.1376e-03,  4.1185e-03,  5.9566e-03,  3.1577e-03,
        -8.6061e-03, -1.0457e-02, -8.4911e-03,  1.0958e-02,  2.6492e-02,
        -3.5524e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0662, -0.6276, -0.0697, -0.0287, -0.0013,  0.0102, -0.0218,  0.0239,
         0.0183,  0.0066,  0.0088, -0.0119, -0.0078,  0.0016, -0.0118,  0.0105,
         0.0044, -0.0026,  0.0038, -0.0011, -0.0081,  0.0062,  0.0049,  0.0294,
         0.0129,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.3917e-03,  7.3157e-01,  1.9495e-03, -9.0272e-04, -2.6085e-04,
        -3.8356e-03,  2.2004e-02,  3.6846e-02, -1.2461e-02, -2.4264e-02,
         3.3936e-03, -1.2566e-02,  1.0073e-03, -2.2400e-03, -6.5694e-03,
        -3.8725e-03,  1.1762e-02, -1.8331e-02, -4.9599e-02, -1.3134e-02,
         3.7040e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0775, -0.6714, -0.0309,  0.0133,  0.0313, -0.0032, -0.0121,  0.0010,
         0.0219,  0.0150,  0.0115,  0.0102,  0.0032,  0.0052,  0.0204, -0.0186,
         0.0477, -0.0058,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7125e-02, -5.7017e-01,  9.1379e-03, -9.9321e-03, -1.4862e-02,
         2.6943e-02, -1.2697e-02, -1.0004e-02,  5.9935e-03,  2.5561e-02,
        -6.6044e-03, -4.2177e-03, -5.0689e-03,  9.7621e-03, -5.1176e-03,
         8.6199e-03, -6.6429e-03,  3.2945e-02, -2.5631e-02,  4.2888e-03,
         1.3339e-02,  2.1928e-03,  1.0670e-02,  4.9423e-03, -9.1927e-03,
        -2.8752e-02, -6.1922e-03, -8.3580e-03, -6.4931e-04,  7.1296e-03,
        -2.2050e-04, -7.7209e-03,  2.8129e-03, -5.5371e-03,  2.5667e-02,
         4.4522e-03, -4.2598e-03, -1.6590e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4008e-02, -6.6486e-01, -6.1742e-02, -9.1390e-03,  1.4183e-02,
        -2.9875e-03,  7.2214e-03,  1.2377e-04,  1.4302e-02, -1.9802e-03,
        -2.1648e-04,  1.9186e-02, -3.9323e-03,  6.8381e-02,  8.1761e-03,
        -1.5978e-02,  2.4730e-02,  7.0351e-03,  2.9226e-02,  3.2595e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0137,  0.6480,  0.0208,  0.0700,  0.0302, -0.0107,  0.0372,  0.0283,
        -0.0178,  0.0163, -0.0257,  0.0015, -0.0248, -0.0551,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2322e-02,  7.2110e-01,  4.4260e-02,  1.1922e-02,  3.1378e-03,
         1.4696e-02,  1.8426e-02,  1.2847e-03, -3.2315e-03, -3.4236e-03,
        -1.2562e-02, -4.0788e-03,  1.1322e-03,  4.1014e-03,  8.2208e-03,
        -1.0726e-02,  1.4129e-02,  3.8438e-03,  1.1924e-04, -3.3875e-03,
        -6.8706e-03,  1.2651e-02,  1.5074e-03,  1.1431e-02, -5.9063e-03,
         5.4385e-04,  1.4372e-03, -2.6256e-03, -1.7020e-03,  8.1841e-03,
        -8.9504e-04,  4.0141e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3469e-02, -6.5962e-01, -6.0589e-04, -4.0098e-02, -3.3266e-02,
        -1.0649e-02, -1.2692e-02,  2.5023e-02,  1.9748e-02,  3.6593e-02,
         1.0459e-02,  6.8289e-03, -4.9191e-03, -8.7634e-03,  1.3026e-02,
        -2.9470e-05,  8.0612e-03,  1.8047e-02,  1.1116e-02,  4.6982e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9665e-02, -4.5838e-01, -1.8952e-02,  1.0205e-01,  2.6931e-02,
         2.3705e-03,  8.5059e-03, -9.9109e-03,  6.7589e-03,  1.6276e-02,
         7.2873e-03, -6.7742e-03, -3.0716e-03, -3.2885e-03, -4.9951e-03,
         4.3127e-03, -2.1526e-03, -4.2712e-03, -5.5389e-03,  8.1769e-04,
        -6.8007e-04,  4.9392e-03,  1.1430e-03, -1.0288e-02,  2.1335e-02,
         5.3476e-03,  1.9110e-02,  1.0752e-02,  2.3621e-05, -1.1034e-02,
        -1.2384e-02,  1.4464e-02, -4.2647e-03,  4.0498e-03,  2.3832e-03,
         5.6226e-03, -1.1380e-02,  2.6903e-03, -5.5897e-04,  1.0850e-03,
        -5.7733e-03, -1.4001e-02, -2.3433e-02, -9.7361e-03,  7.8143e-03,
         4.5677e-02, -1.7719e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0120, -0.7438,  0.0216,  0.0050, -0.0486,  0.0016,  0.0020,  0.0086,
         0.0071,  0.0164, -0.0116, -0.0045,  0.0331,  0.0583,  0.0257,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.7307e-03, -6.5082e-01, -6.4249e-03, -5.7231e-03, -6.0679e-03,
        -4.7831e-03,  1.8194e-02,  1.5116e-02,  1.1352e-02,  9.4668e-03,
        -8.6357e-03,  4.4996e-03,  1.0622e-02, -1.1302e-02,  1.0623e-02,
         1.6540e-02,  7.4475e-03, -1.6343e-02,  1.8517e-02, -4.1130e-03,
        -9.7622e-04,  7.6226e-03,  1.3070e-03,  1.6227e-02,  2.8932e-03,
         9.7820e-03, -7.1012e-03,  3.6333e-02,  3.1831e-04, -2.8996e-03,
         7.4360e-04,  2.9810e-03,  1.9443e-03,  3.0402e-03,  5.3210e-03,
        -7.7745e-03,  7.6808e-05,  1.6432e-03, -4.6358e-03,  2.7987e-04,
        -1.1511e-03, -2.1499e-03, -1.7135e-03,  2.0289e-03, -8.3465e-03,
         3.3676e-03, -1.5975e-02, -6.0473e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-9.2167e-02, -5.9860e-01, -3.1618e-02, -2.7369e-02, -8.0726e-03,
         3.0280e-04, -1.6580e-02,  2.3200e-02, -4.0433e-02, -7.5491e-03,
         2.8037e-03, -6.3154e-04,  9.3614e-03, -7.1640e-03,  3.6699e-03,
         9.0378e-03, -2.6609e-04,  1.7841e-02,  1.6918e-02, -7.8928e-03,
         1.2671e-02,  1.4652e-02, -5.4095e-03, -4.6948e-03, -3.8524e-03,
        -3.7246e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0509,  0.6242,  0.0173,  0.0135,  0.0223,  0.0211, -0.0041,  0.0041,
         0.0311, -0.0115,  0.0018, -0.0227, -0.0042,  0.0025, -0.0051,  0.0170,
        -0.0258, -0.0240, -0.0358, -0.0043,  0.0566,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.2043e-02,  6.5511e-01,  2.7949e-03, -3.2182e-03, -1.6051e-02,
        -1.2608e-02,  6.6921e-03, -2.0744e-02, -3.1468e-03, -6.1728e-03,
         7.9880e-04,  9.1352e-04, -1.1146e-02, -1.2692e-03, -1.6113e-02,
        -8.7116e-03,  1.6499e-03, -1.7828e-02, -3.1914e-03, -2.0492e-02,
        -1.8975e-02, -6.0860e-03,  8.0929e-03,  6.0764e-04, -1.0254e-02,
        -4.9859e-03, -9.7164e-03, -6.8502e-03, -1.5325e-03, -4.4515e-03,
        -6.4897e-03, -3.1142e-03, -2.7372e-03, -3.9869e-03, -1.2588e-02,
        -3.8841e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0234,  0.6939,  0.0560,  0.0054, -0.0160, -0.0063, -0.0135, -0.0036,
         0.0210,  0.0080, -0.0073,  0.0067, -0.0130, -0.0016, -0.0063,  0.0073,
        -0.0042, -0.0231, -0.0239, -0.0593,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4524e-02, -7.1041e-01, -3.7945e-02,  1.5871e-02, -1.7975e-02,
        -3.7055e-03,  9.5923e-03,  3.9659e-02,  8.1742e-03, -1.3403e-02,
        -3.0655e-03,  2.3753e-02,  4.3195e-04,  4.0967e-03, -3.2150e-03,
         8.0626e-04, -1.1008e-03, -2.3865e-03,  3.2663e-02, -1.4068e-02,
         1.3153e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0766,  0.7408,  0.0734,  0.0060, -0.0232,  0.0129, -0.0013, -0.0076,
         0.0115,  0.0027, -0.0113, -0.0165, -0.0161,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6552e-02, -5.9572e-01, -7.5392e-02, -9.2744e-03,  2.1661e-02,
         7.3351e-03,  3.2567e-02,  1.0383e-02, -2.6476e-03,  2.4417e-03,
        -4.8714e-03,  1.9648e-02,  1.8455e-02,  1.3925e-02,  1.8842e-02,
        -2.5561e-03,  1.3400e-02,  4.9828e-03,  1.5783e-03,  4.6725e-03,
        -1.0126e-02, -5.9806e-03,  9.9724e-03,  2.5610e-03, -2.2523e-03,
         1.8356e-02,  6.4184e-03, -2.9523e-03, -3.6805e-03, -2.7503e-03,
        -1.0516e-02, -2.7003e-03,  1.4390e-04,  1.4129e-03, -1.0605e-02,
        -1.0296e-03, -1.1642e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0708,  0.5571, -0.0336,  0.0020, -0.0673, -0.0074,  0.0172, -0.0010,
         0.0011, -0.0207, -0.0132, -0.0016,  0.0327,  0.0117, -0.0034,  0.0022,
         0.0300,  0.0193,  0.0074,  0.0070,  0.0029,  0.0229,  0.0287,  0.0390,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4157e-01,  5.8223e-01,  6.3687e-02,  7.6326e-03,  1.0972e-02,
        -3.2252e-03,  1.7327e-02,  1.1952e-02,  3.8670e-04, -2.5295e-03,
         1.1198e-02,  7.6612e-03,  1.2257e-02, -2.6810e-03,  5.5791e-03,
         7.3228e-03,  9.4831e-03, -8.5107e-03, -3.5305e-03,  1.0928e-02,
         1.7171e-04,  3.4960e-03,  3.3240e-03,  2.2705e-03, -6.3726e-03,
        -7.6467e-04, -6.4889e-03,  4.1199e-03,  1.3829e-02,  4.6240e-03,
        -1.2518e-02, -2.1350e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4678e-03, -7.4715e-01, -1.3330e-02, -3.0483e-02,  1.5305e-02,
         2.7807e-02,  2.1383e-04,  1.4378e-02, -2.7038e-03,  8.1294e-03,
        -2.1826e-03,  5.2837e-03,  2.6134e-03,  6.4965e-03,  2.1529e-03,
         6.7439e-03,  8.5916e-03, -6.2500e-03, -7.9925e-03,  8.0366e-03,
         1.0056e-02,  1.5523e-02, -6.2131e-03,  1.5699e-03,  1.9154e-02,
         1.3489e-02,  8.0790e-03,  1.1036e-03, -1.1320e-03, -1.9349e-03,
         1.4307e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.3496e-02,  5.7966e-01,  5.4171e-02,  2.3865e-02, -1.2946e-02,
         3.7481e-03,  4.1600e-02,  1.1592e-02,  6.9604e-03,  8.8019e-03,
         3.8387e-02,  7.9228e-03, -1.0639e-02,  3.5538e-03,  1.8513e-02,
         1.7024e-02,  1.6744e-02,  1.6824e-02,  2.8244e-03,  1.9685e-02,
        -1.6587e-02,  4.4639e-04,  4.0119e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.2019e-02,  6.2534e-01, -1.0312e-02,  2.3558e-02,  7.7499e-04,
         2.3624e-03,  1.8348e-03,  2.4245e-02,  1.3433e-02, -4.7649e-03,
        -9.0691e-03, -1.5627e-02,  4.6852e-04, -7.6313e-03,  1.6178e-02,
         5.7470e-03, -1.0428e-02, -1.1008e-02,  2.2400e-03,  1.4985e-02,
         1.6608e-04,  3.3002e-03, -3.2406e-03, -5.0502e-03,  1.0353e-02,
         3.7697e-03,  9.3391e-03,  9.2762e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-2.5279e-02,  5.7653e-01, -2.9613e-02, -4.8238e-03,  1.4937e-03,
         2.4880e-02,  2.6061e-02,  2.0479e-02, -2.3908e-02, -8.9018e-03,
         2.6631e-02, -3.3692e-03, -2.9786e-03, -8.7033e-03, -1.3700e-03,
         7.2899e-03,  2.8984e-04,  2.3609e-02, -1.1868e-02,  1.1260e-02,
         8.2581e-03,  7.2781e-03, -2.2645e-02, -9.9378e-04, -2.2527e-02,
         1.5386e-02,  2.7671e-02, -1.7423e-02, -3.8477e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.1398e-02,  5.7208e-01,  8.7098e-03, -1.1368e-02,  4.1206e-02,
         1.3500e-02, -2.2397e-02, -1.3751e-02, -9.5693e-03, -1.9301e-02,
        -2.0053e-02, -1.0753e-02, -1.9720e-02, -1.6205e-02, -3.6248e-03,
        -7.4341e-03,  1.1391e-02,  1.0587e-02, -2.7210e-04, -1.0659e-02,
        -9.6607e-03,  1.4144e-02, -1.3625e-02, -4.8590e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0203,  0.5997, -0.0149, -0.0702,  0.0085,  0.0103,  0.0006, -0.0025,
         0.0030, -0.0250, -0.0082,  0.0032, -0.0069, -0.0022, -0.0125, -0.0010,
        -0.0019, -0.0015, -0.0298,  0.0180,  0.0114, -0.0062, -0.0141,  0.0276,
         0.0089, -0.0137, -0.0058,  0.0060, -0.0071, -0.0034,  0.0054, -0.0154,
        -0.0049,  0.0300], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9235e-02,  7.2345e-01, -8.8816e-02, -2.0341e-02, -6.0122e-03,
        -2.2953e-02, -1.4950e-03, -4.9369e-03,  1.0799e-03, -6.8332e-03,
         4.7577e-03, -9.2349e-04,  2.8742e-03, -1.4277e-02, -9.1317e-03,
         9.1037e-03,  2.5630e-03,  1.4554e-03,  7.4336e-05,  9.9598e-03,
        -1.1593e-02,  8.1333e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.6808e-02, -6.2226e-01,  1.0656e-02, -3.4125e-02,  8.7099e-02,
         7.3698e-02, -2.0415e-03, -3.3433e-05,  2.7938e-03, -6.0576e-03,
        -2.8342e-02, -1.8837e-03, -5.1576e-03, -7.8256e-03, -1.1181e-03,
         2.7368e-03,  1.3598e-02, -3.7637e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0333,  0.6298,  0.0647,  0.0051, -0.0198,  0.0028,  0.0345,  0.0095,
         0.0052, -0.0077,  0.0096, -0.0153, -0.0124, -0.0105, -0.0013, -0.0077,
        -0.0054, -0.0148,  0.0128,  0.0051,  0.0072, -0.0408, -0.0444,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0362,  0.6763, -0.0286,  0.0383,  0.0423,  0.0162, -0.0064, -0.0044,
        -0.0045, -0.0071, -0.0164, -0.0043,  0.0024,  0.0039, -0.0418, -0.0256,
         0.0453,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1411,  0.5844,  0.0453,  0.0152,  0.0076,  0.0108,  0.0069,  0.0029,
        -0.0197,  0.0141,  0.0062,  0.0047,  0.0096, -0.0014,  0.0355, -0.0234,
         0.0141,  0.0572,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0872,  0.6230,  0.0650, -0.0105,  0.0069,  0.0324,  0.0018,  0.0042,
         0.0103,  0.0225,  0.0052, -0.0138,  0.0032,  0.0061, -0.0077, -0.0150,
        -0.0192, -0.0130, -0.0036, -0.0016, -0.0017, -0.0050, -0.0063, -0.0139,
         0.0031,  0.0178,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7949e-02, -6.8678e-01, -2.1483e-02,  3.3621e-02,  1.8416e-02,
         1.2154e-03, -1.9082e-02, -6.6928e-03,  3.9153e-03,  1.8555e-02,
        -1.6082e-02, -7.5109e-03,  6.1968e-03,  7.1640e-03,  7.4931e-03,
         2.0725e-03,  1.8546e-02,  3.8352e-03,  5.0862e-03, -3.0663e-03,
        -4.1972e-03, -6.8617e-03,  1.3029e-02,  9.6017e-04, -1.8360e-03,
         2.6753e-03, -4.1383e-03,  4.8162e-03, -2.3020e-04, -2.3748e-02,
         2.2750e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.5865e-02,  5.9983e-01,  4.1961e-02,  1.5411e-03, -1.3414e-02,
         3.9150e-03, -1.4910e-02,  2.2120e-02, -1.3241e-02,  5.9397e-02,
         1.1386e-02, -1.5049e-02,  2.7154e-02,  1.6576e-02,  1.5235e-04,
         2.5992e-02,  1.1990e-02, -5.5510e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0151,  0.8391, -0.0125,  0.0194,  0.0413, -0.0057, -0.0216,  0.0452,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 1.0757e-01,  6.6608e-01,  1.7460e-02, -1.4280e-02, -4.8664e-03,
         6.5508e-03, -1.1834e-03,  6.8930e-04, -1.4235e-02,  6.2626e-03,
         7.9516e-03, -1.8898e-02, -1.7975e-03, -3.1841e-03,  8.5653e-03,
         1.1340e-02,  2.7649e-03, -3.1892e-03, -3.1200e-03, -8.2137e-03,
        -2.5838e-03, -2.3396e-03, -1.6350e-02,  3.4692e-03, -1.8000e-03,
         6.4687e-05,  4.3099e-03,  1.3621e-02,  8.5924e-03,  7.2552e-03,
         6.0553e-03,  3.5595e-03,  2.2214e-03, -3.5818e-04, -9.8714e-03,
         5.5711e-03,  2.9202e-03,  6.8193e-04, -1.6953e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0167,  0.5282, -0.0596, -0.0123,  0.0172,  0.0120, -0.0214,  0.0313,
         0.0084,  0.0115,  0.0018,  0.0121,  0.0253, -0.0027,  0.0179,  0.0109,
        -0.0060, -0.0065,  0.0019, -0.0275, -0.0214, -0.0465,  0.0043, -0.0028,
        -0.0024, -0.0029, -0.0529, -0.0356,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0129, -0.7605,  0.0106,  0.0210,  0.0042,  0.0149, -0.0237, -0.0234,
        -0.0191, -0.0211,  0.0378, -0.0130, -0.0377,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0216,  0.7486, -0.0615,  0.0022,  0.0013,  0.0140, -0.0186,  0.0023,
        -0.0026, -0.0106, -0.0044, -0.0075, -0.0250, -0.0051, -0.0747,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.9610e-02,  4.8074e-01, -1.5691e-02, -4.5599e-02,  2.0275e-02,
        -1.0356e-02, -2.9564e-05, -1.9174e-02, -9.6045e-03, -1.7552e-02,
         7.9978e-03,  1.3725e-02, -1.7609e-02,  7.7746e-03, -1.5810e-02,
        -5.4325e-03,  1.3122e-02,  8.4221e-04, -8.6710e-03, -3.0522e-03,
         1.6357e-02, -6.6071e-03,  1.0245e-02,  2.3909e-04, -1.6765e-02,
         6.5815e-03, -1.8868e-03,  7.1143e-03,  1.3490e-03, -7.6225e-03,
         4.3797e-03, -4.1196e-03,  6.6920e-03,  3.9044e-03,  1.7005e-02,
        -7.0870e-03, -2.5737e-03, -5.2531e-03, -1.1951e-03, -3.0122e-03,
         2.3210e-02, -1.5269e-03,  2.8997e-03, -2.1115e-03, -2.1133e-04,
        -3.7781e-02, -2.9605e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3206e-02,  6.7351e-01,  3.8022e-02, -3.7240e-03,  1.1144e-02,
        -3.4880e-02, -1.6009e-02, -2.0200e-03, -2.3119e-02,  3.8118e-03,
        -1.2604e-02, -1.8154e-02,  3.3888e-03,  2.6813e-03,  1.4346e-05,
        -7.1643e-03,  8.5891e-03, -4.1251e-03,  3.2977e-03,  6.0829e-03,
         1.9315e-03,  1.2448e-04,  3.0351e-03,  2.9759e-03,  9.2394e-03,
         2.4408e-03,  7.8409e-03, -2.0938e-03,  3.0440e-03, -2.7543e-03,
         5.1651e-03, -7.5062e-03, -2.0537e-03,  7.7224e-03, -9.3314e-04,
         2.3846e-03, -1.0567e-03, -2.2164e-03, -1.8506e-03, -1.6157e-03,
         2.4692e-03,  2.4001e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0384,  0.6344,  0.0614,  0.0027, -0.0155, -0.0044,  0.0085,  0.0014,
        -0.0193, -0.0112,  0.0131, -0.0186,  0.0137, -0.0012,  0.0209,  0.0083,
        -0.0125,  0.0083, -0.0088,  0.0030, -0.0204, -0.0036, -0.0246,  0.0228,
        -0.0228,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1045,  0.5720,  0.0993,  0.0031, -0.0131, -0.0115, -0.0103, -0.0060,
        -0.0045, -0.0101,  0.0021,  0.0052,  0.0027,  0.0168, -0.0137, -0.0251,
         0.0046,  0.0061, -0.0039, -0.0211, -0.0206,  0.0021,  0.0415,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0307,  0.4639,  0.0460, -0.0134,  0.0131, -0.0080, -0.0113, -0.0306,
         0.0868,  0.0071,  0.0073, -0.0073,  0.0036,  0.0189,  0.0519, -0.0017,
        -0.0367,  0.0005,  0.0271, -0.0089, -0.0124,  0.0259,  0.0869,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0320,  0.6941, -0.0332, -0.0178, -0.0227, -0.0018,  0.0331, -0.0195,
         0.0453, -0.0083, -0.0260,  0.0179, -0.0483,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1169,  0.6684, -0.1020, -0.0039,  0.0097, -0.0151, -0.0091,  0.0213,
         0.0146,  0.0050, -0.0146,  0.0099, -0.0035,  0.0061,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0379,  0.7354,  0.0261, -0.0052,  0.0304,  0.0077, -0.0058,  0.0239,
         0.0033, -0.0253,  0.0059, -0.0204, -0.0031, -0.0040,  0.0240,  0.0009,
        -0.0145,  0.0224, -0.0035,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-2.0138e-02,  7.6649e-01,  1.5462e-03, -3.0551e-02, -7.5682e-03,
         5.6286e-03, -4.9124e-03, -1.2354e-03, -2.8556e-04,  1.5748e-02,
         5.5138e-03, -1.3001e-02, -7.0419e-03, -6.6388e-03, -2.2650e-03,
        -3.1000e-02,  8.8962e-03, -4.2607e-03, -2.3274e-03,  8.9099e-03,
        -6.1424e-03,  1.0197e-03,  2.3228e-03, -6.1075e-03,  1.1715e-02,
        -2.8732e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7102e-02, -7.7372e-01, -2.5401e-02, -4.5377e-02, -5.4823e-03,
        -1.2285e-02,  1.4264e-02,  1.5188e-03, -1.7162e-03,  3.2437e-02,
        -1.5437e-02, -6.7884e-04,  6.7577e-03, -1.9200e-02, -1.1677e-02,
         1.6949e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0206,  0.6820,  0.0178,  0.0625, -0.0158,  0.0289,  0.0231, -0.0205,
         0.0136, -0.0064, -0.0120, -0.0244, -0.0066, -0.0009,  0.0193, -0.0252,
         0.0206,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5261e-01,  6.2171e-01,  5.6798e-02,  5.1380e-04, -8.0067e-03,
         6.4770e-02,  7.7442e-03,  3.4710e-02, -5.3135e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0117,  0.5178,  0.0614,  0.0216, -0.0224,  0.0024,  0.0234, -0.0016,
         0.0387, -0.0143, -0.0028,  0.0025,  0.0018,  0.0095,  0.0071,  0.0031,
         0.0081,  0.0078,  0.0040,  0.0179,  0.0107,  0.0161, -0.0009,  0.0008,
         0.0025,  0.0125, -0.0157,  0.0020,  0.0123,  0.0042,  0.0067,  0.0049,
        -0.0034, -0.0024, -0.0032, -0.0119,  0.0058,  0.0059, -0.0093,  0.0010,
        -0.0082, -0.0055,  0.0042, -0.0096, -0.0054,  0.0549], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.1296e-02,  6.1929e-01,  6.8752e-03,  1.3757e-02, -1.5945e-02,
         3.3277e-05,  1.3648e-02, -2.7829e-02, -3.1867e-02,  1.5897e-03,
        -9.2265e-03, -5.8543e-03,  2.1442e-02, -4.5088e-03, -6.7012e-03,
        -1.7875e-02,  6.5933e-03, -7.2116e-03, -1.1417e-02,  1.0319e-02,
        -2.0155e-03, -5.5419e-03,  1.3285e-02, -7.1744e-03,  1.0975e-02,
         2.4180e-03,  2.4197e-02, -9.1660e-03, -4.9824e-03, -6.9674e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0729, -0.7679, -0.0404,  0.0296,  0.0111,  0.0168, -0.0613,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1219, -0.7253, -0.0538,  0.0125, -0.0061, -0.0128, -0.0216, -0.0162,
        -0.0121, -0.0054,  0.0124,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0292, -0.6899, -0.0871, -0.0167, -0.0067, -0.0228, -0.0074,  0.0098,
         0.0007,  0.0201,  0.0023, -0.0186,  0.0017,  0.0871,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1034e-02,  6.8493e-01,  7.2122e-02, -1.8572e-02,  1.1820e-02,
         4.2342e-03,  5.3913e-03, -3.5361e-03,  2.3668e-03,  6.8868e-04,
         1.3028e-03, -5.3442e-03, -2.9190e-03, -1.3895e-02,  3.5194e-03,
        -3.2253e-03,  2.4019e-03, -6.5579e-04,  1.4283e-03,  8.2467e-03,
        -7.9092e-03,  1.5530e-02, -1.4118e-02, -5.9517e-04,  1.7683e-03,
         5.5472e-03,  2.1300e-03,  2.2586e-03, -1.3686e-02,  4.5942e-03,
        -2.7478e-02,  5.9924e-03,  4.8938e-04,  2.0273e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0274,  0.7047, -0.0087,  0.0106, -0.0124, -0.0017,  0.0113, -0.0061,
         0.0070,  0.0155, -0.0015, -0.0008, -0.0024,  0.0322, -0.0188, -0.0103,
         0.0016,  0.0399,  0.0030,  0.0108,  0.0246,  0.0008,  0.0045, -0.0082,
        -0.0078, -0.0012, -0.0107,  0.0155,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3086e-02,  6.9101e-01, -4.3843e-02, -3.0999e-02, -2.5125e-02,
         9.4056e-03, -2.8890e-03, -4.1036e-03, -1.0469e-02,  8.8123e-03,
        -9.1880e-03, -1.4188e-03,  4.3921e-02, -8.3176e-03, -2.0902e-02,
         6.4241e-03,  1.3271e-04, -6.7142e-03, -9.1459e-03, -8.3027e-03,
         2.4271e-03, -3.2869e-03,  1.1214e-02,  8.8667e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 6.8479e-02, -5.9789e-01,  1.4391e-02,  3.7046e-03,  1.4446e-02,
        -3.0335e-03,  2.3185e-02,  3.2860e-02,  2.6647e-02,  3.9974e-03,
         3.9884e-02,  6.3670e-03,  1.8415e-03, -1.0098e-02,  6.9174e-03,
         1.1236e-02,  1.4558e-02,  1.0583e-02,  1.9070e-02,  8.2777e-03,
        -1.0573e-03,  4.2832e-03,  1.7851e-03,  1.3570e-02,  6.6505e-03,
        -5.6047e-03,  3.2727e-03, -5.5463e-04,  7.7394e-03, -1.1701e-02,
        -2.6316e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0280e-01, -5.9573e-01, -2.7226e-02,  1.6102e-02,  1.2688e-02,
        -3.4952e-02,  2.5845e-04, -3.6921e-02, -2.2973e-02, -2.9787e-02,
        -5.9623e-03,  1.1460e-01,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1706, -0.4420,  0.0178, -0.0503,  0.0092,  0.0037,  0.0301,  0.0016,
        -0.0108, -0.0167,  0.0063,  0.0539, -0.0634,  0.0406, -0.0829,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.5219e-03, -6.7660e-01, -2.7728e-02, -1.1654e-02, -2.6248e-02,
        -3.9163e-03,  3.0217e-02,  5.4797e-04,  1.7852e-03, -1.1686e-02,
        -2.4623e-02,  2.5436e-02, -1.2223e-02, -8.7071e-03, -5.4927e-03,
        -7.1815e-03, -6.9394e-03, -4.6008e-03,  1.3341e-02,  1.3188e-02,
        -3.9322e-03, -5.7524e-03, -1.0907e-02, -1.0223e-02, -4.0652e-05,
         1.0466e-02,  4.0924e-03,  2.3258e-03,  4.3440e-03,  1.0471e-02,
         3.6916e-03, -2.0967e-03,  1.1016e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0021, -0.6735,  0.0271,  0.0288, -0.0268, -0.0181,  0.0141,  0.0164,
         0.0066,  0.0111, -0.0042, -0.0099,  0.0299,  0.0553, -0.0036,  0.0725,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.2097,  0.5460,  0.0409, -0.0097,  0.0455,  0.0249,  0.1007, -0.0030,
        -0.0195,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2170,  0.7355,  0.0045, -0.0326, -0.0104,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1424, -0.5127,  0.0123,  0.0120, -0.0103,  0.0039, -0.0150, -0.0009,
        -0.0164, -0.0305, -0.0141, -0.0263,  0.0122, -0.0227, -0.0110, -0.0103,
        -0.0057, -0.0125, -0.0131, -0.0256, -0.0069, -0.0135, -0.0045, -0.0110,
        -0.0055, -0.0175, -0.0312,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1151, -0.6595, -0.1181,  0.0151,  0.0186,  0.0347, -0.0388,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0261, -0.7675, -0.0615, -0.0022, -0.0115,  0.0106, -0.0103, -0.0015,
         0.0094,  0.0376, -0.0013, -0.0052,  0.0074,  0.0120,  0.0131,  0.0077,
        -0.0008,  0.0040, -0.0019, -0.0018, -0.0039,  0.0028,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0200, -0.7154, -0.0134,  0.0088,  0.0708,  0.0024,  0.0018, -0.0359,
        -0.0014, -0.0129, -0.0123,  0.0164,  0.0419,  0.0090,  0.0011, -0.0051,
        -0.0017, -0.0184,  0.0113,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.0465e-02, -7.5565e-01, -9.0193e-03, -6.6544e-03, -3.6565e-02,
         1.2542e-03, -1.1315e-02, -9.4166e-03, -1.7884e-02, -1.1345e-02,
        -3.8163e-04, -1.7508e-03,  4.4457e-04,  3.4394e-03,  5.1300e-04,
         8.1678e-03,  1.4715e-02,  2.1733e-03,  3.7772e-03,  5.3818e-04,
        -7.4829e-03, -6.6423e-04, -3.9588e-03, -7.1347e-04,  2.3211e-03,
         8.0602e-04, -2.3513e-03, -8.2413e-03, -6.2417e-03,  3.7455e-04,
        -1.3789e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-6.0621e-02,  5.2795e-01,  1.2413e-01, -1.9477e-02,  7.8579e-03,
         3.1200e-03, -5.4622e-03,  3.7704e-03,  3.0400e-03, -1.2750e-04,
         2.8988e-03,  7.9748e-03,  6.0374e-03, -1.8719e-02,  5.9612e-04,
         2.5152e-02, -1.2867e-02,  6.7775e-03, -2.9852e-02,  3.6797e-03,
        -9.8221e-03,  5.9687e-03, -1.0170e-02, -1.7221e-02, -7.3374e-03,
         1.4437e-03, -3.3581e-03, -3.0720e-02, -1.0743e-02,  6.8040e-04,
         2.5591e-02, -6.8337e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2879e-02, -6.9165e-01, -1.3427e-02, -7.3544e-03,  2.1988e-02,
        -2.3653e-05, -1.5947e-02, -2.6111e-03, -2.3014e-02, -1.6286e-02,
        -2.5769e-02, -2.0502e-02, -7.2114e-04,  1.5900e-02, -1.1537e-02,
         3.0214e-03,  9.3308e-03, -1.2336e-02,  9.9800e-03, -5.9577e-03,
        -2.3758e-02,  4.6002e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6959e-02,  7.0079e-01, -2.4035e-02,  1.4985e-02, -3.0925e-02,
        -5.0832e-04, -4.5220e-02,  1.3441e-02,  2.1471e-03,  1.2136e-02,
         4.0140e-03, -1.3423e-02, -8.4417e-03,  1.6400e-03,  1.9807e-03,
         3.4791e-03,  8.0543e-04,  9.0841e-03,  3.1335e-03,  7.3568e-03,
         1.1758e-02,  3.8905e-06,  2.3253e-03, -6.7963e-03, -3.3943e-04,
         1.0440e-02,  3.5099e-03, -1.2799e-02,  2.6855e-02,  6.6653e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0342,  0.7832, -0.0019, -0.0363,  0.0016, -0.0120, -0.0175, -0.0176,
        -0.0009, -0.0164, -0.0223, -0.0042, -0.0060,  0.0054, -0.0105, -0.0019,
         0.0026,  0.0255,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0662e-02,  5.7468e-01, -8.1459e-02, -2.4521e-02, -7.2162e-03,
        -4.8571e-02,  6.8522e-03,  1.1981e-02, -2.7615e-02, -2.8158e-03,
         3.6994e-04,  4.5357e-03, -1.0866e-03,  3.5639e-03,  9.7277e-03,
        -1.7543e-02, -8.6205e-03, -6.7653e-03, -5.9384e-04, -6.9252e-03,
        -9.5253e-03, -3.4325e-03, -3.1480e-02,  1.0090e-03, -2.6964e-03,
         1.3113e-04,  1.2194e-02,  1.3333e-03, -5.2881e-03, -1.9851e-02,
         3.5240e-03, -6.2347e-03,  7.1245e-03,  2.9448e-03,  8.6767e-03,
        -4.3415e-03, -1.4111e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0409, -0.6164, -0.0433,  0.0032, -0.0029,  0.0069, -0.0024,  0.0046,
         0.0055, -0.0026, -0.0073,  0.0966, -0.0074, -0.0140, -0.0048,  0.0034,
        -0.0033,  0.0008,  0.0155, -0.0101, -0.0157, -0.0173, -0.0651, -0.0099,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1538e-03, -6.2686e-01, -1.7037e-02, -1.7042e-02, -2.2635e-02,
        -1.4686e-03, -3.8505e-03, -3.3446e-03, -2.3907e-02, -7.1919e-03,
         1.1492e-02, -3.7149e-03, -2.6541e-03, -3.6124e-03, -1.1695e-02,
         7.3570e-03,  2.1924e-03,  7.3435e-03,  7.2865e-04,  1.0609e-02,
        -7.3470e-04,  1.2484e-02, -7.0684e-03, -5.1801e-03, -1.3585e-02,
        -2.0018e-02, -5.4451e-03, -1.0423e-02, -2.0279e-03, -2.6475e-03,
        -1.4159e-02,  2.6062e-05, -1.2062e-04,  6.6178e-03, -5.0876e-03,
         5.6347e-03,  1.3998e-03, -7.0497e-04, -4.7041e-03,  4.3508e-03,
         2.0401e-02,  6.4588e-03,  6.0688e-03, -5.5010e-03,  4.3796e-02,
         7.4666e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0090,  0.5728, -0.0151,  0.0194, -0.0062,  0.0271,  0.0319, -0.0127,
        -0.0037, -0.0242,  0.0220,  0.0229,  0.0251,  0.0017,  0.0030,  0.0042,
        -0.0010, -0.0646, -0.0128,  0.0076,  0.0031,  0.0184,  0.0915,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0074, -0.8052,  0.0087, -0.0087, -0.0367,  0.0327, -0.0169,  0.0483,
         0.0074,  0.0190, -0.0090,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0986,  0.7036,  0.0203, -0.0506, -0.0148,  0.0528,  0.0077,  0.0017,
        -0.0080, -0.0045, -0.0026, -0.0012,  0.0008,  0.0327,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0498, -0.6353, -0.0866, -0.0191,  0.0164, -0.0094,  0.0266,  0.0058,
         0.0007,  0.0274, -0.0101, -0.1128,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.5378e-02,  5.5396e-01, -2.4128e-02, -4.6430e-02, -9.8807e-03,
        -1.3364e-02, -1.0127e-02,  6.7012e-03,  4.4151e-04,  1.1384e-02,
        -2.4414e-03, -4.3821e-03,  4.9513e-03, -3.8704e-03, -5.5949e-04,
        -6.6948e-03, -6.0819e-03, -9.5239e-03, -8.9613e-03, -1.2793e-02,
        -2.2212e-03,  4.1497e-04,  3.3868e-03, -6.5300e-03,  1.3453e-02,
        -3.2676e-02,  1.5344e-02,  3.5533e-03, -6.8042e-03,  2.5617e-03,
        -5.0016e-03,  1.1561e-02,  4.2448e-03,  1.3174e-02,  1.0208e-02,
         9.3726e-05,  4.5878e-03,  3.3749e-03, -9.0247e-04,  4.9199e-03,
        -1.3805e-03,  3.4345e-03,  5.2599e-03,  1.7356e-03, -1.0091e-03,
         4.2419e-03,  3.6581e-03, -3.0636e-03,  1.5937e-04,  1.3605e-03,
         5.6208e-03, -2.1103e-03,  1.5839e-04, -2.3546e-04, -1.9502e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-4.8004e-03, -7.0326e-01, -5.3068e-02, -3.2053e-02, -5.2820e-04,
        -8.3596e-03, -2.4707e-02,  4.8484e-04, -4.2505e-03, -1.4293e-02,
        -1.6916e-04, -7.0719e-05, -1.8940e-02,  8.8424e-04,  2.6654e-02,
         7.6656e-03,  2.3768e-02, -1.9178e-02, -3.9703e-04, -1.2883e-02,
         1.5912e-02, -2.5809e-02,  1.8704e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0036, -0.7352, -0.0358,  0.0038,  0.0015,  0.0039,  0.0087,  0.0229,
         0.0041, -0.0397,  0.0217, -0.0128,  0.0064,  0.0043,  0.0142, -0.0073,
         0.0385, -0.0355,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1173,  0.5217,  0.0388, -0.0118,  0.0387,  0.0211, -0.0130,  0.0252,
         0.0280, -0.0204,  0.0128,  0.0113,  0.0249, -0.0125, -0.0138,  0.0122,
         0.0441, -0.0325,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1177, -0.6414,  0.0143,  0.0024, -0.0143,  0.0271, -0.0036,  0.0471,
         0.0210,  0.0080, -0.0007, -0.0320, -0.0244,  0.0180,  0.0282,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7503e-02,  5.5855e-01, -8.3652e-03,  7.2824e-03, -8.2396e-03,
         1.3715e-02,  2.5361e-02,  4.3443e-03,  4.1963e-02,  2.0307e-02,
        -1.7849e-03,  3.4023e-03,  5.4931e-03,  3.5969e-04,  1.2857e-02,
         8.8928e-03,  2.5507e-03,  1.6660e-02,  1.0801e-02, -1.9976e-02,
         4.6933e-03, -2.1269e-02,  3.3190e-03,  2.0444e-03,  6.3578e-03,
         2.4509e-02,  5.0837e-02,  7.5840e-03, -3.5018e-02,  3.4782e-02,
         1.9173e-02,  2.0072e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0256, -0.5329,  0.0675,  0.0143, -0.0348,  0.0043, -0.0007,  0.0091,
         0.0013, -0.0020, -0.0031, -0.0046, -0.0115,  0.0247,  0.0010,  0.0155,
        -0.0032,  0.0127,  0.0211,  0.0164,  0.0268,  0.0272,  0.0202, -0.0068,
         0.0059,  0.0026,  0.0023, -0.0315, -0.0006,  0.0062, -0.0023, -0.0144,
        -0.0474,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0268e-02, -6.5542e-01,  5.4534e-03, -5.0512e-02, -4.0328e-02,
        -1.6938e-02, -3.2885e-03,  1.2816e-03,  6.3858e-03, -1.0440e-02,
        -5.3721e-03,  1.7251e-02,  6.1617e-03,  6.9634e-03,  2.4514e-03,
        -4.7597e-04, -6.6138e-04,  1.2560e-03,  1.2595e-02,  2.8242e-03,
         1.0619e-03,  1.7442e-03,  8.8841e-03, -2.2958e-03, -4.9506e-03,
         7.7448e-03, -7.9623e-03, -1.0442e-03, -7.1293e-04, -1.2407e-03,
        -4.7098e-03, -2.9445e-03, -2.1562e-03,  1.5665e-03, -8.8931e-04,
         1.0702e-02,  1.1696e-03,  8.0894e-04, -1.0155e-03, -7.6047e-03,
         6.1853e-04,  1.5697e-04,  1.9351e-03,  5.6620e-03,  3.1932e-03,
        -2.5449e-03,  2.6073e-03,  2.9629e-03, -8.2360e-04, -3.4959e-04,
         9.5134e-03, -3.1985e-03, -2.3020e-03, -1.1142e-03, -8.3768e-04,
        -2.1187e-03,  5.6092e-03, -6.9133e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7088e-02, -7.5029e-01, -6.9039e-02, -7.8589e-03, -2.8221e-03,
        -1.3251e-03, -1.6276e-02,  4.3091e-03,  1.2288e-03, -2.8836e-03,
         1.0629e-03,  5.1886e-03, -4.6371e-03,  3.0508e-02,  7.7196e-03,
        -3.0601e-03,  2.5524e-03,  3.9012e-03, -4.9213e-06, -1.7521e-02,
        -3.2854e-03, -4.1573e-03,  1.6440e-02, -2.9047e-03,  2.7740e-03,
         1.1159e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1709e-02,  7.0346e-01, -3.2469e-03, -2.0363e-02,  5.3308e-03,
         3.0126e-02,  1.2346e-02, -1.0665e-02,  1.5496e-02, -3.6952e-03,
         4.4337e-03, -2.2521e-02, -8.4418e-04, -1.2561e-02,  5.4200e-03,
        -1.9997e-03, -4.2521e-03, -1.1389e-02,  1.8894e-02, -3.9624e-04,
        -2.1286e-03, -3.5288e-02, -6.7107e-03, -1.3442e-03, -5.5377e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4647e-03, -6.3758e-01,  1.2170e-02,  3.8505e-03, -1.3616e-02,
         2.0646e-02, -1.2837e-02, -6.5855e-03, -9.3244e-03, -2.6091e-02,
        -1.7902e-03,  5.7706e-03, -1.5214e-02, -2.6355e-02,  4.0864e-02,
         1.0557e-02,  6.0175e-03, -7.5622e-03, -3.7431e-03, -2.5320e-03,
        -1.1953e-04,  1.5047e-03,  8.1925e-03,  2.4460e-03,  4.3615e-05,
        -1.5385e-03,  3.0884e-03, -3.2010e-04,  1.9200e-03, -8.1766e-03,
         7.9694e-04,  1.2996e-02, -1.1522e-02, -3.8736e-03, -1.0698e-02,
         1.2062e-03,  2.5262e-03, -1.0479e-02, -2.6115e-03,  1.2801e-03,
         1.4906e-03, -4.2893e-03, -1.1358e-02,  6.4445e-03,  2.5509e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.4308e-02, -6.3868e-01, -5.7385e-02, -4.3743e-02,  2.6002e-03,
         6.0682e-04, -3.5606e-03, -1.6986e-02,  1.1579e-02,  8.3425e-03,
         2.1441e-02,  6.3160e-03, -9.4871e-03,  3.1410e-03, -3.1534e-03,
         1.5669e-02,  1.1581e-02,  7.3151e-03,  8.5830e-04,  5.1607e-03,
         1.0618e-02,  1.7745e-02,  9.7240e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0685,  0.8175,  0.0201, -0.0084, -0.0056,  0.0154,  0.0246, -0.0066,
         0.0044, -0.0030,  0.0013, -0.0244,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 2.2829e-03,  6.6522e-01,  7.2996e-03, -4.1307e-02, -2.1805e-02,
        -5.8221e-03,  2.3594e-02,  2.0595e-02,  4.4363e-02, -1.8449e-02,
        -1.0311e-02,  6.4392e-04,  7.4689e-03,  8.3686e-04,  4.0827e-03,
         2.3104e-03, -2.2384e-03, -1.8238e-02, -8.5087e-03, -3.9419e-03,
        -4.2501e-03, -6.6303e-03, -7.9800e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.0560e-02, -6.8553e-01,  3.0335e-02,  4.2434e-03, -5.7628e-03,
         1.0761e-02, -3.1565e-03,  1.6913e-02, -5.5768e-04, -7.9694e-03,
        -2.0326e-03,  5.4317e-03,  6.3066e-03, -1.1554e-03, -6.2993e-03,
         4.1718e-03,  6.2501e-03,  2.6641e-02,  1.4500e-02, -3.4953e-03,
        -2.3080e-02,  3.1413e-03, -8.6531e-04,  1.2762e-02,  2.1216e-04,
        -3.0333e-03,  6.6570e-03,  3.5813e-03,  1.3675e-03, -5.0670e-03,
         7.2287e-03,  1.6596e-02,  4.3390e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6777e-03,  7.5111e-01, -5.3380e-02, -3.0233e-03, -1.0483e-02,
         1.1694e-02, -1.6739e-02,  2.5168e-03,  2.0117e-03, -1.3445e-02,
        -2.7032e-02,  4.9858e-03, -2.4625e-04,  2.5686e-02,  2.2662e-02,
        -2.0230e-02, -1.7355e-02, -1.1953e-02, -1.7655e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0801, -0.6015,  0.0202, -0.0392, -0.0337, -0.0461, -0.0292,  0.0056,
        -0.0295,  0.0266,  0.0194, -0.0242, -0.0237,  0.0210,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0276, -0.7244,  0.0324,  0.0173, -0.0134, -0.0067, -0.0111,  0.0136,
         0.0098,  0.0138,  0.0058,  0.0010, -0.0104,  0.0070,  0.0039,  0.0022,
         0.0186,  0.0070,  0.0067,  0.0009, -0.0249, -0.0412,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0370e-02,  7.5466e-01,  3.1385e-02, -2.5126e-02, -2.7995e-03,
         1.3459e-02, -3.5664e-03, -3.1032e-03, -1.5966e-03, -2.2690e-02,
        -1.8208e-04, -7.9054e-03, -2.8697e-02, -2.3644e-03, -2.7105e-04,
         4.4638e-03,  1.5793e-02, -5.3312e-03, -1.7226e-03,  1.8160e-02,
        -5.7430e-03, -1.6171e-03, -6.4017e-04,  1.2697e-03, -8.8585e-04,
         1.1419e-03, -6.3368e-03,  1.6146e-02, -9.8804e-03, -2.6977e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0336e-03,  7.4270e-01,  5.7002e-02, -1.5483e-02,  1.8779e-02,
        -1.7435e-02, -4.7007e-02,  3.0340e-03,  8.4992e-03,  7.0126e-04,
        -1.1275e-02, -1.6060e-02,  6.5954e-03, -8.8833e-03, -1.7714e-02,
        -5.5783e-03,  1.7224e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0090, -0.7428,  0.0173, -0.0954,  0.0379,  0.0016,  0.0112, -0.0134,
         0.0301,  0.0181,  0.0234,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5947e-02,  7.3797e-01,  2.7022e-02, -1.6636e-02, -1.0882e-02,
        -4.9539e-02, -1.0013e-02, -1.2946e-02, -7.4724e-05,  3.4766e-03,
        -5.3922e-04,  1.1835e-02, -9.4748e-04,  1.6217e-02, -4.0884e-02,
        -2.4113e-03, -2.6572e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7880e-02,  6.9759e-01, -1.8211e-02,  1.3369e-02,  7.6686e-03,
         1.4391e-02, -1.2176e-02,  7.6038e-03, -2.3257e-03, -1.6405e-02,
        -1.3658e-02,  9.5916e-03, -5.8064e-03, -1.1190e-02, -6.4966e-02,
        -2.5970e-03,  6.2798e-03, -9.9198e-03, -1.2001e-04,  8.2100e-04,
        -5.5061e-03, -4.9608e-03, -7.3406e-03,  3.5057e-03,  1.3970e-02,
         2.1505e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0503,  0.5992,  0.0810, -0.0153, -0.0052,  0.0205,  0.0093,  0.0249,
         0.0137,  0.0037, -0.0053, -0.0175, -0.0055,  0.0128,  0.0010, -0.0137,
        -0.0014,  0.0121, -0.0046, -0.0023, -0.0034, -0.0067,  0.0045, -0.0038,
         0.0043, -0.0081,  0.0025,  0.0055, -0.0302,  0.0316,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0799, -0.5856, -0.0329,  0.0603, -0.0144,  0.0049, -0.0160,  0.0196,
         0.0036,  0.0327,  0.0138, -0.0280, -0.0162,  0.0087,  0.0111,  0.0052,
        -0.0055,  0.0440, -0.0175,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([-0.0107, -0.6339, -0.0529, -0.0115, -0.0260, -0.0222, -0.0397, -0.0038,
        -0.0163, -0.0140, -0.0015, -0.0048, -0.0107, -0.0075, -0.0063,  0.0131,
        -0.0061,  0.0140,  0.0052, -0.0089,  0.0028,  0.0067, -0.0061,  0.0009,
         0.0177,  0.0335,  0.0230,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0183, -0.7257, -0.0274,  0.0042,  0.0429, -0.0280,  0.0015,  0.0083,
         0.0190, -0.0196, -0.0146,  0.0268, -0.0553, -0.0085,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0068, -0.6528, -0.0038,  0.0204, -0.0204, -0.0258, -0.0200, -0.0081,
        -0.0313,  0.0103, -0.0076, -0.0130,  0.0016, -0.0089, -0.0044,  0.0291,
         0.0047, -0.0239,  0.0023, -0.0425,  0.0154, -0.0059,  0.0022, -0.0041,
         0.0150,  0.0196,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.7823e-02, -7.0679e-01, -6.3106e-02,  4.1816e-03,  8.2557e-03,
         4.3580e-02, -2.3468e-03,  2.0332e-02,  1.0041e-02, -1.8878e-02,
         1.1966e-02,  4.6125e-02, -5.9089e-03, -6.6591e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8149e-02, -6.7085e-01,  7.1883e-02,  1.0786e-03,  7.3515e-03,
         1.4104e-02,  1.5628e-02,  9.4119e-03, -6.7986e-03,  5.3323e-03,
         1.0784e-02,  1.0953e-02,  5.6670e-04, -1.8420e-02, -1.9275e-02,
         9.6530e-03,  6.5879e-03,  2.0576e-03,  5.8948e-03,  2.6146e-03,
         6.4750e-03,  7.9374e-03, -4.7450e-03,  1.7024e-03,  8.1700e-03,
         1.2598e-02, -2.5769e-03,  5.1493e-03,  4.7868e-03,  3.5836e-03,
         1.8380e-02, -1.6501e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.3392e-02, -6.4431e-01, -3.7504e-02, -4.4439e-04, -4.7942e-02,
        -7.0018e-03,  1.8629e-02,  1.9454e-02, -1.3543e-02,  1.8769e-02,
         2.2388e-02,  6.1263e-03,  2.5702e-03,  1.0327e-03,  4.9739e-04,
        -2.7587e-03, -9.7668e-04, -3.5696e-03, -7.5125e-03, -7.8662e-04,
         7.2984e-03, -1.8658e-03,  1.7540e-03, -6.0188e-03,  4.3854e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5338e-02,  7.8421e-01,  1.9981e-02,  4.8481e-04, -9.6924e-02,
        -8.3063e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.8259e-02,  7.7923e-01, -8.9823e-03,  1.4515e-02,  9.9580e-04,
        -6.4776e-03,  8.2621e-03, -3.7503e-03,  2.5107e-03,  4.6793e-03,
        -1.5435e-03,  1.2446e-03, -1.6634e-03, -1.2780e-02, -4.8490e-03,
         5.0788e-03,  2.4393e-04, -3.6986e-04, -4.6926e-04,  2.4203e-03,
        -2.6314e-03, -1.1462e-03,  1.0722e-02, -2.7645e-03, -6.0367e-03,
         9.8610e-03,  1.2141e-03,  3.3248e-03, -2.3422e-03, -5.1193e-03,
         5.9960e-03,  6.6383e-03, -2.1100e-03, -2.2939e-03, -2.3951e-03,
         6.3520e-03,  1.0730e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.1836e-03, -6.0406e-01, -4.6768e-02, -7.3900e-02, -1.0267e-02,
        -1.3738e-02, -8.5401e-03,  4.5890e-03, -3.3408e-03, -6.9363e-03,
         7.7236e-03, -5.0534e-03, -4.8300e-03,  6.8446e-03, -4.0512e-03,
         2.0366e-02,  1.5314e-02, -2.6959e-02, -2.9468e-03, -1.4190e-03,
         3.2199e-02,  3.3309e-04,  1.3955e-02, -1.9096e-03,  2.4140e-03,
        -2.5359e-03, -2.8416e-03, -8.2417e-03,  5.6583e-03,  7.6920e-03,
         4.4507e-03,  4.3291e-03, -5.5091e-03, -1.5191e-02,  1.5910e-02,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0537e-01, -5.9276e-01,  4.4998e-02,  1.6856e-02, -2.2763e-02,
        -5.3715e-03, -8.6940e-03, -4.4588e-04,  1.2109e-02, -1.6253e-02,
         2.2365e-02, -4.3794e-02,  1.9783e-02, -2.1361e-02,  6.7077e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3603e-02,  7.0574e-01,  5.7858e-02, -3.3064e-02, -7.3230e-03,
         6.7935e-03,  1.4490e-03,  3.3860e-02, -7.7870e-03,  1.1790e-02,
        -2.0743e-03, -4.2266e-03, -3.9931e-03, -1.2168e-03, -3.6877e-03,
         7.1494e-03,  5.3570e-04, -1.2806e-04,  7.1106e-03, -1.0066e-02,
        -1.0194e-02, -5.4042e-03, -1.2431e-03,  2.9301e-03,  9.5509e-03,
        -7.8463e-03, -2.9958e-03, -4.8725e-03, -7.1752e-03, -3.1358e-03,
         2.0368e-03, -1.0274e-02,  1.2882e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5528e-02, -6.3072e-01, -4.1163e-02,  4.4820e-02,  1.7117e-02,
         1.7701e-04, -1.4211e-02,  2.1989e-03,  1.1301e-02,  3.1232e-03,
        -5.9613e-03, -2.1604e-02,  2.7892e-03,  2.1061e-02, -9.2095e-03,
         4.0000e-03, -1.6537e-02,  9.6269e-03,  5.1500e-03, -8.3454e-04,
         5.5957e-03, -5.5727e-03, -1.0674e-02,  3.4526e-03, -2.9557e-03,
        -2.1210e-03,  9.0370e-03, -3.2643e-03,  4.0192e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 8.3390e-03,  7.8223e-01, -9.2721e-03, -1.9335e-02,  1.5424e-02,
        -1.1837e-02,  6.4967e-03,  1.7515e-03, -2.1075e-02, -4.9698e-02,
        -7.4862e-03, -1.1340e-02,  2.7935e-04,  6.7653e-03, -4.8668e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9734e-02, -7.2542e-01,  4.5938e-02, -1.9062e-03, -1.5746e-02,
        -2.7331e-02, -7.8190e-03, -5.0999e-03, -7.6142e-03, -9.2747e-03,
        -7.9734e-04,  5.3688e-03,  5.9107e-03, -6.8057e-03,  4.9486e-04,
        -1.2729e-03,  2.5748e-03, -4.4483e-03,  1.1356e-03,  8.7487e-04,
        -4.5861e-03,  2.9690e-03, -9.9026e-04,  6.0691e-04, -1.4111e-03,
        -1.4266e-03,  2.1598e-03, -8.7068e-03,  2.9141e-03, -2.3053e-03,
         5.3133e-04,  9.8579e-04, -4.9122e-03,  6.1312e-03,  2.3178e-03,
        -1.6734e-03, -3.1179e-03,  2.4650e-03,  2.7324e-04,  3.6849e-03,
        -4.2450e-03, -1.6468e-03, -4.8163e-03,  6.9818e-03,  2.4772e-04,
         9.1286e-03, -1.9021e-03,  1.9622e-03,  1.9333e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4522e-03,  6.3298e-01, -1.1779e-01, -3.5497e-02, -1.1049e-02,
         2.4243e-02,  8.9830e-03,  8.3502e-03, -9.1145e-03,  1.9403e-03,
        -1.0482e-02, -5.7122e-03, -2.8721e-03,  3.8967e-03, -9.3046e-03,
        -1.2927e-02, -1.0861e-02,  1.6933e-03,  4.8649e-03,  9.1611e-05,
        -1.1951e-03, -6.4312e-04, -9.7286e-03,  6.3712e-04,  5.5804e-03,
        -4.0334e-03, -1.6981e-02, -3.2584e-03,  4.1166e-03, -9.0670e-03,
        -5.7426e-03,  1.3675e-03,  1.1067e-03,  1.5685e-02, -2.7534e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0253, -0.7350, -0.0243, -0.0128, -0.0071,  0.0140,  0.0042,  0.0009,
        -0.0071,  0.0125,  0.0239,  0.0136,  0.0056,  0.0079, -0.0074, -0.0164,
        -0.0041,  0.0664, -0.0114,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1993e-02, -6.5371e-01,  3.7675e-02,  9.6441e-03, -4.6433e-02,
        -1.6712e-02,  1.5652e-03, -1.4323e-02, -4.6378e-02,  2.1808e-02,
         5.6034e-04, -7.2534e-03, -2.4496e-02,  7.4749e-02,  2.2699e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2833e-02,  7.6092e-01, -3.7279e-02, -6.3452e-03, -5.0187e-04,
        -1.1425e-02, -4.3361e-03,  2.4040e-02,  1.5870e-03, -9.3171e-03,
        -4.0275e-03, -3.4333e-04, -9.6375e-03,  7.3784e-03, -7.5056e-03,
        -7.4848e-03, -9.0668e-03, -2.9803e-03, -7.2976e-03,  2.1187e-02,
        -8.5015e-03, -1.6009e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0258,  0.6820, -0.0549,  0.0030,  0.0053, -0.0062, -0.0125,  0.0172,
        -0.0153,  0.0105, -0.0008,  0.0095, -0.0044, -0.0032, -0.0141, -0.0014,
         0.0015,  0.0113,  0.0009, -0.0058, -0.0022,  0.0034,  0.0104,  0.0137,
         0.0102, -0.0038,  0.0052, -0.0095,  0.0142,  0.0418,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0193, -0.7102,  0.0379,  0.0050,  0.0031,  0.0622,  0.0193, -0.0145,
         0.1061, -0.0037,  0.0185,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9389e-02,  7.8872e-01,  3.6831e-02,  7.7654e-03, -5.1002e-03,
         2.3328e-02,  2.7149e-02,  5.3841e-04, -5.0964e-03,  6.2513e-03,
        -1.3234e-02,  1.7333e-03, -4.7584e-03,  2.7151e-03,  7.2785e-03,
         1.1607e-02, -2.7236e-03, -9.4309e-03, -4.3062e-03, -5.5359e-03,
         1.0956e-02, -5.5516e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0012,  0.7172,  0.0093,  0.0077,  0.0086, -0.0053,  0.0172,  0.0076,
        -0.0016,  0.0021, -0.0296, -0.0142, -0.0050, -0.0088,  0.0064, -0.0088,
         0.0053,  0.0008, -0.0043, -0.0043, -0.0062,  0.0087, -0.0135, -0.0085,
        -0.0146, -0.0111,  0.0062, -0.0109, -0.0085, -0.0095, -0.0030,  0.0026,
         0.0313,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4339e-02, -6.4503e-01, -4.9483e-02,  1.2492e-02,  1.2660e-02,
        -1.5280e-02,  9.4226e-03,  2.6387e-02, -4.2794e-03,  9.7515e-04,
         1.2393e-03, -4.7145e-03,  3.5491e-03,  1.4355e-02,  2.0800e-03,
        -9.1062e-02, -1.5966e-03, -3.5983e-03,  5.7589e-04, -1.1425e-02,
         1.7713e-02,  6.1933e-03, -2.4621e-03, -1.5008e-03, -4.1251e-03,
         4.9108e-03,  8.1773e-03,  3.0378e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1117e-02, -6.6959e-01, -9.9998e-02,  2.4185e-02, -9.0671e-03,
        -9.9808e-04,  2.3721e-03,  4.7986e-02,  8.7872e-03,  7.6578e-03,
        -2.0846e-03,  3.5175e-03,  9.0118e-04,  1.9596e-03,  1.7350e-04,
         1.3126e-03, -7.2585e-03, -1.3629e-03, -5.4906e-03, -1.4917e-02,
        -3.3263e-04,  1.1933e-03, -8.3443e-03,  4.3909e-02,  5.4801e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-4.0633e-02, -6.8663e-01, -4.5868e-03,  1.6171e-02,  1.7420e-02,
         1.4167e-03,  1.7345e-02,  2.3388e-02, -1.0994e-02, -6.1265e-05,
         1.2581e-02, -7.2292e-04,  5.8062e-03,  1.2444e-02,  3.8422e-03,
         2.4009e-02,  5.3600e-03,  2.4209e-03,  2.7532e-03,  7.2753e-03,
        -5.6776e-04,  9.9063e-03,  1.1303e-02,  6.6861e-03,  3.5911e-03,
         2.5500e-03,  1.1991e-02,  5.6977e-03,  1.0047e-02,  1.2436e-02,
        -3.3723e-03,  2.9044e-03,  4.0526e-04,  2.6215e-04,  3.6259e-04,
         1.2933e-03,  4.2480e-04, -4.0525e-03, -4.9636e-03, -8.6506e-03,
         2.6714e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0257,  0.7345,  0.0243, -0.0083,  0.0290,  0.0092,  0.0108, -0.0063,
        -0.0040,  0.0047, -0.0036,  0.0147,  0.0231,  0.0041,  0.0173,  0.0033,
         0.0021, -0.0040, -0.0018,  0.0178,  0.0063, -0.0188,  0.0071, -0.0134,
        -0.0058,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0606,  0.7064,  0.0202,  0.0056,  0.0148,  0.0202, -0.0027,  0.0159,
         0.0023, -0.0108,  0.0042,  0.0046,  0.0082,  0.0060, -0.0029,  0.0009,
         0.0164, -0.0070, -0.0152, -0.0085, -0.0665,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.9545e-02,  6.9308e-01, -1.6096e-02,  1.7953e-02, -1.9782e-04,
         9.1176e-03,  4.2622e-02,  5.6308e-03, -2.5317e-02, -1.4967e-02,
        -1.7253e-02, -8.8787e-03,  2.1855e-03, -4.3961e-03, -1.0692e-02,
         5.2296e-03, -4.0275e-02, -6.5630e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.3415e-02,  5.8156e-01,  5.5853e-02, -3.1681e-02,  1.6777e-02,
         4.3700e-03,  6.4459e-03,  3.8569e-03, -5.4587e-03,  4.2788e-03,
         1.5468e-02,  1.1609e-02, -8.2022e-03,  1.7170e-03,  3.1501e-03,
         4.9588e-03, -1.1292e-02,  4.0895e-03,  1.4985e-02,  7.4430e-03,
        -2.9840e-03, -3.0086e-03, -1.1137e-02,  5.7804e-03,  5.5907e-04,
         3.3575e-03,  6.1413e-03,  7.2588e-03, -5.1377e-03,  1.0406e-03,
         4.8841e-03, -8.3636e-04,  6.5381e-03,  1.1653e-02, -6.2777e-03,
         5.0208e-03,  2.6446e-02,  1.5330e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0876,  0.6460, -0.0032,  0.0140,  0.0103,  0.0174,  0.0097, -0.0320,
         0.0115,  0.0141, -0.0134, -0.0136,  0.0178, -0.0199, -0.0086,  0.0191,
        -0.0121, -0.0073, -0.0026, -0.0398,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0870, -0.6771,  0.0087, -0.0295, -0.0143, -0.0257, -0.0213, -0.0443,
        -0.0136,  0.0089,  0.0383, -0.0162,  0.0117, -0.0035,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5710e-02,  6.7480e-01,  2.5282e-02,  3.8753e-02,  1.3855e-02,
         1.0070e-02,  3.2472e-02,  6.8868e-03,  6.7404e-03, -4.6352e-03,
         1.5030e-02,  1.1548e-03,  7.2780e-03, -1.6402e-03,  4.3324e-03,
        -7.5683e-03, -4.5988e-04, -3.9237e-03, -5.8711e-03, -4.4122e-03,
         4.3981e-03,  2.6129e-02,  2.8642e-03,  2.5703e-03,  1.0086e-02,
         2.8868e-03,  7.4891e-04,  1.1802e-02,  8.1126e-03,  1.8263e-02,
        -3.0795e-03,  1.8184e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3540e-02, -6.7584e-01, -4.9654e-04, -3.2067e-02,  7.2038e-03,
        -1.7971e-02,  1.3672e-02,  4.3476e-03, -1.9331e-02, -5.8218e-04,
         5.5210e-03,  3.5695e-03, -7.4777e-04,  1.5146e-02,  1.4891e-02,
        -1.8920e-03, -4.7665e-03, -3.3646e-03, -2.0132e-02,  7.4917e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2002e-02,  5.9142e-01,  6.8067e-02, -1.8906e-02, -1.3837e-02,
         2.7141e-02,  2.9156e-03, -1.0252e-02, -5.8754e-04, -7.3985e-03,
        -1.0429e-02,  3.0173e-03,  8.8239e-04,  3.3222e-04,  9.3789e-04,
        -8.3560e-03, -2.4178e-03,  8.0656e-04, -5.9388e-03,  9.0263e-04,
         3.4648e-03,  3.8338e-03,  1.2515e-02, -4.4083e-04, -1.7703e-02,
        -3.0605e-03, -1.5063e-02, -1.0675e-02, -4.3943e-03, -3.8651e-03,
         2.1930e-03, -6.2325e-03, -1.7230e-02, -4.8072e-03,  5.5848e-03,
        -1.6990e-03,  4.5076e-03, -5.3754e-03,  2.8146e-04,  1.6897e-04,
         1.4522e-03,  2.8235e-03,  5.9511e-03,  1.4115e-03,  6.0937e-03,
        -3.2257e-02,  3.0371e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.9715e-02,  6.7684e-01,  5.3000e-03, -4.1048e-02,  4.0520e-02,
        -1.4119e-02,  7.2062e-03, -4.2271e-03, -5.8357e-02,  1.3232e-02,
        -7.7558e-03,  6.4313e-04,  2.2522e-03,  9.0882e-04, -2.7878e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.0659e-02,  6.5218e-01,  8.1834e-03, -2.9528e-03,  1.0410e-03,
        -6.9794e-03, -1.3647e-03,  1.8203e-02,  7.7713e-03, -3.7398e-03,
         7.8711e-03, -6.3849e-03, -2.5365e-03,  1.3825e-02,  5.3075e-03,
         2.7697e-03,  2.3183e-03,  1.9783e-03,  1.6526e-03, -3.9881e-04,
         9.2971e-04,  5.3912e-03,  7.9448e-03,  6.7940e-03, -2.8969e-03,
        -8.6668e-03,  1.4612e-02,  7.0755e-03, -3.6365e-03, -2.9254e-03,
         3.9236e-03,  9.5608e-04,  7.2860e-03,  1.5470e-03,  3.8217e-03,
         1.5234e-02,  2.9839e-03,  8.4608e-04, -7.3294e-03,  4.6741e-03,
         5.4433e-03,  2.8856e-03, -1.0304e-02, -4.6921e-03, -6.8060e-03,
         4.2335e-03,  8.1128e-03,  2.9932e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 9.2532e-03, -7.2786e-01,  5.7127e-03, -2.8992e-02,  2.3443e-02,
         6.1323e-03,  6.9397e-04, -1.7456e-02,  1.2076e-02, -8.9256e-03,
        -3.2772e-03,  1.8874e-02,  9.1716e-03,  1.4261e-04,  4.2649e-03,
        -2.2214e-03,  5.4761e-03,  7.5193e-03,  4.2959e-03, -2.6499e-03,
        -1.3219e-03,  1.2705e-02,  8.7373e-03, -1.8763e-02, -8.6358e-03,
         5.1396e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.4224e-03, -7.7371e-01, -2.6346e-02, -7.9287e-03,  3.6638e-03,
         9.6525e-03,  6.0359e-03,  9.2761e-03,  1.3451e-02,  2.8664e-03,
        -1.3075e-02,  1.8133e-02,  2.7615e-03, -4.0379e-04,  3.3416e-03,
        -1.0487e-02,  5.5375e-03, -8.4637e-03,  2.6814e-02, -6.2625e-03,
         4.4371e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4152e-02,  5.8854e-01, -3.3932e-02,  7.8543e-02,  3.2305e-02,
        -2.3811e-02, -1.1024e-03, -5.8128e-03,  5.9903e-03, -5.5781e-03,
        -1.5719e-03,  3.9185e-03, -1.2592e-02,  1.7398e-03, -8.5351e-03,
        -8.2962e-03,  1.9671e-02, -1.4710e-02,  4.3232e-03, -8.8999e-03,
         3.9202e-03, -1.1895e-03, -1.5786e-02,  6.4941e-03, -1.0183e-03,
        -8.2552e-04, -4.1947e-04,  4.5360e-03, -2.7256e-03, -2.4418e-03,
        -1.4538e-02,  4.7731e-03, -2.8615e-04, -8.4896e-04, -1.7234e-02,
        -2.8944e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2614e-02,  6.8266e-01,  7.9346e-03,  3.0993e-04, -1.7844e-02,
         2.9036e-02,  7.1450e-03, -2.2345e-02,  1.8700e-02, -2.2150e-03,
        -6.1530e-02, -1.1880e-02, -1.0738e-02,  8.4078e-03, -3.2030e-04,
        -9.2326e-03, -5.6937e-03, -1.1854e-02,  1.0499e-02, -6.9040e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0400,  0.6145,  0.0611,  0.0207,  0.0134,  0.0035, -0.0475, -0.0044,
        -0.0033,  0.0286,  0.0321,  0.0104, -0.0159, -0.0025,  0.0078,  0.0074,
         0.0101, -0.0072, -0.0193,  0.0151, -0.0350,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.4673e-02,  7.1156e-01,  7.7149e-02,  1.0958e-02, -7.6327e-03,
         4.3990e-03, -9.4112e-04, -2.9419e-02,  3.4856e-03, -4.6395e-03,
         1.7126e-02, -2.0504e-04,  3.7809e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3762e-02,  6.5656e-01,  2.8755e-02,  3.2841e-03, -2.0245e-02,
        -1.1537e-02, -7.2626e-03, -1.2376e-03,  9.7547e-03,  2.6083e-03,
         2.8684e-02, -1.1237e-02,  1.7920e-03, -1.0801e-02,  2.4952e-03,
         4.2879e-03, -1.6318e-02, -4.5695e-03, -8.5220e-03,  4.3977e-04,
        -8.5160e-03, -1.0514e-03, -9.0840e-03, -5.5618e-03, -5.0356e-03,
         7.1467e-03, -6.0285e-03,  4.7003e-03,  4.3961e-03,  4.5583e-03,
         3.3243e-03,  1.5657e-02,  4.1562e-03,  2.4464e-03,  1.2224e-03,
         3.0444e-02,  2.2517e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3488e-02,  7.1018e-01,  5.8505e-02,  4.0497e-02,  1.0190e-03,
         8.7707e-03, -6.2894e-03,  1.3195e-03,  1.9757e-02, -1.0874e-02,
         1.5177e-03,  1.4933e-03,  3.4741e-02, -7.6602e-03,  5.3038e-03,
         8.9559e-03,  1.5640e-02,  1.2346e-02, -2.6186e-03,  2.5197e-04,
        -1.6889e-02, -6.7765e-03, -1.2937e-02,  2.1737e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0491e-01,  5.5611e-01,  9.5140e-02,  4.1393e-02, -5.9986e-03,
         1.2170e-02, -4.6393e-03,  7.3607e-04, -5.9344e-03, -4.1892e-04,
         1.2350e-03,  1.1535e-02,  2.0800e-02,  1.5659e-02, -1.1319e-02,
        -1.6967e-02,  9.6883e-03, -5.9678e-04, -7.3299e-03,  3.7341e-03,
         2.8041e-03,  2.2153e-03, -2.5932e-03, -3.5502e-03, -4.8849e-03,
        -6.0830e-03, -3.0583e-04, -4.5320e-03,  2.2154e-03,  1.9070e-03,
         2.8179e-02,  1.4421e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0520,  0.6904,  0.0037, -0.0053, -0.0217,  0.0016, -0.0032, -0.0365,
         0.0053,  0.0146, -0.0033, -0.0095, -0.0059, -0.0073, -0.0023,  0.0024,
        -0.0178, -0.0109, -0.0054,  0.0055,  0.0033, -0.0159, -0.0022,  0.0072,
        -0.0041, -0.0054, -0.0161, -0.0034, -0.0076,  0.0098, -0.0204,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.2761e-02, -6.7601e-01, -7.0090e-03,  1.5720e-02,  1.9284e-02,
         4.2927e-02,  2.6600e-03,  8.2565e-03,  1.0575e-02,  7.0796e-03,
        -2.8903e-02,  3.8419e-04,  6.1776e-03,  3.5710e-02,  2.7794e-03,
         3.7284e-04, -2.7724e-03,  3.8748e-03,  3.8916e-03, -1.1984e-02,
         2.0555e-02, -2.2921e-02, -7.3892e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0357,  0.6862, -0.0496,  0.0039, -0.0059,  0.0065, -0.0159,  0.0153,
        -0.0042, -0.0013,  0.0057, -0.0010, -0.0150, -0.0216, -0.0015, -0.0216,
        -0.0159, -0.0032, -0.0024,  0.0021, -0.0018, -0.0105, -0.0039,  0.0082,
         0.0044, -0.0056, -0.0345,  0.0165,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #100: [tensor([-4.6557e-02,  6.7398e-01, -4.3309e-02,  6.6351e-04,  1.0300e-02,
         1.0254e-02,  5.6193e-03,  5.3499e-03, -9.1956e-03, -9.4681e-03,
         1.6676e-02, -2.4548e-03,  1.3869e-03,  1.1106e-03, -2.4815e-04,
        -4.7393e-03, -1.1815e-03,  1.4531e-02, -7.6493e-03, -1.6815e-02,
         2.1990e-02,  4.3357e-03,  2.4509e-04,  3.9413e-03, -4.6882e-03,
         1.3205e-02,  1.7709e-02, -4.8856e-03, -4.7511e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0225,  0.6633, -0.0113,  0.0205, -0.0265,  0.0007,  0.0147, -0.0149,
        -0.0206, -0.0012, -0.0160,  0.0103, -0.0030, -0.0167, -0.0068,  0.0014,
        -0.0079, -0.0052,  0.0103, -0.0061, -0.0103, -0.0212, -0.0657,  0.0231,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0253,  0.6270,  0.0386, -0.0099,  0.0176, -0.0190,  0.0233,  0.0058,
         0.0128, -0.0172, -0.0131,  0.0201, -0.0071, -0.0124,  0.0060, -0.0063,
        -0.0129,  0.0022, -0.0079, -0.0093,  0.0094, -0.0027, -0.0061,  0.0124,
        -0.0060, -0.0014, -0.0114,  0.0034, -0.0099,  0.0032, -0.0151, -0.0027,
         0.0063, -0.0162], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2356e-03, -6.7441e-01, -6.1573e-02,  2.2234e-02, -9.2883e-03,
        -1.7751e-02, -3.7226e-02, -1.9148e-02, -1.4866e-02, -6.1832e-03,
        -8.2095e-03, -4.9988e-03,  5.4431e-04,  4.4048e-03, -4.4646e-03,
        -6.9907e-03, -1.1306e-02, -7.8949e-03, -3.5661e-03, -3.9628e-02,
         1.1376e-02,  2.9696e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0023, -0.7653, -0.0582, -0.0113,  0.0178, -0.0060, -0.0013, -0.0092,
        -0.0161,  0.0117,  0.0102, -0.0060,  0.0045, -0.0106, -0.0155, -0.0246,
        -0.0028, -0.0266,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6738e-02,  6.8802e-01,  4.0172e-02, -2.1929e-02, -1.4546e-02,
         3.3839e-02,  4.2845e-03,  2.9726e-02,  1.6096e-02,  1.1927e-03,
         2.6953e-02,  1.0387e-02, -5.7567e-03, -9.9286e-03, -4.7480e-03,
        -2.1261e-02,  1.0502e-03, -5.9643e-03,  2.3925e-03,  4.2333e-03,
        -2.7993e-04, -6.4923e-03,  4.0104e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.4022e-02,  6.8652e-01,  3.1071e-02,  3.3245e-03,  2.4555e-02,
        -6.5715e-03, -2.6327e-02, -4.9449e-03, -2.2534e-02, -9.0483e-04,
         2.1147e-02,  1.9200e-02, -6.5405e-04, -7.1464e-03, -4.3592e-02,
        -1.7404e-03, -1.5746e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0654, -0.7368,  0.0332, -0.0048, -0.0270, -0.0061,  0.0047,  0.0018,
        -0.0030,  0.0080, -0.0100,  0.0062, -0.0092, -0.0093, -0.0161, -0.0159,
         0.0247, -0.0178,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.2439e-02, -7.3346e-01,  1.3101e-02, -9.8504e-03, -1.9391e-03,
        -8.7126e-04, -5.9155e-04,  1.0358e-03,  7.6118e-03, -3.7946e-02,
        -1.2759e-03,  3.8042e-03,  7.8544e-03, -1.4406e-02,  1.0203e-03,
         3.1050e-03,  8.0166e-03,  1.4670e-02,  2.9105e-03,  1.3263e-02,
         8.3061e-03, -4.3985e-03,  9.9006e-03,  1.2073e-02,  1.2002e-02,
         2.4145e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5847e-02, -6.2598e-01, -9.3882e-02,  3.5592e-02, -4.4698e-03,
         1.7750e-03,  4.6551e-03,  1.0963e-02,  1.9420e-02,  9.4360e-03,
         1.3300e-02,  9.2176e-03,  4.4321e-03,  3.8689e-03,  7.8624e-03,
         4.4842e-04, -4.9918e-03,  1.0641e-03, -1.0926e-03,  8.6091e-03,
         1.3950e-02,  4.0579e-03,  4.3394e-03,  2.2941e-03, -3.7842e-04,
         7.6422e-03,  1.1923e-02, -3.8490e-03, -2.6709e-03, -2.4233e-02,
         2.7753e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2552e-02,  7.3745e-01, -4.3982e-04, -1.8401e-03, -1.8615e-03,
         2.5447e-02,  1.2559e-02,  8.1818e-03, -4.2580e-03,  4.5724e-03,
        -6.3552e-03, -1.6472e-02, -2.9937e-02, -4.6484e-03, -2.7192e-03,
        -1.4754e-02, -2.6782e-02, -5.9167e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0096,  0.7924, -0.0403, -0.0480,  0.0483,  0.0352, -0.0016, -0.0246,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 5.7697e-02,  6.7445e-01,  2.3114e-02, -1.3692e-02, -1.7011e-02,
         7.6534e-03, -9.3460e-03, -2.9526e-03, -6.8963e-03, -1.3829e-03,
        -6.1562e-03, -1.6311e-03, -6.0697e-03, -2.3762e-04, -5.6208e-03,
         7.5791e-03,  2.2282e-03, -7.1973e-03, -2.1702e-03,  8.5732e-03,
         1.3125e-03, -7.2463e-03, -2.7023e-02, -9.2364e-03, -1.1168e-02,
         1.4228e-03, -5.6604e-03,  6.5696e-03, -9.8348e-03, -2.6071e-03,
         2.1425e-02,  6.4810e-03,  5.2356e-03,  7.4433e-03, -4.2387e-03,
         3.9263e-03, -3.0862e-03,  1.3753e-03, -3.0490e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9644e-02,  5.9736e-01, -1.0056e-02, -2.7511e-02,  1.3990e-02,
         2.1283e-02, -2.0875e-03,  3.8876e-03,  3.8349e-02,  1.7326e-02,
         1.7279e-02,  3.7716e-02,  1.0396e-02,  2.3888e-02,  3.4227e-02,
         4.8120e-03,  8.1153e-03, -4.9954e-03,  2.7017e-03,  9.1033e-03,
         2.9819e-02, -2.6025e-02, -2.6279e-03,  7.9097e-04, -1.2862e-03,
        -3.6125e-04,  3.2218e-02, -2.1473e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0403, -0.7423,  0.0196, -0.0560, -0.0037,  0.0127, -0.0030,  0.0034,
        -0.0057, -0.0187,  0.0024, -0.0161,  0.0762,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8498e-02,  8.1775e-01, -4.7181e-02,  3.2932e-03,  4.9661e-03,
         7.9124e-03,  3.2625e-03,  2.6492e-02, -4.2614e-03,  1.0474e-02,
        -1.5124e-02, -2.3891e-02,  1.3746e-03,  1.4987e-02,  5.2963e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0087e-02,  6.6576e-01,  2.5217e-02, -4.4808e-02, -4.5224e-04,
         1.1191e-02,  8.4471e-03, -8.8168e-03,  1.9155e-03, -6.4478e-03,
         5.4168e-03,  1.1693e-03,  5.7827e-03,  2.1116e-02, -1.5878e-03,
        -2.1529e-04,  8.9187e-04,  1.9393e-03, -1.5144e-03, -1.8981e-02,
        -1.0741e-02, -4.0246e-03,  5.5543e-03, -3.6472e-04, -3.4895e-03,
         5.2890e-03,  1.5925e-03,  6.9051e-04,  4.3248e-04, -3.5805e-03,
        -7.0910e-03, -3.3654e-03, -8.4117e-03,  1.4299e-02,  1.0988e-02,
         6.6534e-03, -2.2198e-03, -2.8921e-03,  1.5209e-03, -2.5969e-03,
         4.4837e-03, -8.6165e-04,  3.7220e-04, -2.2839e-03, -8.6228e-03,
        -1.9396e-02, -2.6430e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.3106e-03, -7.4440e-01,  8.0312e-03,  1.5245e-03,  2.5106e-03,
         2.2419e-02,  1.6365e-02, -6.8693e-03,  4.4184e-03,  2.7692e-02,
         1.1191e-02,  1.3574e-02, -4.2643e-03,  2.9760e-03, -2.2801e-03,
        -1.0429e-03, -3.2490e-03, -3.9509e-03, -8.5899e-03, -2.3743e-03,
        -3.2713e-04,  1.4532e-03, -5.5528e-03, -2.5962e-03, -2.4286e-03,
        -2.9198e-03, -1.6250e-02, -2.6885e-03, -1.8354e-03, -1.6823e-03,
        -6.8311e-03,  4.5572e-03,  4.0481e-03, -4.3392e-03, -5.0779e-03,
        -4.5144e-03, -2.2004e-04,  3.5864e-04,  1.6372e-03,  4.5132e-03,
         1.9991e-03,  2.8132e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6764e-02,  7.0635e-01,  3.1845e-02, -3.7205e-02,  2.0889e-03,
         1.9923e-03,  2.6281e-02,  5.7959e-03,  7.0771e-03, -1.0776e-03,
        -4.8563e-03, -1.3393e-02,  3.1081e-03,  1.4051e-04,  7.9662e-03,
        -1.1831e-02,  4.0133e-04,  3.5195e-03,  5.6926e-03,  2.6527e-03,
        -1.5125e-02,  5.3793e-04, -3.6122e-03,  1.1187e-02, -4.9505e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0837e-02,  6.0232e-01,  1.4302e-02, -9.6683e-04,  3.4919e-03,
         1.9567e-03, -4.7586e-03, -3.4253e-04,  1.4994e-02,  3.0962e-03,
         8.5461e-03, -2.6896e-03,  9.3297e-03, -1.4112e-02, -2.8031e-03,
        -9.4755e-03, -8.3413e-03,  1.0762e-02, -8.1949e-03, -1.0291e-01,
         2.6022e-02,  2.5495e-02, -4.4249e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0443,  0.6802,  0.0321, -0.0110,  0.0258, -0.0078, -0.0168, -0.0109,
        -0.0082,  0.0262,  0.0311,  0.0019, -0.0080,  0.0024, -0.0194, -0.0014,
        -0.0063,  0.0285, -0.0040, -0.0121,  0.0057, -0.0108, -0.0052,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1398, -0.6528,  0.0190,  0.0049,  0.0022, -0.0134, -0.0130, -0.0021,
        -0.0172,  0.0176,  0.0362,  0.0169, -0.0649,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0688,  0.7092,  0.0200, -0.0201, -0.0106, -0.0183, -0.0013, -0.0355,
         0.0062,  0.0083,  0.0048, -0.0150, -0.0488,  0.0331,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0388,  0.7748,  0.0073, -0.0035,  0.0246, -0.0114, -0.0028, -0.0044,
        -0.0026, -0.0155, -0.0215, -0.0067,  0.0060, -0.0063,  0.0138, -0.0067,
        -0.0287, -0.0024, -0.0221,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 0.0217,  0.7302, -0.0663, -0.0166, -0.0039, -0.0084,  0.0014, -0.0027,
        -0.0019,  0.0057,  0.0024, -0.0012, -0.0040, -0.0061, -0.0025, -0.0112,
        -0.0033, -0.0047, -0.0043,  0.0228, -0.0033,  0.0021, -0.0038, -0.0038,
         0.0040, -0.0615,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0304, -0.6961, -0.0499,  0.0209,  0.0114,  0.0349, -0.0030, -0.0091,
        -0.0076,  0.0463, -0.0008,  0.0096, -0.0289, -0.0082,  0.0254, -0.0174,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2982e-02,  7.0517e-01,  2.1268e-02,  1.3408e-02,  8.9152e-03,
         2.4026e-02,  1.9562e-02,  1.2710e-02,  2.1877e-02,  1.7180e-02,
        -1.5033e-02, -2.6164e-02, -1.8116e-04, -2.1463e-03,  3.5336e-02,
         1.3686e-02,  2.0354e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0431,  0.7267,  0.0848, -0.0641, -0.0089, -0.0191,  0.0219, -0.0117,
         0.0199,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.4242e-02, -6.0613e-01, -3.3569e-02, -2.2350e-02,  5.6084e-03,
        -4.3939e-04,  2.3242e-04, -7.4957e-03, -1.1439e-02, -4.4863e-03,
        -1.1738e-02,  7.1554e-03,  2.0561e-03,  1.0090e-02, -8.8622e-03,
        -1.7202e-02,  1.1330e-02, -3.4926e-03,  2.0122e-04, -3.2110e-03,
        -8.9097e-04, -4.8870e-03, -8.8184e-03,  3.1907e-03, -1.6613e-04,
         9.5029e-03, -5.0235e-03,  2.6688e-03, -1.5801e-03,  3.8220e-03,
         4.0274e-03, -5.7733e-04,  6.2612e-04,  2.5964e-03, -4.0344e-03,
         1.3137e-02,  5.8649e-03, -1.3349e-03,  7.3316e-03,  4.7997e-03,
         1.0276e-02,  6.0566e-03,  5.4739e-03,  5.7076e-03, -6.4225e-04,
        -2.5629e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.1652e-02,  6.4019e-01, -5.8231e-02,  3.2535e-03,  7.8576e-03,
        -1.0722e-02, -3.7316e-02, -4.2881e-03,  2.1182e-03,  4.0384e-03,
        -5.2492e-03, -4.6340e-03, -3.7884e-03,  1.7642e-03,  8.8506e-03,
         1.1501e-03,  3.2853e-03,  3.6703e-04, -4.5074e-04,  9.7994e-03,
        -1.1213e-02, -2.8341e-03, -2.2420e-02, -1.4355e-02,  1.8200e-03,
        -4.7297e-03,  6.6281e-03,  1.9554e-02, -3.4985e-03, -1.3939e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0048, -0.8508, -0.0227, -0.0281, -0.0318, -0.0165,  0.0453,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1691e-02,  8.2472e-01, -4.0232e-02,  7.4713e-03, -1.9606e-04,
         1.2592e-02, -1.5638e-03,  4.0047e-02, -4.3175e-03,  1.4637e-03,
        -5.5701e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0837, -0.6798, -0.0169, -0.0181, -0.0273,  0.0008, -0.0432,  0.0178,
        -0.0125, -0.0111, -0.0020,  0.0166,  0.0254,  0.0448,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9586e-02, -6.8826e-01, -4.0753e-02, -1.5788e-02, -1.8732e-02,
         2.2345e-03, -4.6219e-03, -3.1299e-03,  1.2472e-02,  1.3619e-02,
        -5.5049e-03, -6.8695e-03, -2.6102e-03,  1.0206e-02, -1.6453e-03,
         1.2264e-02, -1.7345e-03, -6.8988e-03,  1.8046e-03, -1.4460e-02,
         6.0541e-03, -9.7501e-03,  4.7730e-03, -1.3341e-03,  9.7109e-03,
         3.0067e-03,  2.5945e-04, -6.6374e-03, -8.4317e-05, -1.1713e-02,
         2.2902e-02,  6.8975e-03,  9.5259e-03, -4.1580e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0085, -0.6236,  0.0295,  0.0198, -0.0082, -0.0145, -0.0164, -0.0094,
        -0.0060, -0.0057,  0.0020,  0.0191,  0.0110, -0.0030, -0.0042,  0.0069,
         0.0024,  0.0328,  0.0046, -0.0055,  0.0274,  0.0092,  0.0299, -0.0056,
        -0.0140, -0.0097,  0.0285,  0.0426,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0168, -0.7244, -0.0251,  0.0338,  0.0078,  0.0059,  0.0075,  0.0105,
         0.0096, -0.0018,  0.0211,  0.0239,  0.0221,  0.0169, -0.0009,  0.0083,
        -0.0061, -0.0060, -0.0100,  0.0041,  0.0044,  0.0069, -0.0119, -0.0145,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([-1.1167e-02, -7.2468e-01, -1.9514e-02, -9.1728e-03, -8.1841e-03,
         1.4013e-02,  5.7508e-03, -1.0931e-04, -2.2933e-03, -8.7927e-03,
         3.1507e-02,  7.9774e-03, -1.8825e-02, -5.0056e-04, -8.5776e-03,
         5.0161e-03,  3.6013e-03,  1.1494e-02,  1.3976e-03,  7.2157e-03,
         1.4345e-05, -1.5806e-03,  1.1165e-02,  3.8700e-03, -1.3446e-03,
         1.8647e-02,  4.7316e-03, -2.2231e-03,  2.5047e-03,  2.4139e-02,
         2.9991e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.7478e-02,  7.7299e-01,  7.9073e-03, -1.3388e-02,  2.4950e-02,
         3.4379e-04,  5.8389e-03, -1.4277e-02,  4.8771e-02, -6.9135e-03,
         1.6038e-02,  3.1100e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0301, -0.6838, -0.0115,  0.0343,  0.0103, -0.0232, -0.0184, -0.0032,
        -0.0176, -0.0171,  0.0452, -0.0188, -0.0512, -0.0203, -0.0150,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6273e-03, -6.7276e-01, -7.2745e-02, -1.7619e-02,  2.2240e-02,
        -1.3611e-02, -1.0145e-03, -4.0947e-03, -1.4002e-02,  1.2524e-02,
        -3.7541e-03,  1.1817e-02,  3.7380e-03,  7.8930e-03,  5.5269e-03,
        -3.4028e-03,  1.4716e-02,  1.6061e-02,  4.5162e-03,  2.0097e-02,
         3.8652e-03,  2.5330e-03,  2.4213e-03,  1.1675e-03, -6.1156e-03,
         2.6180e-03,  3.4217e-04, -1.1801e-03,  2.1442e-03, -8.7910e-03,
         8.5720e-03,  7.8756e-03, -2.8613e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0102,  0.6770,  0.0564, -0.0519, -0.0088, -0.0008, -0.0225, -0.0261,
         0.0057, -0.0169, -0.0164,  0.0042, -0.0180, -0.0323,  0.0418, -0.0110,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1529, -0.6399,  0.0696,  0.0725,  0.0141, -0.0240, -0.0162,  0.0030,
        -0.0077,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0590,  0.7151, -0.0975,  0.0574,  0.0710,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3059e-04, -7.8450e-01,  5.7248e-03, -5.1232e-03,  4.1655e-03,
        -3.7047e-03,  3.6849e-03, -1.0331e-02, -9.3774e-03,  2.2691e-02,
         5.7132e-03, -1.3764e-02,  3.1744e-02, -1.5717e-02,  1.7585e-02,
        -1.4301e-03,  1.8569e-03,  4.1213e-03, -7.7096e-03,  1.4767e-03,
        -7.1753e-03, -1.4656e-03, -3.4359e-03,  5.6281e-03,  8.7092e-05,
        -3.1428e-02, -1.3066e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.7448e-02,  8.2409e-01,  2.2323e-02, -1.8030e-02, -1.9235e-02,
        -5.8740e-02,  1.3600e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0648, -0.6444, -0.0165, -0.0204, -0.0094, -0.0108, -0.0206, -0.0080,
         0.0163,  0.0479,  0.0073, -0.0049, -0.0038,  0.0125,  0.0247,  0.0152,
        -0.0053,  0.0058,  0.0262,  0.0024,  0.0130,  0.0198,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0788, -0.6996, -0.0447,  0.0113,  0.0596,  0.0048,  0.0086,  0.0020,
         0.0050, -0.0030, -0.0058,  0.0103,  0.0029, -0.0162,  0.0054, -0.0082,
         0.0117,  0.0021,  0.0201,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0911e-04, -7.9852e-01,  1.1216e-02, -3.5188e-03,  3.4865e-03,
        -3.1054e-02,  5.4062e-04, -1.2020e-02,  4.0548e-03, -6.6754e-03,
         7.7302e-03,  9.9218e-04, -1.5914e-03,  3.3338e-03, -2.6679e-03,
        -2.9639e-03,  2.0432e-02,  5.5385e-04,  5.5641e-03, -6.8953e-03,
        -1.0540e-02, -1.3502e-02,  3.2621e-03, -3.5172e-03, -2.7834e-03,
        -7.2318e-03, -1.1797e-03, -3.3651e-03, -1.8270e-02, -8.4408e-03,
         3.2923e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 0.0988, -0.4653, -0.1269,  0.0199,  0.0073,  0.0079, -0.0112,  0.0027,
        -0.0022, -0.0034,  0.0044,  0.0063, -0.0191,  0.0154, -0.0095,  0.0022,
         0.0139, -0.0123,  0.0190, -0.0021, -0.0054, -0.0109,  0.0074, -0.0041,
         0.0164,  0.0038, -0.0039,  0.0409, -0.0012, -0.0105,  0.0089,  0.0371,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0250,  0.6420,  0.0291, -0.0093,  0.0067,  0.0054,  0.0401,  0.0338,
         0.0540,  0.0074,  0.0059,  0.0114, -0.0200, -0.0341, -0.0032, -0.0053,
         0.0014,  0.0097, -0.0210,  0.0044,  0.0028, -0.0278,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3329e-02,  6.2650e-01,  1.3573e-02, -1.4843e-02, -4.3703e-02,
         2.2524e-03, -4.9251e-02,  7.4624e-03, -3.6603e-03,  2.4078e-02,
         3.5795e-03, -6.0859e-03,  1.3504e-02, -4.1865e-03,  2.2708e-02,
         3.9731e-03, -1.2877e-03,  7.0010e-03,  4.4681e-03, -7.6325e-03,
         3.2959e-03,  2.3133e-04,  3.8711e-03,  5.0529e-03, -1.2210e-02,
         9.4312e-03, -3.4572e-03,  2.8651e-02,  2.9815e-04, -4.0420e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1529e-01,  7.0267e-01,  2.4507e-02, -2.6669e-02,  3.9735e-03,
        -6.8492e-03,  1.8091e-02,  3.7822e-03, -4.7185e-03, -1.5953e-02,
        -1.5302e-02, -8.1468e-03, -6.8940e-03,  5.0177e-03,  2.1252e-03,
         1.3599e-02,  8.2383e-05,  2.6329e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8670e-02, -7.5927e-01, -9.3497e-03, -4.1563e-03,  4.4265e-02,
         9.2422e-03, -9.0716e-03, -8.1609e-04,  1.4231e-02,  9.1645e-03,
         2.9302e-03,  3.3184e-03,  3.8293e-03,  1.2560e-02, -3.3391e-03,
         1.3491e-02,  3.3972e-03,  1.2899e-03,  6.5780e-03,  8.3384e-03,
         5.8166e-03,  2.2169e-03, -4.0455e-03, -2.3097e-03,  1.5158e-03,
         4.8975e-04,  1.8899e-02,  9.3491e-04,  1.8871e-03,  5.8598e-04,
         2.7600e-03, -3.6776e-03,  1.7997e-04,  1.2023e-03,  1.5938e-04,
         9.0247e-03,  6.9895e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9628e-02, -6.0987e-01,  5.8634e-02,  5.0051e-02, -5.5019e-03,
         5.4394e-03,  5.6489e-03,  2.7936e-04,  1.0322e-02, -7.5500e-03,
        -1.1588e-02,  6.0138e-02,  1.4796e-02, -2.8069e-02,  5.0498e-03,
         9.2233e-03,  2.6279e-03, -8.3102e-03,  3.6200e-03,  5.0648e-03,
         5.0768e-03, -7.8890e-03, -2.7219e-02, -2.8402e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.3801e-03,  6.5966e-01, -1.2976e-02, -1.3127e-02,  1.2758e-02,
        -3.2848e-03, -1.1148e-02,  1.9519e-02,  1.4684e-02,  1.4436e-03,
        -2.4179e-03,  9.8453e-04, -6.5465e-03,  9.6364e-03, -8.3103e-03,
         1.1074e-02, -4.5145e-04,  1.1479e-03,  1.0476e-03, -2.8815e-04,
         6.7165e-04, -3.2514e-03, -7.0115e-04, -4.6053e-03,  2.0323e-02,
         3.0697e-02,  3.0088e-03,  8.5654e-03, -1.3521e-03,  7.8102e-03,
         2.1208e-02,  1.8406e-02,  5.7471e-03,  9.9699e-03,  8.5676e-03,
        -7.7363e-03, -2.8667e-03,  4.2718e-03, -1.3141e-03,  3.1724e-03,
         5.4389e-04,  4.9751e-03,  1.6546e-03,  1.5520e-02,  3.9462e-03,
         1.1223e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.7349e-03,  7.4878e-01,  2.2285e-02,  9.1648e-04, -1.5762e-02,
        -3.2294e-03,  1.4648e-02, -7.8188e-04, -1.8183e-03, -1.1192e-02,
         4.5497e-03, -1.1738e-02, -3.8842e-05, -4.1868e-03,  4.3334e-03,
         6.2582e-04, -4.6361e-04, -2.9774e-02, -1.4589e-02,  1.1339e-02,
         2.3234e-02,  5.3612e-02,  1.3369e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9172e-02, -7.9565e-01,  1.7987e-02, -9.2631e-03, -1.2743e-02,
         6.7938e-04, -2.2835e-03,  6.1402e-02,  1.1043e-02,  9.8878e-03,
         2.9885e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0682, -0.7292, -0.0258,  0.0207, -0.0010, -0.0247, -0.0145,  0.0200,
         0.0040, -0.0116, -0.0084,  0.0048, -0.0474,  0.0198,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0655, -0.7461, -0.0286,  0.0259,  0.0319,  0.0045,  0.0125,  0.0287,
        -0.0030,  0.0214, -0.0266,  0.0053,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6812e-03,  6.7400e-01,  5.5026e-02, -1.8306e-02, -2.5438e-03,
        -2.7986e-03, -1.7170e-02,  3.9447e-03, -3.4166e-04,  3.6279e-04,
        -7.9463e-04, -7.9049e-03,  4.6291e-04, -8.9655e-03, -5.2750e-03,
        -9.4022e-04,  6.0761e-04, -4.7353e-03, -3.0910e-03, -1.6185e-02,
         2.8670e-03, -2.6843e-03, -4.7126e-03,  1.2730e-03,  9.4839e-03,
         2.1465e-02,  2.4822e-02, -9.2183e-04, -3.6246e-04, -2.9966e-03,
         4.8754e-03, -1.9193e-03,  8.1108e-03,  8.3111e-03,  7.4801e-03,
        -9.7504e-03, -8.6375e-04,  3.6062e-04,  5.7987e-03,  1.1466e-03,
        -1.7794e-03,  1.5418e-03,  4.4995e-03,  3.1174e-03, -3.3019e-03,
         3.8799e-03, -7.1881e-05,  2.3566e-04,  2.4076e-03, -4.4879e-03,
        -2.8514e-04, -9.2611e-04, -2.7318e-03,  8.6663e-03,  1.3723e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-7.7196e-02, -7.2986e-01, -2.4620e-02, -2.5334e-03,  1.7992e-02,
        -3.0099e-03, -4.7656e-03,  8.0231e-03, -2.1580e-03, -1.7370e-05,
         1.0612e-02, -5.0168e-03, -7.0869e-03, -1.6910e-02, -6.5644e-03,
        -2.3840e-03,  2.0470e-02, -2.1599e-04,  1.1064e-03, -5.7458e-03,
         2.3086e-03, -1.5225e-02,  3.6180e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0748,  0.6021,  0.0668,  0.0088,  0.0174,  0.0012, -0.0128,  0.0258,
        -0.0337,  0.0198, -0.0199, -0.0121, -0.0115, -0.0172, -0.0011, -0.0111,
        -0.0351,  0.0288,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0459,  0.6500, -0.0145, -0.0106,  0.0083,  0.0487,  0.0525, -0.0011,
         0.0144,  0.0068,  0.0303, -0.0020,  0.0141, -0.0111,  0.0111,  0.0271,
         0.0110,  0.0406,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0124,  0.6907,  0.0589, -0.0235,  0.0118, -0.0205, -0.0096, -0.0102,
        -0.0108, -0.0066, -0.0013, -0.0213,  0.0251, -0.0171, -0.0802,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0939, -0.4957, -0.0094,  0.0240,  0.0236, -0.0292, -0.0172,  0.0082,
         0.0175, -0.0086, -0.0093, -0.0087, -0.0095, -0.0013, -0.0152,  0.0042,
         0.0090,  0.0008,  0.0030,  0.0220,  0.0037,  0.0196,  0.0063,  0.0093,
         0.0029,  0.0067,  0.0086, -0.0030,  0.0270, -0.0326, -0.0425, -0.0279,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.2742e-02, -5.1242e-01,  6.9710e-02,  4.1885e-02, -1.2599e-02,
         5.3509e-03, -1.3902e-02, -7.8003e-03, -1.9268e-02,  2.2492e-04,
         4.0260e-03, -1.5518e-02,  6.3447e-03, -5.2529e-03, -2.2233e-03,
        -1.2826e-03,  2.2941e-03,  1.4209e-02, -1.8709e-02, -3.3562e-03,
        -6.9664e-03,  1.1982e-02, -7.8596e-03,  1.6299e-02, -4.2671e-04,
        -1.5781e-02,  6.1336e-03, -8.5124e-03, -6.9913e-03,  7.2883e-03,
        -2.0821e-02, -3.6965e-02, -4.8601e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7302e-03, -6.4875e-01, -5.5307e-02, -3.9043e-02, -2.3068e-02,
        -1.1782e-02, -5.2209e-03,  5.0025e-03,  6.6806e-03,  3.7778e-03,
         2.9014e-03,  1.0791e-02,  1.1317e-02, -7.0964e-04,  1.8871e-02,
        -4.6545e-04, -9.5032e-03,  1.6555e-03,  3.0763e-03, -1.0779e-03,
         5.3200e-03,  1.1555e-02, -4.2937e-04,  1.2736e-02, -1.6459e-03,
        -1.2232e-03, -7.4214e-03, -1.6555e-03, -1.5141e-03,  5.7994e-03,
        -4.3452e-04,  2.4926e-03, -3.1721e-03, -1.1933e-03, -2.1975e-03,
         5.0756e-03, -6.3504e-04, -9.7099e-04, -8.2041e-04,  2.8274e-03,
        -2.1120e-03,  1.5547e-03,  3.1236e-03, -5.7133e-03,  1.1869e-03,
        -4.2905e-03,  4.7516e-03, -2.3799e-03, -2.0708e-03, -5.5442e-03,
         6.9245e-03, -2.7881e-03,  9.5237e-04, -1.4282e-03, -5.2739e-03,
         3.7893e-03, -6.6029e-03, -7.6687e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0380, -0.7203, -0.0540, -0.0253,  0.0027,  0.0146, -0.0019, -0.0020,
         0.0059, -0.0038, -0.0086,  0.0029,  0.0267,  0.0187,  0.0014,  0.0081,
         0.0015,  0.0072,  0.0070,  0.0054,  0.0013,  0.0102,  0.0051,  0.0043,
         0.0037,  0.0197,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0757e-01, -6.0247e-01,  2.9946e-02, -6.6486e-03,  1.2019e-02,
         2.8726e-03,  5.3403e-02,  1.0688e-02,  4.3950e-02, -5.9621e-04,
         7.3373e-03, -3.4173e-03,  6.6177e-03,  1.8286e-03,  7.0184e-03,
         1.9327e-03, -2.6540e-03,  1.1105e-02, -6.8661e-03, -3.0793e-03,
         7.2325e-03,  5.4155e-04,  1.0818e-02, -3.1045e-02,  2.8342e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7430e-02,  6.5445e-01,  6.3184e-03, -1.3788e-02, -1.2621e-03,
        -1.9659e-02, -9.9336e-03, -1.3869e-02, -1.5296e-02, -5.0935e-03,
        -1.5626e-02, -2.6031e-02,  1.9549e-02,  1.0151e-02, -3.0563e-04,
        -5.3223e-03, -2.6002e-02,  1.2614e-03,  2.5926e-04, -1.0969e-02,
        -3.9652e-03,  9.0381e-03, -7.5427e-03, -1.7750e-02,  1.0983e-02,
        -2.3815e-03, -6.9041e-03, -1.9111e-03, -7.1450e-03, -3.3987e-03,
         3.1749e-03,  4.9091e-04,  1.3032e-03, -2.8623e-03,  2.6891e-04,
        -1.8512e-03, -9.2407e-03,  3.1687e-03,  3.1321e-04,  2.7852e-03,
        -6.8766e-04, -6.2574e-04, -4.5175e-03, -9.3660e-03, -1.5752e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0830,  0.6631,  0.0543,  0.0082, -0.0050,  0.0259, -0.0031,  0.0166,
         0.0046,  0.0092,  0.0324,  0.0113, -0.0025, -0.0060,  0.0109, -0.0145,
        -0.0126,  0.0050, -0.0026, -0.0043, -0.0088,  0.0120,  0.0041,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2254e-01,  6.4568e-01, -3.9300e-02, -2.2807e-02, -9.6842e-03,
         3.0748e-02,  5.0474e-02, -5.5255e-04, -2.3998e-02,  9.8762e-04,
         2.0780e-03, -5.1148e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 8.9064e-03, -7.3277e-01, -5.5175e-02,  1.9056e-02,  1.3429e-02,
         7.6444e-03, -4.8215e-03, -1.1790e-02, -1.4631e-02,  7.8135e-03,
        -2.1379e-02, -1.7184e-02, -3.6188e-03, -1.2619e-02,  6.0593e-03,
         4.1796e-04,  8.3664e-03, -4.1256e-03,  4.8973e-03,  1.4220e-03,
         1.2170e-02,  8.2917e-03,  2.3413e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5104e-02, -6.2176e-01,  9.8831e-03,  1.2543e-02, -1.5250e-02,
         8.3405e-03, -8.3388e-03,  8.4347e-03, -4.0771e-03, -8.1064e-03,
        -5.6698e-03, -1.5571e-03,  7.1381e-03, -8.8540e-03, -2.2167e-02,
         1.3244e-02,  5.6151e-03,  1.5888e-02, -8.6953e-03, -3.7531e-03,
        -3.6745e-03,  1.2882e-03,  2.8318e-02, -5.9186e-03,  1.9400e-03,
         8.3577e-03, -2.9833e-03,  2.2383e-03,  1.7645e-03,  3.0319e-04,
         4.1883e-03, -3.6055e-02,  5.8553e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3083e-02, -7.2153e-01,  6.1821e-02,  3.7609e-02,  3.5027e-02,
        -4.7042e-03,  2.6403e-02,  9.6494e-03,  8.7516e-03,  9.4400e-05,
         1.0437e-03, -9.8336e-03,  1.1575e-02, -8.7458e-04, -8.6074e-03,
         3.5128e-03,  1.7624e-03, -1.6629e-02,  7.4861e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0068, -0.7359,  0.0856, -0.0237, -0.0191, -0.0396, -0.0058, -0.0023,
        -0.0089, -0.0105, -0.0011, -0.0313, -0.0188, -0.0105,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1279e-02, -7.1034e-01,  3.2950e-02,  6.1493e-04, -3.3329e-02,
        -1.7320e-02, -4.6427e-03,  1.6718e-03, -2.0666e-02, -4.7020e-03,
        -7.5871e-03, -3.8293e-03, -2.3699e-03,  4.9779e-02,  7.1786e-03,
        -2.1078e-03,  2.2172e-02, -1.0715e-03,  1.0960e-02,  1.6433e-02,
        -8.6572e-03,  2.0337e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.6415e-03,  7.4537e-01, -1.9897e-02,  1.5737e-02,  8.3027e-03,
         6.1945e-03,  3.8050e-04, -3.0973e-02,  4.6131e-05, -1.0181e-02,
        -1.8136e-02,  9.9161e-03,  1.3260e-03,  6.3644e-04,  6.0275e-04,
         4.0191e-03, -1.2481e-02,  1.6777e-02,  6.5298e-03,  1.2279e-02,
         1.6394e-03,  1.1227e-02, -2.2208e-03,  5.8152e-03, -8.8864e-03,
        -6.2210e-03,  1.2867e-03,  4.0314e-03,  1.8700e-02,  1.0542e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3916e-02,  5.6731e-01,  8.5691e-03, -3.7500e-02, -2.2656e-02,
         5.1153e-02,  4.8390e-02,  1.0677e-02,  6.9898e-04, -1.0098e-02,
        -2.6867e-02, -2.1950e-02, -1.8923e-02, -9.6832e-03, -4.7050e-04,
        -2.5688e-02, -8.5445e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0150, -0.7293, -0.0416,  0.0675,  0.0233, -0.0034,  0.0012, -0.0280,
         0.0490, -0.0136,  0.0281,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0084e-01,  5.7734e-01,  7.8215e-02,  1.8156e-02,  3.5055e-02,
        -3.4065e-02, -1.7655e-02, -1.2089e-02, -1.4590e-02, -2.9985e-05,
         4.9730e-03,  1.4396e-02, -2.7939e-02, -3.7167e-03, -3.2288e-02,
        -5.5891e-03,  2.3067e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.7663e-05,  7.2007e-01, -2.7791e-02,  8.5242e-03,  2.5210e-03,
        -1.6694e-02, -1.6610e-02,  9.0713e-03, -1.2678e-02, -1.0177e-02,
        -8.5339e-03,  1.3998e-02, -2.8858e-03,  2.3300e-03, -1.7959e-02,
         1.2062e-02, -2.1467e-03, -2.9056e-03, -7.6422e-03, -2.1515e-03,
        -4.4830e-03, -1.1993e-02,  1.4046e-02, -3.1911e-03,  1.8416e-04,
        -6.9267e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9303e-02,  6.1246e-01,  6.9058e-02, -1.5773e-02,  8.3369e-03,
         2.8777e-02, -2.1353e-02,  3.2089e-02,  2.3912e-02, -1.6777e-03,
        -1.0722e-02, -1.6266e-02, -2.1116e-03,  6.5113e-03,  3.4931e-04,
        -9.2237e-03, -1.0197e-03,  3.6460e-03, -1.1725e-03, -7.9001e-03,
        -9.0798e-03, -7.2819e-03, -4.1758e-03,  1.0502e-03,  5.1739e-03,
         6.5198e-04, -3.3355e-03,  8.3079e-03, -1.8172e-02,  1.1111e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0997,  0.6639,  0.0116, -0.0301, -0.0047, -0.0156,  0.0360, -0.0066,
         0.0121, -0.0178,  0.0071,  0.0149,  0.0122,  0.0191,  0.0034, -0.0049,
        -0.0180, -0.0145, -0.0076,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([-0.0104, -0.6678, -0.0245,  0.0232,  0.0323,  0.0089,  0.0011,  0.0187,
        -0.0107,  0.0037,  0.0038,  0.0060, -0.0026, -0.0024,  0.0093,  0.0022,
        -0.0128,  0.0300, -0.0072, -0.0064, -0.0012,  0.0123, -0.0023,  0.0126,
         0.0094,  0.0600, -0.0183,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1118, -0.6372, -0.0016,  0.0317,  0.0239, -0.0218, -0.0152, -0.0008,
         0.0250,  0.0044, -0.0244, -0.0303, -0.0339, -0.0381,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0191e-01,  6.9721e-01, -1.3581e-02,  4.4987e-04, -9.5345e-03,
        -1.7277e-02,  1.2230e-02,  4.2416e-03, -1.0292e-02, -2.1321e-03,
        -1.2675e-02, -1.7807e-02, -1.7275e-02,  3.9899e-03, -1.0183e-03,
         3.1467e-03,  3.8989e-03, -1.6736e-05,  1.6247e-02, -6.8091e-04,
        -1.4659e-02, -1.0023e-02, -4.7959e-03,  3.5201e-03,  1.6743e-02,
        -4.6399e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4428e-02, -7.5321e-01, -1.4132e-04, -1.2862e-02,  2.6982e-03,
         2.5432e-02, -5.8659e-03, -1.5752e-03,  2.1387e-02,  3.3468e-02,
         3.7280e-02,  3.6204e-02, -3.5133e-02, -3.1687e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.0633e-02,  6.8087e-01, -1.0676e-03, -7.4678e-03,  1.8398e-02,
        -8.4113e-03,  4.5510e-03,  7.0899e-03,  1.6134e-02,  1.4946e-02,
         4.9333e-04,  1.6981e-02, -7.8313e-03,  1.7152e-02,  7.0543e-03,
         1.4978e-03,  7.2944e-03,  1.3262e-02, -3.1265e-03, -8.5936e-03,
         4.2765e-03,  1.1029e-03, -4.3896e-03,  4.7876e-03,  1.8136e-03,
        -1.6364e-03, -1.1371e-02,  2.3377e-03,  1.5145e-02,  5.7094e-03,
         3.0692e-02,  3.8801e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3771e-02, -7.5944e-01,  3.4246e-03,  6.2716e-02,  5.1818e-03,
         1.3021e-02,  6.4086e-03,  1.6644e-02, -1.7553e-02,  1.6438e-02,
        -6.2602e-04,  5.5144e-04,  2.4249e-04, -1.1201e-02, -4.7645e-03,
         4.9493e-03, -1.7551e-02, -6.5482e-03, -1.2630e-02, -4.2190e-03,
         1.7205e-03,  1.6295e-03,  9.5080e-03, -7.1269e-03,  2.1355e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0334, -0.7454, -0.0355,  0.0776,  0.0464, -0.0618,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.6382e-03,  7.3188e-01,  2.5671e-02,  2.5224e-03,  1.8764e-02,
         2.3907e-02,  9.2150e-03,  4.0590e-03,  1.7259e-03,  5.2092e-03,
        -1.7967e-03,  1.4316e-02,  3.5338e-03, -1.3301e-03,  6.0863e-03,
        -3.1652e-03, -9.1120e-03,  1.0752e-02, -6.4408e-03,  1.5986e-04,
        -8.5126e-03, -1.2193e-03,  1.4848e-02, -6.1060e-03,  1.5089e-03,
        -3.7246e-03,  2.4940e-03,  4.6395e-03, -2.1997e-03, -1.5754e-03,
         4.6020e-03, -9.6195e-04, -1.2859e-03,  1.7735e-03,  4.1288e-03,
        -3.7361e-02,  1.5778e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3943e-02, -5.9598e-01, -2.2269e-02, -4.9986e-02,  2.8600e-02,
        -3.6023e-03, -1.9205e-03,  1.1054e-02, -1.1307e-03,  1.8265e-03,
         5.0892e-03, -5.4946e-03, -4.0528e-03,  5.1362e-03, -2.0256e-03,
        -6.9799e-03,  1.6480e-02,  2.5813e-03, -5.0612e-04, -5.8341e-03,
         2.6191e-02,  2.1485e-02,  4.8232e-02,  2.8116e-03,  1.2575e-02,
        -6.5974e-03, -6.5896e-03, -1.1105e-02,  9.1395e-03,  4.3195e-03,
        -9.7773e-04,  5.1149e-04, -1.2811e-02, -4.2348e-03,  7.9284e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0063,  0.6204, -0.0852,  0.0516, -0.0041,  0.0106, -0.0095,  0.0034,
        -0.0069, -0.0236, -0.0337, -0.0027,  0.0232,  0.0683,  0.0504,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1021e-02, -6.8461e-01, -5.1476e-02,  5.5483e-02,  3.6254e-03,
        -2.0061e-02,  1.4783e-02, -2.4125e-02, -3.2629e-03, -5.7535e-03,
        -2.7299e-03,  1.2770e-02,  1.1804e-02,  1.8772e-03, -1.0599e-03,
         3.8849e-03,  4.2052e-03,  2.5132e-03,  9.1051e-03,  1.0659e-02,
        -2.8688e-04,  1.0789e-02,  3.8008e-03,  8.3573e-04, -4.1471e-03,
         9.6847e-03, -3.0430e-03,  9.9757e-04,  3.9168e-03,  5.3526e-03,
         1.0405e-03, -2.9325e-03, -8.3604e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8717e-03, -7.9484e-01,  1.3163e-02, -3.0808e-03, -1.0786e-02,
         2.3597e-03,  1.0580e-02,  4.2331e-03,  1.0397e-02,  6.2940e-03,
         5.3351e-03, -6.4080e-04,  9.6559e-03,  1.0826e-02,  1.3835e-02,
         1.3377e-03,  4.8830e-03,  7.4782e-04, -1.8369e-02, -2.4624e-03,
         5.4732e-03, -2.0385e-03, -1.7065e-03,  1.0922e-04, -1.9308e-03,
         6.6533e-03,  1.5406e-02, -1.9764e-02,  2.1218e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 0.0868,  0.6921, -0.0437, -0.0073, -0.0177,  0.0065,  0.0507, -0.0045,
        -0.0121, -0.0168, -0.0047,  0.0052,  0.0078,  0.0077, -0.0366,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6905e-02,  7.4325e-01,  9.3259e-03, -2.3247e-02, -1.3890e-02,
        -3.9090e-03,  1.4378e-02, -2.4310e-03, -8.4422e-03, -5.0086e-03,
        -1.3610e-02, -1.0201e-02,  3.3380e-03, -1.0339e-02, -5.3297e-03,
        -4.7681e-03,  4.9145e-04, -5.7811e-04, -1.0353e-04, -3.2178e-03,
         1.9308e-03, -1.7634e-03, -5.4504e-03, -4.0850e-03, -6.1890e-03,
        -1.0343e-03, -1.2413e-02,  4.2384e-04, -1.2412e-03, -3.2497e-03,
         1.5431e-03, -9.9122e-04,  1.7228e-03, -3.9408e-04, -4.2152e-03,
         3.5081e-03,  1.0357e-03,  2.8653e-03, -4.5411e-03, -3.2894e-03,
         1.9476e-03,  1.0648e-03,  1.1837e-02, -4.0259e-03,  3.7592e-03,
         2.2991e-04,  4.9127e-03,  2.4328e-03, -1.5142e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8886e-03, -5.3296e-01,  9.6813e-02,  4.2747e-02,  4.4553e-03,
        -2.3456e-02, -6.8516e-03, -1.8618e-02,  3.7334e-03, -9.0684e-03,
         4.1130e-04, -7.7899e-04, -4.5103e-03,  1.8858e-03,  2.5460e-03,
         6.2372e-03,  3.0215e-04, -4.2732e-03, -8.7895e-03,  2.5472e-03,
         5.8444e-04, -9.5938e-03,  8.1943e-03,  2.7998e-03,  3.0100e-03,
         1.4811e-02, -5.6729e-04,  3.8498e-03, -2.9424e-04,  8.5793e-03,
         3.5094e-03,  1.2707e-02,  8.3400e-03, -6.6381e-02, -8.3902e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0036, -0.7469, -0.0618,  0.0014, -0.0216,  0.0104, -0.0056, -0.0075,
        -0.0026, -0.0154, -0.0120, -0.0013, -0.0036, -0.0082, -0.0183,  0.0139,
        -0.0044,  0.0064,  0.0548,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0643,  0.6493,  0.0458, -0.0096,  0.0230,  0.0186,  0.0183, -0.0008,
         0.0041, -0.0090,  0.0209, -0.0053,  0.0559, -0.0152, -0.0597,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9883e-02, -7.5120e-01,  4.2233e-03,  3.1455e-02,  2.0790e-02,
        -5.6389e-03,  1.0346e-02,  8.9284e-03,  1.1856e-02,  4.0636e-03,
         2.7679e-03, -1.9733e-03,  2.4999e-02,  1.0513e-02, -1.0999e-02,
        -3.5227e-03,  7.2465e-03, -6.9274e-03, -2.4584e-04, -8.5302e-03,
         4.1767e-02, -2.1295e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0056, -0.6934,  0.0178, -0.0182, -0.0063, -0.0066,  0.0419,  0.0225,
         0.0382,  0.0019,  0.0167, -0.0117,  0.0011,  0.0050,  0.0092,  0.0074,
         0.0118, -0.0109,  0.0014,  0.0073,  0.0074,  0.0089,  0.0156, -0.0084,
         0.0022,  0.0066, -0.0026, -0.0017, -0.0009, -0.0109,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0485,  0.6941, -0.0801, -0.0051,  0.0252, -0.0097,  0.0072, -0.0148,
         0.0091,  0.0743,  0.0319,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3692e-02,  7.2892e-01, -7.3816e-02, -7.4653e-03,  1.4685e-02,
        -2.4638e-02,  6.7586e-03, -6.3645e-03, -9.7847e-03,  1.5380e-03,
        -3.3139e-02,  1.0946e-02, -6.6404e-03, -4.7547e-03, -1.6717e-03,
        -4.9292e-03, -3.8508e-03, -1.6842e-02, -2.2482e-03, -9.2433e-04,
        -1.5819e-02, -5.7680e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1804e-02, -6.4347e-01, -5.5937e-02,  1.1934e-02, -1.1603e-03,
        -1.4449e-02, -1.2525e-02, -6.7658e-03,  4.2340e-03, -3.9822e-04,
         4.3211e-02,  6.7309e-03,  4.9887e-03, -3.1803e-03,  7.7319e-03,
         1.5561e-02,  3.4027e-03, -1.2046e-02,  6.4745e-05,  2.3390e-03,
         3.8998e-03,  1.2696e-03,  2.0586e-02,  4.5639e-03, -1.5436e-04,
         3.0644e-03, -3.5598e-05,  2.0191e-03,  2.6944e-02,  1.2844e-03,
         2.6992e-03, -2.0663e-02, -2.0882e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1883e-02, -6.8774e-01, -8.1365e-02,  1.2751e-02,  1.5835e-02,
        -7.5671e-04, -1.1117e-02,  1.4905e-02,  1.2575e-03, -2.0705e-03,
         4.3205e-03, -6.6194e-03,  4.5307e-03,  1.6525e-02, -1.6373e-04,
        -3.7612e-02,  4.9126e-03, -2.8071e-03,  6.8400e-03, -4.0533e-03,
        -7.9250e-03,  8.0402e-03, -6.7341e-03,  1.8546e-04, -1.1626e-02,
         6.5504e-03,  1.0611e-02,  2.0265e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6750e-02, -7.3090e-01, -3.4403e-02, -1.8899e-02, -2.1033e-04,
        -6.7748e-03,  1.4568e-03,  5.0670e-03,  1.0040e-02, -1.3637e-02,
         1.9491e-02,  3.6579e-03,  6.5084e-03,  4.2278e-03, -1.5960e-03,
        -2.2055e-03,  1.3737e-02,  6.5148e-03,  1.1695e-03,  1.2714e-02,
        -3.9963e-03, -2.8365e-03,  9.7445e-03, -3.2039e-02,  2.1425e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-4.1707e-02,  6.5590e-01,  3.5446e-03, -3.0168e-03, -4.3566e-03,
         3.9528e-03, -1.1035e-03,  1.6853e-03, -8.3366e-03, -8.3075e-03,
        -5.9622e-03, -2.4795e-02, -3.4698e-03, -1.3293e-02, -9.3774e-03,
        -1.7778e-02, -6.1028e-03,  5.5875e-03,  5.3971e-03,  5.9100e-03,
        -8.7159e-03, -1.1082e-02,  6.2754e-04, -5.0898e-03, -1.7235e-02,
         6.0220e-05, -1.0210e-02, -4.8671e-03, -1.0251e-02, -1.9845e-02,
         6.6439e-03, -6.8961e-03, -3.3374e-04,  2.8196e-04,  2.4364e-03,
        -5.1048e-03, -8.5833e-03, -4.7603e-03, -1.1138e-02, -9.1846e-03,
         2.7070e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.3600e-02, -6.4532e-01,  5.8827e-02,  9.6318e-03, -4.2450e-03,
         3.6358e-02, -3.6750e-03, -1.7968e-02,  3.6679e-03, -9.1198e-06,
         6.0710e-03, -1.3803e-02,  1.3979e-02,  1.6316e-02, -7.7059e-03,
        -1.7208e-03,  1.2555e-02, -2.8769e-03,  3.6346e-04,  2.7331e-03,
         4.7863e-03,  2.8697e-02,  1.9097e-03,  2.2452e-02,  1.0733e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0507, -0.7175, -0.0603,  0.0125,  0.0156,  0.0028,  0.0041, -0.0223,
         0.0125,  0.0050, -0.0139,  0.0019,  0.0033,  0.0027,  0.0073,  0.0007,
         0.0062, -0.0105,  0.0187,  0.0236,  0.0079,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.1546e-03,  7.1797e-01, -2.4658e-02, -1.0053e-02, -1.6708e-02,
         7.0581e-03,  1.3227e-02, -7.3921e-03,  3.3223e-04,  8.8931e-03,
         1.6980e-03, -1.7640e-02,  1.4177e-02, -1.0833e-02, -1.1893e-02,
        -4.4864e-02, -2.7546e-02, -5.5908e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6740e-02, -6.7755e-01,  7.1770e-03,  1.1302e-02, -2.6500e-02,
         3.0178e-03,  5.1430e-03, -1.7076e-02,  5.7968e-03,  2.0624e-03,
        -1.4310e-03,  5.0201e-03, -2.7380e-04,  4.4861e-03, -2.1636e-03,
         1.9810e-02,  2.3677e-02, -2.7229e-05, -2.2129e-02, -2.6077e-02,
         2.9686e-03, -1.5485e-03,  1.7505e-02, -1.0966e-03, -3.6107e-03,
         1.1411e-03,  3.7196e-03,  5.5794e-03,  1.0187e-03,  8.0062e-03,
         1.0490e-03, -1.2168e-03,  1.0836e-02,  3.5492e-03,  1.7879e-02,
        -4.5494e-03,  9.5627e-03,  7.6994e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.8609e-02,  7.0466e-01, -5.3100e-02, -7.5652e-03,  1.9412e-02,
         5.5258e-03,  2.6132e-03,  4.6120e-03,  6.8604e-03, -7.6859e-03,
        -4.6480e-03, -5.4891e-03,  9.9061e-03,  1.4066e-02,  1.8614e-03,
         5.9204e-03, -1.3385e-05, -1.7697e-02, -5.4452e-03, -3.4306e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1017e-02, -7.3376e-01,  7.1387e-02, -5.1866e-04, -1.3878e-02,
        -1.0416e-02, -2.3940e-03, -1.4424e-02, -2.2908e-02,  1.2108e-02,
         1.0451e-02, -2.2988e-03,  3.0562e-02, -1.3882e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5238e-03, -6.6927e-01, -1.3024e-02, -4.8186e-03, -2.6387e-02,
        -1.9385e-02, -9.3238e-03, -5.5141e-03, -2.4038e-03,  1.4440e-03,
         1.2272e-02, -1.9990e-03, -3.2441e-03,  1.1127e-02, -2.3131e-04,
        -1.1778e-02,  1.2731e-02, -2.5313e-02, -8.6106e-03, -7.1684e-03,
        -2.9361e-03, -4.0429e-02,  6.4712e-03, -2.5592e-02,  9.1187e-03,
        -1.5125e-03, -6.5410e-03, -7.2544e-03, -1.8526e-03, -3.2659e-05,
        -7.4311e-03, -4.1262e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0512, -0.6590, -0.0467, -0.0408, -0.0287, -0.0089, -0.0090,  0.0051,
        -0.0033,  0.0187,  0.0037,  0.0127,  0.0026,  0.0099,  0.0057, -0.0012,
         0.0016,  0.0012,  0.0244,  0.0659,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.8394e-02, -6.0996e-01, -4.3183e-03, -7.8731e-03, -1.2207e-03,
         1.9471e-02, -3.9818e-03, -6.1079e-03,  6.5916e-03, -9.0740e-03,
         2.1593e-03,  5.1632e-03,  6.3185e-04, -8.4980e-03, -8.2323e-04,
        -1.2103e-02,  5.7924e-03, -3.3444e-04, -3.9233e-03, -1.1414e-03,
        -4.6919e-03, -1.9701e-02,  6.5994e-04,  1.6918e-02, -1.3160e-03,
         1.1183e-02,  1.9462e-02,  2.6053e-02,  2.0101e-02,  9.6730e-04,
        -7.2634e-03,  3.5383e-03, -9.3495e-03, -3.4914e-04,  2.4693e-03,
         4.0153e-04,  9.1929e-04, -3.8456e-03,  6.7421e-04, -5.2982e-03,
        -1.3042e-02, -1.6405e-03, -3.7797e-03, -3.4041e-03, -5.7263e-03,
        -2.9414e-02,  2.7023e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0173,  0.7858, -0.0026, -0.0239, -0.0130,  0.0082,  0.0154, -0.0023,
         0.0292,  0.0397, -0.0080,  0.0271, -0.0102,  0.0008,  0.0165,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0401e-02, -6.1918e-01, -4.6608e-02,  2.8561e-02, -8.8856e-03,
         1.7860e-03,  6.3134e-03, -1.8282e-02,  1.0991e-02, -1.1467e-02,
        -2.4114e-03, -8.3218e-03, -5.5218e-03,  1.9482e-02,  2.0348e-02,
         6.8475e-03, -2.1678e-03,  8.7091e-03,  5.4457e-03, -2.5692e-03,
        -2.5619e-03,  6.1413e-03, -3.2183e-03, -2.0749e-03,  9.7788e-03,
         6.0290e-03, -9.7081e-03,  4.5832e-03, -1.6821e-03, -2.3824e-03,
         4.8879e-03,  2.3570e-03, -1.3886e-03,  3.7718e-04, -1.9935e-03,
        -9.9388e-03,  1.2373e-03,  9.4711e-04,  6.8853e-03, -2.6681e-03,
         1.9472e-03,  1.8647e-03, -1.9527e-03, -2.5210e-03, -3.9576e-03,
         3.5222e-03,  9.1192e-03, -3.9975e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-3.0186e-02, -6.5846e-01, -4.2076e-02, -5.8623e-02,  4.3616e-03,
         1.8326e-02,  5.7808e-03,  3.4463e-02,  3.9935e-02,  6.7764e-04,
        -6.1744e-03, -2.4891e-03,  2.6760e-03, -1.2055e-04,  5.7654e-03,
         2.9545e-03, -3.6977e-03,  1.2086e-02,  1.7876e-03, -6.3613e-03,
         3.2962e-05,  1.6015e-02,  1.4516e-02,  2.5956e-03,  1.6266e-02,
         1.3577e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5323e-02, -7.0928e-01, -3.8473e-02,  1.8479e-02, -2.0961e-03,
         9.8322e-03, -4.4247e-03,  9.6329e-04,  6.0044e-02, -9.8745e-04,
        -8.1952e-03,  4.0229e-04, -8.5035e-04, -7.1060e-03,  9.0168e-03,
        -2.9457e-03,  8.7922e-03,  8.5072e-03,  2.7529e-02,  1.0749e-02,
         1.6009e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9034e-02,  6.1464e-01,  1.0105e-01,  3.6613e-02, -2.8378e-02,
        -7.5377e-03,  3.3870e-03,  3.7787e-03,  1.3750e-03,  3.1437e-04,
        -4.5764e-03,  1.5238e-04, -8.9396e-04,  1.9514e-03, -1.7676e-03,
         7.3349e-03,  1.3302e-02, -1.4245e-03, -1.4292e-02,  2.7318e-03,
         7.3151e-04, -1.7191e-03, -4.5603e-03,  5.6407e-03,  1.9011e-02,
        -1.2007e-02, -1.2716e-03,  5.6302e-04,  1.7474e-03,  7.5627e-03,
        -7.5590e-03, -6.0611e-03,  1.6100e-02, -2.2021e-03,  1.0266e-02,
        -1.8458e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0312,  0.6706, -0.0301, -0.0008,  0.0090,  0.0171,  0.0635, -0.0313,
         0.0015, -0.0043,  0.0193, -0.0015, -0.0093,  0.0056, -0.0024,  0.0033,
        -0.0021, -0.0134,  0.0090, -0.0749,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9449e-02, -6.5235e-01, -7.2581e-02, -1.1222e-02, -2.5070e-02,
         4.7219e-04,  1.3638e-02,  9.6812e-03, -7.3377e-03, -2.8516e-02,
        -1.1513e-02, -1.0263e-02,  5.4437e-03, -1.7845e-03, -2.3677e-02,
         1.4959e-02,  4.3523e-03,  4.8877e-03,  2.1230e-02,  9.0333e-04,
         5.0672e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0342,  0.6752,  0.0695, -0.0899, -0.0056,  0.0081, -0.0025, -0.0475,
         0.0150, -0.0036, -0.0046,  0.0112, -0.0332,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9935e-02,  7.0626e-01,  6.4308e-04, -1.2985e-02, -1.0854e-02,
        -1.3307e-02, -4.5345e-02, -5.3514e-03,  6.5170e-03,  3.2339e-03,
         3.9967e-03, -2.0032e-03, -5.3150e-03, -1.0268e-02,  1.1964e-02,
         6.7065e-03, -2.6293e-02,  9.2083e-03, -4.7019e-04, -2.0103e-03,
         1.2543e-02,  3.6806e-03, -5.6657e-03,  1.3862e-03,  3.5623e-03,
        -5.2279e-03,  2.7438e-03,  9.0478e-03, -3.1370e-03,  5.5316e-03,
         5.1856e-03, -4.5136e-03,  7.0247e-03, -8.3391e-04,  1.4451e-03,
        -1.5184e-02,  6.1571e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2508e-02,  6.9392e-01, -1.6853e-02,  4.2231e-03, -3.5110e-02,
        -2.0981e-02, -6.2537e-03, -2.2711e-02,  7.7538e-03, -1.3027e-02,
        -1.3220e-02, -6.8547e-05,  1.8557e-02,  2.0701e-02,  6.2147e-03,
        -9.4621e-03,  6.6195e-03,  6.0023e-03,  1.2233e-03, -4.6112e-03,
        -1.0557e-02,  1.0519e-02, -1.7026e-02,  1.1880e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0197,  0.6029, -0.0147,  0.0204, -0.0047,  0.0171, -0.0178,  0.0319,
        -0.0096, -0.0360, -0.0018, -0.0220,  0.0280, -0.0072, -0.0055, -0.0175,
        -0.0025, -0.0147, -0.0065,  0.0052, -0.0028, -0.0357, -0.0029, -0.0032,
        -0.0177, -0.0106, -0.0081, -0.0100,  0.0026,  0.0060,  0.0023, -0.0126,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6677e-03, -7.0953e-01, -7.3898e-02, -5.5463e-04,  8.8832e-03,
         2.7567e-02, -3.3806e-03,  2.6883e-03, -8.3625e-03,  8.6547e-03,
         7.2221e-03,  1.2452e-03,  5.1574e-03,  1.1681e-03,  4.8103e-03,
         5.7723e-03, -2.9431e-03, -4.1571e-03,  2.6257e-03,  7.6381e-03,
         2.1242e-02,  1.0553e-02,  5.8016e-03, -5.6874e-03, -2.2936e-03,
        -1.5825e-02,  2.7426e-02,  9.0284e-03,  6.5290e-03,  3.2126e-05,
         5.6574e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7631e-02,  6.8934e-01,  1.0511e-01,  6.5788e-03, -7.6414e-03,
         7.4187e-03, -1.0760e-02, -1.2871e-02, -1.3322e-02,  7.9316e-03,
         3.0615e-02,  3.1089e-03, -1.3817e-02, -2.3962e-02,  1.0633e-02,
         2.4152e-04, -9.1747e-04, -2.9008e-03,  6.6072e-04,  6.1458e-03,
        -6.9308e-05,  9.8647e-04, -7.3409e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.3914e-03,  7.5314e-01, -2.6077e-02, -3.3941e-02,  5.4754e-03,
         9.3172e-04,  5.9656e-04,  4.4226e-03,  8.1830e-03, -9.6259e-03,
        -4.1529e-03, -2.2908e-02,  1.5771e-03, -1.9965e-02, -4.0644e-03,
        -7.8860e-04, -2.8133e-03, -1.8020e-03, -5.1831e-03, -7.0669e-03,
        -2.1312e-02, -6.7502e-03,  5.7038e-03, -8.5917e-03, -1.1251e-02,
         4.2679e-03,  1.0751e-02, -1.0266e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-2.3234e-02,  7.5192e-01, -5.8679e-03,  2.2708e-02,  7.9858e-03,
         3.3634e-03,  1.4879e-02,  1.8934e-02, -3.6850e-05, -1.4648e-02,
         1.9636e-03,  1.0536e-03, -2.1551e-03, -9.4552e-03,  1.5537e-03,
        -5.4553e-04, -5.4931e-03,  6.6078e-03, -2.9642e-03, -3.9311e-03,
         1.3707e-02, -3.6703e-03, -1.7258e-02, -6.5302e-04, -8.5873e-03,
         2.7923e-03,  1.2635e-02, -1.8473e-02, -2.2923e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.4679e-02,  6.8751e-01, -1.4402e-02,  8.4049e-04,  1.4212e-03,
        -8.1924e-03,  4.3952e-03, -5.7468e-03,  9.4394e-03, -1.8751e-02,
        -1.8803e-02, -4.0789e-05, -1.6862e-02, -7.5281e-03, -7.7689e-03,
         8.3242e-03, -1.5333e-03, -4.8291e-03,  2.7619e-04, -4.1119e-03,
        -1.1925e-02, -8.3533e-03, -3.8568e-02, -5.5694e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0314, -0.5418,  0.0073, -0.0133, -0.0121, -0.0189,  0.0047, -0.0083,
         0.0062,  0.0053, -0.0185, -0.0359,  0.0172,  0.0083,  0.0132,  0.0117,
         0.0182,  0.0024,  0.0188, -0.0007, -0.0148,  0.0261,  0.0096, -0.0104,
         0.0009, -0.0024,  0.0204, -0.0010,  0.0098,  0.0019, -0.0027, -0.0178,
         0.0340, -0.0538], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7963e-02, -7.2523e-01,  8.0918e-02,  1.1149e-02,  5.1822e-03,
         1.0033e-02, -2.8140e-02, -2.2961e-02, -1.4323e-02, -6.4638e-03,
         1.2015e-04, -3.1462e-03, -2.2460e-03,  2.7104e-03, -1.2930e-03,
        -1.0467e-02, -1.5316e-02, -6.7808e-03,  1.0639e-03, -7.5160e-03,
        -6.8406e-03,  1.0137e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0154, -0.6953,  0.0028,  0.0889,  0.0142,  0.0127,  0.0056,  0.0095,
        -0.0093, -0.0162, -0.0454, -0.0231,  0.0172, -0.0035,  0.0061,  0.0046,
        -0.0170,  0.0132,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9828e-02, -7.1233e-01, -3.9935e-02,  1.4442e-02,  7.7840e-03,
         7.2534e-03,  6.4854e-05, -1.6830e-02, -6.9795e-03, -1.3442e-02,
        -1.3862e-02,  1.1187e-02,  2.0696e-02,  1.1168e-02,  1.3434e-03,
         1.7035e-02,  3.9254e-03,  4.2228e-03, -2.9104e-03, -9.3094e-04,
        -6.5058e-04,  1.7384e-02,  2.5795e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0575, -0.7159, -0.0332, -0.0258,  0.0122,  0.0034,  0.0062, -0.0215,
         0.0398,  0.0041,  0.0021, -0.0119, -0.0009, -0.0129,  0.0237, -0.0147,
         0.0143,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2460e-02,  7.2447e-01,  5.7954e-02, -1.9730e-02, -1.6843e-02,
         1.9746e-02,  1.4246e-02,  2.7473e-02, -2.0964e-02, -6.2454e-03,
         1.8911e-02, -5.8336e-03, -9.3173e-04, -8.6318e-04, -1.1716e-04,
        -2.8984e-02,  1.1709e-02, -1.2517e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0127,  0.5978,  0.0267, -0.0213, -0.0123, -0.0247, -0.0034,  0.0129,
        -0.0155,  0.0533,  0.0020,  0.0175, -0.0098, -0.0073, -0.0021,  0.0137,
        -0.0057, -0.0101,  0.0018, -0.0077,  0.0063, -0.0082, -0.0253, -0.0168,
        -0.0174, -0.0678,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1486e-03, -6.3816e-01,  1.1846e-01,  1.8439e-02, -2.2725e-02,
        -1.6558e-02, -4.6286e-03,  4.0848e-03, -1.1305e-02, -5.8744e-03,
        -2.4293e-03,  1.4326e-02,  5.3109e-03,  1.7225e-03, -2.4904e-03,
         7.0748e-03,  6.2908e-03,  2.3052e-03, -1.2541e-03,  5.1105e-03,
        -2.0157e-03, -1.3083e-02,  4.4088e-03, -5.4404e-03,  4.3954e-03,
         3.2100e-03,  4.0381e-03,  3.3333e-04, -7.1911e-03, -1.2480e-03,
         5.9941e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2654e-03,  7.4923e-01,  4.5114e-02, -2.1345e-02, -5.1564e-03,
         2.8675e-03,  1.4658e-02,  1.0184e-03, -1.4337e-02,  2.5934e-02,
        -2.5348e-02, -3.6937e-02,  8.9435e-03,  1.2629e-02,  3.3509e-03,
         5.4963e-04,  2.3063e-02, -8.2569e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1500e-02,  7.1200e-01,  5.1639e-02, -8.5570e-02, -5.0977e-02,
         2.5302e-02, -1.2660e-02,  3.4995e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 4.0928e-02,  7.1122e-01,  3.1696e-02,  9.6705e-04,  3.3774e-02,
         1.6407e-03, -7.7444e-03, -3.3138e-03, -8.5512e-03,  1.6075e-02,
        -2.2092e-03, -6.8212e-03,  7.3672e-03,  6.1014e-04,  7.5456e-03,
         8.5518e-03,  7.9567e-03, -3.7492e-03, -2.0577e-03, -9.3368e-04,
        -9.8749e-03, -1.4541e-02, -1.0946e-02, -3.8651e-03, -9.2457e-03,
         7.6705e-03, -5.6952e-03,  4.3628e-03,  1.6379e-03,  1.2280e-04,
         5.7349e-03,  6.4370e-03,  2.1195e-03,  3.5236e-03, -3.0865e-03,
         5.9054e-03, -4.8484e-04, -6.9288e-04,  3.3827e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0209e-02,  7.0044e-01,  1.6415e-03, -2.0637e-02,  5.2072e-03,
        -5.0410e-03,  2.2622e-02,  1.0462e-02,  2.2210e-02, -1.8711e-02,
         5.7037e-03,  1.3390e-02,  2.6516e-02,  3.1580e-03,  1.3763e-03,
         7.1933e-03, -4.8238e-03, -7.3241e-03,  3.3043e-03, -4.2358e-03,
         7.5633e-04, -1.3515e-02, -8.2905e-05,  7.0781e-04, -4.4120e-03,
        -4.5257e-03, -3.0771e-03, -6.8719e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0518, -0.6323, -0.0556, -0.0241, -0.0251,  0.0160, -0.0319,  0.0329,
         0.0026, -0.0202, -0.0066, -0.0320,  0.0689,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6999e-02,  7.5011e-01, -5.5324e-02, -2.5767e-02,  1.3343e-02,
        -1.0575e-02, -1.9916e-02,  1.9105e-02,  9.1543e-04, -1.1290e-02,
         4.1658e-04, -1.9844e-02, -3.0451e-04,  1.5895e-02, -2.0193e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.3185e-03, -6.4760e-01, -4.1756e-02,  4.4892e-02, -3.7846e-02,
         3.3599e-03,  1.1723e-02,  7.6919e-03, -6.2139e-03,  8.1401e-04,
         4.7146e-03,  3.5068e-03,  8.7794e-04, -4.4587e-03, -5.2561e-03,
        -4.3581e-04,  3.2104e-03, -2.2758e-02,  4.5980e-03,  6.5978e-03,
         9.0501e-03,  4.6767e-04,  5.0286e-04, -2.8191e-03,  8.8890e-04,
         1.6666e-02, -8.9793e-04, -9.3012e-03,  6.8086e-04, -5.5740e-03,
        -1.2210e-02,  3.2390e-03, -1.0637e-03,  7.3245e-03, -9.5324e-03,
         6.8644e-04, -1.8236e-03, -5.8451e-03,  3.2180e-03,  3.9640e-04,
         5.1559e-03,  3.5772e-03, -1.5574e-03,  1.8332e-03, -4.0377e-03,
         1.5753e-02,  8.2667e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4477e-02,  7.5240e-01,  2.6276e-02, -4.3224e-03,  1.1958e-02,
        -2.5448e-03, -9.8179e-03,  4.2481e-03, -3.5174e-04, -1.0584e-02,
        -4.2317e-03, -3.1217e-03,  2.8032e-04,  1.4352e-03, -1.7284e-03,
         2.8544e-03, -6.4186e-03, -6.6279e-03,  3.9047e-03, -1.0042e-03,
        -2.6469e-03, -2.4998e-04,  4.2390e-03,  5.7056e-03, -3.6270e-03,
         1.1023e-03,  3.3645e-02,  7.2542e-03,  5.7209e-04,  6.9094e-03,
         9.3172e-03,  4.6719e-04, -1.4243e-03,  9.7158e-04, -4.1719e-04,
         1.1673e-04, -2.0308e-03,  2.9759e-04, -4.3340e-03, -6.9896e-03,
         1.5627e-02, -2.3468e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.2654e-02, -6.5817e-01, -3.1480e-02,  4.1371e-03,  2.4482e-03,
        -1.6346e-02, -7.4603e-03,  1.2186e-04,  9.7371e-04, -3.0693e-02,
        -1.7218e-02,  2.1937e-02, -1.2426e-02, -4.3966e-04, -1.0286e-02,
         2.2410e-02, -3.6773e-03, -1.2532e-02,  1.0775e-02, -6.5351e-03,
         9.7964e-03,  3.2447e-03,  1.0714e-02, -8.1112e-03, -4.5413e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9475e-02,  7.9792e-01, -2.4813e-02, -9.6787e-03, -1.1376e-02,
         9.1796e-03,  6.1707e-03, -4.9953e-03,  4.4027e-03, -1.2469e-04,
        -3.5155e-03, -1.1962e-03, -3.0329e-05,  3.2858e-03,  1.7654e-02,
         7.5051e-03,  2.8388e-03, -2.1620e-03,  1.6081e-03, -3.1764e-02,
        -6.2762e-03,  2.6229e-03,  1.4013e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5179e-03, -7.7963e-01, -3.5556e-02,  2.4536e-02,  1.3407e-02,
        -1.0214e-02,  1.1792e-02,  2.8723e-03,  1.7586e-02, -1.2775e-02,
        -1.0729e-02,  1.5077e-02,  8.7304e-03,  1.3080e-02,  5.4938e-03,
         1.0025e-03, -3.8282e-04, -8.0285e-03, -4.7450e-03,  4.7168e-03,
        -4.6111e-03,  7.0707e-03,  5.4494e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0488e-01,  7.4261e-01, -5.2505e-03, -3.2883e-02,  5.0671e-03,
        -2.2329e-04,  1.5279e-03, -3.0414e-02, -2.5485e-03, -1.3723e-02,
        -9.7870e-03,  5.7305e-03, -4.5357e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1091,  0.7721,  0.0083,  0.0069, -0.0047,  0.0026, -0.0100,  0.0032,
        -0.0011,  0.0111,  0.0094,  0.0417,  0.0185, -0.0014,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.7944e-03,  8.1953e-01, -2.7387e-02, -1.3408e-02,  2.1719e-04,
        -2.5171e-03,  8.9161e-03, -5.2059e-03,  1.3428e-02,  6.3824e-03,
         8.2184e-03,  9.3247e-03,  7.6731e-03,  6.1234e-03,  1.4409e-02,
         2.3314e-03, -1.8301e-02, -7.8100e-03, -1.9028e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-3.1972e-02, -7.7641e-01,  4.2354e-02,  6.5423e-03,  2.1206e-04,
        -2.5330e-02, -8.5623e-03,  6.6956e-03,  5.2137e-03, -4.1989e-03,
        -4.7497e-03,  8.4178e-03,  9.1121e-04, -9.7482e-03,  8.8234e-03,
         2.0473e-03,  2.6899e-04,  1.8580e-04,  2.3069e-03, -1.1696e-02,
         1.4013e-02, -6.3977e-03, -4.4006e-03,  4.0163e-03, -7.5509e-03,
        -6.9741e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0191, -0.7709,  0.0330, -0.0269, -0.0331,  0.0125, -0.0177,  0.0090,
        -0.0183,  0.0203,  0.0011, -0.0015, -0.0086, -0.0204, -0.0041, -0.0035,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1058,  0.6139,  0.0144,  0.0841, -0.0103,  0.0055,  0.0031,  0.0185,
        -0.0178,  0.0113, -0.0080, -0.0424, -0.0097,  0.0037, -0.0100, -0.0210,
         0.0206,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4003e-01,  6.8753e-01,  1.8822e-02,  3.5283e-02,  1.5835e-02,
         6.4483e-03,  5.0845e-03,  6.3057e-04, -9.0342e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2709e-02,  5.1118e-01,  3.6202e-04, -3.0847e-02, -1.1545e-01,
        -7.1783e-03, -5.6954e-03,  2.9977e-02, -7.7345e-03,  5.1902e-03,
         2.9528e-02,  3.5970e-03,  5.6013e-03,  1.1256e-02,  3.6702e-03,
         7.8314e-03,  4.4095e-03,  2.5244e-03, -1.1633e-03,  1.7414e-02,
         1.1117e-02,  4.8509e-03,  5.0436e-03, -2.2963e-03, -8.8288e-05,
         2.6732e-03,  3.4043e-03,  4.4877e-03,  2.3216e-02,  7.6187e-03,
         4.0492e-03,  1.8697e-02, -4.5646e-03,  4.6345e-03,  7.1120e-03,
         2.5584e-03, -5.1038e-04, -7.5805e-03, -7.9832e-03,  1.1202e-03,
        -2.9884e-04, -4.8837e-03, -2.9569e-03, -1.0760e-02,  4.6719e-04,
         2.3710e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7403e-02, -6.9659e-01,  2.7121e-02,  3.1725e-02, -1.5476e-02,
         2.4410e-03, -1.4954e-02,  1.1650e-04, -1.2603e-02,  1.2821e-02,
        -1.0941e-02, -7.9495e-04, -2.0696e-02, -5.3191e-03, -6.3149e-03,
        -1.3302e-03, -1.0464e-02, -6.5728e-03,  3.1215e-03, -9.0278e-03,
        -8.8505e-03, -9.9938e-03, -1.6719e-02, -1.3140e-03,  1.4868e-03,
        -7.6615e-03, -1.4803e-02, -3.0123e-03, -7.8701e-03,  2.4619e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0482e-01,  7.0466e-01, -6.0267e-02, -1.9424e-06, -4.2610e-02,
         8.5436e-02,  2.2116e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0633,  0.7967,  0.0201,  0.0085,  0.0162, -0.0044,  0.0147, -0.0008,
        -0.0291,  0.0255,  0.0206,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0513, -0.6821, -0.0656,  0.0504,  0.0128,  0.0252,  0.0061,  0.0171,
         0.0052,  0.0121,  0.0100, -0.0133, -0.0217,  0.0272,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.3379e-02, -7.3608e-01, -1.5967e-02, -3.4092e-03, -3.0353e-02,
        -3.9316e-04, -2.9526e-04,  3.7828e-04, -9.5128e-03,  1.5173e-02,
         2.7941e-03, -4.0503e-03, -2.1641e-03, -4.0625e-03, -5.3240e-03,
         1.4508e-03,  6.5197e-04,  6.6068e-03,  2.2900e-03, -1.1542e-02,
         3.9869e-03, -4.8376e-03, -2.9042e-03, -8.5113e-03, -1.3812e-06,
        -5.5421e-03,  8.3159e-03, -7.9146e-03,  6.6976e-03, -5.3745e-03,
        -1.2438e-02, -2.5472e-03, -1.2684e-02, -2.2365e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0346,  0.5908, -0.0270,  0.0076,  0.0354, -0.0183, -0.0132, -0.0022,
         0.0140, -0.0042, -0.0135, -0.0215, -0.0032,  0.0334,  0.0375,  0.0088,
         0.0068, -0.0228, -0.0019, -0.0076,  0.0037,  0.0022, -0.0228,  0.0100,
         0.0209,  0.0135,  0.0133, -0.0096,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4447e-02, -7.0254e-01, -3.4026e-02,  3.4188e-02, -4.9063e-03,
         2.6858e-02, -7.2589e-03, -6.9232e-03, -7.8267e-03,  1.0017e-03,
        -9.3493e-03,  3.3628e-03, -1.6364e-02, -2.9509e-02,  1.1712e-02,
        -6.8285e-04,  3.1635e-03, -1.5887e-02,  9.3820e-04, -6.4235e-04,
        -6.6802e-03, -5.0407e-03, -3.4339e-02,  2.3567e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-7.5423e-02,  6.8141e-01,  1.8864e-02,  4.9744e-02,  1.1842e-02,
        -5.0922e-04,  8.1091e-04, -1.3453e-02, -5.8706e-03,  1.5024e-03,
        -1.4053e-02, -4.1178e-03,  3.3576e-03,  1.4392e-02,  1.1952e-03,
         1.7808e-03,  7.2208e-04, -1.9067e-02,  5.9557e-03,  3.3350e-04,
        -1.0121e-02, -6.6106e-03, -4.2834e-03, -4.4546e-03,  2.3234e-03,
        -4.0383e-03, -6.8837e-03,  1.1391e-02,  7.5870e-04, -2.1223e-02,
        -3.5033e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0862, -0.6507,  0.0340, -0.0064, -0.0304, -0.0269, -0.0323, -0.0286,
        -0.0162, -0.0091, -0.0130,  0.0661,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0546, -0.7746,  0.0219, -0.0312, -0.0290,  0.0048,  0.0101, -0.0022,
         0.0079, -0.0019,  0.0106,  0.0099, -0.0222,  0.0125, -0.0067,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9399e-03,  7.6751e-01, -9.1788e-03, -2.9149e-02, -1.6426e-02,
        -2.1394e-03,  5.9861e-03,  1.0709e-02,  7.0101e-03, -1.1524e-03,
        -2.1153e-02, -4.0791e-03, -8.3799e-03,  2.1799e-03, -4.5833e-03,
         1.4646e-02, -8.2523e-03, -6.7667e-03,  4.0567e-03, -9.2033e-03,
         2.9682e-03, -7.5471e-03, -3.5858e-03,  3.0064e-03, -5.8679e-03,
        -1.4718e-03,  4.8860e-03, -6.2586e-04, -6.3669e-03,  9.6323e-03,
         4.8172e-03, -3.1649e-03,  1.0561e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5416e-02, -7.3707e-01, -3.8809e-02,  2.7467e-02, -5.9302e-03,
         3.2973e-02, -1.6262e-04,  1.6592e-02, -6.7843e-03, -8.4092e-03,
        -3.6675e-04, -8.7182e-03, -5.3263e-03,  4.8609e-02,  1.7482e-02,
         2.9880e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1482,  0.6192,  0.0966,  0.0045,  0.0201,  0.0117,  0.0307,  0.0543,
         0.0147,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0219,  0.7497,  0.0141,  0.0994, -0.1149,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8295e-02, -7.5558e-01,  1.7752e-02, -1.6920e-02,  1.0772e-03,
         2.2585e-02,  1.4188e-03, -2.6775e-03,  2.5629e-02, -1.1104e-02,
         6.4066e-03,  1.2695e-02,  2.7775e-02, -7.5365e-03,  9.6452e-04,
         2.1491e-03,  1.0061e-03, -4.5345e-04, -1.1607e-02, -4.8857e-03,
        -4.3810e-03, -8.8093e-03, -4.6176e-03,  1.3222e-02,  1.4845e-03,
        -1.0600e-02,  8.3665e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([0.0492, 0.7899, 0.0398, 0.0023, 0.0255, 0.0657, 0.0276, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.8028e-02, -6.6971e-01, -2.8365e-02, -4.9594e-03,  5.2551e-04,
         3.2883e-03, -2.3104e-02,  9.2253e-05,  1.6608e-02,  3.5391e-02,
         6.3618e-03, -2.6595e-02, -9.9237e-03,  2.4014e-03,  5.0348e-03,
        -1.2859e-02,  7.5917e-04, -1.4123e-02,  1.1389e-02, -1.8603e-02,
         2.7424e-02, -3.4453e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0232, -0.6895,  0.0081, -0.0127,  0.0575,  0.0046,  0.0223, -0.0440,
        -0.0090,  0.0015, -0.0171,  0.0145,  0.0064, -0.0058, -0.0068, -0.0114,
        -0.0127, -0.0298, -0.0231,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5414e-02,  7.8143e-01,  3.2028e-02, -1.2233e-02,  1.1079e-02,
         2.9409e-02,  6.0612e-03,  5.1422e-03, -1.1601e-03, -1.2655e-02,
        -6.5786e-03,  5.0408e-05, -3.2796e-03, -2.6952e-03,  5.6053e-03,
         2.3993e-03, -6.8187e-04,  3.2934e-03, -5.7026e-03,  7.0192e-03,
         6.0842e-03, -1.1871e-02,  5.9153e-03,  5.5836e-03,  2.1756e-04,
         9.5072e-03, -3.3350e-04,  5.6016e-03, -1.1276e-03, -7.5233e-04,
         9.0929e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 1.0463e-01, -5.6718e-01, -3.3144e-02,  6.1699e-02,  2.1790e-02,
         1.9781e-02,  3.6833e-03, -9.4830e-04, -9.9961e-04,  7.1418e-03,
         2.3979e-03, -7.1377e-03, -1.1689e-02,  7.0914e-03, -6.8060e-03,
         3.3374e-04,  4.7102e-03, -1.8416e-02,  3.0422e-03, -9.0511e-03,
         5.3681e-03,  7.6436e-03,  1.1360e-02, -7.3303e-03, -7.2731e-03,
         4.3820e-03, -3.0719e-03,  9.9389e-03,  4.4094e-03,  2.0247e-03,
        -9.2227e-03,  3.6301e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0052, -0.7196, -0.0625,  0.0054, -0.0032, -0.0071, -0.0239, -0.0133,
        -0.0199,  0.0092,  0.0439, -0.0029,  0.0119, -0.0016,  0.0040, -0.0034,
        -0.0157, -0.0113,  0.0030,  0.0018, -0.0175, -0.0137,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.0926e-02,  6.6711e-01,  3.1346e-02, -1.7117e-02, -2.2110e-03,
         3.5039e-02, -3.5030e-02,  2.4241e-02,  1.1331e-02,  7.2133e-03,
        -1.8063e-02, -5.9387e-03, -4.7810e-03, -5.4935e-03,  6.2104e-03,
         1.5308e-03, -2.4270e-03,  8.4608e-03,  7.0233e-03,  5.7202e-03,
         4.6574e-03, -1.5937e-02,  1.0006e-02, -4.1770e-03, -3.9613e-03,
         6.8005e-03, -3.0675e-04,  1.4312e-02, -2.4128e-03, -2.2009e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0822,  0.7442, -0.0123,  0.0314, -0.0183, -0.0015, -0.0100, -0.0055,
         0.0119, -0.0111,  0.0009, -0.0054,  0.0052,  0.0113,  0.0020, -0.0066,
        -0.0052,  0.0351,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5110e-02,  6.9669e-01,  3.9325e-02,  1.5952e-02, -2.7467e-03,
        -3.9964e-02,  8.6499e-04,  3.5376e-05, -9.0238e-03, -7.2845e-04,
         2.4267e-03, -5.4466e-03, -8.2047e-03,  4.8636e-03, -1.3948e-03,
         1.4321e-03,  4.0394e-03, -3.8310e-03, -4.5663e-03, -7.5111e-03,
        -5.4891e-03,  6.2480e-05, -1.6284e-02,  7.0412e-04,  3.6070e-03,
        -5.1828e-04,  9.8759e-03, -5.6274e-03,  2.0010e-03, -1.7472e-03,
        -2.0317e-03, -5.5281e-03,  9.4603e-03,  5.3737e-03, -3.1800e-03,
        -4.7552e-02,  1.6802e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0123, -0.6640,  0.0394,  0.0457,  0.0059, -0.0014, -0.0118,  0.0195,
         0.0014, -0.0104, -0.0048,  0.0188,  0.0039, -0.0089, -0.0101, -0.0074,
        -0.0051, -0.0082, -0.0067, -0.0185,  0.0110, -0.0171, -0.0276, -0.0402,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.1568e-03,  6.2445e-01, -3.3759e-02,  1.1599e-02,  1.4852e-02,
        -1.2383e-02, -2.3558e-02,  1.6158e-03,  1.1141e-02, -1.2907e-02,
        -4.5381e-03, -1.3265e-02, -1.7665e-02,  2.5688e-03,  5.4532e-05,
         2.4543e-03,  5.9069e-03, -4.6563e-03,  5.1095e-03, -2.5702e-03,
         5.6048e-04, -3.2386e-03,  2.6105e-03,  4.5909e-03,  1.5324e-02,
         2.1947e-02,  2.7225e-03,  2.6755e-03,  8.0403e-04, -7.9261e-03,
         7.1463e-03,  4.4846e-04,  1.6070e-03, -1.3552e-03,  1.2302e-02,
        -5.6440e-03, -4.3851e-03, -8.1764e-05,  2.8290e-03,  4.1266e-03,
        -6.5414e-03,  3.2373e-03,  2.1530e-03,  2.3278e-02, -2.4270e-02,
        -2.3980e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0488,  0.6130,  0.0827, -0.0158, -0.0654,  0.0139,  0.0049, -0.0053,
         0.0043,  0.0164,  0.0159,  0.0035,  0.0311, -0.0018,  0.0068, -0.0022,
        -0.0013, -0.0195, -0.0067,  0.0060,  0.0117, -0.0150, -0.0080,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0647, -0.7112,  0.0778, -0.0309,  0.0049,  0.0268,  0.0050, -0.0045,
         0.0233,  0.0395, -0.0115,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.3715e-02,  7.6289e-01,  1.0403e-02, -3.1714e-02, -7.3091e-03,
        -1.9245e-03,  9.8450e-03, -5.5064e-03, -9.4121e-03, -7.0319e-03,
         1.2489e-03,  3.7537e-04,  4.4169e-02, -2.4457e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1399, -0.6535, -0.0207,  0.0283,  0.0197,  0.0020,  0.0186, -0.0050,
         0.0485, -0.0157, -0.0316, -0.0165,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4998e-02,  6.8342e-01,  1.6831e-02, -1.3139e-02, -1.4476e-02,
        -4.8882e-04, -9.0477e-03,  1.0891e-02, -4.4837e-03,  1.3695e-02,
        -2.0200e-03,  6.1536e-03, -8.1188e-03, -1.4125e-03,  1.4698e-03,
         2.3159e-03, -3.2413e-03, -5.9954e-03, -4.7419e-03, -1.3506e-02,
        -4.2545e-03,  1.9584e-03,  6.4819e-03, -3.9376e-03,  1.0085e-02,
        -1.6862e-02, -5.9658e-03, -1.9102e-03, -6.6792e-03,  1.8119e-03,
        -7.7724e-03, -2.8640e-04, -3.6003e-03,  1.0873e-02,  4.0523e-03,
         3.8699e-03, -5.9704e-04, -3.7681e-03, -1.0734e-03, -3.7340e-03,
        -2.1431e-03, -3.5324e-05, -2.3065e-03,  1.8187e-03, -1.2976e-03,
         5.0272e-03,  8.3677e-04, -1.3847e-03, -1.5012e-03, -5.1546e-03,
         6.8588e-04,  4.2047e-03, -2.9802e-03,  1.2736e-03,  1.9326e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 0.0467, -0.5300,  0.0245, -0.0036,  0.1008,  0.0353, -0.0301, -0.0238,
         0.0126,  0.0173, -0.0140, -0.0080, -0.0093,  0.0100, -0.0324,  0.0129,
         0.0016, -0.0108,  0.0015,  0.0053, -0.0097, -0.0170, -0.0428,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8265e-02,  6.6303e-01, -5.8867e-02,  3.9145e-03, -5.4315e-04,
        -1.1440e-02, -2.1351e-02,  1.7747e-02, -1.3019e-03,  2.5136e-02,
        -1.1712e-02, -9.3438e-03, -1.7221e-02, -1.2125e-02, -1.4243e-02,
         8.6739e-03, -4.7840e-02, -5.7244e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1449,  0.5000, -0.0133, -0.0851, -0.0306, -0.0290, -0.0100, -0.0325,
        -0.0065, -0.0163,  0.0110,  0.0024,  0.0356, -0.0157,  0.0061,  0.0154,
        -0.0307,  0.0148,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0615, -0.6729, -0.0100,  0.0042, -0.0139,  0.0138,  0.0095,  0.0337,
         0.0192, -0.0171,  0.0245, -0.0149,  0.0031,  0.0013,  0.1005,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5025e-02,  6.4668e-01, -2.7637e-02,  4.4359e-02, -1.8721e-02,
         6.4917e-03,  8.4586e-03,  1.2922e-02,  2.6474e-02,  4.5063e-03,
         1.0203e-02, -8.6601e-03, -7.0989e-04, -4.0124e-03,  7.9290e-04,
         2.7694e-03, -1.2520e-03, -1.8621e-02, -9.1926e-03, -8.0565e-03,
         1.1709e-02, -1.3036e-02, -3.1094e-03,  1.3527e-03,  3.8056e-03,
         2.0776e-02,  1.0827e-02, -2.0674e-03,  4.2365e-03,  3.1302e-04,
         4.2933e-02, -2.8921e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9917e-02, -6.4519e-01, -6.0228e-02, -9.6830e-03, -1.3200e-02,
        -3.1326e-03,  7.5915e-03,  9.9280e-03,  1.2028e-02, -3.7287e-03,
         1.7333e-02,  5.4445e-03,  6.5233e-04,  1.0934e-03, -2.1669e-03,
         2.2511e-02,  3.7116e-03,  1.3129e-02, -5.3174e-03,  5.7470e-03,
         2.1071e-02,  9.2428e-03,  1.4263e-02,  2.3417e-03,  8.3011e-03,
         5.3086e-04, -2.6350e-04, -1.6023e-02, -1.6119e-03, -5.1511e-03,
         6.6916e-03,  3.5305e-03, -2.9242e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3489e-02,  6.5647e-01,  3.1657e-02,  1.9039e-02,  2.2665e-02,
        -1.1085e-02, -1.6971e-04, -7.1797e-03,  1.0792e-02,  8.4922e-03,
        -1.4554e-03, -1.0392e-02, -1.5961e-02, -4.5130e-03, -2.6023e-03,
         1.1899e-02, -7.1951e-03, -5.4024e-03, -4.7524e-03, -7.5480e-03,
        -8.9867e-03, -4.9843e-03, -6.2429e-03, -3.1074e-03, -2.6785e-03,
        -1.0696e-02,  5.5096e-03,  8.8303e-03,  4.8973e-03,  3.0952e-03,
        -3.2847e-03,  5.4416e-03, -9.0051e-04, -3.7181e-03,  9.5506e-04,
        -2.2384e-03,  7.2951e-04, -1.1055e-04, -4.3455e-04,  8.5457e-04,
        -1.6033e-03, -5.1524e-04, -3.0589e-03, -2.2681e-03, -3.9283e-03,
         3.1179e-03,  1.7895e-03, -5.8680e-03,  3.5172e-03,  1.7267e-03,
         4.9767e-03,  1.0697e-03,  1.2630e-03,  3.9317e-05,  2.6715e-03,
         7.9926e-04,  1.1246e-02, -8.2346e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2152e-02, -6.9088e-01, -6.4445e-02, -2.1167e-04, -4.4587e-03,
         5.5201e-03, -5.2765e-04,  2.0210e-02,  6.8143e-03,  9.1004e-03,
         9.1232e-03,  7.8357e-03,  3.4442e-03, -2.6627e-02,  5.7800e-03,
         9.6952e-03,  2.3059e-03,  4.7860e-03,  5.2219e-03,  3.5393e-03,
         7.4185e-04,  1.2350e-03,  6.1753e-03,  2.7960e-03, -1.5488e-02,
         5.0886e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0307,  0.7748, -0.0027, -0.0209, -0.0061,  0.0146, -0.0072,  0.0054,
        -0.0105, -0.0101,  0.0073, -0.0024,  0.0019,  0.0084, -0.0078,  0.0048,
         0.0130, -0.0114,  0.0084,  0.0198,  0.0070, -0.0050, -0.0050, -0.0066,
         0.0081,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9121e-02, -6.9453e-01, -1.6818e-02, -6.6520e-04, -1.8464e-02,
        -1.4567e-02,  5.6733e-03,  1.1788e-02,  8.1950e-03, -1.5285e-02,
         1.0119e-02, -7.1317e-03, -1.6232e-02,  2.6940e-03,  1.3552e-02,
        -7.7432e-03,  6.7904e-03, -3.6256e-03, -1.9581e-03, -6.5716e-03,
         1.2047e-03,  7.4731e-03,  7.3904e-04,  6.1075e-03, -5.8742e-03,
        -1.9559e-03,  6.3106e-05, -3.6571e-03,  7.2987e-03,  2.4471e-03,
        -1.0041e-03,  4.4267e-03,  2.3176e-03,  1.8776e-03, -4.8519e-03,
        -4.1124e-04,  4.5779e-03, -7.0971e-03, -1.5314e-02, -5.7879e-03,
         3.5485e-03,  2.0859e-03, -4.2744e-04,  9.3797e-03,  8.5478e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0176,  0.6887,  0.0356,  0.0696, -0.0035,  0.0065,  0.0085, -0.0010,
         0.0012, -0.0126, -0.0243,  0.0034,  0.0021, -0.0104, -0.0081, -0.0116,
        -0.0201, -0.0145,  0.0030,  0.0021,  0.0074, -0.0075, -0.0407,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0946,  0.7269, -0.0499,  0.0144,  0.0293, -0.0015,  0.0292, -0.0052,
         0.0052, -0.0134,  0.0204,  0.0099,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-4.8337e-02,  6.2354e-01,  8.0175e-02, -2.4300e-02, -1.2223e-02,
         1.0779e-02,  3.2239e-02, -7.9484e-03, -7.4758e-03,  6.2503e-03,
         1.8713e-02,  7.5367e-03, -1.3625e-02,  1.3036e-02, -5.1251e-03,
         3.0381e-04, -6.0504e-03,  1.4615e-02,  8.8673e-03, -2.5387e-02,
        -1.2723e-02, -2.9975e-04, -2.0451e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5409e-02, -7.0482e-01, -9.1798e-02,  3.3729e-02, -6.9419e-03,
         2.6431e-03, -6.6403e-03, -9.2547e-03, -4.1562e-03, -5.8590e-03,
        -2.4309e-03, -5.0835e-03,  4.7945e-03, -2.8709e-03,  1.0304e-03,
        -4.1891e-03,  3.1200e-03,  8.2689e-03,  7.2082e-03, -3.6605e-04,
        -3.6725e-03,  1.5026e-03,  3.3476e-03,  5.9275e-03,  9.4290e-03,
        -3.5689e-03, -2.5905e-03, -7.6500e-03, -1.1329e-03, -5.6716e-03,
         8.0923e-04, -2.6873e-02, -7.2124e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0784e-02,  7.7882e-01,  1.1711e-03, -5.4059e-03, -6.6025e-03,
         1.6658e-02, -2.4661e-03, -7.0783e-03,  8.6665e-03, -1.2883e-03,
        -2.2082e-03,  1.5977e-02,  9.4768e-03,  1.9889e-02,  2.5932e-02,
        -1.6719e-02, -8.9034e-05,  1.4895e-02,  3.5878e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0072, -0.8256,  0.0272,  0.0071,  0.0271, -0.0068,  0.0214,  0.0025,
         0.0013, -0.0188,  0.0096, -0.0144, -0.0059, -0.0252,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9173e-03,  7.8433e-01,  4.1676e-02,  1.8479e-02, -1.6272e-02,
        -5.0348e-03,  3.3169e-03,  2.3158e-03,  1.7876e-02,  6.1975e-03,
         6.8730e-03,  1.4132e-03,  1.1142e-02, -2.7405e-02, -9.7175e-03,
        -5.3749e-03, -1.3876e-02,  2.9000e-03,  2.4066e-03,  1.3488e-03,
         7.0788e-04, -1.9416e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6206e-02,  7.5651e-01,  3.2503e-02,  1.5530e-02, -6.9082e-03,
         1.1190e-02,  5.8936e-03, -1.2603e-03,  7.7287e-04, -2.9076e-03,
        -6.4366e-03, -2.1317e-02, -1.7133e-02, -2.7683e-03, -4.8449e-03,
         6.1786e-04,  4.2788e-03, -4.8409e-03,  5.0565e-03,  6.4249e-03,
        -1.2301e-02,  5.7116e-03,  6.2359e-04, -8.7114e-03, -7.6744e-03,
         1.4236e-03,  2.5735e-03,  6.8148e-03,  1.3971e-02,  6.7890e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0606,  0.6595,  0.0021, -0.0386, -0.0048,  0.0414,  0.0049,  0.0124,
         0.0207, -0.0187, -0.0192, -0.0029, -0.0187, -0.0025, -0.0010,  0.0171,
        -0.0750,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0843, -0.7538, -0.0188, -0.0115, -0.0087, -0.0157,  0.0169, -0.0057,
        -0.0326, -0.0325,  0.0197,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2399e-01, -5.0431e-01, -5.7232e-02,  1.7592e-02, -5.8007e-03,
         7.6425e-02, -6.5795e-02, -8.5712e-03, -1.0071e-02, -6.7626e-03,
        -9.0344e-03, -1.0239e-02,  1.5913e-02, -1.2688e-02, -1.9804e-04,
         2.9402e-02, -4.5974e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.7190e-02,  6.7204e-01, -9.8486e-03,  1.5723e-02, -8.0000e-03,
         1.0018e-02, -5.4791e-03,  6.3690e-03,  1.0216e-02, -2.4454e-03,
         5.1279e-03,  2.5129e-03,  2.7068e-03, -1.1898e-02, -9.2777e-03,
         4.4288e-03, -9.7693e-03, -8.1374e-05, -2.5855e-02, -8.8094e-03,
        -7.1100e-03,  1.2545e-02,  6.8386e-03,  4.8336e-03,  9.4946e-03,
        -5.1382e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.8176e-03,  7.1995e-01,  4.9230e-02, -1.3964e-02,  6.9830e-03,
         4.6373e-02, -1.4882e-02,  2.7117e-02,  1.5588e-02,  3.0720e-03,
        -4.7147e-03, -1.5596e-02,  1.0493e-03, -2.2146e-03,  5.1291e-04,
         1.0466e-02, -6.8416e-03, -3.0854e-03,  1.4872e-03, -3.7025e-03,
         7.5820e-03, -2.9975e-03, -1.1018e-03,  2.8369e-03,  6.8568e-03,
        -1.0713e-03,  2.1443e-03,  2.5845e-03, -1.0261e-02,  7.9159e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0835e-02,  8.3876e-01, -1.2967e-02, -5.8703e-03, -1.7920e-02,
        -1.3553e-02,  1.7248e-04,  1.3701e-02,  3.0308e-03, -1.6804e-03,
        -1.4984e-02,  1.8744e-03,  4.9022e-03,  4.1307e-03, -1.2142e-02,
        -4.9411e-03, -8.7455e-03,  2.1444e-03,  7.6503e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 2.0062e-02,  7.2989e-01,  7.9244e-03,  2.7075e-02,  9.0880e-03,
        -2.5270e-02,  1.9148e-02, -5.5880e-03,  2.0136e-03,  3.5603e-03,
        -1.0076e-02, -7.5450e-03, -9.2965e-03,  5.4483e-03,  1.8260e-03,
        -3.5671e-02,  8.1057e-03,  3.7696e-03,  1.8802e-03,  3.0547e-04,
        -1.1303e-02, -1.0751e-02,  2.5571e-03, -5.5959e-03, -1.7567e-02,
        -8.9831e-03, -9.7001e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0196,  0.7758,  0.0021, -0.0580, -0.0276, -0.0133, -0.0100,  0.0114,
        -0.0058,  0.0036, -0.0238,  0.0180,  0.0275, -0.0036,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0147,  0.7348,  0.0277, -0.0059, -0.0026, -0.0241,  0.0210, -0.0085,
         0.0124, -0.0087, -0.0082, -0.0026, -0.0128,  0.0154,  0.0029, -0.0141,
        -0.0044, -0.0088, -0.0106,  0.0120, -0.0068, -0.0025, -0.0060,  0.0088,
         0.0159, -0.0078,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.2889e-02, -6.3168e-01, -1.7860e-01,  1.6343e-04, -1.3716e-02,
         2.8408e-03,  2.4910e-03,  1.3975e-02,  5.3159e-03,  6.9758e-03,
        -8.6924e-04,  1.4724e-02, -3.8960e-02,  1.6794e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6511e-02,  7.4604e-01, -3.0539e-02, -8.3203e-03,  7.7416e-04,
        -5.2100e-03, -9.4815e-03, -1.3219e-03,  4.9728e-03, -2.3677e-04,
         1.9355e-04, -4.8296e-03,  4.8319e-03, -1.4152e-02,  1.0112e-02,
         4.3196e-03, -1.5100e-02, -9.0562e-03,  5.1761e-03,  1.8529e-02,
        -2.5667e-03, -2.8659e-03,  1.2495e-02,  1.2913e-03, -1.4181e-03,
        -1.5110e-03, -7.0712e-03, -1.4867e-02, -7.1564e-03, -2.8713e-03,
         5.2514e-03,  3.0930e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5800e-02,  7.0348e-01, -1.9840e-02, -3.5759e-02, -2.0251e-02,
        -2.3255e-02, -1.0503e-02, -1.3038e-02,  1.0880e-02, -3.2253e-03,
        -3.9808e-03,  4.5004e-03, -1.5747e-02,  4.8362e-03,  1.6805e-02,
        -3.9391e-04, -8.2009e-03, -4.3299e-03,  1.5770e-02,  5.6341e-03,
         8.4678e-04, -1.4172e-03,  1.3328e-03, -4.6948e-03, -2.5483e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0127,  0.8628, -0.0072, -0.0393, -0.0179,  0.0601,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2755e-02,  7.7495e-01,  3.9275e-02, -1.5841e-02,  1.7775e-02,
        -9.0162e-04,  1.1168e-02, -4.3728e-03,  3.2412e-03,  1.9149e-03,
        -6.4493e-03,  1.2347e-03, -1.0865e-03, -3.3910e-03,  1.4414e-03,
         7.3722e-03, -1.3180e-02, -6.4011e-03, -1.1605e-03,  3.5764e-03,
        -2.4970e-03,  9.4360e-04, -3.9486e-03, -1.0258e-03, -4.4689e-05,
         7.3474e-03, -3.2428e-03, -3.2924e-03, -3.1852e-04, -3.7657e-03,
         1.0847e-02,  3.0760e-03,  6.1272e-03, -2.8536e-04,  9.8492e-04,
         1.0151e-02,  1.4618e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.2561e-02, -6.3169e-01, -3.3463e-02, -2.2058e-02,  1.2954e-03,
        -1.6593e-02,  7.1204e-04, -8.5142e-06,  2.1128e-03, -5.5452e-04,
         3.8640e-03, -4.2873e-03, -6.4187e-03,  1.4496e-03,  1.8022e-02,
         1.8308e-02,  6.7321e-03, -1.3650e-02,  3.8440e-03,  1.4222e-04,
         3.6734e-02,  2.8754e-03,  1.3863e-02,  1.8773e-03,  4.0668e-03,
         5.0810e-03,  1.3391e-02, -1.5067e-02,  5.0386e-03,  3.2766e-03,
         2.6918e-03, -1.3176e-03,  8.3713e-05, -1.4646e-02,  2.2226e-02,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4183e-02,  6.5364e-01, -1.6435e-01,  6.5480e-03, -2.5618e-03,
        -1.8749e-03, -1.1099e-02, -4.1205e-03, -1.6171e-02,  1.7316e-02,
        -1.0860e-02,  4.2165e-04, -1.6316e-02, -5.0500e-02, -3.0038e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1486e-02,  7.2608e-01,  1.3635e-02, -2.5339e-02, -4.6567e-03,
         1.3928e-02,  2.4010e-02,  1.7796e-02,  8.1727e-04,  7.1782e-03,
        -3.1140e-03, -3.8066e-03,  1.9966e-03, -7.8155e-03,  2.4352e-03,
        -5.8305e-03,  2.9454e-03,  1.8791e-02,  2.3143e-02,  6.7339e-03,
         9.2131e-04, -3.4727e-03, -6.0592e-03, -5.1519e-03, -2.8718e-03,
         4.7015e-03, -5.1910e-03,  2.8756e-04, -6.5369e-03, -3.3457e-03,
         6.1431e-03,  1.2338e-02,  2.1437e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0407, -0.7203,  0.0048, -0.0096,  0.0039,  0.0187,  0.0190,  0.0065,
         0.0079,  0.0088,  0.0086,  0.0140,  0.0018,  0.0120,  0.0070, -0.0017,
        -0.0054, -0.0135,  0.0069,  0.0007,  0.0144,  0.0014, -0.0058,  0.0025,
        -0.0036, -0.0126,  0.0012, -0.0041,  0.0424,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #500: [tensor([-0.0900,  0.6844, -0.0047,  0.0028,  0.0016,  0.0061, -0.0015,  0.0103,
        -0.0065, -0.0496, -0.0065, -0.0016, -0.0039,  0.0628, -0.0677,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7167e-02,  7.0804e-01, -2.2652e-03, -9.8934e-03, -1.3188e-02,
        -1.7584e-02, -1.1224e-03,  7.4505e-03, -2.1942e-03, -2.3388e-02,
        -1.5266e-02, -1.0738e-02, -6.1514e-03, -1.3112e-02,  2.8260e-03,
         1.3300e-02,  4.7609e-03, -1.7342e-03,  1.4897e-03, -1.3056e-03,
         1.4080e-02, -1.7869e-04, -5.2137e-03,  3.1200e-03,  6.3936e-04,
         8.0708e-04,  1.0336e-02,  1.8150e-02, -4.0706e-03, -8.2123e-04,
         1.5232e-03,  8.4882e-05, -7.2127e-03, -3.5846e-03, -1.0751e-02,
        -6.7715e-04, -1.0591e-03,  1.1383e-03, -4.5699e-03,  3.1115e-03,
         8.5808e-03,  7.3540e-03,  1.2701e-03,  5.6893e-04,  4.7613e-04,
        -2.8991e-04,  7.4045e-03, -8.5765e-03,  1.3777e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.6083e-03, -6.2656e-01,  1.4122e-01,  2.7544e-02,  1.7839e-02,
        -9.4498e-03,  1.2152e-02,  1.2317e-02, -4.3169e-03,  5.3840e-03,
         3.8839e-03, -6.9155e-04,  2.5565e-03, -1.5974e-04,  2.2181e-03,
         3.4976e-03, -1.1901e-02, -2.5759e-04, -1.3444e-03, -1.7979e-02,
        -2.2523e-03,  8.0045e-03,  4.0332e-03,  1.1581e-03, -4.2094e-03,
         5.9183e-03,  6.6134e-03, -3.6685e-03, -1.7242e-03,  5.3628e-03,
         5.3227e-03, -8.3493e-03, -7.1483e-03, -1.9448e-02,  6.9036e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0304,  0.7714,  0.0312, -0.0083,  0.0216,  0.0158, -0.0036, -0.0038,
         0.0015, -0.0170,  0.0198, -0.0041, -0.0057,  0.0097,  0.0123,  0.0024,
         0.0008,  0.0137, -0.0267,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0203, -0.6662,  0.0268, -0.0040, -0.0112, -0.0152, -0.0100,  0.0021,
        -0.0308,  0.0086, -0.0151,  0.0441, -0.0719,  0.0248,  0.0490,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.5797e-02,  6.8617e-01, -4.4274e-02,  1.9555e-02, -4.4985e-04,
        -1.1586e-02, -3.8255e-02, -3.1844e-03,  1.6569e-02, -4.9833e-04,
         1.6351e-03, -1.7843e-02,  1.9732e-03, -1.7460e-03, -9.2574e-03,
        -8.7160e-03, -4.5156e-03, -3.7160e-03, -3.3538e-03, -1.1153e-02,
        -1.6874e-02,  3.2876e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.5338e-03,  7.4636e-01, -1.6369e-02,  1.1692e-02, -4.4123e-03,
        -9.7761e-03, -1.9814e-02, -7.4206e-03, -1.9864e-02,  8.1869e-03,
        -5.7609e-03,  1.3565e-02,  2.9305e-03,  2.5203e-03,  1.0793e-02,
         5.6997e-03, -3.6178e-03,  1.1341e-03, -2.8722e-04, -2.1927e-03,
         1.4228e-02,  5.5724e-04, -1.2896e-04,  1.4868e-02,  8.8575e-03,
        -6.9534e-03,  9.7007e-03, -9.3274e-04, -1.1765e-02,  3.3077e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0775,  0.6677, -0.0402, -0.0500,  0.0367,  0.0278,  0.0032,  0.0263,
        -0.0499,  0.0090, -0.0117,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.9591e-03,  8.2680e-01,  1.2510e-02,  1.2289e-02, -4.8220e-03,
        -1.5874e-02,  8.1442e-03, -4.3861e-03, -7.7818e-03,  2.0801e-03,
        -1.3915e-02,  4.5236e-03,  3.0915e-03, -5.9225e-04,  2.1967e-03,
         1.7065e-02, -3.8328e-03, -5.8312e-03, -5.9096e-03, -2.9977e-03,
         3.5410e-02,  3.9888e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6145e-03,  6.9259e-01,  4.7749e-02, -1.9115e-02,  8.5694e-03,
         1.0630e-02,  1.5588e-02,  4.2548e-03, -3.6579e-04,  4.2547e-03,
        -8.3304e-03, -8.1203e-03,  6.2833e-03, -8.9483e-03,  1.1349e-02,
         9.2218e-03, -2.8260e-03, -2.0016e-03,  4.0417e-03,  3.4184e-03,
        -3.7574e-03, -9.9826e-04, -1.2753e-02, -6.4912e-03,  1.2577e-02,
        -2.0672e-03,  8.7201e-03, -2.4809e-03, -1.9015e-02, -7.4005e-03,
        -2.4629e-04,  3.4850e-03,  5.0739e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.5970e-03, -7.0709e-01, -3.2376e-02,  5.5351e-02, -4.1024e-03,
        -1.6287e-02,  6.7465e-03,  7.3458e-03, -1.3267e-02,  1.1161e-02,
         5.0266e-03,  4.6020e-03, -4.8636e-04,  1.8345e-03,  1.2104e-02,
        -1.9041e-02, -2.6545e-03, -1.0259e-02,  1.6754e-03, -3.9139e-03,
        -6.4183e-03,  1.2209e-02, -7.9066e-03,  1.7907e-03,  5.3669e-03,
         1.4911e-03,  2.1486e-02,  2.1406e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0398, -0.7123, -0.0769, -0.0106, -0.0060,  0.0105, -0.0116,  0.0014,
         0.0170, -0.0026,  0.0028, -0.0028, -0.0082,  0.0008,  0.0024,  0.0104,
         0.0091,  0.0090,  0.0039, -0.0034,  0.0030, -0.0048,  0.0042,  0.0231,
         0.0234,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #550: [tensor([-3.5657e-02,  6.4873e-01,  3.0151e-02, -1.1854e-02, -1.6827e-02,
         8.4440e-03,  2.2898e-03, -2.2813e-02, -2.4311e-03, -6.9417e-03,
         2.9479e-03, -4.1785e-03,  5.8448e-03, -4.5533e-03, -4.9648e-03,
        -7.9106e-03,  3.6880e-03, -5.9866e-03,  3.0430e-03, -1.6069e-02,
         6.2579e-03, -5.3374e-04, -1.5930e-02, -1.6642e-02, -8.5129e-03,
         5.2877e-03, -6.5630e-03, -1.1698e-02, -4.1587e-03, -7.0026e-04,
        -9.9122e-04, -2.6350e-03, -9.1458e-04,  2.7174e-03, -1.7563e-03,
        -2.7294e-03, -2.6839e-03, -3.4445e-03, -2.1867e-03,  3.8923e-02,
         1.9406e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.6877e-02,  6.6914e-01,  4.6618e-03, -2.6913e-02,  1.3442e-03,
         4.1744e-03, -7.7445e-03,  3.5485e-03, -2.4209e-02, -6.0415e-03,
         1.2702e-02,  5.8296e-03, -6.2913e-03,  3.0173e-04,  2.5909e-03,
        -1.2630e-03, -1.3388e-02, -1.2395e-02, -1.8563e-03,  2.8961e-03,
        -3.6859e-03, -6.3286e-03,  1.7752e-04, -4.0115e-02, -5.5527e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1377,  0.5476, -0.0639, -0.0428,  0.0019, -0.0084, -0.0169, -0.0057,
        -0.0252, -0.0384, -0.0253,  0.0073, -0.0048, -0.0009, -0.0112, -0.0011,
         0.0042,  0.0168, -0.0175,  0.0007, -0.0215,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5704e-03,  7.5527e-01,  3.5761e-03, -7.9228e-03, -5.9887e-03,
         1.6987e-02,  2.8343e-02,  1.4636e-02, -5.4147e-03, -9.3892e-03,
         3.5036e-02,  9.9079e-04,  1.5519e-04, -3.8889e-03, -6.0122e-03,
        -1.6191e-02, -1.5942e-02, -7.2683e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0223e-02,  5.9386e-01,  4.9152e-02,  2.3747e-03,  3.0238e-02,
        -1.5409e-02, -1.0900e-02, -1.0924e-02, -3.9400e-02, -3.3650e-03,
         1.5237e-02,  7.5332e-03, -7.2308e-03,  4.5509e-03,  1.9188e-03,
         5.7347e-03, -1.6150e-02, -3.6842e-03,  1.0169e-02, -7.6606e-03,
        -1.2456e-02, -1.1123e-03, -5.4401e-03, -5.0624e-03,  5.7079e-03,
        -8.0926e-03, -6.1137e-03, -1.7008e-03, -1.3670e-02,  1.6799e-02,
        -4.5534e-03,  4.4370e-05, -6.4243e-03, -1.6892e-02, -1.0061e-02,
         1.2538e-02, -5.5814e-03,  1.2040e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0204,  0.7304,  0.0086,  0.0215, -0.0078,  0.0022,  0.0082, -0.0073,
        -0.0066,  0.0156, -0.0064, -0.0087, -0.0068, -0.0226,  0.0052,  0.0493,
        -0.0038,  0.0188, -0.0207,  0.0289,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0678, -0.6292, -0.0530,  0.0370, -0.0079, -0.0222,  0.0038, -0.0187,
         0.0023, -0.0136,  0.0539,  0.0363, -0.0323, -0.0221,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0153,  0.6557, -0.0174,  0.0233,  0.0060,  0.0149, -0.0136, -0.0058,
        -0.0065, -0.0198,  0.0074, -0.0250,  0.0108,  0.0072, -0.0023, -0.0331,
         0.0094, -0.0025, -0.0071, -0.0068, -0.0038,  0.0275,  0.0077,  0.0012,
         0.0154, -0.0121,  0.0051,  0.0073,  0.0115,  0.0009, -0.0038, -0.0138,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2963e-01,  6.1007e-01,  3.0427e-02,  4.8626e-03, -2.2182e-02,
         1.0409e-02,  2.1885e-02, -2.0529e-02,  7.4382e-03, -1.1828e-02,
         2.9726e-03,  4.1487e-03,  4.0454e-04,  5.3086e-03,  7.8281e-03,
         6.0874e-03, -3.9021e-03, -1.1555e-02, -1.1228e-03, -8.7412e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2301e-02,  5.5731e-01,  3.3260e-02, -4.0339e-02,  1.8295e-02,
        -1.2728e-02, -7.7261e-03, -8.8067e-03, -1.0607e-02, -1.6403e-02,
         4.3472e-04, -9.9382e-03, -2.3802e-03, -4.6775e-04,  6.8181e-04,
        -1.0853e-02, -8.0616e-03,  1.2380e-03,  3.7406e-03,  9.1663e-04,
        -1.0032e-02, -1.9913e-03,  7.1331e-03, -1.0547e-02, -1.5672e-02,
        -1.2404e-02, -1.5462e-02, -1.7759e-02, -1.7348e-02, -8.6673e-03,
         7.4741e-04, -6.6826e-04,  1.0459e-02,  2.8281e-03,  2.3042e-03,
        -5.4996e-05, -1.2341e-03, -4.0432e-03, -6.0293e-03, -9.6893e-03,
         9.0692e-03,  2.9826e-03, -7.7287e-03, -8.6081e-03, -1.8215e-02,
         7.5231e-03,  1.4314e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0293,  0.7608,  0.0123, -0.0161,  0.0282,  0.0098, -0.0138,  0.0060,
        -0.0480, -0.0127, -0.0106,  0.0053, -0.0083,  0.0091, -0.0296,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5941e-02,  3.5111e-01, -7.6468e-03, -3.3477e-02,  1.8432e-03,
         7.6422e-03,  1.1754e-02,  1.4964e-02, -1.6298e-02,  6.5751e-03,
        -2.0908e-03,  7.2374e-03, -5.4881e-03,  2.0812e-02,  1.4409e-02,
         4.4191e-04, -1.1300e-02, -1.2415e-02, -3.9102e-03,  1.4525e-03,
        -1.0663e-03,  4.7132e-04, -2.9030e-04,  2.4064e-03,  1.1115e-02,
        -2.8568e-02,  7.0759e-03,  1.3946e-02,  4.1930e-03,  7.9144e-03,
         8.8316e-03,  3.8133e-03,  5.8622e-03,  1.3407e-03, -3.1153e-03,
         1.5585e-03,  5.9982e-03,  1.5025e-02,  4.0025e-03,  4.0746e-02,
         2.3640e-02,  2.0672e-02, -2.9012e-03,  3.4513e-02, -3.8323e-03,
        -7.1372e-03, -9.1417e-02,  9.1739e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 0.1831,  0.3109,  0.0262,  0.0071, -0.0135, -0.0012, -0.0059, -0.0085,
        -0.0162, -0.0021,  0.0092, -0.0093,  0.0390,  0.0035, -0.0054,  0.0207,
        -0.0185,  0.0213,  0.0123, -0.0045, -0.0047, -0.0138, -0.0113,  0.0168,
         0.0635, -0.1713,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6229e-02,  7.0972e-01,  6.5085e-02, -2.5522e-02, -3.4529e-03,
         1.8620e-03,  1.8372e-02,  6.3233e-03, -2.0736e-02, -8.1987e-03,
        -4.6307e-03,  1.9962e-03, -9.2265e-04, -7.0709e-03, -1.6135e-02,
        -9.8842e-04,  2.3357e-03,  5.8672e-04,  4.9240e-02,  6.7284e-05,
         4.0523e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1883,  0.1965, -0.0179,  0.0168, -0.0071, -0.0575,  0.0226,  0.0157,
         0.0069, -0.0104,  0.0122,  0.0015, -0.0295,  0.0066, -0.0216, -0.0101,
         0.0322, -0.0449, -0.0022, -0.0375, -0.0037, -0.0086,  0.0426, -0.0229,
         0.0052,  0.0092,  0.0037,  0.0014, -0.0029,  0.0025, -0.0056,  0.0017,
         0.0299, -0.0096, -0.0344, -0.0779,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1070,  0.5473,  0.0166, -0.0174, -0.0029,  0.0402, -0.0927, -0.0089,
        -0.0073, -0.0008,  0.0006, -0.0084, -0.0197, -0.0332, -0.0342,  0.0039,
        -0.0104, -0.0183, -0.0270,  0.0031,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.1055e-04,  6.9208e-01,  4.6874e-02,  2.0140e-02,  1.3443e-02,
        -9.1789e-03,  5.0210e-04, -2.6751e-02, -4.3434e-03, -2.9562e-03,
         3.0807e-02, -2.5312e-03, -2.7589e-02, -5.1399e-03, -3.2741e-03,
        -5.0887e-03, -1.0942e-02, -8.1074e-03,  1.0831e-03,  3.6374e-03,
        -8.4620e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1041e-02,  7.9742e-01,  5.4334e-03, -9.2619e-03,  2.0675e-02,
         3.6357e-04, -1.5657e-02, -1.5030e-02,  3.5608e-02, -4.9654e-03,
         9.2613e-03,  1.9612e-02, -4.5669e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6464e-02,  6.0923e-01,  7.8025e-02,  2.6426e-03, -2.2591e-02,
        -7.1323e-03, -1.5511e-02, -5.6254e-04, -9.6236e-04,  5.2638e-03,
         4.5010e-03,  2.4748e-03, -4.1458e-03, -6.3067e-03,  1.3078e-02,
         1.1152e-02, -7.3237e-04, -1.2076e-03,  4.1117e-03,  2.1701e-03,
         3.7748e-03,  1.6708e-02,  1.9130e-03,  7.6161e-03,  2.1456e-03,
        -7.3507e-04, -2.0264e-02, -1.3657e-02,  9.8668e-04,  5.1184e-03,
         5.3751e-03, -2.1046e-03, -4.3401e-03, -2.2257e-04, -3.0778e-03,
        -5.9541e-03,  6.7742e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0329,  0.5853,  0.0673, -0.0186, -0.0330, -0.0155, -0.0162, -0.0138,
         0.0023, -0.0090, -0.0199, -0.0389, -0.0032, -0.0053, -0.0128,  0.0082,
         0.0188,  0.0029, -0.0071,  0.0112, -0.0096, -0.0092, -0.0352,  0.0238,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1269,  0.4306,  0.0082, -0.0283, -0.0105, -0.0062, -0.0201,  0.0079,
        -0.0189, -0.0114, -0.0093,  0.0381,  0.0051, -0.0171, -0.0042, -0.0140,
        -0.0400, -0.0396, -0.0179, -0.0114,  0.0028, -0.0254, -0.0076, -0.0123,
        -0.0155, -0.0073,  0.0070, -0.0040, -0.0041,  0.0029, -0.0407, -0.0047,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1643e-02,  6.0907e-01,  6.0065e-02, -1.3683e-02,  1.5778e-05,
         9.8696e-03, -1.4629e-02, -1.9773e-02,  5.8691e-03, -1.9894e-03,
        -6.9920e-04, -6.2032e-03, -8.4308e-03, -3.0531e-03, -1.0003e-02,
         3.8429e-03, -4.2188e-03, -9.3205e-03, -3.9133e-03, -1.4985e-03,
        -1.9757e-02, -9.6927e-03, -2.1358e-03, -2.0013e-03,  3.2541e-03,
        -1.2695e-02, -1.4176e-02, -5.7100e-03,  1.0360e-03, -3.1123e-02,
         1.0063e-01,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1093,  0.5542, -0.0579,  0.0168,  0.0176,  0.0194,  0.0099, -0.0291,
         0.0091,  0.0065,  0.0061,  0.0067, -0.0097, -0.0057, -0.0019,  0.0151,
        -0.0015,  0.0087,  0.0023, -0.0061, -0.0335,  0.0400,  0.0330,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1104,  0.0375,  0.0343,  0.0228, -0.0494, -0.0219,  0.0239, -0.0130,
         0.0245, -0.0505, -0.0224, -0.0037,  0.0077, -0.0073, -0.0428, -0.0164,
        -0.0394, -0.0364, -0.0083,  0.0031, -0.0060, -0.0136, -0.0213, -0.0167,
         0.0621, -0.0405, -0.2152,  0.0486,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #100: [tensor([-1.7811e-04,  7.9576e-01,  3.7573e-02,  5.8053e-03, -5.9672e-04,
         3.6381e-03,  9.0889e-03, -8.7271e-03, -1.5689e-03, -2.8920e-03,
         7.4397e-03,  4.9749e-03, -3.5598e-03, -9.0890e-03, -3.1732e-03,
        -1.6906e-02, -3.5224e-03, -3.0498e-03, -5.0555e-03, -1.5183e-02,
         8.7459e-03,  4.4558e-03, -6.9842e-03, -1.4733e-03,  2.3610e-03,
         1.0374e-03,  7.8137e-03, -1.1989e-02,  1.7362e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0181, -0.5361,  0.0395,  0.0005,  0.0231,  0.0639,  0.0098,  0.0410,
         0.0090,  0.0133,  0.0144,  0.0174,  0.0262,  0.0175,  0.0147, -0.0025,
        -0.0084,  0.0007,  0.0007,  0.0040,  0.0080, -0.0320,  0.0368, -0.0623,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9941e-02,  5.5816e-01,  6.2627e-03,  1.3294e-02,  3.4564e-02,
         1.3075e-02,  9.7590e-03,  3.4835e-02,  2.4539e-02, -1.4935e-02,
        -4.2265e-03,  2.4556e-02, -1.2814e-02, -1.3742e-02,  4.1321e-03,
        -5.3839e-03, -1.5591e-02,  1.5141e-03,  2.9893e-02, -5.5988e-03,
         1.2352e-03, -1.5997e-02,  3.8163e-04,  2.0274e-02,  7.9323e-03,
         5.1600e-04, -4.8506e-04,  3.4475e-03, -5.1891e-03,  4.3528e-03,
        -1.1768e-02,  2.4154e-02, -2.4035e-02, -2.3412e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3875e-02,  5.0585e-01, -2.2764e-01, -4.0952e-02,  8.5584e-03,
        -7.6454e-03,  9.0969e-03,  5.1741e-03,  6.7477e-03, -2.4567e-02,
        -1.3580e-02,  4.9520e-03, -1.0583e-02, -3.0716e-03, -2.3172e-02,
        -2.2240e-04,  8.0425e-04,  1.7866e-02, -2.6974e-03,  2.7992e-02,
        -2.8259e-02,  1.6691e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0194, -0.6380,  0.0639,  0.0081, -0.0189, -0.0054,  0.0041,  0.0053,
        -0.0068,  0.0164, -0.0247,  0.0199,  0.0059,  0.0038, -0.0017,  0.0093,
         0.1029, -0.0457,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0382e-03,  7.3273e-01,  4.3091e-02, -7.9254e-03, -1.0856e-02,
         3.4007e-02, -1.4016e-03, -7.3135e-03, -5.0206e-03, -1.7916e-03,
         2.5550e-03, -2.7175e-03, -1.3347e-02, -1.5463e-02,  8.0062e-03,
        -3.0235e-03, -3.9828e-03, -2.6398e-03, -1.1307e-02, -1.0869e-03,
         7.9613e-05,  3.1879e-02, -5.7738e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.1961e-02,  7.9557e-01,  7.3613e-02,  9.9140e-03,  1.6656e-05,
        -5.4904e-03, -1.2552e-02, -7.5854e-03,  6.2893e-03,  8.3354e-03,
         6.6224e-03,  2.3750e-03,  2.9353e-03,  6.7815e-04,  2.6044e-03,
        -3.3614e-03, -1.0100e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0133, -0.6627, -0.0042, -0.0200, -0.0458, -0.0463, -0.0007, -0.0276,
         0.0117, -0.0014, -0.0056, -0.0093, -0.0183, -0.0187, -0.0217,  0.0007,
         0.0080,  0.0839,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0447,  0.6960, -0.0160,  0.0123, -0.0251, -0.0024, -0.0094, -0.0319,
        -0.0113, -0.0058, -0.0118,  0.0043, -0.0093,  0.0008, -0.0051, -0.0225,
         0.0017, -0.0034,  0.0035,  0.0031,  0.0009,  0.0034, -0.0029,  0.0156,
        -0.0243, -0.0324,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3469e-03, -6.5442e-01, -5.9394e-02,  1.4881e-02, -2.5769e-02,
         1.3126e-03,  3.1534e-02, -1.4230e-03, -4.6734e-03,  1.3813e-03,
         1.5198e-02, -7.9491e-03,  1.7646e-03,  4.0586e-03,  2.0245e-02,
         1.2240e-02, -2.7331e-03,  1.5915e-02,  1.5912e-02,  5.8968e-03,
         1.2807e-02, -1.0315e-02,  1.3053e-02, -1.5153e-02,  3.2917e-04,
        -3.4792e-03, -4.9790e-03,  5.6966e-04, -1.0426e-02,  2.5961e-02,
        -4.8791e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0695,  0.6671,  0.0214,  0.0074,  0.0381,  0.0009,  0.0144,  0.0011,
         0.0182,  0.0118, -0.0249, -0.0106, -0.0081, -0.0061,  0.0214, -0.0309,
        -0.0429, -0.0051,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0949, -0.6192, -0.0102,  0.0648,  0.0890, -0.0404, -0.0412,  0.0403,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 1.3303e-02,  6.8526e-01,  2.6473e-02, -1.9549e-02, -5.9899e-03,
         2.6426e-02,  2.0691e-02,  4.0604e-03,  5.9868e-03,  1.7687e-03,
        -1.4035e-02, -5.4031e-03,  3.4130e-03, -8.9304e-03, -7.6617e-04,
         3.6043e-03, -5.4871e-03, -3.1565e-03,  3.8207e-03,  1.2355e-02,
         1.2244e-02,  1.3284e-02,  4.6296e-03,  2.0235e-03, -4.4336e-04,
        -1.1816e-03, -4.9589e-03,  1.1230e-03,  2.3699e-03, -7.2414e-03,
        -8.7733e-04,  1.3761e-03, -8.2918e-03, -8.3939e-03, -6.9290e-03,
         2.8836e-03,  4.5032e-03,  1.3744e-02, -3.3027e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2019e-03,  6.6571e-01, -7.2830e-03,  3.1544e-02,  1.2984e-02,
         1.7611e-02, -2.5995e-04,  8.1082e-03,  1.3486e-02, -9.9602e-03,
         3.9349e-03,  2.2554e-02,  2.2182e-02,  8.5694e-03,  1.1352e-02,
         1.2610e-02, -3.8651e-03, -1.4697e-03, -1.1093e-02, -3.3906e-02,
        -4.2603e-03,  1.2219e-02, -5.5175e-03,  1.5190e-03,  9.1510e-03,
        -1.7664e-02,  6.0797e-03, -4.1908e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1097, -0.6885, -0.0531, -0.0289, -0.0206, -0.0120,  0.0132, -0.0103,
        -0.0145, -0.0080,  0.0026, -0.0176,  0.0210,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1086, -0.4962,  0.0333,  0.0149,  0.0390,  0.0386,  0.0030,  0.0367,
        -0.0580,  0.0428,  0.0015,  0.0236,  0.0145,  0.0648, -0.0243,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3836e-02,  5.4280e-01,  1.3619e-03, -5.2874e-02,  5.8708e-03,
         4.9518e-02,  6.8960e-03,  2.4143e-02,  1.4668e-02, -2.1542e-02,
        -1.7866e-02, -1.6050e-02, -5.8936e-03,  2.6170e-02,  1.2711e-02,
         6.7985e-03,  4.6412e-03, -1.4832e-03,  1.1013e-03, -1.0792e-02,
         3.4556e-03, -6.3188e-03,  3.0722e-03,  1.1789e-02,  1.7997e-02,
         2.9564e-03, -1.0161e-02,  1.1273e-03, -4.6526e-03, -5.7926e-03,
        -2.0275e-03, -3.2138e-03, -4.8495e-03, -1.1429e-03,  2.0647e-02,
        -8.0120e-03,  4.1527e-03, -2.1733e-03,  2.2294e-03, -4.0081e-04,
         4.0398e-03, -3.6801e-03, -5.8177e-03, -2.5544e-03, -1.3036e-03,
        -7.4871e-03, -1.9290e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2072e-02,  5.3288e-01, -1.4271e-02, -1.6142e-03, -1.0404e-02,
         5.2813e-02, -1.2861e-02,  5.5550e-03,  1.4130e-02,  5.7937e-03,
        -7.3777e-03, -6.0923e-03, -6.0396e-03, -3.1717e-02,  3.5442e-03,
        -2.6880e-03, -3.6225e-03, -3.9085e-03, -2.1843e-02,  2.2189e-03,
        -2.9942e-03,  5.2884e-03,  7.6837e-04, -1.9645e-03,  9.8754e-03,
        -1.2986e-02, -3.4622e-03,  1.0522e-03, -8.2252e-03,  2.4646e-03,
         1.4879e-03, -3.3730e-03, -6.6359e-03, -2.0330e-02,  1.6777e-03,
         2.3403e-02, -6.2005e-05, -2.6612e-03, -7.5327e-03,  2.0864e-02,
         3.6290e-02,  6.5155e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.7234e-02,  6.0451e-01,  7.6204e-02,  1.8569e-02,  4.3288e-03,
         8.3889e-03, -3.4539e-03,  1.2673e-02, -4.9434e-03,  8.0771e-03,
         2.8382e-02, -3.9402e-02,  2.6373e-03,  1.6691e-03,  3.1608e-04,
         5.4100e-03, -8.0360e-03,  3.6491e-03,  1.0456e-02, -1.6461e-02,
        -1.2241e-02, -1.4171e-02, -6.4782e-03,  1.3138e-02, -9.1694e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1427,  0.5684,  0.0163,  0.0156, -0.0165, -0.0017, -0.0065, -0.0061,
         0.0066, -0.0112, -0.0144, -0.0117, -0.0106, -0.0371, -0.0410, -0.0154,
        -0.0065, -0.0008, -0.0152, -0.0182, -0.0184, -0.0151,  0.0041,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0250,  0.5960,  0.0495, -0.0355,  0.0157,  0.0253, -0.0089, -0.0293,
        -0.0203,  0.0147,  0.0274, -0.0152, -0.0038, -0.0155, -0.0052, -0.0086,
        -0.0062, -0.0068,  0.0039, -0.0064,  0.0033, -0.0419, -0.0358,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1413, -0.5747,  0.0559,  0.0517,  0.0218,  0.0113, -0.0358, -0.0021,
        -0.0184,  0.0128, -0.0084,  0.0316,  0.0344,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4830e-01, -6.0924e-01, -3.0575e-02,  4.6125e-02,  5.3285e-02,
        -2.8412e-03, -3.4347e-02,  2.3729e-04, -4.7528e-03,  4.2210e-03,
         1.3584e-02, -9.0152e-03,  3.9945e-02,  3.5358e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0127, -0.7108, -0.0976, -0.0230, -0.0271, -0.0122, -0.0147,  0.0022,
         0.0036, -0.0015, -0.0076,  0.0085,  0.0008, -0.0083, -0.0075, -0.0008,
         0.0397, -0.0136,  0.0078,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 0.0821,  0.5812, -0.0246,  0.0038,  0.0134,  0.0176, -0.0040, -0.0155,
         0.0212, -0.0036,  0.0078, -0.0092, -0.0047, -0.0036,  0.0044, -0.0348,
        -0.0070, -0.0058, -0.0041,  0.0183,  0.0299, -0.0043, -0.0032, -0.0060,
         0.0459, -0.0440,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0917,  0.6848,  0.0339,  0.0068,  0.0136, -0.0125, -0.0176, -0.0184,
        -0.0062, -0.0244, -0.0027, -0.0116, -0.0160,  0.0167, -0.0157, -0.0273,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0870,  0.5735, -0.0245,  0.0591, -0.0045,  0.0273,  0.0295, -0.0039,
         0.0100, -0.0124, -0.0239, -0.0571,  0.0065, -0.0313,  0.0022, -0.0146,
         0.0327,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0049,  0.7643, -0.0316, -0.0479, -0.0050, -0.0291, -0.0011, -0.0584,
         0.0577,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.0244e-02, -3.8213e-01,  1.6617e-02,  1.7881e-02, -3.5328e-02,
        -5.4965e-03, -2.3597e-03, -1.6812e-02,  1.5559e-02, -5.0365e-03,
         2.8135e-02, -1.7356e-03,  1.4322e-02,  9.1397e-03,  3.0853e-02,
         9.3069e-03,  1.0986e-02,  1.5871e-02,  5.3021e-03, -3.6226e-04,
        -3.2486e-03,  1.6945e-02,  5.4291e-03, -1.3208e-04,  3.4010e-03,
         6.6408e-03, -1.5796e-02,  1.8877e-02, -6.1862e-03,  1.4660e-02,
         1.0939e-02,  3.9016e-05,  1.8490e-03,  1.1604e-02, -3.6050e-03,
         2.1148e-02,  1.0031e-02, -1.8685e-03,  7.5473e-03,  9.4965e-03,
         7.7177e-03, -8.0779e-03,  1.3136e-02,  1.9073e-02,  8.2091e-03,
        -7.0873e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2105e-02,  6.1060e-01, -3.8693e-02,  1.5141e-02,  1.1601e-02,
         1.9447e-02,  2.1133e-02, -2.6826e-02, -1.8270e-02,  1.0028e-02,
        -1.4127e-02, -1.2613e-02, -6.6895e-03, -1.4228e-02,  2.4272e-05,
         2.6313e-02, -1.5105e-03, -1.7437e-02, -8.7251e-03,  1.2099e-02,
        -3.4409e-03, -1.1467e-02, -1.2104e-02, -6.3573e-03, -1.2796e-04,
        -2.3405e-03, -1.4880e-03,  1.6882e-03,  7.9414e-03,  4.5437e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0032e-01, -7.6505e-01,  1.3541e-04,  1.5518e-02,  2.0453e-02,
        -1.7435e-02, -8.1085e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0383,  0.7268, -0.0177, -0.0263,  0.0109, -0.0309, -0.1001,  0.0182,
         0.0225, -0.0018,  0.0065,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0331,  0.6290,  0.0731,  0.0722,  0.0214, -0.0166, -0.0089,  0.0054,
         0.0058, -0.0149, -0.0012,  0.0053, -0.1055,  0.0074,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0158,  0.5967,  0.0237,  0.0186,  0.0239,  0.0063, -0.0017,  0.0043,
        -0.0012, -0.0128, -0.0011, -0.0113, -0.0180, -0.0274, -0.0082,  0.0017,
        -0.0140, -0.0056, -0.0060,  0.0077, -0.0235, -0.0047, -0.0067,  0.0076,
        -0.0115, -0.0041, -0.0087, -0.0047, -0.0121, -0.0036, -0.0007, -0.0158,
        -0.0455,  0.0448,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.5058e-02,  5.5174e-01, -2.6976e-02,  4.6659e-02,  2.9049e-02,
         8.0943e-03,  1.8026e-02,  3.7278e-03,  6.9929e-03,  1.6345e-02,
         2.5645e-03, -1.9935e-03, -5.6438e-03,  5.6208e-04, -2.4944e-02,
        -3.0063e-03, -1.6619e-02, -3.9908e-02, -1.5249e-02,  3.9072e-04,
         1.3179e-02,  2.1499e-04,  5.6842e-03, -2.8266e-03,  3.1499e-03,
         2.7387e-03,  4.2100e-02, -1.6561e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6938e-01,  4.5305e-01,  5.6726e-02, -6.9462e-03, -2.0004e-03,
        -2.2656e-02, -1.0173e-02, -7.2239e-03, -6.4858e-03, -1.4784e-02,
        -5.8640e-03, -6.7833e-03,  5.4125e-03,  6.9428e-03,  1.7689e-02,
        -1.4817e-02, -1.0175e-02, -1.3812e-02,  7.7993e-03, -1.0308e-04,
         6.4420e-03, -7.3962e-03, -8.0570e-02,  6.6776e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 0.0451,  0.6384, -0.0473,  0.0047, -0.0280, -0.0074, -0.0070, -0.0066,
        -0.0064,  0.0099, -0.0104, -0.0174, -0.0018,  0.0118, -0.0068,  0.0104,
         0.0032, -0.0039, -0.0029,  0.0060,  0.0060,  0.0074, -0.0079,  0.0039,
        -0.0049,  0.0063, -0.0115, -0.0047,  0.0047, -0.0358, -0.0315,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0323, -0.7677,  0.0054,  0.0539,  0.0185, -0.0296, -0.0086, -0.0020,
        -0.0057,  0.0196, -0.0248,  0.0319,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0873, -0.6760,  0.0064,  0.0112, -0.0258, -0.0178, -0.0074,  0.0011,
        -0.0058, -0.0054,  0.0106,  0.0247, -0.0205,  0.0617,  0.0384,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4976e-02,  7.1498e-01, -4.1058e-02, -8.2693e-03, -8.6633e-05,
        -4.6619e-03,  1.0215e-02,  2.0859e-02,  3.1341e-03, -6.2127e-03,
        -1.0597e-02,  9.7681e-03, -2.1855e-03, -9.6183e-03, -3.4533e-03,
         1.0357e-02,  4.2907e-03,  1.1790e-03, -8.6006e-03,  4.2898e-03,
         4.4967e-04, -1.5835e-03, -2.1584e-03, -1.7527e-03,  4.0926e-03,
        -2.4824e-03, -6.2728e-03,  9.9063e-04,  5.7650e-03,  1.7410e-03,
         1.5484e-02,  1.3270e-02,  2.5162e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0176, -0.7734,  0.0116,  0.0205,  0.0103,  0.0129, -0.0161, -0.0016,
        -0.0087,  0.0079,  0.0190,  0.0027,  0.0185,  0.0527, -0.0161,  0.0104,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0273,  0.6469, -0.0996,  0.0210,  0.0370,  0.0503,  0.0307,  0.0347,
        -0.0524,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([0.0360, 0.6235, 0.1900, 0.0772, 0.0732, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7659e-02, -7.1074e-01, -9.0159e-02,  2.2262e-03,  4.5178e-03,
        -2.0458e-02,  1.2857e-02, -2.1994e-03,  5.3928e-03,  1.3654e-02,
         1.4568e-02, -8.0459e-03,  1.3890e-02,  1.4318e-02,  1.8008e-03,
        -2.6276e-03,  4.3563e-03, -3.3514e-03, -1.5204e-03,  8.6496e-04,
         3.6249e-04,  9.5158e-04, -3.9022e-03, -3.4202e-03, -1.0539e-02,
        -5.1860e-03, -2.0434e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0724,  0.7282,  0.0101,  0.0131, -0.0922,  0.0227, -0.0612,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2667e-02, -6.9424e-01,  3.3586e-03,  6.3739e-03, -4.7300e-03,
        -4.4449e-03, -2.3994e-02,  4.7775e-03,  1.2489e-02,  3.5385e-02,
        -9.3549e-04,  1.4481e-02,  5.3972e-05,  5.5052e-03,  1.4733e-02,
         1.3187e-02,  2.7772e-02, -4.2185e-03,  9.2394e-03,  4.2608e-02,
         4.3966e-02, -2.0842e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0074,  0.7570,  0.0370, -0.0057, -0.0280, -0.0178, -0.0054, -0.0066,
        -0.0124, -0.0262, -0.0029, -0.0211, -0.0292, -0.0208, -0.0069, -0.0046,
        -0.0018, -0.0037, -0.0055,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3971e-02,  6.4477e-01,  3.3147e-02, -4.2983e-02,  8.9651e-03,
        -3.5391e-03, -4.7760e-02,  1.3843e-02, -7.8172e-03, -9.2762e-03,
        -9.2239e-03, -5.2529e-03, -6.6684e-03,  6.3135e-03, -5.3374e-04,
        -1.3548e-02, -4.9846e-03,  2.4856e-03, -2.7012e-03,  9.4652e-03,
         1.5639e-03,  1.0788e-02,  1.3590e-04,  2.9910e-03, -5.1545e-03,
        -2.4487e-03,  1.6619e-03, -7.9122e-03,  3.3014e-03, -2.6566e-02,
        -1.0230e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-3.2067e-02, -6.3371e-01, -8.0963e-02,  3.9333e-02,  3.6998e-03,
         9.1630e-03, -2.7335e-03,  3.4196e-03, -3.0157e-04, -3.4156e-03,
        -3.3492e-03, -5.7965e-03,  1.0846e-02,  8.6542e-03, -5.8642e-03,
        -4.3815e-04,  2.2674e-02, -3.4587e-03,  1.3555e-02, -4.4581e-03,
        -4.4004e-03,  1.5172e-03,  1.0764e-03,  1.3042e-02, -5.5730e-03,
         5.3855e-03,  2.6102e-03,  1.4050e-02,  1.1346e-02,  2.1288e-02,
        -1.3704e-02,  1.8108e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0507, -0.6810,  0.0407, -0.0071,  0.0122,  0.0164,  0.0380, -0.0162,
        -0.0166,  0.0032,  0.0024,  0.0086,  0.0035,  0.0048,  0.0137,  0.0197,
         0.0009, -0.0020,  0.0022, -0.0180,  0.0069, -0.0352,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.0968e-03,  6.1983e-01,  4.4633e-02, -1.9935e-02, -8.4393e-03,
         4.2764e-03, -4.1443e-03,  2.4860e-02, -1.1223e-03, -1.0922e-02,
        -2.2803e-03, -2.4416e-03, -3.5039e-02, -2.2923e-02, -5.7265e-03,
        -3.9183e-03,  2.0048e-03,  3.4655e-03, -7.7231e-03, -6.6538e-03,
        -1.2655e-02, -8.1736e-03,  4.5672e-04,  4.1629e-04, -6.0404e-03,
         1.5693e-02, -1.7694e-03, -3.2050e-02,  7.7003e-03, -7.7615e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0087e-02,  6.6518e-01, -1.1083e-01, -1.3784e-02, -5.1928e-03,
        -1.7613e-02,  3.0117e-02, -1.0076e-02,  8.8652e-03, -7.9268e-04,
        -2.6762e-02,  4.3703e-03,  6.8404e-04,  1.0125e-02,  1.7193e-02,
        -4.2608e-02, -1.5080e-02,  6.4222e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2079e-03,  6.4632e-01,  6.5153e-02,  1.1042e-02,  1.3247e-03,
        -7.3035e-03,  2.2547e-03, -9.5340e-03, -1.8869e-02,  7.6121e-04,
        -8.3063e-03, -6.7237e-03, -9.6149e-03, -1.4233e-03,  3.0043e-03,
         1.8895e-02,  6.7408e-03, -1.7187e-02, -9.5040e-03, -3.0797e-03,
        -7.0571e-03, -1.9316e-02, -7.7378e-03, -6.5070e-03, -7.5434e-03,
         3.0911e-04, -1.0744e-03, -1.9048e-03,  7.1857e-04,  4.2942e-03,
        -4.2415e-03, -6.4729e-04, -2.8440e-03, -9.9987e-03,  1.6306e-03,
         8.8226e-03,  6.7104e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0644,  0.5938, -0.1018, -0.0542, -0.0197,  0.0013, -0.0074, -0.0041,
        -0.0085,  0.0031, -0.0065, -0.0039, -0.0023,  0.0238, -0.0046,  0.0007,
        -0.0092,  0.0252, -0.0028,  0.0071, -0.0011,  0.0076, -0.0456,  0.0014,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0055e-02, -5.7213e-01, -2.2827e-03, -2.9425e-02, -1.5393e-02,
         4.7457e-03,  6.2229e-03, -5.5672e-03,  1.2554e-02,  1.3654e-02,
        -3.4393e-03,  1.8577e-02,  3.6971e-02, -6.3578e-03, -2.5256e-03,
        -2.9178e-03,  1.4475e-03,  3.3076e-03,  1.0493e-03,  7.1189e-03,
         2.1891e-03, -1.2535e-03,  2.7255e-03,  4.9570e-03, -9.1459e-03,
        -8.6355e-03, -9.0193e-04, -4.3946e-03, -4.7935e-03,  5.6987e-04,
         1.9184e-02,  3.4599e-03,  3.3455e-04, -4.0432e-03, -4.1315e-03,
         1.4385e-02,  3.8227e-03,  2.1011e-03, -3.8129e-03, -6.7916e-04,
         1.0021e-02,  5.2115e-03,  4.7778e-03, -5.5220e-03, -1.8221e-03,
         9.1388e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1011e-02,  6.9326e-01,  2.9983e-02, -1.4805e-02, -6.6021e-02,
         2.8797e-03, -7.2237e-06, -1.7678e-02,  1.4135e-03, -2.6313e-02,
         1.3782e-02,  1.6318e-02,  3.1695e-03, -1.0843e-02,  5.6922e-04,
        -3.9935e-03,  1.6248e-03,  1.4189e-03, -3.7139e-03, -3.6341e-03,
         2.4141e-02, -2.5524e-02,  1.7895e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0056, -0.6445,  0.0386, -0.0944, -0.0196, -0.0391, -0.0317,  0.0322,
         0.0108, -0.0308,  0.0527,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.3194e-02,  7.5510e-01,  5.1009e-02,  1.3111e-02, -1.2774e-02,
        -1.2566e-02, -1.2484e-02, -2.1970e-03,  9.8195e-04,  4.0846e-05,
         1.7317e-02,  5.1331e-03, -2.1324e-02,  5.2771e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0971,  0.6805,  0.0270, -0.0022, -0.0144,  0.0123, -0.0028,  0.0071,
        -0.0182, -0.0278,  0.0662, -0.0445,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4944e-03,  6.1106e-01,  8.3097e-02,  4.0351e-02,  1.7830e-02,
        -1.2889e-02,  1.1993e-02,  1.0445e-02,  4.6201e-03, -5.8055e-03,
        -5.0118e-03, -1.1911e-03,  6.3701e-04, -4.7689e-03,  3.1957e-03,
        -5.9354e-03,  2.5111e-03, -4.1132e-03,  2.6824e-03, -1.0922e-02,
         1.2646e-04,  5.2508e-07,  2.4832e-03, -2.1326e-03, -2.2269e-03,
         1.6973e-02,  1.4766e-02,  1.4986e-04, -7.5511e-03,  4.9643e-05,
         4.5834e-03,  5.5085e-03, -4.2616e-03, -3.2021e-03,  4.0721e-03,
        -4.3281e-05,  8.7054e-04, -4.2164e-03,  2.6389e-03,  2.9125e-03,
         1.9136e-04, -2.2941e-03,  2.9509e-03, -4.2292e-03, -1.1697e-02,
        -4.4604e-03,  2.2446e-03, -5.8630e-03, -2.2753e-03, -5.8613e-03,
         5.0565e-03,  6.5342e-03,  2.7923e-03,  9.4748e-03,  1.3749e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 1.8690e-02,  6.5740e-01,  9.9937e-02,  1.3104e-02, -2.4254e-02,
         1.6376e-02, -1.5409e-03, -2.0622e-02,  7.9460e-03,  1.0126e-04,
        -1.3679e-02,  1.9928e-03, -8.6625e-04,  2.4695e-02, -2.7306e-03,
        -3.2073e-03, -2.6681e-02,  1.0178e-02,  2.6602e-03,  4.4133e-03,
        -9.4847e-04, -4.3484e-02, -4.4963e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0697,  0.6521,  0.0345,  0.0418,  0.0428,  0.0020, -0.0049,  0.0079,
        -0.0082,  0.0312,  0.0065, -0.0208, -0.0299, -0.0016,  0.0063,  0.0160,
        -0.0174,  0.0064,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0510,  0.4821,  0.1135, -0.0179, -0.0652,  0.0044,  0.0868,  0.0117,
        -0.0090,  0.0054, -0.0043,  0.0096, -0.0082,  0.0029, -0.0031, -0.0490,
        -0.0233, -0.0527,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0433,  0.5965,  0.0219,  0.0214, -0.0104, -0.0120, -0.0175, -0.0346,
        -0.0594, -0.0035, -0.0171,  0.0158,  0.0152,  0.0210, -0.1106,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9335e-02,  7.0512e-01,  2.0816e-02,  4.0726e-02,  4.3949e-03,
         2.7213e-02, -9.3424e-03,  7.0351e-03,  4.5655e-03, -2.0764e-03,
        -1.3408e-02, -1.9786e-03,  6.3062e-03, -9.1894e-03,  9.8454e-03,
         3.8680e-03,  2.9810e-03,  1.4171e-02, -3.6512e-03,  9.2420e-03,
         4.1614e-04, -1.0124e-02, -6.5550e-04,  5.7866e-03, -8.4046e-03,
        -1.8575e-03,  8.0980e-03,  9.8078e-04, -1.3867e-02,  2.4571e-02,
        -9.3989e-03, -5.7622e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.2633e-02, -6.5823e-01, -5.3924e-02, -1.2858e-02, -5.0517e-03,
         1.1762e-03,  1.9518e-04, -4.6206e-03,  6.4217e-03, -3.7681e-03,
         1.1352e-02,  2.9694e-03, -3.3955e-03,  3.2374e-03,  6.8025e-03,
         4.1295e-03,  6.1765e-03,  2.3681e-02,  2.2310e-03,  5.1031e-03,
         2.9743e-03,  8.6906e-03, -1.5941e-05,  1.0286e-02, -2.0324e-03,
         9.7852e-03,  7.9638e-03, -2.1813e-02, -7.1569e-03,  1.6295e-02,
         4.3204e-03,  5.7172e-03, -3.4990e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6924e-02,  5.8882e-01,  4.1241e-02,  5.1775e-02,  3.8122e-02,
         1.1995e-02, -2.9217e-03, -6.0556e-03, -1.2224e-02,  4.3561e-03,
        -4.7339e-03, -1.7585e-02, -1.9999e-03, -3.6213e-03,  6.4619e-03,
        -4.5934e-03,  6.1122e-03,  4.9953e-03,  2.8121e-03, -1.0679e-02,
        -4.2021e-03, -6.5163e-03, -1.6564e-03, -1.0266e-02,  2.5428e-03,
        -3.0017e-03,  2.5259e-03,  4.5799e-04, -2.9195e-03, -9.0156e-03,
         5.8054e-04, -8.8463e-03, -3.0357e-03,  1.6816e-03,  2.7924e-03,
        -4.7640e-03, -1.7418e-03, -6.0700e-03,  4.2466e-04, -2.2113e-03,
        -5.4754e-04, -1.6690e-03, -4.7339e-03, -4.8364e-03,  2.6477e-03,
         5.9619e-04, -4.0701e-03,  2.4170e-03, -4.8810e-04,  7.5348e-03,
        -3.2104e-03, -5.1263e-04,  4.9036e-03,  1.0496e-03,  3.2405e-04,
         5.4240e-03,  7.4507e-03,  1.4301e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5205e-02,  6.1164e-01, -5.0450e-03,  4.9908e-02,  6.8541e-03,
         4.9188e-03,  4.1524e-03, -1.0651e-02, -4.3786e-04, -3.7593e-03,
        -1.0300e-02, -8.0293e-03,  2.6922e-02, -1.1440e-02, -8.7040e-03,
        -3.4751e-03,  3.3904e-03, -7.0792e-03, -7.9021e-03, -7.4454e-03,
        -1.9721e-02, -6.5809e-03, -2.1769e-02,  2.7105e-03, -4.6385e-02,
         5.5569e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0758,  0.4339, -0.0369, -0.0050, -0.0380,  0.0267, -0.0445,  0.0052,
        -0.0824, -0.0015, -0.0005, -0.0013,  0.0043, -0.0027, -0.0320, -0.0077,
        -0.0011,  0.0049, -0.0198,  0.0036, -0.0114, -0.0024,  0.0047,  0.1525,
         0.0013,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1714e-02,  5.8880e-01,  8.4741e-02,  5.1119e-02,  3.4721e-02,
         1.2189e-02,  1.5896e-03, -5.9905e-03,  5.9811e-03, -2.7051e-03,
        -1.9733e-02,  5.9157e-03,  4.3400e-03, -1.1327e-02,  9.9297e-03,
         6.0953e-03, -2.5519e-03, -8.8625e-03, -3.6342e-03, -1.0425e-02,
         1.3425e-04,  7.9852e-04, -7.6884e-03,  1.9713e-04, -3.3778e-03,
        -1.2947e-02,  4.1897e-03, -4.1538e-04, -1.4679e-03,  6.7048e-03,
         1.3533e-03, -6.3239e-03, -4.8583e-04,  6.6509e-04,  5.6157e-03,
        -5.6040e-03,  2.8735e-03,  7.5041e-04,  1.1899e-03,  3.2165e-03,
        -7.4486e-03, -5.1727e-03, -5.7441e-04,  1.0163e-02, -1.8275e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0285,  0.7269,  0.0167,  0.0172, -0.0280, -0.0010,  0.0131,  0.0149,
        -0.0163, -0.0176, -0.0074,  0.0072, -0.0033, -0.0042, -0.0050,  0.0204,
        -0.0186, -0.0163, -0.0019, -0.0025, -0.0133,  0.0150, -0.0047,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0880,  0.4830,  0.0223, -0.1284, -0.0522,  0.0327,  0.0295,  0.0070,
        -0.0059, -0.0163, -0.0804,  0.0545,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 1.1115e-01,  5.3479e-01,  3.1597e-02, -7.2878e-03, -6.3882e-02,
         8.4744e-03, -2.8979e-02,  2.4280e-04,  7.8476e-03, -1.7215e-02,
         2.0960e-02,  9.5308e-03,  1.5725e-03,  1.3250e-02,  9.6352e-03,
        -8.1824e-03, -1.2852e-03, -6.2784e-03, -6.0207e-03,  1.3919e-02,
        -2.4860e-02, -1.7118e-02, -5.5928e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0262e-02,  7.6614e-01,  5.9167e-03,  2.0419e-02,  1.4926e-02,
        -1.0510e-03, -3.7006e-03,  8.1016e-03,  1.0003e-02, -1.2218e-02,
        -1.9591e-03,  6.2246e-03, -6.5711e-03, -5.1666e-03, -1.4498e-02,
        -3.0885e-03, -6.9853e-04, -1.8136e-02,  3.2314e-03, -1.1455e-04,
         5.3380e-03,  4.5416e-03, -2.7522e-02, -3.9429e-03,  3.1356e-03,
         3.4417e-03,  2.4801e-03, -1.2569e-02,  5.5438e-03, -3.4642e-04,
         1.7581e-03, -1.2807e-02,  4.1520e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0427,  0.7571,  0.0702,  0.0021,  0.0077,  0.0073, -0.0044,  0.0051,
        -0.0168, -0.0077, -0.0087, -0.0011, -0.0083, -0.0118,  0.0021, -0.0137,
        -0.0038, -0.0019,  0.0276,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0476, -0.7278,  0.0098, -0.0592,  0.0378, -0.0125, -0.0028,  0.0202,
         0.0078, -0.0464, -0.0029,  0.0052,  0.0161,  0.0038,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1488e-02, -6.9642e-01, -1.7843e-02,  1.3400e-03, -4.7174e-02,
        -3.4666e-03,  1.2035e-02,  1.2632e-02, -2.4187e-02, -1.5931e-03,
         2.1017e-03,  5.9974e-04, -1.6260e-03,  3.0075e-02, -4.9252e-03,
        -8.3813e-04,  1.1859e-02,  1.0428e-03, -2.4181e-03,  1.1303e-02,
        -2.3199e-02,  4.1837e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.5881e-03,  6.5850e-01,  5.7209e-02,  5.3518e-02, -6.6877e-03,
         3.1585e-02,  3.8660e-03,  2.3368e-02, -6.0953e-03, -1.7377e-02,
        -9.0078e-03, -1.6035e-02, -1.1101e-03, -5.1626e-03, -6.2712e-03,
         1.2523e-04,  6.1640e-03, -1.5123e-02, -2.5695e-03,  5.5745e-03,
        -5.5870e-03, -1.1869e-02, -1.8293e-03, -1.6919e-02, -1.0470e-02,
         1.1639e-03,  3.0370e-03,  6.9292e-03, -7.8249e-03, -2.4307e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.4435e-03,  7.0695e-01, -1.2777e-02, -1.9940e-02, -2.7908e-02,
         3.1262e-02,  3.3996e-04,  7.5606e-03, -2.3012e-02, -3.0671e-02,
        -1.0225e-02,  1.1748e-02, -2.0481e-02, -1.5241e-02, -3.0237e-03,
         5.1070e-03,  6.4309e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0889,  0.6599,  0.0676,  0.0330, -0.0397,  0.0222, -0.0011,  0.0100,
        -0.0059, -0.0108, -0.0611,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8179e-04,  7.3620e-01,  5.2659e-02, -2.4680e-02, -4.4441e-02,
        -1.9061e-02, -7.4354e-03, -6.7579e-03, -7.5483e-03,  2.7415e-03,
         1.1313e-02,  7.6759e-03, -2.6798e-02,  1.5762e-03,  2.5988e-03,
        -2.2732e-02,  2.5500e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9882e-02,  4.3189e-01,  3.2160e-01, -4.9951e-02, -9.1465e-03,
         2.4236e-03,  1.3100e-02, -1.4143e-02,  8.4400e-03, -6.0391e-03,
        -1.0287e-02,  2.1516e-02, -7.0983e-03,  2.8238e-03, -3.3890e-03,
         2.9797e-03, -4.5102e-03, -9.0397e-03, -1.7450e-02,  3.2445e-04,
         5.7730e-03,  3.7953e-03,  1.3981e-02, -4.3058e-04,  7.2880e-03,
         1.2696e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0688,  0.6585,  0.0429,  0.0019,  0.0115, -0.0022, -0.0076,  0.0025,
        -0.0195, -0.0022, -0.0061, -0.0101, -0.0071,  0.0036,  0.0017, -0.0062,
        -0.0070,  0.0050, -0.0052, -0.0033, -0.0073, -0.0092, -0.0048, -0.0075,
         0.0345, -0.0076, -0.0024,  0.0132, -0.0121, -0.0285,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0319, -0.7686,  0.0052,  0.0159, -0.0079,  0.0113,  0.0075, -0.0110,
         0.0087, -0.0107,  0.0117, -0.0025, -0.0146, -0.0060,  0.0187,  0.0038,
         0.0189, -0.0010,  0.0441,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([-5.8575e-02,  5.6555e-01,  1.4491e-01,  2.8958e-02,  3.4326e-02,
        -5.9422e-03,  3.6450e-02,  1.8652e-03, -2.6089e-03,  3.0475e-03,
        -5.0867e-04, -7.7047e-03, -2.6241e-04,  1.1665e-02, -9.1221e-03,
        -8.4578e-03,  1.2361e-02,  7.2350e-03, -5.5181e-03,  1.2123e-02,
         2.5202e-03, -4.9672e-03,  3.4247e-03, -4.1960e-03, -9.8629e-03,
        -2.3168e-03,  1.5528e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1111, -0.6406, -0.0476,  0.0182,  0.0369,  0.0164, -0.0102,  0.0159,
        -0.0046, -0.0062,  0.0112, -0.0120, -0.0616,  0.0077,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.6246e-03,  7.0405e-01,  1.8008e-02,  2.5878e-02,  1.0181e-02,
         1.8412e-02,  5.4564e-03,  1.4140e-02,  1.6763e-02, -3.5978e-03,
         4.3573e-02, -8.0131e-03, -6.7839e-03,  2.4305e-03, -6.7874e-03,
        -5.4909e-03, -3.7955e-04, -3.2772e-03, -2.1194e-02,  8.4603e-03,
        -8.9956e-03, -9.5681e-03, -2.2588e-02, -3.7191e-03,  2.3210e-02,
        -4.4160e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1055,  0.6154,  0.0121, -0.0106,  0.0234, -0.0310,  0.0169, -0.0040,
         0.0248, -0.0206,  0.0442,  0.0192,  0.0248, -0.0475,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5624e-02,  6.6731e-01,  2.2931e-02,  1.9066e-02, -5.3313e-03,
         2.3823e-02,  1.0598e-02,  1.2029e-02,  2.6104e-03, -6.1296e-03,
         3.8849e-03, -4.2895e-03,  4.7063e-03,  2.9953e-02,  2.5936e-03,
         3.9115e-04,  5.9560e-03, -2.1064e-03, -6.4817e-03, -2.5828e-03,
        -1.8632e-03,  1.7715e-03,  7.7781e-03, -2.6725e-04, -1.4768e-02,
        -6.5479e-03, -6.9130e-03,  1.8074e-03,  1.3828e-02,  1.0490e-02,
         4.3194e-03, -4.1244e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6629e-02,  6.3331e-01,  3.2410e-02,  5.8523e-02, -1.6002e-02,
         2.6741e-02, -2.2224e-02, -3.0655e-03, -2.9313e-03, -1.0962e-02,
        -7.3149e-03, -8.9090e-04, -6.8309e-03, -8.0043e-04, -1.3489e-02,
        -3.1885e-04, -8.0703e-03,  8.1706e-03, -4.0354e-03, -7.3237e-03,
        -4.7289e-03, -4.1543e-03,  9.5261e-04, -5.9101e-02, -1.1020e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0039,  0.7433,  0.1300,  0.0482,  0.0259, -0.0486,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9434e-02,  6.9403e-01,  3.2101e-02,  9.4961e-03,  2.0255e-02,
         3.2112e-03,  1.8795e-03,  1.6866e-03,  3.5255e-04,  5.8151e-03,
         3.6006e-03,  2.5747e-02,  1.2678e-02,  6.7138e-03,  4.2984e-03,
        -3.2107e-03, -1.8081e-02, -1.1549e-03, -9.4552e-04, -7.6881e-04,
        -2.7103e-03, -2.3254e-03,  5.0448e-03,  3.3707e-04, -3.6579e-03,
        -8.5026e-03,  3.5868e-03, -1.7241e-03, -3.8717e-03, -4.7632e-03,
         1.6112e-02,  2.3778e-03, -1.2335e-03,  5.8154e-04, -4.0526e-03,
         4.4541e-02, -1.9116e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6413e-02, -6.2913e-01, -3.4827e-02, -2.4456e-03, -2.7750e-02,
        -5.8173e-03, -9.8618e-03,  7.9614e-03, -3.6545e-03,  1.2782e-02,
         4.5044e-03,  4.7607e-04, -1.1035e-02,  2.7783e-02,  1.8587e-02,
         4.9308e-03,  2.3664e-02,  7.5445e-04,  2.2832e-03,  6.9421e-03,
         1.9214e-02, -3.0686e-03,  6.7720e-03,  2.9055e-03,  2.1105e-03,
         5.4662e-03, -7.8492e-03,  8.4453e-03,  1.6912e-02,  1.0384e-02,
        -4.4283e-04,  2.2668e-03, -1.2282e-02,  1.4894e-02, -5.3855e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1113,  0.4557, -0.0036,  0.0071, -0.0082,  0.0131,  0.0140,  0.0313,
         0.0259,  0.0133, -0.0243, -0.0519, -0.0326,  0.1029, -0.1047,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4075e-02, -6.7289e-01, -1.8870e-02,  2.9600e-02,  1.3006e-02,
        -1.2119e-02,  3.1275e-03,  8.0893e-03, -2.8495e-03,  6.7830e-03,
        -1.9463e-03,  9.7787e-03,  6.1559e-03,  2.1724e-05, -3.5789e-03,
        -5.9931e-03,  2.1307e-03, -5.7748e-03,  7.0685e-03,  1.5359e-02,
         4.8200e-03, -7.0837e-04,  6.3079e-03,  9.1660e-04,  8.0077e-03,
         4.5610e-03,  1.1004e-02,  4.8414e-03,  4.7920e-03, -3.6601e-03,
         4.4928e-03,  3.8777e-02,  2.7897e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7499e-01,  5.6006e-01,  4.5021e-02, -3.6405e-03, -1.0045e-02,
        -2.3017e-02, -2.8847e-03, -3.8980e-04, -8.0200e-03,  2.0597e-03,
        -5.5193e-04,  2.9850e-03, -5.1596e-03, -8.8170e-03, -7.1061e-03,
        -3.1453e-03, -4.2368e-03, -3.7864e-04, -2.6414e-02,  6.7545e-03,
        -9.4772e-03, -7.5664e-03, -7.9714e-03,  2.9764e-03,  8.6599e-04,
        -2.5946e-03, -1.4616e-02, -4.0611e-02, -1.7643e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([-6.4137e-04,  7.7189e-01,  3.3054e-02, -4.9170e-02, -3.4239e-02,
        -1.0709e-02,  2.8344e-02,  1.6590e-02, -1.0115e-02, -1.6508e-03,
         1.2269e-02, -1.1968e-03, -1.0741e-03, -2.1615e-02, -7.4475e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8570e-02,  7.1163e-01, -2.7152e-02,  7.6882e-03,  1.9587e-02,
         5.5351e-03, -2.4377e-03,  9.4005e-03, -7.8640e-03, -3.1397e-04,
        -1.4323e-02,  2.5712e-03, -2.6350e-03, -6.5044e-03,  3.2075e-03,
         8.3765e-03,  3.8444e-03,  6.4368e-03,  2.9264e-03,  5.0808e-03,
         1.0353e-02, -3.3295e-05,  2.5739e-03, -9.6071e-04, -2.7501e-03,
        -2.0236e-04, -6.9523e-04,  3.9676e-03,  3.8384e-03,  4.4062e-03,
        -2.6961e-03, -3.4945e-04, -2.6779e-03,  2.4900e-03, -4.5948e-03,
        -2.3846e-03,  4.0436e-03, -6.2047e-04,  1.6068e-03, -2.0536e-03,
        -6.9235e-05,  2.2702e-03, -2.5902e-03, -1.4620e-03,  5.0839e-04,
        -6.3721e-03,  6.4330e-03,  8.7792e-03, -5.2137e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0105,  0.5781, -0.0899, -0.0178,  0.0166, -0.0086, -0.0036,  0.0188,
         0.0286, -0.0393,  0.0058, -0.0009,  0.0008, -0.0011, -0.0053, -0.0057,
        -0.0194,  0.0014, -0.0075, -0.0021,  0.0113, -0.0084, -0.0069,  0.0028,
         0.0044,  0.0038, -0.0014,  0.0144,  0.0008, -0.0022,  0.0022,  0.0017,
         0.0007,  0.0728, -0.0043,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0433, -0.7735, -0.0301,  0.0019, -0.0040,  0.0100,  0.0108, -0.0011,
         0.0019,  0.0054,  0.0126,  0.0113,  0.0060,  0.0048, -0.0052,  0.0067,
        -0.0071,  0.0191,  0.0454,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8014e-03, -7.8266e-01,  5.2852e-02, -9.0911e-03,  2.3065e-03,
        -7.7665e-03, -1.9963e-02,  7.1828e-06, -4.8042e-03,  5.4773e-03,
         8.5450e-03, -1.0927e-02, -4.9932e-02, -2.6238e-03,  4.1247e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1625e-03,  7.2569e-01, -6.1094e-02, -1.3137e-02,  8.9830e-03,
        -1.9699e-02, -3.0351e-02, -9.0963e-03,  4.2745e-02,  7.8497e-04,
        -6.7084e-03,  6.0329e-03, -3.0746e-04,  3.5622e-04, -1.7670e-02,
         8.6501e-03, -2.4155e-03,  5.6178e-03,  9.4500e-03,  2.1394e-02,
        -2.2289e-03, -6.4302e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.9001e-02,  6.3696e-01,  3.2894e-02, -1.8220e-02, -1.5536e-03,
        -1.0800e-02, -8.5469e-03,  1.1536e-02, -3.1423e-03,  1.0500e-02,
         5.1725e-03,  3.2595e-02,  1.6593e-03,  1.6239e-02,  5.6031e-04,
         6.7960e-03, -3.6450e-03, -3.3626e-03, -1.3065e-03, -3.3948e-03,
         2.6265e-03,  3.5201e-03, -3.5225e-03, -4.4916e-03,  2.9967e-02,
         1.1297e-03, -9.7052e-03, -6.0547e-03,  9.5329e-03,  3.1568e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0247,  0.5809, -0.0501, -0.0252,  0.0991,  0.0371,  0.0083,  0.0295,
        -0.0040,  0.1134, -0.0277,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1876e-02,  7.4621e-01, -6.5789e-05, -9.2018e-03,  2.8417e-03,
        -8.3840e-03,  1.0770e-03,  6.3079e-03, -2.8209e-03,  3.9678e-03,
        -3.6148e-02, -3.7272e-03, -5.8446e-03,  8.4412e-03, -1.5705e-03,
         7.2797e-03, -8.6690e-03,  4.0462e-04,  3.2425e-03, -1.2243e-02,
         3.9480e-02, -6.0197e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9526e-02, -7.0638e-01, -4.6828e-02, -2.3515e-02,  2.1501e-03,
         1.0731e-02, -5.0759e-05,  5.9950e-04,  3.2939e-03,  1.5872e-02,
         2.9435e-03, -1.9369e-02,  1.4362e-03,  2.1556e-02,  2.0221e-03,
         1.6479e-02, -3.1020e-03,  5.6249e-03, -9.5283e-03,  6.1427e-03,
        -1.6115e-03,  7.1097e-03,  1.2658e-02,  9.7987e-03,  5.5202e-04,
         1.1498e-03,  6.1088e-03, -2.8807e-03,  7.3419e-03, -3.8148e-03,
         1.3034e-03, -2.1803e-03, -6.3363e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5406e-03,  7.1545e-01,  6.0552e-02,  1.8051e-03,  3.4426e-03,
         2.1811e-02,  3.8594e-03,  7.6914e-03,  7.8160e-03, -9.4884e-03,
        -1.6992e-02,  9.1355e-03, -3.3476e-03,  1.4329e-02,  4.5636e-04,
        -5.6933e-03,  4.1458e-03, -4.8830e-03, -8.0904e-03, -1.7452e-02,
        -1.7733e-02, -1.0669e-02,  7.2161e-03, -3.6996e-03,  6.0955e-03,
         1.2920e-02,  1.6699e-02, -4.9841e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8551e-02, -7.1494e-01, -9.4866e-02, -1.3314e-02,  3.0273e-03,
         4.2354e-03,  3.3090e-03, -8.3249e-03,  1.6256e-03,  1.1148e-03,
         9.9006e-04,  2.8572e-03,  1.0261e-02,  1.4223e-02,  1.3063e-02,
         1.1565e-02,  1.5673e-02,  9.0241e-03,  8.0015e-03,  7.0235e-04,
         6.9095e-03, -1.6112e-03,  6.8896e-03, -1.0262e-02, -4.6621e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-6.3366e-03,  5.7437e-01,  6.6047e-02, -2.1642e-02,  1.4461e-02,
        -1.2207e-02, -1.9471e-02, -1.9147e-02, -7.0749e-03, -4.2365e-03,
         7.4235e-04, -5.5255e-03, -6.2672e-03,  4.0200e-03,  6.9315e-03,
        -1.5021e-02, -1.1198e-02, -6.9921e-03, -3.4275e-03, -7.0560e-03,
         3.6441e-04, -4.3874e-03, -1.6450e-02,  1.3113e-02,  2.9746e-02,
         7.4804e-03,  9.6318e-03, -1.7250e-03,  9.2693e-04, -6.3186e-05,
         3.5440e-03, -7.7144e-03, -1.3557e-02, -5.0252e-03,  4.7175e-03,
        -9.4997e-03, -4.7474e-03, -1.0753e-03, -1.2755e-02,  2.2827e-02,
        -1.8479e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3063e-02,  6.8760e-01, -9.6938e-03, -1.1563e-02,  2.6488e-02,
        -3.1373e-02, -1.8531e-02, -1.5752e-03, -5.4805e-03,  3.2748e-03,
        -1.1824e-03, -2.0667e-02,  1.4740e-02, -1.4661e-03,  1.6301e-02,
        -5.2256e-03, -4.9277e-03, -4.1854e-03, -7.3586e-03, -2.1392e-02,
        -3.4716e-04, -2.1961e-03, -4.1584e-03, -4.8400e-02, -2.8814e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.3855e-02,  7.0223e-01,  1.9812e-02, -6.9433e-02, -1.0014e-02,
         2.3496e-02, -1.7937e-02, -4.3450e-03, -4.9420e-03, -1.5742e-02,
         9.6558e-04,  9.2228e-03, -6.2030e-03,  2.8268e-04,  3.0968e-03,
        -2.0420e-03,  2.7875e-03,  8.6151e-03, -1.3600e-02, -6.5650e-04,
        -7.2069e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0429,  0.6905,  0.0332, -0.0025, -0.0105, -0.0040, -0.0039, -0.0397,
        -0.0117,  0.0237,  0.0231, -0.0044, -0.0009, -0.0139, -0.0098,  0.0082,
        -0.0538, -0.0233,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.8381e-02,  5.6918e-01,  2.2813e-02,  5.3795e-02,  2.5771e-02,
        -1.3602e-02, -4.6871e-04, -1.3674e-02,  1.2068e-02,  8.2354e-03,
        -1.3839e-02, -6.9155e-04, -1.0572e-02, -1.6275e-02, -9.7547e-03,
         4.6905e-03, -2.0228e-02,  1.2230e-02, -1.7611e-02, -1.1770e-02,
        -1.1522e-02, -9.6528e-03, -4.4150e-03, -1.0903e-02,  2.0943e-03,
         8.2115e-03,  7.0916e-03,  4.2533e-03, -5.0735e-03,  4.1967e-03,
         2.5703e-03,  1.8193e-03, -1.0999e-03, -3.4127e-03, -7.9345e-03,
         1.6913e-03,  1.3848e-02,  1.4562e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0814,  0.6119, -0.0478,  0.0185, -0.0102, -0.0016, -0.0069, -0.0018,
        -0.0081,  0.0012, -0.0034, -0.0056,  0.0019, -0.0194,  0.0008,  0.0052,
         0.0032, -0.0385, -0.0494, -0.0831,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2541e-02,  7.2224e-01,  5.4853e-02, -6.9126e-04,  2.1296e-02,
         6.5891e-02, -1.0318e-02, -2.1674e-04,  1.7314e-02,  2.4873e-02,
         1.4746e-02, -1.6527e-02, -9.8614e-03,  1.8631e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0674,  0.5656, -0.0724, -0.0203,  0.0084,  0.0151,  0.0129,  0.0056,
         0.0103,  0.0127,  0.0044,  0.0085,  0.0108, -0.0213, -0.0034, -0.0106,
         0.0059,  0.0012,  0.0045,  0.0018,  0.0049,  0.0218,  0.0044,  0.0246,
        -0.0012,  0.0167,  0.0046,  0.0054,  0.0049,  0.0201, -0.0087,  0.0196,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0326, -0.7324, -0.0238,  0.0132, -0.0018,  0.0220, -0.0010, -0.0042,
         0.0062,  0.0325,  0.0124,  0.0147,  0.0051, -0.0136, -0.0081,  0.0019,
         0.0089,  0.0142,  0.0038,  0.0478,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0401,  0.5879,  0.0538, -0.0234, -0.0069,  0.0145, -0.0114, -0.0072,
        -0.0073,  0.0078, -0.0045, -0.0061, -0.0074, -0.0087, -0.0077, -0.0079,
        -0.0055, -0.0045,  0.0023,  0.0018, -0.0059, -0.0116, -0.0112, -0.0045,
        -0.0009, -0.0048, -0.0151, -0.0126, -0.0074, -0.0080,  0.0023, -0.0025,
         0.0016,  0.0028,  0.0023,  0.0030, -0.0035, -0.0127, -0.0063,  0.0049,
         0.0067,  0.0165,  0.0107, -0.0010, -0.0216,  0.0020, -0.0010,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6118e-02,  7.4335e-01, -1.2521e-02,  1.2288e-02,  7.1925e-03,
         1.5900e-02, -3.9378e-04, -1.5503e-02, -2.4382e-02,  1.4060e-02,
        -2.6149e-02,  1.9656e-02, -1.3868e-02, -1.9129e-02,  5.9485e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4591e-02,  6.1036e-01,  5.7300e-02,  5.2857e-03, -1.1249e-02,
        -1.3198e-03,  5.5959e-04,  3.7748e-03,  6.9872e-03,  1.9750e-04,
        -1.1260e-02, -4.7371e-03, -6.3283e-03,  1.3490e-02,  8.7106e-03,
         2.6388e-03,  3.0048e-03,  3.6279e-03,  9.8587e-03, -9.5129e-03,
         1.3250e-04, -9.9182e-03, -3.7513e-03,  1.2009e-02,  2.7813e-03,
        -2.4826e-02,  1.0490e-02,  1.2064e-02, -7.9390e-03,  4.0950e-03,
        -1.9730e-03, -3.2508e-04,  7.4086e-03, -3.7904e-03,  4.1096e-03,
         8.1480e-03, -2.1710e-04, -2.6911e-03, -1.2649e-02, -1.3831e-05,
        -2.5326e-04,  1.0005e-03, -2.1475e-03, -1.3482e-03, -1.0998e-03,
         5.5169e-04,  1.1406e-02,  1.8066e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-0.0535,  0.6342,  0.0305, -0.0016,  0.0158,  0.0057,  0.0141,  0.0058,
        -0.0290,  0.0019,  0.0018,  0.0058, -0.0043, -0.0087,  0.0065, -0.0034,
        -0.0173, -0.0077, -0.0120,  0.0054,  0.0360,  0.0033,  0.0078,  0.0213,
         0.0627, -0.0038,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0121e-02, -7.3600e-01, -2.7950e-02, -1.3369e-03,  6.5892e-03,
        -1.1405e-02, -3.0316e-02, -3.2528e-02, -6.6964e-04,  4.3269e-03,
        -2.3005e-03, -6.8573e-03,  3.1469e-03,  1.1838e-02,  8.4293e-03,
         7.4079e-03,  1.1567e-02,  1.4278e-02,  3.3444e-02, -9.7403e-03,
        -1.9751e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.8423e-03,  6.4293e-01,  1.0015e-01, -1.2034e-03,  1.4561e-02,
        -1.8049e-02, -1.0424e-02, -1.4492e-03, -1.1027e-02, -3.4735e-03,
         4.1380e-03,  4.8884e-03, -2.0116e-02, -1.2576e-02,  1.5094e-03,
        -3.5913e-04,  2.3823e-03, -1.6042e-02,  6.7225e-03, -1.3399e-02,
        -5.9139e-03, -6.4300e-03,  5.7463e-03, -6.1254e-03,  1.8457e-02,
        -4.8570e-03,  3.3589e-03, -2.4101e-04, -3.8234e-03,  1.6311e-02,
        -7.9011e-03, -2.5178e-03,  2.9405e-03,  5.0979e-03, -6.8996e-03,
         9.1435e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4992e-03,  7.1112e-01,  2.5904e-02,  4.2279e-03,  1.6736e-02,
         1.0037e-02, -2.9485e-02, -5.2462e-03, -2.2641e-02,  2.5704e-02,
         1.4353e-02,  4.4901e-03,  2.6495e-04, -3.1513e-03, -6.2579e-03,
         3.0613e-02, -1.6703e-02,  2.0167e-02,  1.6161e-02, -3.5242e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.5478e-02,  6.6777e-01,  1.3896e-02,  1.2225e-02,  1.0861e-02,
        -1.0937e-02, -1.1658e-02, -2.1486e-02, -3.6159e-02,  3.6526e-03,
         1.6950e-02,  1.3154e-02, -1.4160e-02, -7.3394e-03, -3.6537e-04,
         6.2125e-03, -1.0833e-02, -6.5438e-03,  5.2333e-03, -9.0190e-03,
        -5.6071e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0616,  0.6493,  0.0541, -0.0050,  0.0026, -0.0062, -0.0218, -0.0529,
         0.0056,  0.0282,  0.0267,  0.0250,  0.0611,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3664e-03,  7.0985e-01, -1.4648e-02,  1.7497e-02,  5.3767e-03,
         1.5118e-02,  1.2583e-04,  7.0009e-03,  5.3756e-03,  4.2915e-03,
         1.4094e-02, -7.3601e-03, -6.9259e-03, -9.6989e-03,  1.0993e-02,
         1.6717e-02,  3.9624e-03, -1.3036e-03,  1.1232e-03,  2.3340e-03,
         1.1749e-02, -6.9757e-03,  4.6757e-03,  1.8092e-02,  4.4330e-03,
         7.9810e-03, -1.7509e-02, -9.0671e-03,  4.4201e-03,  7.7522e-03,
         2.4418e-03,  3.3486e-03,  2.3361e-03, -5.4181e-03,  1.0587e-02,
        -5.8090e-04,  2.6466e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0678,  0.6993,  0.0023,  0.0023, -0.0384, -0.0197, -0.0064,  0.0023,
         0.0120, -0.0319, -0.0044, -0.0185,  0.0133, -0.0127, -0.0051, -0.0092,
         0.0133,  0.0028,  0.0057, -0.0031, -0.0020, -0.0080,  0.0046,  0.0151,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0377,  0.5673,  0.1228, -0.0054,  0.0118,  0.0081, -0.0065, -0.0152,
        -0.0100,  0.0012, -0.0039,  0.0269,  0.0099,  0.0013, -0.0216, -0.0072,
        -0.0044, -0.0065, -0.0071, -0.0061, -0.0057,  0.0229, -0.0046, -0.0073,
         0.0026, -0.0008, -0.0014,  0.0015, -0.0009,  0.0104, -0.0055, -0.0553,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8192e-02,  7.0957e-01, -3.7959e-03,  1.2478e-03,  3.4286e-02,
         1.0322e-02, -1.1413e-02, -1.4921e-02,  1.3284e-02, -2.9379e-02,
         1.7131e-03, -4.1649e-03, -8.9249e-03, -6.0492e-04, -2.7411e-03,
         9.5948e-03, -8.1439e-05, -5.6853e-04,  1.6212e-02,  1.3653e-02,
        -1.6120e-02,  5.9316e-03, -8.1760e-03, -1.8732e-02,  9.7785e-04,
         9.3573e-03, -1.7591e-02, -3.7719e-03, -2.0689e-03,  4.2053e-03,
        -8.3978e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5635e-02,  7.3997e-01, -2.9999e-02, -7.0841e-03,  3.3093e-02,
         2.2650e-02, -5.5562e-03,  2.6708e-03,  1.7010e-02,  2.3867e-03,
        -5.5264e-04,  1.6516e-02, -1.0879e-02, -7.7423e-03,  1.9907e-03,
        -7.7162e-03, -4.3214e-03,  1.1295e-02,  2.8207e-03, -2.2183e-03,
         1.8138e-02, -3.6094e-02,  3.6607e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0260, -0.6086, -0.0249,  0.0295, -0.0085,  0.0217,  0.0044,  0.0160,
         0.0132,  0.0183,  0.0099,  0.0349,  0.0098,  0.0137,  0.0107,  0.0127,
         0.0178,  0.0046,  0.0141,  0.0010,  0.0168,  0.0073, -0.0087, -0.0031,
         0.0043, -0.0054,  0.0487,  0.0053,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #100: [tensor([-3.5108e-02,  6.8722e-01,  2.6690e-03,  1.5135e-03, -6.7907e-03,
         4.1591e-04,  4.2160e-03,  9.7868e-03,  6.1802e-03, -3.3730e-02,
        -1.1099e-02, -1.3895e-02, -6.2134e-03, -1.8571e-03, -6.3743e-03,
        -4.2374e-03, -3.8982e-03, -2.0655e-03, -7.1021e-03,  9.9354e-04,
         9.9098e-03,  6.5577e-03, -6.8129e-03,  4.3341e-03, -5.4414e-03,
         3.9998e-03,  3.6084e-02, -4.0936e-02, -4.0560e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7081e-03, -7.5239e-01, -1.4728e-02,  2.8910e-03, -1.3676e-02,
        -2.7484e-02, -1.4412e-02, -9.2719e-03,  4.4769e-03,  8.1960e-03,
        -1.0568e-02, -3.6298e-03,  1.5731e-02,  8.0306e-03, -5.4213e-03,
         5.3423e-03, -7.6638e-03, -9.0229e-03,  6.3517e-03,  2.2638e-02,
        -5.5148e-03,  5.0408e-04,  7.9441e-03, -3.4402e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0058,  0.5541,  0.0571,  0.0424,  0.0119,  0.0007,  0.0012, -0.0163,
         0.0195, -0.0074,  0.0083,  0.0212, -0.0158,  0.0091, -0.0281,  0.0057,
         0.0017, -0.0008,  0.0027, -0.0072,  0.0075,  0.0030,  0.0007,  0.0055,
         0.0118,  0.0050, -0.0125,  0.0043, -0.0025,  0.0056, -0.0052, -0.0208,
         0.0083, -0.0903], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.4101e-03,  7.8186e-01,  1.6705e-02, -8.4213e-03, -2.7223e-03,
        -1.1789e-03,  3.3716e-02,  1.2261e-02,  1.2356e-02,  4.3125e-02,
        -4.4981e-03, -9.7651e-03, -3.1838e-03, -8.1541e-03,  7.3290e-05,
        -1.4185e-03,  7.5109e-03, -8.9285e-03,  6.7614e-03,  1.7198e-02,
        -3.3310e-03,  9.4204e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0036, -0.7706, -0.0340,  0.0251, -0.0321,  0.0300, -0.0078, -0.0055,
         0.0022,  0.0009, -0.0081,  0.0075,  0.0182,  0.0022, -0.0073,  0.0074,
         0.0207,  0.0168,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2041e-02,  7.0685e-01,  1.1042e-02, -2.0600e-02,  3.1183e-02,
        -1.2886e-03,  8.1739e-03,  2.1777e-02,  1.5862e-04, -2.0815e-02,
         5.8525e-03, -3.3610e-03, -5.4642e-03,  4.4432e-04, -7.7906e-03,
        -1.7039e-02, -1.9977e-03, -6.9968e-03, -2.4018e-02, -1.3532e-03,
        -1.6207e-02, -4.0550e-02, -1.5000e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1689e-02,  7.7414e-01,  4.2784e-02,  2.9361e-02,  4.9108e-03,
        -1.8787e-03,  1.5848e-03, -7.7980e-03,  1.7532e-03,  1.6323e-03,
         1.7568e-03, -8.9569e-03,  4.5639e-04, -1.2768e-02, -8.6326e-03,
        -7.0242e-02, -1.9656e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3965e-02, -7.7268e-01, -2.1250e-02,  2.1372e-03,  4.0043e-03,
        -6.4442e-03,  2.1314e-02,  2.2558e-02,  1.1301e-02,  2.0031e-02,
         1.6083e-02,  1.0338e-02,  6.3749e-03,  3.0251e-04, -9.3514e-03,
         4.1172e-03, -3.2785e-02,  4.9603e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0028,  0.7350,  0.0259, -0.0162, -0.0012,  0.0068,  0.0010,  0.0025,
        -0.0133,  0.0510, -0.0016,  0.0095,  0.0012,  0.0049, -0.0074, -0.0010,
        -0.0078, -0.0015,  0.0229, -0.0109, -0.0038, -0.0084, -0.0100,  0.0103,
         0.0068,  0.0363,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.4667e-02,  6.0712e-01,  7.6347e-02,  7.0739e-03,  2.2676e-02,
         4.6282e-03,  4.2023e-03,  2.6946e-03, -1.6678e-02, -1.4630e-03,
        -7.9198e-03, -1.2621e-02, -1.8958e-03, -3.2772e-03, -1.3818e-02,
        -6.6163e-03, -9.8117e-03, -1.1705e-02, -4.6898e-03, -2.0118e-02,
         1.6744e-03, -7.2725e-03, -9.7332e-03, -1.1649e-02, -1.0302e-02,
         2.6439e-03,  3.2930e-03, -3.7852e-04,  8.3734e-03,  2.2881e-02,
        -1.1775e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9471e-03,  7.7564e-01,  1.6412e-02, -3.1074e-04,  3.2966e-02,
         2.5808e-02,  1.8734e-02, -3.6027e-03, -8.2889e-03,  9.5393e-03,
         1.9923e-02, -1.0302e-02, -1.7189e-03, -9.8604e-03,  8.3460e-04,
        -2.5392e-02, -1.2459e-02, -2.5266e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0845,  0.7734, -0.0316, -0.0112,  0.0089, -0.0133,  0.0355, -0.0417,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 5.6090e-02,  6.6982e-01,  5.0098e-02, -1.9614e-02,  1.6010e-02,
         1.6045e-02, -4.1292e-03,  2.7716e-03, -9.1386e-03, -4.0542e-03,
         7.6377e-04, -6.8084e-03,  1.1350e-03,  9.1605e-04, -4.7115e-03,
         1.5037e-02,  3.9891e-03,  4.6738e-04, -5.4176e-03, -1.4843e-02,
         9.9071e-04, -8.1434e-03,  1.0387e-04, -3.3402e-04, -3.3895e-03,
        -2.0951e-03, -1.0139e-02,  7.1484e-03, -2.1764e-03,  6.8767e-04,
        -8.2426e-03,  1.7150e-03,  7.1990e-03,  2.6587e-03,  4.7915e-03,
         1.9826e-03,  1.5881e-03,  1.3016e-02,  2.1741e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0110,  0.6654,  0.0270, -0.0051, -0.0275, -0.0274, -0.0135, -0.0021,
        -0.0131,  0.0024,  0.0042,  0.0091,  0.0206, -0.0131,  0.0110,  0.0021,
        -0.0036, -0.0118, -0.0074,  0.0017, -0.0190, -0.0171, -0.0084, -0.0050,
         0.0026,  0.0018, -0.0499,  0.0172,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.0067e-02, -7.2082e-01,  1.7340e-02, -2.3799e-02, -2.0265e-03,
         1.9093e-02,  1.3657e-02, -2.9015e-03, -6.4018e-04, -1.1381e-02,
         6.4317e-02,  1.8127e-03,  4.2150e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0565,  0.6867, -0.0331, -0.0234,  0.0140, -0.0328, -0.0050, -0.0063,
         0.0094, -0.0248,  0.0081, -0.0388,  0.0115,  0.0417,  0.0078,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.9210e-02,  6.1788e-01,  1.1299e-02,  2.4283e-02,  1.0060e-02,
        -1.4034e-02,  5.3841e-03, -2.3006e-02, -4.7287e-03, -1.3605e-02,
        -2.2954e-03, -4.4917e-03, -1.7052e-02,  1.8441e-02,  5.4579e-03,
         4.7274e-03, -2.7141e-03, -1.0465e-03,  3.7436e-03, -3.6599e-03,
         9.3797e-03, -2.7425e-03,  4.6505e-03, -1.0693e-03,  6.4754e-04,
        -4.0802e-03, -7.1194e-03,  2.8871e-03, -5.5382e-04,  7.6337e-03,
        -3.0658e-03,  3.1316e-03,  2.5163e-03, -7.8037e-03,  1.3861e-03,
        -2.3740e-02, -3.1055e-03,  2.1243e-03, -8.5050e-03, -2.6801e-03,
         2.9037e-03, -5.8475e-03,  6.2880e-03,  1.8130e-03,  3.1307e-03,
         1.2893e-02,  1.1180e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6505e-04,  7.2767e-01,  4.0527e-02,  1.6460e-02,  1.5239e-02,
        -1.0319e-02,  6.1686e-03,  1.0343e-02,  4.1337e-03,  4.8152e-03,
         3.6920e-03, -1.6781e-02, -5.0976e-03, -1.0297e-02, -1.0097e-03,
        -2.2923e-03,  4.4201e-03, -1.6715e-02, -8.0864e-03, -1.6639e-03,
         4.3962e-04, -1.7830e-03,  3.7913e-03, -1.0034e-02, -2.1181e-03,
        -7.2592e-03, -1.7560e-02, -3.5850e-04, -5.2423e-03, -2.0465e-03,
        -7.4867e-03, -4.4726e-03, -1.3116e-03, -7.0878e-04,  2.8017e-03,
        -6.5321e-03, -4.7584e-03, -2.0890e-03, -6.0535e-03, -1.3582e-03,
         3.5202e-03,  2.2842e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4087e-02,  6.3955e-01,  4.2284e-02, -1.7996e-02,  2.1661e-02,
         1.4367e-02,  5.0317e-02,  1.2568e-02, -7.8151e-03,  2.1290e-02,
         1.6271e-02, -4.1229e-03, -2.3535e-04,  4.5648e-03, -4.3269e-03,
         1.0990e-02,  2.4066e-02,  9.7324e-04, -2.7397e-03,  7.4685e-03,
         8.8035e-04, -4.5075e-03,  3.6952e-03, -2.2411e-02,  2.0810e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3378e-02,  6.9308e-01,  9.0291e-02, -1.1613e-02, -3.4260e-03,
        -1.1307e-02, -5.6908e-02, -1.2089e-02,  1.9993e-03,  5.8097e-03,
         2.1946e-03, -4.3873e-03, -3.9069e-04, -8.7501e-03, -2.2597e-02,
        -9.7430e-03, -5.4550e-03,  6.4630e-04, -7.4607e-03,  6.2399e-03,
        -2.5381e-03,  6.5521e-03,  2.3141e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5277e-03,  7.8297e-01, -1.2920e-02,  7.2446e-03, -4.4377e-02,
         5.4646e-03,  2.7627e-03, -4.7223e-03, -1.0161e-02,  6.0141e-04,
        -1.2231e-03, -2.2508e-02, -2.1003e-02, -3.5891e-03, -3.8531e-03,
         1.0993e-03, -2.1380e-02, -7.3757e-03, -4.1501e-03, -8.0947e-03,
         3.4921e-03, -7.3980e-03, -2.2079e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0616,  0.6810, -0.0732,  0.0026, -0.0216, -0.0019,  0.0181, -0.0329,
         0.0165, -0.0166,  0.0474, -0.0178,  0.0087,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.1722e-03,  6.1819e-01, -4.4609e-02,  5.1481e-02, -2.5429e-02,
         2.8216e-02,  9.4507e-03,  7.9452e-02,  3.3201e-04,  1.5709e-02,
        -1.8315e-02,  3.2352e-02,  2.2515e-02, -4.6779e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0238,  0.7203,  0.0659, -0.0199,  0.0110, -0.0127, -0.0085,  0.0126,
        -0.0076, -0.0143, -0.0067, -0.0060,  0.0019, -0.0302,  0.0072,  0.0055,
        -0.0031, -0.0102,  0.0327,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 1.6781e-01,  4.4762e-01,  1.3743e-02, -1.6959e-01, -1.6763e-02,
        -1.0543e-02, -1.8652e-03, -9.8055e-04, -2.4418e-02,  1.6068e-05,
         6.2979e-03, -4.7298e-04, -1.1723e-04,  7.9708e-04, -5.6544e-04,
         1.5253e-02,  1.8751e-04, -2.0366e-02, -3.0120e-03,  6.2386e-03,
        -2.6276e-02, -1.2925e-03,  1.8241e-03,  7.0068e-04, -6.2278e-02,
        -9.7889e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9005e-02, -7.4095e-01, -4.7068e-04,  8.5799e-04, -2.7902e-02,
         2.5406e-02, -9.1075e-04,  2.0261e-02, -5.5514e-03,  1.3193e-02,
        -1.3573e-02,  1.3597e-03, -1.8468e-02, -1.7895e-03,  3.0575e-02,
        -9.7265e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2238e-01,  6.2855e-01, -3.3112e-02,  1.9652e-02, -4.3225e-03,
         2.2150e-02, -1.1201e-02,  1.7403e-02, -4.6119e-02, -5.0591e-03,
        -2.3828e-04, -2.1512e-03, -1.0892e-02,  1.8982e-02,  2.2062e-02,
         7.2930e-03,  2.8430e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1242e-05,  8.1022e-01, -1.4502e-02, -2.0688e-02,  1.1835e-02,
        -1.5982e-02, -4.2410e-02,  2.2818e-02, -6.1505e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0014, -0.4785, -0.1031,  0.0037, -0.0192, -0.0093, -0.0182,  0.0006,
        -0.0024, -0.0114, -0.0041, -0.0041,  0.0021,  0.0105,  0.0053, -0.0067,
         0.0088, -0.0049,  0.0087, -0.0009, -0.0028,  0.0022,  0.0212,  0.0130,
         0.0101,  0.0020,  0.0018,  0.0074,  0.0091,  0.0086, -0.0010,  0.0039,
         0.0043,  0.0061, -0.0044, -0.0045, -0.0023, -0.0094,  0.0074,  0.0041,
         0.0051,  0.0022,  0.0185,  0.0124, -0.0007, -0.1316], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6666e-02,  7.0949e-01,  2.0012e-05,  7.4277e-03,  2.0265e-02,
        -4.9891e-03,  3.1532e-02, -1.7659e-02, -1.6734e-02,  8.7637e-03,
        -5.0224e-03, -7.8727e-03,  2.4213e-02, -3.1925e-03, -2.8422e-03,
        -1.0934e-02, -1.7583e-03, -6.1822e-03,  2.6557e-03, -2.0576e-03,
         1.0909e-03,  2.3113e-03, -3.1986e-04, -8.7377e-03,  8.4026e-04,
        -1.3468e-03, -5.9466e-03,  2.3253e-03, -2.0032e-02, -3.6767e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([0.0191, 0.8339, 0.0284, 0.0530, 0.0052, 0.0127, 0.0478, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0983,  0.6225, -0.0292, -0.0363,  0.0219,  0.0231, -0.0174,  0.0401,
        -0.0256,  0.0774, -0.0084,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0266,  0.6816,  0.0233,  0.0236,  0.0217,  0.0367, -0.0087, -0.0109,
         0.0164,  0.0063, -0.0171, -0.0072, -0.0908, -0.0292,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0262,  0.6900,  0.0400,  0.0130,  0.0232,  0.0074,  0.0123,  0.0158,
         0.0042,  0.0011, -0.0073, -0.0122, -0.0035, -0.0110,  0.0043, -0.0026,
         0.0023, -0.0034,  0.0016,  0.0045,  0.0195,  0.0063,  0.0074,  0.0094,
        -0.0096, -0.0098, -0.0186,  0.0032,  0.0021,  0.0016, -0.0019, -0.0052,
        -0.0008,  0.0188,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0901,  0.4994,  0.0087, -0.0438, -0.0032, -0.0132,  0.0240,  0.0144,
        -0.0246,  0.0107, -0.0257,  0.0092,  0.0151,  0.0451,  0.0030, -0.0180,
         0.0035, -0.0105, -0.0115, -0.0018,  0.0332, -0.0108, -0.0187, -0.0010,
         0.0052, -0.0114, -0.0410, -0.0031,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0877,  0.6465,  0.0508, -0.0047,  0.0168, -0.0124, -0.0059, -0.0055,
         0.0032, -0.0084, -0.0021, -0.0110,  0.0021,  0.0074,  0.0205,  0.0121,
        -0.0064, -0.0052, -0.0073, -0.0073, -0.0038,  0.0033, -0.0573, -0.0123,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([-0.0266, -0.4223,  0.0053, -0.0065,  0.0531,  0.0067, -0.0019, -0.0199,
        -0.0260, -0.0107,  0.0154,  0.0170, -0.0106, -0.0069, -0.0127,  0.0143,
         0.0062,  0.0090, -0.0108, -0.0133,  0.0096,  0.0104, -0.0062, -0.0048,
         0.0046,  0.0096, -0.0242, -0.0033,  0.0129,  0.1224, -0.0968,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0677,  0.4250, -0.0967,  0.0367,  0.0438,  0.1289,  0.0573, -0.0169,
        -0.0258, -0.0144, -0.0383, -0.0486,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5856e-02,  3.9856e-01,  5.9632e-02, -5.4285e-02,  2.6760e-02,
         8.1105e-03, -5.2431e-05, -1.2826e-02,  1.5358e-02, -6.5663e-03,
        -9.8373e-02, -6.0924e-02,  6.8899e-03, -5.6050e-02,  1.3975e-01,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6803e-02, -5.0582e-01,  1.0355e-02,  4.6970e-02,  8.4198e-05,
        -8.3448e-03, -1.8566e-03, -3.4455e-03, -7.9566e-03, -2.0731e-02,
         3.4087e-02, -3.6154e-03, -2.2113e-02,  8.4106e-03,  1.6553e-02,
         1.2297e-03, -9.1680e-03,  5.7250e-03, -2.6413e-02,  1.1571e-02,
        -1.4302e-02, -4.3888e-03, -9.2653e-03, -5.2235e-03, -1.7802e-02,
         3.7026e-02,  1.3745e-02,  5.6208e-03,  5.7963e-03,  2.0893e-02,
        -5.0600e-03, -5.9342e-02,  1.0288e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1016, -0.2944,  0.0145,  0.0115,  0.0329, -0.0116,  0.0151, -0.0505,
         0.0512, -0.0152,  0.0798,  0.0198,  0.0124,  0.0492, -0.0881,  0.1523,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0281, -0.6150,  0.0919, -0.0439, -0.0539, -0.0011,  0.0527,  0.0771,
        -0.0363,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1138, -0.6536, -0.1010, -0.0675,  0.0641,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9349e-01,  4.9858e-01,  9.9262e-02, -2.9503e-02, -1.0540e-02,
        -5.0367e-03, -1.7757e-02,  3.7101e-04,  7.5067e-03, -1.3403e-02,
         4.7319e-03,  9.2281e-03, -1.0659e-02, -1.7299e-02, -2.5383e-03,
        -5.1666e-03, -8.1437e-03, -3.8222e-03,  9.0557e-03,  8.5500e-05,
         6.9590e-03, -1.4200e-03, -2.1543e-03,  4.8842e-03,  3.6622e-03,
        -1.9510e-02, -1.5234e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0777,  0.6574,  0.0495, -0.0051,  0.0796, -0.0823,  0.0485,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5159e-01, -5.5568e-01,  7.9115e-03, -1.2386e-02, -1.0166e-02,
        -2.8422e-03, -1.1943e-02,  4.9216e-03,  2.4337e-03,  5.7629e-03,
        -6.4064e-03, -7.6199e-03,  2.2254e-03,  3.5104e-04, -6.8594e-03,
         3.0096e-03, -3.1933e-02,  3.1474e-03, -9.2340e-03, -4.6565e-02,
         1.9949e-02,  9.7061e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1752e-02, -5.5902e-01, -1.6632e-02, -9.9994e-03, -3.1321e-02,
         2.9308e-03,  1.3475e-02, -6.2651e-02, -3.5627e-03,  1.9517e-03,
        -1.5951e-02,  3.0211e-02,  2.1487e-02, -1.3607e-04,  2.1186e-02,
         1.1896e-02, -3.8434e-03,  8.3984e-02, -4.8011e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3655e-01,  4.3481e-01,  1.1902e-02, -1.2574e-02,  2.0468e-02,
         3.1766e-02,  2.5275e-02,  1.2374e-03, -4.0640e-03,  8.5825e-03,
        -7.9246e-04,  5.6967e-03,  3.4420e-03,  1.2558e-02, -6.1472e-03,
        -6.8840e-05, -1.9767e-02, -2.2876e-03,  6.1092e-03,  2.7928e-02,
         1.3270e-02,  2.2353e-02,  7.4193e-03,  1.0971e-02,  6.5727e-03,
        -6.3307e-03, -9.7730e-04, -1.1125e-02,  2.9009e-03,  3.7816e-02,
        -8.2422e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-3.5210e-02,  5.3824e-01,  7.8335e-02, -4.5019e-02, -3.9079e-04,
        -1.0850e-02,  9.7920e-03,  1.1405e-03, -1.0876e-02,  8.8435e-03,
         1.6904e-03,  6.5796e-03, -1.4401e-02, -7.7884e-03, -1.6108e-03,
        -2.0071e-03, -1.7553e-02, -5.4183e-03, -2.1520e-02,  3.0155e-03,
        -4.5979e-03, -1.1472e-03, -1.1463e-02, -2.4250e-02, -1.1759e-02,
        -1.9507e-03, -6.5368e-03, -1.7587e-02, -6.3194e-03,  5.1133e-03,
         3.4171e-02, -5.4827e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.8947e-02,  6.9975e-01,  2.2844e-02,  3.8308e-03, -4.2484e-03,
         1.3614e-02,  2.7169e-02, -1.2671e-02, -1.7718e-02, -2.1314e-02,
        -5.7453e-03, -3.6841e-03,  1.4461e-03,  6.7544e-03, -1.1057e-02,
        -1.7764e-02,  1.0967e-02,  4.4715e-03, -1.9514e-04,  5.6858e-03,
         1.9036e-02, -1.0891e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.7710e-03,  6.4735e-01,  6.0381e-02,  2.3697e-02, -4.0103e-03,
         2.4902e-03, -2.6431e-02,  3.4582e-03,  8.6304e-03, -1.1511e-03,
        -5.6844e-03,  5.9730e-03,  1.1279e-02, -1.0622e-02, -1.5510e-02,
        -1.3092e-02, -2.3421e-02,  6.4521e-03, -7.5461e-03, -1.1081e-02,
        -2.5548e-04, -1.3518e-02, -8.3691e-03, -1.1011e-02, -2.2115e-02,
        -1.7307e-02, -3.3187e-03,  1.5005e-02, -1.3266e-02,  2.8002e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.2145e-02,  5.8910e-01, -7.2555e-05, -1.4430e-02,  2.1467e-02,
         1.8324e-02,  4.4557e-02, -2.1123e-03,  1.4380e-02, -4.5621e-02,
         1.8184e-03,  1.7695e-02, -5.9579e-03, -1.4183e-02, -1.1066e-02,
        -1.7345e-02,  8.5686e-02,  1.4038e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7602e-02, -5.5332e-01, -9.8244e-02, -2.7956e-02, -7.8033e-03,
        -1.7903e-02,  6.5727e-03, -7.1011e-03,  1.2223e-02, -3.0591e-03,
         1.2237e-02,  1.2123e-02,  1.2193e-02,  1.0954e-02,  1.5207e-02,
         1.1720e-02,  4.8229e-04,  2.3236e-02, -6.0821e-03,  3.8921e-03,
         5.4703e-03,  1.1563e-02,  1.2581e-02,  1.4253e-02,  3.2758e-03,
         5.8606e-03, -5.1041e-03, -2.0509e-03, -1.1570e-02, -1.1867e-03,
        -2.4004e-03, -2.2659e-03, -4.1665e-04, -1.1301e-03,  7.2801e-03,
         1.8330e-02, -3.5350e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1279,  0.0365,  0.0139,  0.0067, -0.0058, -0.0084,  0.0005, -0.0385,
        -0.0169,  0.0059,  0.0535,  0.0327,  0.0165,  0.0006,  0.0076,  0.0438,
        -0.0099,  0.0033, -0.0495,  0.0018, -0.0180,  0.0835,  0.4171,  0.0016,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6881e-02,  5.5345e-01, -2.5476e-03,  1.5215e-03,  2.3851e-02,
         3.3724e-02,  3.4977e-03, -7.1598e-03, -3.1281e-03,  3.2342e-03,
         6.4592e-03,  6.2243e-04, -8.9983e-03, -1.1223e-02,  5.3198e-03,
         7.3073e-04, -6.1600e-03, -6.2946e-04, -1.2850e-03, -1.3969e-02,
        -1.2773e-02, -6.0996e-03, -8.4260e-03,  3.6888e-03, -7.2521e-03,
         1.4979e-02,  8.1487e-05,  7.8034e-03,  1.3185e-03, -3.7272e-03,
         3.0630e-02, -1.2522e-03, -4.1170e-03,  2.7819e-03,  5.9125e-03,
        -1.0138e-02, -8.1792e-03,  4.2372e-03, -5.3312e-03, -2.4655e-03,
        -4.1535e-02, -1.5306e-03,  9.0066e-04, -3.6597e-03, -4.3050e-02,
        -2.3736e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0996e-02,  5.8023e-01,  1.5030e-02,  2.8269e-02, -8.8915e-02,
         2.5000e-02, -1.8098e-02,  3.0417e-03,  1.7438e-02, -8.1765e-03,
        -1.6657e-03, -8.8703e-03,  5.7775e-02, -1.2728e-02, -1.4440e-02,
        -9.4289e-03,  1.3207e-02, -3.4443e-02, -4.0473e-03,  2.8164e-03,
        -6.0518e-03, -3.9220e-02,  1.1307e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2974e-02,  7.6337e-01, -2.1529e-02,  3.6679e-02, -1.0529e-02,
        -2.5574e-02, -2.4031e-04, -4.9303e-02,  7.8231e-03, -2.3632e-02,
         4.8349e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1389,  0.6167, -0.0162,  0.0137,  0.0274,  0.0564, -0.0011, -0.0222,
         0.0074,  0.0159, -0.0132, -0.0141, -0.0254, -0.0313,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0076,  0.7214, -0.0019, -0.0148,  0.0303,  0.0188,  0.0061,  0.0179,
         0.0164,  0.0671,  0.0601,  0.0376,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5019e-02,  5.5427e-01,  6.9782e-02,  2.9281e-02, -5.3819e-03,
        -6.6920e-03,  3.1384e-03,  1.7080e-02, -2.7937e-03, -2.5715e-03,
        -9.9599e-03,  2.1762e-03, -1.4314e-02, -1.1134e-02, -1.7181e-02,
        -4.3613e-03, -2.3726e-04, -7.0128e-03,  9.7372e-04, -7.8953e-03,
        -3.1257e-03,  2.6815e-03,  1.8347e-03,  2.8901e-03,  1.5474e-02,
        -2.4748e-02,  1.4615e-02,  1.4063e-03, -9.5835e-03,  1.3483e-04,
         2.2582e-03,  1.3477e-02, -2.0473e-03,  1.1291e-02,  2.2327e-03,
        -1.8599e-03, -4.2966e-04, -1.3458e-03,  3.0903e-03, -2.1302e-03,
        -8.8310e-04, -1.4791e-03, -1.9768e-03, -4.1032e-03, -2.3061e-04,
         5.9177e-03,  5.9831e-03,  6.5242e-04, -3.7947e-03, -2.7035e-03,
        -2.1037e-03, -2.5262e-03, -2.7157e-03, -2.5769e-02,  1.2495e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 9.1407e-02,  6.4032e-01,  4.7045e-02,  6.9310e-05,  3.8573e-03,
         2.7055e-02,  1.0237e-02, -5.8909e-03,  9.1558e-03,  1.7358e-02,
        -1.0722e-02,  1.5165e-02, -1.0186e-03, -2.9715e-02, -2.1752e-02,
        -1.0722e-02, -2.5457e-02,  1.9112e-03,  4.5873e-04,  1.0444e-03,
         9.4807e-03,  7.4842e-03, -1.2675e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3331e-02, -6.0144e-01, -7.1091e-02, -6.5572e-02,  4.6952e-03,
         2.0954e-04, -3.6968e-03, -4.5491e-03,  2.3666e-02, -1.4561e-02,
         2.6675e-02, -2.5283e-02, -1.6417e-03,  9.3689e-03,  8.6153e-03,
        -1.1778e-02,  7.5558e-02,  1.8264e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0545, -0.5739, -0.0422, -0.0470,  0.0069,  0.0031, -0.0392, -0.0446,
        -0.0320, -0.0006, -0.0095, -0.0121, -0.0471, -0.0136,  0.0073, -0.0250,
        -0.0136,  0.0279,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0049,  0.6703,  0.0400, -0.0070,  0.0738,  0.0113, -0.0124, -0.0354,
        -0.0264, -0.0286, -0.0211,  0.0058,  0.0097,  0.0149, -0.0385,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.8791e-02,  5.8558e-01,  4.9001e-02,  3.4029e-02, -6.1397e-03,
         9.0023e-03,  1.3435e-02,  1.2842e-03,  5.8033e-03,  1.9889e-02,
         9.3681e-04, -5.4767e-04, -1.2468e-03, -2.3658e-03, -3.9294e-03,
         1.0624e-02,  1.5023e-02, -3.6202e-02,  4.1596e-04, -1.9436e-02,
         7.6264e-03, -4.8778e-03, -1.2848e-03,  2.3236e-03, -2.3537e-02,
         7.6635e-03,  3.5294e-03, -1.5295e-03, -7.6182e-03,  5.3970e-03,
         5.4133e-02,  1.6795e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.0570e-02,  7.0515e-01, -4.2214e-02,  2.3288e-02,  1.5643e-02,
        -2.8938e-03, -5.7311e-03, -6.4606e-03,  8.4702e-03, -2.1842e-03,
        -6.0218e-03, -3.5597e-04, -1.0666e-02,  6.4384e-03, -1.7612e-03,
        -1.3379e-02, -7.7671e-03, -3.3216e-03,  2.4494e-03,  5.8470e-04,
        -6.8094e-04, -7.1087e-03, -2.9707e-02,  7.8828e-03,  1.2659e-02,
         1.8999e-03,  9.4002e-03,  1.7734e-04, -6.1006e-03,  6.8642e-03,
        -6.8750e-04,  5.2105e-03,  6.2663e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4601e-02, -5.5163e-01, -3.7148e-02, -5.6924e-02, -1.3412e-03,
        -2.2633e-03,  8.5299e-03,  3.4629e-03, -4.8559e-03, -3.0910e-03,
        -1.2396e-02, -1.5191e-02, -2.1659e-02,  1.1444e-02,  2.0321e-02,
        -4.4346e-03, -9.5330e-03, -5.4518e-03, -1.3367e-02, -1.2011e-03,
        -4.1734e-03, -7.3118e-03,  4.7325e-03, -1.8684e-02,  1.1800e-03,
         2.3474e-03, -2.1224e-05,  3.1416e-03, -7.3765e-03, -7.9760e-04,
        -5.2929e-03,  3.7555e-03, -8.0868e-03,  8.6631e-03,  1.1732e-03,
         4.7517e-03, -2.5134e-03, -5.1528e-03, -6.3699e-04, -3.9022e-05,
         3.3511e-03, -3.5094e-04, -4.9291e-04, -5.8931e-03, -5.0568e-03,
         2.4062e-03, -1.0371e-03, -2.6659e-03, -2.5929e-03, -3.0093e-03,
        -1.1385e-02, -5.0072e-04, -2.0913e-03, -3.4636e-03, -5.4130e-04,
        -1.8641e-03, -1.4706e-03, -2.3150e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0147, -0.7148, -0.0099, -0.0358, -0.0081,  0.0019,  0.0142, -0.0105,
         0.0172,  0.0101,  0.0137,  0.0059,  0.0044,  0.0284,  0.0043,  0.0119,
        -0.0081,  0.0027,  0.0164, -0.0108, -0.0029, -0.0010,  0.0057,  0.0133,
         0.0107,  0.0226,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.6233e-02,  6.6319e-01,  4.5851e-03, -2.9241e-02, -9.2981e-03,
        -2.1035e-02,  1.5265e-02,  1.8523e-02, -2.7861e-03,  8.1226e-04,
         2.2298e-02,  2.2822e-03,  8.5964e-03,  1.8854e-02, -1.0475e-03,
        -1.6598e-03,  1.2413e-02, -1.5114e-03, -1.8205e-02, -4.1360e-03,
         6.0794e-05, -8.4535e-03, -2.5305e-02,  1.6345e-02,  1.7865e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4313e-02,  5.9845e-01,  4.6438e-02, -2.8537e-02,  1.1195e-02,
         1.8306e-02, -3.3024e-03, -1.0831e-02,  1.2847e-03, -3.6653e-03,
         8.5147e-03, -1.0318e-02, -5.1979e-03,  4.9419e-03, -1.2959e-02,
        -2.8115e-02, -1.3221e-02, -3.0938e-04, -6.3774e-03, -1.2163e-02,
        -2.4808e-03,  3.2616e-03, -6.5013e-03,  7.8986e-03, -6.7088e-03,
        -1.1754e-02, -4.3874e-03,  1.2613e-02,  6.6194e-03, -1.6737e-04,
        -2.0759e-03, -2.3684e-03, -6.4304e-04, -4.6638e-04,  8.9256e-03,
        -1.1145e-04,  8.2939e-03, -6.0738e-03, -2.9961e-03,  4.7786e-03,
        -2.1596e-03, -1.0759e-02,  1.1234e-02,  6.7845e-03, -1.4939e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0599,  0.6595, -0.0040,  0.0353, -0.0122,  0.0140,  0.0031, -0.0068,
         0.0065,  0.0009, -0.0161, -0.0144, -0.0169,  0.0053, -0.0041,  0.0134,
         0.0258, -0.0047,  0.0082, -0.0100, -0.0007,  0.0176, -0.0606,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0029,  0.6069,  0.0089, -0.0644, -0.0084,  0.0561,  0.0143, -0.0125,
        -0.0207,  0.0589,  0.1254, -0.0208,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 0.0335, -0.6330, -0.0990, -0.0021, -0.0074, -0.0611, -0.0013, -0.0222,
         0.0023, -0.0084,  0.0249, -0.0029, -0.0112, -0.0142, -0.0085,  0.0024,
         0.0026, -0.0056, -0.0204, -0.0053, -0.0078, -0.0038, -0.0202,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0508e-01,  4.1588e-01, -7.7588e-02, -5.8073e-03,  3.1073e-02,
        -2.0666e-04,  1.0157e-03,  7.7137e-03,  1.3546e-02, -1.5144e-04,
        -1.0362e-02, -7.8177e-03,  5.4279e-03,  2.5869e-03, -1.9540e-02,
        -4.2482e-03,  1.2837e-03, -9.8969e-03, -7.3754e-05, -1.1720e-02,
         4.2860e-03, -2.2751e-02,  5.1126e-03,  3.9292e-03,  4.9329e-03,
        -8.0298e-03,  8.0999e-03, -7.7594e-04, -1.6484e-02, -1.3388e-02,
         5.9251e-03,  1.2474e-01,  5.0524e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8422e-02, -7.3130e-01,  7.9739e-03,  4.2967e-02, -7.0146e-04,
         1.5668e-02,  3.0859e-02,  1.2483e-02, -2.3117e-02,  1.6176e-02,
        -7.2766e-03,  6.6217e-05,  2.6340e-02, -1.2097e-02,  1.1816e-02,
         1.3411e-02,  7.4005e-03, -1.2695e-03, -6.6101e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1126, -0.6360,  0.0137, -0.0340, -0.0016, -0.0446, -0.0136, -0.0177,
         0.0096, -0.0088, -0.0273,  0.0044,  0.0039, -0.0722,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1185, -0.5895, -0.0480,  0.0152,  0.0682, -0.0164,  0.0025,  0.0106,
         0.0136, -0.0011, -0.0099,  0.0067,  0.0033,  0.0233,  0.0099, -0.0021,
         0.0037,  0.0168,  0.0092,  0.0066, -0.0186,  0.0063,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4682e-02,  6.6024e-01,  7.5043e-03,  1.7074e-02,  2.6753e-02,
         9.5236e-03, -1.6585e-02,  2.1636e-03, -8.5519e-03,  4.3064e-03,
        -1.2576e-02, -2.0649e-02, -2.0955e-02, -1.0796e-02, -1.1278e-02,
         3.8956e-03,  4.7671e-02,  5.8444e-04, -5.4183e-03,  4.6635e-03,
        -2.4541e-03, -1.3017e-03,  3.3482e-03,  1.0030e-03, -1.1786e-02,
         3.9382e-03, -5.3927e-03, -5.8576e-03, -8.1724e-03, -4.0871e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0240,  0.6723,  0.0037, -0.0378, -0.0124,  0.0160, -0.0320,  0.0092,
         0.0180, -0.0031,  0.0051,  0.0119,  0.0099, -0.0381,  0.0137, -0.0336,
        -0.0591,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2431, -0.0113,  0.1271,  0.0350,  0.0168, -0.0128, -0.1014, -0.0707,
         0.0267,  0.1704,  0.1846,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0368, -0.6658, -0.1190, -0.0237, -0.0010,  0.0108,  0.0123, -0.0151,
        -0.0108, -0.0197, -0.0036,  0.0043, -0.0068,  0.0009, -0.0356,  0.0147,
         0.0192,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4658e-02,  5.5161e-01,  9.7486e-02,  3.4789e-02,  1.6008e-02,
         2.6352e-03,  9.4841e-03,  1.8684e-04, -3.1990e-03, -1.2436e-02,
         4.5161e-03,  1.0809e-02, -6.1745e-03,  1.9921e-02,  1.3036e-02,
        -7.6448e-03, -9.2871e-04, -1.3269e-02,  2.3212e-03, -2.2707e-03,
        -6.5113e-03, -1.9022e-02,  6.8412e-03, -8.6269e-03, -8.6131e-02,
         9.4849e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5747e-02,  6.3506e-01,  5.3667e-02, -5.3980e-03,  8.9579e-03,
         1.4715e-02, -2.1949e-02,  3.9018e-02, -2.3852e-02, -1.7418e-05,
        -4.2150e-03, -4.8274e-03,  1.4302e-03,  7.8960e-03, -1.3460e-02,
         8.0377e-03,  1.3646e-02,  5.5849e-03,  1.4527e-03,  1.4361e-02,
        -8.2545e-03, -6.1223e-03, -2.0096e-03,  2.5263e-03,  7.3010e-04,
         9.9881e-03, -2.6989e-03, -1.6033e-02,  3.1741e-03, -3.5172e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0345, -0.6406,  0.0346,  0.0331,  0.0088, -0.0050,  0.0067,  0.0063,
        -0.0313,  0.0135, -0.0174,  0.0486, -0.0059,  0.0091,  0.0334, -0.0021,
         0.0242, -0.0420, -0.0028,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([-3.0367e-02,  7.1721e-01,  4.5179e-02,  1.5158e-02,  1.8133e-02,
        -1.6154e-02,  8.8732e-03, -8.6896e-03,  5.5969e-03,  5.6652e-03,
        -1.0967e-03, -5.3325e-03,  1.0760e-03, -2.3620e-03, -9.3058e-03,
        -1.9996e-03, -7.9905e-03,  1.7613e-02, -4.7938e-03, -1.6512e-03,
        -4.3131e-03, -3.6522e-03,  5.5185e-03, -2.8808e-04, -4.8757e-03,
         2.0923e-02,  3.6181e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1953, -0.4008,  0.0068,  0.0602, -0.0364, -0.0247, -0.0289, -0.0535,
         0.0083, -0.0179,  0.0418,  0.0112, -0.1137,  0.0006,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5195e-03, -7.4496e-01, -6.1173e-03, -2.2820e-02, -2.6098e-02,
        -1.9099e-03, -1.4780e-02,  2.9646e-03,  1.3933e-02,  2.7049e-02,
         7.6164e-03, -3.0233e-04,  3.2321e-03,  1.0628e-02, -1.0707e-02,
        -8.9829e-04,  7.5781e-03,  1.1899e-02,  2.1953e-02, -2.2854e-03,
         6.0180e-03,  2.7475e-02,  3.0551e-03,  7.6149e-03, -1.2971e-03,
        -1.1289e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1342, -0.6729,  0.0142, -0.0022,  0.0054, -0.0027, -0.0065,  0.0060,
        -0.0071, -0.0150,  0.0157, -0.0398, -0.0691, -0.0092,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0386,  0.6490,  0.0148, -0.0033, -0.0082, -0.0025, -0.0191, -0.0182,
         0.0018, -0.0080, -0.0128, -0.0204, -0.0065, -0.0211, -0.0092, -0.0039,
        -0.0229, -0.0077,  0.0014, -0.0073, -0.0082,  0.0053, -0.0154, -0.0095,
        -0.0097, -0.0020, -0.0052,  0.0083, -0.0037,  0.0053, -0.0156,  0.0350,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0496,  0.7229,  0.0144,  0.0081, -0.0062,  0.0201, -0.0075,  0.0068,
        -0.0023, -0.0086, -0.0112,  0.0062, -0.0050, -0.0040,  0.0042,  0.0012,
        -0.0051, -0.0290,  0.0066,  0.0085, -0.0034, -0.0030, -0.0113, -0.0396,
         0.0152,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0065, -0.8018, -0.1295, -0.0265, -0.0022, -0.0335,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0677e-01,  6.3289e-01,  3.3692e-02, -3.0940e-02,  7.2554e-04,
         1.6380e-04, -9.2661e-03, -6.5096e-03, -2.1441e-03, -2.5087e-03,
        -1.0607e-02,  1.0645e-02, -6.0191e-03,  9.4413e-04, -1.9687e-02,
        -1.5831e-02,  5.0031e-03,  3.9777e-03, -3.1016e-03,  7.4235e-04,
        -1.0451e-02, -1.3642e-02, -4.7750e-03, -5.3793e-03, -3.8453e-03,
        -1.0028e-02, -5.7730e-03,  3.4396e-03,  3.4066e-06,  3.2489e-03,
         1.7745e-03,  1.6832e-03,  6.4294e-03, -2.0388e-03, -7.9635e-03,
        -7.6949e-03,  9.6673e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2872e-03, -7.0532e-01, -4.5740e-02, -4.0705e-02, -2.2241e-03,
         3.9187e-03, -3.1766e-03,  1.5068e-02, -3.5403e-03,  3.9716e-03,
         1.2178e-02, -4.2944e-04, -1.0630e-02,  8.0139e-03, -4.3632e-03,
         6.5955e-03,  4.8985e-03,  9.5471e-03, -6.6864e-04,  2.1938e-03,
         1.9540e-02,  4.0192e-04,  9.0094e-03,  1.1870e-02,  4.0592e-03,
        -1.5848e-03, -8.8332e-03, -5.2931e-03,  5.0775e-03, -9.5350e-04,
         1.7576e-03, -3.7419e-03, -6.2608e-03,  1.5382e-02,  1.4768e-02,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7298e-01,  5.0127e-01, -1.8653e-02, -1.4076e-02,  6.1114e-03,
         2.0611e-02, -1.8243e-02, -3.3967e-04, -3.2910e-02, -1.9044e-02,
         3.9439e-03,  1.6136e-03, -6.0800e-03,  5.0590e-02, -3.3538e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.0457e-02, -7.0225e-01, -3.1306e-02,  1.4900e-02, -6.8543e-04,
        -2.2600e-02, -7.4275e-03, -1.7884e-02,  1.3139e-02,  6.2702e-03,
         7.7688e-05,  1.3041e-02,  4.7901e-03,  6.3214e-03, -6.0777e-04,
        -1.5993e-03, -2.5139e-03,  1.9373e-02, -3.2356e-03, -5.7146e-04,
         2.1477e-03,  2.1009e-02,  4.0994e-03, -2.5303e-03, -2.1650e-03,
        -5.4222e-03,  4.1499e-03, -5.9104e-03,  1.0959e-02, -4.7457e-03,
         5.3868e-03,  1.6966e-02, -5.4609e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1759e-04, -6.0493e-01,  1.6575e-02,  3.1240e-02, -1.1495e-02,
         3.5335e-02,  1.5012e-02, -5.6695e-03,  2.2425e-02, -5.5576e-03,
        -2.0755e-02, -2.0929e-02,  1.1430e-02,  2.6185e-02,  2.2782e-02,
         7.3221e-04,  1.0076e-02,  1.0371e-02,  4.8876e-03, -1.1894e-02,
         1.4315e-02,  1.0657e-02,  3.8930e-03, -5.0746e-03, -1.6388e-03,
         4.7056e-04,  3.6563e-03,  5.2176e-02, -1.9224e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 0.0308,  0.6774, -0.0135, -0.0418,  0.0280, -0.0099, -0.0195, -0.0077,
        -0.0183, -0.0370, -0.0041,  0.0012, -0.0384, -0.0115, -0.0609,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.2237e-02,  4.3684e-01,  8.8092e-03, -4.5500e-03, -3.7456e-03,
        -9.0740e-03,  2.2171e-03,  2.5967e-02,  1.0134e-02, -1.5584e-02,
         1.6103e-02,  5.4829e-03,  3.6478e-03,  6.5207e-04,  9.4312e-03,
        -3.1656e-03,  4.5305e-03,  2.7962e-02, -2.6493e-04,  8.4833e-04,
         5.2891e-03, -4.7787e-03,  4.6897e-03, -2.1512e-02,  5.5374e-03,
        -1.3712e-02, -1.8223e-03,  8.4957e-03,  4.7460e-02,  3.2611e-03,
         1.9170e-03, -4.3277e-03,  1.8665e-02, -2.1276e-03,  2.3043e-03,
         8.9624e-03,  1.0723e-03,  1.7776e-03, -8.5346e-03,  2.3719e-03,
        -9.1164e-04, -2.2518e-03,  8.8420e-04,  9.7226e-03, -1.0478e-02,
         9.1396e-05, -4.7255e-03, -1.2683e-01,  1.4240e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7762e-02, -6.0328e-01,  1.4425e-02,  6.0353e-02,  2.3619e-02,
         5.5776e-03, -2.8703e-02,  9.0591e-03,  1.5467e-02,  3.1015e-02,
        -4.8689e-03,  1.3324e-02,  3.9850e-03, -2.6777e-03, -1.1970e-03,
         9.8200e-04,  5.5547e-03, -9.2218e-04, -6.9979e-03,  1.1620e-02,
         7.7693e-03,  2.3350e-03, -2.8416e-04,  8.8140e-03,  4.0137e-03,
         5.6682e-03, -3.3595e-03, -6.3699e-03,  3.3669e-04,  1.5744e-02,
        -5.6559e-03, -4.6016e-03, -5.2212e-03,  4.2013e-02,  6.4262e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0016,  0.7527, -0.0072,  0.0212,  0.0099, -0.0174, -0.0390, -0.0222,
         0.0165, -0.0092,  0.0037,  0.0047, -0.0100,  0.0090, -0.0061,  0.0195,
        -0.0164,  0.0236, -0.0102,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8996e-02,  7.2864e-01, -6.3019e-03,  3.9329e-02,  1.0418e-02,
         5.1446e-03,  8.2019e-05, -8.0019e-04,  1.8075e-02,  1.3169e-02,
        -2.6539e-02, -1.0482e-02,  2.6398e-02, -3.9717e-02, -3.5908e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.2294e-02,  6.5490e-01, -4.2005e-02,  1.2366e-02,  4.9500e-03,
        -1.0845e-02,  5.2513e-02,  1.3781e-02, -1.4827e-04, -1.8065e-02,
        -1.4011e-02, -1.9291e-05, -1.7294e-02, -1.1317e-02, -3.1381e-03,
        -5.2156e-03, -1.4127e-02, -4.4991e-03, -2.7640e-03,  2.7328e-03,
         2.3422e-02,  2.9597e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.6916e-02,  6.5660e-01,  2.6859e-02,  2.4685e-02, -2.0574e-03,
        -9.6264e-03, -2.9433e-02, -8.8496e-03, -1.7585e-02,  1.2476e-02,
        -1.5094e-03,  8.5023e-03, -5.1925e-04, -6.7392e-04,  4.5413e-03,
        -7.4749e-04, -1.5374e-02, -1.5368e-02,  1.2980e-03,  4.2793e-03,
         6.9819e-03,  1.5726e-02,  3.6426e-03, -6.2131e-03, -2.2442e-02,
         8.8750e-03,  2.1510e-02, -4.9860e-03, -6.3759e-03,  1.5348e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0653,  0.6021, -0.0747, -0.0463,  0.0123,  0.0084, -0.0104,  0.0103,
        -0.0482, -0.0794, -0.0427,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0930, -0.6043, -0.0270,  0.0060,  0.0305, -0.0025, -0.0134,  0.0106,
         0.0170,  0.0119,  0.0221, -0.0111,  0.0135,  0.0027, -0.0025,  0.0049,
        -0.0151,  0.0109,  0.0011,  0.0015,  0.0557,  0.0428,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2856e-02, -6.3943e-01, -8.0676e-02, -1.7573e-02, -1.3017e-02,
        -1.7253e-02, -1.5913e-02,  6.0238e-03, -7.3160e-04, -1.2785e-03,
         6.9966e-03,  2.5233e-02, -1.5531e-02, -1.0291e-02, -1.7238e-03,
         1.5409e-03, -8.0040e-03, -6.2596e-04, -7.0363e-03, -3.6890e-04,
         4.7008e-03, -3.0206e-04,  1.5148e-02,  4.8213e-03,  4.0366e-03,
        -2.7338e-03, -4.4968e-03,  9.5238e-03, -1.9769e-03,  1.8500e-03,
        -1.4952e-03,  4.4444e-03, -2.2362e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0396, -0.6398, -0.0602,  0.0102,  0.0275, -0.0188,  0.0275,  0.0013,
        -0.0043,  0.0185,  0.0084,  0.0075,  0.0019,  0.0017,  0.0071, -0.0095,
        -0.0098,  0.0148, -0.0028, -0.0011, -0.0066,  0.0010, -0.0055,  0.0112,
        -0.0047, -0.0110,  0.0270,  0.0204,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5729e-02,  6.4692e-01,  9.5221e-02,  1.9524e-02,  9.8206e-03,
         8.2661e-03,  1.1667e-02,  2.1134e-02, -2.7118e-02,  8.1529e-03,
        -1.4509e-02,  9.3037e-03, -5.3148e-04, -1.2788e-02, -5.7204e-04,
        -5.9677e-03, -6.1199e-03, -7.0688e-04, -4.9108e-03,  3.8352e-03,
        -3.5818e-04,  7.1278e-03,  9.5325e-04, -1.4849e-02, -2.3910e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 6.9275e-03,  6.2084e-01,  2.2317e-02,  7.1572e-02,  4.0782e-03,
        -1.1624e-02, -2.4325e-02, -4.5384e-03,  7.3348e-04, -9.0198e-04,
         3.4093e-03, -9.5035e-04, -5.9861e-03, -2.9251e-02,  1.1338e-03,
        -2.3714e-02, -1.6115e-02, -1.0683e-02, -4.8953e-03,  1.1828e-03,
         7.1293e-05,  2.4558e-03,  2.4379e-03,  1.1053e-02, -3.8565e-03,
        -1.1425e-02,  7.8027e-03, -3.4506e-03, -8.0212e-03, -6.3393e-03,
        -5.8545e-03, -5.7085e-03,  1.8127e-03, -4.3415e-03,  3.0676e-03,
        -7.3551e-03, -5.1885e-03, -4.3790e-03, -6.4006e-03,  1.1608e-02,
         2.2193e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0708,  0.7384,  0.0068, -0.0090,  0.0259, -0.0072,  0.0051,  0.0056,
         0.0028,  0.0017, -0.0204,  0.0092,  0.0083, -0.0018,  0.0032,  0.0015,
        -0.0086, -0.0020,  0.0009,  0.0032, -0.0121, -0.0321, -0.0066, -0.0122,
         0.0041,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.9868e-04,  6.4561e-01,  1.0125e-01, -2.0771e-03,  2.3759e-02,
        -5.0728e-04, -2.2626e-02,  3.5977e-02,  1.3955e-02,  1.0550e-03,
         8.1891e-03, -1.0896e-02, -9.7115e-03,  1.1610e-03, -5.9578e-03,
         2.8098e-03, -7.3733e-04, -5.1299e-03, -2.6439e-02,  5.2897e-02,
         2.8453e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6570e-02,  7.7841e-01, -2.7854e-03, -1.1603e-02, -3.3313e-02,
        -1.2049e-02, -3.0764e-03, -6.4710e-03, -1.1950e-02,  7.9780e-03,
         8.2195e-04,  2.8348e-03, -1.0385e-02, -1.8057e-02, -1.8916e-02,
        -2.9477e-02, -6.0548e-04,  1.4694e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4271e-02,  5.9206e-01,  3.7106e-02, -8.6609e-03,  1.5069e-02,
        -3.0637e-03, -3.7874e-03, -1.3054e-02, -3.4967e-02, -5.8207e-03,
        -3.3439e-03,  4.2075e-03, -6.3872e-03,  3.2697e-03,  4.3890e-03,
        -2.3383e-03, -3.1802e-02, -1.0679e-02, -8.8175e-03,  8.2143e-03,
        -5.0892e-03, -1.8927e-03, -3.6801e-03, -3.5270e-03, -4.0599e-03,
        -1.1261e-02, -8.2499e-03,  6.9699e-03,  3.7555e-03,  6.1387e-03,
         4.6136e-04,  4.2237e-03,  4.4201e-03,  1.3610e-03, -1.5361e-02,
         7.3333e-03, -2.6673e-02,  2.4238e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.2408e-02, -7.3034e-01,  3.1833e-05, -2.1531e-02, -4.8170e-03,
        -9.1593e-03,  2.2868e-02, -1.0680e-02, -3.4569e-02,  1.8086e-02,
        -1.6689e-02,  4.2453e-03, -9.5318e-04,  8.3336e-03, -2.3230e-02,
        -6.1979e-03, -5.1598e-03, -3.7564e-03,  9.9542e-03,  6.9952e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0043,  0.7405,  0.0464, -0.0150, -0.0227,  0.0341,  0.0036,  0.0037,
        -0.0061, -0.0103,  0.0129, -0.0056, -0.0910, -0.0038,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0530e-01,  5.6392e-01, -4.9051e-02, -1.4801e-02, -3.0173e-02,
        -7.6270e-03,  1.1744e-02,  5.3461e-03,  9.6842e-03,  3.0727e-02,
         1.1919e-02,  2.1498e-03,  7.6625e-04, -3.7161e-03,  2.5060e-02,
        -1.3306e-02, -3.4805e-03, -1.2500e-02, -5.3168e-04, -9.2132e-03,
         7.9439e-03,  5.4453e-03,  4.4167e-03,  1.4316e-02, -2.4257e-03,
        -5.9227e-03, -3.3832e-03, -5.4430e-03,  4.7000e-03,  3.2215e-03,
        -1.3582e-02, -1.8184e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.7959e-02,  7.0003e-01, -4.5063e-04,  2.9243e-02,  3.7299e-02,
         1.4034e-03,  1.8149e-02,  2.9394e-02,  3.1876e-03, -2.8992e-02,
        -3.3045e-03, -1.9522e-02,  4.6780e-03,  1.3626e-02, -1.4658e-02,
         1.1217e-02,  6.1022e-03, -3.5735e-03, -9.6730e-03,  7.5377e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5752e-01,  4.3019e-01,  3.2624e-02,  2.8737e-02, -4.6341e-03,
         2.4234e-02, -9.8270e-04,  1.3211e-03, -5.2502e-03,  9.3643e-04,
         2.6203e-03,  4.2328e-03,  6.7376e-04,  2.0579e-03,  8.9228e-04,
         4.5240e-03, -1.7044e-03,  3.2868e-03,  6.2409e-04,  7.0979e-04,
         3.5943e-03,  6.9703e-02, -1.2101e-02, -8.1010e-03, -7.1185e-03,
         1.0569e-03,  1.7629e-04,  6.7126e-03, -5.8250e-03, -3.7146e-03,
        -9.7224e-03,  2.2256e-03,  8.1807e-03, -1.2656e-03,  1.3666e-03,
         1.9221e-03, -4.5529e-03,  3.3676e-03, -3.9308e-03, -8.0430e-04,
        -1.3297e-02,  4.4583e-03,  1.0321e-02,  2.0122e-04,  1.6455e-02,
        -4.1597e-02,  5.0474e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.9748e-02, -8.0033e-01, -1.1606e-02,  1.7436e-03, -7.4331e-03,
        -6.1398e-03,  9.1861e-04,  3.2215e-03,  4.3013e-02, -1.4433e-03,
         3.0330e-02,  8.2582e-03,  1.3502e-02, -1.6950e-03, -6.2039e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.8498e-02,  6.1123e-01,  7.9972e-02,  9.9529e-03,  4.2989e-02,
         1.0697e-02, -1.3148e-02,  8.7563e-04, -4.3349e-03,  6.1475e-04,
        -8.3600e-03, -4.6262e-03, -2.5609e-03, -3.6444e-03, -9.4355e-03,
        -4.1169e-03, -1.3589e-03, -2.7818e-04,  8.1688e-03, -6.5744e-03,
        -5.0864e-04, -7.1139e-03, -3.1580e-03,  4.1951e-03, -2.6338e-03,
        -8.1840e-03, -7.1610e-03,  3.2889e-02, -1.0472e-02,  5.0585e-03,
         5.4058e-03,  6.4797e-04,  4.3529e-03, -8.0408e-05,  5.6334e-03,
         2.9526e-03, -3.6217e-03, -1.7372e-03, -3.1610e-03,  1.2210e-03,
        -2.1252e-03, -2.9070e-03, -5.1514e-03,  4.8217e-04,  3.2918e-03,
        -3.4655e-03,  7.0386e-03, -3.9168e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-1.1158e-03, -7.0012e-01, -6.3726e-02, -3.3172e-02, -2.1826e-02,
        -7.6962e-03,  5.8222e-03,  1.5989e-02,  7.2626e-03, -1.5382e-02,
         4.5269e-03,  2.0861e-03,  1.4272e-02, -5.7672e-04,  9.5992e-04,
         2.7752e-03,  1.5247e-02, -3.2997e-03,  1.6652e-02, -3.1450e-03,
        -3.0564e-03, -2.3495e-04, -9.3970e-03,  9.0586e-03,  2.4958e-02,
        -1.7642e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0569, -0.5742, -0.0717, -0.0245, -0.0344,  0.0097,  0.0209, -0.0188,
         0.0039,  0.0044, -0.0233,  0.0113,  0.0060,  0.0044,  0.0138,  0.0066,
         0.0020,  0.0094,  0.0271, -0.0604,  0.0162,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3253e-02,  5.6693e-01,  3.6596e-02, -2.7990e-02,  1.5877e-02,
        -3.2195e-02, -1.6855e-02, -1.7526e-02, -1.6795e-02, -1.8499e-02,
        -4.1382e-03, -2.9995e-04, -7.0941e-03,  5.6426e-04, -1.0319e-02,
        -1.8800e-02, -1.6020e-02, -1.6536e-02, -1.2638e-02, -2.7637e-03,
        -3.0649e-03, -3.9120e-03, -1.0275e-02, -9.8739e-04,  2.0919e-02,
        -1.8922e-02, -5.9020e-03,  2.1101e-03,  2.8231e-04, -3.3170e-03,
        -2.4779e-02, -1.1721e-02,  2.3227e-03,  2.5207e-03, -2.5669e-02,
         1.1611e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8623e-02, -7.0425e-01, -1.5448e-02, -3.4165e-02, -3.1583e-02,
        -4.0678e-03, -2.1267e-02,  3.8504e-03,  1.4782e-02,  8.7233e-03,
         1.4980e-02, -3.1924e-03,  3.3275e-04,  5.0047e-03, -2.3794e-03,
        -1.4365e-02,  4.9093e-03, -1.6549e-03, -2.2139e-02,  4.4283e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0407, -0.6712, -0.0467, -0.0027, -0.0385,  0.0129,  0.0018,  0.0050,
        -0.0103,  0.0184, -0.0126, -0.0099,  0.0192,  0.0246, -0.0030,  0.0085,
         0.0238,  0.0076, -0.0080,  0.0072,  0.0273,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0378, -0.7162, -0.0649,  0.0298,  0.0178,  0.0198,  0.0299,  0.0393,
        -0.0065,  0.0170, -0.0037,  0.0013,  0.0160,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.2368e-02,  6.2757e-01,  7.0213e-02,  1.7233e-03,  1.2409e-02,
        -2.0883e-02, -4.5574e-03, -5.4063e-03, -5.8545e-03, -3.4894e-03,
         2.3054e-02, -4.8181e-03, -3.8462e-03, -8.7287e-03,  7.8226e-03,
         3.5509e-03, -2.3927e-02,  1.6881e-02, -3.2780e-03, -1.1598e-03,
        -6.3812e-03,  6.3324e-03, -6.1259e-04,  1.0578e-02,  1.7437e-03,
         4.7387e-03, -1.0569e-02,  3.8607e-03, -5.0976e-04, -1.9833e-03,
        -6.6734e-04,  2.0897e-03,  4.3257e-03, -7.1617e-03,  1.3351e-02,
        -4.0700e-03,  1.9492e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0311,  0.7295, -0.0207,  0.0200, -0.0012,  0.0017,  0.0023,  0.0213,
        -0.0092, -0.0201, -0.0058, -0.0019, -0.0226, -0.0096, -0.0083, -0.0056,
         0.0061, -0.0016, -0.0059,  0.0095,  0.0181,  0.0066,  0.0144,  0.0269,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.0494e-02,  6.8053e-01,  6.6050e-02, -1.2218e-02,  1.4265e-02,
         1.3617e-02, -5.5163e-04, -8.4941e-03,  8.3820e-03, -4.2647e-04,
        -1.7175e-02,  4.9735e-03,  2.2801e-03,  3.1845e-03,  7.0370e-03,
         5.7783e-03,  2.6294e-03, -5.1068e-03, -4.3827e-03, -1.1277e-03,
         5.0076e-03, -8.4849e-03,  8.1032e-03,  3.0843e-03, -9.6510e-03,
         3.9004e-03, -1.0558e-03,  2.6669e-03,  3.4558e-03,  1.4602e-03,
        -1.6609e-02, -7.8202e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5290e-02,  6.1702e-01,  4.9055e-02,  1.2870e-02,  3.8223e-02,
         1.0995e-02,  1.1332e-02, -8.4463e-03,  1.4382e-02, -9.1057e-03,
        -7.8617e-03, -8.2433e-03, -8.2500e-03, -1.2547e-03,  2.4878e-04,
        -2.2987e-03, -2.1344e-02, -8.0210e-03,  3.8558e-03,  6.3628e-03,
         2.6174e-02, -2.4187e-02, -6.0232e-03,  7.6686e-03,  3.4720e-03,
        -7.8455e-03, -2.2637e-03,  8.5136e-04, -8.8815e-03,  2.6231e-02,
         1.9453e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0455, -0.6762,  0.0296,  0.0058, -0.0323,  0.0085,  0.0226,  0.0198,
        -0.0044,  0.0056, -0.0064, -0.0010,  0.0055,  0.0178,  0.0066,  0.0035,
        -0.0247,  0.0134, -0.0028,  0.0098,  0.0050, -0.0101, -0.0431,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1382,  0.3213, -0.0068, -0.0131,  0.0207,  0.0039,  0.0344, -0.0153,
        -0.0180,  0.0086, -0.0054, -0.0312,  0.0020, -0.0349, -0.0048,  0.0145,
         0.0539,  0.0065, -0.0161, -0.0074,  0.0069, -0.0015, -0.0033,  0.0075,
        -0.0057, -0.0080,  0.2072, -0.0029,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #100: [tensor([-0.1059,  0.6034,  0.0142,  0.0074,  0.0249,  0.0180,  0.0136,  0.0010,
         0.0016, -0.0186,  0.0065, -0.0090, -0.0008, -0.0008, -0.0032, -0.0150,
         0.0107,  0.0222, -0.0028,  0.0017,  0.0052, -0.0028, -0.0141, -0.0047,
         0.0090,  0.0087,  0.0178,  0.0323, -0.0239,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.5020e-03, -6.8045e-01,  2.6151e-02,  1.1606e-03, -1.6231e-02,
        -2.7926e-02, -4.5928e-03,  9.4568e-03,  5.0762e-03,  4.1830e-03,
        -1.0714e-02,  3.5786e-02,  6.4420e-03,  6.6716e-03,  6.6574e-03,
         1.1543e-02,  2.6502e-04,  8.4085e-03, -9.3244e-03,  1.5167e-02,
         2.0400e-02,  3.2002e-02,  2.1633e-02, -3.3256e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0184,  0.6133,  0.0577,  0.0449, -0.0035, -0.0010,  0.0010,  0.0253,
         0.0033, -0.0110, -0.0139,  0.0273,  0.0035,  0.0037, -0.0059, -0.0111,
        -0.0023,  0.0068, -0.0026,  0.0089, -0.0040,  0.0075, -0.0059,  0.0016,
        -0.0011, -0.0112, -0.0071, -0.0028, -0.0075, -0.0103, -0.0122,  0.0034,
        -0.0407, -0.0193], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0521,  0.7156,  0.0580, -0.0298,  0.0177,  0.0218,  0.0042,  0.0063,
         0.0103,  0.0114, -0.0056, -0.0101, -0.0022,  0.0036,  0.0035, -0.0008,
        -0.0130, -0.0071,  0.0019, -0.0148,  0.0008,  0.0094,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.8165e-02,  6.4838e-01,  4.6135e-02,  1.0929e-04,  2.5269e-02,
        -1.5917e-02, -9.9204e-03, -1.1740e-02, -2.5766e-02, -1.1691e-02,
         3.8488e-02, -1.6282e-02, -1.8235e-02,  3.0079e-03, -6.9269e-03,
        -4.7037e-03, -4.6036e-02, -3.2294e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0991,  0.0212,  0.0418,  0.0234,  0.0024,  0.0418, -0.0817, -0.0139,
         0.0263, -0.0784, -0.0072, -0.0064, -0.0201,  0.0329, -0.0451,  0.0398,
        -0.0655, -0.0116, -0.0396, -0.0197, -0.0253,  0.2554, -0.0012,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0451,  0.1675, -0.0353,  0.0800,  0.0130, -0.1400, -0.0558, -0.0303,
         0.0360, -0.1129,  0.0748, -0.0379,  0.0308,  0.0657,  0.0348, -0.0171,
         0.0231,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0818,  0.6305,  0.0899, -0.0062,  0.0113,  0.0129,  0.0144, -0.0070,
        -0.0019, -0.0011,  0.0155, -0.0131, -0.0031, -0.0033,  0.0060,  0.0205,
         0.0537, -0.0280,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0524e-02,  6.8766e-01, -4.9283e-02, -1.9481e-02,  1.6187e-03,
        -1.5576e-02, -4.2995e-03, -1.9621e-02,  2.2901e-03,  3.3643e-03,
        -1.1145e-02,  1.4060e-02,  2.4448e-03,  2.6020e-04, -4.2281e-03,
        -1.0615e-02, -7.8999e-03,  3.5170e-03,  1.3633e-02, -2.0795e-02,
         7.6809e-03, -5.8865e-03,  4.3008e-03, -2.8321e-02, -2.5613e-02,
        -1.5887e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0531,  0.5990,  0.0313, -0.0206, -0.0095, -0.0134,  0.0153, -0.0157,
        -0.0318, -0.0189, -0.0281, -0.0015,  0.0056,  0.0098, -0.0039, -0.0052,
        -0.0106, -0.0052, -0.0016, -0.0020,  0.0147,  0.0042, -0.0013, -0.0016,
         0.0013, -0.0043, -0.0141, -0.0090,  0.0107, -0.0311,  0.0254,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3336e-02, -6.9624e-01, -4.9376e-02,  1.8551e-03, -7.0350e-03,
        -1.9122e-02, -2.4934e-02, -2.0016e-02,  3.9389e-05,  3.3212e-02,
         7.8142e-03, -3.8949e-03,  2.9006e-02, -9.7262e-03,  1.3901e-02,
        -7.2348e-03, -1.2679e-02, -2.0576e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0855,  0.6828,  0.0413, -0.0839, -0.0026,  0.0392, -0.0604, -0.0043,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 6.7755e-02,  6.3050e-01,  3.7271e-02, -2.0954e-02, -3.3218e-02,
         2.8671e-02,  5.6728e-03,  2.4243e-03, -6.1603e-03,  1.1045e-02,
         5.2101e-04,  2.9450e-03,  1.0625e-03,  7.1633e-03,  9.6884e-03,
         2.0516e-02,  1.1608e-02, -3.3874e-03,  5.6576e-03,  2.9361e-03,
         8.5834e-03,  5.2676e-03,  6.0043e-03, -2.8084e-03, -1.8804e-03,
        -4.9467e-03, -3.3551e-03, -1.5158e-03, -2.2157e-03,  1.0934e-02,
        -2.9995e-03, -5.8004e-04, -2.2571e-03,  8.5457e-04, -7.6237e-03,
        -1.9113e-02,  1.5732e-03, -7.8927e-03,  4.3866e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3246e-02, -6.8338e-01,  4.5281e-02, -8.2702e-03, -2.2498e-02,
         5.2777e-03, -5.5499e-03, -2.1272e-03, -2.2756e-03, -4.0818e-04,
        -4.7229e-03, -2.6707e-03, -3.3954e-02, -3.3232e-03, -1.0445e-02,
         2.1968e-03,  3.8616e-03,  1.9383e-02,  3.6237e-03,  1.0460e-02,
         9.5418e-04,  5.0597e-03,  5.6244e-03,  1.7267e-03, -2.1485e-02,
         2.1654e-02, -1.8337e-02, -2.2065e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0513, -0.6735,  0.0338,  0.0199,  0.0177,  0.0172,  0.0082, -0.0029,
         0.0010,  0.0010, -0.0402, -0.0803,  0.0529,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0945, -0.6635,  0.1011,  0.0153,  0.0058, -0.0026,  0.0009, -0.0025,
         0.0146,  0.0138,  0.0046,  0.0113,  0.0192,  0.0255,  0.0248,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.4868e-02,  5.5905e-01,  2.1787e-02, -5.8175e-02,  2.1640e-02,
        -9.8038e-03, -5.4555e-03, -9.0540e-03, -3.4641e-02, -7.0115e-03,
         7.5789e-03,  2.5622e-03,  4.1706e-03,  1.3773e-02,  1.1148e-02,
         6.2307e-03,  4.9767e-03, -2.0353e-03,  1.2131e-02, -5.6415e-03,
        -2.6091e-03, -5.4863e-03,  1.5468e-03, -5.8239e-03, -7.4124e-03,
         2.4571e-04, -4.8063e-03, -1.0271e-03, -4.7217e-03, -1.4613e-03,
         8.2990e-04, -1.7854e-03, -1.7319e-03,  1.0332e-03,  1.0859e-03,
        -4.5158e-03, -4.5824e-03, -5.6161e-03, -4.5909e-04, -8.6501e-03,
        -1.0383e-02, -2.9856e-03, -5.8101e-03, -3.8033e-03, -1.0939e-03,
         8.5896e-03,  1.0172e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9848e-02,  7.6634e-01,  2.6643e-02,  2.2319e-02,  2.9323e-04,
         5.9872e-03,  2.4834e-03, -2.7997e-03,  1.3395e-02,  5.7476e-03,
         2.3341e-03,  2.7059e-03, -2.8606e-03, -5.7927e-03, -3.5245e-03,
        -6.3108e-03,  7.0772e-04, -1.5003e-03,  1.2573e-03,  1.6072e-03,
        -1.0481e-03, -2.2612e-03,  1.1354e-02,  5.8071e-03,  4.5627e-03,
        -6.5626e-04,  1.1382e-02,  1.0097e-03, -1.0015e-02, -2.2223e-03,
         4.8684e-03, -1.1637e-03,  6.4814e-03,  5.6187e-03,  4.7273e-03,
         2.9569e-03,  4.7091e-03,  3.2718e-03, -1.4439e-03,  5.3082e-03,
         1.0686e-02, -3.9930e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0603,  0.6445, -0.0011, -0.0258, -0.0077, -0.0063, -0.0211, -0.0186,
        -0.0085, -0.0042,  0.0124,  0.0108, -0.0222, -0.0061, -0.0010, -0.0025,
        -0.0018, -0.0095, -0.0049, -0.0134,  0.0055,  0.0007, -0.0125, -0.0446,
        -0.0542,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1369e-02,  7.6571e-01,  2.7694e-02,  3.1112e-04, -9.1101e-03,
        -7.7982e-03, -2.3646e-02, -2.7782e-03, -9.2665e-03,  1.1968e-04,
         3.4236e-03,  4.2119e-04,  3.3882e-03, -9.4411e-03, -3.0271e-02,
        -1.4358e-02, -7.4476e-03, -8.1767e-03, -7.1551e-03, -1.6682e-03,
         4.0340e-03,  3.4367e-03, -8.9724e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4470e-02,  7.2766e-01,  8.1878e-02, -3.0804e-03, -7.0668e-03,
         3.8435e-03,  3.5595e-03, -3.4006e-03,  2.4990e-03,  1.7485e-02,
         1.1582e-02,  1.2002e-02,  1.1943e-02,  2.8608e-03,  7.1139e-03,
         6.8165e-04, -2.8060e-03,  1.3658e-02,  1.7076e-03, -1.2074e-02,
         5.8562e-03,  3.2457e-02,  3.1392e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0704, -0.6042,  0.0191, -0.0138,  0.0040,  0.0023,  0.0534, -0.0610,
         0.0061, -0.0196,  0.0257,  0.0940,  0.0265,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0961, -0.7197, -0.0308,  0.0385,  0.0141,  0.0039,  0.0065,  0.0099,
         0.0161, -0.0155,  0.0068, -0.0023,  0.0085,  0.0312,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0012, -0.7042, -0.0610,  0.0204,  0.0057, -0.0008,  0.0138,  0.0141,
         0.0150,  0.0126,  0.0187,  0.0197,  0.0056, -0.0038,  0.0012,  0.0088,
         0.0222,  0.0115,  0.0597,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 1.6262e-02,  6.9681e-01,  6.2919e-03, -4.7168e-02, -1.9817e-02,
         7.5929e-03, -2.5019e-03,  4.6721e-02, -1.8097e-02,  1.2353e-02,
        -7.9138e-03, -1.4107e-04, -1.2469e-03, -7.0228e-03, -4.9181e-04,
         3.8860e-03,  9.6438e-03,  5.9121e-03,  7.8055e-03,  6.2197e-03,
        -9.3713e-03,  4.6200e-03, -9.0154e-03,  7.1076e-03, -3.8543e-02,
        -7.4452e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0635, -0.7229, -0.0648, -0.0187,  0.0010,  0.0083,  0.0098,  0.0331,
         0.0062,  0.0153,  0.0079,  0.0036,  0.0145, -0.0081, -0.0134,  0.0089,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0646, -0.5558, -0.0398, -0.0341, -0.0038, -0.0639, -0.0026, -0.0477,
         0.0191, -0.0067,  0.0013, -0.0138, -0.0070,  0.0035, -0.0275, -0.0569,
        -0.0517,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0216,  0.6667, -0.0763, -0.1456,  0.0276,  0.0191, -0.0125, -0.0091,
         0.0216,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7151e-02, -4.9233e-01, -8.9805e-02,  5.5078e-02, -2.0928e-02,
        -1.4845e-03, -2.0168e-03,  2.0875e-03, -1.0359e-02, -7.0419e-03,
        -7.3090e-03, -1.3282e-03,  6.3263e-03,  1.8650e-04,  5.3556e-04,
        -5.3042e-04, -2.4896e-03,  1.7026e-03, -2.3106e-03,  1.5058e-02,
         3.7247e-03,  9.0072e-03,  7.8351e-03,  1.6490e-02,  1.0112e-02,
         1.1751e-02,  4.9613e-05,  8.1460e-03, -4.3841e-03,  2.4560e-03,
        -3.3919e-03, -6.7262e-03, -3.7300e-03,  2.1283e-03, -1.9315e-02,
        -1.2572e-02,  6.4333e-03, -3.3509e-03,  1.0612e-03,  6.3601e-03,
         1.1385e-02, -5.8497e-03,  1.3998e-02,  1.3169e-02,  5.3041e-04,
        -4.9981e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0287e-01,  5.6999e-01, -5.1213e-02,  2.5893e-02, -8.2642e-03,
        -2.8612e-03,  1.0036e-02,  1.3890e-02, -2.6665e-04,  7.0692e-03,
        -1.6175e-03,  7.0526e-03, -1.1796e-02,  1.8648e-03,  5.8083e-03,
         9.2188e-03,  9.2012e-03,  1.9654e-02, -3.0636e-03,  7.3322e-03,
         3.2963e-03, -1.8941e-03, -7.9352e-03, -4.8937e-03, -9.3846e-04,
         1.6160e-03,  1.2807e-02, -1.0001e-02, -5.4075e-02,  3.3581e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0119, -0.7578, -0.0447, -0.0850,  0.0254, -0.0285,  0.0468,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1000, -0.7286,  0.0484, -0.0121,  0.0047, -0.0175, -0.0074, -0.0561,
         0.0081, -0.0129, -0.0044,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0344, -0.7948, -0.0547, -0.0219, -0.0105, -0.0215,  0.0056,  0.0025,
         0.0038,  0.0160,  0.0056, -0.0214, -0.0043,  0.0029,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7314e-02, -6.8542e-01, -4.0069e-02, -2.3615e-02, -2.1229e-02,
        -6.0068e-03,  3.2677e-03,  9.9510e-03,  6.4076e-03,  9.9077e-03,
         4.7355e-03, -3.0336e-03,  1.3154e-02,  1.6471e-02, -5.7841e-04,
         9.3340e-03, -3.4855e-03, -8.4243e-03,  5.8331e-03, -6.8316e-05,
        -4.1965e-03, -2.9269e-03, -7.3593e-03, -3.4460e-04,  9.0667e-04,
        -2.6773e-03,  7.9380e-03, -5.8915e-03, -2.2321e-03,  1.2228e-02,
         2.7462e-02,  5.5574e-03,  9.7213e-03, -2.2522e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.2446e-02,  5.5117e-01,  8.2270e-02,  1.5694e-02,  1.2841e-02,
         1.9126e-02,  3.2120e-03, -5.7813e-04,  7.6901e-03, -1.0677e-02,
         4.7333e-03,  5.7956e-05, -9.9239e-03,  6.7294e-03, -1.9230e-02,
        -1.5356e-02, -3.7229e-03, -2.6299e-02, -1.1862e-02, -5.9606e-03,
        -1.3986e-02,  2.3345e-03,  9.2047e-04,  1.8973e-02,  2.0373e-02,
        -4.4239e-03,  6.9906e-02, -9.4996e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0138,  0.6670,  0.0372,  0.0225,  0.0072, -0.0034,  0.0037,  0.0031,
        -0.0029,  0.0041,  0.0086, -0.0142,  0.0358,  0.0544,  0.0121, -0.0134,
         0.0053,  0.0023, -0.0095, -0.0057, -0.0092,  0.0139, -0.0169, -0.0338,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 0.1132,  0.5993, -0.0262,  0.0298,  0.0038,  0.0030,  0.0141, -0.0163,
        -0.0032, -0.0032, -0.0081,  0.0059,  0.0202,  0.0172,  0.0008,  0.0106,
        -0.0049, -0.0075, -0.0022, -0.0014, -0.0046, -0.0088, -0.0019, -0.0021,
        -0.0036,  0.0262, -0.0016, -0.0018,  0.0127, -0.0277, -0.0180,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.7143e-04,  8.0327e-01, -9.2981e-04, -6.9129e-02, -7.9781e-03,
         5.5432e-02,  1.7495e-04,  1.1401e-02,  8.2166e-03,  1.0964e-02,
         1.5843e-02,  1.5689e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.0593e-03,  7.9093e-01,  2.9254e-02, -5.5555e-02,  4.6679e-03,
        -3.5720e-03, -5.5020e-03,  9.3495e-04,  1.5679e-02, -3.7325e-02,
         5.0103e-03, -2.0748e-02,  1.1636e-02, -1.1492e-02, -6.2918e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7419e-02,  6.8364e-01,  6.7337e-02, -1.6511e-02,  2.8961e-02,
        -1.4287e-02, -1.5257e-02, -1.3683e-03, -3.9591e-03, -1.7737e-02,
        -9.7971e-03, -3.4811e-03,  1.6947e-03,  4.3033e-03, -3.3974e-04,
        -9.5054e-03, -3.0672e-04, -8.9348e-03, -1.0399e-03,  2.8917e-03,
        -6.4443e-03, -1.0422e-03, -5.4878e-03, -1.2734e-03,  4.7941e-03,
         1.5623e-03,  3.8156e-03, -8.5737e-03, -1.4093e-03, -1.4391e-02,
        -1.5547e-03, -6.6231e-03, -4.2629e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3304e-02,  7.2118e-01,  8.4310e-02, -9.0300e-03,  1.0985e-02,
         2.0013e-03, -1.8561e-02, -1.2520e-02, -7.7326e-03, -2.4952e-04,
        -2.4104e-03, -1.3720e-02, -7.6148e-04,  3.8101e-02, -8.8701e-03,
        -3.6268e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0420, -0.6202,  0.2143,  0.0144, -0.0122, -0.0007, -0.0810,  0.0008,
        -0.0145,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1432,  0.6745, -0.0730,  0.0842,  0.0251,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.3237e-02,  7.1936e-01,  1.5946e-02,  1.3429e-02, -8.7647e-03,
        -1.8284e-02, -1.8904e-02, -8.2413e-03,  8.6269e-03, -6.2091e-03,
        -4.6911e-03,  7.9053e-03,  5.8249e-03, -5.1243e-02,  7.1198e-03,
         4.1922e-03,  1.6854e-03,  3.4498e-03,  9.6152e-04, -1.2571e-04,
         5.1709e-03, -6.3402e-03,  3.4412e-03,  1.3866e-02,  1.4727e-03,
         1.7798e-02,  3.7121e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0909, -0.6695, -0.0682, -0.0351, -0.0416, -0.0837,  0.0108,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2883e-02,  6.9989e-01, -2.0662e-02,  9.8311e-03,  3.0952e-02,
         9.0934e-03,  4.5073e-02,  1.0889e-04,  2.7995e-03, -3.2863e-02,
         9.8569e-04,  3.8785e-04,  3.7305e-03, -1.2120e-02, -5.1401e-03,
         9.5312e-03, -1.8062e-02,  8.5336e-03,  9.8017e-03,  2.6325e-02,
        -9.2434e-03,  1.9778e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0275,  0.7708,  0.0136,  0.0261,  0.0230,  0.0113,  0.0033, -0.0035,
        -0.0046, -0.0185,  0.0239,  0.0074,  0.0050,  0.0098, -0.0097,  0.0078,
        -0.0020, -0.0178, -0.0145,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0611e-02,  7.2529e-01,  9.5441e-03, -4.6701e-04,  1.9675e-02,
         1.5904e-02,  7.0668e-03,  2.5919e-02,  2.1163e-02, -3.8546e-03,
         3.0505e-03, -5.0169e-03, -5.4741e-03,  1.3848e-02,  8.8692e-03,
        -3.8262e-03, -4.6706e-03,  4.3525e-03, -6.7916e-03,  2.3830e-03,
         8.9343e-03, -8.6467e-03, -3.6923e-03,  1.7125e-02,  2.9992e-03,
         1.8289e-03, -6.2950e-03,  7.9964e-03, -8.2868e-03, -1.3239e-02,
         1.3181e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 1.3463e-02, -5.9925e-01, -1.1309e-01,  4.0045e-02, -2.7696e-02,
         7.5575e-03, -7.0417e-03,  1.1282e-02,  2.1217e-03,  6.4359e-03,
        -3.0262e-03,  2.4212e-04, -7.8413e-04, -9.0373e-04, -2.2967e-02,
         3.4430e-03, -4.1193e-03, -2.5382e-02,  2.0742e-02, -2.2296e-03,
         4.4081e-03,  6.6073e-03, -2.5185e-03,  4.3769e-03, -2.5935e-02,
         1.3132e-03,  5.3461e-03,  1.4156e-02, -5.0528e-03,  1.4049e-02,
        -1.1401e-03,  3.2788e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0566,  0.6798,  0.0177,  0.0016, -0.0438,  0.0071,  0.0304,  0.0019,
         0.0410,  0.0017, -0.0230,  0.0259,  0.0080, -0.0029,  0.0053, -0.0113,
         0.0012,  0.0028,  0.0049, -0.0028,  0.0038,  0.0266,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.8983e-02,  6.3034e-01,  4.8710e-02,  7.3185e-03, -1.8570e-03,
         5.6748e-03,  8.4594e-03,  1.0947e-02,  6.9756e-04, -6.8687e-03,
        -1.1124e-02, -7.6467e-03, -4.0767e-02, -8.3516e-03, -2.0390e-02,
         4.1598e-03, -5.2970e-03,  5.7369e-03, -7.5648e-03,  1.9757e-03,
        -1.1149e-02, -8.5866e-03,  5.6675e-04, -7.3895e-03, -6.5686e-03,
         8.5907e-03, -7.9739e-03, -3.0521e-02,  2.9166e-03, -2.2873e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0048,  0.7652,  0.0386, -0.0149, -0.0048,  0.0164, -0.0270, -0.0034,
        -0.0135, -0.0068, -0.0102,  0.0161, -0.0369,  0.0114, -0.0091,  0.0097,
         0.0050, -0.0063,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.5465e-02,  6.7067e-01,  4.1916e-02,  1.3835e-02,  1.0366e-02,
        -6.9102e-04,  2.0477e-03, -1.8184e-03,  8.2239e-05, -2.5118e-03,
        -1.3019e-03, -2.8333e-03, -3.1828e-03,  1.2542e-02, -6.9208e-03,
        -1.4580e-02,  3.2348e-03, -1.3173e-02,  2.2455e-03, -8.2015e-03,
        -4.6818e-03,  7.2620e-04, -7.8897e-03, -4.3471e-03,  3.0342e-03,
        -1.3976e-03,  2.9813e-03, -1.5719e-03, -3.8131e-03,  5.0164e-04,
        -2.0758e-03, -1.4282e-03,  2.3058e-03, -1.0053e-02,  5.3163e-03,
        -3.0345e-02,  9.9121e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1025,  0.5438, -0.0911, -0.0532, -0.0272,  0.0118, -0.0166,  0.0061,
         0.0068,  0.0058, -0.0090,  0.0306, -0.0143,  0.0020, -0.0069,  0.0006,
        -0.0013, -0.0032,  0.0055, -0.0051, -0.0048,  0.0013, -0.0473, -0.0033,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3015e-02, -7.1873e-01, -7.8042e-03, -3.0110e-02, -2.0479e-02,
        -8.6912e-04, -8.4690e-03, -3.7545e-03,  2.4874e-03,  5.3896e-03,
         1.0870e-02, -6.4756e-03,  9.7537e-03,  9.4481e-03,  8.8452e-04,
         6.5351e-03,  9.1531e-04, -5.2945e-04, -1.2420e-03,  1.0214e-03,
         1.9098e-04,  9.5762e-04, -3.0673e-03,  6.6066e-03, -7.0602e-03,
        -1.5797e-02, -6.8964e-03, -4.5677e-03, -6.8272e-03,  5.7705e-03,
        -1.1163e-03, -1.4331e-03, -6.8866e-03, -2.4411e-03, -6.9608e-03,
         4.2053e-03,  1.4800e-02,  9.2147e-03, -4.1948e-03,  8.5865e-04,
         1.4209e-03,  5.1311e-03,  7.5447e-04,  2.5436e-03, -5.8711e-04,
        -2.0925e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0829,  0.6388, -0.0308,  0.0018, -0.0131, -0.0068, -0.0049, -0.0058,
         0.0060, -0.0158,  0.0082,  0.0069,  0.0252, -0.0012, -0.0086,  0.0062,
         0.0244, -0.0092,  0.0018, -0.0031,  0.0154, -0.0652,  0.0181,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1014, -0.7051,  0.0053, -0.0630,  0.0211, -0.0268,  0.0155,  0.0297,
         0.0009,  0.0165, -0.0147,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0809,  0.6397,  0.0569, -0.0495, -0.0231, -0.0303,  0.0244, -0.0260,
         0.0094,  0.0081, -0.0010, -0.0015, -0.0288,  0.0203,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0533,  0.7016,  0.0280, -0.0312,  0.0110, -0.0109,  0.0142,  0.0291,
        -0.0075,  0.0325,  0.0596,  0.0212,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0883e-02, -6.0338e-01, -8.3140e-02, -3.0962e-02, -1.3118e-03,
         2.0830e-02,  2.6481e-03, -7.4952e-03, -3.6019e-03,  5.1557e-03,
         7.2550e-03,  3.0135e-03, -1.4554e-04, -3.8001e-03,  3.4347e-03,
         6.5012e-03,  8.7637e-04,  5.0583e-03, -5.8785e-04,  1.7975e-02,
         1.2355e-05, -2.2463e-03, -3.0014e-03, -6.5171e-03, -2.0814e-03,
         7.1897e-03, -1.9716e-02,  6.8325e-03, -9.8523e-04,  2.4597e-03,
         4.1495e-03, -5.7782e-03,  3.7458e-03, -1.0745e-02,  6.0217e-04,
         1.5030e-03, -8.2685e-04, -4.5718e-04,  3.8788e-03,  3.1812e-03,
        -1.3209e-04,  1.2765e-03, -2.8857e-03, -8.2726e-04,  2.5785e-03,
         2.3868e-03, -6.1340e-04,  9.0237e-04,  2.5896e-03,  3.5521e-03,
        -7.4351e-04,  7.5969e-04, -3.0860e-04,  5.5752e-03,  3.0898e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 8.6724e-02, -6.2274e-01, -2.4289e-02, -3.9719e-02, -2.4492e-02,
        -2.3990e-02,  1.7569e-04,  1.6283e-04,  1.9545e-02, -3.1910e-03,
         8.0893e-03, -2.2694e-02, -1.0736e-02, -5.9773e-03, -1.7670e-02,
        -1.2053e-02,  1.0956e-02, -1.4231e-02,  2.9739e-03, -1.3954e-02,
         5.8538e-03, -1.2424e-02,  1.7360e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4146e-03,  6.1557e-01,  1.3020e-01,  8.4622e-02,  1.0869e-02,
         1.3552e-02,  1.1228e-02,  1.9804e-02, -1.4823e-02, -2.2467e-03,
        -1.7330e-02,  9.9283e-03,  3.4192e-03, -2.1060e-02,  9.7381e-04,
         5.8385e-04, -2.4523e-02,  1.3856e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0088,  0.7095,  0.0266, -0.0253,  0.0205, -0.0118,  0.0062,  0.0298,
        -0.0183, -0.0044,  0.0103,  0.0224,  0.0438,  0.0037,  0.0046, -0.0076,
         0.0065, -0.0398,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1145,  0.6568, -0.0271, -0.0343, -0.0139,  0.0019, -0.0044, -0.0600,
        -0.0183, -0.0097, -0.0111, -0.0145, -0.0037, -0.0180, -0.0117,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0698e-03,  6.7652e-01,  5.3651e-02,  2.8611e-02,  2.8781e-03,
         2.0411e-02,  1.8854e-02,  8.1068e-03,  2.2347e-02,  9.6507e-03,
         9.4050e-03, -7.8905e-03,  7.6595e-03,  1.2591e-03, -1.0354e-02,
        -1.0060e-02,  6.8522e-03,  1.7312e-02,  1.5673e-04, -1.1450e-02,
         6.2889e-03, -1.9716e-02, -3.0124e-03,  2.0381e-03,  3.1366e-03,
        -6.3632e-03,  1.2017e-02, -9.6838e-04, -4.0388e-03,  1.3900e-03,
         8.3795e-03, -3.1523e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4050e-02,  6.9032e-01,  5.3553e-02,  1.9052e-02,  3.3767e-02,
         5.7259e-03, -2.1460e-03,  8.5822e-04, -7.8462e-03, -4.5814e-03,
        -1.0339e-02, -9.6678e-03, -1.9350e-03, -2.5357e-02, -4.6459e-03,
        -4.5030e-04, -5.2062e-03,  2.3394e-03, -1.5827e-02, -1.3216e-02,
        -8.5005e-03, -1.5531e-03, -2.0751e-02, -8.7280e-03,  1.7146e-03,
         7.1980e-03,  4.0847e-03,  1.0424e-03,  1.2252e-04, -2.9955e-03,
         6.4843e-03,  2.1663e-04,  1.5729e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1767e-02, -6.0247e-01, -2.8886e-02, -3.9739e-02, -2.0072e-02,
        -7.7594e-03, -1.1081e-02, -5.8817e-03,  2.4074e-03,  1.4975e-03,
        -9.6634e-04, -2.5842e-03,  2.1235e-02,  1.4855e-02,  1.3679e-02,
         1.9602e-05,  3.4941e-03,  5.2851e-03, -6.5205e-03,  1.7539e-03,
         3.4473e-03,  4.6170e-03,  3.7657e-03, -1.1430e-02, -4.2540e-03,
         6.8192e-03,  2.0856e-03, -3.6516e-03, -1.5812e-03, -2.6189e-03,
        -7.1173e-03, -4.4683e-03,  1.7662e-03, -2.7141e-03,  3.6308e-03,
         2.8308e-03,  5.9472e-04,  1.6246e-03, -4.6091e-04, -3.4897e-03,
         1.8722e-04, -2.2053e-03, -8.4805e-05,  5.5860e-03, -4.7799e-04,
        -5.1835e-03, -3.6193e-03, -1.7352e-03, -7.0420e-03,  3.2421e-03,
         1.8545e-03, -5.2298e-03, -3.4772e-03,  2.0573e-03, -5.3845e-03,
        -4.6641e-03,  6.7998e-03,  1.6245e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0568e-01,  7.4795e-01, -6.3009e-03, -2.5439e-03, -9.5922e-04,
         1.2055e-02,  1.0365e-02,  6.3883e-03,  2.8849e-03, -3.9213e-03,
        -5.6224e-03,  3.1293e-03, -1.4622e-03, -1.0002e-02,  1.8511e-03,
        -4.2337e-03, -4.2858e-03, -2.1015e-03, -5.4704e-03, -2.0920e-02,
        -7.9666e-03, -1.6952e-04, -5.3512e-03, -2.1810e-03, -1.7860e-02,
        -8.3475e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0282,  0.7272,  0.0215, -0.0093, -0.0139,  0.0241,  0.0052, -0.0139,
         0.0092, -0.0097, -0.0012, -0.0098,  0.0040, -0.0025,  0.0024,  0.0036,
         0.0112,  0.0086,  0.0052,  0.0084, -0.0024, -0.0157, -0.0043, -0.0231,
         0.0355,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.2404e-03,  5.9645e-01,  3.7374e-02,  1.6903e-02,  5.3611e-03,
        -5.9526e-03, -1.8584e-02, -8.2622e-03,  3.1692e-03, -2.9349e-02,
        -1.4324e-02, -7.0293e-03,  1.6989e-02,  8.8744e-03, -2.3617e-03,
        -1.5234e-02, -7.0452e-03,  1.0458e-02,  1.5007e-04,  1.4110e-03,
         8.1921e-03, -7.1900e-03, -1.1373e-03,  9.6163e-03, -9.9170e-03,
        -1.1334e-02,  1.3836e-02,  1.5173e-03,  8.8431e-03, -1.2717e-02,
         1.9295e-03,  2.1796e-04, -4.8818e-03, -1.1455e-03, -7.7234e-03,
         4.8392e-03,  1.4170e-02, -1.0787e-02,  8.4457e-04,  1.0369e-02,
        -8.2146e-04, -3.7394e-03,  6.5395e-05,  1.9879e-02, -1.9761e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2884e-02,  6.7836e-01, -1.1370e-02,  3.5447e-02,  4.1850e-03,
         1.3161e-02,  6.5480e-03,  1.5580e-02, -3.7972e-02, -1.4166e-02,
        -7.9521e-03,  7.3828e-03, -1.4582e-02, -6.3657e-03, -6.2239e-03,
         8.6370e-03, -1.6623e-02, -4.1098e-04,  5.6007e-03, -2.6992e-03,
         6.6475e-03, -4.3566e-02, -2.3638e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0771,  0.7450, -0.0615,  0.0021,  0.0073,  0.0088, -0.0015, -0.0064,
         0.0077, -0.0143, -0.0265,  0.0419,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-2.6515e-01,  6.6424e-02, -1.5000e-04,  1.8455e-02, -6.3622e-03,
         1.2161e-02, -3.6479e-02, -2.6723e-02,  1.2222e-02,  9.4530e-03,
         1.1070e-02, -2.6427e-03,  1.1301e-02,  9.6689e-03, -3.0499e-04,
         1.0549e-01,  1.2812e-02, -3.7666e-02,  3.0369e-03,  4.5478e-03,
        -1.8026e-02,  2.4671e-01,  8.3145e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0943e-02, -6.7279e-01,  4.1041e-02,  1.2222e-02, -4.8116e-03,
         8.8753e-03,  1.8471e-03, -1.0355e-02,  4.8664e-03, -1.2126e-02,
         1.0437e-02,  6.0523e-03, -6.0017e-03, -8.2592e-03,  2.1883e-02,
         5.7755e-03,  7.3978e-03,  2.2429e-02,  5.4497e-04, -9.2135e-03,
        -9.1440e-03,  4.5422e-04,  8.1077e-03,  8.0748e-03, -3.9820e-03,
        -2.0921e-03, -6.4546e-03, -4.6023e-03,  8.9917e-03, -2.7175e-04,
        -3.1661e-03,  1.0195e-02, -3.6588e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0271, -0.7617, -0.0160,  0.0095,  0.0054,  0.0094,  0.0254, -0.0141,
        -0.0191, -0.0021,  0.0158,  0.0048,  0.0018, -0.0087, -0.0108,  0.0140,
        -0.0044,  0.0271,  0.0229,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1124,  0.6804, -0.0545,  0.0122, -0.0331,  0.0459, -0.0064, -0.0111,
         0.0041, -0.0131,  0.0085, -0.0038,  0.0106,  0.0038,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0091, -0.7151, -0.0724, -0.0196,  0.0339,  0.0126,  0.0091, -0.0077,
        -0.0050, -0.0036,  0.0019, -0.0007, -0.0064, -0.0122,  0.0057,  0.0049,
        -0.0192, -0.0030,  0.0021, -0.0116,  0.0287,  0.0154,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5690e-02, -6.3794e-01, -4.2366e-02, -5.1203e-02, -3.9533e-02,
        -3.2156e-03,  1.1569e-03,  3.9426e-02, -1.8807e-03, -1.1739e-02,
         1.2760e-02, -8.4463e-03,  2.3598e-02,  4.1002e-03,  6.0424e-04,
        -3.4957e-04,  7.9905e-03, -1.4871e-02, -6.3092e-03,  3.8513e-03,
        -3.0669e-03, -1.2502e-03,  1.9457e-03,  2.0850e-03,  5.5977e-03,
        -1.5791e-05, -4.6598e-03,  1.0675e-02, -8.1862e-03, -1.5486e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1187e-03,  6.3729e-01,  2.4833e-02, -8.2484e-02, -4.7159e-02,
         5.1758e-02,  1.0319e-02, -1.7471e-02, -2.6025e-02, -6.8808e-03,
        -6.0936e-03, -3.6273e-03, -1.2601e-03, -4.2926e-04,  1.0382e-02,
        -3.9315e-02, -3.1557e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0624, -0.6719, -0.0277, -0.0476,  0.0294, -0.0121,  0.0090, -0.0196,
        -0.0435,  0.0071,  0.0698,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.3942e-02,  6.9948e-01,  8.4183e-02,  1.3302e-02, -8.8109e-03,
        -2.9391e-02, -8.0296e-03, -1.5693e-02,  1.8935e-02,  1.9979e-03,
        -2.7192e-03, -4.8584e-03,  2.9518e-03, -1.5113e-03,  4.3079e-02,
         8.8607e-04,  2.2525e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3668e-03,  7.2821e-01,  3.5770e-02,  1.0828e-02,  5.1618e-02,
         1.2709e-02,  1.1580e-02,  6.7476e-03, -3.4334e-03, -7.5812e-03,
        -1.4349e-03, -1.8050e-03, -2.5625e-03,  2.2979e-03, -2.4833e-02,
         3.6105e-03, -2.3155e-04, -5.3097e-03,  3.6186e-03,  2.9807e-03,
         7.0937e-04, -3.2925e-03,  1.0508e-02,  6.7702e-03,  2.0729e-02,
        -3.4462e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1561e-01,  5.6678e-01,  5.6059e-03, -6.7335e-02, -3.3274e-02,
         1.4966e-02, -2.3606e-02, -8.0680e-03, -9.9882e-03, -2.0450e-03,
         8.2698e-03, -1.0431e-02, -1.0287e-02, -5.1379e-03, -2.7336e-03,
        -6.5408e-03, -1.6396e-03, -2.4224e-03,  6.1979e-04, -7.9586e-03,
        -2.6530e-03, -3.3005e-03, -9.1385e-03,  5.8707e-03, -1.2374e-02,
        -3.1903e-03, -4.7337e-03, -5.5492e-03, -4.9772e-02,  9.4380e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0103, -0.8022,  0.0087,  0.0238,  0.0208,  0.0047,  0.0084,  0.0065,
        -0.0125, -0.0088, -0.0154,  0.0230, -0.0054,  0.0069, -0.0030, -0.0068,
         0.0046,  0.0178, -0.0103,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([-2.9278e-02,  6.9567e-01,  3.1824e-02,  9.2654e-03,  2.4847e-02,
         4.9664e-03,  3.1005e-03,  3.4004e-03,  2.2706e-02,  3.7154e-03,
        -1.1169e-03, -9.7265e-03, -4.2636e-03,  5.9482e-03, -5.1177e-03,
        -1.8993e-02,  1.2176e-02,  2.0641e-02,  3.8067e-03,  1.3569e-02,
        -6.6444e-04, -1.2335e-02,  7.7105e-03, -8.2915e-03, -1.2580e-02,
        -2.2878e-02,  1.1414e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.9277e-02, -6.7756e-01, -1.8907e-03,  3.9428e-02, -4.3466e-02,
        -2.8633e-02, -2.0086e-02, -1.3312e-02,  4.3877e-02,  1.4614e-02,
         4.4495e-04, -1.0171e-03, -4.0970e-02,  1.5421e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0180,  0.7039,  0.0402,  0.0286,  0.0049,  0.0266, -0.0060, -0.0114,
         0.0370,  0.0059,  0.0024, -0.0077, -0.0024,  0.0143,  0.0039, -0.0051,
         0.0048,  0.0067, -0.0021,  0.0045, -0.0088, -0.0082, -0.0029, -0.0045,
         0.0354, -0.0041,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0644e-01,  6.1974e-01, -1.8684e-03, -2.3626e-02, -3.9515e-03,
        -8.6982e-03, -6.1634e-03, -1.3071e-02, -1.7760e-02,  7.8554e-02,
        -2.6687e-04, -4.7112e-02, -4.0194e-02,  3.2551e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7514e-02,  7.6910e-01,  2.4993e-02, -4.6964e-03,  1.9292e-03,
        -2.3944e-03,  1.9377e-03, -1.4581e-02, -2.8256e-03, -4.0192e-03,
        -1.0788e-02, -9.8994e-03, -3.9081e-03,  3.1300e-03,  6.4871e-03,
        -1.2260e-03, -3.2170e-03, -1.0015e-03,  6.0873e-04, -1.2348e-02,
        -2.4286e-03,  6.0490e-03, -8.3597e-04,  1.6598e-03,  5.7855e-03,
        -7.3055e-03, -7.7673e-03, -1.0175e-03,  5.2288e-03,  1.0808e-03,
         2.0895e-02,  2.3338e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1680e-02, -7.2410e-01, -2.7733e-02,  2.9525e-02,  9.4617e-03,
        -1.2047e-02,  2.4603e-02,  7.6969e-03, -4.8306e-03, -1.7486e-05,
        -1.4128e-02, -1.0466e-03, -5.4911e-03, -2.6333e-03, -2.4909e-03,
         1.3552e-03,  1.3988e-03,  1.4881e-02, -2.3818e-03, -1.3694e-03,
        -1.5326e-03, -3.9688e-03,  5.7642e-03, -2.8702e-02, -1.1157e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1647,  0.7464, -0.0059, -0.0171, -0.0469, -0.0190,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.5049e-02,  6.5384e-01,  5.9253e-02, -9.7367e-03,  2.1098e-02,
         6.7607e-03,  1.0270e-02, -1.4847e-02, -3.9397e-03, -2.5257e-03,
         7.7594e-05, -1.8618e-03,  4.0697e-03,  3.7271e-02, -3.2205e-03,
        -3.3278e-03, -2.7599e-03,  5.0407e-03, -1.5696e-04, -5.1030e-03,
        -8.0533e-03, -8.6945e-03,  3.1352e-03, -2.7657e-03, -7.2588e-03,
        -2.5376e-03,  2.4343e-03,  9.9518e-03, -1.3660e-03, -9.1472e-04,
        -1.1283e-03,  3.4709e-03, -2.5455e-03, -3.7401e-03, -2.9620e-03,
         2.8265e-03,  1.6007e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0375, -0.6537, -0.0164, -0.0436,  0.0111, -0.0115, -0.0274, -0.0053,
        -0.0056, -0.0030,  0.0044, -0.0057,  0.0037,  0.0143, -0.0051,  0.0074,
         0.0049, -0.0137, -0.0073,  0.0051,  0.0118,  0.0086,  0.0072,  0.0054,
         0.0079, -0.0061, -0.0046, -0.0074,  0.0076,  0.0029,  0.0031,  0.0010,
        -0.0047, -0.0133,  0.0220,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0305,  0.5428, -0.1828, -0.0372,  0.0086,  0.0094,  0.0082,  0.0023,
         0.0088,  0.0121, -0.0055, -0.0150, -0.0359,  0.0767, -0.0242,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1893e-02, -7.0867e-01, -1.6471e-02,  2.2893e-02, -1.3884e-03,
         1.5403e-02, -2.3801e-02, -2.9212e-02,  1.3975e-03,  1.4573e-04,
        -5.8188e-03,  7.2550e-04, -3.2549e-03,  5.5034e-03, -1.7599e-03,
         5.2904e-03, -2.8660e-03,  1.7229e-02,  1.4232e-02,  1.9994e-02,
         3.4928e-03,  1.3324e-02, -3.8579e-03,  9.8170e-04,  1.0208e-02,
         1.5533e-02,  6.7225e-03,  4.4837e-03,  8.0048e-03, -4.0682e-03,
         2.7216e-04,  3.2677e-03, -7.8318e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8839e-02, -6.2686e-01,  7.1464e-02,  3.0135e-03,  4.1597e-02,
        -2.7075e-03,  1.3849e-02, -1.3900e-02,  1.0299e-03,  5.8622e-03,
        -6.5269e-03, -6.4994e-03,  1.1002e-02,  2.3263e-03,  1.2075e-02,
         2.7425e-03, -8.1859e-03, -1.0231e-02,  1.9749e-02,  5.3586e-04,
         2.3302e-03, -1.8691e-04, -1.0505e-03,  3.0975e-03, -1.4160e-03,
        -1.0505e-02,  1.6429e-02,  1.1298e-02,  6.4692e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 0.0197,  0.6391,  0.0222, -0.0523,  0.0147, -0.0008, -0.0210,  0.0259,
         0.0075, -0.0493, -0.0065, -0.0036,  0.0016,  0.0565, -0.0794,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2988e-02,  7.3223e-01,  1.5373e-02,  9.0015e-03, -8.5537e-03,
        -1.1343e-02, -1.6051e-02, -4.0910e-03,  1.1682e-03, -1.5672e-02,
        -3.6379e-04, -3.7146e-03, -1.5135e-03,  9.5494e-04, -2.4220e-03,
        -4.0744e-03, -8.6505e-03, -5.0882e-03,  1.1229e-04, -3.7657e-03,
        -5.6743e-03, -5.7981e-03, -7.0317e-03,  4.2720e-03, -3.1708e-03,
        -3.0999e-03, -4.2773e-03,  7.6046e-03, -1.3622e-02, -1.9444e-03,
         1.2435e-04, -6.5245e-03, -4.9678e-03, -4.3378e-03, -5.9439e-03,
        -8.5971e-03, -3.0712e-03,  2.4159e-03,  5.4432e-04, -3.3572e-03,
        -1.3073e-02, -2.1386e-03,  1.5400e-03, -6.7502e-03, -2.3774e-03,
         4.7278e-03,  2.5574e-03, -1.2843e-02,  4.8415e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1005e-02,  7.5485e-01,  2.7185e-02,  3.5496e-03,  3.1560e-03,
        -2.5958e-02,  3.1252e-03, -3.2889e-03, -5.7433e-03, -2.9598e-03,
        -1.2508e-02, -2.6041e-03,  1.6776e-03, -9.9717e-04, -8.9698e-03,
        -3.2099e-03, -1.5181e-02, -2.6026e-03,  9.9146e-03, -8.1762e-03,
        -6.5118e-03,  1.3985e-04, -8.0180e-03, -2.5581e-04, -7.2743e-03,
        -2.1534e-03, -1.7075e-02,  1.2252e-02,  5.4675e-03,  7.8801e-03,
        -1.6654e-03, -2.2614e-03, -5.6426e-03, -3.6908e-03, -3.0521e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3093e-02, -6.6294e-01, -6.4185e-02, -3.1236e-02, -9.8802e-03,
        -2.4343e-02, -6.1711e-03, -7.7504e-03, -3.8589e-03,  5.2931e-03,
         4.4907e-03, -3.1405e-03,  1.6060e-02,  6.3752e-03, -1.3149e-02,
        -4.9332e-04,  5.9080e-03, -3.0113e-02, -4.1518e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0488,  0.6738, -0.0440,  0.0207,  0.0269,  0.0031,  0.0346, -0.0051,
         0.0008,  0.0080, -0.0187, -0.0241, -0.0176, -0.0683,  0.0054,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.0323e-02,  6.2817e-01, -1.0607e-02,  2.4877e-04, -7.8514e-03,
        -1.1456e-03, -2.7679e-02, -1.3867e-02, -1.7304e-02, -1.0114e-02,
        -1.7617e-02, -1.0346e-02, -1.9041e-02, -9.4637e-03,  1.0138e-02,
         9.5682e-03, -4.6616e-03, -1.1749e-03,  1.3440e-02,  2.5131e-02,
        -6.1931e-02, -5.0176e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0177,  0.6459,  0.0338, -0.0269,  0.0023,  0.0037, -0.0442,  0.0033,
        -0.0069, -0.0080, -0.0223, -0.0037, -0.0230, -0.0103, -0.0077, -0.0116,
        -0.0024, -0.0032, -0.0085, -0.0049,  0.0029,  0.0115, -0.0054,  0.0126,
        -0.0313, -0.0047,  0.0057,  0.0118, -0.0208,  0.0028,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1072, -0.5953,  0.0628, -0.0024,  0.0472,  0.0078,  0.0439, -0.0325,
        -0.0744, -0.0061, -0.0205,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3455e-02,  8.1041e-01,  4.3479e-02,  6.9348e-03,  1.6676e-02,
        -2.5277e-03,  9.4765e-03,  1.6791e-03,  1.2327e-02,  1.8961e-03,
        -6.4808e-03,  6.7190e-03, -1.1197e-04,  8.7721e-03,  3.2252e-03,
         3.2450e-03,  1.1836e-02,  7.6849e-03,  5.2596e-03,  9.3558e-03,
        -4.2847e-03,  4.1648e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.0676e-03, -6.8191e-01, -7.3896e-02, -5.2390e-02, -1.1037e-02,
        -2.4089e-02,  3.1827e-03,  1.0068e-02,  1.2955e-02, -5.2563e-03,
        -1.9292e-03, -8.1921e-04, -1.1700e-02,  9.1858e-03, -2.2951e-03,
         9.5103e-04,  7.1330e-04,  2.8973e-03, -9.3501e-03,  3.9471e-03,
         4.4417e-03, -5.1585e-03, -4.8414e-03, -1.6451e-03, -6.6886e-03,
         4.2560e-03,  7.9665e-03,  3.8176e-04,  3.8306e-04, -3.2659e-03,
        -8.5613e-03,  9.5914e-03, -1.5177e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1464e-02, -6.6547e-01, -7.2439e-02,  3.5564e-02, -8.4691e-03,
        -5.7616e-03,  9.8645e-03,  1.2604e-02,  1.2541e-02,  3.1954e-03,
         7.6775e-03,  2.4075e-03,  1.6840e-02,  2.1509e-02,  1.4493e-02,
        -1.9318e-02, -1.7805e-03, -4.9685e-03,  1.4236e-03,  1.7589e-03,
        -4.1917e-03, -2.2506e-03,  1.9776e-03,  1.0648e-02, -5.3211e-03,
        -3.9090e-04,  1.0721e-02, -4.9552e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0017e-03, -6.6779e-01, -9.5729e-02, -1.8248e-02, -4.2612e-03,
        -1.6543e-02,  7.2803e-03,  1.8263e-02, -4.1166e-02,  1.3805e-02,
         1.2351e-02,  3.0677e-03, -1.6507e-02,  1.0781e-02, -2.3356e-03,
        -6.4495e-03, -4.7473e-03, -3.3553e-03, -4.4556e-03, -5.6496e-03,
         1.2284e-02, -6.6982e-03,  2.2377e-02, -4.7999e-03, -5.7586e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-3.3070e-03,  6.0924e-01, -2.0841e-02,  5.1354e-02,  3.8189e-04,
        -4.1280e-03, -2.8638e-03, -4.2004e-03, -1.2002e-02, -1.3275e-02,
         1.0267e-03, -6.8818e-03,  2.6850e-03,  5.2405e-03,  2.7616e-04,
        -2.8506e-02,  4.9853e-04,  3.6882e-03, -5.6113e-03,  1.1277e-03,
         3.9451e-03,  1.4394e-02, -2.3468e-03,  9.3906e-03,  4.1081e-02,
        -9.0498e-03, -1.1024e-02, -1.3672e-02,  1.4546e-03, -1.5887e-02,
        -1.0565e-02, -6.5341e-03,  5.0874e-03, -8.9762e-03, -6.2739e-03,
        -1.8870e-02, -1.3150e-02, -1.3005e-02,  1.6398e-03,  4.8823e-03,
        -1.1634e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0687, -0.7153, -0.0096, -0.0110, -0.0266, -0.0230, -0.0060, -0.0226,
         0.0103,  0.0019,  0.0009,  0.0117, -0.0144,  0.0009, -0.0040,  0.0068,
         0.0032,  0.0029, -0.0053, -0.0055, -0.0023, -0.0086,  0.0140,  0.0234,
        -0.0012,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2082e-02,  7.0835e-01, -3.8468e-03, -5.5694e-02, -4.9936e-03,
        -3.0324e-02, -1.9290e-02,  7.2489e-04, -1.5257e-02, -8.1027e-04,
         2.0946e-02, -5.7263e-03, -9.2105e-03, -1.1817e-02, -7.0292e-03,
        -1.1859e-02,  3.7641e-04, -6.5119e-03, -2.4474e-02, -2.4453e-02,
        -6.2226e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0128, -0.7612, -0.0369, -0.0047, -0.0081, -0.0406, -0.0067, -0.0127,
         0.0042,  0.0228,  0.0102,  0.0055,  0.0179, -0.0024,  0.0107, -0.0050,
         0.0352,  0.0023,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4040e-02,  7.1564e-01, -2.0145e-02,  5.2524e-02,  2.2836e-02,
         7.1497e-03,  1.5804e-02, -8.4609e-03, -5.5983e-03, -9.7542e-04,
         1.3172e-02, -2.8802e-03, -4.2566e-04, -1.0718e-02,  8.3177e-04,
        -7.3824e-03,  1.0463e-03, -1.2790e-02,  1.1305e-02,  1.3468e-03,
        -1.7358e-03,  9.9328e-03,  4.5087e-03,  3.1313e-03,  1.3162e-03,
         4.0226e-03, -8.3872e-03,  1.5120e-03, -4.1682e-03, -2.5583e-04,
         2.9892e-03,  3.4927e-03,  1.5019e-03,  2.1663e-03,  7.5895e-03,
        -3.0297e-04, -1.0701e-03,  6.8494e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0807e-01, -6.9001e-01,  3.9130e-02,  1.6037e-02,  1.3452e-03,
        -1.1227e-03, -3.7244e-05, -1.6288e-02,  9.1266e-03,  4.6509e-03,
         3.9659e-03, -2.9494e-03, -4.0008e-03, -3.5654e-02, -1.9923e-03,
        -2.3005e-02,  9.3855e-03,  1.5743e-02,  1.5679e-02,  1.8143e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0366,  0.7077, -0.0158, -0.0082,  0.0141,  0.0346,  0.0111,  0.0258,
        -0.0074, -0.0147,  0.0188, -0.0114, -0.0342, -0.0595,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0991, -0.5663,  0.0495,  0.0508, -0.0281,  0.0018,  0.0169,  0.0170,
         0.0104, -0.0085,  0.0123,  0.0018, -0.0033, -0.0049, -0.0117,  0.0122,
        -0.0107,  0.0109,  0.0046,  0.0037,  0.0085,  0.0007, -0.0050, -0.0182,
        -0.0011,  0.0172, -0.0035, -0.0030, -0.0011, -0.0006, -0.0134, -0.0033,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0294, -0.7733, -0.0153, -0.0107,  0.0200, -0.0031, -0.0132, -0.0012,
        -0.0018,  0.0034,  0.0170,  0.0088,  0.0023, -0.0141,  0.0121,  0.0045,
         0.0033,  0.0231,  0.0347,  0.0087,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0535,  0.5695,  0.0434, -0.0441, -0.0341,  0.0122,  0.0097, -0.0075,
        -0.0070,  0.0011,  0.0069, -0.0106,  0.0046,  0.0012,  0.0059,  0.0118,
         0.0025, -0.0109, -0.0028, -0.0064,  0.0014, -0.0013,  0.0058,  0.0063,
        -0.0060, -0.0028, -0.0009, -0.0180, -0.0056, -0.0013,  0.0092,  0.0060,
         0.0048,  0.0045, -0.0022,  0.0008, -0.0089,  0.0072, -0.0033,  0.0029,
        -0.0036, -0.0006,  0.0155,  0.0117,  0.0211, -0.0021, -0.0009,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1481e-02, -7.3204e-01,  3.3715e-03, -8.4789e-03, -3.4683e-03,
        -5.7964e-03, -2.2067e-02, -5.2602e-04, -5.9820e-02, -1.7565e-02,
         9.3672e-03, -1.6922e-02,  2.1562e-02, -2.9543e-03,  3.4581e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5513e-02,  6.1563e-01,  2.9391e-02,  2.9743e-04,  2.6113e-03,
         1.1888e-02,  9.6360e-03,  1.4657e-03,  9.6331e-04,  2.9761e-04,
         4.6840e-03, -4.8243e-03,  5.9300e-04,  1.1986e-02, -2.8037e-02,
         1.0501e-02,  3.4443e-03, -8.6335e-04,  6.2213e-04, -1.1081e-02,
        -2.0816e-03, -6.1134e-03,  3.9166e-03,  7.1867e-03, -7.4605e-03,
        -3.9447e-03,  2.4462e-02,  1.3021e-02, -9.1646e-03,  5.4381e-03,
         1.0535e-03,  2.4757e-03,  8.7555e-03,  2.8418e-03, -8.4135e-05,
         6.5067e-03, -4.2838e-03, -2.9302e-03, -3.7297e-03,  3.6953e-03,
        -2.2199e-04, -5.8745e-04, -1.1821e-03, -5.0877e-05, -2.5334e-03,
        -1.0239e-02, -4.0029e-02, -2.1681e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 1.3718e-01, -5.2408e-01, -6.5982e-02, -8.3143e-03,  1.2149e-02,
         2.0370e-02, -1.1876e-02, -7.8659e-04,  2.8475e-03, -5.5114e-03,
         1.1057e-02, -7.9246e-03, -5.7801e-03, -4.0292e-03, -1.0249e-04,
        -1.2030e-02,  1.4512e-02, -9.0094e-03, -2.6932e-03,  7.2266e-03,
         2.1735e-03,  8.0110e-03,  8.9433e-03, -1.3184e-02, -6.8107e-02,
        -3.6121e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.6312e-02, -6.6087e-01,  2.8709e-02, -4.1684e-02, -4.0037e-02,
        -2.9068e-03,  2.4606e-02, -9.3485e-03, -6.3801e-03, -9.6367e-03,
        -1.4645e-02,  8.5338e-03,  8.7579e-04,  7.4460e-03,  1.8206e-02,
         9.3687e-03,  3.5662e-04, -2.0214e-03,  1.4201e-02, -1.2892e-02,
         9.6332e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.7097e-02, -5.7750e-01,  1.1432e-02, -1.0043e-02, -3.0680e-03,
         2.5987e-02,  9.3781e-04, -2.6750e-03,  6.6596e-03,  4.0398e-03,
        -6.5574e-03,  6.1875e-03,  2.7156e-03,  3.4405e-03,  1.4033e-02,
        -8.8902e-03, -2.8273e-03,  2.6967e-02, -1.1374e-02,  1.5049e-02,
         2.6467e-02,  6.6956e-03, -2.3202e-02,  1.9313e-03,  5.4433e-03,
         8.1182e-03,  5.8216e-03,  5.7755e-03, -4.4989e-03,  1.5961e-03,
         2.8869e-02,  5.0438e-03, -1.6063e-04,  1.8906e-03,  1.9393e-02,
         2.7613e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2507e-02,  6.4672e-01,  3.8926e-03, -2.9470e-02,  2.4722e-02,
         1.9695e-02,  3.2883e-02,  1.0859e-03,  1.4467e-03,  1.1570e-02,
        -7.5978e-03, -5.4625e-03, -7.1674e-03, -8.0252e-03,  7.7990e-03,
         9.8587e-03, -5.8131e-04, -9.9644e-04, -5.4268e-02, -1.0425e-01,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0112,  0.7193,  0.0189,  0.0216,  0.0293,  0.0170,  0.0017,  0.0074,
         0.0561,  0.0042,  0.0299, -0.0052, -0.0051, -0.0027, -0.0015, -0.0063,
        -0.0037, -0.0138, -0.0194, -0.0017, -0.0242,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.1089e-02,  7.7562e-01,  3.7268e-02, -9.4665e-03, -1.4750e-02,
         2.2293e-02,  1.4290e-02, -2.1833e-02,  4.5011e-03,  1.4183e-03,
        -1.5643e-02,  1.5263e-03,  3.0095e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.6491e-02, -6.2477e-01, -3.0930e-02, -6.9845e-03,  6.8877e-03,
         7.8555e-03,  6.8342e-03, -1.1920e-04,  1.3616e-03, -7.4464e-03,
        -1.3546e-02,  1.0826e-02, -2.9390e-03,  2.0541e-03, -1.3668e-02,
         2.5416e-03, -1.1026e-02, -9.6538e-03, -6.4009e-04, -5.6451e-03,
         1.6950e-03, -5.0533e-03, -9.2284e-03, -8.4475e-03, -3.8162e-03,
        -2.9083e-03,  2.0505e-02,  2.3407e-03, -1.2338e-03, -1.6463e-02,
         6.0077e-03, -2.4829e-03, -1.2477e-02, -1.1175e-03, -2.4467e-03,
        -1.6694e-02, -3.4858e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2857e-02, -7.5984e-01,  3.7624e-02, -2.1757e-03,  2.0084e-03,
         1.9845e-03,  4.6894e-03,  1.0381e-02,  1.9428e-03,  8.9599e-03,
         1.1128e-02, -1.5022e-03,  3.1251e-03, -2.1813e-03,  1.4468e-02,
         1.0735e-03,  5.7317e-04, -2.9639e-04, -1.5319e-03,  2.5697e-03,
         3.1463e-03, -5.8982e-03, -6.4929e-02, -5.1142e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.5523e-02,  6.0646e-01, -4.1085e-02,  4.1368e-03, -2.1121e-03,
        -1.6106e-03, -8.5438e-03,  1.1638e-02, -7.8371e-03, -2.4032e-02,
        -5.7608e-03, -1.7756e-02,  3.2189e-03,  1.3166e-02,  1.5357e-02,
        -1.7684e-02, -1.7277e-02, -1.3915e-02,  4.9498e-03, -2.6350e-04,
        -2.0886e-03,  6.7673e-03, -6.1449e-03, -1.0161e-02, -6.2142e-03,
        -7.0034e-04,  3.2091e-03,  2.4055e-03, -3.6542e-03,  1.1341e-03,
         2.8250e-02, -3.6943e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2081e-02,  7.1458e-01, -9.0171e-03, -3.8500e-03,  2.2213e-02,
         1.2807e-03, -8.8152e-03, -2.7101e-02,  1.2592e-02, -1.6914e-02,
         3.6617e-04, -1.2097e-03, -9.0857e-03, -4.6213e-03,  1.6467e-03,
         2.4011e-03, -1.9480e-02, -4.9787e-03,  1.5162e-03,  1.4379e-02,
        -2.3913e-02, -1.2482e-03,  4.5017e-03, -1.9950e-03, -3.4874e-03,
        -1.5006e-02, -7.0478e-03,  1.4116e-03, -4.7709e-03, -1.0176e-02,
        -1.8308e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2847e-01,  5.0282e-01, -3.0904e-02, -2.2677e-02, -1.9082e-02,
        -2.0983e-02, -2.1749e-02, -1.2884e-02,  1.5112e-03,  4.5782e-02,
         1.2217e-03, -2.4823e-02,  3.9709e-03, -9.1441e-03, -3.7464e-04,
         1.9122e-02, -6.1074e-03, -1.3274e-02, -1.0500e-02,  1.1123e-02,
         1.3116e-02,  5.9670e-02,  2.0685e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.8346e-02,  6.3888e-01,  4.1740e-02, -8.3538e-03,  1.9370e-02,
        -9.6798e-03,  1.1341e-03, -2.4104e-02, -1.4542e-02, -7.9070e-03,
        -5.3288e-03, -2.9989e-02,  6.1118e-03,  8.2900e-05, -4.1309e-03,
         1.7943e-03,  1.8166e-02,  1.3436e-03, -5.8451e-03, -3.7358e-03,
        -1.0689e-02,  6.5952e-04, -5.5214e-04,  4.1416e-03, -3.0570e-03,
         5.1898e-03,  6.5720e-02,  9.4094e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 1.3882e-02, -8.1969e-01, -7.5583e-03,  3.2346e-03,  1.7912e-02,
        -5.3988e-03, -3.3715e-03,  1.5047e-02,  1.7483e-02,  8.5876e-03,
        -6.7663e-03, -6.4009e-04,  1.5199e-03, -1.5363e-03,  5.8520e-03,
        -2.5615e-03,  7.0275e-03,  5.1650e-03,  9.4287e-03, -4.4196e-03,
        -4.0976e-04, -1.4762e-03,  1.1691e-02,  6.9932e-03, -5.0525e-03,
        -1.0878e-03,  4.2765e-03, -2.0330e-03, -9.8931e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0515e-02, -6.0970e-01, -2.0013e-02,  6.7292e-03, -4.7757e-02,
         1.0473e-01, -6.2546e-03,  2.7823e-02,  1.7588e-02, -8.7396e-05,
         9.2294e-03, -1.3544e-02,  1.8781e-02,  3.9636e-03,  1.1070e-02,
        -4.8777e-04,  3.3333e-03,  1.0579e-02, -7.9219e-03, -2.9400e-03,
        -2.1665e-04,  4.2341e-02,  2.1327e-02, -3.0643e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0844e-01,  5.3755e-01,  9.9627e-02,  4.5011e-02,  2.1221e-02,
        -1.0519e-02,  1.6546e-03, -1.2060e-02,  5.7872e-03, -7.0941e-03,
        -7.6040e-03,  1.9107e-02,  6.2022e-03,  4.5393e-03, -4.4343e-03,
        -5.6177e-03, -5.4624e-03, -4.3716e-03, -1.6221e-02, -5.3268e-03,
         2.0600e-02, -8.8612e-05,  3.6936e-03,  9.0206e-03, -2.6814e-03,
        -6.7531e-03, -1.5556e-03,  2.7191e-04, -5.9872e-03,  1.0289e-03,
        -6.7035e-03, -3.7614e-03, -2.0702e-03, -7.9336e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2646e-02, -7.3262e-01, -5.6463e-02,  2.5945e-02, -2.5743e-02,
         5.7930e-03, -1.8769e-02,  4.3875e-03, -1.6771e-02, -1.3370e-02,
        -7.1292e-04, -1.3243e-02, -1.9055e-03,  1.2359e-03, -4.7373e-03,
         2.4723e-04,  7.6530e-04,  1.9642e-02,  3.2093e-03,  1.3828e-02,
         6.1453e-03,  1.1822e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.1358e-02,  6.5992e-01,  2.8857e-02, -2.8521e-02,  5.3947e-02,
         8.1866e-03, -1.3252e-02,  4.6042e-03, -1.9303e-04, -7.7152e-03,
        -1.2392e-02,  1.3575e-04, -8.4223e-04, -6.8098e-03,  2.0314e-02,
         1.2666e-02,  6.2637e-02,  2.7647e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0617e-01,  6.2782e-01, -5.0351e-02, -4.6394e-03, -8.3524e-03,
        -3.4145e-02, -1.3017e-03,  1.3749e-02, -3.9845e-03,  2.5585e-03,
         4.1369e-03,  7.4982e-04, -6.6661e-04,  2.5070e-03, -7.6945e-03,
         6.4173e-05, -7.1626e-03, -1.1890e-02, -5.2566e-03, -1.2832e-03,
        -2.4672e-03, -6.7452e-02, -3.5601e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.0068e-02, -8.0370e-01,  1.8677e-02, -2.9707e-02,  5.1044e-03,
        -4.9458e-03,  2.3113e-03,  7.2915e-03, -1.6333e-02,  2.2654e-03,
         4.8592e-03, -8.5599e-04,  3.7219e-03, -1.7679e-03, -1.1112e-02,
        -6.2291e-04, -3.6660e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.0138e-02, -6.9672e-01, -2.1168e-02, -6.7711e-03, -2.1361e-02,
        -3.9310e-02,  1.9895e-02,  2.8081e-02,  1.5940e-02,  1.1513e-02,
        -1.7719e-02,  5.0514e-03,  6.0152e-03, -1.8581e-02,  2.9795e-03,
         7.5329e-03, -1.1139e-02,  8.2219e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.8490e-03, -6.6114e-01, -2.0281e-02, -1.5622e-03, -1.0081e-02,
         1.5700e-02,  9.0858e-03, -1.5430e-02,  6.0403e-03, -3.5833e-02,
         2.6546e-02,  2.2701e-02, -2.5301e-03, -3.0948e-03, -8.0382e-03,
        -2.3924e-03, -7.9718e-03,  5.6601e-03, -4.4064e-03, -3.1040e-03,
         9.3813e-05,  9.5481e-03,  2.6190e-03, -1.0980e-02,  3.6463e-02,
         6.8847e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0213,  0.5792,  0.0626, -0.0144,  0.0108, -0.0229,  0.0508,  0.0097,
        -0.0130, -0.0109, -0.0337,  0.0033,  0.0020,  0.0010, -0.0016,  0.0117,
         0.0037,  0.0093,  0.0047, -0.0008, -0.0071,  0.0061,  0.0061, -0.0049,
         0.0014,  0.0085, -0.0115, -0.0013, -0.0072, -0.0626,  0.0160,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.6982e-02,  5.9545e-01,  4.6770e-02,  2.0642e-02,  9.7410e-03,
         1.5044e-02,  3.1751e-02, -1.0430e-02,  9.8587e-05, -1.9139e-02,
        -4.9372e-03, -8.9718e-03,  1.0227e-02, -3.9832e-03, -1.5297e-02,
        -6.8860e-03,  9.2765e-02,  3.0884e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1174, -0.6266,  0.0669,  0.0739,  0.0011,  0.0128, -0.0417,  0.0597,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([-1.2056e-01,  5.9383e-01,  2.1191e-02, -1.7113e-02, -9.8290e-04,
         2.0784e-02, -2.7744e-02, -7.0640e-03, -2.3717e-04,  8.3348e-03,
        -5.3712e-04, -2.1793e-02,  9.5038e-03,  7.5493e-03,  6.7986e-03,
         1.3749e-02,  5.6257e-03, -2.5153e-03,  3.1690e-03,  6.8779e-03,
         5.3189e-03,  7.0459e-03, -5.0337e-03, -5.5090e-03,  1.3955e-02,
        -3.6325e-03, -3.4202e-03,  5.9905e-03,  1.2749e-03, -6.2024e-03,
        -1.6666e-02, -2.6376e-03, -3.1160e-03, -4.0363e-03, -6.2168e-03,
         8.1198e-03,  2.8027e-03, -1.4841e-03,  1.5715e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9800e-02,  6.2440e-01, -3.1100e-02,  7.7121e-03,  5.1228e-02,
        -6.1373e-04, -1.0799e-02, -1.4661e-02, -1.8355e-02,  1.2441e-02,
        -7.1013e-03,  7.7099e-03,  2.4689e-03, -1.1193e-03,  9.4364e-03,
        -2.9998e-03, -2.4109e-03, -1.1255e-02, -3.4017e-02, -7.8791e-03,
        -9.0395e-03,  3.9892e-03,  4.6436e-03,  4.3952e-03, -2.4082e-03,
         1.6273e-03, -5.2859e-02, -3.3528e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0826, -0.7618,  0.0089, -0.0176, -0.0096,  0.0106,  0.0086,  0.0245,
        -0.0094, -0.0132, -0.0320, -0.0155,  0.0058,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0262, -0.7978,  0.0236,  0.0113, -0.0328,  0.0183,  0.0042, -0.0142,
        -0.0011,  0.0138,  0.0128,  0.0045, -0.0036,  0.0094,  0.0263,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9436e-02,  6.2282e-01, -3.7577e-02, -7.1488e-03,  2.0958e-02,
        -1.5638e-03, -1.2737e-02, -7.1052e-03,  9.2780e-03, -6.0782e-03,
        -2.3756e-02, -9.5726e-03, -2.8765e-03,  4.9485e-03,  1.8848e-02,
        -1.1267e-02, -5.8433e-04,  1.3818e-03, -4.5698e-03, -2.0914e-03,
        -8.9656e-03,  2.8199e-03,  6.6333e-03,  8.4902e-04,  1.0097e-02,
         1.7038e-02,  9.8457e-04,  6.5742e-03, -2.3592e-03, -2.5720e-03,
        -1.7256e-03, -5.1995e-03, -4.5199e-03, -9.0386e-03, -5.8061e-03,
        -6.2110e-03, -1.8069e-02, -2.8034e-03, -2.2971e-04, -9.9716e-04,
        -2.2509e-03, -2.9281e-03, -3.4427e-03,  6.5261e-03, -2.1158e-03,
         1.2997e-03, -2.3344e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.5239e-02, -5.9688e-01, -2.5341e-02,  8.1997e-04,  1.0810e-02,
         2.0821e-02,  3.5157e-03,  6.8663e-03, -9.5076e-03,  5.3006e-03,
        -6.3743e-03, -2.0087e-02, -6.8464e-03, -7.4087e-04, -8.5374e-04,
         1.1137e-02, -4.2823e-03,  2.7660e-03,  3.5163e-03, -1.9035e-03,
        -6.4504e-03,  8.2088e-03, -3.6120e-03, -9.6378e-03,  6.5049e-04,
         8.4812e-04,  4.7221e-03,  4.1950e-03,  3.0054e-03, -4.8716e-03,
         7.0003e-03,  1.6982e-03, -1.1093e-02, -1.1058e-02, -4.3670e-03,
         3.7825e-04,  9.1989e-03,  2.3822e-03,  4.0011e-03, -6.6797e-03,
        -6.2297e-02,  3.0038e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6489e-03,  7.2195e-01,  6.8179e-02, -1.2756e-02,  1.9861e-03,
         1.1532e-02,  1.6353e-02, -6.7348e-03,  1.7572e-02,  9.1633e-03,
         1.1716e-02, -1.3795e-02, -3.7131e-03,  1.9007e-04,  1.3391e-02,
         5.1048e-03,  2.1395e-02,  8.8903e-04, -3.4337e-03,  9.1891e-03,
        -1.0231e-02, -9.5194e-03, -8.7303e-03,  2.0783e-03, -1.7746e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.8026e-02,  7.0481e-01, -7.0827e-03, -2.1402e-02,  1.1090e-03,
        -1.3286e-02,  4.1762e-03,  1.3968e-03,  1.0457e-03,  3.4519e-03,
         7.0918e-03,  6.1102e-03,  9.2863e-03, -9.0559e-03, -7.1255e-03,
        -1.7971e-02,  5.0338e-04,  8.9010e-03, -3.1790e-03, -1.1708e-02,
         3.2728e-02, -3.3869e-02, -2.6687e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.8318e-03, -7.4497e-01, -8.3224e-02, -1.0158e-02,  1.7660e-02,
        -9.7992e-03,  5.8087e-03,  1.8065e-02,  3.4441e-03,  2.0230e-02,
        -1.0027e-02, -3.5345e-03, -7.6590e-04, -1.2879e-02,  2.0246e-03,
         6.2074e-03,  1.0318e-02, -7.0519e-03, -1.5689e-03, -3.5389e-04,
         5.8982e-03,  6.1904e-03,  9.9936e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1638, -0.5381, -0.0802, -0.0010,  0.0095,  0.0160,  0.0029,  0.0352,
        -0.0413, -0.0116,  0.0184, -0.0466,  0.0354,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0810, -0.6751,  0.0579,  0.0323, -0.0110,  0.0143, -0.0009,  0.0109,
        -0.0340, -0.0331,  0.0310, -0.0102, -0.0007,  0.0074,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1603e-02, -7.4820e-01, -2.6093e-02,  7.7409e-03, -4.6611e-02,
         1.0296e-02, -6.0934e-03,  2.7429e-02,  8.9938e-03,  2.2931e-03,
        -1.9735e-04,  2.9992e-03,  2.6032e-03,  4.8389e-03, -9.7863e-03,
        -3.9168e-03,  2.2221e-02,  1.4821e-02,  3.3268e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 2.1699e-02,  7.4282e-01, -3.1723e-02,  2.3538e-03, -3.2507e-02,
         1.8489e-02,  4.8232e-03, -1.9375e-02,  5.5225e-03,  2.1152e-02,
        -5.8843e-04,  1.5444e-03, -1.6639e-03, -1.4178e-02,  4.3620e-03,
        -8.7424e-03,  2.0657e-03, -4.1563e-03,  6.0081e-03,  1.2710e-02,
         1.0977e-02, -9.2251e-03, -8.9588e-03,  4.2002e-03,  3.6585e-03,
         6.4934e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0060, -0.7454, -0.0450,  0.0031, -0.0156,  0.0157,  0.0090,  0.0142,
         0.0089,  0.0353, -0.0192,  0.0040,  0.0042, -0.0161,  0.0118, -0.0465,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9536e-03, -6.0636e-01, -3.6727e-02, -8.4215e-02, -1.6275e-02,
        -1.1304e-02, -2.7592e-02, -1.5776e-02,  2.7720e-02, -2.6069e-03,
        -5.1246e-04, -1.4732e-03, -1.4902e-03, -9.6628e-03, -3.0624e-02,
        -1.0057e-01, -2.1147e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1324,  0.6644, -0.0146, -0.0655,  0.0435, -0.0272, -0.0127, -0.0221,
         0.0176,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0329e-02, -4.5344e-01,  2.9593e-02,  2.8374e-02, -2.6419e-02,
        -2.7055e-02,  1.0597e-03, -3.8322e-02, -9.8438e-03, -9.9873e-03,
        -1.1021e-02,  4.2893e-04,  5.6611e-04,  8.9383e-03, -1.2623e-02,
        -1.4941e-03,  2.8149e-03, -2.4826e-03, -9.2694e-03, -1.4725e-03,
        -5.2340e-03,  9.7290e-03,  7.3415e-03, -1.0618e-02,  1.7667e-03,
         6.0452e-03,  2.3870e-02, -8.3935e-03,  1.9293e-02,  1.2822e-02,
         8.1785e-03, -4.4798e-03,  1.6081e-03,  2.3743e-03, -5.0805e-03,
        -2.0982e-02, -1.8820e-03,  1.2029e-04, -3.7069e-03,  2.2179e-03,
         8.6205e-04, -5.1673e-03,  1.5809e-02,  1.0379e-02, -4.5362e-03,
        -9.1970e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4283e-02,  6.7954e-01, -1.5059e-02, -1.0650e-02,  7.4358e-03,
        -1.7069e-02, -6.1417e-03,  1.6789e-02,  4.9317e-04,  9.0923e-04,
        -1.1818e-03,  9.9160e-03,  1.0611e-02,  6.0843e-03,  9.2791e-03,
         2.2153e-02,  5.0812e-03,  8.4586e-03, -7.1081e-03,  1.5124e-02,
        -2.0985e-02,  1.2335e-02, -9.4229e-03,  2.6984e-03,  1.2605e-02,
         4.2508e-04, -8.4219e-03,  4.0948e-03, -1.8365e-02,  3.7276e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0270, -0.7644,  0.0022, -0.0372, -0.0443, -0.0461, -0.0787,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9545e-02, -8.5043e-01, -3.9311e-03,  3.4646e-03,  8.0845e-04,
         1.4546e-02, -2.0220e-02, -1.4162e-02, -1.5078e-02, -1.9941e-02,
        -1.7879e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0390,  0.6101,  0.0657,  0.0831, -0.0066,  0.0212,  0.0136, -0.0035,
         0.0104,  0.0393,  0.0180,  0.0369,  0.0282, -0.0244,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1787e-02, -6.8230e-01, -1.0737e-02,  6.7573e-04, -1.0761e-02,
        -2.1756e-02, -4.3008e-03, -1.5408e-02,  3.9263e-02,  1.6340e-02,
        -2.5365e-03,  1.8544e-03,  1.9581e-03,  8.3586e-03, -1.2634e-03,
         1.0138e-02, -4.2743e-03,  9.1940e-03,  1.6628e-03, -7.1915e-03,
        -7.6644e-04, -9.2672e-03,  4.3214e-03, -1.5355e-03, -3.6304e-03,
         3.4575e-03,  9.7695e-03, -4.8277e-03,  8.3068e-04,  2.0197e-03,
         1.5509e-02,  9.0444e-04,  2.0314e-02, -4.1084e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3614e-02,  7.4772e-01,  3.2599e-02,  1.1656e-02,  3.5668e-03,
         2.1664e-03, -2.3021e-03,  1.2164e-02, -2.0994e-04,  9.9603e-03,
        -9.5308e-04, -1.3550e-02, -6.0454e-03,  2.5827e-02, -8.3874e-03,
        -5.5004e-04,  1.4312e-02, -2.1235e-02, -1.1615e-03, -8.2345e-03,
         1.1000e-03,  1.0239e-03, -1.4038e-03,  1.4939e-02,  1.3891e-03,
        -7.1050e-03, -5.1297e-03, -3.1698e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5835e-02,  7.4489e-01,  1.7757e-02,  1.9882e-02,  1.0587e-02,
         6.0630e-04,  7.5700e-03,  8.2654e-03, -1.1175e-03,  8.2298e-03,
         1.3543e-03, -3.3743e-03, -8.8436e-03,  1.2593e-02, -3.4089e-03,
        -1.6740e-02,  1.2411e-02,  1.8408e-02, -1.5140e-02, -4.9874e-04,
        -1.0891e-03, -3.7414e-03,  8.6958e-03, -3.8962e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 1.0339e-01,  5.0853e-01, -8.7952e-02, -1.8432e-02, -1.7831e-02,
        -2.9345e-02, -1.3226e-02, -5.0766e-03,  9.4166e-03, -4.8738e-04,
        -2.0678e-02, -1.0035e-02, -1.2073e-03,  4.1468e-03,  4.0747e-03,
        -8.9793e-03, -1.1072e-02, -1.3543e-02,  2.7691e-03,  1.9768e-02,
        -7.7122e-03,  8.1679e-03, -6.7094e-03, -7.1370e-03,  3.5955e-03,
         1.3155e-02, -3.3384e-03, -1.0619e-03, -1.0422e-02, -2.8245e-02,
         2.0494e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0412,  0.7373, -0.0293,  0.0199,  0.0123,  0.0388,  0.0212, -0.0066,
         0.0091,  0.0209,  0.0307, -0.0328,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1970e-02,  7.3328e-01,  5.1495e-02, -4.6939e-02, -1.2186e-03,
        -1.6640e-02, -5.9978e-03,  2.0975e-02, -1.8113e-02, -3.5862e-03,
        -2.3759e-03, -6.4020e-03,  5.1382e-02, -2.8998e-02,  6.3194e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.7803e-02, -6.6291e-01,  2.9011e-02, -5.1411e-03, -3.2114e-04,
         8.1107e-03, -1.2912e-02, -1.7873e-02,  1.2200e-03,  5.4944e-03,
         1.4283e-03, -6.5992e-03, -4.7382e-03,  1.0285e-02, -1.1557e-03,
        -8.3717e-03, -5.6423e-03,  4.8530e-03, -3.3708e-03,  1.0096e-02,
        -3.2338e-03, -6.2045e-03, -1.2126e-02, -3.6513e-04, -1.4178e-02,
        -5.9819e-03, -9.1547e-03,  8.8764e-04, -2.4587e-03, -5.7888e-03,
        -5.6653e-03, -2.9999e-02, -3.6624e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0228, -0.6174,  0.0234,  0.0725, -0.0326, -0.0080,  0.0163,  0.0048,
         0.0099, -0.0097, -0.0009, -0.0083,  0.0124,  0.0325,  0.0077,  0.1209,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0578, -0.6806,  0.1386,  0.0182, -0.0158,  0.0017, -0.0503,  0.0086,
         0.0283,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1460,  0.5761, -0.1148, -0.0767,  0.0864,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.4198e-02,  6.2891e-01,  7.6128e-02, -3.2746e-02, -3.7200e-03,
        -1.9988e-03, -1.2922e-02,  1.2095e-03, -7.7550e-03, -2.2479e-02,
         6.0457e-04,  2.5277e-02, -6.2949e-04, -3.4048e-02, -3.2647e-03,
         2.5345e-05,  1.6581e-03, -2.5832e-03,  7.8381e-03,  5.8480e-03,
        -6.5781e-03, -5.0745e-03, -9.4663e-03,  7.1090e-03, -5.2476e-04,
        -2.0471e-02, -6.9354e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0304, -0.7225, -0.0879,  0.0523, -0.0248, -0.0791, -0.0031,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.6995e-02, -6.5922e-01, -6.4918e-04, -2.1824e-02, -2.9028e-02,
        -2.6432e-03, -9.5679e-03, -7.0152e-03,  5.6248e-03,  2.3813e-02,
        -3.6521e-03, -8.0347e-03, -7.4011e-03, -1.1331e-02, -7.1175e-03,
        -1.1946e-02, -2.4830e-03, -2.2640e-02, -3.0716e-02,  1.5415e-02,
         6.1193e-03, -3.6759e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5733e-02,  7.1643e-01,  3.9476e-02, -2.9870e-02, -1.0179e-02,
         2.5201e-02, -2.1448e-02,  2.1986e-02, -2.4505e-03, -1.0101e-02,
         1.8050e-02,  2.5940e-03,  7.7173e-03,  2.4554e-03, -4.2368e-03,
        -5.5560e-04, -1.4221e-02,  2.9878e-02,  7.4182e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.4670e-02,  6.6150e-01,  5.2163e-02,  5.7721e-04,  9.6257e-03,
         2.2853e-02,  1.1920e-03, -1.5018e-02, -2.9106e-03, -8.4762e-03,
         4.5127e-04, -2.0354e-03, -5.4889e-03,  2.8306e-03, -1.4729e-03,
        -6.3407e-03,  1.1724e-02,  9.1257e-03,  1.0412e-03, -1.8199e-03,
         8.8096e-03,  1.1181e-02, -3.4894e-04,  1.8151e-02,  1.5748e-02,
         2.8844e-03, -4.9884e-03, -3.6580e-03,  1.0678e-02, -1.6583e-02,
         5.6526e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-3.0796e-02,  7.1408e-01,  1.3025e-02, -1.5161e-02,  1.1838e-04,
        -4.8099e-03, -3.3138e-04, -5.9063e-03, -3.6427e-03,  3.8743e-03,
         4.8497e-03,  3.6755e-03,  1.7651e-02,  1.8945e-02, -2.2640e-04,
         9.2694e-04,  1.2793e-02,  1.1631e-02, -3.7804e-03,  3.8441e-03,
         2.6090e-03, -6.7633e-03, -1.2053e-03,  2.0521e-02, -6.6854e-03,
        -1.1607e-03,  4.2743e-03, -3.9668e-02, -5.5091e-03,  1.5933e-02,
         1.2805e-02, -1.2797e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.6705e-02,  7.1483e-01,  2.4620e-02, -7.4259e-04, -3.5015e-02,
        -1.3085e-02, -3.5236e-02, -5.9388e-03,  5.5348e-03, -1.3199e-02,
         1.7494e-03,  1.7127e-02, -1.1080e-02, -4.3571e-03, -4.6967e-04,
         4.2482e-03,  1.2000e-03, -1.1809e-02, -5.7114e-03, -7.8277e-03,
        -8.0922e-04, -8.7110e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0143e-01,  7.0440e-01, -1.5681e-02,  1.2266e-02,  9.1128e-03,
        -1.0054e-02,  4.2548e-03, -4.5337e-04,  8.8458e-03, -5.1477e-03,
        -6.1374e-04, -6.0842e-03, -7.4167e-04,  5.1514e-03, -6.5596e-03,
         2.8714e-04,  4.0075e-03,  1.2735e-02, -7.4774e-03,  1.2313e-02,
        -7.9446e-03, -1.2746e-02, -1.1585e-03, -1.1119e-02,  1.2110e-02,
        -4.2678e-03, -3.2278e-03, -7.7929e-03, -1.1236e-02, -7.8367e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0322,  0.6992, -0.0376, -0.0101, -0.0111, -0.0342,  0.0063, -0.0187,
        -0.0266, -0.0213,  0.0184,  0.0060, -0.0011,  0.0143, -0.0101,  0.0101,
         0.0256, -0.0171,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9499e-02, -5.9730e-01, -9.1015e-02, -3.2113e-02, -1.1906e-02,
         3.0847e-02,  1.5440e-02, -4.8249e-03,  2.0128e-02,  3.2252e-03,
         9.5949e-03,  4.1359e-03,  9.2463e-03, -9.9657e-04,  2.2727e-03,
         1.2968e-02, -1.8360e-04, -2.4819e-03, -1.3227e-03,  1.1333e-02,
         2.4720e-03,  8.5326e-03, -7.8981e-04, -4.2360e-03,  8.5444e-04,
         3.6687e-03, -1.0190e-02,  5.6109e-04,  6.2863e-04, -1.4458e-03,
         1.8218e-03, -4.8001e-03, -2.5715e-03, -2.5116e-03, -1.6430e-03,
         4.2873e-02,  9.5632e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8115e-02, -6.6331e-01,  9.8842e-02, -6.6331e-02, -6.8048e-03,
        -2.5370e-03,  7.3361e-03, -1.8215e-03,  2.9821e-03, -2.0305e-03,
         2.8731e-04, -7.8553e-03, -1.9667e-03,  8.2806e-03, -3.3202e-03,
        -4.1622e-03, -2.4282e-04,  8.6562e-04,  8.4928e-03, -1.6583e-02,
         3.9259e-03,  1.4115e-02,  9.6902e-03, -4.0098e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3035e-01,  6.0447e-01,  3.3150e-02,  3.1962e-03,  1.5035e-02,
        -1.8128e-03, -1.7482e-02,  6.9782e-03, -6.2409e-03,  8.7254e-03,
        -5.3990e-03,  1.5784e-03, -1.0733e-02,  6.6425e-03,  1.9576e-03,
         2.5404e-04, -3.6970e-03, -4.3389e-05,  2.0587e-03,  4.0750e-03,
         3.2157e-03,  3.1373e-03,  5.9914e-03, -3.2434e-03,  1.0385e-02,
         8.9439e-03,  3.4213e-03,  1.2233e-02,  1.6668e-03, -2.3822e-03,
        -5.5441e-03,  1.3600e-02,  5.1101e-03,  3.4625e-03, -7.4009e-04,
        -1.1250e-02, -4.6990e-03, -3.6106e-03,  1.3381e-03, -2.5661e-03,
         1.5132e-02, -1.3978e-03,  2.4154e-03, -6.0169e-04, -1.8639e-03,
         8.1673e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5375e-02,  6.1957e-01,  8.2223e-02,  1.1034e-02, -5.5009e-02,
         2.0018e-02, -1.8280e-02, -9.9977e-03,  6.7713e-03, -1.0107e-02,
         1.5896e-02, -1.2252e-02, -4.6392e-03, -5.9682e-03,  4.8199e-03,
        -2.1288e-03, -7.5203e-03, -3.6693e-02, -2.8761e-02, -1.1281e-02,
         1.0494e-04,  5.9825e-03,  5.5719e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0446, -0.7636,  0.0136, -0.0285,  0.0249, -0.0243,  0.0156,  0.0174,
         0.0082,  0.0381,  0.0213,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0641,  0.7184,  0.0449, -0.0424,  0.0186, -0.0206,  0.0043,  0.0252,
        -0.0038,  0.0114,  0.0229,  0.0085, -0.0121,  0.0028,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0721,  0.7340, -0.0132, -0.0039, -0.0044, -0.0091, -0.0521, -0.0354,
         0.0179, -0.0366,  0.0164, -0.0050,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9225e-02,  6.5346e-01,  4.4437e-02,  4.4246e-02,  2.5745e-03,
        -5.3874e-03, -2.3858e-03,  1.0877e-02, -1.1043e-02,  5.2623e-04,
        -6.9202e-03, -9.6219e-03, -7.8151e-03, -1.7667e-03,  5.6412e-06,
        -8.8794e-04, -2.1381e-03, -4.8439e-03, -4.4619e-03, -9.7750e-03,
        -3.9186e-03,  1.7731e-03,  1.0426e-03, -1.1276e-03,  4.2193e-03,
         4.2418e-03,  9.3656e-03, -1.7343e-03, -4.5123e-03, -7.7608e-03,
        -2.5598e-03,  4.6029e-03, -4.7402e-03,  9.2505e-03, -1.7600e-03,
        -1.7240e-03, -1.5995e-04, -2.3569e-03, -3.4689e-03, -1.2848e-03,
         1.8500e-03,  1.6183e-03,  6.4962e-03, -2.2262e-03, -4.6440e-03,
        -3.5411e-03, -3.4350e-03, -1.1083e-03, -3.2846e-03, -6.9543e-03,
         5.2674e-03, -2.0216e-03, -1.3716e-03, -2.9468e-03, -9.2373e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 1.1156e-02,  7.4272e-01,  2.1582e-02,  1.6332e-02,  3.8172e-03,
         2.8480e-02, -4.1795e-03, -5.9412e-03, -5.4468e-03,  5.9209e-03,
        -5.5279e-03,  7.3692e-03,  1.5434e-02,  1.2006e-03, -3.5259e-03,
        -3.7993e-03, -2.4913e-02,  7.1835e-03,  1.4997e-04,  1.7306e-02,
        -2.1218e-02,  2.0237e-02, -2.6562e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.7210e-03, -7.1277e-01, -8.6344e-02, -3.4867e-02, -1.1854e-02,
        -9.8245e-03,  2.9512e-03, -4.8308e-03,  2.3659e-02, -2.0067e-02,
        -1.0413e-02, -1.3562e-03, -8.0020e-03,  1.6793e-02,  4.0126e-04,
        -1.2047e-04,  3.8109e-02, -1.0915e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1003, -0.5642, -0.0891,  0.0162,  0.0223, -0.0287, -0.0390,  0.0232,
         0.0152,  0.0167,  0.0104, -0.0063,  0.0093,  0.0191,  0.0182, -0.0150,
        -0.0043,  0.0024,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7166e-02, -7.0438e-01, -1.0582e-02, -5.1313e-03,  3.0973e-02,
        -1.7565e-02,  2.7193e-02,  1.2714e-02,  8.4021e-03,  1.0511e-02,
        -2.9512e-04, -4.5371e-04, -1.2458e-02, -1.7500e-03,  1.0042e-01,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2406e-02,  6.4056e-01,  1.9862e-02,  7.7140e-02,  2.1955e-02,
        -1.7163e-03, -5.7033e-03, -1.1946e-02, -1.1062e-02,  6.2865e-03,
         5.8334e-04, -2.0420e-03,  1.3987e-02,  8.0487e-03,  7.2969e-03,
         1.2745e-02,  1.5893e-02, -1.6290e-03, -2.2558e-03,  8.0250e-03,
         8.9504e-03,  6.5196e-04,  1.5095e-03, -3.1698e-03, -1.4819e-02,
         1.8488e-02,  3.5713e-02, -3.3270e-03, -3.2695e-03,  6.3487e-03,
         5.3124e-04, -2.2079e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2883e-02, -7.3945e-01, -2.0634e-02, -8.0676e-03, -1.8750e-02,
        -1.1222e-02,  9.0028e-04, -6.7854e-03, -2.1732e-03, -5.2586e-03,
        -1.1578e-02,  3.9633e-03,  2.5362e-03,  1.1576e-03,  2.8364e-03,
        -5.8951e-03,  1.2858e-03,  3.2464e-04,  1.2951e-03,  6.7534e-03,
         1.5867e-02,  1.2855e-02,  2.6653e-02, -4.3131e-03,  3.9068e-03,
         4.5188e-03,  3.3665e-03, -8.0878e-03, -3.0688e-03,  2.4813e-03,
         1.1972e-02, -1.8096e-03, -3.7351e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.6010e-02, -4.8483e-01, -9.1521e-03,  3.3607e-03, -1.8250e-02,
        -1.8155e-02,  2.0317e-03,  8.4616e-03,  5.6750e-03,  2.3834e-02,
        -3.4028e-03,  4.0984e-03, -1.7980e-02, -1.8191e-02, -2.6211e-02,
         6.0650e-03,  1.2274e-03,  1.3521e-03, -3.9166e-04,  2.1213e-03,
         3.6352e-03,  4.1275e-03,  3.0941e-03,  1.8499e-02, -1.6481e-03,
        -1.4385e-03, -5.8377e-03, -1.3746e-03, -3.8948e-03,  1.0355e-04,
         1.2389e-03,  4.6254e-03, -1.3097e-02, -8.1496e-03, -6.2305e-03,
         7.5315e-03, -2.9957e-03,  2.1926e-03,  1.5010e-03,  1.4279e-03,
        -4.6512e-03, -1.3603e-03, -5.4958e-04, -4.5198e-03, -1.8233e-03,
        -5.5302e-03, -6.4607e-04, -9.6581e-04, -3.8086e-03, -3.2875e-03,
        -1.1002e-03, -2.2739e-03, -1.6253e-03, -1.6369e-03, -9.0645e-04,
        -4.1552e-03, -2.0261e-02,  9.7450e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0074,  0.4561,  0.0532, -0.0617,  0.0156,  0.0048,  0.0069, -0.0023,
        -0.0031,  0.0166,  0.0291, -0.0172, -0.0148,  0.0356,  0.0570, -0.0019,
        -0.0032,  0.0108,  0.0069,  0.0175, -0.0288, -0.0030, -0.0214, -0.0288,
         0.0465,  0.0500,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3770e-02,  4.7906e-01, -2.2923e-01, -1.4339e-02, -4.7264e-03,
        -1.6000e-03,  2.2236e-02,  8.1285e-04,  2.6095e-02,  3.0708e-03,
        -6.4375e-03,  6.6770e-03,  1.1000e-03,  3.4042e-03,  6.9911e-03,
        -1.4662e-02,  1.0561e-01, -2.0808e-03,  9.0514e-03,  3.4589e-04,
         7.6553e-03, -3.8088e-03, -6.7194e-03, -4.2553e-03, -2.6261e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2363e-02,  4.9604e-01,  1.0965e-01, -1.8567e-02, -2.0703e-02,
         2.5278e-02,  1.1712e-02, -3.3398e-02, -2.0128e-02, -3.8748e-02,
        -2.3506e-02,  4.7504e-03, -2.3313e-03,  2.6529e-03, -1.6503e-03,
         1.6320e-02, -1.0088e-02,  3.2200e-03, -1.3539e-03, -6.8524e-03,
         2.8165e-03,  3.5779e-03,  2.4885e-03,  2.6957e-03, -1.5282e-02,
        -7.1667e-03,  1.0274e-04,  1.6995e-02, -1.6279e-03, -2.0546e-03,
         3.2311e-03,  2.1395e-03,  2.4872e-03,  2.1566e-06,  5.4838e-03,
         1.9947e-03, -6.0127e-03, -2.8017e-03,  5.1175e-03,  2.7003e-04,
        -5.2944e-03, -4.4978e-03, -6.9033e-03, -1.1465e-02, -1.8181e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6416e-02,  6.5528e-01,  2.3331e-02,  1.1109e-02, -5.4190e-02,
         3.8439e-03, -1.1814e-02,  1.2963e-02,  1.6988e-02, -3.2699e-03,
        -9.7724e-03, -1.1374e-02, -3.0424e-02, -2.0419e-04,  2.2168e-03,
         1.9655e-02, -1.9828e-02,  5.8405e-03,  4.1588e-03,  1.4999e-03,
         7.6336e-03, -5.5494e-02,  2.6970e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.2244,  0.5948, -0.0227,  0.0143,  0.0334,  0.0169, -0.0166,  0.0033,
        -0.0051, -0.0348,  0.0040,  0.0298,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-0.0660,  0.7048,  0.0401,  0.0174, -0.0051,  0.0052,  0.0463,  0.0101,
         0.0112, -0.0198, -0.0019, -0.0032,  0.0026, -0.0070, -0.0083, -0.0092,
        -0.0059, -0.0011, -0.0029,  0.0021, -0.0074,  0.0082, -0.0144,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0562,  0.6156,  0.0134, -0.0038, -0.0093, -0.0055, -0.0180,  0.0225,
         0.0184, -0.0053, -0.0018, -0.0024,  0.0130,  0.0127, -0.0066, -0.0060,
        -0.0173, -0.0163,  0.0126, -0.0059,  0.0121, -0.0033,  0.0008, -0.0079,
         0.0017,  0.0020,  0.0062,  0.0013, -0.0053,  0.0044,  0.0085,  0.0590,
         0.0249], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0095,  0.7641,  0.0717,  0.0014, -0.0023,  0.0192, -0.0091, -0.0162,
        -0.0066,  0.0177,  0.0047, -0.0011,  0.0099, -0.0071,  0.0162, -0.0147,
        -0.0125,  0.0120,  0.0040,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0825, -0.7179,  0.0247, -0.0345,  0.0212, -0.0226, -0.0187,  0.0245,
        -0.0047, -0.0017, -0.0066,  0.0093,  0.0197,  0.0114,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5417e-01, -5.8614e-01, -2.6234e-02, -1.2167e-02, -5.4782e-02,
        -1.4783e-02, -2.4297e-03,  2.6939e-03, -1.4092e-02, -5.6320e-03,
        -6.5356e-03, -3.6323e-03, -1.4234e-02,  5.1179e-04, -2.9438e-03,
        -7.3602e-03,  2.6229e-03, -6.7506e-03, -8.8586e-03,  5.5003e-03,
         3.8742e-02, -2.9177e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9311e-02,  7.0284e-01,  1.7750e-02,  2.5410e-02, -8.9989e-03,
         7.8723e-03, -4.2668e-03, -1.8950e-02,  1.0832e-02, -1.0129e-03,
         1.4376e-02,  1.4772e-02,  7.9657e-04, -8.4596e-04,  1.1695e-04,
        -1.9671e-03,  2.6620e-02,  9.3428e-03, -7.2187e-03,  6.0198e-03,
         3.9621e-03,  4.9466e-03, -4.7056e-04, -8.8448e-03,  1.5621e-03,
         2.7269e-03, -2.1892e-03,  1.0268e-02,  1.7845e-02,  4.7871e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6678e-01,  3.8716e-01, -1.3218e-03, -2.8763e-04,  3.9689e-03,
         1.5384e-02, -2.9131e-03, -2.9200e-02,  4.0775e-03, -7.5633e-03,
         1.1144e-02,  3.4974e-02,  1.2696e-02,  1.2354e-02,  7.3130e-03,
         1.0639e-01,  9.6473e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0188,  0.5808,  0.1032,  0.0319, -0.0329,  0.0044, -0.0351,  0.0580,
        -0.0784,  0.0449,  0.0116,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0435,  0.7121,  0.0231, -0.0120,  0.0229, -0.0287,  0.0081, -0.0049,
        -0.0067, -0.0033, -0.0085, -0.0164,  0.0026,  0.0242, -0.0134, -0.0445,
         0.0250,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0064, -0.7262, -0.0376, -0.0095, -0.0243,  0.0093,  0.0122,  0.0013,
        -0.0120,  0.0080,  0.0046, -0.0070,  0.0129,  0.0146,  0.0044, -0.0102,
        -0.0012,  0.0031, -0.0007, -0.0030, -0.0009, -0.0066,  0.0051, -0.0011,
        -0.0298,  0.0479,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1226e-02,  6.2674e-01,  7.6955e-02, -1.2565e-02,  2.0654e-02,
         2.2209e-02, -1.5720e-02,  2.5212e-03, -2.1761e-02, -4.7569e-03,
         8.0423e-03, -2.2820e-03, -1.3621e-02, -1.1068e-02, -1.3956e-02,
        -2.3079e-03, -1.3375e-02,  2.7682e-05, -3.3547e-03, -1.1142e-02,
         6.9322e-03,  1.0647e-04,  1.8546e-03,  2.8414e-03,  1.0766e-02,
        -4.3455e-03, -1.4980e-03,  1.0505e-02, -3.0880e-02, -5.9810e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0324, -0.7403,  0.0121,  0.0340, -0.0260, -0.0037, -0.0312, -0.0070,
         0.0019, -0.0059,  0.0178,  0.0385,  0.0017, -0.0028,  0.0040, -0.0067,
         0.0008,  0.0090,  0.0243,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 1.4148e-02,  7.0514e-01,  9.3718e-02,  2.4570e-02,  8.6802e-03,
        -3.4411e-03,  1.7509e-03, -2.4484e-03,  8.7211e-03, -2.7679e-04,
        -6.6378e-03, -7.2748e-03, -3.2457e-03,  8.9921e-04, -8.2130e-03,
        -3.7718e-02,  4.1279e-03, -2.3027e-03,  2.5409e-03,  4.4614e-03,
        -3.0590e-03, -1.5964e-02,  3.0440e-03,  1.6228e-03, -2.6633e-03,
        -2.1693e-02, -1.1634e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2842, -0.2240, -0.0462, -0.0621, -0.0382, -0.0047, -0.0016, -0.0230,
        -0.0086, -0.0255,  0.0645, -0.0545, -0.1340,  0.0289,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.2145e-03,  6.9190e-01,  4.2132e-02, -8.4636e-03, -3.6348e-03,
        -1.6277e-02,  1.3444e-02,  9.3716e-03, -7.2408e-03,  1.5570e-03,
         3.1698e-04,  9.8931e-03, -2.4788e-03,  1.6332e-02, -1.7397e-03,
        -1.2245e-03, -3.5714e-04, -2.4592e-02, -7.7372e-04, -2.5874e-04,
         2.2758e-03, -2.7058e-03,  1.2763e-02, -1.9033e-03, -1.2288e-02,
        -1.0786e-01,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0868,  0.6982, -0.0115,  0.0097, -0.0281, -0.0205, -0.0084,  0.0086,
        -0.0161, -0.0177,  0.0036, -0.0270,  0.0380, -0.0257,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2533e-01,  6.1202e-01, -4.1898e-03,  1.3302e-02, -2.6785e-03,
        -5.7944e-04, -3.1288e-02, -1.1544e-02, -5.2962e-03, -9.6608e-03,
        -1.0579e-02, -1.7598e-02,  4.2199e-03, -6.0391e-04,  1.4202e-02,
        -4.3514e-03, -5.9632e-03, -2.4542e-03, -4.2261e-03,  4.4510e-03,
        -1.7236e-03, -3.0814e-03, -1.1086e-02, -2.1978e-02, -3.1547e-03,
        -5.0530e-03,  2.8550e-03, -6.1633e-03,  2.2992e-03,  6.1244e-03,
        -5.0501e-02,  1.4524e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0128,  0.1193,  0.0100,  0.0555,  0.0388,  0.0377,  0.0470,  0.0196,
         0.0250,  0.0046,  0.0265, -0.0341,  0.0119,  0.0334, -0.0353,  0.0434,
         0.0199,  0.0193,  0.0063,  0.0004,  0.0535,  0.0076, -0.0030,  0.0745,
        -0.2606,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0344,  0.6974,  0.0768, -0.1154,  0.0509, -0.0251,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.3738e-02, -6.8553e-01, -3.6198e-02,  4.1191e-02, -2.9628e-02,
         3.5047e-03,  9.5504e-03,  4.7668e-04, -8.7636e-04,  3.6890e-03,
         8.1641e-04, -8.1579e-03,  4.7477e-03,  2.9275e-03,  1.5998e-02,
         9.0597e-03,  3.5717e-03, -1.9789e-03,  6.7750e-03, -1.0464e-03,
         2.0579e-02,  1.8869e-03,  5.5366e-04,  1.6545e-03,  1.6332e-03,
        -7.9853e-03,  2.4774e-03,  2.7375e-03,  1.9977e-03, -4.0650e-03,
        -2.6162e-03, -2.9674e-04, -2.8858e-03, -3.1580e-04,  8.4988e-04,
         3.4058e-03, -4.5974e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.9851e-02,  7.0382e-01,  2.3323e-02,  2.1275e-02,  5.2491e-03,
        -3.2212e-03,  8.2061e-03, -9.7600e-03,  1.0038e-02, -7.5374e-03,
        -5.8976e-03, -3.4508e-03,  7.4403e-03, -3.5459e-03, -1.5564e-03,
         3.2179e-03, -9.3959e-03, -2.0764e-03, -9.5790e-03, -7.6930e-03,
        -1.0114e-02, -5.9886e-03, -7.6504e-03, -9.6493e-03,  1.9264e-04,
        -4.3405e-03,  1.0571e-02,  1.7601e-03, -2.7618e-03, -1.2469e-02,
        -2.2315e-03,  1.2728e-03, -1.7540e-03, -2.0526e-02,  2.5869e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1675e-02, -6.6256e-01,  9.1526e-02, -3.7466e-03, -5.5221e-02,
        -2.6458e-02, -1.5270e-03, -4.5679e-04,  1.1178e-02,  2.2126e-02,
        -1.6406e-02,  1.1753e-02, -1.0706e-03,  1.2368e-02,  2.1927e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4145e-02, -7.2700e-01, -3.4474e-03,  2.4229e-02, -3.4472e-03,
        -1.7108e-02,  1.4456e-02,  3.8987e-02, -1.0045e-02, -2.9590e-03,
        -6.6488e-03,  1.2389e-03, -6.3281e-03, -5.2016e-03, -4.8617e-03,
        -5.9630e-03, -1.0831e-02, -5.9735e-03, -2.1608e-03, -2.8619e-03,
        -4.7977e-03, -3.7783e-03,  1.0274e-03,  3.5648e-04, -7.1583e-03,
         1.7273e-03,  8.4675e-03, -3.3558e-03, -9.0279e-04, -1.0928e-03,
        -1.3758e-03, -1.0848e-02,  1.7219e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1880e-02, -5.7549e-01,  4.4857e-02,  1.5828e-02, -3.4836e-02,
         3.1195e-02, -2.3884e-04, -1.1745e-02,  3.4284e-03,  1.3056e-02,
        -3.1962e-02, -1.5120e-02, -2.3075e-02, -1.8862e-03,  1.0067e-02,
        -4.1933e-03, -6.6829e-03,  5.5000e-03, -2.7220e-02, -1.6169e-03,
        -1.0706e-02,  1.0974e-02,  2.1667e-02, -1.2967e-03,  3.4624e-04,
         6.7495e-04,  4.3629e-03, -1.5997e-02,  1.4096e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 1.1472e-01,  6.7342e-01,  1.7224e-02, -1.5794e-02,  2.4370e-02,
         3.1888e-03, -1.7042e-02, -1.0058e-03, -9.3379e-03, -6.6966e-04,
        -1.0695e-02, -1.1665e-02, -9.4247e-03,  1.3104e-02, -7.8336e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.1378e-03,  7.8841e-01,  1.1597e-02,  4.2663e-04, -1.4862e-03,
         2.0342e-03,  4.7934e-03, -2.4708e-03,  6.9359e-03, -1.0651e-02,
        -1.9903e-02, -3.9069e-03, -2.3314e-03, -5.6952e-04, -5.9659e-03,
        -3.1691e-03, -8.8363e-03, -8.1767e-03, -1.9641e-03, -1.3607e-04,
        -5.9434e-03, -1.7719e-03, -7.1812e-03, -2.9120e-03,  1.4799e-03,
         4.5542e-03,  9.2136e-04,  2.2486e-04, -7.4833e-03, -3.7417e-04,
         2.3556e-03, -4.3542e-03, -4.7946e-04, -1.0949e-03, -9.4070e-03,
        -1.9068e-03,  1.5854e-03,  7.0536e-04,  5.7178e-04, -5.0322e-03,
         3.2022e-03,  8.5459e-04, -4.4792e-03, -7.3103e-03, -1.3259e-03,
        -1.2082e-02,  1.9581e-04,  5.1793e-03,  1.3128e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0197,  0.6450, -0.0269, -0.0284,  0.0193,  0.0210, -0.0133,  0.0062,
        -0.0042,  0.0186,  0.0043, -0.0071,  0.0057,  0.0007, -0.0053, -0.0092,
         0.0054,  0.0048,  0.0063,  0.0029,  0.0009, -0.0007, -0.0111, -0.0043,
        -0.0014, -0.0071, -0.0021,  0.0170, -0.0017,  0.0015, -0.0021, -0.0017,
        -0.0055, -0.0688, -0.0198,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4149e-02, -7.9813e-01, -3.8259e-02,  3.0401e-05, -1.1962e-02,
        -4.2748e-03,  1.4796e-02, -1.8479e-02, -1.6561e-02,  9.6414e-03,
         6.0690e-03,  1.4817e-02,  4.7302e-03,  9.6465e-03, -1.6937e-02,
        -5.6549e-03, -1.0309e-02, -3.2886e-03,  2.2672e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.3265e-02,  6.9515e-01,  2.3283e-02,  5.7669e-02, -1.8442e-04,
         2.2095e-02,  1.6279e-02,  2.1235e-02,  4.5897e-04,  1.8554e-03,
        -3.0591e-03, -2.3170e-03,  3.0959e-02, -2.4014e-02, -2.8174e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0246,  0.7355, -0.0309,  0.0122,  0.0115,  0.0109, -0.0052,  0.0035,
        -0.0100, -0.0115, -0.0099, -0.0040, -0.0254,  0.0048,  0.0088, -0.0111,
        -0.0113, -0.0120, -0.0047,  0.0057, -0.0393,  0.0072,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.4591e-02, -5.6021e-01, -4.8240e-02, -2.2120e-02, -1.4512e-02,
        -1.4828e-02,  2.6748e-02, -3.6099e-03,  7.9806e-03, -5.9766e-03,
         1.6522e-02, -9.3717e-03,  1.8163e-02, -8.6959e-03, -3.8725e-03,
        -8.8201e-03, -6.9278e-04,  7.4513e-03,  5.5330e-03,  2.2065e-03,
         4.2553e-03, -1.9872e-02, -5.1387e-04,  7.4440e-03,  4.4737e-03,
        -4.1303e-03,  3.0002e-03, -8.9128e-03,  8.6323e-02, -9.3382e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1797,  0.5212, -0.0514,  0.0587, -0.0017,  0.0520, -0.0137, -0.0006,
        -0.0116,  0.0630,  0.0464,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1979e-02, -7.1352e-01, -2.1163e-03,  5.3380e-03,  2.5544e-02,
         2.4910e-03,  5.5317e-04, -7.6561e-04, -2.6808e-02, -2.0158e-02,
         7.8931e-03, -4.2381e-04, -3.3425e-02, -8.5463e-03, -1.6857e-04,
        -1.3177e-02, -7.5623e-03,  1.4711e-02, -1.8291e-03,  2.9849e-03,
        -5.9173e-02,  8.3133e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0033, -0.6471, -0.0499, -0.0023, -0.0192,  0.0353, -0.0025, -0.0105,
         0.0017,  0.0084,  0.0266,  0.0092,  0.0129,  0.0102,  0.0078,  0.0078,
         0.0059,  0.0038, -0.0102,  0.0091,  0.0018,  0.0043,  0.0078, -0.0040,
         0.0074,  0.0035,  0.0109,  0.0064,  0.0126,  0.0174, -0.0074,  0.0230,
        -0.0096,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0249, -0.5253, -0.1111, -0.0036,  0.0126, -0.0194,  0.0176,  0.0217,
         0.0049, -0.0173,  0.0043, -0.0044,  0.0191,  0.0356,  0.0151,  0.0198,
        -0.0074,  0.0114,  0.0106, -0.0066, -0.0088, -0.0033,  0.0058,  0.0013,
         0.0250,  0.0011, -0.0517,  0.0102,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5098e-01, -4.6095e-01, -1.7836e-01, -1.7796e-03,  1.9736e-02,
        -1.6579e-03, -5.7274e-03, -4.7115e-02, -3.1656e-03,  6.7536e-03,
         5.1466e-03,  4.9117e-03,  6.0568e-03,  6.6472e-05,  1.9949e-03,
        -1.3752e-02, -7.3405e-04,  5.7746e-03,  5.1517e-03,  4.2781e-03,
         7.2859e-03, -6.7943e-03,  7.8824e-03,  3.5925e-02,  1.8014e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-1.1283e-01, -3.6057e-01,  2.1140e-02, -7.5052e-03, -5.3002e-03,
         1.2446e-02,  2.2438e-02,  1.1827e-02, -8.1286e-03, -1.3037e-02,
        -1.3149e-02,  9.7958e-03,  8.5832e-04, -2.5264e-04,  2.2301e-03,
         1.8665e-03,  5.7178e-03,  1.4899e-03,  1.1498e-02, -4.5582e-03,
        -1.0613e-02,  2.9436e-02, -1.3228e-02, -3.4907e-03,  4.0673e-02,
         7.5400e-03, -2.3013e-04,  2.1787e-02,  5.8656e-03,  1.9307e-03,
         1.1076e-02,  9.8844e-03,  8.9483e-03,  1.0338e-03, -3.2885e-03,
         9.9441e-03,  3.1152e-03, -3.1449e-03, -7.8581e-03,  1.7340e-01,
         6.8636e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0235, -0.5734, -0.0312,  0.0398, -0.0142,  0.0471, -0.0024, -0.0391,
        -0.0105,  0.0182,  0.0291,  0.0380,  0.0128,  0.0043, -0.0238, -0.0033,
         0.0164,  0.0066, -0.0127, -0.0078,  0.0041,  0.0052, -0.0027,  0.0147,
        -0.0191,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0545, -0.4431, -0.0511,  0.0578, -0.0392,  0.0261,  0.0220,  0.0066,
         0.0302,  0.0294, -0.0208,  0.0077,  0.0021,  0.0113, -0.0058,  0.0006,
        -0.0250, -0.0408,  0.0343, -0.0013, -0.0904,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8730e-02, -5.7794e-01, -3.6983e-02,  2.6606e-02,  1.3008e-02,
        -2.4756e-02, -5.2086e-02,  3.8314e-02, -1.3789e-03,  5.5255e-02,
        -1.5583e-02,  1.9055e-02, -2.2164e-02, -8.1978e-03, -7.6083e-03,
        -2.4667e-04,  1.0090e-02,  6.2001e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1401e-01, -3.8039e-01, -2.1834e-03,  1.0874e-01,  2.4441e-03,
        -5.1538e-03, -3.2614e-02,  2.7036e-02,  2.0570e-02, -1.4513e-02,
         1.0669e-02,  1.6987e-02,  9.4181e-03,  5.8704e-03,  1.0233e-02,
         3.7696e-03, -1.3041e-02,  1.0745e-02,  2.6358e-03, -4.8512e-03,
        -1.4902e-02, -1.2333e-02, -8.2436e-03, -3.7369e-03, -8.4778e-03,
        -6.5244e-03, -1.0769e-04, -2.5779e-02, -1.3871e-02, -9.6479e-03,
        -4.3120e-03, -2.5860e-03,  7.0534e-04,  5.9199e-03,  1.4572e-02,
        -2.1730e-02,  1.7357e-02, -3.3319e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1819, -0.3356,  0.0151,  0.0148, -0.0029, -0.0185, -0.0013, -0.0341,
        -0.0566, -0.0289,  0.0178,  0.0055, -0.0062,  0.0762, -0.0336,  0.0050,
        -0.0245,  0.0533,  0.0655, -0.0227,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.6093e-02, -5.1016e-01,  1.0650e-01,  1.4639e-02,  7.9043e-02,
        -2.3178e-04,  2.9299e-02, -5.6015e-02, -3.5519e-02, -8.4044e-03,
        -6.5051e-02,  4.0112e-03,  4.9474e-03, -8.8292e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0355, -0.5746,  0.0150, -0.0018,  0.0074, -0.0054, -0.0132,  0.0201,
         0.0151, -0.0216,  0.0390,  0.0083, -0.0032, -0.0151, -0.0045, -0.0123,
        -0.0306, -0.0139,  0.0251,  0.0019,  0.0122, -0.0073,  0.0021, -0.0270,
         0.0050, -0.0078, -0.0047, -0.0013, -0.0124, -0.0111, -0.0113, -0.0340,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0878, -0.5510, -0.0351, -0.0157, -0.0543, -0.0072,  0.0110,  0.0161,
         0.0030, -0.0236, -0.0083,  0.0211,  0.0041, -0.0195,  0.0092, -0.0054,
        -0.0036, -0.0356,  0.0213,  0.0669,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0408e-02, -2.5411e-01,  1.4601e-02,  4.9620e-02,  2.5778e-02,
         2.1073e-02, -1.7355e-02,  1.1127e-02,  1.3824e-02, -4.2213e-02,
        -9.1887e-04, -9.2463e-03,  1.0357e-02, -1.4663e-02, -9.2284e-03,
         2.3164e-03, -3.5630e-03, -8.5181e-03,  9.2274e-03,  3.9814e-03,
        -1.9695e-02,  6.9193e-03,  1.6366e-04, -1.0780e-02, -1.1878e-02,
         1.5190e-02, -1.7583e-03,  5.4249e-03, -9.8807e-04, -1.4364e-02,
        -4.7593e-02, -1.8866e-02, -3.2012e-02, -1.2358e-02, -1.4670e-02,
         6.9464e-03, -3.0060e-02,  3.9816e-03,  1.0464e-02, -1.3644e-02,
        -1.7316e-03,  1.3004e-03, -3.1660e-02, -2.7030e-02, -2.3199e-02,
         6.9703e-02,  2.5497e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1138, -0.5862,  0.0713, -0.0042, -0.0291, -0.0023,  0.0078,  0.0606,
         0.0092,  0.0137,  0.0214, -0.0176, -0.0150,  0.0343, -0.0135,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1302e-01,  2.5280e-01, -2.8220e-02,  3.3643e-02, -8.8694e-03,
         1.8799e-02, -8.2242e-03,  8.6060e-04,  5.5288e-03,  1.5016e-02,
        -8.9984e-03, -4.5170e-03, -4.8177e-04,  2.0464e-02, -1.4029e-02,
         9.4162e-04,  1.5605e-05, -4.5578e-03,  2.3719e-03, -2.0431e-03,
         2.5356e-03,  1.7513e-03,  5.8161e-03,  5.2314e-03, -1.2604e-02,
        -4.9397e-02, -3.1307e-03,  8.3220e-03,  6.5630e-04, -2.2594e-03,
         2.9535e-03,  7.2766e-03, -7.8237e-03,  3.2160e-03, -1.5399e-02,
        -7.2607e-03, -7.1764e-03, -8.5862e-03,  3.4498e-03, -4.5213e-03,
        -3.5797e-04, -1.6387e-03, -1.7344e-03,  6.0988e-03,  6.1355e-04,
         4.3971e-03, -2.3852e-01, -4.3873e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-3.6592e-02,  7.3582e-01,  4.3314e-02,  1.3416e-02,  1.3121e-02,
         3.6729e-02, -5.8496e-03, -5.0292e-03,  5.8637e-04,  3.6028e-03,
         3.3903e-03,  2.5069e-02,  4.4290e-03,  1.9986e-03, -6.6891e-04,
         2.5448e-03, -5.0674e-03, -5.5854e-03, -1.7337e-03, -1.1076e-03,
        -2.3825e-04,  1.6535e-03,  6.5286e-03,  5.3555e-03,  1.0816e-02,
        -2.9752e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.9463e-02,  5.7445e-01,  2.3308e-02,  5.2055e-02,  3.6901e-02,
        -6.4172e-03, -2.0638e-02,  4.5133e-03,  4.1410e-02, -1.7688e-03,
         1.5160e-02, -2.5565e-04,  1.6858e-03,  2.3135e-03, -1.1835e-02,
         6.8054e-03,  1.4049e-02, -2.5436e-02, -1.1112e-02,  5.5098e-02,
         1.5322e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2035e-02,  6.6889e-01,  2.9596e-03,  1.5824e-03, -4.8197e-03,
         3.5515e-03, -9.7462e-03,  1.2271e-02,  2.9690e-03,  7.7032e-03,
        -4.3042e-03, -4.8833e-03, -6.4960e-03, -1.0649e-02, -3.6636e-03,
        -1.9903e-03, -1.4696e-02,  2.1831e-03, -1.9192e-02, -1.1224e-02,
        -6.4511e-03,  1.8384e-03,  2.3756e-02, -6.8346e-03,  6.5879e-03,
        -2.3500e-03, -1.6110e-02, -4.1972e-03,  2.8133e-05,  1.0660e-02,
        -8.8225e-03,  4.0872e-03,  6.6476e-03,  2.1405e-03, -1.9138e-02,
        -4.4544e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0934e-01,  6.2245e-01, -8.4572e-04,  3.0062e-02, -1.0523e-02,
         5.8312e-04,  8.9684e-03, -4.5682e-02,  2.1071e-02,  3.4822e-03,
        -3.7855e-03,  1.5005e-02,  7.3125e-03, -5.1672e-03, -1.9369e-03,
         1.7163e-02, -1.5642e-02, -9.8186e-03,  2.1203e-02, -4.9955e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0011,  0.6567,  0.0218, -0.0027,  0.0254,  0.0017,  0.0173,  0.0065,
         0.0243, -0.0726,  0.0221, -0.0182, -0.0064, -0.0048, -0.0216, -0.0030,
        -0.0155, -0.0128,  0.0087, -0.0362, -0.0205,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0859,  0.7023,  0.0311, -0.0096, -0.0578, -0.0028,  0.0169,  0.0036,
         0.0123,  0.0040,  0.0073, -0.0559,  0.0106,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3906e-02,  6.3643e-01,  2.8266e-02, -5.4605e-03, -1.9188e-02,
        -1.9328e-02,  6.3901e-03, -3.8320e-03, -2.1113e-02,  8.1523e-03,
        -1.6645e-02, -8.0834e-03, -5.4384e-03, -4.4173e-03, -4.5871e-03,
         1.0428e-02, -9.1560e-03, -5.4636e-03, -7.8054e-04,  1.9275e-03,
        -5.5599e-03, -5.0611e-03,  3.4073e-03,  6.5137e-03,  4.6575e-03,
        -5.1038e-04, -1.6719e-03,  3.2080e-03,  2.5845e-04, -1.2737e-02,
        -3.1075e-03,  3.0053e-03,  1.8979e-03, -2.3535e-03, -6.7411e-03,
        -4.0757e-02, -2.9557e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0868e-02,  7.4905e-01,  3.0008e-02,  1.4479e-02, -3.5887e-02,
        -6.3848e-03, -5.3486e-03, -5.3529e-03, -1.2760e-02, -3.9163e-03,
        -7.8908e-03, -3.9580e-03,  1.9338e-03, -2.1818e-03, -1.6384e-02,
        -2.8564e-03, -3.4701e-03, -3.3991e-04, -3.6008e-03, -2.2168e-03,
        -1.0010e-02, -3.3808e-03,  1.4940e-02,  4.2784e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0063,  0.6719, -0.0038, -0.0157, -0.0190, -0.0092, -0.0042, -0.0109,
        -0.0282, -0.0051, -0.0159, -0.0454, -0.0129, -0.0072,  0.0134, -0.0138,
        -0.0101, -0.0090,  0.0026, -0.0099, -0.0023, -0.0054,  0.0027,  0.0009,
         0.0054, -0.0044, -0.0020, -0.0018, -0.0052, -0.0007,  0.0222, -0.0324,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6835e-02,  6.8156e-01,  1.2499e-02,  1.8985e-02,  4.0641e-03,
        -7.8541e-03, -7.4386e-03, -9.8672e-03,  3.0583e-03, -1.0192e-02,
         3.0109e-03, -1.5412e-03, -3.7916e-03, -6.6450e-03, -1.5323e-03,
        -9.0861e-03, -1.2921e-03, -2.9144e-03,  7.5708e-04,  1.6837e-04,
        -2.9590e-02,  2.3365e-02, -7.1421e-04,  8.9059e-04, -1.1576e-02,
         1.3333e-02, -8.3678e-03, -4.2500e-03, -9.2575e-03, -4.0733e-02,
        -2.4832e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5658e-02,  6.8827e-01, -5.3591e-02,  5.1452e-04, -1.8422e-02,
        -2.2364e-02, -3.4689e-02,  1.5399e-03,  1.9929e-02,  1.8895e-03,
         8.7580e-03, -4.1415e-03, -7.8051e-03, -9.9765e-04, -7.4175e-03,
         2.2274e-03,  8.9446e-03,  1.2631e-02,  4.8826e-04,  1.1356e-02,
         3.9346e-02,  2.4936e-02,  1.4089e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0451e-02,  6.8777e-01, -1.4113e-03, -3.6608e-03,  1.3901e-02,
         1.0632e-03, -8.3679e-03, -3.9745e-02, -1.2394e-03, -5.1853e-03,
        -1.9977e-02, -9.5013e-03, -1.8983e-03, -1.9423e-02, -1.0919e-02,
        -2.7231e-03, -1.4738e-03, -5.1028e-04, -7.4328e-03, -1.1694e-02,
        -1.6611e-02, -8.1765e-03, -4.3071e-03,  5.6428e-03,  7.8975e-04,
        -1.9009e-02,  3.6045e-02,  1.0662e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 8.9039e-02, -6.2545e-01,  1.7712e-02, -3.8048e-05,  1.8597e-02,
        -8.8573e-03, -9.6241e-03,  1.9093e-02,  1.1552e-02,  1.7859e-02,
         9.5100e-03,  4.6370e-03,  6.3293e-03,  5.0971e-03, -1.2575e-03,
         7.8223e-03,  7.6792e-03, -1.5383e-02,  2.0857e-03,  1.7968e-02,
        -1.0104e-02, -1.4024e-03,  1.5687e-02,  8.8056e-03,  1.1238e-02,
        -3.8038e-03, -3.9057e-03,  2.0210e-02,  2.9257e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1902e-04,  7.2621e-01, -3.5897e-03,  1.5544e-02, -1.6537e-03,
        -1.5567e-02, -1.7537e-02,  2.4722e-03,  1.7564e-03,  1.2720e-02,
        -1.3855e-02, -1.0679e-02, -1.0831e-02, -7.0456e-03,  1.2464e-02,
         1.1110e-02,  1.6665e-02,  1.6000e-02,  1.4438e-02, -2.6193e-02,
        -1.7143e-02, -2.3447e-02,  5.2981e-03, -1.7561e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0116,  0.6075,  0.0134, -0.0043,  0.0287,  0.0324,  0.0021,  0.0137,
         0.0035, -0.0059,  0.0053,  0.0190, -0.0024, -0.0140, -0.0080, -0.0028,
        -0.0094, -0.0020, -0.0143,  0.0074, -0.0032, -0.0058, -0.0024,  0.0019,
         0.0136, -0.0078, -0.0027,  0.0028, -0.0105, -0.0038, -0.0154, -0.0058,
         0.0745, -0.0420], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9073e-02,  7.3820e-01,  3.0968e-03,  7.7935e-03, -2.1792e-02,
         1.8949e-02,  2.3839e-02,  2.5366e-02, -4.3109e-02, -2.3327e-03,
        -2.1213e-04,  8.3688e-04,  1.4747e-03, -4.8808e-03, -6.2616e-03,
        -1.0005e-02, -1.1465e-02, -4.2705e-04,  8.6630e-03,  1.9527e-02,
         1.6961e-02,  1.5729e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.4782e-02,  6.9518e-01, -3.0506e-03, -2.2246e-03,  1.3380e-02,
        -3.3528e-02, -2.2982e-02,  2.3187e-02, -3.1729e-02, -1.0294e-02,
        -4.0612e-03, -6.6717e-03, -1.0150e-05,  8.7338e-03, -3.2906e-03,
        -7.3886e-03, -6.6758e-02, -1.2752e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6672e-02,  7.0986e-01, -2.6684e-03,  1.1535e-02,  4.0098e-02,
         5.0435e-03,  3.6842e-04,  1.6680e-02,  1.0215e-02,  7.4436e-03,
        -7.9292e-03, -2.8875e-03, -1.4527e-02, -1.6425e-02,  1.4111e-03,
        -8.2355e-04,  6.1098e-03, -4.8492e-03,  1.2189e-02, -3.8036e-03,
        -4.5976e-03, -6.5712e-02, -2.8154e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.5499e-02,  6.4178e-01,  1.5235e-03, -1.9252e-04, -2.1824e-02,
        -2.6640e-04, -4.5740e-02,  1.0078e-02,  3.7299e-02,  8.8473e-03,
        -1.6622e-02, -1.5319e-03, -7.8724e-03,  2.0151e-02,  3.4875e-02,
         1.8281e-02,  5.7621e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.6860e-02, -5.8759e-01, -2.2138e-02, -2.8581e-02, -7.5875e-02,
         1.8741e-02,  4.4867e-02,  4.2826e-03,  2.3794e-02,  1.2011e-02,
        -1.1459e-02,  1.8615e-03, -1.1149e-02, -4.0370e-04, -6.9634e-03,
         2.0959e-02, -4.9297e-02, -1.3170e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0545,  0.6181, -0.0510, -0.0268, -0.0073, -0.0051,  0.0020,  0.0027,
        -0.0099,  0.0264,  0.0203, -0.0029, -0.0239, -0.0110, -0.0138, -0.0020,
         0.0063, -0.0097,  0.0193,  0.0070,  0.0132,  0.0038,  0.0029, -0.0146,
        -0.0025, -0.0428,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.4842e-02,  5.7192e-01,  7.3358e-02, -4.1158e-02,  2.2849e-02,
         1.3842e-03,  4.4779e-03,  6.3590e-03,  2.6882e-03, -2.3906e-02,
         1.4486e-03,  1.3455e-02, -4.9302e-03, -1.6886e-03,  1.9364e-03,
         8.1258e-03, -4.1922e-03, -1.3902e-02,  5.5866e-03, -6.4302e-03,
        -4.4310e-03,  1.5395e-03,  1.3205e-02, -7.2068e-03,  2.3151e-04,
         3.7091e-04, -2.7743e-03,  5.7645e-04, -5.0567e-03,  4.4445e-02,
        -2.5526e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0426, -0.6102, -0.0144,  0.0471,  0.0226, -0.0088, -0.0118, -0.0309,
        -0.0043,  0.0213,  0.0031, -0.0028, -0.0328,  0.0017, -0.0113,  0.0722,
         0.0220,  0.0399,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0235,  0.8374, -0.0512, -0.0151, -0.0263,  0.0257,  0.0082,  0.0127,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 4.8653e-02,  7.3619e-01,  3.4017e-02, -1.0883e-02,  2.8590e-03,
         1.7121e-02, -9.3922e-03, -7.7437e-03, -5.0633e-03, -2.6380e-03,
         3.1545e-04,  6.4367e-03, -3.9847e-05, -2.6100e-04, -3.2415e-03,
         1.4352e-02,  4.3706e-03,  4.1785e-03,  5.8218e-03, -3.5799e-03,
         2.0221e-02, -1.0644e-02,  2.1766e-03,  2.9894e-04, -3.1768e-03,
         2.4029e-03, -6.9045e-03, -1.7961e-03, -4.5215e-03, -4.6641e-04,
         1.6082e-03, -2.6534e-03, -3.0974e-03,  4.1109e-04,  1.7756e-03,
        -8.3925e-03, -2.1696e-03, -9.1818e-03,  9.4364e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.4340e-02,  6.5680e-01, -2.5971e-02,  9.8165e-03,  2.6168e-02,
         5.1995e-03,  7.8262e-03, -1.8050e-02, -2.3354e-02,  1.0720e-02,
        -5.4763e-03,  7.2827e-03, -1.5896e-03,  9.5142e-03,  8.4530e-03,
         7.8590e-03, -7.6972e-05, -1.5914e-02, -1.5147e-03,  5.4651e-03,
        -3.4493e-03,  1.2455e-02, -1.1729e-02,  2.5724e-03, -1.4673e-02,
         7.8516e-03, -1.0534e-02,  3.5345e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0964, -0.7272, -0.0100,  0.0439, -0.0079,  0.0048,  0.0050,  0.0068,
        -0.0080, -0.0126, -0.0073,  0.0256,  0.0445,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0008,  0.7750, -0.0312, -0.0042,  0.0140, -0.0190,  0.0054,  0.0113,
        -0.0103,  0.0194, -0.0027, -0.0116,  0.0088,  0.0367, -0.0494,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.0379e-03,  5.5422e-01,  3.9349e-02, -6.3150e-02,  1.5180e-02,
         1.0659e-02, -8.5262e-04, -8.6431e-03, -3.1378e-03,  8.8344e-03,
        -1.1347e-02,  8.4258e-03,  9.2403e-04,  7.5496e-03,  8.2918e-03,
         2.9563e-03, -5.9420e-03,  3.9330e-03, -3.6987e-03, -2.1579e-03,
         4.6286e-04,  6.4248e-03,  6.7234e-03,  5.6023e-04,  1.5858e-02,
         1.6076e-02,  5.2810e-03,  8.5436e-03, -3.4970e-03, -2.0856e-04,
        -1.7074e-02,  5.3636e-03,  7.2582e-03,  7.5479e-04,  1.0374e-02,
        -2.0340e-03, -9.4358e-03,  1.7829e-03, -6.6372e-03, -3.8736e-03,
        -4.9921e-03, -5.0744e-03,  2.3244e-03,  7.8668e-05,  1.1279e-02,
        -2.0042e-02, -6.0695e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9641e-02,  7.1064e-01,  2.3974e-02,  9.9750e-03,  3.8779e-03,
        -4.4953e-04,  7.4900e-03, -1.2506e-03,  2.0721e-02, -9.3377e-04,
         1.6494e-02, -1.5592e-03, -3.9501e-03, -6.4385e-03, -2.3626e-03,
        -6.6531e-03,  1.7177e-04, -6.3060e-03, -4.4155e-03, -4.7201e-03,
         4.9444e-04, -1.9373e-03,  4.9464e-03,  2.1415e-03, -2.5742e-03,
         1.0934e-03,  8.4412e-03,  3.1177e-03, -5.3372e-03,  9.1711e-03,
        -3.0384e-03,  5.0076e-04,  8.4336e-03, -3.1089e-03,  2.9769e-03,
        -4.8217e-03,  2.7728e-03, -1.2957e-03, -5.7145e-03, -2.5730e-04,
        -8.0450e-03, -4.7757e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0222,  0.6681,  0.0336,  0.0054,  0.0246,  0.0316,  0.0250,  0.0066,
        -0.0051,  0.0307,  0.0138,  0.0315,  0.0062, -0.0042,  0.0143,  0.0200,
         0.0213,  0.0096, -0.0045,  0.0026, -0.0025, -0.0016, -0.0027, -0.0103,
        -0.0021,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0814,  0.7380, -0.0068, -0.0120, -0.0017, -0.0036, -0.0049,  0.0015,
        -0.0074, -0.0062,  0.0020,  0.0056, -0.0021, -0.0022, -0.0167, -0.0143,
         0.0091, -0.0219,  0.0044, -0.0047, -0.0155, -0.0116, -0.0264,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4094e-02,  7.0453e-01,  4.5165e-02, -1.6481e-02, -1.4108e-02,
         1.8429e-02, -2.8038e-03, -4.8151e-03, -3.7791e-03,  2.8030e-02,
         5.0786e-03,  1.2421e-03, -4.7350e-03,  2.1623e-02, -3.2831e-02,
        -2.8684e-03, -1.0959e-02,  2.4941e-03,  5.4026e-04, -7.0350e-03,
         6.6265e-03,  1.8109e-02, -2.3627e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0387,  0.6535,  0.0619, -0.0419,  0.0366,  0.0028,  0.0108,  0.0063,
         0.0166,  0.0052,  0.0303,  0.0331,  0.0624,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0521, -0.7216,  0.0163,  0.0083,  0.0443, -0.0047,  0.0182, -0.0079,
        -0.0081, -0.0301, -0.0111,  0.0354,  0.0022,  0.0397,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6220e-02,  8.0649e-01,  3.6607e-02, -9.8781e-03, -1.0338e-03,
        -3.2748e-03, -4.8057e-03, -3.0163e-03,  8.7227e-03, -1.6713e-02,
         1.4305e-04,  1.7912e-02,  5.9240e-03, -7.5343e-03,  1.2830e-02,
        -8.3003e-03, -2.2322e-03, -1.5238e-02,  2.3122e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-5.8061e-02,  6.2008e-01, -8.9052e-02, -2.7607e-02,  1.7986e-02,
         8.8500e-03, -6.9032e-03,  1.7650e-02, -1.9313e-02,  8.3868e-03,
        -7.2130e-03,  1.6643e-04, -1.7150e-03,  4.4460e-03, -1.2098e-02,
        -1.1242e-02,  6.6459e-03, -8.0684e-03, -1.6712e-03,  1.1986e-02,
         2.6714e-03, -1.0729e-02, -8.5643e-03, -8.9992e-03,  2.8936e-02,
        -9.5941e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0359,  0.6876,  0.0313,  0.0076, -0.0040,  0.0047,  0.0089, -0.0077,
         0.0036, -0.0273,  0.0176,  0.0082, -0.0393, -0.0624, -0.0052, -0.0485,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7480e-02,  6.4524e-01, -5.6243e-02, -1.4147e-02, -9.8106e-03,
         2.6470e-02,  2.9731e-02,  5.4566e-03,  3.0129e-02, -2.5598e-02,
        -2.5316e-03, -7.5535e-03, -6.5921e-05,  1.7787e-02, -2.4754e-02,
         4.1140e-02,  2.5859e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0866,  0.6913,  0.0546, -0.0406,  0.0346, -0.0007, -0.0047, -0.0832,
        -0.0037,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.4659e-02, -4.7121e-01, -7.1177e-02,  1.4971e-02,  1.7959e-02,
        -4.6344e-02, -2.5053e-02, -1.1552e-02,  1.8201e-02, -7.1351e-03,
         1.1933e-02,  1.1072e-02, -5.1140e-03, -5.6437e-03, -1.1188e-02,
        -1.3502e-02, -1.1377e-02, -1.3138e-02, -2.5705e-03, -9.0993e-04,
        -1.1047e-02, -1.3726e-02,  3.1243e-03, -3.7197e-03, -1.6764e-02,
         1.3519e-02, -1.9386e-03,  2.1456e-03, -8.1161e-03, -8.1738e-05,
        -2.1882e-03, -6.0176e-03, -2.8526e-03,  5.9496e-03, -3.6090e-03,
         4.0394e-03,  2.5403e-03, -8.7449e-03,  3.9506e-03,  6.1288e-03,
         5.9450e-03,  7.9620e-03,  1.1383e-02,  1.8615e-02, -1.7131e-03,
        -2.9469e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3358e-03, -6.9588e-01, -5.5799e-02,  2.4186e-02, -1.2000e-02,
        -4.2853e-03, -1.6313e-02, -1.2083e-02,  1.0924e-02, -5.1677e-03,
         2.4453e-03,  4.7490e-04, -1.4509e-02, -1.0717e-02,  5.2690e-03,
        -1.8312e-02, -3.8376e-04, -1.1578e-02, -1.8245e-03, -9.8747e-03,
        -6.9112e-03, -6.9790e-03,  1.0514e-03, -3.4740e-04, -3.8762e-03,
        -3.3828e-03, -4.8551e-03, -2.0205e-03, -4.5165e-02,  1.0048e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([0.0445, 0.8450, 0.0170, 0.0025, 0.0164, 0.0592, 0.0154, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0324,  0.7385, -0.0241, -0.0508, -0.0076, -0.0126,  0.0206,  0.0153,
         0.0256, -0.0299, -0.0425,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0242,  0.7576, -0.0045,  0.0258,  0.0282, -0.0089,  0.0100,  0.0101,
        -0.0206, -0.0084, -0.0177,  0.0202, -0.0244, -0.0395,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.3308e-02,  6.7752e-01,  5.0663e-02, -1.9167e-03,  4.3074e-03,
        -1.6813e-03,  2.4390e-03, -4.2342e-03, -9.8080e-03, -9.3652e-03,
        -9.0635e-03,  1.4874e-02, -8.5328e-03, -1.7099e-02,  1.5915e-03,
        -9.6401e-05, -2.9666e-03,  2.5646e-03,  1.8830e-03,  9.5708e-04,
         2.8389e-03,  8.4724e-03, -5.4839e-03, -4.9708e-03, -3.6572e-03,
         3.0949e-04, -4.4824e-03,  1.8237e-03,  9.3414e-03,  1.6451e-02,
         2.2383e-05,  2.2397e-02, -1.0868e-02,  2.4007e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3154e-02,  6.9516e-01,  4.2588e-02, -9.2162e-04, -7.9495e-03,
         2.2074e-03,  1.0278e-02, -1.7774e-02,  1.0027e-02,  1.3586e-02,
         1.5446e-02, -3.2546e-03, -9.2332e-04,  6.4233e-03,  6.7944e-03,
        -1.4799e-03, -1.2932e-02, -1.2295e-02,  2.8904e-03, -8.8033e-04,
         1.1020e-02, -2.6562e-03, -1.3956e-03,  4.7295e-04,  1.1499e-02,
         2.2532e-03, -8.3882e-03,  5.5348e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0181,  0.6904, -0.0076, -0.0312, -0.0233,  0.0108,  0.0073,  0.0008,
         0.0047, -0.0058, -0.0140, -0.0055, -0.0151,  0.0318,  0.0177,  0.0032,
         0.0095, -0.0103,  0.0064,  0.0009,  0.0079,  0.0122, -0.0375, -0.0279,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([-1.1501e-02,  5.2240e-01,  6.9415e-02,  4.3112e-02,  2.0108e-02,
         1.8110e-02, -3.2980e-03, -5.1443e-03,  3.6550e-03,  1.9116e-03,
        -2.3647e-02,  9.4067e-04, -6.8025e-03, -1.7792e-02,  1.2519e-02,
         1.0062e-02, -9.7728e-03,  4.9945e-03,  1.2993e-02,  1.1635e-03,
        -3.2991e-04,  4.9847e-04,  1.1214e-02, -1.2501e-03, -1.2773e-02,
        -1.0640e-02, -6.6888e-03, -4.0291e-03,  6.2332e-03,  1.2331e-01,
         2.3690e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.1916e-02,  8.0758e-01,  2.3301e-02, -3.4972e-02,  2.1666e-02,
         8.3782e-03, -7.1540e-03,  1.6059e-02,  4.7177e-03, -1.1005e-03,
         4.1271e-04,  2.7432e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.2806e-02,  6.6675e-01,  4.0353e-02,  2.8743e-02,  3.0388e-02,
         2.6795e-02, -1.9948e-03,  1.0082e-02, -3.8171e-04,  1.3445e-02,
         2.5277e-02, -2.9723e-02,  1.9084e-02,  3.3913e-02,  2.0265e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.3917e-02, -7.3235e-01, -2.4557e-02,  9.1577e-03,  6.4532e-03,
        -4.0410e-03, -1.5790e-02, -1.2076e-02,  7.5027e-03,  1.1941e-03,
         1.0165e-02, -7.2693e-03, -1.0606e-02,  8.2054e-03,  7.7264e-03,
        -4.0957e-03, -3.6683e-04,  9.7049e-03, -1.1542e-02,  1.0188e-02,
        -9.1473e-04,  3.6270e-03,  4.5910e-03,  4.5022e-04, -9.4404e-03,
         8.0839e-03, -3.1296e-03,  2.4814e-03,  8.6803e-03,  8.5278e-03,
         2.4460e-03,  1.0549e-02, -1.6722e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0015, -0.7039, -0.0855,  0.0392, -0.0160, -0.0025,  0.0174,  0.0086,
         0.0063, -0.0025,  0.0277,  0.0022,  0.0175, -0.0133,  0.0156,  0.0404,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0157,  0.7870, -0.0102,  0.0604,  0.0403,  0.0285,  0.0333,  0.0090,
        -0.0155,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0433,  0.8650, -0.0453, -0.0300, -0.0163,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1103,  0.5972, -0.0142, -0.0319, -0.0340, -0.0120,  0.0051, -0.0115,
         0.0079, -0.0036,  0.0101,  0.0105,  0.0097,  0.0122, -0.0011, -0.0016,
         0.0140, -0.0027,  0.0081, -0.0075,  0.0016,  0.0061, -0.0018,  0.0057,
        -0.0060, -0.0162,  0.0573,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1133,  0.7407,  0.0208, -0.0079,  0.0813,  0.0322,  0.0038,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0314,  0.7420,  0.0137, -0.0189, -0.0112, -0.0055,  0.0059,  0.0142,
        -0.0188, -0.0053, -0.0019,  0.0135,  0.0105,  0.0071,  0.0188,  0.0129,
         0.0116, -0.0211,  0.0103, -0.0026, -0.0080,  0.0146,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2401e-02,  8.0244e-01,  6.4800e-03, -7.4983e-03, -1.5573e-02,
         1.9591e-02,  6.2062e-03, -7.9120e-04, -4.4015e-03, -1.9071e-02,
        -5.1499e-03,  1.9244e-02, -2.3054e-02,  1.1913e-02, -4.2800e-03,
         1.6751e-02,  3.2790e-03, -6.9632e-03, -4.9121e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.0309e-03,  7.2812e-01,  2.5813e-02, -1.3438e-02, -6.1523e-03,
         9.8228e-03, -2.8464e-02,  7.3431e-03, -1.1067e-04, -4.1466e-03,
        -3.1195e-03, -1.3558e-03, -5.0396e-03, -6.2456e-03, -3.0230e-03,
        -5.2406e-03,  1.6882e-02, -1.1296e-03, -5.2370e-03, -2.3983e-02,
         4.1865e-04,  7.8534e-03, -1.3043e-03,  6.9122e-03,  1.0688e-02,
        -5.6069e-03, -9.5337e-04,  1.4776e-03,  2.2313e-04, -4.6873e-02,
        -1.5992e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 6.5724e-02, -5.8190e-01, -1.3078e-01, -4.0309e-03, -3.6677e-03,
         3.2715e-03,  6.3967e-03, -2.2722e-03,  1.9775e-02, -4.2155e-04,
        -4.0703e-03, -6.9755e-03, -1.1076e-02,  2.2085e-03, -9.9369e-03,
        -6.7499e-05, -1.0774e-03, -5.6657e-03,  9.5091e-03, -1.2500e-02,
         4.0735e-04,  1.7169e-04,  1.6502e-03, -9.3573e-03,  2.6954e-03,
         2.0689e-03, -1.5662e-03,  2.0693e-02,  7.3051e-03, -1.7955e-03,
        -5.5593e-02, -1.5368e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0760, -0.6833, -0.0319, -0.0205, -0.0103, -0.0041, -0.0029,  0.0026,
        -0.0106,  0.0143,  0.0490, -0.0116,  0.0032,  0.0034, -0.0043,  0.0016,
        -0.0063, -0.0080,  0.0048, -0.0015, -0.0241,  0.0257,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5818e-02,  6.7113e-01, -6.0325e-03, -3.1678e-02,  3.5919e-04,
         9.4093e-03, -3.4346e-02,  7.0478e-03, -2.7465e-04, -9.4370e-03,
         3.8357e-03,  1.5607e-02,  9.4815e-03,  6.0351e-03, -1.3672e-02,
        -5.1971e-03, -8.9667e-03,  1.2632e-02,  6.0125e-03, -3.0695e-03,
        -6.6367e-03, -1.8867e-02,  3.7256e-03, -1.1744e-02, -1.2858e-02,
         1.1834e-02, -4.1984e-03, -6.2435e-03, -2.7058e-03,  2.1141e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0215,  0.8267, -0.0377, -0.0066, -0.0042,  0.0069,  0.0009, -0.0043,
        -0.0059,  0.0042, -0.0144,  0.0114, -0.0106,  0.0060,  0.0081, -0.0044,
        -0.0189, -0.0075,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9464e-02, -6.6196e-01, -8.3950e-02, -5.1277e-03, -1.5477e-02,
         2.0370e-02,  7.1641e-04,  3.5141e-04,  1.6738e-02, -1.0275e-02,
        -9.8941e-04, -5.8974e-03, -5.8751e-03, -6.4350e-04,  2.1086e-03,
         2.6802e-02,  2.4074e-03,  2.0670e-02,  1.8015e-03,  1.0028e-02,
         1.4922e-03,  8.6554e-03,  1.8844e-03, -8.2015e-04,  3.2263e-03,
         9.5163e-03, -5.2272e-03, -6.7523e-04, -1.0119e-03, -6.0158e-04,
         9.9519e-04,  2.8700e-03,  1.1678e-03,  5.9245e-03, -2.4097e-03,
         1.6063e-02,  1.5806e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7458e-02, -6.0719e-01,  1.4353e-01,  2.4835e-02,  1.3009e-02,
         6.5320e-03,  2.2912e-04, -1.4767e-02,  3.7763e-03, -9.5521e-03,
        -4.7332e-03, -3.0550e-03, -6.8214e-03,  5.5360e-03, -4.3539e-03,
        -5.9855e-03,  4.9125e-03, -1.3109e-02, -7.4055e-03,  1.7637e-02,
        -1.8030e-02,  1.2674e-03, -3.3986e-02, -3.2282e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.7669e-02,  6.6803e-01, -1.4500e-02, -4.3781e-03,  2.0268e-02,
         7.0581e-03, -3.4384e-03,  4.6233e-03, -1.6611e-02, -5.6678e-04,
         6.3749e-03, -1.0506e-02, -1.0972e-02,  6.4314e-04,  3.3560e-04,
        -1.8738e-03, -1.7115e-06, -8.0974e-04,  1.0153e-03,  1.8932e-03,
        -3.9102e-03,  7.7379e-03, -3.1685e-03,  1.1969e-02,  1.5847e-02,
         5.3155e-03,  5.1960e-03,  5.2562e-03, -9.6116e-03, -9.3331e-03,
        -2.2234e-02, -7.4828e-04, -2.0154e-03,  4.0152e-03,  5.4542e-04,
        -1.2438e-03,  3.6304e-03, -5.9851e-03,  4.2941e-03,  5.3013e-04,
        -1.2390e-02,  6.3744e-03,  9.2216e-05,  1.3853e-02,  1.6852e-03,
         1.4455e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.7559e-03,  6.6594e-01,  3.4411e-02,  2.6054e-02, -2.4373e-02,
         1.6795e-02, -1.1821e-02, -1.2301e-02, -6.7563e-03, -1.8557e-02,
         1.9942e-02,  1.0003e-02,  4.5119e-02, -1.2145e-04,  1.2672e-03,
         1.1441e-02, -6.4163e-03, -2.0703e-02,  5.5638e-04,  8.7780e-03,
         1.3613e-02, -1.8156e-02, -1.7120e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8925e-02, -8.2509e-01,  4.5216e-02, -2.1217e-02, -4.7306e-04,
         1.3188e-02,  1.2539e-02,  2.2327e-02,  2.8764e-03, -1.0586e-03,
         1.7090e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1813,  0.5748, -0.0224, -0.0429,  0.0042, -0.0259,  0.0042,  0.0225,
         0.0211, -0.0023,  0.0140, -0.0037, -0.0391, -0.0417,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0354,  0.6994,  0.0646, -0.0633, -0.0164, -0.0098,  0.0157,  0.0104,
         0.0211,  0.0382,  0.0085,  0.0170,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3470e-02, -7.3276e-01, -2.9510e-02, -1.5438e-02, -1.5956e-03,
         1.9238e-02, -3.5821e-03, -1.4417e-03,  8.5930e-03, -7.6718e-03,
        -4.6086e-03,  1.4960e-03, -4.8692e-03, -1.1788e-03,  1.4334e-03,
         1.2360e-03,  3.2908e-03,  4.8809e-03,  6.3174e-04,  1.4212e-02,
         3.6056e-05, -1.6433e-03, -3.4481e-04,  9.9098e-03,  4.4406e-04,
         2.8731e-03, -1.4019e-02,  1.7972e-03, -6.4625e-03,  7.5164e-03,
        -4.5553e-03, -1.1010e-02, -3.7162e-04, -2.4085e-04, -7.2192e-03,
        -3.9490e-03, -4.8930e-03,  5.4506e-03,  4.3286e-03,  4.9994e-04,
        -1.8929e-03, -2.5093e-03, -1.5907e-03, -1.1272e-03, -3.8462e-03,
        -2.9688e-04, -3.0709e-03,  1.3130e-03,  2.3519e-03,  4.3784e-03,
        -2.2948e-03, -1.3337e-03,  2.3811e-04, -4.0245e-03, -1.0311e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-0.0775,  0.6920,  0.0177, -0.0014,  0.0033,  0.0022, -0.0074, -0.0081,
        -0.0154,  0.0084, -0.0156,  0.0253,  0.0155,  0.0037,  0.0195, -0.0102,
        -0.0038,  0.0176,  0.0031,  0.0121, -0.0045,  0.0272, -0.0084,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9164e-02, -7.0739e-01, -6.3944e-02, -2.8151e-02, -2.0902e-02,
        -8.2717e-03, -1.0896e-02, -1.5911e-02,  1.1020e-02, -1.7689e-02,
         1.9141e-06, -4.7211e-03,  2.2468e-02,  1.1766e-02,  1.1100e-02,
         8.2013e-03,  3.1896e-03, -2.5210e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2561e-02,  6.3859e-01,  4.6056e-02, -6.5231e-03,  6.4755e-04,
         5.2927e-02,  6.3764e-03,  1.1745e-02,  2.1203e-02,  3.8563e-03,
        -4.6078e-03,  3.1887e-03,  9.7725e-02,  3.5938e-04, -6.0024e-03,
         6.3965e-03,  1.3396e-02,  3.7840e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0575,  0.6870,  0.0263,  0.0010,  0.0186,  0.0442, -0.0054,  0.0168,
         0.0245,  0.0057,  0.0067,  0.0260,  0.0226, -0.0143, -0.0435,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8364e-02, -6.2874e-01, -7.1385e-03, -2.6102e-02, -1.0054e-02,
        -7.9944e-03, -1.1522e-02, -1.7696e-02, -2.1059e-02, -5.6078e-03,
        -1.4036e-02, -7.8596e-03, -1.6853e-02, -4.5343e-03, -6.6595e-03,
         1.3103e-04,  2.5026e-03, -4.7697e-02, -1.8839e-03, -1.1021e-02,
         3.4481e-03,  4.1069e-03,  4.1008e-03, -6.0768e-03, -3.9746e-03,
        -1.1474e-02, -3.0636e-02, -7.5976e-03, -2.4064e-03, -1.8937e-02,
         1.9484e-02,  1.0307e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7496e-02,  7.0152e-01,  2.4987e-02,  1.9861e-02, -1.2012e-02,
         5.0670e-04,  2.5471e-03, -8.0637e-04, -2.4145e-03, -7.9898e-03,
        -1.4220e-02,  7.8254e-04,  3.6683e-03, -3.9046e-03, -2.4216e-03,
        -4.3749e-03, -6.1250e-04, -1.3154e-02, -1.5741e-02, -1.8610e-02,
        -7.5310e-03, -1.5522e-02, -1.7700e-02, -3.5360e-03, -9.9135e-03,
         1.1606e-03, -7.7553e-03,  1.0769e-02, -4.1395e-03, -2.7175e-03,
         1.3859e-02, -6.1909e-03, -1.1572e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8605e-02,  6.4010e-01,  5.4806e-02,  4.5378e-02,  2.4359e-02,
        -1.1427e-02,  4.1834e-03, -1.6022e-02,  2.0494e-03, -2.6497e-03,
         4.6130e-04, -1.1173e-02,  1.9245e-02,  4.3905e-04,  2.7682e-03,
        -3.8161e-03,  7.6282e-03, -4.1773e-03, -1.4455e-03, -2.6856e-04,
         4.5617e-04, -5.6087e-03, -1.6409e-03,  3.1982e-04,  4.8963e-04,
        -4.6330e-03,  4.5379e-04,  5.8108e-03,  4.9605e-03, -3.3695e-03,
         1.1614e-03,  7.7054e-04,  1.8815e-03, -8.8954e-03, -5.6934e-04,
        -1.6135e-04, -2.9864e-04, -1.9873e-03,  1.9696e-03,  4.3349e-03,
        -1.0012e-03,  1.8825e-03, -4.1822e-04,  9.6736e-04,  6.1267e-03,
         3.4194e-03,  1.7420e-04, -1.8299e-03,  3.5153e-03,  8.8593e-04,
         1.0354e-03,  6.1308e-03,  1.4658e-03, -1.9682e-03,  8.7856e-03,
         4.8980e-03,  8.0554e-03, -1.6668e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.0987e-02, -6.5206e-01,  1.1562e-02, -1.2281e-02, -6.7577e-03,
        -6.5262e-03, -3.9465e-03,  3.1192e-04,  1.2842e-02, -2.7831e-02,
         4.3198e-03, -2.1588e-02,  6.5071e-03, -6.6025e-03,  7.8930e-04,
         5.8027e-03, -1.5586e-02,  9.8050e-03,  6.2370e-03, -1.0592e-02,
         2.7292e-03,  5.4178e-04,  1.8116e-02, -1.2434e-02, -2.5652e-02,
         7.7590e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5919e-02,  7.5714e-01,  2.1840e-02, -2.2012e-02,  1.5782e-03,
         1.5900e-02, -3.5616e-03,  7.6195e-03, -1.1481e-02, -5.3732e-03,
         1.5952e-03, -2.7580e-03,  1.5793e-02,  1.4162e-02, -6.7723e-04,
        -1.3367e-03,  3.4500e-04,  1.1928e-02,  3.3025e-04, -4.6501e-03,
         1.9287e-03, -3.3839e-04, -3.2397e-02, -1.9082e-02, -2.0255e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5929e-02,  6.0798e-01, -2.6395e-02, -3.9419e-02, -2.2192e-02,
         1.0277e-02, -2.7619e-03, -1.4663e-02, -1.1014e-02, -2.9673e-02,
        -2.2105e-02, -1.6275e-02, -7.2882e-04,  1.9498e-02, -8.9260e-04,
         1.4505e-02, -1.2321e-02, -5.0256e-03,  1.7811e-03,  5.7716e-03,
         2.1062e-03, -3.8796e-03, -1.3711e-04,  7.1344e-03, -4.7517e-03,
         1.4980e-03,  1.9685e-03,  6.4389e-03,  3.7915e-03,  2.0228e-03,
         4.7441e-03, -5.6160e-03,  1.2194e-03,  4.3050e-03, -2.6179e-03,
         1.2002e-03,  5.1481e-03,  1.5565e-04,  1.0662e-03,  9.1364e-03,
         5.1782e-03,  1.3389e-03,  2.3264e-03, -2.8915e-03, -2.0123e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.8446e-02,  6.2370e-01, -4.9516e-02,  1.1487e-02, -1.0053e-02,
         5.1811e-04, -7.3612e-03,  8.9553e-03,  1.9167e-02, -5.2232e-03,
         8.7485e-03,  7.5569e-03, -6.8487e-03,  3.2624e-03, -1.1089e-02,
        -1.2126e-02, -3.9093e-02, -1.1307e-02,  1.4724e-02, -1.5308e-03,
        -1.1947e-02, -2.2221e-02, -5.5117e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0384,  0.6684, -0.0217, -0.0284,  0.0075,  0.0211,  0.0269,  0.0315,
         0.0180,  0.0175, -0.0695,  0.0511,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-0.0817,  0.6683,  0.0184, -0.0197, -0.0177,  0.0033,  0.0038, -0.0163,
         0.0149, -0.0190,  0.0191, -0.0058, -0.0068,  0.0239,  0.0047,  0.0117,
         0.0046, -0.0084,  0.0025,  0.0146,  0.0039, -0.0039,  0.0270,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2998e-02, -6.7700e-01, -2.3320e-02,  9.5748e-03, -1.7968e-02,
        -2.6523e-03,  1.4246e-02, -1.0186e-02, -1.2641e-02,  1.6158e-02,
        -4.8366e-03, -1.8613e-03, -1.1138e-03, -9.9243e-03,  4.9242e-03,
         4.6233e-03,  7.5731e-03,  1.2252e-02,  2.2021e-03,  1.2474e-03,
         4.2186e-03, -6.1468e-03,  1.3406e-02,  1.0225e-02, -1.0691e-05,
         4.1538e-03,  8.8921e-03,  3.8201e-03,  1.3713e-02,  4.5960e-04,
        -9.7400e-03, -3.2938e-02, -3.4977e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0655,  0.7299,  0.0311, -0.0087,  0.0044,  0.0043, -0.0281, -0.0042,
         0.0055, -0.0134,  0.0060,  0.0140,  0.0059, -0.0221,  0.0014, -0.0120,
        -0.0026,  0.0249, -0.0162,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2364e-02, -7.8845e-01,  6.2544e-04, -1.1728e-02, -2.7246e-04,
        -2.8116e-02, -3.7632e-02, -1.3645e-02, -1.0127e-02,  2.0122e-02,
         9.1625e-03,  1.7344e-02, -2.3066e-02, -2.7340e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9128e-02,  6.7610e-01,  1.7387e-02, -2.1502e-02, -7.0873e-02,
         2.1909e-02, -5.7418e-03, -1.5500e-02,  1.1536e-02,  1.1922e-02,
        -1.9716e-04, -8.8969e-03,  1.4556e-02,  5.1753e-03, -1.3006e-02,
        -3.9656e-03,  7.2246e-04, -2.5021e-03, -1.7186e-04,  8.4098e-03,
        -8.2729e-03,  4.2525e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0375e-02,  7.1930e-01,  1.4853e-02, -4.0618e-02,  7.2477e-03,
         1.3962e-02, -1.8809e-02, -1.3111e-02, -1.0364e-02, -2.4759e-03,
         1.7665e-03, -4.3931e-03, -1.6556e-02,  5.4191e-03, -3.1084e-04,
         1.1743e-02,  1.9471e-02, -3.0005e-03,  3.4141e-03,  9.0391e-03,
         3.4831e-03, -2.4680e-03,  2.2127e-03, -1.1039e-02, -5.4151e-03,
        -1.6414e-03, -1.3637e-03,  1.2998e-02,  1.1833e-02,  1.3141e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0425,  0.6374, -0.0384,  0.0056,  0.0458,  0.0035,  0.0550, -0.0414,
        -0.0218,  0.0082,  0.0236, -0.0016,  0.0049,  0.0097, -0.0097,  0.0284,
        -0.0224,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0432,  0.7482,  0.0591,  0.0274, -0.0217,  0.0110,  0.0079,  0.0294,
         0.0177, -0.0089, -0.0254,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0393,  0.7269,  0.0190, -0.0216,  0.0099, -0.0356,  0.0244,  0.0072,
         0.0107,  0.0043,  0.0043, -0.0317,  0.0196,  0.0209, -0.0119, -0.0057,
        -0.0070,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3190e-02,  6.6107e-01, -1.5577e-03,  2.9049e-02,  1.1328e-02,
        -5.1765e-03, -4.7021e-03, -3.0785e-03, -1.5074e-02, -4.6650e-04,
         1.5461e-02, -5.7882e-03,  3.4032e-03, -1.9158e-02,  1.8412e-02,
         1.3082e-02, -5.5290e-04, -1.0127e-02,  6.8843e-03, -4.0495e-03,
         3.2626e-03,  1.3806e-02, -1.1286e-03,  5.2506e-03, -3.9822e-02,
        -5.5121e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.4859e-02,  6.1903e-01, -2.1025e-02, -7.7993e-02, -2.8140e-02,
        -7.0327e-03, -2.1301e-02, -4.6097e-03, -7.7847e-04,  1.0629e-02,
         6.4484e-03, -2.6202e-03, -2.8220e-03, -1.5775e-02,  4.3662e-04,
         1.1773e-02, -4.4516e-03,  3.6465e-03, -2.7671e-03, -1.7598e-03,
        -1.6302e-03, -5.2726e-03, -2.0517e-03, -1.6161e-02, -2.2336e-02,
        -1.6646e-03,  2.3602e-03, -7.4690e-03,  1.9937e-02,  3.2209e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0125, -0.6144, -0.0179,  0.0773,  0.0315,  0.0117,  0.0011, -0.0166,
         0.0031,  0.0320,  0.0034,  0.0274, -0.0023, -0.0015, -0.0063,  0.0119,
        -0.0024, -0.0840, -0.0425,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 0.0045,  0.6631,  0.0714,  0.0285,  0.0080,  0.0198,  0.0391,  0.0007,
         0.0105,  0.0009, -0.0044, -0.0037, -0.0022,  0.0041, -0.0209, -0.0095,
         0.0097,  0.0211,  0.0124,  0.0124, -0.0059, -0.0060,  0.0008, -0.0077,
        -0.0018, -0.0263, -0.0045,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5021e-02, -7.4100e-01, -8.9848e-03,  5.8127e-02, -2.5888e-02,
         5.3108e-03, -6.0444e-03, -5.0702e-03, -9.0441e-06,  1.4501e-02,
         8.7192e-03, -2.6520e-02, -2.1464e-02,  5.3339e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.0135e-02, -7.3889e-01, -2.0286e-02, -8.4628e-03,  1.0440e-02,
         1.9267e-03,  8.2427e-03, -2.1518e-02, -2.3747e-03,  3.9058e-03,
        -7.2473e-04, -2.9067e-02,  1.8240e-02, -1.4008e-02,  5.7819e-03,
        -2.6316e-03,  2.0070e-03,  7.4331e-03, -1.5595e-03, -1.5709e-02,
         5.9512e-03,  1.3066e-02,  2.2320e-03,  7.8356e-03,  1.2749e-02,
         4.8250e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.4769e-03,  7.3992e-01, -1.6435e-02, -7.8263e-03,  2.1562e-02,
        -3.6007e-02,  3.0700e-02, -4.9393e-04, -1.5534e-02, -9.5357e-03,
        -2.2401e-02, -1.9928e-02,  5.1740e-02, -2.2437e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8523e-03,  7.2888e-01,  3.7449e-02, -5.0630e-03, -6.0431e-03,
        -5.4600e-03, -1.0715e-02,  2.2747e-03,  7.8991e-03,  3.4826e-03,
         8.8042e-03,  2.6451e-04, -6.4222e-04, -6.7166e-02,  1.6387e-03,
         2.5338e-03, -1.5540e-03, -9.1417e-03, -8.0945e-04,  3.6551e-03,
        -5.8625e-03, -1.5290e-03, -2.1246e-03,  5.3681e-03,  1.1103e-02,
         2.6407e-03, -2.3075e-03, -8.3774e-03, -1.4098e-02,  1.1324e-02,
         6.3509e-03,  2.1586e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.7557e-03,  7.4364e-01,  7.3948e-03, -2.0527e-02,  2.5938e-03,
         7.7779e-03, -3.1228e-04, -2.7731e-02,  5.5996e-03, -9.9854e-03,
        -1.1270e-03, -6.0578e-03, -2.7963e-02,  4.3714e-03, -1.0061e-03,
        -1.0272e-02, -5.7617e-03, -1.5604e-02, -1.5246e-02, -9.5108e-03,
        -2.3149e-03, -1.4288e-02, -2.8180e-02,  1.0420e-02,  1.5560e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0516,  0.7535,  0.0343,  0.0347, -0.0307, -0.0951,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4533e-01,  6.0773e-01,  1.8701e-02, -2.8937e-02, -8.7932e-03,
        -4.6929e-02, -2.0361e-03, -2.2953e-03, -2.4451e-04, -2.1776e-03,
        -1.1861e-02, -1.0472e-02, -8.8949e-03,  2.3126e-03, -6.9531e-03,
        -5.8844e-03,  9.5842e-03,  2.9905e-03, -3.4005e-03, -2.0517e-03,
        -3.8355e-04, -3.8349e-03, -1.0772e-02,  7.2511e-04, -3.6214e-04,
        -6.0086e-04, -1.1598e-02, -1.0312e-03,  6.0095e-03, -1.2904e-03,
         2.9455e-03, -4.4703e-03,  5.5120e-03, -3.4984e-03, -3.3175e-03,
        -1.5306e-02,  7.6910e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.9481e-03, -7.6018e-01, -2.1561e-02, -3.2747e-02, -1.1521e-02,
        -7.0666e-03, -3.8496e-03,  5.9152e-03,  1.3895e-03,  3.0558e-04,
         4.2214e-03,  5.1633e-03,  5.9073e-03,  1.1715e-02, -2.1712e-03,
         1.5162e-02,  5.2662e-03, -5.6930e-03, -5.7077e-04,  4.1452e-04,
         1.3188e-02, -2.8907e-03,  4.3207e-03, -6.5184e-03, -7.9385e-05,
         1.9746e-03,  1.1040e-03, -6.9019e-03,  7.3532e-03,  5.5459e-03,
         6.3927e-03, -2.2645e-03,  1.6332e-03, -1.1041e-02,  1.9021e-02,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9103e-03,  5.9830e-01, -1.2058e-01, -4.8573e-02,  2.0279e-02,
        -3.5251e-04,  3.6391e-04, -1.9149e-02, -1.6846e-02,  3.5872e-04,
        -1.8279e-02, -4.0845e-02, -1.3832e-02,  6.1499e-02, -3.6829e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7366e-02, -5.9437e-01, -2.5264e-02,  2.8421e-02,  4.7505e-03,
         1.5061e-02,  7.5423e-03, -4.4367e-03,  1.2563e-02, -2.3337e-02,
        -7.3325e-03,  1.3401e-02, -1.7507e-03,  3.8548e-03, -3.3862e-03,
        -1.0899e-03,  3.8632e-03,  3.5358e-02, -8.5306e-05,  8.7667e-03,
         1.3602e-02,  7.5324e-03, -1.5041e-04, -4.2790e-03,  1.1363e-03,
         2.7932e-03,  7.0077e-03,  4.0710e-03,  1.6114e-02, -9.0515e-05,
         3.4421e-03,  6.6814e-02, -6.0969e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0108,  0.7819, -0.0014,  0.0102,  0.0010,  0.0059, -0.0056,  0.0138,
         0.0237,  0.0017, -0.0030, -0.0141,  0.0093,  0.0034,  0.0024, -0.0073,
         0.0083,  0.0017, -0.0108, -0.0093, -0.0131, -0.0098, -0.0050,  0.0038,
         0.0027, -0.0020,  0.0019,  0.0130,  0.0229,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 3.4800e-02,  7.3303e-01, -4.4529e-02, -3.4718e-02,  2.1314e-02,
        -4.3181e-03, -8.1540e-03,  2.0454e-05,  1.8992e-03, -3.3751e-02,
        -1.6372e-02,  4.6112e-03, -4.3765e-03,  3.9711e-02, -1.8398e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.4509e-02,  6.8689e-01,  6.6768e-03,  3.5251e-03, -1.2680e-03,
         5.7029e-03, -1.2905e-02, -1.2774e-02,  1.2200e-02, -8.8582e-04,
        -1.6201e-03,  6.0887e-04, -2.8074e-03, -9.7661e-03,  5.5635e-03,
        -2.6284e-03, -4.7794e-03, -1.3319e-02,  2.3193e-03, -1.6348e-03,
         8.5616e-03, -2.3515e-04,  5.3391e-03,  8.2753e-03,  4.9928e-03,
         2.9288e-03,  1.5658e-03,  1.5390e-02,  8.7393e-03,  5.3667e-04,
        -5.1164e-03, -2.8601e-03, -5.3769e-04, -5.3551e-04,  5.2942e-03,
         9.1102e-03,  1.3165e-03, -8.8659e-04, -1.0402e-03, -1.9393e-03,
         4.4054e-03,  2.8530e-03,  9.3243e-04,  1.4647e-03,  2.0467e-03,
        -1.1994e-02, -5.1931e-03,  3.0761e-04,  1.3218e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1083e-02,  7.4051e-01, -7.0593e-03,  2.4363e-03,  9.7117e-03,
         6.7844e-03,  8.2920e-03, -1.0201e-02, -2.4638e-03, -1.6032e-02,
         7.2766e-03, -3.0911e-03, -5.7556e-03,  7.1573e-03, -1.4137e-03,
         5.1094e-03,  3.9310e-03, -3.9317e-03,  2.1581e-04, -1.9139e-02,
        -2.1859e-02,  6.5134e-03, -3.4099e-03,  1.0331e-03, -5.6955e-03,
        -5.3566e-03, -1.5185e-02,  6.0715e-03, -9.1733e-04,  4.0785e-03,
         1.0224e-02,  3.1903e-03, -1.8657e-03, -4.2873e-04, -2.2576e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4566e-02,  8.1079e-01,  4.9680e-02, -1.2695e-02, -7.4963e-03,
         5.5164e-03,  9.5065e-03,  3.0590e-03, -1.7481e-02, -3.1779e-03,
        -9.3827e-03, -7.9327e-03,  1.1202e-03, -3.1481e-04, -1.6300e-03,
        -7.5772e-03, -1.3361e-02, -6.7233e-03,  1.7985e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0714,  0.7759,  0.0184, -0.0061, -0.0022,  0.0120,  0.0105, -0.0013,
        -0.0136, -0.0030, -0.0115, -0.0402, -0.0040, -0.0238,  0.0059,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0745, -0.6557,  0.0112, -0.0228,  0.0215,  0.0156,  0.0304,  0.0081,
        -0.0250,  0.0093,  0.0113,  0.0047,  0.0211,  0.0026,  0.0029,  0.0106,
         0.0080,  0.0084,  0.0125, -0.0084,  0.0233,  0.0122,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7606e-02,  6.7071e-01, -1.5799e-02,  2.8020e-03,  1.7528e-02,
         5.7465e-03, -3.5380e-02, -8.6342e-04, -4.3103e-03,  4.3435e-03,
         7.3089e-04,  1.0884e-02,  3.8068e-02, -7.4145e-03,  2.5770e-03,
         3.7894e-03,  1.5625e-02, -5.0340e-03,  1.0425e-02,  8.1756e-03,
         8.4340e-03,  8.9767e-03,  4.9859e-04, -8.6231e-04,  9.4326e-03,
         6.4963e-03, -2.4629e-03, -7.5476e-03,  2.3729e-02,  3.3751e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1321,  0.6023,  0.0123, -0.0684,  0.0666, -0.0114, -0.0083,  0.0152,
        -0.0442, -0.0374,  0.0018,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0322,  0.7367, -0.0048, -0.0176,  0.0195,  0.0065,  0.0054, -0.0129,
        -0.0131, -0.0033, -0.0118,  0.0191,  0.0021,  0.0008, -0.0041,  0.0031,
        -0.0034, -0.0092, -0.0022, -0.0069, -0.0738, -0.0117,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2286e-04, -7.5680e-01, -2.4382e-02, -1.6313e-02, -1.9394e-02,
         9.5632e-03, -1.0735e-04, -3.2974e-03,  3.4135e-03,  1.3866e-02,
         1.4525e-02,  9.5928e-03,  1.9295e-03,  5.4576e-03,  8.1806e-03,
         4.7546e-03,  2.8544e-03,  6.4805e-03, -1.1732e-02,  6.7091e-03,
        -2.0278e-03, -1.4403e-03, -6.4106e-03,  2.8571e-03,  5.1751e-03,
        -1.1906e-03, -2.0588e-03, -1.1506e-03,  3.6905e-03,  8.0718e-03,
        -2.9711e-03,  1.1752e-02, -3.1530e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0042,  0.7489,  0.0578,  0.0020,  0.0058,  0.0250, -0.0046,  0.0055,
        -0.0071,  0.0021, -0.0008,  0.0067,  0.0056, -0.0029, -0.0076,  0.0161,
         0.0042, -0.0030, -0.0011, -0.0093,  0.0017, -0.0171,  0.0069, -0.0045,
         0.0017,  0.0084, -0.0082, -0.0315,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0261, -0.6916, -0.0355,  0.0290,  0.0304,  0.0067, -0.0117, -0.0143,
        -0.0035,  0.0143, -0.0010, -0.0025,  0.0018, -0.0098, -0.0039, -0.0027,
         0.0054,  0.0136,  0.0069, -0.0139,  0.0067, -0.0016,  0.0092,  0.0470,
         0.0109,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 1.7877e-02,  7.1270e-01,  3.5561e-02,  1.3012e-02,  7.8497e-03,
         1.2096e-02, -5.0923e-03, -1.0276e-02,  5.0489e-03,  4.9021e-03,
         6.9694e-03,  1.1283e-02,  2.0067e-03,  4.2457e-03, -8.1552e-03,
        -7.8421e-03, -2.1764e-03, -1.6563e-03, -4.8504e-03,  2.7525e-04,
         4.1566e-03, -1.0439e-02, -1.3213e-03,  2.3625e-03, -1.3150e-02,
        -1.4706e-02, -4.8638e-03, -4.7674e-03, -7.5266e-03, -3.8859e-03,
         1.3281e-03, -7.2973e-03, -4.7029e-03, -6.1611e-03,  3.2322e-03,
         1.7014e-03, -5.4217e-04, -7.6686e-03,  2.3729e-03, -2.3357e-02,
        -5.7772e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.2607e-03, -7.7688e-01, -5.1513e-02,  2.6444e-02, -4.6272e-02,
        -4.2251e-03,  7.8111e-03, -7.1644e-04, -5.4105e-04,  5.7651e-03,
         2.6843e-03,  1.3722e-02, -3.5236e-03,  3.7590e-03, -1.4188e-02,
         3.2011e-03,  1.1294e-04, -2.7159e-03, -1.6212e-04,  6.7758e-03,
         2.3645e-03,  5.8109e-03,  9.4797e-03,  5.2600e-04, -5.5447e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0443, -0.6593, -0.0472,  0.0564, -0.0015,  0.0065,  0.0372, -0.0065,
        -0.0063,  0.0024,  0.0032,  0.0188,  0.0141, -0.0064,  0.0166, -0.0048,
        -0.0061, -0.0113, -0.0085,  0.0216,  0.0209,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0453, -0.7774, -0.0314, -0.0121,  0.0294, -0.0107,  0.0033, -0.0125,
        -0.0140, -0.0095,  0.0113, -0.0017,  0.0082,  0.0024,  0.0043,  0.0059,
         0.0132, -0.0077,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3108e-02,  6.1563e-01,  2.2265e-02,  4.5435e-02,  9.3512e-06,
        -2.1601e-02,  3.6964e-03, -2.3859e-02, -5.3070e-03,  5.8842e-03,
        -1.0572e-02,  1.4812e-03, -3.3300e-03,  5.0096e-03,  7.5733e-03,
         2.2639e-02, -2.4526e-02,  2.5888e-03, -2.1869e-03,  1.5688e-03,
         2.6552e-03, -3.0780e-03, -2.8139e-04, -4.9010e-03, -5.1787e-03,
         2.1566e-03, -7.0255e-03, -2.3319e-03, -4.9127e-04, -1.9920e-03,
         4.7234e-04, -3.6873e-03,  1.6826e-04, -2.5125e-03, -1.8104e-02,
         3.8224e-04, -1.0501e-03,  6.5265e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4536e-02, -7.2551e-01, -3.0584e-02, -5.2584e-02, -1.0629e-02,
        -4.4102e-03,  9.7631e-03,  1.2278e-02, -1.0238e-03, -1.6135e-02,
         1.4834e-02,  8.4374e-03, -4.1347e-03, -2.8880e-02, -1.8008e-03,
        -1.6564e-02, -2.6560e-04,  2.2080e-02, -3.5674e-03, -1.9792e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0491, -0.6846, -0.0484, -0.0096, -0.0440, -0.0340, -0.0125, -0.0148,
        -0.0056,  0.0109, -0.0232,  0.0155,  0.0306,  0.0173,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0039e-01,  5.4010e-01, -4.1625e-02, -3.6976e-02,  1.8729e-02,
         9.4504e-03,  3.3872e-02, -1.5447e-03, -1.2720e-03, -1.0163e-03,
         1.3351e-02, -1.2385e-02,  2.0971e-03,  1.5793e-04,  9.0962e-03,
        -6.4590e-03,  8.3696e-03,  2.4254e-02,  7.5315e-03,  7.6834e-03,
         1.3735e-02,  1.6778e-02, -1.8840e-03,  1.6738e-02,  3.1187e-03,
        -6.5987e-03,  3.1561e-04,  8.8847e-04, -5.6909e-03,  1.5949e-02,
         1.9777e-02,  2.2172e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0454, -0.7011,  0.0076, -0.0382, -0.0141,  0.0127, -0.0301,  0.0159,
         0.0031, -0.0010,  0.0074,  0.0246,  0.0010, -0.0058,  0.0069,  0.0037,
         0.0035, -0.0011,  0.0298,  0.0469,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2521e-01,  4.8055e-01,  2.6757e-02,  1.4478e-03,  7.4517e-03,
        -2.6711e-02,  9.6700e-05, -5.1568e-03, -8.9227e-03,  7.0918e-03,
        -6.6621e-04,  5.1019e-03, -1.9689e-03,  2.9004e-03,  1.1363e-03,
        -2.1638e-04,  7.9886e-03, -9.9422e-04, -1.8052e-03, -2.7945e-04,
         2.0575e-03, -2.0897e-03,  6.4293e-03, -7.0062e-03, -7.9887e-03,
         3.1767e-05, -1.6788e-02, -1.8700e-02, -1.4591e-02,  3.6127e-03,
         1.5238e-02,  3.5997e-03,  1.4592e-02,  2.4136e-03,  2.9844e-03,
         3.3718e-02, -7.1480e-03,  2.1120e-03,  4.1271e-03, -1.0475e-03,
        -7.7446e-03, -1.3477e-04,  1.1987e-02, -3.1507e-03,  1.0967e-02,
         4.7909e-02,  3.9381e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0808,  0.6493,  0.0120,  0.0149,  0.0338,  0.0132,  0.0125, -0.0118,
         0.0979,  0.0049,  0.0059,  0.0294, -0.0151, -0.0044,  0.0141,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.7968e-02, -6.7743e-01, -1.6533e-02,  1.9365e-03, -2.1231e-03,
        -4.4152e-03, -1.5952e-03, -2.5139e-02, -7.4675e-03,  3.8414e-03,
         1.9421e-03, -6.7811e-03,  1.8421e-03, -8.4180e-03,  1.4887e-02,
         5.9650e-03,  1.1026e-02, -8.9315e-04, -1.6963e-02,  8.5679e-03,
        -4.2618e-03,  9.3433e-03,  5.1636e-03, -6.2276e-04,  3.4047e-03,
         6.4051e-03, -1.2313e-02, -4.6138e-03, -7.6137e-03, -2.9934e-03,
        -3.9348e-03,  2.4712e-03,  2.8194e-03,  2.7610e-03, -5.2659e-03,
        -4.1175e-03,  2.7359e-03,  1.0128e-03,  4.2172e-03, -9.9861e-05,
        -2.4634e-03,  1.1121e-03,  3.1633e-03,  4.1143e-04, -9.9610e-04,
         2.2154e-03,  5.2780e-03, -2.4562e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-3.4044e-02,  6.9820e-01,  4.9252e-02,  2.5364e-03,  2.0441e-02,
        -4.1275e-03,  6.8794e-03,  2.6982e-03,  8.3237e-04, -3.6284e-04,
         4.1111e-03,  1.4986e-03, -9.9104e-03, -4.3028e-04, -3.1171e-03,
         5.4860e-03, -3.0681e-02,  2.5329e-03, -1.2803e-02, -8.7311e-03,
        -6.8950e-03, -5.1202e-03,  6.3575e-03,  9.6132e-03,  3.7640e-02,
        -3.5700e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8946e-02,  6.4195e-01, -2.9745e-02, -5.1889e-03,  5.0072e-03,
         1.5229e-02, -3.1932e-02,  1.2019e-04, -3.0597e-02, -1.3890e-02,
         2.7761e-02,  9.0323e-03, -2.2326e-03, -4.9455e-03,  1.5992e-02,
         1.2465e-02, -1.2583e-02, -5.4627e-03,  4.2873e-02, -5.2887e-02,
        -1.1162e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0248,  0.6513,  0.0234,  0.0287, -0.0018,  0.0146, -0.0024,  0.0289,
        -0.0049, -0.0063, -0.0024, -0.0052,  0.0024, -0.0062, -0.0230, -0.0052,
         0.0008, -0.0110, -0.0069, -0.0109, -0.0123, -0.0031, -0.0058, -0.0042,
         0.0025,  0.0007, -0.0070,  0.0032,  0.0012, -0.0008, -0.0212, -0.0070,
         0.0127, -0.0169,  0.0009, -0.0392,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1968e-01,  7.0106e-01, -2.7206e-02, -5.7682e-03, -6.6489e-03,
         6.0846e-04, -2.9915e-02,  1.0068e-02,  3.0358e-04,  7.2514e-04,
        -1.7758e-02, -2.2822e-03, -9.3765e-03, -9.6614e-03,  4.5537e-03,
         3.0992e-03, -5.2010e-03, -1.4875e-02,  1.5947e-02, -1.5261e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0314, -0.7589,  0.0044, -0.0118, -0.0136, -0.0248, -0.0036,  0.0029,
         0.0151, -0.0099, -0.0030,  0.0043,  0.0054, -0.0144, -0.0126, -0.0037,
         0.0117,  0.0028,  0.0061, -0.0578, -0.0018,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0678,  0.7026,  0.0246,  0.0319,  0.0090,  0.0214,  0.0255,  0.0107,
         0.0265, -0.0138, -0.0026,  0.0491,  0.0144,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6869e-02,  5.8183e-01,  3.3485e-02,  1.3551e-02, -1.0428e-02,
         3.5018e-03, -3.6354e-02, -9.5069e-03, -1.0880e-02, -1.1589e-02,
        -2.2836e-02,  4.9558e-04, -6.2430e-03,  1.7945e-03,  1.8227e-04,
         4.3589e-03,  3.7161e-02,  2.1533e-02, -8.3965e-04, -7.5612e-04,
         8.9907e-03,  7.6940e-03,  8.3457e-03,  1.1323e-03, -2.4051e-04,
         1.0235e-02, -2.8073e-03,  3.9674e-03,  7.0797e-04, -1.0676e-02,
        -1.8568e-03,  2.4396e-04,  1.2152e-02,  1.6163e-03,  4.9410e-03,
        -6.8906e-02, -3.1295e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0133e-01,  6.0095e-01, -1.5186e-02, -2.7483e-02, -3.7530e-02,
         6.4961e-03, -5.9735e-03,  3.0528e-02, -1.4345e-02, -8.5877e-03,
        -4.1920e-03, -1.7201e-03, -6.2098e-03, -3.9274e-03, -8.5516e-03,
        -3.7465e-03,  6.4123e-03,  1.2857e-04, -7.2974e-03, -2.1286e-02,
        -6.5733e-04, -2.8387e-03,  3.1292e-02,  5.3339e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9486e-02,  6.6953e-01, -3.2256e-03, -1.2053e-02, -9.5848e-03,
         5.7395e-03, -1.2374e-02,  1.5028e-02,  2.5029e-03,  1.3979e-04,
        -2.3031e-02, -3.4378e-05,  7.1668e-03,  2.0768e-03, -5.5109e-03,
        -3.4458e-02, -2.8733e-03, -1.2301e-03, -8.1247e-03,  5.5296e-03,
         1.6079e-04, -8.7413e-03,  9.6313e-03, -3.4070e-03, -5.9767e-03,
        -9.1078e-03,  5.2160e-05,  2.8686e-03,  4.2997e-03, -4.5114e-03,
        -4.8815e-02, -5.2734e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2727e-02,  6.7519e-01,  2.2868e-02, -7.5560e-03,  2.8088e-02,
        -3.2300e-04, -1.3341e-02, -2.0682e-03,  1.7203e-02,  3.0663e-03,
         7.1084e-03,  1.6260e-03, -2.7926e-03,  4.1076e-03,  2.6132e-03,
         2.7286e-03, -2.3456e-03,  2.8587e-03,  1.4311e-02,  3.7675e-03,
         2.3808e-02, -5.3709e-02, -1.6955e-02, -1.5580e-03,  3.2160e-03,
        -3.0353e-03, -1.2884e-03,  4.8848e-03, -1.2514e-02,  4.1489e-02,
         1.0851e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.3818e-02, -5.6699e-01,  1.4034e-01, -1.3719e-02, -3.7712e-04,
        -3.0425e-02,  2.1343e-02, -7.5454e-05, -4.9896e-03,  1.0516e-02,
        -3.8726e-02,  2.5413e-04, -2.5963e-04, -7.7212e-03, -1.6084e-03,
         1.6841e-02,  7.8421e-03, -7.6571e-04,  1.1254e-02,  4.9557e-03,
         3.1892e-02, -7.3300e-03, -1.7965e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.1730e-02,  6.5390e-01, -2.1197e-02,  1.9730e-02, -2.2694e-02,
        -1.6200e-02, -1.0779e-02, -1.1244e-02, -3.8815e-03,  3.3740e-03,
        -5.0777e-03, -4.9518e-03, -9.2349e-03, -1.3761e-02, -6.2695e-04,
        -4.1861e-03, -2.7066e-02, -5.2685e-03, -7.3449e-03, -4.3653e-03,
         2.3325e-03,  3.5062e-03, -3.6761e-04,  6.5361e-03, -2.8571e-03,
         7.1308e-03, -3.7142e-02, -3.5095e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-1.6378e-02,  8.5486e-01, -2.8745e-04,  4.0589e-03,  1.3812e-02,
         7.7388e-03,  7.3066e-03,  3.4335e-03, -2.4375e-03, -1.3385e-02,
         1.2374e-02,  2.1749e-03, -3.3923e-04, -5.0630e-03, -7.1998e-03,
        -3.7057e-03, -3.0448e-03, -2.8521e-04, -2.3985e-03, -3.1293e-03,
         2.7952e-03, -7.7098e-04, -4.9438e-03, -6.5027e-03,  7.7635e-03,
         4.2010e-03, -2.6839e-03, -3.3607e-03, -3.5681e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5877e-02, -7.5367e-01, -9.0507e-03, -5.7710e-03, -4.8421e-02,
         9.9000e-03, -3.9583e-03,  3.7228e-03,  1.3909e-02,  1.9638e-03,
        -3.6782e-03,  9.3952e-05,  2.0184e-03, -2.8929e-03,  9.4508e-03,
        -1.2250e-03,  4.6827e-05, -4.8836e-03,  8.1317e-03, -1.8298e-04,
        -4.8856e-03, -4.9470e-02,  4.6102e-03, -2.2183e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.0736e-02,  5.2095e-01, -4.1149e-02, -2.7272e-02,  6.9947e-03,
        -2.3744e-02, -2.7270e-03, -2.4942e-02, -8.8969e-03, -6.1983e-03,
        -1.1496e-02,  3.9535e-03,  2.6612e-02,  7.7514e-03, -7.3388e-03,
        -2.7647e-03, -1.8599e-02,  1.1349e-02, -2.2075e-02, -5.4518e-04,
         1.0815e-02,  8.2655e-03,  4.5248e-03, -4.7074e-03,  1.0487e-02,
         7.9890e-03,  3.0790e-03,  3.5333e-03, -4.1465e-04,  3.4924e-03,
         8.5338e-03,  4.5787e-02,  2.5745e-02, -6.5314e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0474, -0.7232,  0.0394,  0.0163,  0.0146, -0.0095, -0.0080,  0.0017,
        -0.0160, -0.0166, -0.0027,  0.0043, -0.0038,  0.0128,  0.0018, -0.0042,
        -0.0110, -0.0025,  0.0119, -0.0188, -0.0200, -0.0135,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0087, -0.7127, -0.0323,  0.0322, -0.0130,  0.0266,  0.0128, -0.0034,
        -0.0049, -0.0053, -0.0363,  0.0009,  0.0025, -0.0011, -0.0040, -0.0084,
         0.0813,  0.0136,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0519e-03,  7.9198e-01,  1.2144e-02,  3.2205e-02, -4.5700e-03,
         1.2678e-02,  1.2084e-02, -2.8021e-02,  2.6997e-03,  1.1796e-02,
         4.6807e-03,  7.4883e-03, -1.0051e-02, -1.2750e-02, -1.2870e-02,
        -1.2601e-03, -1.2959e-03, -7.7013e-03,  4.4767e-04, -7.0699e-03,
        -1.3452e-02, -5.5891e-03, -4.1170e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.5515e-02, -7.2572e-01, -4.5578e-03, -2.5935e-02,  1.4824e-02,
         6.9632e-03,  1.3059e-02,  1.5994e-04,  1.6600e-02, -8.7264e-03,
        -2.7639e-02, -1.0533e-02, -2.8206e-03, -6.5099e-03, -6.0121e-03,
        -1.3314e-02, -3.1114e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7810e-02,  6.7621e-01, -2.4891e-02,  2.2583e-02,  1.7898e-02,
        -1.5865e-02,  3.0498e-02, -1.3959e-02,  1.0363e-02,  2.4186e-02,
         3.5476e-03, -4.5477e-04,  1.6672e-03,  3.0108e-03, -1.1013e-02,
         2.5669e-02, -2.0116e-02,  2.5837e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1695e-02, -6.1716e-01,  2.3455e-02,  4.3648e-03, -1.0557e-02,
        -1.8246e-02,  4.0710e-03, -2.2127e-02,  2.1536e-02, -3.3862e-02,
        -9.0638e-04,  2.1991e-03,  4.7107e-03,  1.9052e-02,  1.5348e-02,
         1.7076e-02,  1.2976e-02,  1.7734e-02, -2.5187e-02, -1.3294e-02,
        -1.4507e-02, -5.6336e-03, -1.2405e-02,  1.2505e-02,  5.3787e-04,
         3.8855e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0747,  0.6515,  0.0607, -0.0232, -0.0195,  0.0102, -0.0058, -0.0043,
        -0.0279, -0.0111,  0.0078,  0.0123, -0.0031, -0.0075,  0.0018, -0.0033,
        -0.0056,  0.0059,  0.0073,  0.0070,  0.0041, -0.0069,  0.0036,  0.0007,
         0.0018,  0.0013,  0.0036,  0.0044,  0.0044, -0.0016, -0.0171,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0575, -0.7603, -0.0013,  0.0109,  0.0172,  0.0092, -0.0138, -0.0047,
         0.0095, -0.0224,  0.0145,  0.0122,  0.0036,  0.0067,  0.0173, -0.0078,
         0.0098,  0.0216,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0459,  0.7313, -0.0463, -0.0367,  0.0198,  0.0539, -0.0337, -0.0324,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 8.6820e-02,  7.2067e-01, -1.6513e-02, -1.6815e-02, -2.8688e-03,
         6.5163e-03, -4.9019e-03,  8.6301e-04,  3.6947e-03,  2.6199e-03,
         1.9861e-03, -1.6001e-02,  2.6487e-04,  5.5356e-03,  3.0498e-03,
         4.9552e-03, -2.8281e-03, -4.2936e-03, -4.0819e-03, -9.3879e-03,
         1.9537e-02, -6.3325e-03, -5.0831e-03, -4.3804e-03, -3.0574e-03,
        -2.6974e-03, -7.3193e-03,  2.4527e-04, -1.2738e-02,  1.9264e-03,
        -2.0289e-03, -1.3222e-03,  1.7078e-03,  1.1789e-03, -2.7673e-03,
         3.4534e-03,  1.2100e-03, -4.9530e-03, -3.3959e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.6665e-02,  7.2909e-01, -6.0638e-03, -6.1178e-03, -5.3814e-03,
        -3.0668e-03,  2.2220e-03, -3.0362e-03, -2.6240e-03,  7.4015e-03,
         1.7493e-03,  4.7244e-04,  1.1513e-02,  9.9556e-04,  8.8786e-03,
         3.0974e-03, -9.3116e-03, -8.5165e-03,  8.4509e-03,  1.9924e-02,
         6.3063e-03, -4.6220e-03, -1.2019e-02,  3.9487e-03,  2.5622e-03,
        -2.5082e-02,  8.2890e-03, -2.2592e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7743e-01, -4.9242e-01,  1.1732e-02,  6.7278e-02, -1.0588e-02,
         7.6378e-03, -1.2959e-02, -3.4441e-02, -3.5066e-03, -1.6187e-02,
        -2.3086e-04, -7.4191e-02, -9.1405e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0620,  0.6785, -0.0259,  0.0418,  0.0080,  0.0302,  0.0041,  0.0216,
         0.0123,  0.0027, -0.0023, -0.0143, -0.0174,  0.0476, -0.0311,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.9303e-02,  6.0847e-01,  2.0408e-02,  7.1567e-03,  3.0592e-02,
        -2.8955e-02,  2.4717e-03, -1.1280e-02, -1.5783e-03,  1.9817e-03,
        -1.8922e-02,  4.2073e-03,  2.8391e-03,  1.2359e-02,  4.1634e-03,
         6.3919e-03, -5.3603e-03,  5.4722e-03, -6.2582e-03, -7.9488e-03,
        -2.9714e-03, -4.9967e-03, -3.7431e-04,  9.4682e-03, -3.9483e-03,
        -3.4911e-03, -3.1624e-03, -3.4536e-03, -1.2702e-03, -3.9331e-03,
        -9.3728e-03, -6.2233e-03, -1.1161e-02, -1.1745e-02,  9.1820e-03,
        -5.5407e-03, -2.0771e-03, -4.2895e-03, -1.4164e-03, -3.8562e-03,
         2.2631e-03, -4.1409e-03, -2.0338e-03, -9.1635e-04, -5.7536e-03,
        -3.8667e-04,  6.4524e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.6370e-02,  6.5887e-01,  4.3114e-02, -4.0676e-03,  1.2707e-02,
        -3.7254e-03,  3.5224e-03, -6.5614e-03, -8.6612e-03, -5.4706e-03,
         2.1394e-03,  2.9622e-03,  5.0878e-03,  5.2572e-04,  1.3088e-04,
        -2.2485e-03,  3.1867e-03, -9.4903e-03,  9.9504e-03, -2.4334e-04,
         5.1108e-03,  1.9232e-03,  5.0349e-03,  1.7960e-03,  1.5773e-02,
         7.8626e-03,  1.2005e-02,  1.4100e-03, -3.0331e-03,  1.4547e-04,
         1.6423e-03,  5.0414e-03,  9.2611e-04,  6.5858e-03,  2.8590e-03,
         4.1540e-03, -2.9708e-03,  7.5759e-03,  7.0684e-04, -4.7430e-03,
         2.7262e-02,  3.2401e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0190, -0.6480,  0.0076,  0.0079, -0.0097, -0.0275, -0.0178,  0.0254,
        -0.0024, -0.0048, -0.0078,  0.0110,  0.0206, -0.0080, -0.0178, -0.0338,
        -0.0133,  0.0026,  0.0073, -0.0109,  0.0149,  0.0094, -0.0030,  0.0366,
         0.0329,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.2215e-02,  7.4612e-01,  3.0685e-03, -1.0649e-02, -7.5648e-03,
        -1.5038e-02,  2.0959e-02, -1.5754e-04, -4.2821e-03, -3.6381e-03,
         2.9845e-03,  6.2258e-03, -1.3783e-02,  3.2294e-04, -2.4576e-02,
        -4.6904e-03, -1.0573e-03,  1.0651e-02,  3.4781e-03, -5.3989e-03,
        -9.0352e-04, -7.5887e-03, -4.4650e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4960e-02,  6.6822e-01,  4.2734e-02, -1.4909e-02, -4.7035e-02,
        -3.5887e-03, -1.5805e-03, -5.6275e-03, -3.7582e-02,  3.4439e-03,
        -1.2688e-02,  6.2906e-04, -1.9125e-02, -8.9442e-03, -1.4152e-02,
        -3.5197e-03, -1.5211e-02,  6.3431e-03,  5.4559e-04, -1.1016e-02,
         2.3979e-03, -2.0282e-02, -1.5466e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0608,  0.6731,  0.0050, -0.0280,  0.0326,  0.0188,  0.0403,  0.0578,
         0.0085,  0.0202, -0.0233,  0.0215,  0.0100,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5288e-02,  8.1623e-01,  3.0849e-03, -4.1180e-03, -1.7525e-02,
        -1.7169e-02, -8.1361e-04,  3.0710e-02,  8.5930e-03,  1.8714e-02,
         4.6892e-03, -2.5757e-03,  1.5097e-02,  2.5387e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1027e-02,  7.6165e-01,  5.0547e-02, -3.9557e-02, -1.4751e-02,
         1.2175e-03,  3.9653e-04, -6.1051e-03, -4.0637e-03, -4.3009e-03,
         3.0673e-03, -9.5942e-04,  4.7794e-03,  7.7693e-03,  3.6869e-03,
        -5.3654e-03, -1.6416e-02,  7.2198e-03,  2.7118e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 8.5604e-02,  7.0634e-01,  2.1811e-02,  6.6967e-04, -1.1569e-02,
        -2.4701e-03, -3.8017e-03,  5.0612e-03,  6.9759e-03,  2.4284e-03,
        -1.9201e-03,  6.9559e-03, -1.7700e-03,  1.1609e-02,  3.2307e-03,
        -5.7696e-02,  1.0365e-02, -3.6302e-03, -3.9351e-03,  2.5202e-02,
         4.6572e-03, -1.3619e-03, -5.2564e-03,  1.0769e-03,  1.1966e-02,
        -2.6351e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0225,  0.7133,  0.0534, -0.0283,  0.0150,  0.0263,  0.0045, -0.0061,
        -0.0125, -0.0038,  0.0066,  0.0258,  0.0088,  0.0016, -0.0370,  0.0347,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0287,  0.7104, -0.0579,  0.0497,  0.0110,  0.0318, -0.0106,  0.0120,
         0.0083, -0.0033, -0.0016,  0.0010,  0.0049, -0.0024, -0.0194, -0.0378,
         0.0091,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0577,  0.7126, -0.0122, -0.0121,  0.0798, -0.0210,  0.0190,  0.0631,
         0.0225,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7326e-02,  6.2871e-01,  3.7935e-03,  7.5876e-03, -3.5121e-02,
        -6.3296e-03,  1.2068e-02,  2.0810e-02,  9.5682e-03, -1.4728e-02,
         1.6912e-02, -2.3654e-02, -2.6453e-03,  2.5614e-03, -9.2864e-03,
        -2.1718e-03,  1.0800e-03,  6.4128e-03, -6.3910e-03,  3.0075e-03,
         6.4274e-03,  4.2510e-03,  7.6073e-03,  1.9019e-03,  7.2103e-04,
        -2.1895e-03, -8.1127e-03,  2.9762e-04,  4.4367e-03, -4.1189e-03,
        -5.4598e-04,  3.3745e-03, -1.3499e-04, -3.1704e-03, -8.2037e-04,
        -1.5297e-02, -9.8942e-04,  1.8265e-02,  3.5459e-03,  1.8452e-03,
        -8.0118e-03, -3.0897e-03,  6.6777e-03,  3.1968e-02, -1.0421e-03,
         9.9713e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0221,  0.6491, -0.0481, -0.0193,  0.0329,  0.0137,  0.0057, -0.0053,
        -0.0162, -0.0015,  0.0037, -0.0070,  0.0043,  0.0056,  0.0013, -0.0014,
        -0.0079,  0.0063, -0.0077,  0.0209, -0.0102,  0.0065,  0.0025, -0.0099,
        -0.0158,  0.0021, -0.0035, -0.0015, -0.0333,  0.0347,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0827, -0.7041,  0.0290,  0.0516, -0.0095,  0.0400, -0.0831,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9891e-02,  7.6882e-01, -2.0065e-02, -1.8523e-02, -1.6339e-02,
        -1.7472e-02, -1.8594e-02, -7.4365e-04,  7.1607e-03, -6.2402e-02,
        -9.9898e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0671,  0.7323,  0.0251,  0.0585,  0.0045, -0.0110,  0.0173, -0.0163,
         0.0060,  0.0131, -0.0123,  0.0052,  0.0155,  0.0159,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3741e-03, -7.0431e-01, -5.2455e-02,  2.6739e-02, -2.4097e-03,
        -3.2458e-05, -2.5893e-03, -1.1135e-02,  2.7582e-03,  1.6122e-02,
         6.9911e-03,  7.3651e-03,  1.2344e-02,  5.7495e-03, -1.6094e-04,
         5.4213e-03,  9.3019e-04,  2.2123e-04,  5.5228e-03, -1.2478e-02,
         2.5741e-04, -1.9895e-03, -6.9648e-03,  4.0112e-03,  1.1186e-02,
         7.1899e-03,  1.4130e-02, -8.0132e-04,  1.7147e-03,  7.5658e-03,
         2.0684e-02, -4.8165e-03,  1.6372e-02,  1.8208e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0095,  0.6595,  0.0810, -0.0142,  0.0260,  0.0196, -0.0049, -0.0092,
        -0.0065, -0.0064, -0.0081, -0.0129, -0.0030,  0.0121,  0.0360, -0.0029,
        -0.0011, -0.0126,  0.0017, -0.0012,  0.0067,  0.0044, -0.0020, -0.0016,
        -0.0041, -0.0023,  0.0036,  0.0468,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7309e-02, -6.5488e-01, -1.2845e-02, -3.6225e-02, -3.0946e-03,
         7.9631e-03, -8.8538e-04, -2.6723e-02,  2.1961e-03, -1.2100e-03,
         1.9844e-03, -2.9015e-03, -3.1410e-02,  2.9386e-02,  1.1485e-02,
        -6.4974e-03, -2.1758e-02,  4.3900e-04,  2.6207e-03, -5.6928e-03,
        -5.4925e-03, -2.0449e-03, -3.7648e-02,  4.7313e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 7.5354e-03,  7.4322e-01, -1.3175e-02,  1.7651e-02, -1.1027e-02,
        -9.1154e-03,  1.4099e-02, -7.1862e-03, -1.3189e-02,  4.2077e-03,
        -1.2563e-02, -9.2889e-03,  3.9463e-03, -1.8004e-03,  4.3432e-03,
        -1.1037e-02, -4.6757e-06, -6.3506e-03, -2.0332e-03, -1.4572e-02,
         2.2725e-03, -6.5531e-03, -8.4602e-03, -5.1751e-03,  7.7278e-03,
         2.5083e-02, -7.8936e-03,  3.4618e-03, -4.2780e-03, -1.2695e-02,
         1.0055e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0809, -0.6602,  0.0991,  0.0272,  0.0271,  0.0277,  0.0077, -0.0140,
        -0.0336,  0.0087,  0.0049, -0.0089,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0131,  0.8251,  0.0185, -0.0254,  0.0103,  0.0080,  0.0055,  0.0087,
        -0.0045,  0.0033, -0.0164, -0.0072,  0.0345,  0.0018,  0.0177,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4298e-02, -6.5841e-01,  1.3910e-02,  3.4234e-02, -1.6713e-02,
        -7.2138e-03, -3.4330e-03, -1.2263e-02, -9.1892e-03,  7.1096e-04,
        -3.1581e-03, -1.0173e-02, -3.1193e-03,  7.2325e-03, -1.6156e-03,
        -7.6508e-03, -1.8972e-02, -1.0354e-02, -1.5533e-03,  2.2666e-02,
        -5.8812e-03, -3.3850e-03, -1.2062e-02, -1.9401e-02, -1.4793e-02,
        -2.4448e-03, -5.0426e-03, -3.0980e-03, -6.6322e-03, -1.1968e-02,
        -6.6605e-05, -1.5786e-02, -2.2571e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0571, -0.6044, -0.0235,  0.0752, -0.0166,  0.0129,  0.0097,  0.0128,
         0.0021, -0.0020,  0.0060,  0.0041,  0.0166,  0.0618, -0.0610,  0.0342,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0303,  0.6884,  0.0298, -0.0292,  0.0600,  0.0376,  0.0544,  0.0342,
        -0.0360,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0953, -0.7106,  0.0125, -0.0952,  0.0863,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9007e-03,  8.0354e-01,  1.5748e-02, -6.0375e-03, -1.9550e-02,
        -1.3583e-02, -1.3356e-02, -1.0612e-02, -4.0373e-03, -5.8921e-03,
         2.1339e-02,  7.7407e-03, -4.1102e-04, -1.3264e-02,  5.2695e-03,
        -4.4585e-03,  4.2665e-03, -1.3868e-03,  8.1828e-03, -3.8779e-03,
        -2.7843e-03, -4.6997e-03, -2.1124e-03,  5.4313e-03,  3.9751e-03,
        -1.0829e-02, -5.7161e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1022, -0.7212,  0.0599,  0.0122, -0.0565, -0.0469, -0.0012,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0874,  0.5417,  0.0139, -0.0153, -0.0187,  0.0015,  0.0036, -0.0061,
        -0.0098, -0.0068, -0.0064, -0.0121, -0.0042, -0.0034, -0.0025, -0.0021,
        -0.0368,  0.0557,  0.0006, -0.1311, -0.0189, -0.0213,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4556e-02,  7.5889e-01,  3.7059e-02,  1.9686e-02, -8.5599e-03,
        -1.6334e-02, -1.9148e-02,  1.7702e-02, -1.8159e-03,  1.4359e-02,
         6.7291e-03,  1.0615e-02, -3.1934e-04,  2.1071e-02,  7.1105e-04,
        -9.8608e-03, -1.7214e-02, -1.3423e-02, -1.1944e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.8098e-02,  6.9347e-01,  5.0424e-02, -7.5758e-03, -7.8830e-03,
        -2.7558e-02, -1.8838e-02,  5.0923e-03,  2.0169e-03,  1.0050e-02,
        -1.6482e-03, -1.7357e-03,  1.2315e-03, -1.6717e-03, -2.0986e-03,
         4.8623e-03,  1.5476e-03,  3.8118e-04, -1.8504e-03, -3.3371e-03,
         6.8217e-03, -1.1669e-03, -2.0014e-03,  5.4447e-03,  5.7096e-03,
        -8.0135e-03, -1.3724e-02, -1.1981e-02, -7.0625e-03,  1.4251e-02,
        -2.2456e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-7.3555e-02,  5.1745e-01,  1.0448e-01, -1.2994e-02,  5.3666e-03,
        -1.3292e-02, -1.5230e-02,  2.9293e-03, -5.3234e-03,  5.1138e-03,
         4.4157e-03,  8.0752e-03,  2.0792e-02,  1.2641e-02,  8.8466e-03,
         1.1862e-02,  1.3633e-02,  7.7706e-03, -3.1149e-02,  4.4106e-03,
         3.2222e-03, -3.8669e-03, -5.5790e-03,  8.5269e-03, -1.0552e-03,
         2.3745e-03,  1.4631e-04, -3.1395e-02, -1.5251e-04, -3.9691e-03,
         1.4926e-02, -4.5457e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.3651e-02, -7.5929e-01, -3.1018e-04, -4.2059e-03,  1.6603e-04,
         6.6857e-04,  2.9938e-03, -1.3105e-02, -1.0298e-02, -4.4817e-03,
         1.0553e-02, -2.4836e-03,  1.7224e-02, -2.2352e-03,  1.2465e-02,
        -7.1076e-03, -1.5401e-02, -4.4118e-03, -6.6382e-03, -1.7769e-02,
        -2.1448e-02, -3.0978e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1815e-02,  6.7112e-01,  4.7778e-02,  1.3077e-02,  9.9763e-03,
         4.7184e-03, -3.8675e-02,  5.1602e-03,  1.2941e-02,  1.7301e-03,
        -5.5837e-04, -1.0736e-02, -8.0944e-03, -3.6659e-03,  7.8328e-03,
        -1.2267e-03, -3.5264e-03,  1.7345e-04, -4.3492e-03, -1.2357e-02,
         3.3804e-03, -1.1343e-02,  6.4762e-03, -5.0082e-03,  2.1671e-03,
        -7.3376e-03, -1.5703e-03, -2.0430e-02, -5.6601e-03, -1.7111e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0185e-02,  7.3102e-01, -4.5669e-02, -2.4858e-02, -8.4464e-03,
         5.5691e-04,  2.5430e-02,  5.4447e-03, -7.1224e-03, -2.4758e-02,
        -3.4152e-03,  1.2653e-02, -3.6694e-03,  2.3102e-03,  4.7323e-03,
        -2.9401e-02,  1.7400e-02,  2.9274e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7578e-03, -7.1929e-01, -6.3676e-02,  1.7139e-03,  1.5555e-02,
        -5.9368e-04,  3.2410e-03, -1.0850e-02,  2.4499e-02, -3.4946e-03,
        -4.2357e-03, -2.0793e-03,  3.0035e-03, -6.3525e-03,  1.1455e-02,
         7.7980e-03, -2.5100e-03, -5.2520e-03,  1.2255e-02,  8.0778e-03,
         3.8101e-04,  2.9839e-03,  1.2661e-02, -2.1108e-03,  8.0118e-03,
         6.2446e-03, -1.5104e-03, -3.9726e-03,  1.0734e-03,  1.1211e-03,
         4.5110e-03, -2.5249e-03,  8.9849e-05, -9.8476e-03,  2.7996e-03,
        -2.2697e-02,  7.7704e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4221e-02, -6.6131e-01,  1.2950e-01,  2.7406e-02,  5.9293e-03,
         3.7712e-03,  6.3702e-03,  1.4781e-03, -4.3362e-04, -2.1005e-03,
        -1.0343e-02, -1.3750e-02, -7.4484e-03, -1.1082e-03,  1.2512e-02,
        -7.7908e-03, -3.4640e-03, -5.2648e-03, -5.9230e-03, -8.1872e-03,
        -1.5634e-02,  2.1426e-02,  3.6906e-03,  1.0936e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1866e-02,  6.8422e-01,  1.3765e-02,  1.1056e-02,  4.8277e-02,
         3.8301e-03, -9.3328e-03, -2.6853e-03, -1.3923e-02,  4.0203e-03,
        -1.0915e-02,  3.7998e-03, -2.3044e-02, -1.4907e-03, -6.8539e-04,
         1.8855e-03, -1.6850e-03, -1.5866e-03, -3.5137e-03, -6.9795e-04,
         2.2095e-03,  9.8347e-04,  1.5056e-03,  1.7771e-03,  1.0335e-02,
         3.6724e-03,  5.1781e-03,  6.6142e-03,  1.7576e-02,  2.1105e-04,
         4.3600e-03,  2.0002e-02,  7.2699e-03,  5.3482e-03,  5.7155e-03,
         1.7802e-03, -4.6733e-03, -3.6776e-03,  5.1858e-03, -4.1477e-03,
        -2.4158e-03, -5.2128e-03, -8.0510e-03,  1.2557e-02, -2.7486e-03,
         4.5155e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0185e-02,  6.4430e-01,  2.1008e-02,  4.0076e-02, -3.2770e-04,
         1.0600e-02, -9.3766e-03, -1.5926e-02,  1.3677e-02, -2.5357e-02,
         8.2883e-03,  1.1298e-02,  1.0911e-03, -1.3356e-02,  8.6869e-04,
        -1.1399e-02, -1.1851e-02, -2.0304e-02,  1.2546e-03,  9.9883e-03,
         2.0971e-03, -1.6112e-02, -3.1257e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4603e-02,  8.2714e-01,  5.3861e-03,  2.0361e-02,  1.4499e-02,
         1.9463e-02,  1.8175e-02, -1.5613e-02, -7.4590e-03,  4.7166e-02,
         1.3112e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0388,  0.7064,  0.0481, -0.0302,  0.0038, -0.0046,  0.0264, -0.0208,
         0.0238, -0.0102,  0.0260, -0.0046,  0.0466, -0.0097,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0321,  0.7104,  0.0069, -0.0487,  0.0113,  0.0242, -0.0067, -0.0074,
         0.0180, -0.0111,  0.1032, -0.0200,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6737e-02,  6.5815e-01,  7.4889e-02,  7.9751e-03,  1.4133e-02,
        -5.1775e-03,  3.2895e-05,  4.7820e-03, -3.5284e-03,  1.3985e-03,
        -1.0545e-02,  6.4797e-04,  7.0286e-03,  5.0901e-03, -3.8431e-03,
         2.5879e-04,  2.3910e-03, -9.9492e-03,  5.5336e-03, -1.0543e-02,
         1.1990e-03, -3.9986e-03, -1.5762e-03,  1.5847e-03,  2.1897e-03,
        -6.9767e-03,  2.1338e-02, -4.9520e-03,  1.6291e-02, -3.6264e-03,
         8.6352e-03,  4.2680e-03, -2.5498e-03,  3.2879e-03,  1.6813e-03,
        -2.3596e-03,  2.7361e-03,  1.0632e-03, -3.0047e-03, -2.0557e-03,
         1.3301e-03,  6.2453e-04,  2.1974e-03, -1.3990e-03,  9.4711e-04,
         2.1254e-03,  3.2138e-03, -4.8502e-04, -1.3209e-03, -2.8940e-03,
         9.6832e-04,  2.7709e-03, -1.7859e-03,  1.2721e-02,  7.2063e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-2.9077e-02, -6.4453e-01,  4.3062e-02,  2.9693e-03, -4.6801e-02,
        -3.3946e-02,  5.5530e-03,  1.2308e-02,  3.8628e-02, -9.4565e-03,
         1.2540e-04,  7.1558e-03,  5.4473e-03,  1.1764e-02, -9.0449e-03,
         3.1521e-03,  1.2140e-02,  3.4549e-03, -9.5642e-04, -6.2370e-03,
        -4.5543e-04, -4.7279e-02,  2.6460e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0672, -0.6406, -0.0453, -0.0380, -0.0149, -0.0146, -0.0091, -0.0378,
         0.0039, -0.0312,  0.0033,  0.0038,  0.0101,  0.0141, -0.0025, -0.0128,
         0.0473,  0.0035,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.2074, -0.2308,  0.3100,  0.0024,  0.0371,  0.0298, -0.0037,  0.0325,
        -0.0069,  0.0022, -0.0058,  0.0210,  0.0412,  0.0079,  0.0100,  0.0343,
         0.0015,  0.0156,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5512e-02,  7.3364e-01,  4.1052e-03, -2.0243e-02,  8.6321e-03,
         3.1696e-02,  5.3484e-03, -1.5380e-02, -4.4676e-02, -1.1035e-02,
        -7.0825e-03,  1.3872e-02,  5.7105e-04, -6.3604e-03, -5.1848e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.3990e-03,  6.3223e-01,  4.4847e-02,  3.8789e-02, -8.5964e-04,
        -1.7068e-02, -2.2721e-02,  4.1327e-03,  1.2958e-02,  3.7421e-03,
        -9.2606e-03, -3.6669e-03, -4.5917e-04, -1.0017e-02, -1.0689e-02,
         9.5338e-03,  5.4112e-03,  3.6173e-03,  7.1211e-03, -1.2000e-02,
         5.1243e-03, -2.4024e-02,  1.8523e-03, -3.2578e-03,  8.0872e-03,
         8.9702e-03,  1.0467e-02,  4.0616e-03,  1.6059e-03,  2.2761e-02,
         2.2262e-02, -2.9001e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.8871e-02,  6.6890e-01, -1.6759e-03, -7.5154e-03, -3.7338e-03,
         6.1427e-03,  2.1566e-03, -1.8660e-03,  1.3594e-03, -5.9640e-03,
         9.3870e-03,  3.6332e-03,  4.8223e-05, -8.2623e-03, -5.7955e-03,
        -1.2101e-02, -1.9551e-04,  5.3387e-03,  8.6918e-04,  3.4510e-02,
        -1.2571e-02, -9.2456e-03, -7.2119e-03, -5.6654e-03, -4.3226e-03,
         9.3018e-03, -3.4275e-03, -2.3112e-03, -1.0531e-02, -1.9720e-03,
         1.6397e-03,  1.4362e-03, -7.2034e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.4878e-02,  5.6701e-01,  2.9215e-02,  2.7742e-02,  1.5216e-02,
         1.5259e-02, -7.5390e-03,  1.4380e-03,  3.2805e-03, -1.2848e-03,
        -4.5133e-05, -2.3551e-03, -2.2792e-02, -1.7924e-02,  1.6989e-02,
        -1.3833e-02,  9.0715e-03, -1.9085e-03,  4.4192e-03, -6.1386e-03,
        -4.9760e-03,  3.2107e-03,  1.9947e-03,  8.6489e-03, -8.7247e-04,
        -7.1883e-03,  2.7773e-03,  2.7118e-04,  1.2067e-03,  2.1819e-03,
         2.4928e-03, -4.0595e-03,  1.2379e-03,  3.2871e-03,  7.8550e-05,
        -3.5495e-03, -4.8387e-03,  3.4066e-04,  2.9391e-04, -6.1210e-04,
        -3.3477e-03, -1.4023e-03, -1.7235e-03,  1.9963e-03,  5.3697e-03,
         6.1543e-03,  8.9313e-04, -1.5036e-03,  3.4300e-03,  6.2316e-03,
        -1.2375e-03,  3.3894e-03,  4.2519e-03,  8.3297e-04,  8.7048e-03,
         1.6290e-02,  1.1854e-03, -1.9602e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2747e-01,  6.6774e-01, -3.2015e-02, -7.2321e-03,  2.0604e-02,
         7.4906e-03,  6.6052e-03,  3.5089e-03, -9.2207e-03,  1.1010e-02,
        -2.2667e-02,  8.9479e-03, -1.9493e-02, -4.0174e-03,  4.9571e-03,
         4.2915e-03,  2.1824e-03, -4.8479e-03,  5.0026e-04, -3.0140e-03,
         5.6644e-03, -3.5960e-03, -7.7202e-03, -5.0075e-03, -1.2450e-03,
        -8.9489e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0243,  0.6397,  0.0127, -0.0253,  0.0055,  0.0031, -0.0943, -0.0028,
        -0.0370,  0.0015, -0.0030,  0.0176,  0.0191,  0.0062, -0.0073, -0.0017,
        -0.0080,  0.0047,  0.0062,  0.0309, -0.0037, -0.0068,  0.0014, -0.0078,
        -0.0295,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5464e-02,  6.0844e-01,  2.5275e-02, -2.1100e-02,  1.3255e-02,
        -7.9446e-03,  4.1122e-03, -3.2300e-02, -2.2465e-02, -7.5365e-03,
        -2.1418e-03, -3.1559e-02,  7.5632e-03,  1.3034e-02,  1.2012e-03,
        -3.2162e-03, -3.0358e-03,  3.9133e-03,  3.6143e-05, -7.5725e-03,
        -5.6854e-03,  1.1760e-02, -6.3177e-03, -8.2086e-04, -1.0232e-02,
         2.5351e-03,  1.5135e-03,  1.6224e-03, -3.1616e-04, -1.3256e-03,
         4.7935e-03, -6.1185e-03, -2.7969e-03,  2.7590e-04,  2.7211e-03,
         7.0256e-04, -8.7705e-04, -6.5648e-03, -7.2209e-04,  2.3794e-03,
        -5.7767e-03, -9.5881e-05,  3.4421e-03, -9.4338e-04, -5.8496e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0344,  0.6812, -0.0753,  0.0342, -0.0198,  0.0246,  0.0029,  0.0072,
        -0.0074, -0.0065,  0.0156, -0.0032, -0.0204,  0.0075,  0.0022,  0.0115,
        -0.0104,  0.0012,  0.0028,  0.0058,  0.0051, -0.0089, -0.0121,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0567, -0.7276,  0.0528,  0.0232,  0.0026, -0.0405, -0.0243,  0.0040,
        -0.0154, -0.0063, -0.0095,  0.0370,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-0.0025,  0.7135,  0.0912, -0.0255, -0.0051,  0.0145,  0.0013,  0.0121,
         0.0016, -0.0213,  0.0008, -0.0127, -0.0120,  0.0032, -0.0128, -0.0061,
        -0.0079,  0.0016, -0.0035, -0.0220, -0.0150, -0.0028, -0.0113,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.8396e-03,  7.0792e-01,  2.2805e-02, -1.5412e-02,  8.1072e-04,
        -8.2714e-03, -5.6674e-04,  1.3841e-02, -7.2912e-03,  3.5519e-03,
        -5.0608e-03, -1.0146e-02, -8.2928e-03, -7.8746e-04, -1.0413e-02,
        -4.9374e-03,  2.5257e-03, -1.2487e-02, -3.7484e-03, -1.7096e-02,
         1.0191e-02, -3.4120e-03,  6.5728e-04, -4.0169e-03, -4.4416e-03,
        -7.5018e-03, -3.5454e-03, -3.9293e-03, -9.2183e-03,  5.3952e-03,
        -6.1167e-03, -1.5951e-02, -6.2815e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1052, -0.6904, -0.0370, -0.0062,  0.0037, -0.0119, -0.0115, -0.0084,
        -0.0087,  0.0124, -0.0122, -0.0023, -0.0025, -0.0020, -0.0128, -0.0026,
         0.0091, -0.0404,  0.0208,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0591, -0.7736,  0.0160,  0.0045, -0.0091, -0.0252, -0.0104,  0.0093,
         0.0071,  0.0221, -0.0121,  0.0083, -0.0047, -0.0387,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.9425e-04,  8.5651e-01,  8.9999e-03, -1.6460e-03, -2.5102e-02,
         1.6410e-02, -1.7275e-03,  1.1059e-02,  2.1488e-03,  2.7806e-03,
         2.5533e-04, -2.9606e-04,  3.0684e-03, -1.7387e-02, -1.1748e-02,
        -5.7642e-03, -5.7505e-03, -3.6440e-03, -4.1306e-03, -2.2843e-03,
         1.4704e-03, -1.7123e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2698e-02, -7.7625e-01,  8.4356e-03,  5.6759e-03, -5.9794e-03,
        -1.0331e-02,  8.4850e-03,  1.5589e-02,  3.6393e-03, -1.7253e-02,
        -4.9850e-04,  2.3299e-02,  2.6134e-02,  3.8281e-03, -4.5595e-04,
        -2.1424e-03, -3.4687e-03,  1.0904e-02, -4.4930e-03, -3.3455e-03,
         1.1953e-03, -4.4003e-03,  2.5678e-03,  2.2781e-03,  3.8754e-03,
         1.5133e-03, -2.9974e-03, -6.8839e-03,  1.7651e-02,  3.7290e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0970,  0.5925, -0.0345, -0.0122,  0.0322,  0.0219, -0.0019,  0.0118,
         0.0163, -0.0042, -0.0436, -0.0149, -0.0267,  0.0081, -0.0069, -0.0479,
        -0.0273,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3744e-02, -7.6274e-01, -3.6774e-02,  2.9635e-02,  3.4789e-02,
         1.3849e-02,  1.7671e-02, -3.0635e-03, -5.4337e-04,  8.5784e-02,
         1.4124e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.6328e-03, -7.0265e-01, -8.7145e-02,  1.3813e-02,  9.2780e-03,
         1.6134e-02,  9.7117e-03, -1.0606e-02,  9.1789e-03,  5.0231e-03,
         1.4782e-02,  1.6209e-02,  4.7944e-03, -5.5724e-02, -1.1575e-02,
         2.5673e-02,  6.9652e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1810e-02, -7.2124e-01,  1.6563e-02, -2.5107e-02, -1.5708e-02,
         1.0940e-02,  8.9459e-04, -8.0047e-03,  8.1445e-03,  4.5181e-03,
         5.4916e-03, -1.3751e-02,  4.1522e-03, -1.5513e-02, -2.3044e-02,
         4.0141e-03,  4.3474e-03,  1.0422e-02, -4.5892e-04,  8.5633e-03,
         1.2803e-03,  3.8231e-03,  1.6652e-03, -4.6429e-03, -7.0292e-03,
         5.8872e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0237,  0.6408,  0.0442, -0.0299, -0.0332,  0.0264, -0.0303,  0.0077,
        -0.0082, -0.0062, -0.0090, -0.0098, -0.0034,  0.0013, -0.0093,  0.0329,
        -0.0109,  0.0013, -0.0029,  0.0075, -0.0035, -0.0065, -0.0053, -0.0105,
         0.0066, -0.0047, -0.0007,  0.0058, -0.0128, -0.0046,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0216, -0.7623, -0.0058,  0.0208, -0.0267,  0.0128, -0.0138,  0.0128,
        -0.0039, -0.0316,  0.0103,  0.0022,  0.0067,  0.0109,  0.0137, -0.0139,
         0.0056,  0.0217, -0.0028,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([-4.1005e-03,  7.0537e-01,  2.6623e-02,  6.4929e-02,  7.9175e-03,
         5.0687e-04,  1.0915e-02, -2.6379e-02, -3.4052e-03, -5.2398e-04,
        -2.4113e-03, -7.8215e-03, -6.9875e-04,  2.5550e-03, -7.6689e-03,
        -8.7243e-03,  1.0206e-02,  1.1294e-02,  2.0944e-03,  9.9250e-03,
         1.2322e-03,  2.2241e-03,  4.2694e-03,  1.1310e-02, -1.0299e-02,
        -1.0579e-02,  4.6014e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.4110e-02,  7.7828e-01, -2.3252e-02,  1.2424e-03,  1.7974e-02,
         1.3066e-02, -2.5551e-02, -1.5477e-02, -1.0504e-02,  7.0935e-03,
        -5.9175e-03,  5.4757e-05, -5.4269e-03, -2.0512e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.4646e-02, -5.9612e-01, -1.5056e-02,  1.0739e-02, -1.7628e-02,
         5.5394e-03,  4.1600e-03,  3.9358e-02, -7.3178e-03, -1.5299e-02,
        -5.3162e-03,  2.3859e-03,  7.5313e-04, -6.2907e-03, -4.9952e-03,
        -1.1422e-02, -1.5489e-02,  7.7571e-03,  4.9264e-03, -1.6559e-04,
         5.7047e-03,  5.8254e-03, -1.4611e-03,  1.0944e-02, -7.3195e-02,
        -3.7504e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1666e-03,  8.1985e-01,  2.9921e-02,  1.8441e-02,  2.6067e-04,
        -8.8885e-03,  7.1201e-03, -2.2105e-02, -2.9637e-02, -4.4739e-03,
        -3.9883e-03, -3.3249e-02, -1.0646e-02,  9.2501e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9051e-03,  7.4815e-01,  4.4605e-02,  1.9865e-02,  5.7132e-04,
         2.7495e-02,  6.3604e-03, -2.2508e-03,  7.3473e-03, -2.0361e-03,
         2.3945e-03, -1.0730e-03,  4.5928e-03,  2.1772e-02,  1.0241e-02,
        -3.5295e-03,  6.9554e-03, -5.6013e-04, -2.7398e-03, -1.3651e-03,
         6.8232e-03, -8.5905e-03,  4.4021e-03,  6.8776e-03, -3.1752e-03,
        -4.3622e-03, -7.4422e-03,  1.3211e-03,  1.7097e-02, -1.7396e-03,
         1.8539e-02, -8.2426e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.4209e-02,  7.5157e-01,  2.2725e-02, -5.8493e-03,  8.4229e-03,
         3.6776e-03,  1.0871e-02, -6.5886e-03,  6.8715e-04, -5.8270e-03,
        -1.5839e-04, -2.9146e-03, -2.4897e-03,  2.3653e-03,  2.0282e-03,
        -4.3091e-03,  2.7127e-03, -3.6989e-04, -1.3525e-02, -7.2726e-03,
        -7.1275e-03,  1.1040e-03, -9.0439e-03,  1.7371e-02, -4.6777e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0123,  0.8877,  0.0107,  0.0070,  0.0242, -0.0581,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9211e-02,  7.1596e-01,  3.0759e-02, -4.2531e-02,  2.4649e-02,
        -1.5915e-02, -5.1779e-03, -1.0701e-03, -2.2513e-03, -1.2148e-02,
        -1.2845e-02, -5.3183e-03, -6.8390e-03,  1.3141e-03,  5.4459e-04,
        -1.1501e-03,  3.6436e-03,  1.5545e-03, -9.9597e-03, -4.9397e-03,
        -3.6794e-03, -4.1307e-03, -1.0528e-02, -4.7501e-03, -7.6673e-03,
        -9.5842e-03, -4.7014e-03, -1.8639e-03, -2.6703e-03,  3.3502e-03,
        -9.2807e-04, -5.1125e-03, -4.4002e-03, -2.9387e-03, -5.7220e-03,
         6.0953e-04, -9.5827e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2400e-01, -6.1759e-01,  5.4624e-03, -2.6028e-02, -6.6968e-03,
        -4.2621e-03, -1.1529e-02,  1.3298e-02, -4.2583e-03,  1.5133e-03,
         1.4210e-03, -5.2667e-03, -4.4894e-03,  1.4969e-02,  1.0186e-02,
         2.0781e-02, -9.3973e-03, -7.8787e-03, -4.7877e-03, -3.2934e-03,
         1.0481e-02, -2.4624e-03,  4.7647e-03, -1.9876e-02, -3.1825e-03,
         1.0291e-03,  5.5306e-03, -8.1938e-03,  1.6442e-02, -3.1528e-03,
        -2.1460e-03,  4.7655e-04,  1.9887e-03, -7.2317e-03,  1.5934e-02,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.7971e-03, -6.0238e-01,  1.9228e-01, -4.7614e-02,  6.9130e-03,
        -4.0850e-02,  8.9932e-03, -5.9678e-05,  1.8825e-03,  1.2286e-02,
        -6.0414e-03,  2.1556e-02,  1.6411e-02,  1.1542e-02,  2.1391e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.9398e-02,  6.7168e-01,  3.2672e-02, -4.7013e-02, -1.4208e-02,
        -1.1893e-02,  1.5590e-02,  2.2365e-03, -3.7636e-03,  1.4037e-02,
         3.6081e-03, -6.0313e-03,  3.7260e-04, -5.7126e-03, -1.0961e-03,
         1.2948e-03,  8.3125e-04, -2.1076e-03, -7.5862e-03, -3.2997e-03,
        -3.1425e-03, -8.8509e-03, -9.1955e-04, -8.6928e-03,  1.7489e-03,
        -2.0262e-03, -1.8657e-03, -1.3298e-03, -1.6739e-03, -1.5124e-03,
        -7.2526e-04, -4.1196e-03,  1.8959e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.6161e-03, -7.4790e-01, -1.8447e-02,  5.6614e-03,  7.3511e-03,
         6.6001e-03,  5.9209e-03, -7.8469e-03, -2.8888e-02,  5.9136e-03,
        -2.3547e-04, -1.4157e-02, -4.6251e-03, -6.6014e-03,  6.1407e-03,
         4.4183e-03, -3.4577e-03, -2.1746e-03,  1.8691e-02, -4.9256e-03,
         9.2857e-03,  6.3335e-03, -1.2197e-02, -2.9246e-03, -1.6196e-03,
        -6.0323e-03,  7.0436e-03,  2.9099e-02,  1.8894e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 0.0451,  0.7210, -0.0023, -0.0326, -0.0337, -0.0203,  0.0053, -0.0114,
        -0.0039, -0.0515,  0.0019,  0.0050, -0.0043,  0.0188, -0.0427,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.5356e-02,  7.7143e-01,  1.3793e-02, -4.5456e-03, -1.3295e-02,
        -9.1914e-04,  4.7214e-03,  4.8492e-04,  1.3455e-02,  1.8198e-03,
        -5.3876e-03, -2.5641e-04,  7.4342e-03, -5.3453e-03,  4.4227e-04,
         1.7658e-03, -2.8280e-03,  3.8950e-03,  5.7590e-04, -1.7944e-03,
         9.6576e-04,  4.7618e-03, -2.2742e-03, -1.5684e-03, -1.3611e-03,
         2.5056e-04, -1.5333e-03,  5.4302e-03, -1.0395e-03, -2.0984e-04,
         1.4317e-03, -2.6472e-03,  1.3439e-03, -4.1824e-04, -8.1690e-03,
        -6.0412e-03, -3.5429e-03,  9.8314e-04,  4.6777e-03,  3.1445e-04,
        -2.2381e-03, -6.7183e-03, -5.6303e-04,  8.2082e-04, -8.4596e-04,
        -1.2955e-04,  1.2735e-03,  2.7077e-03,  1.6191e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6274e-02,  6.3864e-01,  3.6521e-03, -2.2347e-02, -2.4828e-02,
        -4.6274e-03,  3.5982e-03, -7.3194e-03, -1.7033e-02, -1.1982e-02,
        -2.5271e-03, -4.9740e-03,  9.1586e-03, -4.9943e-04, -6.8373e-03,
        -6.1976e-03,  1.6814e-02, -7.3678e-03,  1.5213e-03, -4.8529e-03,
        -1.7035e-02, -1.3170e-02, -1.6295e-03, -3.7593e-03,  1.3737e-03,
         4.2595e-03, -3.5146e-03,  1.8333e-02,  4.3565e-03, -7.0319e-03,
        -7.2811e-03, -8.6320e-03, -1.6628e-02, -2.9076e-02, -5.2868e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4089e-02, -7.4547e-01, -5.6294e-02,  1.8805e-02,  2.3142e-02,
         2.1145e-03, -3.2402e-04, -2.8328e-03, -4.1962e-03,  4.8457e-03,
         1.6276e-02, -2.5744e-03,  1.0198e-02, -1.5408e-03,  6.9143e-04,
         7.1075e-03,  7.1472e-04,  5.6022e-02,  2.2757e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2700e-02, -7.6777e-01,  2.3786e-02, -2.5127e-02, -8.8249e-03,
        -2.3096e-02, -1.9729e-03,  1.2417e-02, -2.0175e-02,  2.5567e-04,
         7.3996e-04,  4.9402e-03, -1.9219e-02,  4.7383e-02, -2.1596e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.6955e-03,  6.8179e-01,  4.1526e-02,  3.4367e-03, -7.2187e-03,
        -4.4340e-02,  1.6841e-02, -1.2642e-02,  1.6948e-02, -4.4096e-02,
         5.0966e-03, -1.1116e-02, -5.7848e-03, -1.0487e-02,  4.1722e-03,
        -8.1595e-03, -1.4722e-03,  5.4152e-04, -6.8636e-03,  1.9261e-02,
        -4.1193e-02, -8.3143e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8466e-02,  8.0694e-01,  1.5913e-02,  1.4211e-02,  1.1631e-02,
         5.8841e-03, -1.4282e-02,  5.7611e-03, -2.9351e-02,  3.1806e-03,
        -6.3951e-03, -4.1192e-03,  4.0461e-03,  3.3034e-03, -7.7527e-03,
        -1.1656e-03,  6.7910e-04, -2.2706e-03,  5.8389e-04,  1.0388e-03,
         2.4890e-03,  2.5876e-03, -3.5461e-03,  1.3835e-02, -1.8139e-03,
         2.3088e-04,  1.0665e-03,  3.2585e-03,  2.2254e-03, -1.9734e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.2649,  0.4516, -0.0292, -0.0225, -0.0176, -0.0076, -0.0170, -0.0102,
         0.0300, -0.1082, -0.0413,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0198,  0.6842,  0.0095, -0.0203,  0.0026, -0.0145,  0.0121,  0.0138,
         0.0275,  0.0065, -0.0357,  0.0012,  0.0157,  0.0031,  0.0035, -0.0051,
         0.0122,  0.0054, -0.0030, -0.0031,  0.0795,  0.0218,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6756e-02, -6.5621e-01,  9.2251e-03,  1.5607e-03, -6.4543e-03,
        -6.6595e-02, -5.1388e-03,  1.9767e-02,  1.3221e-02,  3.5120e-03,
         1.6192e-02,  4.2641e-02, -9.4660e-03,  4.0080e-03,  3.3447e-03,
         1.3372e-05,  9.8035e-05,  2.7959e-03, -7.2071e-03,  4.7430e-04,
        -1.2604e-03,  4.1121e-03, -9.4219e-05,  1.2768e-02, -9.4621e-04,
        -2.1838e-03, -1.3045e-02, -5.5712e-04, -1.8182e-03,  1.6595e-02,
        -1.1515e-03, -1.1407e-02, -1.9382e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.4630e-02,  6.5546e-01,  5.1283e-02, -2.3452e-03, -2.6402e-02,
         1.0189e-02,  5.9019e-03, -1.5734e-02,  3.5065e-03, -3.6405e-03,
         3.0708e-03,  3.8052e-03,  6.5453e-04, -6.4256e-03, -2.1163e-02,
         2.5396e-02,  8.2556e-03, -1.3881e-02, -5.9575e-03,  2.0258e-04,
         6.2874e-03, -9.7102e-03,  9.1944e-03, -6.4554e-03,  3.1300e-03,
         6.7571e-03, -3.3269e-02,  2.7293e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8079e-02,  6.6125e-01,  7.2040e-02,  3.6487e-02,  3.0098e-03,
        -1.3577e-02,  2.3601e-02,  9.5731e-05, -1.2245e-03, -4.9025e-03,
        -1.2537e-02, -2.1709e-04, -1.6287e-02,  4.6770e-03, -7.2537e-03,
        -1.4441e-02, -9.4769e-03, -6.0934e-03, -2.1418e-02, -1.3351e-03,
        -1.2965e-02,  2.1355e-04, -2.0913e-02, -8.4205e-03, -2.9481e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 1.2256e-02,  6.7429e-01,  5.1539e-02,  4.3841e-03, -1.1294e-02,
         2.9184e-02, -9.9746e-03, -4.9499e-03,  5.2422e-04, -8.7448e-03,
         4.5334e-03, -1.9902e-03,  8.5148e-04, -1.0058e-02, -5.0841e-03,
        -1.8643e-02, -1.0204e-02,  1.1621e-03, -7.5371e-04, -7.8557e-03,
         1.3140e-03,  4.3909e-05, -4.0261e-03,  7.9255e-03, -6.0065e-03,
        -1.1041e-02,  4.3835e-03, -4.7985e-03, -4.5827e-03, -8.5938e-03,
        -2.5569e-03, -1.0275e-02, -6.8585e-03, -7.8758e-03, -2.3388e-03,
         7.7371e-03, -7.5500e-03, -2.2177e-03,  4.6229e-03,  3.0822e-03,
         2.3890e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.9370e-03,  7.6905e-01,  7.0930e-02,  9.7222e-03,  2.6867e-02,
        -5.5511e-03,  4.8167e-03,  1.0724e-02, -1.2529e-03, -2.0343e-03,
        -6.7807e-03, -1.6609e-03,  9.1949e-03, -4.0834e-03,  4.3614e-03,
        -4.2714e-03,  2.6133e-03,  2.4562e-03, -7.5647e-04,  2.6082e-03,
        -7.1814e-03, -1.3855e-02, -6.8908e-03, -1.2719e-02,  1.0685e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4477e-02,  7.4111e-01,  1.6167e-02, -4.6374e-02, -5.4666e-04,
        -2.1228e-02, -3.4790e-03,  7.4104e-03, -8.5449e-03,  4.8499e-03,
        -2.4326e-02,  4.7073e-03,  3.4527e-03,  3.4097e-03, -9.0390e-03,
         1.2825e-02,  7.7377e-04,  1.1985e-02, -4.9115e-03,  1.1503e-02,
         3.8876e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0394,  0.7848,  0.0367, -0.0010,  0.0086, -0.0194,  0.0230, -0.0066,
         0.0023, -0.0212, -0.0063, -0.0065, -0.0103, -0.0221, -0.0015, -0.0036,
         0.0020, -0.0046,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4661e-02,  7.4425e-01,  2.1367e-02,  2.8587e-02,  1.8309e-02,
        -1.4044e-02,  4.4888e-03, -9.2495e-03, -3.2352e-03,  6.4169e-04,
        -1.1883e-02, -2.1158e-04, -4.8288e-03, -9.3556e-03, -6.5846e-04,
        -3.3387e-03, -1.0693e-02, -2.9031e-03,  4.3598e-04,  7.4846e-03,
        -2.7225e-03,  9.8844e-05,  6.4581e-03,  5.2034e-03, -7.7114e-03,
        -2.0212e-03, -5.7954e-03, -1.3232e-04, -3.1689e-04, -8.6176e-03,
        -3.9155e-03,  3.0066e-03,  4.9972e-03, -6.0257e-03, -2.1297e-03,
        -1.1113e-03, -5.5468e-03,  3.5626e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4676e-02,  5.6923e-01, -3.2821e-02,  9.7720e-03, -3.2831e-02,
        -1.5617e-02,  1.2310e-02,  5.1946e-02,  7.1663e-03,  1.9677e-02,
        -1.1049e-02,  2.3319e-04,  1.1734e-02, -2.4085e-02,  1.0366e-02,
        -8.1167e-04, -2.3123e-02, -4.3110e-02, -2.2028e-02, -4.7413e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0189,  0.7589,  0.0555,  0.0133,  0.0080, -0.0239,  0.0017,  0.0116,
         0.0156, -0.0067,  0.0041, -0.0078,  0.0388, -0.0352,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7561e-02,  6.5905e-01,  2.5627e-02, -1.2291e-02, -1.4184e-02,
         2.8366e-03, -3.4991e-03,  5.3505e-03,  1.1496e-02, -1.8154e-02,
        -3.6295e-02, -1.5979e-02, -2.1715e-03,  3.1491e-03, -1.5978e-02,
        -1.2605e-02,  2.3154e-02,  1.2213e-04,  1.5537e-02,  1.2109e-03,
         1.5683e-02,  1.6523e-02,  1.4780e-03,  2.0876e-02,  6.7355e-04,
        -8.1387e-03,  1.6584e-03,  1.8813e-03,  8.1225e-04,  1.6034e-02,
         1.0190e-02,  9.7933e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3434e-03,  8.3455e-01,  1.8065e-02,  9.9475e-03,  1.3810e-02,
         4.1777e-03,  1.0444e-02, -2.0197e-02, -1.1036e-03, -7.5588e-03,
        -7.8756e-04, -7.8921e-03,  4.6107e-03, -1.1965e-02,  1.0596e-03,
        -3.3315e-03,  1.5224e-03,  8.2926e-03, -3.0392e-02,  3.9490e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0408e-02,  5.9969e-01, -5.4273e-03,  1.0303e-02,  3.2442e-03,
        -8.9263e-03, -4.1116e-03, -2.6907e-02, -1.4127e-02, -2.3624e-03,
        -9.3919e-04,  4.2752e-03, -4.2921e-03,  3.1402e-03,  1.9206e-03,
        -6.4389e-03, -6.9907e-03, -3.3157e-04, -4.8623e-03,  5.5247e-03,
        -1.9745e-03,  1.3522e-02, -2.2487e-03, -8.8142e-03, -2.1228e-03,
        -6.4542e-03, -1.3360e-02, -1.3250e-02, -3.0929e-03, -6.8525e-03,
         1.6401e-02,  7.6765e-03,  2.5205e-02,  1.6615e-03,  3.2718e-03,
        -4.8827e-03, -2.6047e-03, -8.0094e-04, -2.7956e-03, -4.2783e-03,
        -1.2517e-02,  3.7650e-03,  2.0753e-02,  9.7176e-03,  1.2317e-02,
         3.2691e-02,  3.2747e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5290e-02, -6.9240e-01,  4.8964e-04, -6.9999e-03, -1.4854e-02,
        -1.3203e-02, -1.3917e-02,  8.6531e-03, -5.5057e-02, -4.0792e-02,
         1.3383e-02, -3.3934e-02, -6.0230e-03, -1.6721e-02, -4.8281e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5342e-02,  7.3757e-01,  2.4589e-02,  1.2883e-02, -7.0419e-03,
        -3.3939e-03,  6.2657e-04,  2.8790e-02, -1.1832e-02, -7.3615e-04,
        -2.6882e-03, -5.6719e-03, -4.0026e-03,  1.6752e-02, -1.6743e-02,
        -5.6733e-03,  1.1101e-03, -5.0440e-03, -3.3764e-03, -6.1519e-04,
        -2.0500e-03, -2.9535e-03,  3.6954e-03,  1.3702e-03,  1.2357e-03,
         2.4121e-03,  1.4461e-03,  1.3254e-02,  5.3610e-03, -2.9428e-03,
        -6.3929e-03, -3.7150e-03,  5.5480e-03,  9.2699e-04,  1.9433e-03,
        -1.2647e-03, -1.3706e-03, -7.3472e-04, -5.2154e-04,  2.7999e-03,
        -2.2089e-03,  5.1828e-04, -4.3565e-03,  2.0355e-04,  5.8793e-04,
        -7.2280e-03, -3.8105e-03,  1.4671e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 0.0395,  0.7793,  0.0364, -0.0088,  0.0014, -0.0111,  0.0101, -0.0135,
        -0.0017,  0.0018,  0.0039, -0.0014, -0.0077, -0.0088, -0.0057, -0.0010,
        -0.0195, -0.0118, -0.0063, -0.0075, -0.0039,  0.0008,  0.0022, -0.0042,
         0.0090,  0.0027,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0291,  0.7298,  0.0291, -0.0260,  0.0371, -0.0107,  0.0055,  0.0018,
        -0.0107,  0.0057,  0.0041, -0.0128, -0.0067, -0.0124, -0.0040, -0.0029,
        -0.0169, -0.0141, -0.0098,  0.0024, -0.0282,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5487e-02,  7.6060e-01,  2.8304e-02, -5.1105e-03,  2.9146e-02,
        -4.1955e-03, -1.3154e-03,  1.3206e-03, -3.2524e-04,  2.0144e-03,
         3.7296e-03,  1.0576e-03, -2.0985e-03, -8.7093e-03,  4.0217e-03,
        -2.4400e-03,  8.5500e-04, -7.0596e-03,  2.1504e-02,  5.0549e-03,
        -6.4422e-03,  7.6285e-04,  1.0875e-02,  8.9713e-03,  8.2220e-03,
        -2.2511e-03, -2.4486e-03,  2.9418e-03, -1.5923e-04, -5.8422e-03,
        -6.8721e-03, -1.3032e-03,  1.5380e-02,  4.6914e-03,  9.1550e-04,
         1.7576e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0074,  0.7399, -0.0244, -0.0089,  0.0075, -0.0232, -0.0251, -0.0031,
        -0.0138, -0.0095, -0.0265, -0.0013, -0.0055, -0.0314,  0.0053,  0.0086,
        -0.0114, -0.0011, -0.0144, -0.0317,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0093, -0.6751, -0.0326, -0.0141, -0.0211,  0.0085, -0.0159,  0.0023,
         0.0188,  0.0092, -0.0130,  0.0156,  0.0261,  0.0116,  0.0171,  0.0197,
        -0.0013,  0.0100,  0.0068,  0.0386,  0.0332,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0815,  0.7708,  0.0336, -0.0024,  0.0424,  0.0016, -0.0030,  0.0167,
         0.0158, -0.0099,  0.0009, -0.0030,  0.0183,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1424e-02, -7.5949e-01, -3.0287e-02,  1.2986e-02,  7.1591e-04,
         3.0968e-03,  2.5964e-02,  8.7560e-04,  1.9676e-03,  1.0537e-02,
         1.3466e-02,  3.8298e-03,  3.4937e-03,  1.1604e-03, -2.6923e-02,
        -1.1615e-02,  1.4158e-02,  6.9327e-04,  2.0942e-03,  5.3399e-03,
        -2.1150e-03, -8.5672e-04,  4.5776e-03, -7.6624e-03, -3.8392e-04,
         1.1433e-03,  1.0440e-03, -2.8022e-03, -7.6572e-04, -1.0470e-02,
        -2.6143e-03, -7.3396e-04, -7.7990e-04,  7.9263e-03,  1.7640e-03,
         1.2240e-02, -2.0101e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.6228e-02,  6.7956e-01, -1.9806e-02, -3.8034e-04, -3.2484e-02,
         1.1652e-02,  4.5309e-03,  1.1927e-02, -1.0539e-02,  2.4426e-03,
        -1.8725e-02, -9.5098e-03,  1.2348e-02,  1.5020e-03,  1.1263e-02,
         5.8095e-03,  7.2853e-03, -9.6778e-03, -7.3515e-05, -3.2601e-04,
         1.2364e-02,  7.1709e-03,  5.0431e-02,  3.9623e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.8077e-02, -6.3028e-01, -8.7895e-02, -6.4964e-03, -1.6454e-02,
        -1.1407e-02,  5.2194e-05,  6.5212e-03,  8.5951e-04, -1.0884e-02,
         2.0202e-02,  4.7185e-03, -1.1911e-02,  2.0993e-03, -1.2510e-03,
        -1.3337e-02, -3.5419e-03, -5.0205e-03, -2.8513e-03,  2.0858e-03,
        -3.1853e-04, -8.1636e-04, -5.9398e-03,  3.8357e-03, -2.7271e-03,
         1.9690e-03, -7.3678e-03,  1.0853e-02, -2.9416e-03,  5.8074e-03,
         1.5585e-02,  1.5891e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6336e-02,  6.7562e-01, -4.3903e-03, -3.5622e-02, -1.3911e-02,
        -3.8998e-02, -8.0866e-03,  2.9200e-03,  2.0641e-02,  1.4344e-02,
        -4.2776e-03, -1.8416e-03, -7.8149e-03, -4.3957e-04, -1.2730e-03,
        -1.7072e-03, -1.2696e-02, -1.6894e-03,  1.0109e-02,  1.1292e-02,
        -5.7903e-03,  1.6496e-02, -4.4074e-03, -4.2974e-03, -8.9629e-03,
        -8.5255e-05,  1.2181e-02,  3.1943e-03, -6.1898e-03, -1.8303e-02,
         2.6083e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1735e-02, -7.2751e-01, -2.4544e-02,  2.2282e-02,  1.2545e-02,
         1.5062e-02,  2.6666e-02,  2.5042e-02,  6.9314e-03, -1.0301e-02,
         8.9772e-03, -3.2892e-03, -1.4064e-02,  3.6722e-03,  8.6526e-05,
        -1.6887e-02, -2.3982e-03,  1.8726e-03, -5.8591e-04, -2.9947e-03,
        -1.9643e-02, -2.0609e-02,  2.2988e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.6909e-02, -6.1276e-01, -2.0729e-02,  5.7007e-04,  2.0972e-02,
         3.5889e-03,  2.4784e-02,  5.6026e-03,  2.6201e-02,  1.0731e-02,
         2.5363e-03,  4.1901e-02,  1.0825e-02,  1.6208e-02,  2.3573e-03,
         1.4830e-02,  9.7275e-03,  1.2386e-02,  4.7142e-03,  5.4362e-03,
         4.6102e-04,  9.5746e-04,  1.8304e-03, -2.2672e-03,  3.2924e-03,
         2.3043e-03, -5.6320e-02,  1.8801e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-8.2644e-02,  6.9606e-01,  2.1024e-02, -4.9358e-04,  8.1091e-03,
         1.4383e-02,  7.4715e-03,  6.9287e-04, -9.3869e-04, -6.2610e-03,
         9.1369e-03, -3.1230e-03, -1.6505e-03, -5.4498e-03, -3.4219e-03,
         6.3327e-03,  3.1654e-03, -5.7527e-04, -9.6172e-04, -6.2121e-03,
         1.6546e-02,  9.7140e-04, -1.2636e-02, -9.6910e-03,  1.5203e-03,
         2.8690e-03,  1.3671e-02,  5.6162e-02,  7.8255e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.6783e-03,  7.6557e-01, -6.7131e-04, -9.9273e-03,  6.5508e-03,
         2.4388e-02,  9.0570e-04, -9.9579e-05,  3.0240e-03, -1.7873e-02,
        -1.2947e-03, -1.6260e-02, -4.0248e-03,  6.1654e-03, -9.5984e-03,
         1.1317e-02,  2.7394e-03,  3.2564e-03,  3.1951e-02, -1.2125e-02,
        -4.4020e-03,  3.0698e-02, -2.0640e-02,  1.0840e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4896e-02,  6.1436e-01, -2.6296e-02,  6.6789e-03,  1.7419e-02,
         4.8264e-03,  1.9294e-03, -6.8721e-03,  1.1133e-02, -1.1598e-02,
        -2.2099e-04,  2.9712e-02, -1.4435e-02,  5.0268e-04,  6.6548e-03,
         1.9209e-03, -6.3425e-03,  6.5761e-03, -1.4349e-02,  1.3255e-02,
         1.6849e-02,  2.6216e-03,  1.1850e-02,  1.6316e-03,  2.7304e-03,
        -4.2977e-03, -5.2949e-03,  7.0789e-03, -3.9519e-03, -4.8908e-03,
        -3.7385e-03,  2.7137e-03,  2.6495e-02, -5.5874e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.6485e-02,  6.8306e-01, -2.2220e-02, -3.5543e-03,  1.2482e-02,
         1.7571e-02,  8.0593e-03,  1.1504e-02,  7.3485e-03, -4.1052e-04,
         1.6941e-02,  6.2616e-03,  9.9403e-03,  6.2189e-03,  9.0724e-03,
         8.0818e-03, -1.7293e-03,  3.3847e-03,  1.6745e-02, -4.3402e-03,
        -1.0257e-03,  7.3569e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.1210e-02, -7.6983e-01,  1.1848e-02,  2.7548e-02, -1.7242e-02,
        -1.9049e-02,  2.4143e-02,  1.0822e-02, -4.5767e-03,  2.0304e-03,
        -9.9839e-03,  1.1218e-02, -2.6846e-04, -6.7053e-03,  3.5159e-03,
        -1.3954e-02,  1.0571e-02, -5.4825e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1035e-02,  7.9368e-01, -2.5164e-07,  2.0195e-03,  8.4991e-03,
         1.7843e-02,  2.6930e-03,  6.2891e-04,  1.2355e-02,  1.7075e-02,
         5.8091e-03, -1.6586e-03, -1.8446e-02, -1.1425e-02, -1.0804e-03,
        -2.1007e-04, -1.3000e-03, -1.5382e-02, -1.8643e-03, -4.2840e-03,
         1.1697e-02, -3.0040e-03, -5.8007e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0285,  0.8206, -0.0089, -0.0443, -0.0162,  0.0013, -0.0055, -0.0111,
         0.0037, -0.0094,  0.0028,  0.0038, -0.0019,  0.0068,  0.0058,  0.0095,
         0.0200,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0735,  0.6809,  0.0678,  0.0033,  0.0364, -0.0192,  0.0089, -0.0092,
        -0.0023,  0.0013,  0.0007, -0.0010, -0.0075, -0.0031,  0.0057,  0.0071,
         0.0206, -0.0513,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0315, -0.7156,  0.0052,  0.0109,  0.0068,  0.0267,  0.0188,  0.0156,
         0.0129, -0.0283,  0.0023, -0.0083,  0.0068,  0.0095,  0.0078,  0.0008,
         0.0041,  0.0160, -0.0067, -0.0034,  0.0071,  0.0073,  0.0048, -0.0060,
         0.0169,  0.0200,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.8441e-02,  6.4550e-01,  2.7746e-02, -3.7127e-02,  8.4195e-04,
         4.1556e-03,  9.5619e-03,  2.4808e-02, -8.9564e-03, -2.1539e-02,
        -8.4216e-03,  6.9305e-03,  5.9936e-03,  8.2574e-03, -8.5336e-03,
         5.9067e-04,  2.8580e-03, -4.6658e-03,  6.2137e-03,  2.8397e-03,
         9.8301e-03,  2.4617e-04,  9.4745e-04,  4.8703e-03,  5.3575e-03,
         1.3831e-02, -4.8391e-03,  3.6070e-03,  2.4635e-02, -3.1226e-02,
        -6.6320e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1226,  0.6606,  0.0177, -0.0026,  0.0305,  0.0157,  0.0429,  0.0141,
        -0.0021,  0.0133,  0.0104, -0.0115,  0.0091,  0.0044, -0.0051, -0.0016,
         0.0247,  0.0112,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.2927e-02,  8.1491e-01, -2.6861e-02,  2.5669e-02, -8.3412e-03,
         4.0964e-02, -1.0230e-02,  9.4822e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([-8.3486e-03,  7.0029e-01,  2.6055e-02, -1.8719e-02,  7.0207e-03,
         1.7444e-02,  1.1621e-02,  8.5892e-03, -1.2001e-02,  3.2132e-03,
         2.1541e-03, -9.1551e-03,  8.3500e-03,  1.8296e-03,  9.0931e-03,
        -1.3919e-04, -6.1407e-04,  3.9122e-03,  1.6290e-03,  2.6201e-04,
         2.3420e-02, -3.3192e-02,  1.1169e-03, -2.9948e-03,  3.6544e-03,
         4.1669e-03, -1.5971e-03,  4.1897e-03, -7.4411e-03, -3.4228e-03,
         6.1345e-03,  1.5315e-03, -1.0988e-03,  2.3529e-03,  3.7045e-03,
         4.2583e-03,  2.1909e-03, -3.4852e-02,  8.2395e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3166e-01,  6.3068e-01,  1.0025e-02,  3.2608e-03,  1.7561e-02,
        -8.4742e-03,  1.6998e-02,  4.0397e-03,  3.6663e-03, -1.4295e-02,
         9.4651e-03,  2.9293e-03,  5.8204e-03,  2.5216e-03,  3.1427e-03,
         3.2875e-03, -6.2752e-03, -7.6113e-03,  5.1373e-04,  6.4149e-03,
         7.5653e-03, -2.2761e-03, -1.3250e-03, -8.8123e-03, -7.1063e-03,
        -1.0696e-02,  5.5621e-02, -1.7958e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0795, -0.7242,  0.0276,  0.0328,  0.0150,  0.0151,  0.0023,  0.0046,
        -0.0133, -0.0095, -0.0163, -0.0345, -0.0254,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1364e-02,  8.0947e-01, -3.3766e-02,  1.9615e-02, -1.2614e-02,
        -2.8536e-03, -1.8561e-02,  2.1473e-04,  3.6954e-03,  1.4440e-03,
        -1.9883e-02, -1.5489e-02,  1.3712e-02,  2.5198e-02, -1.2126e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.5487e-02,  4.7856e-01, -4.2343e-02, -4.4684e-02, -2.1419e-02,
         6.7189e-03, -3.1862e-03, -4.6736e-03,  4.9283e-03, -6.6082e-04,
         1.5460e-03,  6.7718e-03,  9.2542e-03,  1.7620e-02,  3.4767e-03,
         3.0548e-03, -1.0632e-02, -2.6662e-02,  3.5970e-03,  6.2962e-03,
        -9.2605e-03,  2.7175e-03, -4.7259e-03,  3.4789e-03,  1.1687e-02,
         2.3797e-02, -1.9139e-04,  1.3974e-03, -6.8966e-03,  6.8795e-03,
        -7.8752e-03, -2.3608e-03,  1.3370e-02,  3.6215e-02,  2.8583e-03,
        -1.3042e-02, -9.3920e-03,  7.3169e-03, -4.4426e-03,  4.5165e-03,
        -1.4050e-03, -2.6951e-03,  1.9275e-03,  1.3724e-03,  2.9096e-03,
         1.1518e-02, -4.4181e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9843e-03,  6.9368e-01,  5.8484e-02, -1.7849e-03,  2.5311e-02,
         1.3266e-02, -7.4320e-03,  9.4868e-03,  1.4332e-02,  8.0072e-03,
         7.9771e-03,  4.4397e-04,  6.5720e-03, -2.0597e-03, -1.3469e-03,
        -8.1054e-03, -1.7832e-03, -4.0084e-03,  7.0372e-03,  4.4837e-03,
        -3.5986e-03, -7.9937e-03, -7.3111e-03, -5.8342e-03,  8.1885e-03,
         6.1359e-03,  4.7463e-03,  1.1426e-03, -7.0714e-03,  3.1060e-03,
        -7.3125e-03, -4.8108e-03, -1.9335e-04, -9.3898e-04, -1.3458e-03,
        -7.4817e-04,  4.5692e-03, -6.7142e-04, -3.0075e-03, -7.9162e-05,
         1.9705e-02, -2.2906e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3436e-02,  6.8607e-01,  4.6998e-02,  1.2610e-02,  1.4806e-02,
         1.5973e-02,  2.2388e-02,  9.5327e-03,  2.7137e-03,  2.0808e-02,
         2.7944e-02,  1.7680e-02,  2.7154e-03,  2.3850e-03,  8.6192e-03,
         7.1696e-03,  2.1402e-02, -1.4224e-03,  2.0659e-04,  4.0733e-03,
        -7.9623e-03, -4.9332e-03, -7.6527e-03,  2.2625e-03, -3.8239e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0993,  0.7419,  0.0261, -0.0019, -0.0067, -0.0067, -0.0064, -0.0040,
         0.0070, -0.0070, -0.0056,  0.0033, -0.0030, -0.0124, -0.0110, -0.0094,
         0.0069, -0.0025, -0.0017,  0.0126, -0.0101, -0.0020, -0.0125,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0324, -0.7318, -0.0342,  0.0158, -0.0147, -0.0010, -0.0052, -0.0133,
         0.0097, -0.0359, -0.0056,  0.0071, -0.0024,  0.0128, -0.0070,  0.0064,
         0.0028, -0.0094,  0.0057,  0.0107, -0.0026,  0.0051, -0.0285,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0041,  0.7444, -0.0550,  0.0075, -0.0167,  0.0176,  0.0351,  0.0150,
        -0.0056,  0.0054,  0.0274,  0.0410,  0.0252,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2046,  0.5694,  0.0230, -0.0375, -0.0148,  0.0025, -0.0208, -0.0241,
         0.0062,  0.0013,  0.0053, -0.0219, -0.0316, -0.0371,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0157, -0.7517, -0.0319, -0.0159, -0.0200,  0.0135, -0.0095,  0.0120,
        -0.0044,  0.0095, -0.0038,  0.0044, -0.0139, -0.0038, -0.0235, -0.0061,
         0.0341, -0.0128, -0.0135,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-8.6182e-02,  6.1908e-01, -9.3541e-02, -3.6335e-02,  5.9609e-03,
         2.5163e-02, -1.9717e-03,  1.7038e-02,  2.0193e-03,  7.6068e-03,
         3.3813e-03, -6.7353e-03,  4.6899e-03,  6.0343e-03,  6.7228e-03,
        -1.0980e-02,  4.0225e-03,  5.0673e-03,  6.7694e-03,  1.5942e-02,
        -2.2145e-03,  4.5797e-04, -2.3453e-03,  2.9399e-03,  2.6987e-03,
        -2.4105e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.9132e-03,  7.9963e-01,  3.7314e-02,  2.9639e-02,  2.5767e-02,
         1.2083e-02,  1.1634e-02, -5.5271e-03,  3.0476e-03, -1.0627e-02,
         1.8068e-02, -5.7631e-03, -6.4426e-04,  5.0506e-03, -2.1706e-02,
         6.5900e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1201,  0.5583, -0.0655,  0.0385,  0.0128,  0.0179,  0.0069,  0.0240,
        -0.0044, -0.0094, -0.0048, -0.0206,  0.0039,  0.0141,  0.0191,  0.0509,
         0.0289,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0131,  0.8540,  0.0147,  0.0156,  0.0182,  0.0147, -0.0145,  0.0334,
         0.0219,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8646e-02,  5.3616e-01,  3.9691e-02, -2.5516e-02,  1.5330e-02,
         2.6814e-02,  3.2433e-02,  1.0077e-02,  4.7828e-03,  5.3151e-03,
        -1.3412e-02, -2.0803e-02, -6.7828e-03,  6.9344e-04, -6.7697e-03,
         7.9065e-03,  1.4398e-02,  6.0117e-03,  9.1426e-03,  1.7370e-02,
         3.6328e-03,  3.8233e-03, -1.1480e-02,  1.7138e-04,  5.0211e-03,
        -6.1102e-03,  2.3196e-02, -9.4754e-04,  3.7203e-03, -3.7753e-04,
        -8.3498e-04,  1.1681e-02, -2.2084e-03,  1.4978e-03,  1.2653e-02,
        -4.2982e-03, -6.3591e-03,  9.3925e-03,  3.5149e-03, -2.6141e-03,
        -4.2848e-03, -8.5245e-04, -1.2787e-02,  2.7305e-03,  1.2532e-02,
         2.5228e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1499e-02, -6.7833e-01,  4.9576e-03,  2.9502e-02,  1.5322e-03,
         4.2241e-03, -2.5243e-02,  9.8205e-03,  2.7131e-02, -4.1584e-03,
         6.2555e-03, -2.5555e-04, -1.8848e-02, -1.9740e-03, -5.2312e-03,
        -8.7203e-03, -1.4757e-04, -4.4208e-02,  4.9115e-03, -5.0082e-03,
         1.1636e-02, -1.0821e-04,  1.4313e-02,  3.9129e-03,  1.6286e-03,
        -4.0689e-03,  4.7998e-04, -1.0849e-02,  1.9675e-02, -2.1370e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0032, -0.8669, -0.0420,  0.0153,  0.0196, -0.0330, -0.0199,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0091, -0.8219,  0.0273,  0.0450, -0.0084,  0.0029, -0.0048, -0.0222,
         0.0213, -0.0081,  0.0289,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0025,  0.7738,  0.0451,  0.0105, -0.0214,  0.0051, -0.0190, -0.0052,
        -0.0032, -0.0093, -0.0100, -0.0073,  0.0116, -0.0760,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0011,  0.7300,  0.0209, -0.0155,  0.0024, -0.0055, -0.0123, -0.0185,
        -0.0030, -0.0126,  0.0040,  0.0076, -0.0116, -0.0023, -0.0073, -0.0069,
         0.0032,  0.0014, -0.0062,  0.0152,  0.0063,  0.0053, -0.0095, -0.0057,
        -0.0023,  0.0042, -0.0056,  0.0038, -0.0063, -0.0062, -0.0092,  0.0109,
        -0.0363, -0.0008,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4014e-02,  6.3311e-01,  2.7244e-02, -6.0648e-02, -1.0625e-02,
         1.0713e-02,  1.6421e-02, -1.4633e-02,  3.4389e-02,  1.9391e-02,
        -8.7892e-03, -1.3882e-02, -5.1197e-03,  6.1958e-03, -1.8661e-02,
        -8.4592e-03, -2.9761e-02, -1.2893e-02, -6.8903e-03,  3.6376e-03,
        -5.4759e-03, -5.4689e-03,  1.1155e-03, -1.1964e-04, -2.1208e-03,
         1.0162e-03,  1.9720e-03,  2.7231e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0506,  0.7217,  0.0603, -0.0074,  0.0100,  0.0074, -0.0053,  0.0026,
        -0.0055, -0.0059,  0.0047, -0.0066,  0.0088,  0.0017,  0.0139, -0.0042,
        -0.0044, -0.0172,  0.0153,  0.0007, -0.0019, -0.0017, -0.0053, -0.0367,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([-9.8877e-03, -7.7356e-01,  2.1145e-02,  7.9418e-04,  1.8174e-02,
         2.0493e-03, -9.9491e-04,  1.1273e-02,  3.9009e-03,  3.5846e-04,
         2.9857e-03,  5.1287e-03, -9.8844e-03,  6.6828e-04, -4.4121e-03,
        -7.6149e-03, -5.4582e-03,  8.3180e-03,  8.8778e-03, -4.0493e-03,
        -1.0462e-02, -1.7687e-03,  6.3618e-03, -2.3733e-03, -4.6172e-03,
        -1.3619e-04,  6.8569e-03,  3.9516e-03,  2.9726e-03,  2.9727e-02,
        -3.1234e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1016, -0.7536,  0.0488,  0.0047, -0.0128, -0.0031, -0.0062,  0.0052,
         0.0147,  0.0139, -0.0092, -0.0262,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0218, -0.7396, -0.0071,  0.0414, -0.0232,  0.0175,  0.0288,  0.0055,
         0.0018,  0.0057,  0.0097,  0.0122, -0.0407, -0.0088, -0.0362,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.8374e-02,  7.2463e-01, -8.8386e-04, -2.3195e-03, -1.8216e-03,
         1.7204e-02,  1.1969e-02, -8.2125e-03, -3.5414e-03,  1.5756e-02,
        -1.3907e-02, -4.3475e-03, -5.9381e-03,  1.2690e-03,  6.9723e-03,
         8.0761e-03,  3.2524e-03, -1.4046e-02,  1.5883e-03,  1.2136e-02,
        -9.8815e-05, -3.0377e-03, -6.6325e-03, -1.2675e-03,  3.8165e-03,
         2.2536e-03,  4.2225e-03, -4.0875e-03, -9.1633e-03,  4.7820e-03,
         1.2270e-02,  8.6675e-04, -3.1259e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0389, -0.7468, -0.0141,  0.0259, -0.0166,  0.0036,  0.0127,  0.0106,
        -0.0149,  0.0073,  0.0108, -0.0029,  0.0285,  0.0351, -0.0076,  0.0238,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1087,  0.6258, -0.0942, -0.0518,  0.0278,  0.0294,  0.0222,  0.0223,
         0.0178,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0711,  0.8215,  0.0260,  0.0505, -0.0310,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2545e-01,  6.5931e-01,  3.7256e-02, -1.1763e-02,  1.1488e-02,
         1.7190e-02, -4.3562e-03, -1.2974e-02,  5.2913e-03, -7.1925e-03,
         1.4487e-02,  2.3004e-02,  6.0347e-03,  1.4160e-02, -4.9361e-04,
        -8.7622e-05,  1.7296e-03, -2.0136e-03, -1.5955e-02, -1.7940e-03,
        -4.1087e-04, -5.7343e-04,  1.9552e-03, -1.7084e-03,  1.8402e-03,
        -1.5816e-02,  5.6615e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0853,  0.6890, -0.0053,  0.0411,  0.0857,  0.0865,  0.0070,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0738,  0.7077, -0.0389,  0.0114, -0.0058, -0.0038,  0.0138, -0.0077,
        -0.0014, -0.0019, -0.0028,  0.0059,  0.0032, -0.0096, -0.0096, -0.0042,
        -0.0042,  0.0091,  0.0114, -0.0273, -0.0218,  0.0248,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5033e-02,  7.4745e-01,  3.3521e-02,  2.1026e-02,  2.4778e-02,
         4.7944e-03,  1.4995e-04, -1.1261e-03,  4.9890e-03, -1.6999e-02,
        -3.0620e-03,  4.0876e-03, -4.1072e-03,  1.1462e-02, -2.7466e-03,
        -6.0457e-03, -1.5209e-02, -1.2925e-02,  4.0487e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9151e-02, -7.8674e-01, -1.4631e-02,  6.6792e-03, -5.1345e-03,
        -1.5616e-04,  7.6341e-03,  8.3764e-04,  4.6039e-03,  5.6372e-03,
         1.0825e-02, -2.9371e-03,  2.7490e-03, -1.1815e-03, -3.9161e-03,
         3.1401e-03, -1.1265e-02,  2.0077e-03,  4.7074e-03,  1.9432e-02,
         3.6752e-03, -1.7698e-02,  2.0015e-03, -7.4546e-04, -1.2307e-02,
        -1.5267e-03,  6.1223e-03, -1.4541e-04, -9.0209e-04,  2.0726e-02,
         2.0788e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 1.0807e-01, -7.2129e-01,  5.4455e-03,  2.3839e-03, -1.1909e-02,
         1.8632e-03,  6.6746e-03,  4.0662e-03,  1.7137e-03, -3.6734e-03,
         3.0436e-03,  5.5845e-03,  2.6565e-02, -1.1333e-02, -9.1773e-04,
         9.4437e-04, -3.7575e-03, -4.0440e-05,  2.2402e-02, -1.5957e-03,
         1.0084e-02,  2.1710e-03, -7.6293e-05, -1.0176e-02, -3.3963e-03,
        -2.7241e-03,  7.4044e-03,  1.2811e-02,  1.2712e-03, -1.1390e-03,
         5.2930e-03,  1.8567e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1632e-02,  7.9883e-01,  3.2141e-03,  1.3308e-02, -5.7171e-03,
        -7.4319e-03,  5.0212e-04,  1.9019e-03,  2.5828e-02,  1.7229e-02,
        -2.0791e-02, -6.1879e-03, -2.9338e-03,  1.3873e-02, -1.4060e-02,
        -9.4226e-03,  7.3338e-03, -1.3545e-03, -5.8106e-04,  3.4229e-03,
        -6.1255e-03, -8.3222e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3732e-02,  7.2599e-01,  5.2358e-03, -6.1107e-03, -1.0340e-02,
        -1.8069e-02, -4.8073e-02,  3.1940e-02,  6.1113e-04,  4.6650e-03,
        -1.5534e-02,  2.0563e-03,  1.4999e-03,  3.3724e-03, -1.1633e-02,
        -1.1939e-02, -1.9947e-03,  1.3380e-02, -1.4612e-03, -8.1547e-03,
         2.4246e-03,  6.8994e-03, -2.4837e-03, -1.7101e-03, -5.3006e-03,
        -7.1618e-03, -6.1855e-03, -1.2371e-02, -8.4555e-03,  1.1215e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1350e-01,  7.2843e-01,  3.4284e-02, -6.0744e-03,  6.3616e-03,
        -8.6009e-03, -1.3036e-02, -3.3210e-03,  2.9816e-03,  1.1750e-03,
        -1.3609e-02, -1.2068e-02, -1.2733e-02,  1.2817e-02, -7.0110e-03,
        -1.6260e-02, -2.8488e-05,  7.7049e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9278e-03, -7.0345e-01, -1.0559e-02, -3.8374e-02,  1.0926e-02,
         2.0912e-02,  5.0207e-04, -1.2964e-02,  1.5400e-02,  1.2721e-03,
         6.2614e-03, -3.0332e-03,  2.9733e-03, -7.5821e-03, -3.6358e-03,
         1.2313e-02, -1.8758e-02,  1.3752e-03,  1.1922e-02,  9.8629e-03,
         3.7135e-03,  5.1649e-03,  4.5962e-03, -1.0209e-03, -5.4009e-03,
         3.5128e-03,  9.7148e-03, -2.4163e-03, -2.6120e-03, -3.2759e-03,
         3.9538e-03,  5.4420e-03, -9.3456e-03,  4.7098e-03,  3.7051e-03,
        -1.0988e-02, -2.5421e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4955e-02,  6.2406e-01, -1.1533e-01, -6.5038e-02, -2.3442e-02,
        -8.5421e-03, -1.0105e-02, -1.7112e-02,  2.7201e-03, -2.6360e-04,
         2.3889e-03,  2.2013e-02,  5.8244e-03, -2.2711e-04, -3.7326e-03,
         3.5846e-03,  6.8972e-03,  5.8816e-03,  1.9521e-04,  1.1807e-02,
         9.5186e-03, -1.0902e-02, -1.6464e-02, -9.0045e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8482e-02,  7.5935e-01, -4.1019e-03, -1.4715e-02,  3.7366e-03,
         4.4186e-03,  1.8881e-03,  1.5649e-03, -8.8112e-03,  3.4634e-04,
         4.5225e-03, -2.6989e-03, -2.8607e-04,  4.5099e-03,  7.0865e-03,
        -5.5962e-04, -6.8254e-03,  3.0553e-03,  1.7231e-03, -6.5847e-03,
         8.5144e-04,  4.6710e-03, -2.3086e-03, -5.7349e-03,  1.3267e-02,
         1.9424e-02,  3.0666e-03,  7.9091e-03,  7.5260e-03,  1.2933e-04,
         1.1244e-02,  5.2478e-04,  3.7872e-04,  4.2876e-03,  6.1909e-03,
        -1.0486e-02, -4.6745e-03, -5.6544e-03, -2.3939e-03,  4.8216e-03,
        -4.5910e-04,  3.4490e-03, -8.4811e-04,  1.1802e-02,  1.6816e-04,
        -2.4621e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0178,  0.7011,  0.0100,  0.0074, -0.0396,  0.0156, -0.0311,  0.0070,
        -0.0088, -0.0358,  0.0084, -0.0015,  0.0126,  0.0048,  0.0114, -0.0031,
        -0.0037, -0.0295,  0.0013,  0.0133,  0.0276,  0.0012,  0.0074,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0625, -0.7310,  0.0707, -0.0543, -0.0121, -0.0164,  0.0048,  0.0122,
         0.0115,  0.0019,  0.0227,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0093,  0.7184,  0.0338, -0.0297, -0.0008, -0.0238, -0.0080,  0.0096,
         0.0156,  0.0393,  0.0143,  0.0059,  0.0725,  0.0190,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0180,  0.8390,  0.0075, -0.0224, -0.0273,  0.0158, -0.0153, -0.0018,
        -0.0143,  0.0122,  0.0165,  0.0099,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2011e-02,  6.9563e-01,  4.9992e-02,  6.4091e-03,  1.4634e-02,
        -4.9296e-03, -3.4694e-03,  1.2257e-02,  3.2456e-03,  6.0825e-03,
         3.1297e-03,  1.2655e-03,  8.0007e-03,  7.0727e-03, -6.5348e-04,
         5.2838e-03, -5.5778e-04,  7.9634e-03,  1.3849e-03, -1.1641e-02,
         9.8574e-04,  1.1526e-03,  5.5167e-03, -5.6796e-04,  4.6240e-04,
         8.0888e-03,  1.2735e-02,  1.0734e-03, -1.4913e-03, -3.0505e-03,
        -4.6431e-03,  8.0158e-03, -3.2879e-03,  5.3029e-03, -5.0170e-04,
        -8.9716e-05,  1.3348e-04, -2.7164e-03, -1.6852e-03, -9.0270e-04,
         3.1935e-03,  3.3175e-03,  3.3839e-03,  4.7251e-04, -1.8239e-04,
         2.2722e-03,  2.9559e-03,  4.4974e-04,  1.4438e-03, -1.0956e-03,
        -7.2487e-04, -1.1529e-03, -3.4512e-03,  8.5399e-03,  2.9351e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-0.0358, -0.7901,  0.0009,  0.0111,  0.0022, -0.0016, -0.0127,  0.0023,
        -0.0086, -0.0027,  0.0139, -0.0097,  0.0014, -0.0031,  0.0050, -0.0054,
         0.0044, -0.0206, -0.0063, -0.0142,  0.0123, -0.0061,  0.0296,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2783e-01, -6.6137e-01, -1.2989e-02, -5.2588e-02, -1.6812e-02,
        -1.7655e-02, -9.2298e-03, -6.8913e-03,  5.4978e-06, -1.1902e-02,
         2.3294e-03, -9.0672e-03,  3.5752e-03,  1.3322e-02,  1.2833e-02,
        -2.5243e-02, -8.9351e-03,  7.4314e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0787, -0.6421, -0.0179, -0.0263, -0.0100, -0.0081, -0.0488, -0.0130,
         0.0048,  0.0137,  0.0048, -0.0085, -0.0494,  0.0063,  0.0028, -0.0094,
        -0.0031, -0.0522,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8712e-02,  7.4025e-01, -2.7902e-04, -2.3510e-02,  2.3117e-02,
        -2.5349e-02,  4.4210e-03,  1.3789e-02, -3.0356e-02,  4.0333e-03,
         7.5798e-03,  3.3307e-03,  2.2703e-02,  1.0641e-03, -7.1503e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.9974e-02,  6.1824e-01,  2.1600e-02,  2.9546e-03, -4.8536e-02,
         7.4786e-03, -1.2842e-02, -1.8375e-02, -1.7610e-02,  3.9011e-03,
        -2.2647e-03, -3.9165e-03,  3.6351e-03, -2.4786e-03,  9.5070e-04,
         1.9746e-02, -8.3184e-03, -7.6683e-03, -4.6039e-03, -3.0472e-02,
         4.3671e-04, -1.9843e-02, -3.4619e-03, -7.2454e-03, -6.5432e-03,
         1.6360e-03,  1.1364e-02, -8.6948e-04,  2.6291e-02,  1.4823e-02,
        -1.3876e-03, -5.2955e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2930e-02, -7.1401e-01, -2.7387e-02, -2.2749e-02, -6.8941e-03,
        -4.5738e-03,  4.0650e-03,  7.6442e-03,  3.1188e-04,  7.5885e-03,
         1.8243e-03,  7.5565e-03, -7.9244e-03,  1.1114e-02, -1.2408e-03,
         9.1537e-03,  1.4661e-03, -6.8693e-04,  2.0791e-03, -1.3728e-03,
         4.7963e-03, -2.2014e-03, -5.8299e-03,  4.0282e-04,  1.9355e-03,
        -1.1704e-02, -6.9691e-03, -8.4893e-03,  1.2428e-02,  9.0184e-03,
         1.5211e-02,  5.0046e-03, -5.3438e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5842e-02,  6.5981e-01,  2.1493e-02,  3.5754e-02,  3.1496e-02,
         1.7716e-02,  1.5329e-03,  6.5329e-03,  2.5062e-03,  1.5232e-02,
         5.5160e-03, -5.4927e-03, -5.3563e-04,  3.9782e-03,  1.8200e-03,
        -6.4361e-03,  9.6790e-03, -9.3705e-06,  5.5068e-03,  4.0576e-03,
        -1.6291e-03,  1.0094e-02,  1.0772e-03, -1.5281e-03,  2.2222e-03,
         7.5159e-05,  1.0705e-03,  1.2966e-04,  8.7953e-03,  9.0904e-03,
         8.2314e-03,  8.1042e-03,  2.7398e-03,  1.0833e-03,  1.7108e-03,
         7.3725e-05,  3.4953e-03,  5.9283e-04,  4.2820e-03,  9.7817e-03,
         1.1112e-03,  2.1642e-03,  9.2020e-05,  2.5861e-03,  6.1450e-03,
         2.6074e-03,  3.2572e-03,  2.3189e-03,  2.9767e-03,  8.2896e-03,
         2.6572e-05,  1.9892e-03,  4.4542e-03,  1.3640e-04,  7.0730e-03,
         1.2456e-02,  1.6476e-03,  3.9169e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4348e-02, -7.3520e-01, -3.3676e-02,  2.0142e-02, -2.1602e-02,
        -1.4148e-02,  1.0697e-03,  1.1130e-02, -6.7318e-04, -2.0722e-02,
        -6.2655e-03, -4.2752e-03, -3.9182e-03, -1.3394e-02,  4.4395e-04,
        -1.2301e-03, -8.6992e-03,  8.1969e-03,  6.1603e-04,  1.1295e-03,
         2.0772e-03, -7.7324e-03,  9.2967e-03,  5.6586e-03,  1.3660e-02,
        -6.9928e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0386e-02,  8.1408e-01, -1.8157e-02, -3.7796e-03,  3.9708e-03,
         9.7419e-04,  5.1591e-04, -8.1831e-03, -1.0552e-02, -9.4357e-03,
        -1.5115e-02,  1.4987e-03,  3.0218e-03, -2.6037e-03,  5.8856e-04,
        -1.3774e-03, -8.6282e-03,  1.5380e-03,  6.3929e-03,  7.5001e-04,
        -2.2376e-03, -1.9312e-02,  1.6123e-02, -6.7001e-03, -1.4079e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.0016e-02,  6.2928e-01,  4.2405e-02, -1.3510e-02,  2.0625e-02,
         1.9486e-03,  1.1832e-03,  4.5354e-04,  2.1600e-02,  1.5979e-02,
        -2.1429e-02, -8.9974e-03, -1.2083e-02, -7.1253e-03,  3.7535e-03,
        -4.5358e-03, -4.1067e-03, -1.3318e-03,  9.3514e-04, -9.4616e-04,
        -7.6159e-03,  4.0045e-03, -4.6366e-03, -2.9927e-03,  9.4423e-03,
        -1.2523e-03, -1.1108e-03, -2.6672e-03, -9.1266e-04,  2.3631e-03,
        -5.0879e-03, -4.0000e-03,  1.0056e-02,  5.6226e-03,  1.1364e-02,
         1.9746e-03,  2.8340e-03,  5.0579e-03,  3.5697e-03,  7.2374e-03,
        -1.1035e-02, -1.6042e-03,  8.7043e-03,  3.0942e-02,  1.6639e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0607,  0.7109, -0.0178,  0.0220, -0.0162,  0.0059, -0.0111, -0.0182,
        -0.0038, -0.0054, -0.0143, -0.0048,  0.0027,  0.0046,  0.0027,  0.0087,
         0.0038,  0.0036, -0.0012,  0.0030, -0.0150,  0.0459, -0.0176,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0146,  0.7211, -0.0928, -0.0328,  0.0096,  0.0153, -0.0028,  0.0072,
         0.0036, -0.0144, -0.0432,  0.0427,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 2.9660e-02,  7.4254e-01,  2.9651e-02, -1.6516e-02, -1.3627e-02,
         7.0994e-03, -1.4841e-02,  2.3944e-04,  1.1556e-02, -1.3802e-02,
         8.3523e-04,  6.4873e-03,  6.5838e-03,  9.1152e-03, -1.6399e-02,
        -1.2465e-02, -1.9273e-03,  5.7047e-03, -1.3173e-03,  3.5760e-03,
         2.1079e-03, -1.3737e-02, -4.0215e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.4129e-03, -7.3438e-01, -3.4752e-02,  1.0344e-02,  1.9144e-02,
         5.3017e-03,  9.5287e-03, -1.0368e-02,  1.3825e-02, -9.7404e-04,
         4.0765e-03,  9.8817e-03,  3.6874e-03, -1.6464e-03, -3.9314e-03,
         9.3515e-03,  5.7062e-03,  2.0811e-02,  5.0119e-03, -5.3400e-03,
         6.5830e-03,  1.3774e-02,  9.1202e-03,  1.4974e-02,  3.6194e-03,
        -4.6150e-04,  6.7998e-03, -2.0276e-03,  9.9281e-03,  3.2983e-04,
         3.3449e-03,  3.8392e-04, -1.2183e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3735e-02, -7.9946e-01, -4.4380e-02, -1.2705e-03,  2.6757e-02,
        -1.0455e-03,  1.1926e-02, -1.5839e-03, -9.1732e-03,  4.7777e-03,
         1.0840e-02,  3.6190e-03, -1.0554e-02,  1.5378e-03, -3.8192e-03,
         5.3629e-03, -2.4558e-03, -2.8094e-04,  3.7421e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0448e-02, -7.7897e-01,  7.6192e-03, -5.8338e-03, -1.7872e-02,
        -3.5521e-02,  4.7445e-02, -2.2312e-04,  1.9746e-02,  6.4672e-03,
        -3.2277e-02,  2.9976e-03,  3.4328e-03, -1.1496e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5384e-02, -7.9309e-01, -1.4757e-02,  1.7914e-02, -1.0357e-02,
        -6.3204e-03, -5.6246e-03,  8.5874e-03, -2.3065e-03, -1.5043e-02,
        -6.9712e-03, -2.8579e-03, -1.7194e-03,  2.4654e-02, -5.4516e-03,
        -4.0112e-03,  1.6256e-02, -2.9774e-04, -1.0114e-02,  2.9024e-03,
         1.4757e-02, -2.0624e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4322e-02, -7.3291e-01, -3.2210e-02,  3.0720e-02, -2.7248e-03,
        -1.0453e-02,  1.6158e-02,  1.4453e-02,  1.4580e-02, -2.6082e-04,
         1.8830e-03, -7.8200e-04,  1.7551e-02,  2.6801e-03,  2.4650e-03,
        -4.0320e-03, -2.5341e-02,  7.4368e-04, -1.1191e-03, -7.3370e-03,
        -2.8541e-03, -3.9515e-03,  5.7567e-03, -4.2922e-03,  6.6199e-03,
         7.5604e-04, -2.3468e-03, -7.5822e-03, -5.4661e-03,  7.6499e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1133,  0.7067, -0.0015, -0.0035, -0.0054, -0.0164,  0.0081, -0.0022,
        -0.0087, -0.0009, -0.0121, -0.0271, -0.0232,  0.0089,  0.0121, -0.0228,
        -0.0269,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0405,  0.6849,  0.0217,  0.0261,  0.0008,  0.0026,  0.0024,  0.0851,
        -0.0612, -0.0150, -0.0596,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0323,  0.7512,  0.0359,  0.0042, -0.0300,  0.0179,  0.0132,  0.0095,
         0.0016,  0.0172, -0.0018,  0.0038,  0.0032,  0.0213,  0.0037, -0.0032,
        -0.0502,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8693e-02,  7.0016e-01,  2.0343e-02,  4.7566e-02, -9.6404e-03,
        -1.4068e-02, -3.6751e-03,  1.8714e-02, -1.7529e-02, -4.2350e-04,
        -5.6343e-04,  4.6022e-03, -3.6870e-03,  1.6688e-02,  6.2628e-03,
        -5.0939e-04,  1.3558e-04, -1.8405e-03, -1.7764e-02, -1.1614e-03,
        -1.9129e-03, -1.3256e-02,  1.4059e-02,  8.1867e-03,  3.1238e-03,
        -3.5431e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2112e-02,  7.0229e-01, -5.2622e-02, -3.1173e-02, -1.1587e-02,
        -1.6037e-02, -7.9976e-03,  1.1236e-02,  3.1735e-03,  2.4792e-03,
        -4.2909e-03,  7.3350e-03, -2.6404e-03, -5.1405e-03,  6.6171e-04,
         1.8166e-02, -4.6519e-03,  2.2285e-03, -2.5184e-03,  7.9414e-03,
         6.0847e-03,  4.4193e-03, -3.1426e-04,  3.9993e-04,  1.2048e-02,
        -7.8530e-04, -3.9275e-03, -2.6860e-03,  4.5280e-03,  2.8525e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0057, -0.7866,  0.0411,  0.0412, -0.0010, -0.0166,  0.0139,  0.0052,
         0.0092,  0.0190,  0.0056, -0.0031, -0.0122,  0.0015,  0.0008,  0.0098,
         0.0172, -0.0023,  0.0079,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 1.2680e-01, -6.5703e-01,  3.7376e-04, -4.9541e-02,  1.5733e-03,
         9.5283e-03, -2.8656e-02,  2.1957e-02,  3.3776e-03, -6.9184e-03,
        -1.1614e-02, -4.5818e-03, -1.9915e-03,  1.4064e-04,  1.8934e-02,
        -4.2015e-03, -4.6345e-03, -2.0630e-02, -6.8828e-03, -3.1565e-03,
         1.5172e-04,  3.0428e-03, -2.4849e-03, -1.1857e-04,  8.1318e-03,
         2.6460e-03, -9.0899e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0230, -0.7441,  0.0243,  0.0477,  0.0075,  0.0182, -0.0330, -0.0112,
        -0.0081, -0.0283,  0.0062, -0.0292, -0.0097,  0.0093,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.5808e-02, -7.0767e-01, -3.6601e-02,  6.6123e-03, -1.7491e-02,
        -2.5955e-03,  9.9961e-03,  3.8095e-02, -9.5541e-03, -2.1230e-03,
         1.0193e-02,  2.0803e-03, -6.3777e-03, -7.3441e-03, -1.2135e-02,
        -5.5359e-03,  8.8703e-03, -8.6523e-04,  6.8691e-03,  8.1424e-03,
         2.3672e-06, -2.0190e-03,  4.8939e-03, -6.3448e-04, -3.0873e-02,
         6.6163e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0631,  0.6298,  0.0422,  0.0421,  0.0248, -0.0186,  0.0180,  0.0057,
        -0.0196,  0.0017,  0.0026, -0.0229,  0.0432, -0.0657,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0122e-02,  6.9164e-01,  1.4028e-02, -2.2806e-02, -3.2476e-03,
        -9.5555e-03, -1.9254e-02, -1.2195e-02,  3.3736e-03, -4.5838e-03,
        -2.7748e-03, -6.4420e-03, -2.3683e-03, -1.5023e-04,  6.7071e-03,
         6.4461e-03, -2.0803e-02, -1.3125e-02,  2.7157e-03, -1.1407e-02,
        -5.9604e-04, -6.1337e-03,  1.4399e-02,  1.1312e-02, -2.6648e-03,
        -6.9367e-03, -1.3168e-02, -8.0402e-03, -9.2457e-03,  4.5728e-03,
         1.0367e-02,  1.8822e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6871e-02,  7.4527e-01,  2.1575e-02, -2.3423e-02, -2.9077e-02,
        -7.6237e-03, -7.1262e-03, -4.3267e-03, -1.7545e-04, -7.9933e-03,
        -1.6635e-03,  1.7209e-03, -6.0886e-03,  2.4848e-03, -8.6131e-03,
        -3.8517e-03,  4.2997e-03, -9.0350e-03,  5.2546e-03,  5.6958e-03,
        -3.1019e-04, -2.0262e-03, -4.9152e-03, -1.6829e-02, -2.3750e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6209e-02,  7.9461e-01, -9.1304e-05,  4.6700e-02,  9.3439e-02,
         3.8948e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5591e-02, -6.6871e-01, -5.0691e-02,  6.4974e-02, -1.4820e-02,
         2.4859e-02,  1.6386e-02,  2.1655e-03, -2.8619e-03,  9.1512e-03,
         8.0773e-03, -1.5793e-02,  1.4403e-03, -4.7841e-03,  7.2168e-04,
         3.5998e-03,  5.3218e-03, -7.9148e-04,  2.0119e-03,  2.8561e-04,
         1.1785e-03, -4.8414e-03,  4.7705e-03,  2.1939e-03, -8.6768e-04,
         3.6446e-05,  2.2261e-03,  3.9337e-04,  2.5473e-03,  2.8770e-03,
        -1.5157e-03,  1.2465e-04, -1.9172e-03, -1.5966e-04, -8.8483e-03,
        -1.9559e-02, -2.9120e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6432e-02, -6.9698e-01, -2.8825e-02, -4.9800e-02,  2.1015e-03,
        -1.8384e-02, -1.1219e-02,  1.5607e-02, -3.3305e-03,  1.4372e-03,
         3.4488e-04, -3.0635e-03, -9.5198e-03, -1.2374e-03, -1.7042e-03,
         1.3062e-02,  6.2484e-03, -8.4507e-03,  8.3717e-04, -4.1901e-03,
         1.8326e-02,  7.9188e-03,  1.6474e-03,  3.7848e-04,  2.7267e-03,
         5.7807e-03, -1.3488e-03, -5.9836e-03,  7.8005e-03,  2.7310e-03,
         4.5504e-03, -1.9168e-04, -2.1713e-03,  2.4788e-03,  1.3194e-02,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6873e-02,  6.4062e-01, -2.0446e-01,  1.6942e-02, -1.8336e-02,
         2.5032e-02, -1.1549e-03, -2.2441e-03,  8.9553e-03, -5.3825e-03,
        -1.9323e-03,  3.7137e-04, -2.2335e-02, -1.3006e-02, -2.2352e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9384e-02, -7.5594e-01, -2.8595e-02,  4.1196e-03, -5.0269e-03,
        -6.7463e-04, -1.8333e-02, -8.4506e-03, -4.2564e-03,  7.5619e-03,
         5.0702e-03,  7.0350e-03, -2.5613e-03,  5.6354e-03, -1.5038e-03,
        -1.3274e-02, -6.3620e-03, -6.5353e-03, -1.3407e-03, -6.7200e-04,
        -4.0913e-04,  9.5202e-03,  2.8269e-03,  3.3437e-03,  1.0051e-02,
         4.3207e-03,  9.3387e-04, -3.7068e-03,  8.5794e-04, -8.1957e-04,
         2.0326e-03, -2.3388e-02, -5.4548e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4747e-02, -6.7595e-01,  2.3967e-02, -3.7851e-02,  6.7256e-02,
         5.0929e-02,  8.2583e-03, -5.8543e-03,  1.2764e-03,  1.0093e-02,
        -3.1052e-04, -2.7072e-02,  1.6667e-03, -8.2973e-03, -1.0745e-02,
        -6.3230e-03, -5.1555e-03, -3.1642e-03, -3.4387e-04, -2.3952e-04,
         7.1384e-03, -1.6484e-03,  2.0657e-03, -6.9772e-03,  7.4738e-04,
         4.1144e-03,  4.6942e-03, -8.3960e-04, -1.2272e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 2.8668e-02, -8.2485e-01,  2.9566e-02,  3.8578e-02,  1.1816e-02,
         1.3443e-02,  9.5607e-06,  4.9051e-03, -5.5726e-03,  4.1171e-03,
         1.0375e-03, -4.8564e-04,  9.5752e-03,  3.5297e-03,  2.3841e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7373e-03, -7.3649e-01, -2.8141e-02,  3.9173e-02,  1.4225e-02,
         4.3015e-03, -5.4661e-04, -1.5733e-03, -9.4992e-03,  4.9931e-03,
         5.3772e-03, -8.4366e-03,  5.9050e-03, -6.3268e-03,  2.8118e-03,
        -3.3506e-03, -2.2198e-03, -1.8269e-03,  1.8128e-04,  3.9506e-03,
        -6.7340e-05, -2.6374e-03, -1.0464e-03, -2.8713e-03, -1.8298e-03,
        -4.3902e-03, -4.9114e-03, -6.8104e-03,  1.4528e-02,  2.6251e-04,
         1.4035e-03, -1.3424e-04, -2.8513e-03, -2.2806e-03,  9.7237e-03,
        -1.7381e-03, -3.1435e-03, -1.8383e-03, -2.2682e-03,  4.3780e-03,
         4.2847e-04,  3.9699e-03, -6.0990e-03,  1.8374e-03, -4.7963e-03,
         1.8751e-03, -6.5355e-03,  8.8406e-03,  1.4434e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4676e-02,  7.3821e-01, -1.7027e-02,  1.0165e-02,  1.1181e-02,
         1.0186e-02, -1.2404e-03,  7.0135e-04, -8.2211e-04,  1.3989e-02,
         5.8189e-03, -6.1883e-04,  7.5095e-03, -3.9993e-04,  4.4308e-03,
        -4.1637e-03,  4.2348e-03,  4.8520e-03,  3.8402e-03, -3.3651e-04,
        -3.9747e-03, -4.4357e-03, -3.4508e-03,  1.5967e-03,  6.8221e-04,
         5.0688e-04, -1.2259e-02,  7.7813e-03,  1.6036e-03,  9.0428e-04,
         6.2661e-04, -5.6675e-03, -9.9245e-04,  4.4353e-02,  1.6758e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6981e-02, -7.6322e-01, -1.6796e-02, -4.4887e-02,  3.3109e-04,
        -8.3890e-03, -7.1086e-03,  7.1801e-03, -1.4993e-02,  5.4060e-03,
        -9.8084e-03, -5.1849e-03,  6.5865e-03, -6.7540e-03, -1.9044e-02,
        -5.7221e-04, -2.3294e-03,  1.2106e-02, -2.2326e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0380, -0.7746,  0.0096, -0.0231, -0.0201, -0.0025, -0.0184, -0.0126,
        -0.0377, -0.0045,  0.0100, -0.0085, -0.0105,  0.0237,  0.0061,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0021, -0.7523,  0.0102,  0.0073, -0.0112,  0.0089,  0.0161,  0.0137,
        -0.0024,  0.0083,  0.0219,  0.0111,  0.0172,  0.0152, -0.0186,  0.0269,
         0.0020,  0.0049, -0.0019, -0.0175,  0.0096, -0.0208,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6477e-02,  7.7979e-01, -1.9884e-02,  2.3398e-03,  1.4988e-02,
        -1.1821e-02, -2.8599e-02,  5.4065e-03, -6.2591e-03,  3.1872e-03,
        -6.5955e-03,  6.7283e-03, -3.2144e-03, -4.9504e-03, -9.1337e-03,
        -2.0700e-03, -5.3770e-03, -1.4442e-03, -1.1441e-03, -5.0561e-03,
         1.1505e-02, -7.1857e-04, -1.9137e-03, -1.4451e-03, -1.1531e-02,
        -5.0255e-03, -1.5109e-03, -6.9632e-03, -4.8700e-03,  2.0049e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0254,  0.7387, -0.0740, -0.0304, -0.0089, -0.0268, -0.0396,  0.0059,
         0.0097,  0.0223,  0.0183,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7673e-03,  8.3838e-01,  1.2273e-02,  1.5345e-02,  7.8280e-03,
        -1.5676e-02, -1.0221e-02, -2.1917e-03,  1.3674e-03,  2.0486e-03,
        -1.5379e-02, -5.9159e-04,  1.2978e-02,  4.4448e-03, -3.1648e-03,
        -2.2711e-03, -7.9508e-03, -4.0411e-03, -4.8380e-04, -1.4415e-02,
         1.7711e-02, -9.4715e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3076e-02,  7.6478e-01,  3.1881e-02,  2.4569e-02, -3.1012e-03,
         1.4553e-02,  3.3848e-03,  3.2874e-03, -5.4654e-03, -2.1807e-03,
         1.2122e-02, -1.1594e-02, -6.2546e-03, -8.1599e-03, -2.8179e-03,
         4.5209e-03, -4.7996e-03,  2.4696e-04,  3.7958e-03, -7.1911e-03,
        -2.9761e-04,  1.6256e-03,  7.5131e-03, -1.9727e-03, -8.2192e-03,
         6.9466e-04,  6.6545e-03,  1.6002e-03,  6.8994e-03, -2.2420e-03,
         4.1598e-04,  2.6830e-03,  1.4027e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3817e-04, -7.2150e-01, -1.0303e-01,  2.0752e-02,  8.6495e-03,
        -4.0517e-03,  1.4669e-02,  3.4070e-03, -8.6032e-03,  7.1085e-03,
        -4.0014e-04, -2.1921e-03,  9.3743e-04, -4.0290e-03,  1.3572e-02,
        -1.1561e-02,  3.4879e-03,  8.9669e-04, -1.0318e-04,  5.2118e-03,
        -3.1991e-03, -2.3755e-03, -7.7854e-03, -3.4313e-03,  4.9640e-03,
        -5.2095e-03,  1.4110e-02,  2.4621e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0512,  0.7303,  0.0666,  0.0325,  0.0067, -0.0042,  0.0052, -0.0057,
        -0.0181,  0.0032, -0.0109,  0.0068,  0.0011,  0.0055, -0.0072, -0.0036,
        -0.0057, -0.0044, -0.0086,  0.0025, -0.0008,  0.0019, -0.0035, -0.0065,
        -0.0075,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 1.6376e-02, -7.3928e-01, -2.2716e-02,  7.6267e-03,  4.0464e-03,
        -1.2212e-02, -1.6833e-02,  7.2336e-03, -8.3737e-03, -7.6799e-04,
         3.9099e-03, -1.0849e-02, -2.4338e-03,  4.7548e-03, -8.9340e-04,
         2.2615e-02,  1.1133e-02,  3.8522e-03, -2.1891e-03, -1.3247e-04,
        -6.0590e-04,  1.8209e-03, -9.5787e-05, -3.7698e-03, -8.2962e-03,
         9.0927e-03, -3.5490e-03,  6.6944e-03,  9.0500e-03,  4.4687e-04,
         2.9240e-03,  9.2649e-03,  8.8011e-03,  2.7315e-03,  9.6197e-03,
         3.6184e-03,  4.3006e-03,  2.3587e-05,  9.6131e-04, -8.7659e-03,
        -7.3431e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.3342e-02,  7.4171e-01,  5.0722e-02, -2.9266e-03,  4.0872e-02,
         2.9642e-03, -7.6086e-04,  8.3131e-03, -8.5928e-04,  2.5140e-04,
        -1.1975e-02, -5.4689e-03,  6.1259e-03, -1.5660e-03,  1.1781e-02,
         8.0491e-03, -3.8264e-03, -8.3311e-03, -2.6450e-03, -8.8235e-03,
        -6.2941e-03, -3.6078e-03, -9.5230e-04,  2.2703e-03, -5.5623e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0709, -0.6678, -0.0559,  0.0629, -0.0169, -0.0020, -0.0110, -0.0094,
         0.0035,  0.0070, -0.0034, -0.0168,  0.0090, -0.0061,  0.0022,  0.0051,
        -0.0124,  0.0040, -0.0116, -0.0012, -0.0209,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0190,  0.7712,  0.0331, -0.0137, -0.0273,  0.0080, -0.0016, -0.0055,
         0.0027, -0.0070, -0.0210, -0.0143, -0.0064, -0.0147, -0.0084, -0.0175,
        -0.0101,  0.0187,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1850e-02,  6.4773e-01,  2.4865e-02,  1.9706e-02,  1.1643e-02,
        -6.7561e-03,  1.2000e-02, -2.6511e-02, -1.3834e-02,  1.2952e-03,
         1.5895e-03,  1.3074e-02, -4.6380e-03,  2.5226e-03, -1.2026e-03,
         1.8422e-02, -1.4401e-02, -8.0004e-03,  8.6641e-03,  1.2184e-02,
        -2.3350e-03,  9.2320e-04,  3.0919e-03, -4.9156e-03,  6.3429e-03,
        -3.4225e-03, -9.3969e-03,  1.1373e-02,  4.0169e-03,  3.1839e-04,
         2.4591e-03,  2.7919e-03,  2.1426e-03,  1.3364e-03,  1.5393e-02,
         1.1008e-02,  1.7196e-05,  3.7824e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4837e-01, -6.7482e-01, -2.0910e-02,  5.8588e-02, -1.9478e-03,
        -2.1829e-03,  1.1560e-03, -6.9461e-03, -3.2449e-03, -5.3610e-03,
         9.6367e-03,  1.0437e-03, -2.2482e-03, -5.4717e-03,  1.8601e-04,
        -7.1990e-03,  1.1785e-02,  2.1735e-02, -8.7674e-03,  8.4014e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4532e-02, -7.3758e-01,  1.5906e-02,  9.0735e-03, -4.5324e-02,
        -2.0169e-02, -3.7073e-03, -3.6515e-02, -4.4739e-03,  3.4617e-03,
         3.1330e-02,  1.7368e-02, -5.7493e-05, -1.0504e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.5194e-03,  7.3209e-01, -1.9279e-02, -4.0558e-02,  2.3559e-02,
         5.3534e-03, -2.2948e-02, -1.0233e-02,  7.2061e-04, -1.1780e-02,
        -3.2378e-03, -2.8665e-03, -5.9465e-03, -4.0244e-03,  1.9618e-03,
         1.2949e-02,  2.9819e-03,  7.4364e-03,  1.0147e-02,  9.8345e-03,
         4.8023e-03, -1.3675e-03, -2.6553e-03,  5.4732e-03,  1.0590e-03,
        -1.6569e-02, -1.2959e-04,  9.1527e-03,  7.8931e-03,  3.8373e-03,
         1.6454e-03,  1.2986e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.7246e-02, -7.3527e-01, -1.6463e-02, -1.4000e-02,  3.8386e-02,
        -1.5265e-02, -5.5763e-03,  2.8755e-02, -4.3673e-05, -1.2103e-03,
         5.8386e-04,  1.4203e-02, -1.9810e-04, -4.5442e-03, -1.8756e-02,
         2.4438e-03,  5.6122e-03, -2.5893e-03,  1.8989e-02, -9.8624e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1437e-02,  5.9256e-01, -6.6132e-03, -2.4529e-02, -8.2476e-03,
         9.1815e-03,  2.4251e-04,  2.8620e-03, -3.4598e-03,  6.6834e-03,
        -1.0310e-02,  1.5700e-04, -7.3187e-03, -1.0911e-03,  2.8670e-04,
        -5.1431e-04, -2.6867e-03, -4.3624e-04, -8.7630e-03, -2.8418e-03,
         1.0618e-02, -1.3411e-02, -1.8622e-02,  1.5630e-03, -3.7592e-03,
        -8.1331e-03, -3.2743e-02,  3.0588e-03,  1.4477e-02,  1.2964e-03,
         4.6179e-05, -1.4407e-03,  6.9752e-03,  6.2290e-03, -2.0021e-03,
        -4.5959e-03,  3.9350e-03, -1.3024e-03, -4.4548e-03,  2.6723e-03,
        -4.7460e-03,  7.6604e-03,  5.0470e-04,  1.7189e-02,  9.8416e-04,
        -4.8153e-02, -5.9206e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5446e-02, -7.6908e-01, -7.3380e-03,  2.4168e-02, -7.5713e-03,
         5.9402e-03, -3.8334e-03,  3.7845e-03, -2.1938e-02, -7.4794e-03,
        -6.9138e-04, -2.5406e-02,  1.0922e-02, -9.7687e-03,  4.6631e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4917e-02, -6.9433e-01, -5.1566e-02,  1.1263e-03, -4.8305e-03,
        -2.4788e-02,  1.6872e-03, -2.7170e-02,  4.5085e-03,  3.9555e-03,
         7.9450e-03,  9.4127e-03,  6.3572e-03, -6.4925e-03,  1.0281e-02,
         4.5236e-03,  1.0031e-02,  3.0456e-03, -1.3666e-03, -1.7117e-03,
         3.8635e-03,  7.2158e-03,  3.1021e-03,  2.9231e-03, -1.1666e-03,
         3.8533e-04, -6.9492e-04, -1.4910e-02, -3.2774e-04,  2.5279e-03,
         1.4271e-03,  5.7380e-03, -1.5591e-03,  2.2293e-03,  5.1037e-03,
         2.7287e-03,  2.5187e-03,  2.3453e-03,  7.9414e-03,  9.6371e-04,
         4.9315e-04,  2.3107e-03,  3.1258e-03,  3.6314e-03,  2.0584e-03,
         3.5328e-03,  7.9143e-03,  7.2159e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-0.0717, -0.6973, -0.0053, -0.0094, -0.0109,  0.0116, -0.0025,  0.0071,
         0.0223,  0.0023, -0.0080,  0.0050,  0.0025,  0.0043,  0.0089,  0.0064,
         0.0311,  0.0126,  0.0099, -0.0029, -0.0061, -0.0011,  0.0024, -0.0182,
         0.0198,  0.0203,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.2664e-02,  6.8007e-01,  2.1250e-02,  2.4810e-02, -5.0085e-03,
        -4.6212e-03, -4.2154e-02, -4.8606e-03,  2.3541e-02, -4.5304e-04,
         1.4734e-02,  4.0260e-03,  1.1805e-03, -1.6875e-03, -5.7716e-03,
        -4.8825e-03, -1.5374e-02, -1.2545e-02,  5.6676e-03,  1.2796e-02,
        -4.1897e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2600e-03,  7.3610e-01, -4.6655e-02, -2.3239e-03,  5.2277e-03,
        -2.5845e-02, -7.1939e-03, -9.5415e-03, -5.3738e-03, -1.1745e-02,
        -4.6295e-03, -3.2343e-03,  8.1726e-04, -1.6777e-02, -8.3617e-04,
         5.0142e-03,  2.3873e-03,  7.8421e-03,  1.7541e-02, -1.7422e-02,
         1.2761e-02, -6.8844e-04,  7.7242e-03,  5.6367e-03,  3.4788e-03,
        -2.5782e-03, -4.0506e-03,  1.6193e-03, -6.9382e-04, -1.8476e-03,
        -1.0592e-04, -3.1796e-03,  2.2022e-03, -3.4893e-03, -9.1950e-03,
         1.2980e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2000e-02, -7.6448e-01,  1.2903e-02,  1.7268e-02,  6.3748e-04,
         9.5886e-03,  3.1729e-02,  1.8421e-02, -6.5536e-03, -3.2155e-03,
         2.9658e-02,  1.0772e-02,  1.3480e-03,  1.1439e-02,  8.0793e-03,
         2.4893e-03,  8.3800e-03,  6.2688e-03, -1.1975e-02,  2.2797e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1026e-02,  7.6909e-01,  1.2217e-02,  1.5470e-02,  1.2199e-02,
        -4.5765e-03, -1.2696e-02,  3.9614e-03,  8.1341e-03, -1.4341e-02,
         1.8112e-02, -1.1523e-02, -1.0219e-02, -4.3885e-03,  9.7140e-03,
         1.7523e-04, -8.0367e-03, -2.8447e-03, -1.2467e-02,  4.0524e-03,
        -1.4755e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0532, -0.8091, -0.0494, -0.0096,  0.0059, -0.0020, -0.0208,  0.0097,
         0.0048, -0.0057,  0.0058, -0.0194, -0.0046,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.5726e-02, -7.1016e-01, -1.1366e-02,  2.4961e-02, -4.1582e-03,
         1.0774e-03, -2.4314e-03,  4.1469e-03,  4.6475e-03,  8.4475e-03,
         4.6531e-03,  8.8536e-03, -4.2772e-04,  6.5962e-03, -2.8005e-03,
         3.9730e-05,  1.1940e-02, -1.5085e-03, -2.3591e-03, -5.6645e-03,
         2.0282e-04,  1.3038e-03,  3.1200e-03,  4.7980e-03,  5.2576e-04,
         3.2553e-03, -3.4311e-03, -9.6822e-03,  2.0213e-03, -4.9123e-03,
        -1.3888e-04, -2.2281e-03, -3.4575e-03,  5.4102e-03, -8.8588e-03,
         3.6663e-02,  2.8023e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.8251e-02, -7.4782e-01, -2.1844e-02,  1.9149e-02, -8.8430e-03,
         8.0014e-03, -1.1452e-02, -2.7618e-02, -6.0008e-04, -3.5026e-03,
         1.7967e-03,  2.2014e-03, -1.1447e-02,  4.7820e-03, -5.2482e-03,
        -5.2549e-03, -1.5529e-02,  2.0153e-03,  6.1326e-03,  1.2838e-02,
        -6.1738e-03, -3.2265e-03,  1.9687e-03,  4.3036e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.0526e-02,  6.3851e-01, -1.0790e-02, -8.5212e-03,  1.1296e-02,
        -8.1089e-03, -2.3871e-02, -3.8536e-02, -2.3701e-02, -2.8758e-03,
        -3.1071e-02, -1.1899e-02, -2.6630e-03, -6.5084e-03, -1.9815e-03,
        -2.5461e-03,  7.6351e-03, -3.1548e-03, -4.0174e-03,  8.8778e-03,
         1.4216e-03, -7.6674e-03, -2.6118e-03, -2.6398e-03, -1.2024e-02,
        -4.1604e-03,  1.5168e-03,  4.8354e-03, -6.2549e-04, -6.5856e-03,
        -1.9376e-02, -2.9444e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5034e-02,  7.5296e-01,  3.0912e-02, -1.0993e-03,  3.3865e-03,
         6.7242e-03,  2.9315e-03, -4.2317e-02,  6.7094e-03,  1.4902e-02,
         5.2550e-03, -4.5745e-03, -5.8295e-04, -3.1782e-05, -2.9771e-03,
        -3.2986e-03, -2.3676e-03, -9.4285e-03,  2.5948e-04,  9.8400e-03,
        -8.8199e-03,  6.9202e-04, -2.0911e-03,  9.3816e-03, -3.1044e-04,
         1.0266e-02,  4.4039e-03,  2.2198e-04, -2.4713e-03, -1.4298e-03,
         4.3163e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0153, -0.7925,  0.0215,  0.0050, -0.0115,  0.0088, -0.0133,  0.0248,
         0.0012, -0.0135, -0.0145,  0.0055, -0.0064, -0.0015,  0.0019, -0.0059,
        -0.0133, -0.0043, -0.0021, -0.0048, -0.0240, -0.0057,  0.0027,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9567e-02,  7.6427e-01,  3.0545e-02, -1.5873e-02,  1.3079e-02,
         2.0106e-05,  1.3477e-02,  4.5735e-03,  1.0709e-02, -2.1499e-03,
         8.9144e-04, -1.1260e-02, -2.9386e-03, -1.3177e-02, -2.6205e-03,
        -7.9057e-03, -2.8054e-03, -2.8554e-03,  3.7654e-03, -5.2424e-03,
        -1.2099e-02, -1.6256e-03,  7.4754e-03, -3.8666e-03, -4.1681e-03,
         7.6152e-03,  4.7046e-03,  1.0716e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 5.3739e-03, -8.5158e-01,  2.3874e-02,  1.0686e-02,  1.0959e-02,
         3.2564e-03, -1.1681e-03,  3.9510e-03,  6.7535e-03,  8.4956e-03,
        -2.7920e-03,  1.2334e-03, -1.0985e-03,  2.9569e-03,  2.0092e-03,
         2.2339e-03, -2.9341e-03,  1.6766e-03,  1.3188e-03, -4.8633e-03,
         1.3275e-03, -4.0224e-03,  1.3883e-02,  2.7769e-03,  4.1522e-03,
         2.3088e-03, -2.8757e-03, -1.9058e-02,  3.8177e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8245e-03,  7.0222e-01, -6.4729e-03,  2.1365e-02,  5.2921e-02,
         3.1289e-03,  5.8950e-03, -1.8599e-02, -1.5186e-03,  2.6411e-02,
         7.1190e-03, -3.9751e-03, -2.2726e-04,  8.5325e-03,  5.0276e-03,
         8.2031e-03,  6.0413e-03, -6.1949e-03,  1.6801e-02, -1.4904e-03,
        -2.5421e-03,  3.9771e-02, -3.5039e-02,  1.8685e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.0664e-02, -5.8977e-01, -4.1139e-02,  2.7113e-02, -2.0949e-02,
        -1.7418e-02, -9.3095e-03, -6.6289e-03, -2.4469e-03,  3.6999e-03,
        -3.9238e-03, -2.0811e-02, -1.1348e-02, -2.2128e-03,  1.7025e-03,
         6.0774e-04,  1.5128e-02,  1.6235e-03, -6.6380e-03,  2.6691e-03,
        -7.7173e-03,  1.8535e-03,  3.0782e-04,  2.5567e-03, -7.0951e-03,
        -2.8774e-03,  6.6277e-03, -1.4558e-03, -2.9660e-03, -3.1625e-03,
         3.1507e-03, -2.5605e-02, -1.0723e-02,  7.8102e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.6991e-03,  8.1133e-01, -2.1136e-02, -5.1788e-02,  9.4937e-03,
         2.0890e-03,  1.9888e-02, -3.9700e-04,  1.7902e-03,  3.5313e-03,
        -7.8223e-03, -5.7099e-04,  5.4277e-04, -6.9193e-03,  1.2318e-02,
         6.0042e-03, -9.1438e-04,  2.0106e-03,  6.6057e-03,  7.4926e-04,
         2.3584e-02,  8.1703e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7061e-02, -7.6197e-01, -5.2487e-02,  2.3212e-02, -4.0438e-02,
         8.9345e-03,  1.9105e-02, -6.9953e-04, -1.8411e-03, -9.3235e-03,
        -1.7941e-02, -4.8057e-03,  4.7403e-03,  1.9101e-04, -5.1628e-04,
         3.7011e-03, -9.4279e-03,  2.3605e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.9460e-03,  8.3478e-01,  4.5377e-03, -2.3462e-03, -2.4497e-03,
         1.3873e-03,  8.4827e-03, -1.7139e-02, -3.9180e-03,  6.3184e-03,
         5.0976e-03,  6.8461e-03, -4.7146e-03, -5.3188e-04, -2.8507e-04,
        -1.4604e-03, -3.1062e-03, -1.0148e-02, -5.2723e-03, -1.3271e-02,
        -1.9683e-02, -5.7040e-03, -3.3578e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0120e-02, -8.4921e-01,  2.3929e-02, -1.2674e-03,  2.7087e-02,
         8.6791e-03, -1.1022e-02, -4.1508e-04,  1.1224e-02,  7.3496e-04,
        -2.2019e-03, -3.8752e-03,  2.9054e-03, -7.6106e-03, -6.5909e-03,
         2.2262e-02, -1.0859e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.2396e-02,  7.3469e-01,  2.9725e-02, -5.1272e-04,  1.7212e-02,
         9.5999e-03,  2.9038e-02, -3.2891e-03, -4.3923e-03, -5.2828e-03,
         1.4907e-02, -3.4433e-03, -6.2518e-03,  1.5201e-02,  4.4003e-03,
        -4.7947e-03, -1.9620e-03, -2.2900e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8386e-02, -7.1988e-01, -2.8829e-02,  1.3478e-02,  6.0116e-03,
         6.4626e-03,  5.1650e-03, -1.5026e-02, -1.2735e-03, -2.5923e-02,
        -1.9464e-02,  1.3671e-02, -8.3728e-03,  8.3343e-03,  8.1871e-03,
         1.0493e-04,  7.9144e-04,  9.9301e-03, -1.0250e-02, -4.1382e-03,
        -4.5626e-03,  2.2718e-03, -1.0041e-02,  3.5227e-03,  1.9539e-02,
         1.6387e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.3540e-02, -6.6711e-01, -3.6934e-02, -7.8198e-03, -1.6726e-02,
         1.5394e-02,  9.1973e-03, -3.0462e-03, -1.8196e-02,  3.0729e-02,
         7.2023e-03,  4.3484e-03,  5.3973e-03,  5.6241e-03,  1.3947e-02,
        -7.1759e-04,  7.8569e-04, -8.8598e-04,  6.5683e-03,  1.2986e-02,
         8.4165e-04,  5.1168e-04, -1.1386e-02,  3.6489e-03, -3.3830e-03,
         2.9076e-04,  4.8453e-03,  4.6164e-03,  9.3957e-03, -1.5357e-02,
        -2.8566e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0515, -0.7630,  0.0062,  0.0058,  0.0276,  0.0078, -0.0079, -0.0232,
        -0.0091, -0.0171,  0.0140,  0.0241,  0.0069,  0.0093,  0.0145,  0.0065,
        -0.0019,  0.0038,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0161,  0.7448, -0.0354,  0.0084, -0.0561,  0.0783,  0.0422, -0.0185,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([-4.9979e-02, -7.4962e-01, -6.0182e-03,  6.4424e-03, -2.0889e-03,
        -3.1901e-02,  5.2906e-03,  4.8441e-03,  2.2365e-02,  8.3988e-04,
         3.8079e-03,  9.7967e-03, -6.2204e-03, -1.0901e-03, -4.8461e-03,
         5.0925e-03, -3.3962e-03, -2.6907e-03, -1.3083e-03, -1.7008e-03,
        -5.3402e-04,  2.3000e-03,  5.7624e-03, -2.8349e-03,  4.3687e-04,
        -2.1494e-03,  3.9352e-03, -7.6655e-04,  1.1422e-02,  1.3756e-03,
        -1.8328e-03, -1.2886e-03,  2.1311e-03,  3.1033e-03, -2.4692e-03,
         8.5059e-03, -1.3703e-02, -6.0704e-03, -1.0036e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2566e-02,  7.0693e-01, -2.2043e-02,  3.4155e-04,  3.5092e-02,
         1.2755e-02, -1.9279e-02,  4.1709e-03, -4.8630e-03, -1.1383e-02,
        -5.8064e-03, -6.0239e-03,  4.7496e-03, -1.0091e-02,  4.5692e-03,
         1.2825e-02, -2.4338e-03, -9.8064e-03, -1.3468e-02,  5.8314e-03,
        -7.6470e-04,  2.4054e-02,  6.3703e-03, -7.1643e-03, -1.8910e-03,
        -5.7512e-03, -2.9990e-02, -8.9826e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0178, -0.8126, -0.0143, -0.0327,  0.0063, -0.0036, -0.0175, -0.0126,
        -0.0236,  0.0091, -0.0289, -0.0101, -0.0111,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0340,  0.7178, -0.0043,  0.0269,  0.0120, -0.0101, -0.0446, -0.0055,
        -0.0043, -0.0185, -0.0023, -0.0414, -0.0150,  0.0513, -0.0121,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8360e-02,  5.8177e-01, -3.2880e-02, -2.2839e-02,  9.9616e-03,
         1.1328e-02, -1.3642e-02, -9.5518e-03, -1.0803e-02, -1.3655e-02,
        -1.0783e-02, -6.3609e-03,  1.2866e-02,  5.9600e-03,  5.3592e-03,
         1.0345e-02, -1.8008e-03,  4.6816e-03, -4.3941e-03, -2.7547e-02,
         1.3146e-02, -2.1113e-03, -4.2261e-03,  6.1487e-05,  1.8592e-03,
        -6.6811e-04, -7.1942e-04,  9.8611e-04, -8.6171e-03,  1.4736e-03,
        -1.6661e-02, -5.9089e-03, -7.3220e-03,  2.0938e-03,  6.0967e-03,
        -5.7937e-03,  3.0440e-03,  3.3031e-03, -3.5790e-03,  2.4209e-03,
        -1.9503e-03, -4.2569e-03,  1.0310e-03,  1.1636e-04,  4.5128e-03,
         3.8056e-03, -7.9347e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3979e-02,  7.1277e-01,  2.5305e-02, -8.5117e-03, -1.9724e-02,
         1.0797e-02, -1.4757e-02,  1.0074e-03,  2.2664e-02, -4.9588e-03,
        -8.1339e-03, -2.9032e-03,  6.4932e-03,  5.4666e-03,  2.5689e-03,
        -2.6674e-03,  4.5039e-03,  1.0028e-02,  1.2242e-03, -2.3850e-04,
         4.2214e-03,  3.5499e-04,  7.1366e-03,  1.3478e-02,  1.0952e-02,
        -2.3586e-03,  8.8467e-03, -8.8852e-04,  8.2627e-04,  3.7956e-03,
         5.9805e-03, -2.8956e-03,  3.9218e-04,  7.9969e-03, -4.3169e-04,
         1.0386e-03,  2.0976e-03,  3.2004e-03,  5.3092e-04, -1.9525e-03,
        -1.4554e-03, -2.0472e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.3196e-02,  6.6762e-01,  3.0573e-02,  8.5697e-03,  2.1517e-03,
         1.5296e-02,  1.1886e-02, -3.5001e-03,  1.0133e-02,  1.7010e-02,
         8.9219e-03,  1.8361e-02, -1.2978e-02, -3.5873e-03,  1.2669e-02,
        -1.4365e-02,  1.9352e-02,  5.2326e-04,  3.7086e-03,  7.2116e-03,
        -5.9996e-03, -1.3001e-03, -4.0604e-03, -1.2013e-02, -3.5016e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3957e-02,  7.8128e-01,  2.8794e-03,  1.0043e-02, -1.4137e-02,
         1.8189e-02,  5.2626e-04, -1.8938e-03, -7.9433e-03, -9.1722e-04,
         2.0501e-03, -1.0642e-03,  6.6539e-03, -1.5340e-02, -1.2106e-02,
        -9.5622e-03,  5.6711e-04, -3.3623e-03,  4.9029e-03,  1.5197e-02,
         8.7982e-03,  5.4849e-03, -3.3144e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5538e-03,  7.7127e-01,  8.9086e-02, -3.3033e-04, -8.0652e-03,
        -5.6395e-03, -1.3797e-02,  5.6019e-03, -1.4462e-02,  4.9662e-03,
         4.4223e-03, -5.2041e-03,  3.9121e-03, -1.2283e-02,  4.6070e-03,
        -4.6922e-03, -9.4543e-03, -5.4888e-03,  1.4087e-03, -2.2372e-03,
        -4.3543e-04, -1.2451e-02, -1.5632e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.7671e-02,  6.7256e-01, -1.5736e-02, -7.0516e-02, -3.8361e-02,
         1.2217e-02,  1.3110e-02,  4.4761e-02, -3.1338e-04,  3.2619e-02,
         3.8541e-03, -1.3622e-02,  2.4659e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0290,  0.8605, -0.0089,  0.0038, -0.0092,  0.0011, -0.0027,  0.0180,
        -0.0017,  0.0240, -0.0040, -0.0178, -0.0167,  0.0026,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5437e-02, -8.6418e-01, -7.6025e-03,  1.6326e-02,  5.8234e-03,
         2.4491e-04,  3.7246e-04,  7.2504e-03, -1.7940e-03,  1.7902e-02,
         3.7576e-03,  9.5636e-04,  3.8440e-04,  1.1650e-02,  1.0707e-02,
         9.2696e-03,  1.3862e-04,  2.4266e-03,  1.3775e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 4.5861e-02,  6.5719e-01, -5.3050e-02, -5.6909e-02, -7.3016e-03,
         1.6454e-02, -2.6610e-04, -5.1755e-03, -9.9761e-03,  7.0178e-03,
        -1.4468e-03,  2.2697e-03,  7.9181e-03,  1.0745e-02, -2.6543e-02,
        -1.4088e-02, -1.2733e-03,  1.0112e-03,  4.5620e-03,  5.2316e-03,
        -4.6798e-03, -1.1288e-02,  4.1833e-03,  1.6296e-02, -2.6359e-02,
        -2.9001e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0622, -0.7507,  0.0008,  0.0139,  0.0321,  0.0213,  0.0032,  0.0155,
        -0.0100,  0.0079,  0.0152,  0.0044, -0.0153,  0.0085,  0.0189, -0.0200,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0178, -0.7864,  0.0073,  0.0077,  0.0010, -0.0254, -0.0123, -0.0034,
        -0.0437,  0.0130,  0.0061,  0.0107, -0.0033, -0.0161, -0.0092,  0.0071,
        -0.0296,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2692e-01,  7.5308e-01,  6.5844e-04,  1.4431e-02,  6.5913e-02,
        -2.8798e-03,  2.7978e-03,  1.5705e-02, -1.7614e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8556e-02,  5.0943e-01,  1.2281e-02,  4.5124e-03,  3.1551e-02,
         7.5164e-03,  1.7987e-02, -5.0672e-03, -1.0831e-02,  4.6651e-04,
         5.8248e-03,  4.6975e-02, -5.7283e-03, -1.7777e-03,  1.7734e-03,
         5.7035e-03, -2.3514e-03,  5.6718e-03,  2.2927e-03,  4.9879e-03,
         4.9816e-03,  7.6433e-03, -2.0640e-02, -5.5514e-03,  1.0209e-02,
        -9.9104e-03, -1.0253e-02, -3.7957e-03,  3.3945e-03, -2.6246e-03,
        -7.4577e-03,  2.7164e-03, -7.1762e-04,  8.8461e-04, -2.2906e-03,
        -6.7454e-03, -3.8593e-03,  4.7232e-03, -9.5467e-03, -2.3909e-03,
        -8.9823e-03, -3.8342e-02, -7.1112e-03,  1.6140e-02, -2.1422e-02,
         7.6381e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1526e-02,  7.7297e-01,  2.2729e-02,  3.3902e-02,  8.2230e-03,
        -5.3340e-03, -6.1424e-03,  2.1044e-03, -1.2056e-02, -5.7419e-03,
        -7.6966e-03, -3.8162e-03, -1.1829e-02, -1.1813e-03,  2.1792e-03,
         7.3466e-03,  2.7754e-03, -4.1129e-03,  3.4226e-03,  1.5401e-03,
        -1.7386e-04, -3.6426e-03,  7.0689e-03,  1.1696e-03, -4.7877e-03,
        -5.3867e-03, -3.3156e-03, -1.8557e-03, -1.1061e-02,  2.4909e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0569e-01, -8.0983e-01,  7.4909e-04, -2.3797e-03,  1.0995e-02,
        -5.6072e-02, -1.4285e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9412e-02,  8.8658e-01, -9.4564e-03, -2.6802e-02, -2.2344e-04,
        -1.1378e-02,  4.8051e-03, -4.0614e-03, -4.0083e-03,  8.0966e-03,
         5.1767e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3046e-05,  6.8798e-01,  6.7637e-02,  7.3681e-02, -6.3753e-03,
        -3.5819e-02,  1.0210e-02,  4.5001e-02,  5.7801e-03, -2.2930e-03,
        -3.9346e-03,  8.5188e-03,  1.7119e-02, -3.5556e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7329e-02, -7.8897e-01, -1.3333e-03, -8.2560e-03, -2.0682e-03,
        -7.2995e-03,  7.2198e-03,  7.1281e-03,  1.0940e-02,  4.4523e-03,
        -1.2610e-03,  1.1882e-03,  3.6065e-03, -5.2781e-03, -5.0274e-03,
        -5.0533e-03,  2.3972e-05,  4.9890e-03, -9.8806e-05,  3.9390e-03,
        -1.5564e-02, -1.3541e-02,  4.0115e-03, -4.4119e-03, -1.3372e-03,
         1.1544e-03,  1.0870e-03,  4.4741e-03,  3.8446e-03, -1.5836e-03,
         8.1943e-03, -2.2512e-02, -4.0206e-03, -2.8804e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7151e-02,  7.0550e-01,  6.5621e-02, -6.6974e-03,  1.6903e-02,
         1.4601e-02,  6.5593e-03, -2.8838e-03, -9.7404e-03,  2.1786e-02,
         2.4560e-03, -4.3257e-03, -3.7164e-03,  8.1835e-03,  1.7160e-02,
        -5.5143e-03, -5.6269e-04, -9.0332e-03, -3.4642e-03,  1.9714e-03,
         4.9225e-03,  9.9714e-03, -5.6290e-03, -2.5278e-03,  5.8738e-03,
         1.0375e-02,  1.4093e-02,  2.2778e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.2810e-02,  7.0321e-01, -1.3025e-02, -2.8257e-02, -2.7160e-02,
        -1.3036e-02,  5.9090e-03, -3.1474e-03, -3.0922e-04,  2.4122e-03,
        -2.0826e-04,  8.8846e-03, -1.4621e-03,  2.9120e-02,  1.2783e-02,
         6.3898e-03,  1.1449e-02, -8.4590e-03,  1.5867e-02,  3.5860e-03,
        -2.8353e-03,  2.0613e-03, -2.0214e-02,  1.7405e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-3.5831e-02,  6.5511e-01, -6.7325e-02,  5.6095e-03,  3.7349e-03,
        -1.2395e-02,  2.0287e-02,  9.8198e-03, -5.1877e-03,  5.1841e-03,
        -1.8877e-02,  1.0245e-02,  2.0742e-02,  8.0285e-03,  6.9008e-03,
         8.6481e-04,  1.0598e-03, -3.5125e-03,  6.2775e-03,  5.7703e-03,
         5.8356e-03,  7.9097e-03, -2.2735e-03,  2.0123e-03,  7.4469e-03,
         1.9570e-02, -4.4385e-03, -3.1371e-04, -2.3313e-03, -3.9167e-02,
        -5.9351e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.6662e-03,  8.2120e-01, -5.5489e-02, -5.9776e-02,  6.1257e-03,
         1.1197e-02, -6.8354e-03, -1.0504e-04,  4.3300e-03, -5.0313e-03,
         1.8882e-02, -4.3643e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0024, -0.7705,  0.0147,  0.0502,  0.0011, -0.0168,  0.0247,  0.0108,
         0.0056, -0.0041, -0.0398,  0.0282, -0.0167, -0.0038, -0.0105,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.4772e-02,  7.0431e-01, -1.7599e-02, -2.3937e-02,  1.5535e-02,
        -9.9463e-03,  8.8192e-03, -5.9116e-03, -5.2536e-03,  7.0361e-03,
        -3.4523e-03,  1.3027e-02,  7.5929e-03, -4.0835e-04, -4.5526e-03,
         3.9239e-04,  5.4863e-03, -7.2213e-04,  7.0449e-03, -8.0174e-03,
         3.2944e-03, -9.8521e-04, -6.4289e-03,  6.5177e-04,  7.3128e-03,
        -4.3392e-03, -1.0597e-03, -1.8030e-03, -5.9074e-03,  5.3230e-04,
         9.8321e-03, -1.8797e-03,  1.2155e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2412e-02, -7.6912e-01,  5.3920e-02,  2.9220e-02, -7.0065e-03,
         2.6801e-03,  9.4135e-03, -9.6244e-03,  4.5430e-03,  4.9023e-03,
         2.8002e-02, -4.0258e-03, -1.2516e-04,  3.9669e-02, -4.4847e-04,
         1.4886e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0223,  0.7729, -0.1251,  0.0070,  0.0369,  0.0103, -0.0100, -0.0112,
        -0.0043,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1318,  0.7670,  0.0135, -0.0365,  0.0511,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0653,  0.7476,  0.0140, -0.0080, -0.0080, -0.0155, -0.0053, -0.0165,
        -0.0046,  0.0066,  0.0084,  0.0057, -0.0146, -0.0113,  0.0035,  0.0043,
         0.0040,  0.0070,  0.0130, -0.0028, -0.0017, -0.0023,  0.0029, -0.0064,
        -0.0023, -0.0081,  0.0102,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0439,  0.7674,  0.0317,  0.0065,  0.0491,  0.0335, -0.0679,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5125e-02,  7.0353e-01, -3.2424e-02, -9.9154e-03, -2.5346e-03,
         1.0397e-03,  3.1306e-02,  4.3351e-03, -6.5896e-03, -1.9018e-02,
         1.1557e-02,  5.1779e-03, -6.7067e-03, -4.6499e-03,  3.0837e-04,
        -8.5250e-04, -1.9989e-02, -3.8880e-02,  1.1401e-02,  7.9891e-03,
        -1.1785e-02,  1.4892e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6756e-02, -8.5216e-01,  4.8492e-03,  1.1692e-02, -3.4429e-02,
         6.1828e-03,  1.0601e-02, -1.8493e-03,  1.8597e-03,  2.4483e-03,
         3.3744e-04,  1.2637e-03,  2.0430e-02,  7.5688e-05, -1.4454e-03,
         5.1047e-05, -1.3126e-03, -1.1225e-02, -2.1032e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0387,  0.7426,  0.0082, -0.0054,  0.0036,  0.0021, -0.0061,  0.0310,
         0.0143, -0.0015,  0.0032,  0.0028, -0.0017, -0.0024,  0.0093,  0.0081,
        -0.0061,  0.0055,  0.0022,  0.0023,  0.0069,  0.0058,  0.0018,  0.0081,
         0.0061,  0.0119, -0.0012, -0.0032,  0.0248,  0.0059,  0.0273,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #300: [tensor([-3.7402e-02,  6.8822e-01,  1.8675e-02, -9.2417e-03,  3.1552e-02,
        -8.0584e-03, -9.6844e-04,  8.3694e-03, -9.5485e-03,  1.8953e-03,
        -3.2150e-04,  1.9785e-03, -1.8654e-03, -6.3156e-03,  1.8134e-02,
        -1.0465e-02, -7.8409e-03, -1.9771e-02, -2.4707e-02,  3.6030e-03,
        -7.2295e-03, -9.9855e-03,  1.9735e-03, -1.1188e-02,  2.6092e-03,
        -4.7747e-04, -2.9504e-03, -2.1300e-02, -1.9091e-03, -1.3700e-03,
        -1.5655e-03, -2.8508e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7726e-02,  7.5928e-01, -2.0352e-02, -1.3020e-02,  2.4933e-02,
         6.9165e-03, -2.4404e-02,  1.5828e-02,  1.2714e-02, -8.6040e-04,
        -1.2756e-02,  2.7542e-03,  3.4118e-04,  4.8831e-03, -2.7671e-03,
        -4.9667e-03,  4.4953e-05, -2.7084e-03, -1.1376e-03,  3.4245e-03,
         2.4017e-02, -4.1677e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0569,  0.7150,  0.0210, -0.0061, -0.0013, -0.0031, -0.0069, -0.0123,
        -0.0158, -0.0042, -0.0102, -0.0038,  0.0009, -0.0082, -0.0054, -0.0030,
         0.0014,  0.0076, -0.0068, -0.0126,  0.0032, -0.0068,  0.0022, -0.0120,
        -0.0062, -0.0033, -0.0056, -0.0163,  0.0352, -0.0068,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.6726e-03,  7.6226e-01,  3.8675e-02, -4.0007e-02, -1.0665e-02,
         2.4583e-03, -1.8901e-02, -4.1884e-03, -9.5138e-03, -1.9389e-02,
        -1.8269e-02,  6.7964e-03, -1.2896e-02,  1.2685e-02, -1.1801e-02,
         2.0338e-02,  4.3289e-03, -1.5085e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.9729e-02, -6.6532e-01, -4.9717e-02, -6.9958e-03, -1.0918e-02,
        -1.0550e-02, -2.2592e-03, -1.1341e-03, -8.5218e-05,  1.2256e-03,
         5.7716e-03,  1.2055e-03,  1.6348e-04,  3.4500e-05,  8.5273e-03,
         2.0745e-02,  1.6636e-02,  5.3254e-03, -6.5613e-03,  7.6685e-03,
         4.5102e-03,  2.9852e-03, -1.0001e-03,  1.0287e-03, -4.8132e-03,
         1.1028e-03, -9.6986e-03, -8.6796e-04,  4.3657e-03, -4.7451e-03,
        -1.0967e-03, -3.5645e-03,  1.8558e-03, -1.2112e-03, -1.8978e-03,
         2.0157e-02,  1.4528e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.7074e-02,  6.3668e-01, -1.3188e-01, -3.5195e-04,  2.2216e-02,
         2.1011e-03, -4.4505e-03, -2.7158e-03,  4.7166e-03,  5.9802e-03,
         1.4089e-03,  1.2642e-02, -1.0385e-04,  1.2915e-02,  2.5507e-03,
        -2.3058e-03, -1.7028e-02,  3.2480e-03, -6.4495e-03, -4.2017e-03,
        -6.6527e-04, -2.0743e-02,  4.3309e-03, -3.2466e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5180e-02,  7.1054e-01,  9.3725e-03, -9.0507e-03,  4.4165e-02,
         3.7825e-03, -1.7738e-03,  3.7462e-03,  7.7071e-03,  1.2386e-03,
        -2.7770e-03,  4.4683e-03, -4.8232e-03,  3.0509e-03,  1.1861e-03,
         1.1743e-03, -1.4680e-03, -5.6021e-03, -8.3840e-04, -7.2769e-03,
        -1.9419e-03,  5.6999e-03, -4.4945e-04,  1.7373e-03,  1.0796e-02,
         8.4849e-03,  8.1250e-04,  6.8184e-03,  1.1506e-02, -6.1761e-03,
         5.0868e-03,  1.2676e-02,  2.0018e-03,  2.3696e-03,  5.6757e-03,
        -7.8562e-03, -8.2886e-04,  4.7358e-03,  2.9681e-03,  4.0133e-03,
        -2.8555e-04, -4.3215e-03, -2.1238e-03,  2.7583e-03, -1.0335e-02,
         8.3202e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3908e-02,  7.9860e-01,  8.6941e-03, -2.2798e-03, -2.9654e-02,
         1.4726e-03, -3.0843e-02,  4.9066e-05,  1.7951e-03, -1.5317e-02,
        -1.4830e-02, -6.8115e-03, -4.8205e-03, -2.4699e-03,  3.0215e-03,
         6.6841e-03, -3.6841e-03,  1.4249e-02,  4.0786e-03, -3.1953e-03,
         1.7582e-02,  5.6868e-04,  1.5395e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1069, -0.7257,  0.0223,  0.0280,  0.0253, -0.0117, -0.0280, -0.0020,
         0.0183,  0.0277,  0.0040,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0455,  0.7748,  0.0612, -0.0166,  0.0225, -0.0170, -0.0177, -0.0156,
         0.0018, -0.0018,  0.0020,  0.0011,  0.0036, -0.0188,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4833e-02,  7.4140e-01,  3.3613e-02, -1.6157e-02, -8.9329e-05,
         6.4052e-03, -1.6670e-02, -8.1655e-03,  3.2710e-02, -9.5063e-03,
         5.3217e-02,  2.7233e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2800e-02,  7.2603e-01,  4.5310e-02,  1.9734e-02,  4.9781e-03,
        -1.8653e-02,  2.8585e-05, -3.6453e-03,  1.1910e-03,  6.5747e-04,
        -7.6721e-05, -3.6087e-03,  2.9930e-04,  3.3694e-03, -2.4822e-03,
        -3.9957e-03, -2.6405e-03, -3.7408e-03,  7.2739e-04, -8.0334e-03,
        -4.2407e-03,  1.0124e-03,  2.8122e-03, -1.5651e-03,  3.5755e-03,
         1.1538e-03,  3.5816e-02, -6.2492e-03, -5.3257e-03, -4.7401e-03,
        -3.5729e-03,  2.5661e-03, -3.2863e-03,  1.0194e-02,  1.3900e-03,
        -3.5716e-03,  3.4506e-03, -4.4813e-04, -1.1995e-03, -2.4603e-03,
         2.7502e-03,  1.0203e-03,  4.5044e-03, -6.9328e-04, -2.4487e-03,
         1.4492e-04, -2.4282e-03, -2.5866e-03, -2.4749e-03, -1.6296e-03,
        -9.3308e-05,  4.1845e-03, -3.5882e-03,  2.2956e-03,  8.5290e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-0.0533, -0.6759,  0.0324, -0.0019,  0.0293,  0.0214,  0.0046,  0.0053,
         0.0056, -0.0098,  0.0048, -0.0022, -0.0212, -0.0305, -0.0076, -0.0047,
         0.0204, -0.0255,  0.0047, -0.0150, -0.0033, -0.0184,  0.0022,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0513,  0.7283,  0.0704,  0.0342, -0.0035,  0.0115, -0.0116,  0.0055,
         0.0064, -0.0087, -0.0021,  0.0040, -0.0077, -0.0078, -0.0044, -0.0053,
         0.0158,  0.0215,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.0603e-02, -6.5367e-01, -1.1880e-02, -5.8210e-03, -5.5509e-02,
        -3.7640e-03, -3.0124e-02, -2.0552e-02, -1.8691e-03, -7.5121e-03,
         1.5065e-02,  4.2365e-03, -4.7352e-02, -8.2475e-03,  1.7927e-02,
        -1.3414e-04,  1.0498e-02, -1.5233e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.4914e-02,  7.3642e-01, -2.9578e-02,  6.0934e-03,  1.7129e-02,
        -3.2314e-02,  1.3626e-02,  1.4273e-02, -1.0667e-02,  3.2311e-04,
        -6.8706e-03, -3.5271e-03,  1.0433e-03, -1.9470e-02, -6.3751e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2217e-01,  6.7559e-01,  7.9846e-03,  9.9998e-03, -4.9409e-03,
         3.0945e-03,  7.9496e-04,  1.7127e-02,  1.7858e-02,  5.7378e-03,
         2.9662e-03,  1.0382e-03, -3.2467e-04,  4.6617e-03, -9.3494e-03,
         9.4479e-03, -5.8202e-03,  3.0441e-04, -9.0830e-03,  2.1306e-02,
         3.1820e-03, -5.4426e-03, -2.4478e-03,  5.3754e-04, -1.7183e-02,
         3.6546e-03,  1.6230e-02, -2.4381e-03, -2.4280e-03,  1.9758e-03,
        -1.6104e-03, -1.3277e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2093e-02,  7.6522e-01, -4.4893e-02, -2.9432e-03, -1.7192e-02,
         5.1430e-03, -5.3558e-04,  8.5825e-03, -5.5549e-03, -4.6173e-03,
         2.6458e-03, -6.9755e-03, -1.3119e-02, -3.6137e-03, -3.1208e-03,
        -4.8785e-03, -7.2721e-03, -4.0113e-03, -2.1999e-03, -1.7317e-03,
        -1.8238e-03, -4.1745e-03, -1.3874e-02, -1.3361e-04,  5.0573e-03,
         2.7712e-03,  1.3119e-03, -2.2302e-03,  1.0169e-03, -2.2366e-03,
         5.2315e-03,  1.0948e-02,  3.2850e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5134e-02,  7.2879e-01,  1.9115e-02,  7.2804e-03,  1.6496e-02,
        -1.6708e-03,  6.2365e-03,  1.0229e-02, -3.5770e-05,  1.6093e-02,
         2.7520e-04, -4.1739e-03,  4.2175e-03,  8.1948e-03,  8.5403e-03,
         1.6820e-02, -1.6314e-03, -2.4565e-03,  3.5362e-04, -1.5771e-03,
        -3.2169e-03, -9.2554e-04, -1.3552e-03, -1.7998e-03,  7.3672e-03,
         4.9546e-04,  5.5183e-04,  6.7816e-04,  9.3821e-03, -8.6632e-04,
         8.2784e-03,  3.2663e-03, -2.5176e-03,  3.3667e-03,  3.0624e-03,
        -1.7400e-03,  7.6437e-04, -2.3667e-03, -2.4750e-03,  5.2517e-04,
        -4.1263e-04, -5.7380e-04,  2.4376e-05,  1.6403e-04,  3.1618e-03,
         4.1482e-03,  3.5541e-03,  2.8633e-03,  2.9676e-03,  1.0377e-03,
        -1.8670e-03,  2.3805e-03,  5.9771e-03,  5.7559e-03,  4.0056e-03,
         7.7507e-03,  7.9811e-03, -1.0539e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2713e-02,  7.3678e-01,  4.7955e-02,  1.0427e-02,  5.6988e-03,
         1.2889e-02, -5.5820e-03, -1.7194e-03,  1.4773e-03,  2.6767e-02,
         1.1451e-03,  1.5813e-02,  5.0355e-03, -1.3718e-02,  7.1655e-04,
         9.8942e-03, -9.5261e-04, -5.5170e-03, -2.3314e-03,  3.2780e-03,
         3.1750e-03,  4.0746e-04, -1.2079e-03, -6.9919e-03, -2.7519e-02,
         1.0285e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4088e-02,  7.0174e-01, -1.9618e-02, -5.3372e-02, -2.8215e-02,
         3.3470e-02, -8.5374e-03, -5.4571e-03, -8.5960e-03,  4.8407e-03,
         7.3705e-03, -1.3876e-03,  2.3691e-03,  5.4622e-05,  5.6667e-04,
         1.3725e-04, -1.3729e-03, -8.0892e-03,  4.5267e-03,  3.5689e-03,
        -4.7997e-04, -1.2213e-02, -1.3098e-02, -1.2470e-03, -1.5581e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.8843e-03,  6.5603e-01,  4.4742e-02, -3.1597e-02, -1.0258e-02,
        -1.4764e-02, -1.2162e-02, -2.3734e-02,  8.8991e-03,  9.6449e-03,
        -3.3553e-03, -1.0085e-02,  1.0944e-02,  6.9722e-03, -1.4033e-03,
         1.1442e-03, -7.8060e-03,  3.3471e-05, -2.2841e-04, -3.8523e-03,
         3.5122e-04,  7.7991e-03, -1.4884e-03, -6.0901e-04,  9.0405e-03,
         7.7361e-03, -4.9079e-03, -2.8198e-03,  5.1973e-03,  1.0232e-02,
        -2.3325e-03, -9.2808e-03,  1.0562e-02,  3.2179e-03,  8.2825e-03,
         8.2769e-04,  3.3401e-03, -7.1282e-04, -2.7989e-03, -6.4135e-04,
        -7.5627e-03, -2.5800e-03,  1.8940e-03, -8.1183e-04, -2.8439e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.0176e-03,  7.5460e-01,  1.7706e-03,  4.3370e-02, -1.5809e-02,
        -1.0796e-02, -1.6310e-02,  1.1486e-02,  1.8080e-05, -1.1186e-02,
        -9.0177e-03,  8.8079e-03, -1.0769e-02,  9.0116e-03, -1.5937e-03,
         1.9597e-02,  3.3685e-03, -1.2100e-02, -2.3818e-03,  4.3776e-03,
        -7.1507e-03, -3.6668e-02, -3.7921e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.9008e-02,  6.6588e-01, -4.6281e-02, -9.1849e-02, -8.5954e-03,
         3.3137e-02, -4.9447e-02, -5.4059e-03, -9.7446e-03, -1.6597e-02,
         3.5683e-03,  4.8657e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([-2.6549e-02,  7.0839e-01,  6.9817e-02,  8.6278e-04, -6.9058e-03,
         1.5350e-02, -8.0800e-03, -1.1741e-03,  1.1060e-02, -1.4773e-02,
        -2.6288e-02, -3.3081e-04, -2.2732e-02,  4.1632e-03, -3.9480e-03,
        -4.4611e-03, -2.7150e-03, -1.6759e-02, -5.1160e-03, -2.2547e-02,
        -2.2856e-02, -2.1512e-05,  5.1037e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4987e-02,  7.2625e-01,  1.7569e-02, -1.9001e-02, -8.9421e-04,
        -7.5413e-03,  3.3253e-03, -2.3879e-03, -1.2489e-02,  8.5036e-03,
         6.2294e-03, -3.9347e-03,  1.2785e-03,  2.7353e-03, -6.1358e-03,
        -4.0073e-03, -8.3690e-03, -1.8458e-02, -1.1389e-02,  3.5902e-03,
         9.2201e-03, -6.3613e-03, -8.9911e-03,  1.2625e-03, -4.2782e-03,
         1.8314e-04, -2.7003e-03,  2.8774e-03, -7.1983e-03, -4.5566e-04,
        -4.0438e-03,  1.1881e-02,  2.1475e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.9214e-03, -8.0622e-01, -3.0838e-02,  1.2286e-02,  1.4018e-02,
         1.5360e-02,  8.8153e-03,  6.8418e-03, -1.1534e-02,  1.2111e-03,
        -8.6987e-03, -1.1532e-02,  1.2113e-02, -4.5473e-03, -1.9577e-02,
         7.1541e-03,  1.0058e-02,  1.0963e-02, -3.1252e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0404,  0.7619, -0.0206, -0.0389, -0.0043,  0.0065,  0.0153, -0.0073,
        -0.0068,  0.0228, -0.0094,  0.0107,  0.0089,  0.0461,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4394e-02,  8.4174e-01,  4.0273e-02, -1.2163e-02, -9.2343e-03,
        -6.0712e-03,  8.0713e-03,  5.4304e-03,  1.6980e-02,  4.5455e-04,
         3.6179e-03, -1.3810e-03, -8.9467e-04, -1.4669e-02,  1.4854e-03,
         3.4122e-03, -2.6840e-03,  5.0584e-04, -2.6601e-03, -3.5555e-03,
        -2.5203e-03, -7.8005e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7019e-02,  7.7288e-01, -1.2897e-02, -1.1950e-02, -2.7575e-03,
         1.5524e-02,  2.3254e-04,  7.6907e-03, -8.9407e-03, -1.2122e-02,
         5.6362e-04, -1.5282e-02,  8.4557e-03, -9.5899e-04, -2.8178e-03,
        -4.1453e-03, -6.8319e-03, -1.0933e-02,  1.6188e-03,  7.7350e-03,
         1.4967e-03,  3.9716e-03, -9.2023e-03, -7.8014e-03, -9.4093e-03,
        -2.2270e-03,  4.1496e-03, -9.6846e-04, -4.1556e-03, -1.5262e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1127,  0.5541,  0.0309, -0.0514, -0.0162,  0.0454,  0.0191, -0.0313,
         0.0068,  0.0026, -0.0237,  0.0145, -0.0086,  0.0151,  0.0118,  0.0102,
         0.0456,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1485,  0.7270,  0.0342, -0.0100, -0.0014,  0.0137, -0.0275,  0.0154,
        -0.0062, -0.0106, -0.0053,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4546e-02,  7.4281e-01,  7.1115e-03,  1.9787e-02, -1.5433e-02,
        -3.3184e-02, -3.7041e-03,  4.5653e-03,  5.1751e-04,  1.1779e-02,
        -1.6318e-04, -4.4803e-03, -8.6856e-03,  9.7188e-03,  2.5495e-02,
        -7.3712e-03, -7.0649e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.4815e-02, -7.2977e-01, -2.3832e-03, -4.6396e-02, -2.7704e-02,
        -1.5197e-02, -1.4576e-04, -5.3974e-03, -1.5393e-03,  5.5209e-03,
         2.9198e-04,  1.7814e-03, -9.2865e-06,  6.6320e-03,  1.9219e-03,
        -4.0739e-03,  8.1538e-04, -5.6079e-03,  5.0826e-03,  4.7810e-03,
        -4.7465e-03,  1.5724e-03, -9.6932e-03,  6.6539e-05, -3.9822e-02,
         4.2294e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1511e-02,  6.9404e-01,  4.7381e-02,  1.4819e-03, -2.1620e-02,
        -6.2467e-04, -1.3793e-02, -2.2838e-02, -2.2834e-02,  2.1521e-03,
         5.2245e-03,  5.4992e-03, -3.4293e-03, -3.7318e-03, -2.2850e-03,
         2.2370e-03, -5.9848e-03, -1.4333e-03, -2.7992e-03, -4.3659e-03,
        -6.2406e-03, -1.1884e-03, -2.0117e-03, -5.6210e-03,  1.4112e-02,
        -3.4638e-03, -4.2090e-03,  1.4099e-03, -5.9683e-03, -3.0509e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1604e-02,  7.8830e-01, -1.4288e-02, -9.8277e-03,  3.6900e-02,
        -8.7499e-03,  7.9297e-03, -7.7236e-03,  8.1127e-03,  5.2529e-03,
        -3.3442e-03, -1.2508e-02, -3.4710e-03, -1.1238e-03,  3.9213e-04,
         2.8841e-03,  1.8212e-02, -8.6391e-03,  7.3349e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 7.9239e-03,  8.1820e-01, -1.7250e-03,  8.7719e-03, -2.0120e-03,
         3.0270e-03,  1.4645e-02, -7.8620e-03,  5.9812e-03, -7.5321e-04,
         4.5714e-03, -5.1359e-04,  3.0079e-05,  5.4866e-03, -1.6562e-02,
        -3.3582e-03,  9.7760e-03,  2.3219e-02,  2.0641e-03,  4.8080e-03,
         1.9035e-03, -1.9629e-03,  1.3951e-02,  8.1211e-03, -6.6053e-03,
        -8.7749e-03, -1.7389e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0401,  0.7579, -0.0173, -0.0555,  0.0188, -0.0106,  0.0050,  0.0085,
         0.0111,  0.0082,  0.0053,  0.0156,  0.0427,  0.0032,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1566e-02, -7.4986e-01,  1.5394e-02,  2.5396e-02,  2.6231e-03,
        -7.5324e-03,  2.2282e-02, -4.0573e-03, -1.1391e-03,  3.8174e-04,
         1.6924e-03,  1.5818e-02,  2.2412e-03, -8.2178e-03,  5.9422e-03,
         1.6455e-02,  9.7624e-03, -1.5821e-04,  6.1317e-03,  3.4614e-03,
         4.8755e-03,  1.1141e-03,  7.9516e-03, -5.3557e-03,  6.8817e-03,
         3.3706e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5224e-02,  6.4735e-01,  3.2735e-02,  9.8799e-03,  1.0898e-02,
        -4.9595e-02,  1.7502e-02,  1.0996e-02, -1.2454e-02,  5.1953e-04,
        -8.1959e-03,  2.0688e-03,  1.6452e-01,  1.8071e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1104e-02,  7.1666e-01, -5.1636e-02, -8.9775e-03,  7.8806e-04,
        -7.0605e-03,  8.9000e-03, -5.6788e-03,  2.5460e-03,  4.6699e-03,
         5.8139e-03, -8.9772e-03,  1.1036e-03, -1.2794e-02,  1.5873e-03,
         8.2240e-04, -9.4343e-03, -6.7643e-03,  4.0200e-03, -1.1259e-02,
         1.1878e-03, -5.7982e-03,  4.9942e-03,  4.5454e-03,  1.0237e-02,
        -3.1665e-03,  2.4332e-03,  6.8955e-04, -1.1318e-02, -2.8082e-04,
         1.9050e-02,  3.5703e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0203,  0.7098,  0.0165, -0.0661, -0.0164, -0.0082, -0.0115, -0.0161,
         0.0027, -0.0130, -0.0058, -0.0017, -0.0083,  0.0019, -0.0057,  0.0075,
        -0.0036, -0.0187,  0.0068,  0.0025, -0.0088, -0.0107, -0.0029, -0.0144,
        -0.0201,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0628,  0.6545,  0.1312,  0.0748,  0.0242, -0.0525,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.5501e-02,  7.3445e-01,  3.8486e-02, -5.7826e-03,  1.6423e-03,
        -2.0972e-03, -4.8416e-03, -1.4772e-03,  4.7226e-03,  3.2053e-03,
        -3.8381e-03,  7.2183e-03,  1.1989e-04,  6.2252e-03, -1.5745e-02,
        -4.3836e-04, -7.0952e-03, -6.0658e-04, -6.5836e-04,  6.2375e-03,
        -9.8537e-03, -4.5167e-03, -1.0152e-02, -5.8122e-03, -1.1014e-03,
        -2.6957e-03, -4.5221e-03, -8.0278e-03,  2.0166e-04,  6.6299e-03,
         4.4012e-03,  2.9950e-03, -3.4017e-04, -3.2111e-03,  2.0907e-03,
        -8.6403e-03,  4.4143e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7329e-02, -7.4857e-01, -2.9041e-02, -5.3637e-02, -1.0813e-02,
        -1.9137e-02, -6.6620e-03, -5.6436e-03,  3.2144e-03,  3.5640e-03,
         6.5834e-03,  6.1748e-04,  5.5558e-03,  2.7935e-03, -6.4730e-03,
        -7.9786e-03,  3.2025e-03, -3.6141e-03,  7.7674e-04,  7.4303e-05,
         7.7351e-03,  3.5344e-03,  1.3362e-02,  1.8338e-03, -1.8188e-03,
        -2.3508e-03, -9.5174e-03, -2.5595e-03, -1.0372e-03,  1.5311e-03,
        -5.4981e-04, -1.5009e-03,  1.1237e-03, -1.8387e-03,  1.4427e-02,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0026,  0.7414, -0.0499,  0.0551,  0.0173, -0.0049, -0.0064,  0.0062,
        -0.0043,  0.0033,  0.0262, -0.0012,  0.0157, -0.0355,  0.0302,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0486e-02, -7.8283e-01, -2.1811e-02,  3.4838e-02, -1.0561e-02,
         1.1693e-02, -1.0928e-02, -5.4320e-03, -1.6172e-02,  4.1954e-03,
        -3.2188e-04,  1.7857e-03,  5.3894e-03,  1.0631e-03, -5.7691e-03,
         1.0167e-02, -8.3897e-03,  1.2745e-06, -2.2399e-03, -4.7867e-03,
         2.0479e-03, -4.2784e-04,  3.0172e-03, -8.2999e-03, -1.0988e-02,
        -5.0783e-03,  5.9514e-03, -4.4377e-03, -3.9028e-04, -2.3696e-03,
        -2.0500e-03,  5.6065e-03, -4.7878e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0314, -0.6674, -0.0072,  0.0019,  0.0007,  0.0591,  0.0051, -0.0049,
         0.0206, -0.0091, -0.0157, -0.0305, -0.0076, -0.0056,  0.0123,  0.0103,
         0.0072,  0.0035,  0.0216, -0.0045,  0.0152, -0.0022,  0.0007, -0.0032,
        -0.0025,  0.0094,  0.0137, -0.0175,  0.0096,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 0.0809,  0.7062, -0.0220, -0.0071, -0.0150, -0.0119,  0.0054, -0.0116,
        -0.0052, -0.0249, -0.0036, -0.0025,  0.0071,  0.0443, -0.0523,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7794e-02, -8.0560e-01, -1.2766e-02,  8.3752e-03,  6.3637e-03,
         1.3307e-02,  1.6487e-03, -5.6579e-03, -7.0221e-03, -6.3250e-05,
         4.5853e-03,  6.2388e-04,  1.9931e-03,  1.0382e-03, -3.3295e-04,
         5.0649e-03,  6.1460e-03,  9.2862e-04,  2.1873e-03,  2.0783e-03,
         2.9291e-03,  1.2427e-03,  3.5937e-03, -7.6601e-04,  3.0417e-03,
        -2.2855e-03,  2.1656e-03, -7.2546e-03,  3.9599e-03,  2.4566e-04,
        -1.3332e-03,  4.0904e-03,  7.9676e-03,  2.4698e-03,  2.0421e-03,
        -4.4519e-03, -1.5484e-03,  4.2968e-03,  8.4779e-04,  1.5109e-03,
         2.9112e-03, -1.7057e-03, -4.4085e-03,  3.9779e-04,  1.3835e-03,
         1.5732e-02,  2.2780e-03, -3.8823e-03, -5.6843e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9779e-02,  7.4173e-01, -5.3185e-02,  3.1608e-03,  2.5688e-02,
         7.3523e-03, -1.0333e-02,  1.6078e-02,  3.8654e-03,  3.5130e-03,
        -6.9023e-03, -6.9966e-03, -1.4552e-03,  5.7649e-03,  2.1108e-03,
        -1.6023e-03,  1.4943e-02, -9.0237e-04,  6.3071e-03, -1.0975e-03,
        -5.8624e-03,  5.6332e-03,  3.0305e-03,  5.4780e-04,  9.7389e-04,
         2.6801e-03, -9.3160e-04,  5.4423e-03, -1.7259e-03, -3.5942e-03,
         5.0005e-03,  4.4700e-03,  2.3605e-03, -1.2353e-02, -1.2626e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0632,  0.7151, -0.0506, -0.0193,  0.0225, -0.0086,  0.0045,  0.0011,
         0.0028, -0.0015,  0.0044, -0.0025, -0.0107, -0.0040,  0.0115,  0.0112,
        -0.0019, -0.0486,  0.0160,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4603e-03, -8.1196e-01,  2.0477e-02, -2.4496e-02, -3.0524e-03,
        -1.7291e-02, -6.9387e-03, -7.5855e-04,  1.1302e-02,  9.2614e-03,
         2.0587e-02,  8.9358e-03, -2.9341e-02, -2.2364e-02,  7.7770e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.2371e-02, -7.5758e-01,  5.8372e-03,  3.2081e-03, -3.6932e-03,
         3.1676e-02,  3.4096e-02, -3.9861e-03,  1.6053e-02, -6.4777e-03,
         8.2982e-03,  1.6048e-03,  1.7841e-02,  7.6116e-03,  1.0534e-04,
         1.3717e-03,  4.4148e-03,  2.7083e-03, -1.3973e-03, -1.7666e-03,
         2.0703e-02,  7.2001e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.3086e-02,  7.3325e-01,  1.6956e-02,  3.3429e-03,  9.3312e-03,
         2.5938e-03, -2.5331e-02,  1.8453e-03,  2.7318e-03,  1.4207e-02,
         2.0422e-03,  1.0813e-02,  9.6370e-03, -1.3362e-02, -9.7506e-04,
        -1.5383e-03, -1.9685e-03,  6.5621e-03,  2.8183e-03, -5.8526e-04,
         6.2362e-03, -1.7802e-03,  7.4825e-03,  1.6307e-02,  1.4972e-02,
         4.9072e-03,  4.9277e-03, -1.2696e-02, -4.5887e-03,  3.1308e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0320,  0.7586, -0.0031, -0.0426,  0.0255, -0.0022, -0.0313, -0.0095,
         0.0752, -0.0067,  0.0133,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5833e-02, -8.4463e-01, -1.6476e-02, -4.7220e-03, -5.5415e-03,
        -1.6404e-02,  1.4360e-02, -4.9840e-03, -1.8905e-03, -3.5186e-03,
         1.3283e-02, -1.1211e-02, -7.8075e-03, -4.0474e-03, -3.6876e-03,
        -2.4896e-03, -2.9521e-03,  4.8029e-04, -2.5857e-03,  8.2696e-03,
        -5.5533e-04, -1.4274e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3904e-02,  7.2085e-01,  4.8726e-02, -2.8747e-02,  1.1407e-02,
         6.5484e-03, -7.4381e-03,  1.4481e-03, -7.9981e-03, -8.7421e-03,
        -2.3961e-03, -1.0759e-02, -1.1544e-02, -5.7488e-03, -5.2330e-03,
        -1.0909e-02,  2.8791e-04,  1.1197e-03, -4.9307e-03, -1.7294e-03,
         3.5115e-04, -4.0510e-03, -1.7983e-02, -9.5338e-04, -6.6759e-04,
        -2.9433e-03, -2.7427e-03, -4.3034e-03, -1.0665e-02, -1.0114e-03,
         2.8584e-03,  1.1650e-03,  9.8400e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.4479e-03,  7.2765e-01,  7.2660e-02, -4.6099e-03, -1.1180e-02,
         2.2451e-02, -3.9846e-03, -3.2807e-03,  4.4420e-03, -4.8675e-04,
        -6.8558e-03, -4.4016e-03,  4.3376e-04, -1.5130e-02, -2.2785e-02,
         5.0669e-04,  3.1982e-03,  8.5563e-03, -3.2676e-03, -2.3591e-03,
         1.3352e-02, -8.8702e-03, -2.8790e-03, -5.7481e-03,  4.9476e-03,
         6.3259e-03,  9.3617e-03, -2.0833e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.8437e-02, -7.6873e-01, -5.4845e-02, -2.0637e-02,  6.6880e-03,
        -1.1919e-02,  6.7176e-03, -9.9576e-04,  1.6484e-02, -1.6089e-03,
        -6.7917e-04, -8.0689e-04, -2.0936e-04,  5.0508e-03, -1.3886e-03,
         4.3409e-03, -4.6720e-03, -3.0279e-03,  4.6762e-03, -9.7438e-03,
         2.4038e-03,  7.1688e-05,  1.8495e-03, -2.0496e-02,  3.5235e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-1.4434e-02,  7.5559e-01,  1.9270e-02,  6.7743e-03,  4.8606e-03,
         1.0106e-02,  2.8773e-04, -7.6505e-03,  3.5923e-03, -1.5918e-03,
        -5.8519e-03, -8.8849e-05, -4.2499e-03, -6.4498e-03,  8.2640e-03,
        -4.9916e-02, -6.7794e-03,  4.4687e-04, -4.4156e-03,  2.4200e-03,
         3.9285e-04, -5.4831e-03, -2.0087e-03,  4.7195e-03,  1.3995e-02,
         2.2382e-03,  3.7545e-03,  5.2140e-03, -9.8188e-03, -7.9517e-03,
        -4.3327e-04, -3.1984e-03,  1.7820e-03, -7.1741e-03, -3.6558e-03,
         2.3677e-03, -6.9716e-03,  4.6992e-04, -2.5551e-03,  2.4576e-03,
         3.1803e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0129, -0.7243, -0.0688,  0.0101, -0.0324, -0.0069,  0.0013, -0.0025,
         0.0068,  0.0062,  0.0029, -0.0049, -0.0092,  0.0024, -0.0040,  0.0091,
        -0.0022,  0.0033,  0.0057, -0.0030,  0.0016,  0.0229,  0.0037,  0.0244,
         0.0284,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0145, -0.7272, -0.0074,  0.0236, -0.0222,  0.0071, -0.0120, -0.0045,
         0.0217,  0.0203,  0.0116, -0.0103,  0.0137,  0.0062,  0.0138, -0.0034,
        -0.0195,  0.0134,  0.0038, -0.0333, -0.0103,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0116, -0.8133, -0.0066, -0.0097,  0.0112, -0.0067, -0.0078,  0.0116,
         0.0020, -0.0259,  0.0083,  0.0151, -0.0023,  0.0024,  0.0062,  0.0063,
        -0.0151,  0.0379,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5993e-02,  6.4521e-01,  5.7838e-02,  7.0456e-04,  1.9421e-02,
         1.2520e-02,  6.6534e-03,  2.1485e-02, -8.4909e-03, -2.3690e-04,
         1.9197e-03, -1.0091e-02, -5.5659e-04,  3.2692e-04,  3.2210e-04,
        -1.1861e-03, -1.5382e-02, -4.5942e-03, -3.6367e-03, -1.5083e-03,
        -1.1976e-02, -6.3791e-03, -3.2904e-03, -7.5502e-04, -1.0050e-02,
         4.0206e-03, -7.2089e-03, -1.2294e-05, -3.3626e-03, -7.3318e-03,
        -2.7912e-03,  1.8324e-03, -1.8125e-03, -2.2718e-03, -7.0206e-03,
         2.4439e-03, -1.0896e-02, -4.8464e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.5880e-02, -7.3328e-01,  6.6998e-04,  2.0267e-02, -2.2648e-02,
        -8.7633e-03,  3.1200e-03,  1.0208e-02, -6.5714e-04, -8.2284e-03,
         2.7524e-05,  2.5161e-03, -9.0518e-04, -4.1909e-03,  3.5854e-03,
         6.4380e-05,  1.5128e-02,  8.1475e-02, -1.0302e-02,  1.8081e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0833e-02, -8.3425e-01,  8.4937e-03, -1.6647e-02, -2.1357e-02,
        -9.7628e-03, -1.1594e-02, -2.5156e-02, -6.6020e-03,  3.6880e-04,
        -1.2649e-02,  5.0856e-03, -1.1837e-02, -5.3586e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2428e-04, -7.4316e-01,  3.6605e-02,  2.4540e-02, -2.4071e-02,
        -1.0037e-02,  1.7049e-02,  6.4426e-04, -6.6700e-03, -1.1032e-02,
        -6.9346e-03, -6.3613e-03, -1.3352e-03,  1.8350e-03, -1.1844e-03,
         1.1130e-02, -5.8308e-03,  2.0425e-03,  1.2260e-03, -2.2867e-03,
        -8.5216e-03, -1.3665e-02, -4.3827e-03, -1.5802e-02,  1.5093e-03,
         6.9744e-03, -1.7646e-03, -4.0280e-03, -1.1198e-03, -5.7802e-03,
        -4.2615e-03,  1.8095e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0145, -0.7115, -0.0463, -0.0629, -0.0107,  0.0152, -0.0238, -0.0036,
         0.0058,  0.0105,  0.0027,  0.0098,  0.0008, -0.0012, -0.0106,  0.0079,
        -0.0008, -0.0048,  0.0243,  0.0323,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2960e-02,  6.6722e-01,  7.7542e-03, -2.6091e-02, -5.5951e-03,
        -3.6570e-04, -1.4627e-03, -7.2916e-04,  5.0564e-03, -1.2455e-02,
         7.7668e-03, -3.8756e-03, -1.0627e-02, -4.6219e-03, -1.3107e-03,
        -1.1247e-02, -1.4833e-02,  4.3308e-03, -9.0488e-03, -3.9153e-03,
        -1.8344e-03, -8.3665e-03, -5.6142e-03, -1.0265e-02,  3.8828e-03,
        -3.5402e-03, -7.9226e-03, -1.6799e-02,  4.2298e-03,  9.8377e-05,
         1.2423e-02,  2.8287e-03,  2.1895e-02,  4.7037e-03, -2.4642e-03,
         1.3543e-03,  1.1841e-03, -8.2280e-04,  3.2686e-04,  2.0859e-03,
        -1.1803e-02,  1.4561e-02,  2.5172e-03,  8.3761e-03, -1.4565e-02,
        -8.8504e-03, -1.5419e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7220e-02, -7.8756e-01, -5.6647e-03,  1.8431e-02, -4.5225e-02,
         1.9535e-03, -1.0040e-02,  1.2319e-02,  1.6368e-02, -4.1468e-04,
         2.2374e-02, -1.7468e-02, -8.5945e-03,  4.1819e-03, -2.2184e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.8980e-03, -6.5204e-01,  1.6607e-02,  1.0727e-02,  9.4901e-03,
         1.7906e-02, -1.0538e-02, -2.4117e-02,  6.8041e-03, -3.4336e-03,
         9.8600e-03,  3.0856e-03,  1.5802e-02, -1.3083e-02,  1.1891e-02,
        -6.5924e-03, -6.6700e-03, -1.7022e-03, -5.5878e-03,  6.4167e-03,
        -4.7924e-03,  2.9995e-03, -7.2132e-04, -1.2957e-02, -3.7621e-03,
         5.3533e-03, -2.1877e-03,  3.7973e-02, -8.9433e-04, -7.2666e-03,
         4.0871e-03,  1.1071e-02,  8.3230e-03,  4.6111e-03,  2.8873e-03,
        -4.2451e-03,  1.2615e-03,  1.5039e-03,  1.6032e-02,  1.1512e-03,
        -3.6095e-03,  5.7637e-04,  1.8725e-03,  8.2714e-04,  1.3716e-03,
         6.2070e-03,  5.9380e-03,  4.2675e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-6.7008e-03, -7.5772e-01,  7.7987e-03, -3.5136e-04,  1.0045e-02,
         1.2761e-02, -3.1164e-02,  5.6573e-03,  7.9530e-03,  2.7213e-03,
         2.9957e-03, -5.4845e-04,  1.0060e-02,  2.9430e-03,  1.1531e-02,
         1.2079e-03,  3.4674e-02,  1.1499e-02,  2.8740e-03,  8.0912e-03,
         9.6188e-03, -7.4052e-03, -1.1080e-02, -1.5643e-02,  2.5411e-03,
         2.4412e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.2502e-02, -7.5446e-01, -1.2474e-02, -1.9455e-02, -6.7666e-03,
        -2.5105e-02,  1.2306e-02, -7.4358e-03, -1.2175e-02, -5.7243e-03,
        -7.3052e-04,  3.5741e-03,  4.9072e-03, -2.3986e-03,  6.0107e-03,
         3.7970e-03, -1.0702e-02,  4.7342e-03,  9.6469e-04, -1.6615e-02,
         1.7159e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4638e-02,  6.2465e-01,  5.5571e-02,  5.4826e-02,  7.3348e-03,
        -1.1186e-02, -1.1945e-02,  7.2137e-04, -3.9274e-03, -6.6552e-03,
        -1.3437e-03, -1.8901e-03, -6.8893e-03, -2.6440e-02, -2.2616e-02,
         2.2895e-03,  1.1753e-03, -2.4538e-02,  1.2020e-02, -1.1355e-02,
         1.9100e-03,  3.5878e-03,  1.4567e-02, -5.9490e-03,  3.0999e-03,
        -3.4160e-04, -2.4561e-03,  1.9684e-03, -2.7085e-03, -2.9393e-03,
         2.3138e-03, -5.1013e-04, -1.3516e-03, -3.7464e-03, -2.0373e-02,
        -2.0167e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.6480e-02,  6.6131e-01, -6.1622e-02,  1.3186e-02, -1.3895e-02,
        -3.0400e-03, -2.3825e-02, -1.5713e-04, -3.5364e-02, -2.8290e-03,
         1.0460e-02, -1.6528e-02, -1.6082e-02, -1.5252e-02, -5.1818e-03,
        -1.1575e-02,  2.5139e-04, -6.8556e-03,  1.7236e-02, -1.8873e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9374e-02, -7.2526e-01, -5.9327e-02,  5.0416e-03, -1.6054e-02,
         2.1232e-03, -4.7150e-03,  2.7622e-02, -4.8672e-04, -2.5710e-03,
        -1.3943e-02,  8.4144e-03,  1.2739e-02,  3.2017e-03, -1.0058e-02,
        -7.0375e-03,  1.7414e-02,  4.2267e-03,  1.5103e-02, -8.2032e-03,
         7.0816e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.9641e-02, -8.1353e-01, -3.3818e-02, -8.1372e-03,  1.5664e-03,
         4.8090e-03, -3.5775e-03, -2.5439e-03, -1.3127e-02, -1.3019e-06,
        -1.5426e-02, -3.8068e-03, -2.0014e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6935e-02, -7.1153e-01, -6.3555e-03, -3.6964e-03,  7.9150e-03,
         1.0012e-02,  5.6637e-03, -1.0085e-02,  3.2777e-02,  6.3782e-03,
         1.2312e-02,  2.6730e-03, -2.2934e-03,  1.1052e-03, -1.1487e-02,
        -1.6402e-02, -6.4958e-03, -7.0406e-03, -1.1830e-03,  1.0359e-02,
        -1.2768e-02, -7.2140e-03,  7.4782e-06,  5.1172e-03, -1.0192e-03,
        -5.9989e-03,  1.4572e-02, -4.9845e-03,  2.7081e-04,  2.7990e-03,
         3.7740e-03, -3.7859e-03, -6.5160e-03,  1.9892e-03, -1.0185e-02,
        -3.8667e-03, -1.2432e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5773e-02,  7.9128e-01,  2.9958e-02, -8.1176e-04, -1.8125e-02,
         1.4360e-03, -1.7878e-03, -3.1169e-04, -4.3741e-03, -9.9217e-03,
        -1.2642e-02, -2.9133e-03,  1.7418e-02,  3.1900e-03, -1.2423e-02,
         8.7035e-04,  2.0712e-02,  1.0632e-02, -1.9465e-03, -2.6860e-03,
        -1.0415e-02,  1.0642e-02, -6.4164e-03,  3.3183e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.8409e-02,  6.6889e-01,  4.6928e-02,  6.0931e-03,  1.6171e-02,
         2.9695e-03, -4.7650e-03,  7.1361e-04, -6.2597e-03, -3.6510e-02,
        -6.1828e-04, -2.6576e-02, -4.6952e-03,  1.4600e-03, -2.4380e-03,
        -5.2629e-03, -1.6720e-02,  6.7809e-03,  5.5995e-03,  5.4233e-04,
         2.3610e-03,  3.2242e-03,  1.4116e-03, -4.7427e-05, -1.9713e-04,
        -1.8140e-04, -2.1866e-03,  2.2952e-03,  5.5889e-03,  4.8905e-03,
        -1.3215e-03, -1.7894e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8854e-02,  7.1248e-01,  3.3982e-02, -1.3563e-02,  1.2854e-02,
        -2.3854e-02, -7.2675e-03, -1.6424e-02,  8.4134e-03,  1.1513e-02,
         1.6701e-03,  3.1776e-03, -3.6520e-04, -1.3047e-03,  2.1098e-03,
        -3.2845e-03, -6.1413e-03, -1.9249e-03,  9.4689e-03,  2.5459e-03,
        -9.5588e-04,  1.2442e-02, -1.0889e-02, -1.4663e-03,  1.2634e-02,
         1.5870e-02,  1.9679e-02, -5.1187e-03,  8.7789e-04,  3.0532e-03,
         1.5813e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0518,  0.7716, -0.0410,  0.0146,  0.0060, -0.0076, -0.0125,  0.0021,
        -0.0040,  0.0094,  0.0147,  0.0015, -0.0028,  0.0015,  0.0075,  0.0052,
         0.0091, -0.0030, -0.0068,  0.0035,  0.0012,  0.0111,  0.0116,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5824e-02, -7.8809e-01, -3.2228e-02,  1.4823e-02, -1.5241e-02,
         6.3123e-03,  1.1997e-02, -3.7962e-03, -5.7109e-03,  9.5299e-04,
         2.6024e-03,  2.1668e-02,  1.0164e-02, -9.4592e-03,  4.0793e-03,
         6.8955e-04,  6.9506e-03,  6.3761e-03,  1.7650e-04,  4.1326e-04,
         6.6520e-03,  8.2224e-03, -1.2944e-03, -7.2683e-03,  4.2585e-03,
         1.1786e-03, -2.2904e-03,  1.1277e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-5.6674e-02,  7.5504e-01,  3.2655e-03,  3.4110e-03,  7.1973e-03,
         1.5438e-02,  2.3900e-03, -8.7943e-03, -4.4111e-03, -1.7755e-02,
         2.7661e-02, -8.0582e-03, -1.0771e-03, -9.1295e-03, -8.8038e-03,
        -8.9349e-03, -2.7736e-03,  1.0184e-02,  1.5798e-03,  4.4140e-04,
         1.2282e-02,  1.2620e-03, -7.0842e-03, -4.0555e-03,  9.6236e-04,
         2.5108e-03,  7.5950e-03, -2.9014e-03, -8.3289e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.8413e-03, -7.8593e-01,  2.4516e-02,  3.8914e-04, -7.5767e-02,
        -1.0681e-02,  5.9897e-03,  4.1105e-03, -3.4793e-03, -7.9311e-03,
         1.5002e-02,  8.8709e-04,  1.0654e-02,  3.8308e-03, -6.6274e-03,
        -3.3879e-03,  4.2111e-04, -7.1383e-03, -2.0126e-03,  6.4648e-03,
        -1.0721e-03, -4.3685e-03,  8.7371e-03,  1.7656e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0832e-01,  6.4423e-01,  1.2766e-02, -3.4002e-03,  2.5630e-02,
         1.2396e-02, -6.3315e-03,  3.5406e-04,  6.0443e-03, -5.4964e-04,
        -1.4828e-02,  1.6867e-03,  1.0465e-03, -7.7972e-03,  1.1045e-04,
        -6.8243e-03, -1.3546e-02, -4.4905e-03, -5.7833e-03, -1.0604e-02,
         4.3324e-03,  3.7116e-03, -2.6164e-03,  5.7036e-03,  5.3322e-03,
        -2.8593e-03, -1.0689e-02,  1.5890e-03, -8.6858e-04, -3.1691e-03,
        -1.6374e-02,  1.3266e-02,  1.2207e-02, -3.0540e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.3210e-03, -8.3397e-01,  3.8036e-02, -8.3229e-03, -7.8961e-04,
        -2.5312e-04, -2.0442e-02, -1.2871e-02,  5.2760e-03, -6.9047e-03,
         1.0680e-02,  7.4092e-04,  1.0562e-03,  7.5234e-03,  2.3910e-03,
        -2.4632e-03,  1.8401e-03,  6.2286e-03,  1.0474e-03,  7.8425e-03,
         8.2215e-03, -1.8775e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7906e-02,  7.7357e-01,  2.2373e-02, -1.8388e-02,  5.1913e-03,
         8.0486e-04,  1.8942e-03,  2.2937e-03, -4.3282e-03,  1.1236e-02,
         1.4398e-02, -7.5891e-04, -1.2459e-02, -1.0720e-02, -2.0444e-02,
        -3.1023e-02, -1.0997e-02,  2.1216e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0533,  0.7999, -0.0079,  0.0104, -0.0059,  0.0214, -0.0062,  0.0015,
        -0.0089,  0.0082,  0.0092,  0.0019, -0.0065, -0.0089,  0.0018, -0.0015,
         0.0021, -0.0017,  0.0009,  0.0045,  0.0028, -0.0045, -0.0302,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5713e-02, -8.0994e-01, -1.5842e-02,  3.0542e-03, -1.2562e-03,
         2.2455e-03,  3.5583e-03,  8.2541e-03,  1.3688e-02,  1.1322e-02,
         1.7445e-02,  4.3441e-03,  8.9632e-03, -6.4001e-03, -4.9824e-04,
         1.3761e-02,  4.3719e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0947e-01,  7.0407e-01,  2.5470e-02,  1.5600e-04, -2.6354e-03,
         1.1198e-02,  1.1821e-02, -2.5903e-02, -2.1645e-03,  7.2819e-03,
        -5.0328e-03, -1.0406e-02,  7.6467e-03,  1.4826e-02,  5.3659e-03,
         4.2948e-02, -5.4285e-03,  8.1677e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0709, -0.7003, -0.0106,  0.0122,  0.0165,  0.0072,  0.0008, -0.0034,
         0.0013, -0.0354,  0.0152,  0.0126, -0.0045,  0.0034,  0.0068, -0.0213,
         0.0059, -0.0059, -0.0212, -0.0080,  0.0024, -0.0055,  0.0021, -0.0039,
         0.0033, -0.0196,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9066e-02,  7.2622e-01,  6.5597e-02, -1.5289e-02, -1.1785e-02,
        -3.1499e-02,  8.0658e-03, -2.8041e-03,  1.7189e-05, -8.1330e-03,
        -3.3710e-03, -1.9145e-03, -5.3007e-03,  4.1929e-03, -3.8848e-03,
         8.8671e-04, -3.2952e-03,  1.2565e-03, -3.4187e-03, -6.0857e-03,
        -2.9694e-03,  8.0541e-03,  4.5006e-03,  9.0797e-06,  1.6976e-03,
         1.4008e-03, -4.0058e-03, -3.1708e-04,  1.5253e-02,  2.7876e-03,
        -2.6920e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0410, -0.7837, -0.0419,  0.0161,  0.0031, -0.0093,  0.0096,  0.0028,
        -0.0032,  0.0245,  0.0081,  0.0039, -0.0149,  0.0015,  0.0157,  0.0049,
        -0.0021,  0.0138,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0301,  0.7537, -0.0296, -0.0560, -0.0336,  0.0695, -0.0199, -0.0076,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 4.0092e-02,  7.6134e-01,  1.2615e-02, -2.5370e-02, -1.0441e-02,
         2.5062e-02,  4.4849e-03,  8.3654e-03, -3.5714e-03,  8.4184e-03,
        -1.0486e-03,  1.7462e-03,  4.1277e-03,  2.3982e-03,  1.9252e-03,
         3.4251e-03,  5.2339e-03,  2.5767e-03,  1.1442e-04,  3.0465e-04,
        -3.2529e-03, -3.7263e-03, -7.2574e-03, -4.1534e-03, -3.8279e-03,
        -1.8337e-03, -5.2563e-03,  4.1577e-03, -5.6990e-04,  1.6582e-03,
         6.7762e-03, -1.0419e-03,  3.8850e-03,  6.1068e-03,  1.1808e-03,
         9.0209e-03, -5.2530e-03,  4.7643e-03,  3.6218e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.2848e-03, -7.3612e-01,  5.3505e-02,  5.2141e-02, -1.5753e-02,
         1.0316e-02,  6.6260e-03, -6.3391e-03,  4.5958e-03,  3.8886e-03,
         8.1794e-03, -7.1419e-03,  3.4786e-03,  1.5895e-02,  2.1883e-03,
         6.4965e-04, -7.0865e-03,  1.8569e-03, -1.7179e-04, -6.9114e-03,
        -4.2469e-03, -1.6787e-03, -3.6051e-03,  6.7167e-03, -2.2485e-03,
         5.1256e-03,  2.8647e-03,  2.1388e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0436, -0.8062, -0.0121, -0.0154,  0.0130,  0.0221, -0.0249, -0.0243,
         0.0046, -0.0022,  0.0097,  0.0052, -0.0167,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0852, -0.6211,  0.0311,  0.0377,  0.0165,  0.0247,  0.0237,  0.0033,
         0.0046,  0.0010,  0.0174,  0.0026, -0.0512,  0.0649, -0.0149,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0727,  0.6160, -0.0308, -0.0126,  0.0267,  0.0148, -0.0235, -0.0119,
        -0.0149, -0.0038, -0.0046,  0.0109,  0.0025,  0.0020,  0.0030,  0.0018,
        -0.0026, -0.0108,  0.0069, -0.0039,  0.0068,  0.0013, -0.0020, -0.0026,
        -0.0066, -0.0010,  0.0018,  0.0066,  0.0023,  0.0026, -0.0086,  0.0034,
         0.0070,  0.0067, -0.0010, -0.0080, -0.0023,  0.0017,  0.0015, -0.0014,
        -0.0084, -0.0038,  0.0062, -0.0008,  0.0009,  0.0035, -0.0247],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0800e-02,  7.7119e-01,  2.2480e-02,  1.7586e-03,  2.7565e-03,
        -2.9480e-02,  3.1204e-03, -6.0584e-03, -2.3180e-02,  6.0415e-03,
         4.3889e-03,  1.7608e-02, -3.2091e-04, -1.5129e-03, -6.9032e-04,
        -3.1488e-03, -3.9224e-03, -4.6116e-03,  2.7371e-03,  4.0615e-04,
         4.0865e-03,  4.7462e-04,  3.5189e-03,  6.0060e-03,  1.3544e-03,
         3.4038e-03,  1.1106e-02, -3.5239e-04, -8.3822e-04,  4.0980e-03,
         2.1584e-03, -9.6362e-04, -4.6708e-03, -7.3684e-05, -1.9509e-03,
        -5.5547e-04, -1.4811e-03,  2.3324e-03,  1.7208e-03, -1.3758e-03,
        -6.7270e-03,  2.4537e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6936e-03,  7.0280e-01,  6.5809e-02, -2.6301e-02, -4.7754e-03,
         2.3443e-03,  8.4499e-03, -6.2706e-03,  5.9970e-03,  1.5527e-02,
         8.3073e-04,  2.2789e-03, -1.0491e-02,  4.8326e-03,  1.0077e-02,
        -1.4093e-02,  2.7944e-02, -2.7960e-03,  5.9240e-03, -5.8766e-05,
        -5.1599e-03, -3.1968e-03, -3.3186e-03, -2.2676e-02, -4.6351e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2773e-02,  7.4276e-01,  8.1009e-03, -6.1423e-03, -1.8875e-02,
         5.6763e-03,  2.8781e-02,  1.2901e-02,  1.2624e-02,  1.3851e-03,
         1.3594e-03, -7.7961e-03,  3.2089e-03,  3.5505e-05, -8.1078e-03,
        -1.7110e-02,  1.3579e-02,  1.7276e-02, -1.7134e-02,  1.6934e-02,
         1.1757e-02,  9.3633e-03, -6.3216e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3625e-03,  7.2328e-01,  5.1749e-02, -1.9680e-02,  1.4057e-02,
         9.5532e-03,  3.8638e-03,  1.1748e-02, -1.5467e-02,  2.8986e-02,
         8.5554e-04, -3.3913e-03,  1.1963e-04, -6.4009e-03, -2.5736e-03,
        -8.6397e-03, -1.7230e-02, -9.7612e-03,  6.3566e-03, -7.7215e-03,
         7.4215e-03, -9.8733e-03, -3.5905e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5112e-01, -6.7397e-01, -1.2622e-02,  1.8718e-02, -5.3461e-03,
        -1.5167e-02, -1.1625e-02, -4.9840e-02, -1.4626e-02, -5.8363e-04,
         6.3184e-04, -3.9338e-02,  6.4135e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9631e-02, -8.1528e-01,  2.5904e-02,  4.4038e-02,  2.1317e-02,
         2.8245e-03,  3.9928e-03,  1.8074e-02,  2.4196e-02,  7.5209e-03,
         1.2217e-02,  9.9649e-04,  3.8469e-03, -1.5930e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.2685e-03,  8.3180e-01,  3.0817e-02, -2.0656e-02,  2.1563e-03,
         2.1124e-03, -5.3416e-03,  2.3238e-03, -8.1271e-04, -3.0615e-03,
         2.7038e-03, -2.1301e-03,  1.7335e-02,  5.4528e-03,  1.1935e-02,
         3.9960e-03,  1.0124e-02,  2.5678e-02,  1.5297e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 2.9056e-02, -7.1055e-01,  1.1606e-01,  1.2580e-02, -2.1652e-02,
        -2.0912e-02, -2.1303e-03,  4.3855e-04, -5.8166e-03, -9.1877e-03,
        -2.0388e-03,  5.5234e-03, -1.0296e-03,  7.0661e-03, -4.7901e-03,
         5.4144e-03, -2.2343e-03, -5.5949e-03, -5.2634e-03, -1.3305e-02,
         1.1440e-03, -2.5800e-03,  5.4622e-03, -1.1268e-03,  7.5923e-04,
         8.2847e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0804, -0.7533, -0.0085,  0.0129, -0.0392, -0.0078,  0.0085, -0.0078,
         0.0055,  0.0016,  0.0037, -0.0039, -0.0266, -0.0083,  0.0144,  0.0176,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0037, -0.7509, -0.0451, -0.0590, -0.0074, -0.0379, -0.0084, -0.0187,
        -0.0097,  0.0065, -0.0051,  0.0198,  0.0022, -0.0042, -0.0083, -0.0068,
        -0.0066,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0607,  0.7861, -0.0097, -0.0228,  0.0497,  0.0046,  0.0138,  0.0171,
         0.0355,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6063e-02, -6.3624e-01, -1.7624e-02,  1.5818e-02,  2.2576e-03,
         8.1631e-03, -1.9203e-02, -2.5976e-02,  5.2440e-04, -1.5195e-02,
        -5.6907e-03, -7.7612e-04, -2.4325e-03,  4.5246e-03, -9.0277e-03,
        -1.3812e-02,  5.8083e-04, -1.0960e-02, -1.0167e-02, -5.4338e-03,
        -3.4608e-03, -4.7854e-03,  9.2450e-03,  2.6194e-03, -2.4816e-03,
         9.2165e-03, -1.2078e-02, -2.9847e-03, -8.4649e-05,  3.0687e-03,
         7.5653e-03, -5.9647e-03, -4.9987e-03,  4.0049e-03, -3.4280e-03,
        -7.8381e-04,  2.3071e-03,  5.5197e-03,  5.8744e-03,  3.9255e-03,
        -2.5266e-03, -6.6914e-03,  1.6576e-03, -4.4476e-04,  9.1696e-03,
        -4.4648e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.8999e-03, -7.5463e-01, -4.7242e-02,  1.3228e-02, -8.8094e-03,
        -2.4198e-03,  1.9898e-03,  5.3169e-03,  1.5739e-02, -1.2581e-02,
         9.8297e-03,  4.8735e-03, -6.3333e-03,  3.3431e-03,  1.3630e-03,
         3.8632e-04, -4.8465e-03,  1.8429e-03,  4.9808e-03, -6.1599e-03,
         5.4509e-03, -3.4190e-03, -7.1013e-03, -4.8708e-03, -2.2305e-03,
         1.1010e-03,  1.0965e-02, -6.2150e-03, -1.4205e-02, -2.8629e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.4222e-02, -8.2226e-01, -1.7229e-02, -6.6936e-03,  4.0364e-04,
        -7.6255e-02, -2.9393e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0467, -0.7903,  0.0010, -0.0015, -0.0221, -0.0036, -0.0094, -0.0204,
        -0.0137, -0.0507, -0.0406,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0354, -0.7690,  0.0398, -0.0319,  0.0019, -0.0327,  0.0250, -0.0056,
         0.0011,  0.0089,  0.0037,  0.0133,  0.0014, -0.0302,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.9808e-03, -8.0406e-01,  4.4629e-03, -2.7641e-03, -2.0306e-02,
        -1.3778e-03,  7.6297e-03,  8.8927e-03,  9.4577e-03, -1.4522e-03,
         1.7307e-03, -2.3749e-03,  6.7759e-03, -6.2139e-03,  2.3919e-03,
         4.4271e-05, -4.2276e-03, -6.9103e-03,  7.8485e-03, -1.6658e-02,
        -2.8527e-03, -3.7850e-03, -3.5680e-04, -2.6596e-03,  3.3597e-03,
        -3.0563e-03,  3.1578e-03,  9.4321e-04,  8.7412e-03,  4.6344e-03,
         1.3404e-02, -1.1007e-02,  1.3810e-02, -5.6726e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.0638e-03,  6.8342e-01,  6.5305e-02, -4.6129e-02, -1.1136e-02,
        -2.1954e-03,  1.8111e-02,  7.0603e-03, -1.0224e-02,  1.1845e-02,
        -3.9628e-04,  5.4893e-03, -1.0020e-03,  1.2302e-02, -8.3547e-04,
        -1.0667e-02,  1.4682e-02, -1.8888e-02,  8.0527e-04,  1.3021e-02,
        -1.0390e-02,  2.2162e-04, -2.1918e-02,  2.2037e-03, -2.7890e-04,
         1.2068e-03,  2.0795e-02, -4.1082e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1926e-02, -7.8677e-01, -4.1181e-02,  8.9058e-03, -1.0043e-02,
         6.9525e-03,  3.7860e-03, -9.2033e-03,  5.2314e-03, -2.6568e-04,
        -6.6398e-04,  4.9668e-03,  1.7763e-02, -2.4818e-02, -1.9508e-02,
         7.6580e-05,  8.9559e-03, -2.6120e-04, -1.6032e-02,  3.7522e-03,
         1.3035e-03, -2.2806e-03,  1.2517e-02,  2.8353e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 5.7423e-02,  7.3989e-01,  6.3937e-03, -1.4430e-02, -2.2359e-03,
         8.0901e-03, -2.1331e-02, -1.8226e-03, -3.2806e-03, -6.5004e-04,
        -1.0536e-02, -8.4769e-03,  3.2634e-03, -9.6336e-04, -5.4155e-03,
        -8.4176e-03, -4.5436e-03, -4.0308e-03, -8.2756e-03, -8.5577e-03,
        -2.1549e-03, -3.9529e-03, -7.0610e-03, -3.6116e-03,  2.8632e-03,
         4.2788e-03, -3.8930e-03, -1.4586e-03, -1.4327e-02, -1.5496e-02,
        -2.2878e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0734, -0.8225,  0.0127,  0.0229, -0.0053, -0.0035,  0.0101,  0.0037,
        -0.0162,  0.0071, -0.0077,  0.0150,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0471, -0.7617, -0.0011,  0.0610, -0.0025,  0.0017,  0.0135,  0.0094,
         0.0092,  0.0086, -0.0039,  0.0398, -0.0290,  0.0064,  0.0052,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.7958e-02, -7.2879e-01,  4.9645e-02,  8.1206e-03, -1.2000e-02,
        -6.7464e-03,  4.9317e-03,  8.5221e-03, -1.1389e-03, -1.9033e-03,
        -9.6802e-03,  9.4128e-03, -6.2972e-03,  1.0567e-02,  7.4725e-04,
         3.9137e-04, -4.9894e-04, -6.9984e-04, -7.7642e-03,  9.3882e-03,
         8.6825e-04, -5.5759e-03, -1.9749e-03, -3.7855e-04, -2.5959e-03,
         4.1376e-03,  3.2469e-03,  4.3491e-03,  1.4281e-03,  1.1518e-02,
        -1.5464e-03, -5.4760e-03, -2.1700e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1880e-02,  7.4725e-01, -1.8908e-02, -6.2333e-02,  5.9523e-03,
         2.8570e-04, -5.8026e-03,  1.3904e-02, -1.2690e-02, -3.4045e-03,
        -1.4203e-02, -3.6630e-03, -8.9746e-03, -3.8545e-02,  2.4144e-02,
        -2.8064e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0316, -0.8473,  0.0619,  0.0163,  0.0043, -0.0038, -0.0082, -0.0149,
        -0.0117,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1757,  0.6782, -0.0300, -0.0381, -0.0780,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8871e-02, -7.7971e-01,  1.7235e-02, -1.7214e-02,  2.0347e-02,
        -1.5224e-03,  1.3684e-02,  1.2360e-02,  4.8460e-03,  3.9368e-03,
         1.0996e-02, -1.5572e-02, -4.1465e-03,  6.3826e-03, -4.7188e-03,
        -4.3848e-03, -1.6677e-03,  3.4859e-03, -7.4186e-03,  4.5181e-03,
         3.7665e-03, -2.2295e-03,  4.8538e-03, -1.5171e-03, -5.5304e-04,
        -1.4173e-02,  9.8862e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0396, -0.8394, -0.0210, -0.0057, -0.0486, -0.0245, -0.0213,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7086e-02,  8.1667e-01,  3.1842e-02,  1.0817e-02, -2.1375e-02,
         1.1418e-02,  5.6066e-03, -3.1555e-04, -2.2040e-02,  4.2484e-03,
        -4.3177e-03,  1.5046e-03, -7.6202e-03,  3.2407e-03, -1.6211e-04,
        -2.3693e-03,  1.2772e-02,  6.7219e-03, -1.9314e-04,  9.8310e-03,
         2.4305e-03,  7.4159e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0210,  0.7770,  0.0343,  0.0070, -0.0229,  0.0105,  0.0042, -0.0081,
        -0.0134, -0.0328,  0.0046,  0.0036, -0.0058, -0.0013, -0.0139, -0.0033,
        -0.0172, -0.0035, -0.0155,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4070e-02,  8.1387e-01,  9.8653e-03, -1.0592e-02,  6.3056e-03,
        -8.8861e-04, -8.4476e-03,  8.8486e-03,  1.1738e-02,  3.8215e-04,
        -4.0574e-03,  5.6211e-03, -6.6610e-04,  2.9776e-03,  8.7712e-03,
        -1.8965e-03,  9.8531e-03, -8.1489e-04, -5.2928e-05, -1.0288e-02,
         1.6001e-03, -3.2116e-03,  3.3859e-03,  7.7017e-03,  7.8836e-03,
         2.1115e-03, -1.3155e-05,  2.5246e-03, -2.1962e-02, -9.7460e-03,
         9.8486e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 8.1823e-02, -6.6669e-01, -1.5156e-02,  2.3620e-02, -1.1244e-02,
         1.9289e-03,  3.9187e-03,  1.1028e-02,  1.0727e-02,  3.2991e-03,
        -2.4406e-03, -1.1520e-02, -2.1974e-02, -1.0387e-02,  1.2028e-02,
        -1.9533e-03,  1.2870e-03,  2.6748e-03,  2.3220e-02, -9.0435e-03,
        -1.7393e-03, -4.2628e-03, -3.9456e-03, -8.0240e-03, -5.5024e-03,
         9.6148e-05, -5.7040e-04,  3.2037e-03,  1.3000e-03,  3.6902e-03,
        -1.6034e-02, -2.5669e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4357e-03,  8.0134e-01, -7.3905e-03,  9.4722e-03, -2.6666e-02,
         4.0716e-03,  5.6464e-03,  2.4640e-02,  1.3434e-02,  1.4759e-03,
        -3.2091e-03, -6.2524e-03, -1.7434e-02, -1.3514e-02, -1.2199e-03,
        -7.2187e-03, -1.8581e-03, -8.1319e-03, -7.1231e-04,  8.1168e-03,
        -3.9218e-05,  3.3721e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5807e-02,  7.5976e-01,  1.5445e-02, -1.1139e-02, -2.2202e-04,
        -8.3749e-03, -7.9609e-04,  5.8196e-03, -3.6167e-03, -5.1199e-03,
         1.2220e-02,  3.8700e-03, -9.8112e-04, -6.2220e-03, -6.2109e-03,
        -5.8201e-03, -5.6679e-03,  1.7335e-02, -5.0292e-03, -1.1277e-02,
        -1.1326e-03,  4.3640e-03,  7.5012e-03, -3.9063e-03, -1.5831e-02,
        -3.7571e-03, -6.0764e-03, -8.0016e-03, -1.1732e-02, -1.6961e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1046e-02,  8.1744e-01, -1.1981e-02,  4.0991e-03, -4.6364e-03,
        -1.3329e-02,  5.9164e-02, -2.8895e-03, -3.9721e-03, -5.1034e-03,
        -3.2095e-03,  2.5282e-03, -1.3917e-02, -1.7408e-04, -1.4351e-03,
        -1.5133e-03,  1.6511e-02, -1.7047e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3125e-02, -7.5857e-01, -2.7376e-02, -1.5231e-02,  2.5148e-04,
         3.1578e-02,  2.9332e-03, -5.7546e-03,  1.6821e-03, -6.6024e-05,
         5.0633e-03, -1.7998e-03,  2.5386e-03, -1.7713e-03,  1.1482e-03,
         1.4792e-02, -1.4060e-02, -4.8447e-03,  5.1485e-03, -1.6186e-03,
         3.7710e-03,  5.7951e-03,  2.8196e-03, -4.3637e-03, -4.9982e-03,
         9.7084e-04,  2.4107e-03, -1.5648e-03, -3.6401e-03, -4.1735e-03,
        -7.5532e-05,  2.8294e-03, -3.4192e-03,  4.0902e-03, -1.1810e-03,
         5.3438e-05, -4.4866e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8857e-02,  7.7493e-01, -7.9845e-02, -2.2070e-02, -6.7607e-05,
         1.3839e-02, -8.1901e-03, -1.7253e-02, -4.0880e-04,  1.0588e-02,
        -2.6526e-03, -4.8503e-03,  9.1535e-03,  5.6895e-03, -3.2219e-03,
         2.2008e-03,  9.5065e-03,  6.3409e-03, -2.8007e-03,  5.0199e-04,
         3.5924e-03,  1.8561e-04,  3.9634e-06,  3.2528e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5476e-02,  6.9103e-01,  3.9632e-02,  2.7432e-02,  1.3838e-02,
         1.9195e-02, -1.0218e-02,  2.4323e-03, -2.8277e-03, -3.3025e-03,
         1.4639e-03, -3.4283e-04, -1.1555e-02,  2.0377e-05,  8.5364e-03,
         2.7477e-04, -3.1773e-03,  4.1536e-03,  2.8806e-03,  1.7465e-03,
        -3.5275e-03, -5.6889e-03,  3.0209e-03, -9.3482e-03,  6.2298e-03,
         4.5373e-03,  1.2915e-03, -2.3868e-03, -5.0123e-03,  7.2040e-03,
        -1.7744e-03, -4.3855e-03,  3.3320e-03, -5.9362e-04,  4.8828e-03,
        -1.7437e-03, -7.7882e-04, -7.7567e-04,  2.2719e-03,  2.3187e-03,
        -2.0361e-03, -1.0661e-02, -2.3921e-03,  3.5302e-03,  1.5816e-02,
         4.9254e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1599e-03,  7.8940e-01, -1.8757e-04, -5.0497e-03, -4.1845e-03,
         6.6303e-03, -3.6036e-02, -1.0056e-02, -6.9774e-03, -1.3226e-02,
         9.5200e-03, -8.2152e-03, -2.2414e-03,  1.4962e-03,  1.0467e-03,
         6.6249e-03,  6.3805e-03, -2.0203e-02, -1.5834e-02, -3.3744e-05,
         1.9902e-02,  1.2693e-02, -1.9905e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6051e-02,  8.1477e-01, -1.1091e-02,  1.8191e-02, -1.4526e-03,
         6.8093e-04,  2.7609e-02, -1.8605e-02, -8.3936e-03, -1.9626e-02,
        -2.3528e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0225,  0.8192,  0.0049, -0.0283, -0.0055, -0.0036, -0.0167, -0.0112,
         0.0021,  0.0060,  0.0106, -0.0047,  0.0192, -0.0453,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0088,  0.8370, -0.0326, -0.0296, -0.0043,  0.0065, -0.0033,  0.0222,
         0.0046, -0.0111,  0.0322,  0.0078,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8838e-02,  7.5734e-01,  6.3113e-04,  2.3144e-03,  1.4037e-02,
        -1.8891e-02, -4.7325e-03,  1.5725e-02, -1.1257e-03,  4.5174e-03,
         1.8703e-04, -8.5417e-03, -1.0657e-03,  2.2907e-03, -3.6351e-03,
         1.8627e-03,  8.0918e-04, -1.1367e-03,  6.0101e-03,  6.1064e-03,
         1.7679e-03,  1.7131e-03,  2.2785e-03,  2.5276e-03, -4.8889e-04,
        -1.1262e-02,  2.1799e-02,  8.7104e-04, -1.8358e-03,  2.2309e-03,
         2.5689e-03,  5.5448e-03,  8.9198e-03,  6.7707e-03,  3.2076e-03,
         2.7664e-03,  4.4656e-03, -2.6947e-04, -1.0802e-03,  4.8412e-04,
         3.4409e-03, -3.7595e-04,  1.4051e-03, -9.3009e-04, -2.7621e-03,
        -1.1743e-03,  3.1134e-03, -9.2887e-04,  8.4439e-05,  9.7696e-04,
         7.5343e-04, -1.5890e-03, -4.7318e-03, -2.4943e-03,  1.2592e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 4.4417e-02, -7.2971e-01,  7.1556e-04, -8.2127e-03, -1.3915e-02,
        -1.9999e-02,  1.5284e-03,  1.2327e-02, -6.5168e-03, -5.0838e-03,
         1.7809e-02, -1.5574e-02, -7.1379e-03, -1.1131e-04, -1.7150e-03,
        -1.1741e-02,  5.6387e-03, -1.8150e-02, -1.8773e-03, -1.1760e-02,
         1.1855e-02, -2.8112e-02,  2.6090e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.3237e-04,  7.3795e-01, -2.8545e-02,  2.1241e-02,  1.3730e-02,
        -9.6309e-03, -4.5574e-03,  5.6074e-03, -2.1398e-02,  1.1464e-02,
        -2.2005e-02,  1.2653e-04, -2.4988e-02, -2.6560e-02, -1.8668e-02,
         9.5964e-04, -8.5888e-03, -4.3246e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2579e-04, -6.4720e-01,  6.1146e-02, -8.4409e-03, -1.9539e-02,
        -1.9354e-02, -7.6228e-02, -1.7298e-02, -2.4137e-02, -3.6302e-03,
         8.4931e-03, -2.2836e-03,  2.7036e-02,  7.8797e-03,  1.6181e-02,
        -1.6888e-02, -2.0515e-04, -4.3737e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3327e-02, -7.4455e-01,  2.9080e-02,  1.5656e-02, -7.4191e-05,
         2.9456e-03, -2.6220e-02, -1.1235e-02,  1.7827e-02,  2.2968e-03,
         6.5165e-03,  4.0705e-04, -2.5120e-03,  1.1046e-02,  4.6304e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7345e-02, -7.1444e-01, -3.4417e-02, -2.3074e-02, -1.2653e-02,
        -2.5449e-03,  8.4880e-03,  1.9734e-02, -7.7528e-03, -8.3547e-03,
        -9.4009e-03,  1.8465e-03, -9.1358e-04, -2.2001e-03,  1.0632e-02,
        -2.0176e-03, -6.8281e-03, -4.9299e-03, -3.0159e-03, -6.8557e-03,
        -1.0007e-02,  1.3144e-02, -3.9157e-03,  4.4028e-04, -5.7967e-03,
         1.9251e-04, -2.2806e-02, -1.5289e-03, -1.7564e-03, -8.3494e-03,
        -9.3069e-03,  5.3136e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.0714e-03, -7.6333e-01, -2.5181e-02, -1.9307e-02, -2.3644e-02,
        -4.1305e-03, -6.1243e-03, -3.1701e-03,  5.9570e-03,  6.4414e-03,
        -5.0397e-03,  9.4675e-03,  1.0605e-03,  1.1399e-02, -2.4038e-04,
        -1.8075e-03, -1.2709e-03,  9.2573e-03,  3.1716e-03, -4.4584e-03,
         1.7358e-02,  2.5749e-03,  4.1952e-03,  8.3022e-03,  3.2763e-03,
         3.2182e-03, -8.4414e-05, -2.4871e-03,  4.8312e-03,  7.1034e-03,
        -5.1933e-03,  1.7369e-03, -2.8113e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3449e-02, -7.5568e-01, -1.0378e-02, -2.8340e-02,  1.9816e-04,
         1.9281e-03, -4.2681e-03,  5.8805e-03, -5.0163e-03, -2.4317e-03,
         2.1747e-03, -4.9860e-03,  5.9314e-03,  2.1370e-03, -5.0909e-03,
        -6.9302e-03,  8.7190e-04, -6.1249e-04,  5.7105e-03,  5.5133e-03,
         2.3165e-03,  6.0284e-05,  4.7122e-04, -7.9746e-03,  8.5656e-04,
        -1.2023e-03,  7.9996e-04, -3.1097e-03, -4.9074e-03,  6.9522e-03,
        -5.0659e-03, -5.1110e-03, -1.2095e-03, -1.7405e-03, -1.2747e-05,
         4.6205e-03, -3.0395e-03,  5.1581e-04, -2.0948e-03, -7.0014e-03,
        -8.4514e-04, -8.1887e-04, -2.7668e-03,  1.1909e-03, -2.5779e-03,
        -7.8231e-03, -1.4115e-03, -2.8469e-03, -5.5126e-03, -6.9718e-04,
         9.7424e-04, -2.0030e-03, -3.0707e-03, -3.0392e-03, -7.2343e-03,
        -3.8201e-03, -6.2369e-03, -1.0542e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0025,  0.8016,  0.0241,  0.0117,  0.0125, -0.0057,  0.0059, -0.0042,
        -0.0146,  0.0095,  0.0062,  0.0103, -0.0127, -0.0042, -0.0035, -0.0033,
        -0.0054, -0.0027, -0.0041, -0.0054, -0.0036, -0.0031, -0.0052, -0.0023,
         0.0009, -0.0350,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0327, -0.7275,  0.0419,  0.0166,  0.0069,  0.0051, -0.0049,  0.0155,
        -0.0095, -0.0141, -0.0201,  0.0009, -0.0063,  0.0039, -0.0055,  0.0012,
        -0.0027,  0.0080, -0.0098, -0.0045,  0.0091,  0.0134,  0.0173,  0.0197,
         0.0026,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.0981e-04, -6.8705e-01, -6.1586e-02, -2.1059e-03, -1.9021e-02,
        -4.6201e-03,  3.7641e-04,  5.7294e-03,  1.5126e-03,  2.6756e-04,
        -9.2011e-03,  7.8741e-03, -6.4257e-03,  1.2389e-03,  8.7100e-03,
         6.1680e-04,  7.4034e-03, -1.4789e-03, -8.3790e-04,  1.7788e-03,
         1.2531e-05, -1.3487e-02,  3.6994e-03,  1.0785e-02,  3.6751e-03,
         6.8722e-03,  3.3645e-03,  8.5093e-04,  6.3415e-03,  5.3848e-03,
         8.7205e-04,  4.2065e-03,  5.5979e-03, -3.5958e-04, -9.0225e-03,
         3.8802e-03,  4.1315e-03,  2.8872e-03,  7.2335e-04, -1.6953e-04,
         5.7633e-03, -1.3276e-03,  8.9659e-03,  1.5884e-02,  5.3089e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2658e-02, -8.0653e-01, -2.3515e-02, -2.2922e-02, -5.9198e-03,
        -2.1534e-02, -4.9276e-03, -1.3323e-02,  1.5864e-02,  8.0636e-03,
         3.5537e-03, -2.6191e-03,  9.8375e-03,  2.5739e-03, -8.4986e-04,
         2.8769e-04, -7.6231e-03,  4.1380e-03,  3.3413e-03, -9.0409e-04,
        -2.1566e-03, -1.6276e-03,  5.2301e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0579, -0.7171,  0.0118,  0.0574,  0.0110,  0.0090,  0.0142,  0.0043,
         0.0207,  0.0030,  0.0337, -0.0598,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 2.0925e-02,  7.4572e-01,  3.5514e-02, -8.6716e-03, -1.6091e-02,
         2.3830e-02, -1.7026e-02, -6.2026e-03,  3.7110e-03, -1.7446e-02,
        -1.8147e-02,  6.0272e-04, -3.6310e-03,  5.8592e-03,  4.8936e-03,
        -1.6283e-02,  8.5427e-04,  9.5988e-03,  6.5633e-03, -2.9152e-03,
        -1.0521e-02,  6.1449e-03,  1.8845e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.3049e-02,  7.7087e-01, -7.1117e-03, -5.2145e-03,  1.6411e-02,
         9.4121e-03,  5.8842e-03, -1.9969e-03, -1.8037e-03,  4.5621e-03,
        -1.7186e-03, -5.4910e-03, -2.9276e-04,  2.0965e-03,  1.1763e-02,
         1.0381e-03, -6.7000e-03, -1.9539e-02, -1.6202e-03,  1.1471e-03,
        -1.0899e-02, -7.3251e-03, -7.7552e-03, -6.5964e-03, -4.0992e-03,
        -1.6035e-05, -6.0715e-03, -1.8573e-03,  1.6821e-03, -1.7159e-03,
        -1.5448e-04,  3.0838e-03, -1.0218e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5600e-02, -7.8542e-01, -4.0824e-02,  2.5084e-02, -2.8212e-04,
         1.0667e-02,  7.0792e-03,  2.7991e-03,  1.2729e-02,  5.4166e-03,
         1.4263e-02, -2.5975e-03, -4.6154e-03, -2.8591e-02, -8.7199e-03,
         7.3416e-03,  8.6546e-03,  5.8820e-03,  3.4361e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.6775e-02, -7.5075e-01,  1.7820e-02, -2.3318e-02, -1.1830e-02,
         7.2097e-06, -3.9938e-03,  5.0665e-03,  3.8345e-04,  1.3340e-02,
        -1.9117e-02, -1.9184e-02, -3.7802e-02, -3.0608e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0800e-01,  7.1806e-01,  2.9028e-02, -2.2656e-03, -1.8603e-02,
         2.1918e-02, -8.0795e-03, -3.5460e-03,  8.0187e-03,  4.6418e-03,
         1.1615e-02,  1.9503e-03,  1.4680e-03, -8.8626e-03, -7.7922e-04,
        -4.4699e-03, -1.0359e-02, -1.3073e-04, -7.6383e-03, -7.4551e-03,
         1.8716e-02,  4.4025e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.4886e-03,  7.4581e-01, -1.1004e-02,  1.2018e-02,  1.1764e-02,
         2.7837e-02, -1.3968e-02, -3.0392e-02, -8.4789e-05, -1.0257e-02,
         3.5378e-03,  3.8934e-03, -7.4358e-04,  2.0349e-03, -6.1675e-03,
         3.5933e-03,  2.7785e-02,  1.3542e-02,  8.7914e-03,  3.3447e-03,
        -1.3303e-03,  7.1391e-03,  1.3548e-03, -5.0864e-03, -5.4140e-03,
         4.5590e-03, -1.5863e-03,  7.1908e-03,  1.7508e-02,  3.7782e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2208e-03,  7.5046e-01, -3.0078e-02, -2.6969e-02,  3.0587e-02,
         9.6695e-03, -2.4199e-03,  4.4422e-04, -2.1651e-03,  1.3280e-02,
        -1.2493e-02, -1.3347e-03,  5.9883e-03,  9.2279e-03,  7.8970e-04,
         4.1787e-02, -6.1087e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0184,  0.7394,  0.0141, -0.0061, -0.0552,  0.0158, -0.0199, -0.0013,
         0.0342,  0.0500, -0.0456,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0339, -0.7230, -0.0142,  0.0182,  0.0168,  0.0206,  0.0074,  0.0102,
         0.0028, -0.0178,  0.0011,  0.0085,  0.0047, -0.0375, -0.0058,  0.0291,
         0.0484,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0117, -0.7846,  0.0180, -0.0462,  0.0048, -0.0058, -0.0102,  0.0035,
        -0.0014,  0.0080, -0.0024, -0.0121, -0.0009, -0.0091, -0.0180, -0.0088,
        -0.0022, -0.0021,  0.0009, -0.0035, -0.0045, -0.0013, -0.0248, -0.0085,
        -0.0052,  0.0013,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1637e-02,  7.7112e-01, -4.5786e-02, -1.4795e-02, -1.3288e-02,
         6.0256e-03, -1.3435e-02, -3.3266e-03, -1.1174e-02, -1.4434e-02,
         4.3824e-03, -1.2176e-03,  1.4150e-03,  3.9625e-03,  6.8011e-03,
        -1.4309e-03, -3.2192e-03, -4.1133e-04, -2.9905e-03,  2.4848e-03,
        -3.2852e-03, -5.3682e-03, -4.4661e-03, -2.2750e-03,  5.5854e-04,
        -1.8686e-03,  2.3411e-03,  2.9170e-03,  7.1264e-04,  2.8738e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.1335e-02, -8.1029e-01,  4.5251e-03,  1.6601e-02, -1.9568e-02,
         1.2927e-02, -1.1542e-02,  8.8387e-03, -2.0628e-03,  4.9505e-03,
         1.1653e-02, -1.8912e-04, -5.2804e-03,  2.2948e-03,  1.1555e-02,
        -2.2884e-03, -6.1219e-04,  5.1563e-03,  1.8330e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-3.8772e-02,  7.1930e-01, -1.6714e-02,  1.8274e-02,  2.7572e-03,
        -1.5480e-02,  3.1670e-02, -9.1221e-03,  3.2878e-04,  2.7712e-04,
         2.9554e-03,  7.3022e-03,  1.7633e-03, -1.0683e-03, -1.7892e-02,
         1.4511e-02,  2.8031e-03,  1.9543e-02,  1.0887e-02,  4.6439e-03,
        -1.3133e-03,  3.0511e-03,  5.8435e-03, -1.5411e-03, -2.2122e-02,
        -7.6488e-03,  2.2419e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.2904e-04, -7.8800e-01,  2.7940e-02,  1.6527e-02,  2.9043e-02,
        -2.3784e-02,  1.1027e-02,  1.6380e-02,  1.6773e-02, -6.3359e-03,
        -1.9193e-02, -2.1739e-02,  9.1490e-03,  1.3579e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4894e-02,  7.7079e-01,  4.4613e-02,  1.0533e-02,  2.4822e-03,
         1.0271e-02,  9.2260e-03, -7.5806e-03,  5.6125e-03, -1.6101e-03,
         1.1725e-02, -5.4301e-04, -7.9540e-04, -5.9834e-03,  1.1134e-02,
         4.4251e-03, -2.4520e-03, -2.3128e-03, -4.0644e-03,  2.0143e-03,
        -2.5702e-04, -5.5521e-03,  2.1207e-03, -4.8099e-03,  1.0503e-02,
        -4.3695e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0706,  0.6975,  0.0074, -0.0107,  0.0163, -0.0280, -0.0026, -0.0079,
        -0.0114, -0.0261, -0.0191,  0.0469,  0.0113, -0.0442,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5981e-02,  7.2351e-01,  4.0189e-02, -1.9032e-02, -5.8107e-03,
         6.0910e-04,  2.2922e-03, -1.8616e-02, -1.9677e-04,  2.2637e-03,
        -5.7153e-03,  2.3031e-03, -2.9376e-03,  1.6967e-02,  1.1985e-02,
         3.4485e-03,  1.0759e-02,  9.7770e-03, -2.3320e-03,  6.2173e-03,
         4.2417e-03,  3.4705e-03,  6.3163e-03,  6.7435e-03,  1.3848e-03,
         1.6791e-04, -1.2065e-02,  1.1029e-02,  2.7408e-02, -1.3345e-03,
         9.0366e-03,  1.5859e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4252e-02,  7.6485e-01,  1.1910e-02, -6.4249e-03,  1.7156e-02,
         2.8735e-02, -1.8447e-02, -9.2475e-03,  1.2206e-03, -8.8502e-03,
         1.3977e-02,  3.5266e-03, -9.0020e-03,  1.2948e-05,  6.9194e-03,
         2.8983e-03,  1.8653e-03, -7.2364e-03, -1.0916e-03,  4.8324e-03,
        -2.3968e-03, -4.4415e-03,  4.6046e-04,  2.2212e-02, -2.8030e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0834,  0.7487,  0.0407,  0.0151,  0.0129, -0.0991,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.8846e-02,  7.3097e-01,  1.4966e-02, -8.4280e-03,  1.6885e-02,
        -1.1196e-02, -9.9717e-03, -7.0715e-05,  3.2514e-03, -1.8069e-02,
        -3.1648e-03,  1.5782e-02, -2.9192e-03, -2.5833e-03, -5.7126e-03,
        -7.9294e-03, -4.7865e-03,  6.0377e-03, -1.4954e-02, -6.0348e-03,
        -1.8122e-03, -6.8063e-03, -3.9838e-03, -6.4527e-03, -1.9362e-03,
        -1.5220e-03, -6.1683e-04, -3.6567e-03,  3.1443e-03,  1.3385e-02,
         7.7610e-04,  5.1014e-03,  5.5142e-03,  1.8007e-03,  2.0471e-03,
        -7.7155e-03,  1.1174e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2872e-03, -8.1286e-01,  1.2156e-02, -5.3367e-03, -6.6032e-03,
        -1.3684e-02, -7.0746e-03, -4.6622e-03, -1.5175e-03, -2.7579e-03,
         6.7257e-03, -8.8024e-03, -2.2590e-03,  5.9335e-03,  6.9149e-03,
         5.9008e-03, -3.6377e-03, -1.0152e-04,  4.6245e-04, -3.4442e-04,
         3.3114e-02, -8.7183e-04,  7.9937e-04,  3.1057e-03,  6.6805e-04,
         1.9852e-03, -5.3818e-03,  1.5598e-03,  1.5049e-02, -4.2549e-03,
        -1.4625e-03, -2.0102e-03, -1.6574e-03,  5.9033e-03, -1.2160e-02,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0155,  0.5971, -0.1878,  0.0038,  0.0092,  0.0391,  0.0027,  0.0036,
         0.0021,  0.0044, -0.0063, -0.0126, -0.0421,  0.0098, -0.0641,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.9832e-02,  7.9306e-01,  3.5328e-03,  1.7893e-03,  2.3944e-02,
         1.0733e-02,  1.0530e-02, -1.6349e-02,  1.1499e-02,  1.3402e-02,
         3.9139e-03, -5.7379e-03, -9.5852e-04, -2.4776e-03,  5.2015e-03,
        -2.1640e-03,  2.1563e-03, -2.5199e-03, -6.5363e-03,  1.1475e-03,
        -5.4559e-03, -8.4862e-03, -2.9579e-03, -3.4052e-03, -3.5989e-03,
        -5.3167e-03, -2.0748e-03,  1.1431e-03,  5.5823e-04,  9.7849e-04,
         1.2792e-02,  4.9629e-04,  5.2563e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7766e-02, -7.3574e-01,  2.5155e-02,  9.2190e-03, -3.6581e-03,
         2.2926e-02, -1.8396e-02, -5.9662e-04,  2.5430e-03,  1.8236e-03,
        -6.7694e-03, -5.4636e-03,  2.9385e-03,  1.4058e-02,  1.6904e-02,
         2.1697e-03,  1.8276e-02, -9.4641e-03,  1.0581e-02,  4.5230e-03,
         1.5242e-03,  3.4045e-03, -1.9804e-04,  3.0795e-05, -6.7619e-03,
         4.9353e-03,  5.6244e-03, -1.3720e-02,  1.4834e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 7.4471e-02,  7.7508e-01, -5.2931e-04, -2.3076e-02, -6.3327e-03,
        -2.6109e-03,  1.8184e-02, -1.2677e-02,  3.4552e-03, -2.8168e-02,
        -4.4097e-03, -4.6572e-04, -1.4966e-03,  5.5258e-03, -4.3517e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.1929e-02,  7.6687e-01, -1.8750e-02, -1.5731e-03, -2.3636e-02,
        -4.6331e-03, -3.5758e-03,  7.7899e-04, -4.2044e-04,  7.3647e-04,
        -1.4037e-02, -3.5469e-03, -2.1902e-03, -5.7033e-03, -8.0252e-04,
        -4.5044e-03, -5.4274e-03, -2.9184e-03,  1.5246e-03, -2.9009e-03,
        -4.4423e-03, -5.8122e-03, -5.8953e-03,  5.0817e-04,  9.9624e-04,
         1.2477e-03, -7.1896e-04, -6.1892e-04, -2.9208e-03, -1.8647e-03,
         6.5718e-04, -1.2807e-03, -1.0193e-03, -3.1506e-03, -1.8056e-03,
         4.2175e-03, -5.3780e-03, -4.3973e-03,  1.2238e-03, -6.2033e-03,
        -3.2016e-03, -5.7648e-03,  3.1915e-03, -3.3524e-03,  1.6037e-03,
        -1.1336e-02, -1.9632e-03, -5.4340e-03, -3.3355e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9744e-02,  7.2572e-01, -3.3421e-02, -1.8448e-03, -1.4167e-03,
        -8.9020e-03, -9.3311e-03,  4.4839e-03, -1.6286e-02, -1.3685e-02,
        -6.5194e-03,  3.6738e-04,  8.5682e-03, -2.4722e-03,  1.8468e-03,
        -1.8442e-03,  2.9946e-02,  5.7232e-03,  9.3505e-03,  5.0012e-03,
         1.5070e-03, -3.8373e-03, -4.8833e-04,  3.3096e-03, -4.7680e-04,
        -2.0854e-03, -1.1044e-02,  9.9312e-03,  4.9981e-03, -8.2794e-04,
        -3.7413e-03, -7.5766e-03,  6.7050e-04,  9.3033e-03, -3.3731e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0638, -0.7392, -0.0176,  0.0078, -0.0193, -0.0102,  0.0126, -0.0022,
        -0.0067,  0.0084, -0.0115, -0.0026,  0.0036, -0.0019, -0.0133, -0.0168,
         0.0068, -0.0167, -0.0392,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.1789e-03,  7.2885e-01, -2.9106e-03,  4.1966e-02,  2.9125e-03,
         1.9215e-02,  1.9517e-02,  9.3264e-03,  1.7636e-02,  2.8804e-04,
         9.7970e-03, -1.1004e-03, -7.1619e-03, -4.1331e-02, -8.8805e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.6177e-02,  7.9551e-01, -9.6507e-03,  1.6528e-02, -2.1609e-03,
        -2.5543e-03, -1.5068e-02,  2.9795e-03,  1.0120e-02, -1.0257e-02,
        -4.5753e-03, -1.6741e-03, -5.6111e-03, -3.8001e-04, -1.1295e-02,
         5.0770e-04, -3.5982e-03, -7.1203e-03, -3.3850e-03, -1.8727e-03,
        -1.4691e-02,  1.4285e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2999e-02, -7.5302e-01,  3.5899e-02, -1.4654e-02, -2.1538e-02,
        -1.5620e-03,  1.0805e-02,  6.6116e-03, -4.1332e-04, -5.4852e-03,
        -1.8788e-03,  5.3047e-03, -4.5226e-03,  7.9633e-03,  2.4638e-03,
         2.0055e-04, -4.4784e-03,  3.1099e-03,  9.9329e-04, -8.0401e-03,
        -6.0633e-03, -4.3659e-03,  4.3388e-03,  4.2793e-03,  9.7095e-03,
        -2.5551e-03, -1.3529e-03, -5.1103e-03, -1.4088e-02, -1.6190e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0098, -0.7447,  0.0275,  0.0446,  0.0671,  0.0158,  0.0017,  0.0229,
         0.0169, -0.0325, -0.0166,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.6646e-02,  8.1516e-01, -2.5415e-02, -1.0673e-02,  3.0948e-04,
        -1.8657e-02, -5.6299e-03,  4.2203e-04,  2.5220e-03,  1.3029e-02,
        -1.6865e-02,  4.2262e-03,  3.7787e-04, -1.5734e-03, -8.1461e-04,
         1.5905e-03,  1.0392e-03, -4.9683e-03,  2.9524e-04, -1.1842e-03,
        -2.8536e-03, -5.7510e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3497e-03,  7.6339e-01,  6.2034e-03, -1.7525e-02,  2.5819e-02,
         4.9719e-02,  7.3240e-03,  4.4195e-03, -4.8595e-03, -6.4203e-04,
        -9.4101e-03, -1.5622e-02,  5.1570e-03, -9.1677e-03,  8.7395e-05,
        -3.2745e-03, -6.8397e-03, -4.4145e-03,  8.7367e-03, -1.1961e-03,
         8.6581e-04, -1.6479e-03,  8.9028e-03, -1.5697e-03, -4.1937e-04,
        -1.4460e-03,  2.0946e-03, -5.6479e-03, -7.6211e-03,  2.6138e-03,
        -4.4802e-05,  9.7546e-03,  1.0213e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7532e-02, -7.4362e-01, -4.1399e-02,  5.2345e-02,  8.1410e-04,
        -1.3980e-02, -2.8991e-03, -1.0225e-03,  2.7699e-03,  7.4878e-03,
        -2.9028e-03, -5.4566e-03,  3.6794e-03,  6.9412e-03, -1.0303e-02,
         6.3530e-03, -1.7652e-03,  3.8819e-03,  3.9135e-03,  4.2065e-03,
         5.6144e-03,  1.4604e-02, -9.0937e-03,  7.6810e-03, -1.5997e-02,
         6.7949e-03, -7.4009e-04,  6.2043e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7112e-02, -8.0959e-01, -2.9197e-02, -5.2572e-03, -3.0049e-02,
        -1.0406e-02, -1.8075e-03, -9.2990e-03,  1.0368e-02, -2.3161e-03,
         9.5596e-03, -2.7423e-03, -4.6614e-03,  6.1864e-03,  5.8773e-03,
        -4.7596e-04, -4.4609e-03,  2.6513e-03,  3.0158e-03, -1.0802e-02,
        -1.4223e-03,  1.1345e-04,  7.3744e-03, -4.3679e-03,  8.9156e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-6.0783e-02, -5.7573e-01,  4.5667e-02, -1.3883e-02, -1.7905e-02,
        -8.9181e-03,  7.3166e-04,  6.2514e-03, -6.9854e-03,  2.6345e-03,
        -2.1916e-03, -4.3830e-04,  4.0129e-03,  6.5817e-03,  1.2252e-02,
         3.4257e-02,  1.3181e-02,  7.3592e-03,  6.3683e-03, -6.8110e-03,
         4.5688e-03,  1.1321e-02,  2.0544e-02, -1.7877e-03,  1.3212e-02,
         1.0724e-02,  5.7815e-03, -3.6159e-04,  7.4609e-03,  5.5748e-03,
         1.8601e-03,  5.1101e-03,  6.5709e-03,  2.2333e-03,  3.6368e-03,
        -4.9698e-03,  3.9737e-03,  3.3800e-03, -2.4207e-03,  2.8928e-02,
         2.2641e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3643e-02, -7.1408e-01, -4.5717e-02,  1.1254e-02, -1.5281e-02,
        -3.0064e-03,  7.7264e-03, -8.3677e-03,  8.8982e-03,  3.1639e-03,
         8.8935e-03,  5.9596e-03, -4.4769e-03,  2.4100e-03,  2.8610e-03,
         1.0672e-02,  1.8430e-03,  5.2417e-03, -6.4347e-04, -1.8286e-02,
         2.4985e-03,  1.3696e-02,  1.0457e-03,  1.7357e-02,  4.2981e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4194e-02, -7.5333e-01, -6.0585e-02,  3.8053e-02, -1.6672e-02,
        -3.0826e-03,  4.6898e-03, -1.2977e-02,  4.3168e-03,  8.7010e-03,
        -4.4308e-04, -8.8815e-03,  5.9331e-03,  3.4749e-03,  1.2599e-02,
         7.4686e-03, -9.6546e-04,  4.9781e-03,  2.7065e-02,  3.7118e-03,
        -7.8776e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0081, -0.7681, -0.0494, -0.0011,  0.0062, -0.0078, -0.0032,  0.0238,
        -0.0084,  0.0121,  0.0146,  0.0099, -0.0114,  0.0086,  0.0253,  0.0089,
         0.0190,  0.0141,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8903e-02,  7.8868e-01, -1.8390e-02,  2.7812e-02,  2.9421e-03,
        -1.2685e-02, -7.8475e-03, -5.1117e-03, -6.4080e-03, -6.2193e-03,
        -4.7864e-03, -2.1608e-03, -6.7628e-03,  6.7859e-04, -9.3647e-04,
         6.6233e-03, -5.8595e-03, -6.2541e-03,  2.7657e-03, -1.5958e-03,
         2.4156e-03,  2.9589e-03,  9.9751e-04, -3.1139e-03,  1.5164e-02,
         1.3218e-03,  4.9811e-04, -6.2213e-04,  2.1711e-03, -7.6417e-03,
         1.6921e-03,  1.8098e-03, -3.9333e-03,  5.4703e-03, -8.9311e-03,
         5.8099e-04,  1.2083e-03,  6.0518e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.0842e-02,  7.2522e-01,  4.6689e-03, -1.3493e-02,  5.7493e-02,
         1.6272e-02, -7.7020e-03,  5.3835e-03,  2.0610e-02, -5.1184e-03,
         1.3333e-02,  7.6471e-03,  5.4433e-03, -2.6508e-04,  3.9441e-02,
         1.3274e-02, -6.9006e-03,  6.5748e-03,  4.0383e-03, -6.2855e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0047,  0.8465, -0.0089,  0.0195,  0.0144,  0.0195, -0.0094, -0.0072,
        -0.0070,  0.0013,  0.0075, -0.0088, -0.0228, -0.0226,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0099e-03,  7.7446e-01, -1.0353e-02, -2.4279e-02, -2.5064e-02,
        -6.6691e-03, -2.1904e-02, -8.2196e-03,  8.2117e-04, -5.8497e-04,
         2.0096e-03, -1.3991e-02, -4.1780e-03, -9.5782e-04,  4.9880e-03,
         1.1956e-03,  5.2798e-03,  4.5944e-03,  3.3704e-03,  2.3946e-03,
         7.0859e-03,  2.2386e-02,  4.7388e-03,  1.4375e-04,  4.4296e-03,
        -1.7608e-02, -1.9465e-03, -4.2747e-03, -3.6703e-03,  4.1478e-03,
         1.7957e-03, -8.4477e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0219, -0.8069, -0.0179, -0.0499, -0.0052,  0.0118, -0.0076,  0.0130,
         0.0100, -0.0049, -0.0021,  0.0009,  0.0081, -0.0043, -0.0053, -0.0019,
        -0.0017, -0.0083, -0.0031,  0.0152,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.8226e-02,  7.2725e-01, -1.5061e-02, -5.4633e-03,  1.0416e-02,
         1.1356e-03,  9.4512e-04, -6.1546e-03, -8.6383e-03,  3.8961e-03,
         2.4995e-03, -1.0843e-03,  5.5787e-04,  1.5916e-03, -9.6250e-04,
         1.1898e-03, -9.0613e-03, -5.8302e-03, -6.6674e-03, -1.7437e-03,
         5.4326e-03,  1.3172e-03,  5.4300e-03, -4.2569e-03,  2.0326e-02,
        -6.1881e-03, -7.7959e-03,  6.6760e-04,  6.9478e-03,  9.9054e-04,
         4.0997e-03, -9.3476e-06,  1.6113e-02,  8.8678e-03,  5.2395e-03,
         5.5664e-03, -5.7196e-03,  1.7463e-04, -1.1616e-04,  1.5119e-03,
         2.3319e-03, -5.4324e-03, -3.4267e-03, -3.3488e-03,  2.1826e-04,
         1.3891e-02, -6.2104e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0109,  0.8208, -0.0058, -0.0217,  0.0230, -0.0056, -0.0061,  0.0021,
        -0.0270,  0.0018, -0.0138,  0.0039, -0.0052, -0.0087, -0.0435,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.9274e-03,  6.9174e-01,  9.6527e-02,  9.6245e-03,  7.2264e-03,
         1.2049e-02, -6.1210e-03,  2.4858e-03, -4.6515e-03, -3.0598e-03,
         3.4451e-04, -5.3341e-03, -5.9878e-03,  3.5568e-03, -1.5976e-02,
         1.0267e-02, -2.1635e-04, -5.8981e-03, -6.6303e-03, -6.1932e-03,
         4.0530e-03, -4.6497e-03, -2.9412e-03, -2.7099e-03, -2.2078e-03,
         8.6977e-04, -1.1545e-03,  3.4651e-02,  4.0788e-04, -1.0081e-05,
        -2.3596e-03,  2.5319e-03,  1.3383e-03, -1.2751e-03, -1.9867e-03,
        -2.1726e-03, -1.3835e-03,  7.9756e-04, -1.0625e-03, -1.9080e-03,
         2.0242e-03,  3.2447e-03, -2.3359e-03, -1.1187e-03, -2.0368e-03,
        -3.2330e-04, -1.3380e-02, -4.2516e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 1.2441e-02,  8.4772e-01,  4.0811e-03, -1.3859e-02,  6.5509e-03,
        -9.5172e-03,  1.1027e-03, -1.1203e-03,  6.5199e-03,  3.4366e-03,
         6.9636e-03,  5.8789e-04, -1.3938e-03, -5.9383e-03, -1.2449e-03,
         1.2722e-03, -1.5114e-02, -4.0480e-03, -3.7281e-03, -1.0936e-02,
        -4.3667e-03, -6.5803e-03,  1.1928e-03,  2.4802e-04,  2.7471e-02,
        -2.5701e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0393, -0.7481, -0.0546,  0.0204,  0.0127, -0.0075,  0.0029,  0.0093,
        -0.0008, -0.0081, -0.0261,  0.0050,  0.0026, -0.0112, -0.0019,  0.0053,
         0.0015,  0.0098, -0.0057, -0.0120,  0.0152,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6595e-02,  7.9554e-01,  3.4050e-02, -3.8930e-04, -2.1021e-02,
        -1.6291e-02, -3.9660e-03, -5.1141e-03,  2.0892e-04, -1.2374e-04,
         3.2000e-03,  4.5922e-03,  5.9905e-04,  2.9928e-03, -4.0614e-03,
         3.3335e-03,  6.6235e-03, -1.2689e-03, -5.8983e-03, -1.6725e-02,
        -2.6319e-03,  8.5221e-03,  2.2627e-03, -5.0778e-04,  1.3445e-02,
         1.0049e-03, -1.7304e-03,  2.7904e-04,  1.7557e-03, -3.9315e-03,
        -4.8436e-03,  3.8095e-03,  5.4796e-03,  5.4218e-03, -1.0093e-03,
         7.6630e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2758e-02,  6.7364e-01, -4.9713e-02,  3.2556e-02,  1.9878e-02,
        -6.6055e-03, -9.8727e-03, -1.4535e-02,  5.5782e-04,  4.6579e-03,
         1.8404e-02, -7.4322e-03, -6.3767e-03, -4.9418e-03,  8.1649e-03,
         5.8297e-03, -1.2437e-03, -9.8557e-03,  3.9675e-02, -5.3304e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5545e-03,  8.1019e-01, -5.2603e-03,  2.6219e-02,  6.7251e-02,
         9.3942e-03, -2.1658e-03,  2.2760e-03,  1.4339e-02, -5.9650e-03,
         7.1005e-03,  6.7696e-03, -7.1052e-03,  1.2390e-02,  6.8211e-03,
         4.4298e-03, -3.0062e-04, -2.0498e-03, -4.1994e-04,  3.7415e-03,
         3.2622e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0786,  0.7867,  0.0048, -0.0277, -0.0173, -0.0057,  0.0011,  0.0147,
         0.0401, -0.0154,  0.0042,  0.0021,  0.0017,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6619e-02,  7.6924e-01,  3.1221e-02, -1.6681e-02,  6.1445e-03,
        -1.7473e-03, -2.0665e-02, -1.3362e-03, -3.9262e-03,  5.3084e-03,
        -5.8984e-03, -6.8079e-03, -4.9375e-03, -8.2148e-03,  1.2368e-02,
         3.4809e-03, -6.9711e-03,  2.1203e-03, -4.2712e-04, -3.5698e-03,
         4.2064e-03,  1.1204e-03,  2.8933e-03, -3.7630e-03,  3.1528e-03,
         1.6207e-03, -1.8368e-02,  6.3119e-03, -4.7441e-03,  4.5265e-04,
        -8.3371e-04,  6.8982e-03,  1.6607e-03, -1.3414e-03,  1.3060e-04,
        -3.7921e-03, -1.0295e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8693e-03,  8.1886e-01,  3.8783e-02,  6.7747e-03,  2.3453e-03,
        -5.5180e-03,  6.1440e-03,  8.5247e-03, -7.6274e-03, -1.5374e-02,
        -1.7797e-02, -1.0839e-02, -5.3763e-05,  1.9737e-03, -5.2829e-03,
         7.1357e-03,  7.8757e-03,  4.9919e-03, -2.3379e-03,  8.1928e-03,
        -4.0358e-03, -6.7429e-03, -1.0831e-03,  8.8418e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2911e-01,  5.9421e-01,  5.6075e-02,  1.1679e-02, -1.4726e-02,
        -4.4503e-03, -1.1491e-02,  1.5101e-02,  1.4583e-03,  1.6924e-03,
        -2.0133e-02,  7.6131e-03,  1.6942e-02,  7.0600e-03,  4.6405e-03,
        -1.0561e-02, -6.0181e-03,  6.3660e-03,  7.9000e-03,  4.2314e-03,
        -1.3683e-03, -9.9551e-03, -7.1250e-03, -3.8923e-04, -2.0748e-03,
         4.9581e-04,  6.8186e-03, -7.6911e-03,  5.1105e-03,  6.5174e-03,
        -1.4557e-02,  6.4385e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7985e-02,  7.7672e-01, -3.4225e-02, -9.8055e-03,  5.0403e-03,
        -2.4089e-02,  4.0396e-03,  6.6228e-03,  1.6772e-02, -9.7334e-03,
         2.0467e-03, -1.2689e-03, -1.6408e-03,  4.1875e-03, -8.2097e-04,
         2.7518e-03, -6.5874e-04, -1.5471e-03,  6.7359e-04,  9.6180e-03,
        -9.1911e-03,  6.6013e-03, -4.0444e-03, -5.5777e-03, -7.4401e-04,
         3.4086e-03,  8.0298e-03, -4.4708e-03, -6.3819e-03,  1.0182e-02,
        -1.1119e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9332e-02,  7.2689e-01, -9.4129e-02,  2.2807e-03, -7.1384e-03,
        -4.3235e-03,  8.6185e-03, -7.9433e-03,  2.4321e-03, -6.1497e-03,
         2.1269e-02,  6.9071e-03,  1.2720e-02, -1.5250e-02,  7.2661e-03,
        -3.1347e-03, -1.2767e-03,  6.5209e-04,  3.9602e-03,  3.6683e-03,
        -5.2733e-03, -1.0847e-02,  8.5408e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1858e-02,  7.8465e-01,  4.0860e-03, -1.4724e-02, -1.2728e-02,
        -1.3129e-03, -6.7944e-03, -5.3345e-03, -3.3811e-03, -1.2019e-03,
        -1.0210e-02,  7.4898e-03, -4.4899e-03, -3.3415e-02,  7.1237e-04,
        -1.4884e-03, -4.5886e-03, -8.9007e-03, -4.7929e-03, -1.2243e-03,
        -1.3851e-02, -6.3878e-03,  4.5591e-03, -3.7769e-03,  1.3421e-04,
         3.9279e-03,  4.0270e-03, -1.9953e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-3.3846e-02,  7.6926e-01, -2.3729e-02,  2.7778e-03,  2.0623e-03,
         1.5228e-02,  7.5768e-03, -2.6186e-02, -4.5295e-03, -1.5893e-02,
         6.2730e-03, -5.4601e-03,  8.3521e-04,  9.7658e-04,  7.1934e-03,
         7.2528e-03,  1.5714e-04, -1.7878e-03, -2.9615e-03,  5.7512e-04,
         1.5429e-02,  7.4214e-03, -1.0347e-02,  1.7823e-03, -1.2482e-03,
         2.7071e-03,  1.0375e-02,  9.8061e-03,  6.3235e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1973e-02, -7.6899e-01,  5.1629e-02, -1.7912e-02, -1.0468e-02,
        -1.3170e-02, -9.5206e-04,  7.1180e-03, -2.3620e-03, -1.2114e-02,
        -4.5490e-03, -1.4100e-03,  6.3703e-03, -6.9380e-03,  3.4226e-03,
         3.3229e-03, -5.3196e-03,  4.5825e-04,  8.3847e-03,  1.4409e-02,
         3.1759e-03, -6.5797e-03, -2.1828e-03, -2.6794e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.8672e-02, -6.9473e-01, -3.9064e-02,  1.6502e-02, -3.8250e-02,
        -1.8395e-03,  2.5506e-03, -9.3543e-04, -9.2283e-03,  9.4722e-03,
         1.4587e-03, -1.1195e-03,  9.7029e-03,  8.3577e-03, -8.9936e-03,
         7.4740e-03,  5.1010e-03, -1.0857e-03,  2.3050e-02, -8.9088e-03,
         1.4103e-02, -6.2581e-04,  2.7434e-03,  4.6988e-04,  6.7322e-04,
        -2.1012e-03,  1.9139e-03,  3.0362e-03,  8.3246e-04,  1.4804e-03,
         9.2530e-04, -3.2101e-03,  1.3576e-02,  7.8112e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0047e-03, -8.2658e-01,  1.7335e-02,  3.7500e-03,  3.1823e-04,
        -1.7528e-02, -1.0325e-02, -8.0370e-03, -8.8626e-03, -1.7243e-02,
        -1.2445e-03, -5.1150e-03, -3.1289e-03,  1.9006e-03, -1.8253e-02,
        -1.1230e-02, -1.1281e-02, -5.2926e-03, -3.4000e-03, -1.4055e-02,
         4.0040e-03,  8.1107e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0531, -0.7198,  0.0137, -0.0157, -0.0080,  0.0748,  0.0076, -0.0173,
        -0.0067,  0.0112, -0.0143,  0.0049,  0.0018, -0.0025, -0.0024,  0.0108,
         0.0342, -0.0012,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0951e-02, -7.2294e-01, -1.5107e-02, -1.8827e-02, -1.2401e-02,
        -7.7856e-03, -8.6252e-03, -2.0960e-02,  4.1967e-03, -9.9726e-03,
        -1.0372e-02, -5.1351e-03, -1.5520e-03,  7.3392e-03,  4.6835e-03,
        -5.9683e-03,  4.4461e-04,  3.1227e-03,  3.7636e-03,  7.8020e-03,
         5.1133e-04, -6.3387e-02,  4.1577e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0659e-02, -8.5148e-01,  2.5153e-02,  5.1776e-03,  2.0779e-02,
        -7.4099e-03, -6.2588e-03, -9.1540e-05,  7.8551e-03, -6.9990e-03,
        -1.4321e-03,  1.2283e-02, -3.8289e-03, -7.6807e-03, -5.0071e-03,
        -5.5722e-03, -1.2336e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0898, -0.7186, -0.0100,  0.0061,  0.0067, -0.0051, -0.0170, -0.0160,
         0.0084, -0.0279, -0.0049,  0.0009, -0.0129, -0.0032, -0.0102, -0.0125,
        -0.0127,  0.0369,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0570, -0.7418,  0.0034,  0.0073,  0.0070,  0.0062,  0.0009, -0.0020,
         0.0044, -0.0395,  0.0011,  0.0068,  0.0041, -0.0068,  0.0026, -0.0068,
         0.0147, -0.0126, -0.0122, -0.0057, -0.0074,  0.0010,  0.0159,  0.0062,
         0.0011,  0.0255,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.4866e-02,  7.4028e-01,  3.2054e-03, -1.2289e-02,  1.6546e-03,
        -1.6987e-02, -8.7669e-03, -6.4526e-03,  1.6341e-04, -6.2556e-03,
        -5.8534e-03,  1.1527e-02, -7.5454e-03,  6.1084e-03, -1.8447e-03,
        -2.2329e-03,  3.5276e-03,  5.1211e-03,  6.8624e-03, -5.5303e-03,
         9.5317e-03,  3.4636e-03,  6.1587e-03,  4.9435e-03, -2.0205e-03,
         1.8466e-03, -3.5182e-03, -8.5273e-04,  4.1471e-03,  1.6054e-02,
        -1.0392e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6457e-02, -7.7198e-01, -4.0681e-02,  1.5618e-02, -8.8288e-03,
        -6.9386e-03, -4.5345e-04, -6.2875e-03,  1.7597e-02,  6.9810e-03,
         1.7131e-02,  6.9189e-03,  1.7486e-03,  2.6921e-03,  8.8163e-03,
        -6.4301e-03, -2.8320e-02,  2.6125e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0264, -0.7892,  0.0888,  0.0435,  0.0147, -0.0213,  0.0019,  0.0141,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 8.0205e-02,  6.8327e-01,  2.6710e-02, -1.6706e-02, -2.4841e-02,
         1.6871e-02,  6.3272e-03,  7.5343e-03,  1.0048e-02,  1.7468e-02,
         4.1198e-03, -3.5788e-04,  1.0422e-03,  1.0971e-03, -5.1189e-03,
         1.8012e-02,  6.8176e-03,  2.0952e-03, -6.3285e-04,  1.0357e-02,
        -1.4380e-03, -5.8950e-03, -1.4297e-03, -3.2630e-03, -7.8668e-03,
         5.9093e-04, -6.8016e-03,  2.6751e-03,  4.2048e-03,  2.1017e-03,
        -2.5543e-03,  8.0414e-04,  2.1688e-04,  3.7070e-03, -7.2612e-04,
        -3.8850e-03,  7.3367e-03,  2.2117e-03, -2.6598e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4844e-02,  7.6534e-01, -4.4265e-02,  3.4818e-03,  8.8026e-03,
         9.6762e-03, -3.0083e-04, -2.6328e-04,  1.6113e-03, -5.9799e-03,
        -3.0490e-03,  3.6387e-03,  1.1935e-02,  2.0459e-03,  6.4355e-03,
         1.2349e-02, -8.7562e-03, -3.3316e-03,  5.6336e-03,  3.3890e-03,
         9.6806e-04, -7.9511e-03, -3.0509e-04, -2.1966e-03,  5.2480e-03,
        -6.7055e-03, -1.6964e-02, -2.4534e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0013, -0.8369, -0.0114,  0.0081, -0.0232, -0.0104,  0.0106, -0.0190,
        -0.0086, -0.0299, -0.0155, -0.0243, -0.0010,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0267, -0.7893,  0.0373, -0.0059, -0.0014,  0.0109,  0.0121, -0.0020,
        -0.0019,  0.0167, -0.0032,  0.0196, -0.0096, -0.0219,  0.0415,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1201e-02,  6.5071e-01, -5.1447e-02,  3.4172e-02,  2.5610e-02,
        -1.2972e-02,  7.2251e-03, -5.9133e-03, -4.8957e-03, -3.2223e-03,
        -7.3304e-03, -2.4103e-03,  1.7191e-02, -3.0980e-04,  1.4799e-02,
        -2.8772e-03, -4.0638e-03, -1.4201e-03,  2.1793e-03, -5.9732e-03,
        -6.9496e-03, -2.1604e-03, -6.6957e-04, -6.8897e-03,  2.3227e-03,
        -4.5116e-03,  1.6829e-03,  3.3480e-03, -2.1548e-03,  1.0062e-03,
        -4.3843e-03,  1.8291e-03, -2.0866e-03,  6.1939e-03,  1.0180e-02,
         4.3070e-03, -4.2499e-03, -4.2834e-03, -4.3677e-03, -1.5782e-03,
        -2.4372e-03, -8.0855e-03,  2.5181e-03, -2.0320e-03, -1.1261e-03,
         4.8213e-03, -2.7903e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9731e-03, -7.6166e-01, -3.2800e-02,  6.4293e-03,  4.4911e-03,
         4.5040e-03,  6.7695e-03, -1.0479e-02, -6.0570e-03,  1.6560e-03,
         1.8176e-04, -1.2016e-02, -2.0800e-03,  5.5141e-03, -4.9976e-04,
         5.7994e-03, -8.2080e-03, -7.9110e-03, -1.3581e-03, -9.0927e-04,
        -5.2537e-03,  4.6665e-03, -2.2590e-03, -2.9551e-03, -3.8681e-03,
         7.2816e-03, -1.9593e-02,  5.5280e-04, -5.9547e-03, -3.2899e-03,
        -4.1488e-03, -1.6471e-04, -7.8506e-04, -9.7487e-05,  1.8004e-03,
        -2.6891e-03,  3.2846e-03, -2.8001e-03,  4.3888e-03,  9.9001e-03,
         2.1667e-03, -2.3804e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0395,  0.7061,  0.0266, -0.0074,  0.0176,  0.0170,  0.0358, -0.0009,
         0.0051,  0.0158,  0.0050,  0.0055, -0.0076, -0.0084,  0.0121, -0.0194,
         0.0388,  0.0011,  0.0018,  0.0025, -0.0045, -0.0015, -0.0017,  0.0040,
        -0.0146,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0588,  0.7419,  0.0377, -0.0127, -0.0108, -0.0116,  0.0046, -0.0025,
        -0.0034,  0.0037, -0.0008, -0.0028,  0.0063, -0.0176, -0.0098,  0.0087,
         0.0090,  0.0032,  0.0050,  0.0082,  0.0091,  0.0087, -0.0232,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7342e-02,  7.5954e-01,  3.4943e-02,  4.6540e-03, -8.5448e-03,
         4.5161e-03,  1.9379e-03,  5.4841e-04, -3.6334e-02,  4.9097e-03,
         8.5945e-03, -1.2416e-02,  9.3924e-03, -1.1623e-02, -9.2147e-03,
        -5.5558e-03, -4.9206e-04, -3.4351e-03,  3.0840e-03, -6.6649e-03,
        -8.0526e-03,  1.0216e-02,  7.9930e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1170,  0.6817, -0.0267, -0.0127, -0.0270,  0.0040, -0.0085,  0.0160,
         0.0026, -0.0239,  0.0135, -0.0455, -0.0210,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0047, -0.8615, -0.0237,  0.0194,  0.0081, -0.0059,  0.0183,  0.0111,
        -0.0060,  0.0134, -0.0021, -0.0101, -0.0009, -0.0148,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0176,  0.8594, -0.0077,  0.0102, -0.0077,  0.0020, -0.0028, -0.0087,
        -0.0086, -0.0096, -0.0067, -0.0087, -0.0081,  0.0032, -0.0024,  0.0079,
        -0.0165,  0.0095,  0.0027,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 5.3199e-02,  6.7379e-01, -4.0747e-02, -2.6518e-02,  1.3619e-04,
         1.9770e-02,  2.2671e-03,  2.0371e-02, -4.1458e-03,  1.5316e-02,
        -2.3590e-03,  1.1201e-03, -3.5314e-03,  1.3077e-02,  9.2260e-03,
        -1.6495e-02,  3.6324e-03,  2.0376e-03, -7.8376e-03,  6.4013e-03,
        -6.9248e-03, -1.6079e-03, -3.6046e-03,  4.9749e-03, -6.4502e-03,
        -5.4465e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8808e-02, -7.9663e-01,  2.3308e-02,  1.4724e-02, -2.7332e-03,
         1.8448e-02,  5.9073e-03,  1.3348e-02,  2.4120e-03,  3.5539e-02,
        -3.7364e-03, -1.7483e-02, -1.7746e-02, -1.5932e-02,  2.5762e-03,
        -6.6525e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0377,  0.7883,  0.0044,  0.0177,  0.0155,  0.0182, -0.0097,  0.0165,
         0.0157,  0.0123, -0.0021,  0.0009, -0.0021, -0.0126, -0.0056,  0.0372,
        -0.0036,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0716,  0.8242, -0.0031,  0.0089, -0.0164,  0.0180,  0.0071,  0.0203,
         0.0303,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6746e-05,  7.1239e-01, -2.5122e-02,  1.4878e-03,  2.4227e-03,
         2.7689e-03,  2.9153e-04,  8.8687e-03,  2.6593e-03,  4.0690e-03,
         3.0642e-02, -1.4679e-03,  8.8642e-03, -1.4870e-02,  4.0154e-03,
         3.6519e-03,  1.5626e-03,  2.0705e-03,  8.8065e-03,  1.2529e-02,
         5.7282e-03,  3.5152e-03, -9.6237e-03, -4.6123e-03,  3.8507e-03,
         4.3832e-03, -1.0002e-02, -3.6027e-03, -5.8345e-04, -6.9231e-03,
        -7.2580e-03,  1.7998e-04, -2.6008e-03,  4.4494e-03, -9.1576e-04,
        -1.2523e-02, -3.9932e-03,  1.2222e-02, -6.7153e-03, -5.0904e-03,
        -2.2539e-03,  6.2358e-04, -1.4001e-02,  1.5936e-02, -3.0084e-03,
         6.8127e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.6165e-03,  8.3769e-01, -1.3611e-02,  9.7716e-03,  1.1179e-02,
        -1.3620e-03,  1.0294e-02,  1.4367e-03, -2.0185e-03,  4.9230e-03,
        -2.1851e-03, -8.0161e-04, -4.2983e-03,  4.6593e-03, -5.7457e-03,
         1.0212e-02,  1.7268e-03,  1.4694e-02, -1.4368e-03,  5.9361e-03,
         4.2039e-03,  2.9001e-03, -5.8680e-03,  2.8487e-04, -3.9542e-03,
        -1.6074e-03,  2.1545e-03, -1.6097e-03,  1.3314e-02,  1.0501e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0230, -0.8711, -0.0281,  0.0378,  0.0016, -0.0074,  0.0311,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0614e-02, -8.1545e-01,  2.7424e-02,  1.8679e-02, -3.5602e-02,
         1.2927e-02,  9.4901e-05, -1.4221e-02,  2.3837e-02,  1.9976e-02,
         1.1719e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0033, -0.8392, -0.0146, -0.0137,  0.0037, -0.0356,  0.0052, -0.0150,
        -0.0166, -0.0043,  0.0023, -0.0094, -0.0352,  0.0019,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7941e-02, -7.7105e-01,  4.3471e-03, -5.2856e-03,  6.2282e-03,
        -2.0036e-03,  3.0524e-04,  1.6268e-03,  4.8117e-03,  9.4154e-03,
        -9.1508e-04,  2.4381e-03,  9.6809e-03,  1.8598e-02,  3.1962e-03,
        -2.3947e-03,  3.4114e-04,  1.9509e-03, -6.0612e-03, -1.5078e-02,
        -3.1866e-03, -1.2113e-04,  3.3843e-03,  1.0359e-03,  4.3786e-03,
        -4.8389e-03,  1.0774e-02, -1.2401e-02,  3.7254e-04,  3.8938e-03,
         1.5919e-02, -5.5812e-03, -1.2982e-02, -1.7460e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.8275e-03,  7.1534e-01,  3.1547e-02,  1.1029e-02,  5.7399e-03,
        -4.8556e-04,  1.4924e-02,  6.5610e-03, -7.5102e-03,  8.2066e-03,
        -2.3643e-03, -1.1675e-02, -1.1673e-03,  1.2145e-02,  2.8715e-02,
         1.0510e-03, -2.3040e-02, -1.4836e-02,  7.6088e-04, -1.4557e-03,
        -8.9930e-03,  1.4087e-03, -1.6729e-02,  3.0589e-03, -3.0206e-03,
        -6.3311e-03,  8.5805e-03,  4.5494e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0757,  0.6653,  0.0806,  0.0070,  0.0254,  0.0013, -0.0021,  0.0120,
         0.0087,  0.0117,  0.0048, -0.0095,  0.0068,  0.0069, -0.0062,  0.0091,
        -0.0048, -0.0027, -0.0096,  0.0032,  0.0058,  0.0055,  0.0297,  0.0058,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([-3.8581e-02, -6.8498e-01,  3.2087e-02, -1.8249e-03, -1.0330e-02,
        -1.8558e-03, -2.4806e-02, -1.1235e-02, -1.1540e-02, -1.0851e-03,
         2.7710e-02,  2.8443e-03,  4.5169e-03, -4.4210e-03, -6.2734e-03,
         3.0969e-03,  3.1813e-03,  2.7518e-04,  7.9460e-03, -2.6603e-02,
         1.9716e-03,  1.4016e-02, -7.1990e-04, -2.7665e-03, -4.1800e-03,
        -2.0324e-02, -6.3426e-03, -8.2888e-03, -1.8908e-03,  1.0166e-02,
         2.4143e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0282, -0.8034,  0.0290, -0.0192, -0.0145,  0.0065,  0.0178,  0.0054,
        -0.0075, -0.0098, -0.0038,  0.0549,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0800,  0.7807,  0.0281, -0.0062,  0.0122,  0.0028, -0.0098, -0.0047,
         0.0020, -0.0023, -0.0180, -0.0305,  0.0187,  0.0027, -0.0012,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7171e-02,  7.7082e-01,  1.4363e-02,  1.0835e-03,  4.0514e-02,
         1.4426e-02,  2.5449e-02, -3.8192e-03,  2.2025e-03, -1.0631e-04,
        -5.1505e-04,  1.0332e-02, -1.9288e-03, -2.9977e-03,  8.0894e-04,
         2.7530e-03,  1.3000e-03, -4.3715e-03, -4.3384e-04, -6.1829e-03,
         5.9705e-04,  2.6944e-04,  7.0435e-03, -3.9237e-03,  1.0334e-03,
        -4.5079e-03,  9.1177e-04, -2.3524e-03, -5.9764e-03,  2.2932e-03,
         1.6329e-03, -7.6125e-03,  1.0268e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.1946e-02, -8.1268e-01,  3.0804e-02,  2.9288e-02, -7.2843e-03,
         6.5708e-03, -2.8515e-03,  2.3229e-03,  4.5556e-04, -2.0372e-03,
         7.4111e-03,  5.1912e-03,  1.2753e-02, -1.0859e-02, -1.4927e-02,
         2.6223e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0644,  0.8136, -0.0221,  0.0181,  0.0153, -0.0141, -0.0179, -0.0266,
         0.0078,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0566,  0.7718, -0.0218,  0.0827, -0.0671,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0840,  0.6564, -0.0717,  0.0095, -0.0066, -0.0186, -0.0194,  0.0143,
        -0.0032,  0.0225,  0.0123, -0.0060, -0.0100,  0.0028, -0.0034,  0.0026,
         0.0088, -0.0009,  0.0019,  0.0010,  0.0030, -0.0030,  0.0033,  0.0022,
         0.0055, -0.0149, -0.0123,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0533,  0.7813,  0.0513,  0.0202,  0.0646, -0.0095,  0.0197,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0013,  0.7578,  0.0404,  0.0096, -0.0218,  0.0111,  0.0100, -0.0015,
        -0.0101, -0.0029,  0.0077, -0.0022, -0.0049,  0.0044,  0.0044,  0.0060,
         0.0025, -0.0128, -0.0167, -0.0138,  0.0142,  0.0439,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.8216e-04,  8.4216e-01, -2.7580e-02,  9.0847e-03,  1.6490e-02,
         9.5606e-03, -6.2057e-03,  1.5415e-02, -4.9174e-03,  1.4394e-03,
         1.1441e-02,  4.6369e-03, -2.4682e-04,  1.5340e-02, -9.8286e-04,
         5.9200e-03, -1.6921e-03, -1.7829e-03, -2.4428e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1915e-02,  7.9096e-01, -2.5702e-02, -1.1215e-02, -1.4332e-02,
         3.2365e-03,  5.7963e-03,  9.2507e-03,  2.5476e-03,  1.0663e-02,
         1.1639e-03,  5.4783e-03,  4.6687e-03,  4.5084e-03,  1.5666e-03,
        -6.6522e-04,  7.6026e-03, -1.4898e-03,  2.4056e-03,  6.3142e-03,
         7.2748e-03,  9.0035e-03,  5.6481e-03, -1.9895e-03,  1.1538e-02,
         2.1113e-03, -1.5910e-03,  8.2085e-03,  1.5239e-02,  1.4261e-02,
         1.6578e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 9.2063e-02, -6.9985e-01,  3.4255e-03,  9.1801e-03, -2.1436e-02,
        -1.0968e-02,  1.2561e-02,  3.7855e-03,  1.6457e-02, -1.9030e-03,
        -3.9853e-03, -6.2833e-03, -8.2068e-03, -2.3400e-02,  8.8909e-03,
        -6.8719e-03,  8.8637e-03, -2.0674e-03,  3.2982e-03, -4.7445e-04,
        -2.3591e-03, -9.4837e-04, -6.8076e-03, -2.4572e-03,  8.2263e-04,
        -3.6466e-03, -1.3132e-03,  5.1418e-03,  4.2130e-03, -7.2939e-03,
         6.1319e-03, -1.4898e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5828e-02, -7.6328e-01,  3.3215e-02,  4.1303e-03,  9.8505e-03,
         9.8912e-03,  2.1186e-02, -2.0141e-02, -8.7365e-03,  6.5840e-04,
         4.6493e-02, -8.8666e-03,  7.2091e-04,  1.0736e-02, -2.1484e-03,
         1.3127e-02, -9.5806e-03,  3.3652e-03, -2.5836e-03,  6.0742e-03,
        -2.2515e-03, -7.1345e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0327,  0.7870,  0.0267,  0.0015,  0.0024,  0.0053, -0.0069,  0.0013,
        -0.0044, -0.0085, -0.0057,  0.0013, -0.0023,  0.0015, -0.0063, -0.0044,
        -0.0011,  0.0085, -0.0008,  0.0088,  0.0063, -0.0081,  0.0019, -0.0051,
        -0.0038,  0.0035, -0.0020, -0.0250,  0.0237,  0.0030,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5005e-02,  7.7175e-01, -5.3614e-02, -3.0921e-02,  1.0021e-03,
         1.5500e-02,  6.2524e-03, -4.0598e-03,  3.9600e-03, -3.8323e-03,
        -1.1929e-02,  3.1682e-03, -9.9568e-03,  3.7669e-03, -3.8227e-04,
         5.9032e-03,  2.1990e-02, -7.0091e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4193e-02, -7.5384e-01, -4.0788e-02, -1.1148e-02,  7.1698e-03,
         1.1107e-02,  6.9299e-04, -2.4646e-03,  3.1485e-03,  5.8909e-05,
         1.3841e-03, -3.7847e-03, -1.4304e-03, -1.4264e-03,  1.4355e-03,
         1.3202e-02, -1.3931e-02, -1.8934e-02,  4.5133e-03, -3.1491e-03,
         1.1284e-03,  5.2071e-03, -7.5944e-04, -1.2098e-03, -3.2761e-03,
         1.5745e-03,  1.0341e-03,  3.6654e-04, -1.1282e-03, -3.9003e-03,
        -3.6599e-04, -3.5029e-03, -7.7661e-04,  2.0923e-03,  8.7874e-04,
        -2.3820e-02, -2.1173e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6227e-02,  7.2726e-01, -1.1052e-01, -8.1669e-03, -3.1443e-03,
         7.7726e-03, -4.3455e-03,  7.2968e-03,  3.0114e-03,  4.2703e-03,
         2.9504e-03, -8.6570e-03,  4.6060e-03,  9.4278e-03,  3.8660e-03,
         5.1611e-03,  4.4662e-04,  6.8354e-03,  4.5753e-03, -2.1446e-03,
         2.7394e-02, -7.7515e-03,  1.0763e-02,  1.3407e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.0807e-02,  7.6611e-01,  1.4991e-02,  2.1247e-02,  2.3597e-02,
         5.2650e-03, -1.2114e-02,  1.1372e-02,  2.5715e-03, -1.0059e-03,
         1.9021e-03, -2.9170e-03, -1.0292e-02, -1.9152e-04,  6.7450e-04,
        -9.9498e-04,  7.7357e-04, -1.2130e-03,  1.4092e-03, -6.9877e-04,
         5.0574e-04, -2.6509e-03,  1.3378e-03, -5.3302e-04,  5.1411e-03,
         9.1097e-03, -4.9299e-04,  1.8283e-03, -1.3923e-03,  2.6739e-03,
         1.1916e-02, -4.4403e-04,  5.1287e-03,  6.9094e-03,  4.4701e-03,
        -9.1923e-04, -2.2357e-03, -7.1516e-04,  1.2953e-03,  2.3692e-04,
        -1.7872e-03,  2.4558e-03, -5.2142e-03,  1.0726e-03, -2.2084e-03,
         7.1806e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8953e-02,  7.7243e-01,  5.9294e-03,  1.5017e-02,  2.2893e-02,
         1.2073e-02, -1.2190e-02, -7.2588e-03, -1.1346e-03, -1.3688e-02,
         1.3098e-04, -5.7477e-03,  1.2741e-02, -5.1871e-03, -6.4235e-03,
        -6.0257e-03, -8.0889e-03,  3.8253e-03, -2.6739e-03,  7.1411e-03,
         1.9786e-02,  1.7582e-02,  3.0838e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2485e-02,  8.5759e-01, -3.6025e-02, -1.4216e-02,  1.8385e-02,
         9.8710e-03, -6.5160e-04,  6.2674e-03,  1.1990e-02, -1.1410e-02,
        -1.1042e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6161e-02,  8.0101e-01,  2.4984e-02,  2.1819e-02,  1.7054e-02,
         1.7619e-02,  2.0129e-03,  9.7465e-03, -3.8110e-03,  4.2053e-04,
         5.1485e-03,  9.0500e-03,  4.4566e-02,  6.5930e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0560,  0.8177,  0.0112, -0.0039, -0.0041,  0.0092, -0.0049, -0.0130,
        -0.0072, -0.0105,  0.0145, -0.0479,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9756e-02, -7.3370e-01, -4.1463e-02, -2.3801e-02, -3.4021e-03,
        -6.9105e-03,  1.0025e-02,  4.2741e-03,  6.5873e-04, -2.7748e-03,
         6.0512e-03, -8.5694e-04,  3.4460e-03, -3.9659e-03,  2.7167e-03,
         2.2803e-03, -5.4561e-04,  4.2744e-03, -9.0145e-04,  1.2546e-02,
         1.0956e-03, -2.3625e-03, -4.0050e-03,  3.4778e-03, -5.1667e-03,
         1.6334e-03, -9.0776e-03,  4.3431e-04,  5.6447e-03,  1.4720e-03,
         1.4754e-03, -3.9753e-03,  1.3240e-03, -1.3406e-03,  1.2615e-03,
         1.9871e-03, -1.2993e-04,  3.1928e-03,  3.7093e-03,  2.5978e-03,
        -2.0944e-03, -8.6318e-04,  5.2133e-04,  1.6242e-03, -2.7325e-03,
         2.4914e-03, -2.8477e-03,  2.0255e-03,  1.6999e-03,  4.6050e-03,
        -3.5651e-04, -3.6500e-04,  2.5823e-03,  3.4712e-03,  1.2011e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-4.8668e-02, -7.7640e-01, -1.3249e-02, -1.1120e-02,  5.7336e-03,
         1.4256e-03,  1.1295e-02, -4.1154e-03,  3.8319e-03,  1.6026e-03,
        -4.4034e-04,  2.5974e-03, -3.3256e-03, -1.3433e-02,  4.4422e-03,
        -2.5025e-03,  5.4941e-03, -3.3901e-02, -4.0088e-03, -5.4471e-03,
        -1.7719e-03,  1.1949e-02,  3.3251e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0875,  0.6833,  0.0753,  0.0190,  0.0112, -0.0015,  0.0066,  0.0026,
         0.0207, -0.0284, -0.0014,  0.0140, -0.0211,  0.0014,  0.0010, -0.0031,
         0.0197,  0.0022,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8196e-03, -7.4105e-01, -9.0311e-02, -6.0894e-03, -1.3997e-02,
        -1.4551e-04,  1.2738e-02,  7.0423e-03,  2.1301e-02,  1.5700e-02,
         1.2946e-02,  1.2618e-02,  5.0196e-03,  1.0802e-02,  1.8569e-02,
         1.8112e-03, -1.2642e-02, -1.2399e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0280e-02, -8.8959e-01, -9.6147e-03,  1.2866e-02,  7.9999e-03,
        -7.8815e-03, -4.7172e-03, -2.6590e-03, -9.6007e-03,  4.4539e-03,
         1.3577e-02,  6.0831e-03, -1.4444e-02, -5.7809e-03, -4.5779e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.7224e-02,  6.6006e-01,  1.7123e-02, -1.7562e-02,  4.7634e-04,
         2.3715e-04, -1.1596e-03, -6.5982e-03,  1.5191e-02,  1.3552e-02,
         8.6641e-03,  1.0394e-02, -1.0904e-04, -9.0513e-03, -1.8638e-02,
        -9.8771e-03,  1.1011e-02, -8.5500e-03, -1.5465e-03, -4.1139e-03,
         1.7621e-02, -4.0135e-03, -1.8286e-03,  1.7132e-03, -1.6355e-02,
        -5.4888e-03,  1.1161e-02, -4.2223e-03,  1.5397e-02,  1.9159e-02,
        -2.7245e-03, -9.1823e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.9490e-02, -6.8756e-01, -3.4026e-02, -1.9632e-02,  1.1430e-02,
        -1.0417e-02, -5.6794e-03,  7.6210e-03,  7.3463e-03,  6.1986e-03,
        -1.0148e-03,  9.3795e-03, -7.6472e-03,  9.4947e-03,  1.6253e-03,
         3.7254e-03,  3.3738e-03,  1.2628e-02, -3.6823e-03, -9.9706e-03,
         7.0375e-03,  3.2993e-03,  1.4076e-02,  5.3025e-03, -9.1382e-03,
         1.3004e-03,  5.6642e-04, -7.7775e-03,  3.1316e-03,  6.1291e-03,
        -8.4228e-03,  1.8416e-03, -3.0037e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5177e-03,  7.9183e-01,  3.3920e-03,  4.1699e-03,  1.8414e-02,
         7.4135e-03, -7.6344e-03, -1.1415e-02, -3.9534e-03,  6.2655e-03,
        -1.8234e-03, -9.6034e-04, -2.4153e-03, -7.9583e-03, -3.6186e-03,
         3.1627e-05,  3.6326e-03, -2.7885e-03, -5.7833e-03,  3.1746e-03,
        -2.7357e-03,  3.6156e-03, -1.2739e-03, -2.1263e-03, -1.2726e-03,
        -7.7392e-04,  3.6949e-03, -2.7967e-03,  1.5489e-03, -7.3337e-03,
         1.9203e-03,  4.3863e-03,  4.2845e-03,  4.7792e-03,  3.7213e-03,
        -2.6566e-03, -1.1317e-03,  2.9657e-04,  2.5449e-03,  1.6777e-03,
         1.6178e-03,  7.7000e-04,  9.1829e-04,  2.4516e-03,  1.7876e-03,
         3.8242e-03, -3.0980e-03, -2.1934e-04,  1.1625e-03,  8.6452e-03,
        -2.5566e-03,  3.1495e-03,  4.4206e-03,  1.5526e-03,  2.8486e-03,
         4.2637e-03,  4.0397e-03,  9.9140e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.9839e-02,  7.4914e-01,  6.4492e-03, -3.0702e-02,  3.3232e-03,
        -1.5125e-03,  1.1173e-03, -4.9488e-03, -5.2728e-03,  1.5328e-02,
        -5.6363e-03,  4.7464e-03, -1.7058e-02, -9.2491e-03, -3.8943e-03,
        -2.3024e-03,  5.8365e-03,  2.2857e-04, -9.1096e-05,  4.5591e-03,
         1.9189e-03,  2.2956e-03, -7.6359e-04,  1.2692e-02, -1.1089e-02,
        -5.0004e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4406e-02, -7.1879e-01,  3.3479e-02,  2.0778e-02, -3.4279e-03,
        -1.1747e-03,  1.0798e-02,  6.2219e-04, -4.9409e-04,  4.6975e-03,
         7.1323e-03,  2.2526e-03,  8.8094e-03, -4.0564e-03,  9.3564e-03,
         4.7969e-03, -4.3166e-03, -1.0934e-03, -1.0215e-02,  8.8524e-03,
        -1.3832e-03,  1.3419e-02, -3.9046e-03, -1.5345e-02,  6.6404e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.6519e-02,  7.0202e-01,  4.4063e-02, -9.3856e-03,  1.7983e-02,
        -7.0047e-03,  1.5552e-03,  4.3690e-03, -3.5566e-03, -8.8676e-03,
        -1.7109e-03, -8.2873e-03, -9.4162e-03, -9.4226e-03,  1.9596e-03,
         1.0948e-02, -4.8094e-03,  9.4093e-03, -5.6245e-04, -5.0461e-04,
        -7.8531e-03,  2.9772e-03, -5.3744e-03, -5.3921e-03,  3.8578e-03,
        -5.8655e-03,  4.1325e-04, -2.8373e-03, -1.8568e-03, -1.8768e-03,
         3.6760e-04, -7.6045e-04,  1.6737e-03,  2.9284e-03, -6.8713e-03,
        -1.4966e-04, -7.7794e-04, -3.9296e-03, -8.4456e-04,  1.9137e-03,
        -6.4918e-03, -2.3738e-03,  8.7082e-03,  1.0060e-02,  1.1493e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0231,  0.7795,  0.0388,  0.0057, -0.0022,  0.0031,  0.0050,  0.0065,
        -0.0101,  0.0079, -0.0080, -0.0109, -0.0083,  0.0077,  0.0038, -0.0063,
        -0.0195, -0.0046, -0.0042,  0.0021,  0.0121,  0.0059,  0.0244,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0112,  0.7903, -0.0577, -0.0347, -0.0067,  0.0103, -0.0275, -0.0118,
         0.0234,  0.0223, -0.0020,  0.0022,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-5.9323e-03, -7.2858e-01, -5.7320e-02,  1.0852e-02,  7.0065e-04,
        -3.2841e-02, -6.4991e-03,  1.8185e-02, -1.8274e-02,  1.8373e-02,
         2.7016e-02,  7.4788e-03, -2.9818e-03,  5.4369e-03, -7.7629e-03,
         5.4387e-03,  2.3864e-03,  1.0626e-02,  2.0694e-03,  7.3189e-03,
         6.4135e-03,  1.3571e-02, -3.9409e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7212e-02, -7.9670e-01,  1.0458e-02,  6.8863e-03,  1.3679e-02,
        -2.7633e-03,  1.2327e-02, -8.6058e-03,  1.2550e-03, -6.0862e-03,
        -2.1136e-04,  6.0533e-03, -5.3826e-03, -1.5191e-03,  5.8769e-03,
         1.1635e-03,  2.0738e-03,  2.6351e-02, -2.1903e-03,  2.0885e-03,
         3.1961e-03,  3.0649e-04,  1.0638e-02,  5.1766e-03,  7.3785e-03,
        -5.5670e-04,  1.0833e-03, -1.5907e-03, -1.3702e-03, -9.7333e-04,
         3.6580e-03,  1.6284e-02, -8.9017e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.9168e-04, -8.3440e-01, -2.1466e-02,  1.6039e-02, -2.0648e-02,
        -3.4922e-04,  8.2076e-03,  9.0427e-03, -1.0123e-02, -5.3611e-03,
         2.0062e-03, -4.9567e-03, -8.8453e-03, -3.3733e-03, -1.0256e-02,
         5.1704e-03,  6.4865e-04, -2.8506e-02,  1.0307e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0582, -0.8272,  0.0108, -0.0053,  0.0517,  0.0030,  0.0060,  0.0080,
         0.0054,  0.0077, -0.0064, -0.0032,  0.0011,  0.0059,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.6304e-03, -8.4029e-01, -2.8425e-02,  2.2091e-02,  2.1194e-02,
         3.4255e-03, -7.1239e-03,  4.5006e-03, -4.9545e-03, -4.4972e-03,
        -4.6481e-03, -6.1047e-03,  1.4764e-03,  9.2703e-04, -3.7307e-06,
         3.1240e-03,  2.0718e-03, -5.3786e-03, -6.5282e-03, -1.1796e-02,
        -4.3496e-03,  7.4556e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2372e-03,  8.1835e-01,  6.9467e-03, -1.0713e-02, -1.6038e-04,
        -5.2090e-03, -2.6490e-03,  3.1630e-03, -1.2852e-03, -4.2297e-03,
        -4.1114e-03, -1.4626e-02, -1.8214e-02, -7.9003e-03,  1.2490e-03,
         2.2118e-03,  4.0019e-02,  8.2170e-04, -2.7053e-03,  6.6337e-03,
         4.0320e-03, -1.1844e-03,  9.7442e-04, -8.0679e-03, -6.0781e-03,
         3.8502e-03, -2.3450e-03, -7.3107e-03,  5.0991e-04, -1.1213e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2998e-01,  6.1517e-01, -4.5505e-02, -4.1966e-02,  1.1023e-03,
        -1.6275e-04,  1.3630e-02, -1.9571e-03, -4.6652e-03, -1.5621e-03,
         3.6511e-04, -3.3544e-02, -3.6218e-03,  1.1938e-02,  4.6447e-03,
        -5.9629e-02,  3.0552e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0062, -0.8342,  0.0275,  0.0215,  0.0131, -0.0295,  0.0208,  0.0033,
        -0.0338,  0.0044,  0.0057,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0583, -0.7344, -0.0393, -0.0036, -0.0247,  0.0246, -0.0070,  0.0137,
         0.0099, -0.0090, -0.0035,  0.0090, -0.0102, -0.0162,  0.0018, -0.0092,
         0.0258,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2947e-02,  7.5480e-01, -2.4033e-02,  3.6650e-02,  2.8324e-03,
         4.0158e-03,  8.1092e-03,  3.5788e-03, -7.0336e-03,  6.6531e-03,
        -1.9972e-03,  5.0303e-03, -1.1391e-02, -9.3649e-04,  3.0544e-02,
         4.4102e-03,  1.2479e-03, -6.9557e-03, -4.7731e-03,  4.4633e-03,
         4.6681e-03,  6.9941e-04,  1.0391e-02,  4.1430e-03,  6.1675e-03,
        -1.1526e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1322e-02,  7.7194e-01, -5.1672e-02,  4.7440e-03,  6.3591e-03,
         5.3478e-03, -7.0183e-04, -1.2256e-03,  1.5091e-03, -4.9372e-04,
        -8.1394e-03, -9.7732e-03,  9.6202e-04, -9.1571e-03,  9.6282e-03,
        -2.1873e-03, -6.9412e-04,  5.5562e-03, -2.3125e-03, -4.3392e-03,
         3.4875e-03, -7.0847e-04, -1.7414e-03, -1.2697e-03,  1.1390e-02,
         2.2911e-03,  5.8757e-03, -2.3471e-03,  4.3792e-03,  2.8444e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2003e-02, -7.8998e-01,  6.8421e-03,  2.4795e-02, -1.1290e-04,
         2.3110e-02,  2.1604e-02,  2.4784e-02,  1.6880e-03,  1.6593e-02,
         6.7866e-03, -1.3766e-06, -4.6375e-03,  3.2139e-03,  9.7950e-03,
        -6.5392e-03,  1.6170e-02, -1.4803e-02,  1.6541e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-1.0655e-01,  6.2956e-01,  1.7686e-02,  1.9357e-02,  1.5826e-02,
         5.0253e-03,  2.8321e-02, -3.0819e-03, -1.4400e-02, -1.9610e-03,
         7.1732e-03,  1.1459e-03, -4.9822e-03,  8.9965e-03, -1.8515e-02,
         5.0946e-03,  9.4453e-03, -4.8467e-03,  4.4684e-04,  5.4502e-03,
        -1.9608e-03,  2.9645e-03, -2.7039e-03,  1.0252e-02, -2.0237e-02,
         1.4213e-02,  3.9797e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0098, -0.8268, -0.0031,  0.0486, -0.0164, -0.0262,  0.0087, -0.0024,
        -0.0108, -0.0034, -0.0023, -0.0082,  0.0279,  0.0055,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9475e-05,  7.6007e-01,  3.4445e-02, -1.7746e-02, -9.5578e-04,
         4.8842e-02,  8.9361e-03,  7.1425e-03,  1.6099e-02,  2.6463e-03,
        -8.5770e-03, -3.8970e-03,  4.5188e-03,  1.3898e-03, -4.4278e-04,
        -2.8519e-03,  4.9634e-03, -5.5345e-03,  6.1345e-03,  5.1628e-03,
        -9.1949e-03, -3.1006e-03, -2.0742e-03, -8.6297e-03, -3.0075e-02,
        -6.5083e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0175,  0.7835, -0.0120,  0.0170,  0.0158, -0.0236,  0.0064,  0.0153,
         0.0022,  0.0307,  0.0083,  0.0074,  0.0221, -0.0382,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5407e-02,  8.3718e-01, -9.6767e-04,  8.1818e-04, -7.1508e-03,
        -3.4213e-04, -1.1952e-02, -1.6732e-02, -1.2472e-03,  3.8972e-03,
        -3.3891e-03, -3.7521e-03,  5.5119e-04, -1.3201e-02, -3.0140e-03,
         1.5690e-03, -1.6362e-02, -5.4351e-03,  6.3063e-05, -3.6641e-03,
        -6.3016e-03, -2.4357e-03, -4.8806e-03, -6.0652e-04,  5.7621e-04,
        -3.9020e-03, -2.4199e-04, -1.2478e-02, -7.2424e-03,  5.1278e-03,
        -9.3422e-03,  1.7435e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0074e-02,  7.8597e-01,  4.0846e-02,  5.7944e-03, -7.2083e-03,
         1.5384e-02, -4.7568e-03, -7.4237e-03,  4.1012e-03, -5.9943e-03,
         1.3858e-02, -3.0846e-03, -2.0573e-03,  2.1020e-03, -4.6503e-03,
         1.2586e-03,  2.5303e-03, -1.3719e-02,  5.1364e-03,  4.4745e-03,
         1.3014e-03, -5.0442e-04,  4.8939e-03,  2.3250e-02, -9.6240e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0658,  0.7742,  0.0840,  0.0557, -0.0097,  0.0106,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3711e-02,  7.4713e-01,  1.8820e-02, -1.1951e-02,  1.0285e-02,
        -7.8970e-03, -5.2399e-03, -2.5470e-03, -9.3886e-04, -4.8031e-03,
        -3.1791e-03,  9.4591e-03, -8.7558e-03, -7.1677e-03, -9.0867e-03,
         5.5682e-03,  8.5003e-03,  2.0539e-03, -1.0067e-03, -4.1164e-03,
        -4.3157e-03, -1.0459e-03, -1.3023e-02,  3.0984e-03, -1.0524e-03,
         3.0599e-03, -2.2283e-03, -5.3999e-03, -2.2521e-03,  3.1334e-03,
         3.6054e-03,  1.1689e-03,  7.5478e-03,  6.8923e-04, -2.0216e-03,
        -3.4309e-02, -9.8290e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.6316e-02, -7.4235e-01, -5.5553e-03, -2.3491e-02, -7.2164e-04,
        -3.5285e-03, -7.0724e-03,  3.0791e-03,  6.5481e-04,  2.7159e-03,
        -2.7087e-03, -4.1241e-03,  1.9818e-03,  2.7599e-03, -9.1888e-03,
         2.8958e-03, -3.6068e-04, -5.7197e-03,  1.2273e-04, -1.4533e-03,
         2.1251e-02, -1.6187e-03, -4.4111e-03, -8.2235e-03, -2.7260e-03,
         7.7231e-03,  5.5051e-03, -3.3355e-03,  1.4972e-03,  1.3687e-03,
        -9.8361e-04, -2.9362e-03, -2.0216e-03, -2.7017e-02,  1.2585e-02,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0168,  0.6732, -0.1350,  0.0292,  0.0037,  0.0018,  0.0055, -0.0050,
        -0.0049,  0.0169,  0.0111, -0.0295, -0.0312, -0.0140,  0.0223,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9982e-02, -8.0636e-01, -3.6683e-03, -5.0851e-03, -8.0537e-03,
        -4.5108e-03, -1.9100e-02,  9.2828e-03,  8.2516e-03,  1.8954e-03,
        -1.5309e-04,  3.1096e-03,  2.1383e-03, -2.3242e-03, -7.1728e-05,
        -5.2889e-03, -8.5270e-04, -1.4801e-03,  4.8293e-03,  3.7032e-03,
        -3.1354e-03, -9.0463e-03,  2.2624e-03, -3.3018e-03,  1.8574e-03,
        -4.3433e-03,  6.5565e-05, -4.0768e-03,  4.0779e-03, -3.7780e-04,
        -4.8868e-03,  1.1819e-02, -3.0612e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0801e-02, -6.9005e-01,  2.9536e-02,  9.2703e-03,  4.0046e-03,
        -8.4626e-03,  8.9934e-03,  6.8646e-03, -1.5755e-02,  5.7486e-04,
         8.8166e-03,  8.6778e-04, -2.5513e-03,  2.7747e-03,  3.2585e-03,
         9.3425e-03,  1.5512e-02,  1.2240e-02,  9.8243e-03,  3.1138e-03,
         1.5935e-02,  5.3050e-03,  1.0330e-02, -3.6194e-04, -4.9239e-03,
        -8.4215e-03,  2.1756e-02,  4.7242e-02,  1.3112e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 9.3839e-02,  7.5277e-01, -1.6360e-03, -2.1400e-02, -2.5627e-02,
        -7.3220e-03, -1.3513e-02, -2.6022e-03, -1.0712e-02, -2.7362e-02,
         6.2633e-04, -3.1045e-03, -1.3848e-03, -3.9416e-04, -3.7707e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3523e-02,  7.9784e-01, -2.2138e-02, -3.1675e-02, -5.0101e-03,
         7.1616e-03, -5.4336e-03, -4.0534e-03,  4.7698e-03, -3.7088e-05,
        -5.7656e-03, -2.5454e-03, -5.2903e-03,  3.2425e-03,  5.4063e-03,
        -1.7225e-03, -3.3522e-03,  6.2772e-04,  1.7472e-03, -1.7083e-03,
        -3.0446e-03,  4.0383e-04, -5.9047e-03,  2.0868e-03,  4.0328e-03,
         3.5276e-03,  8.0524e-04,  4.1286e-03, -2.8667e-03,  3.7963e-04,
         3.5324e-04, -3.1603e-03, -3.8086e-03, -3.9890e-04, -5.3293e-03,
        -4.3294e-03, -1.4596e-03,  2.2455e-03, -3.8350e-03, -5.7143e-03,
        -4.9914e-03,  1.9488e-03,  5.9963e-04, -1.1058e-03,  4.3393e-03,
         1.9679e-03,  5.9645e-04,  1.2165e-03,  2.3729e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3311e-02,  7.6828e-01,  7.3108e-03, -5.7249e-03,  1.0820e-02,
        -2.4816e-03,  1.1866e-03,  3.7724e-03,  1.3572e-02, -2.5736e-03,
        -1.1073e-03,  6.6267e-04,  2.5025e-03,  6.7481e-03, -2.0348e-03,
        -3.9273e-03,  3.4768e-03,  5.3212e-03,  4.2727e-03,  2.9019e-04,
        -7.6831e-04,  6.8007e-03, -4.0824e-03, -6.8306e-04, -8.5716e-04,
        -1.6314e-03, -1.1248e-03,  1.0759e-02,  5.3614e-03, -3.3279e-03,
        -1.6936e-03,  2.0670e-03, -3.0003e-03,  1.3460e-02,  1.5011e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.9908e-02, -8.5164e-01, -1.4722e-02, -1.4350e-02, -5.9070e-03,
        -1.3717e-02,  6.6973e-03, -2.8428e-03, -7.7461e-03, -5.9099e-04,
         8.5640e-03,  3.6614e-03,  5.9674e-03, -6.4952e-03, -2.9821e-03,
         1.8932e-05,  7.4654e-04, -2.5544e-03,  8.8569e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8233e-02,  8.4563e-01, -3.9293e-03,  3.5695e-02,  8.2873e-03,
         1.0785e-02,  8.6249e-03,  1.0305e-02,  1.3466e-02,  1.2674e-03,
        -6.8857e-03,  2.4347e-03,  1.8158e-02,  6.1234e-03, -1.7442e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.4653e-02,  7.8264e-01,  2.5096e-02,  2.0475e-03, -2.2333e-02,
        -2.1062e-02, -1.6746e-02, -2.1915e-03, -8.8805e-03, -3.4963e-03,
         1.2295e-03,  1.1393e-02, -8.0176e-03, -4.6044e-03,  7.4779e-03,
        -1.0533e-03, -6.0334e-03, -5.8436e-03, -6.1730e-03,  2.3713e-03,
        -6.1907e-03, -4.6771e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0194e-02,  8.3332e-01,  6.9629e-03, -7.7773e-03,  1.0375e-02,
        -6.8335e-03, -8.8352e-03, -1.7641e-03,  1.4739e-02,  3.4793e-03,
         1.2275e-03,  3.4637e-04, -4.1431e-03, -9.1660e-04, -5.6657e-03,
         4.2772e-03, -2.4785e-03, -4.9041e-04, -4.2049e-03,  7.3045e-03,
         4.5203e-03,  8.7067e-03, -1.0383e-02, -7.2840e-03,  1.4765e-02,
         3.7308e-03, -2.9436e-03, -4.7332e-03, -1.4970e-03,  6.1008e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0168,  0.7493,  0.0168, -0.0345,  0.0775, -0.0296, -0.0082,  0.0233,
         0.0073, -0.0163,  0.0204,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6655e-02, -8.2248e-01,  6.1069e-03,  2.5791e-02, -1.1231e-02,
         1.0744e-02,  6.8161e-03, -4.9635e-03,  1.3634e-03, -3.6547e-03,
         6.5679e-03, -9.8429e-03, -2.3108e-03, -6.8306e-03, -1.3649e-03,
         1.8274e-03, -6.5600e-03,  9.9142e-04, -1.8344e-03, -3.8510e-04,
         4.1019e-02,  6.5821e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2977e-02, -8.0449e-01, -3.2034e-02, -2.0020e-03, -3.6466e-03,
        -1.4632e-02, -2.7580e-03,  3.2005e-03,  5.1292e-03, -2.0998e-03,
         8.8563e-03, -8.0861e-03,  1.0973e-02,  3.9712e-03, -2.8049e-03,
         1.9513e-02, -3.1097e-03,  3.9442e-03, -6.1995e-03,  5.2129e-03,
        -4.4303e-06,  1.4837e-03, -2.6669e-03,  1.3446e-04,  1.2222e-03,
        -5.4777e-04, -1.1652e-03, -7.8596e-03, -6.5616e-03,  2.5807e-03,
        -8.5207e-03, -1.2729e-03, -3.3869e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5249e-02, -7.7493e-01, -5.3354e-02,  1.1922e-02,  1.0207e-02,
        -7.9694e-04,  5.6262e-03,  8.3642e-03,  3.3130e-03,  4.8850e-03,
        -1.9707e-03, -1.1241e-03,  7.9237e-03,  7.5797e-03,  1.7446e-02,
         4.2989e-04,  7.4654e-05,  7.7041e-04,  4.5197e-03,  1.5553e-03,
         1.2463e-03,  8.8098e-03, -7.4581e-03,  2.9489e-03, -1.1856e-02,
        -8.3705e-03,  1.5524e-02,  1.1744e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2781e-02, -7.4287e-01, -7.0929e-02, -8.1667e-03,  9.0819e-03,
        -2.2394e-03, -2.2905e-02,  3.8948e-04,  1.7158e-02,  7.6676e-03,
         5.3833e-03, -7.1756e-03,  1.7496e-02,  1.0231e-03, -4.2903e-03,
         3.4397e-03,  6.1013e-03,  6.4639e-03,  2.5922e-03, -3.5769e-03,
         2.5725e-03,  1.1585e-03, -4.4376e-03,  1.0070e-02,  3.0035e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 2.6601e-02, -7.5827e-01, -8.5196e-03,  3.9317e-02, -9.6935e-03,
        -1.7142e-02, -7.2498e-03, -5.0074e-03, -1.2067e-03, -2.8455e-03,
        -4.7737e-03, -1.1172e-03,  1.1535e-03, -4.2518e-04, -2.9609e-03,
         2.4287e-02,  2.5927e-03, -3.8985e-03, -2.1020e-03, -3.6774e-03,
        -5.2096e-05, -2.2577e-03,  5.0988e-03, -3.0672e-03, -5.4776e-03,
        -6.7490e-04, -1.8492e-03, -1.9512e-03, -1.0658e-03,  1.8265e-03,
        -1.0950e-03,  3.1113e-03,  6.7650e-04, -8.9851e-05, -5.0550e-03,
         3.2317e-03,  4.1546e-03,  5.5740e-03, -1.1184e-03, -2.1337e-02,
        -8.3938e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.8366e-02,  7.9203e-01, -1.3636e-02, -2.0297e-02,  3.0913e-03,
         9.5491e-03, -6.5914e-03,  5.0595e-03, -1.9017e-03, -9.1415e-04,
        -2.4383e-03,  8.2677e-04,  7.4538e-03, -3.9246e-03,  1.1988e-02,
        -3.0323e-03, -2.2109e-03, -1.0361e-02, -2.1614e-03, -5.1644e-03,
        -1.4053e-04, -1.3100e-03,  9.8109e-04,  1.0965e-02,  5.6114e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1466e-02,  7.8936e-01,  3.2757e-02, -2.0527e-02,  1.0024e-02,
        -8.4363e-03, -2.3434e-02, -1.4138e-03, -9.9828e-03,  5.0687e-03,
        -7.2753e-03, -3.2719e-03, -2.8703e-04, -6.1248e-03, -1.1631e-02,
        -4.3134e-03,  1.1293e-02, -2.0887e-02, -1.5737e-02, -6.3240e-04,
        -6.0744e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.0690e-02,  7.9391e-01, -2.5703e-02, -1.6163e-02, -1.5547e-02,
        -1.0241e-02,  8.0121e-03,  1.0076e-02,  1.1015e-02, -2.1349e-03,
         2.9552e-03, -5.6438e-03, -7.1743e-04, -1.3812e-02, -9.6507e-03,
        -9.3431e-03,  1.3681e-02,  7.0620e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9329e-03,  7.4784e-01, -3.0823e-03, -1.3512e-02,  6.8196e-03,
         1.0765e-02,  1.4132e-02,  2.2712e-02,  8.2514e-03,  2.7616e-03,
        -4.0839e-03, -8.5906e-03,  1.6756e-03, -1.3785e-03,  1.2334e-03,
         3.1767e-04,  3.9327e-03,  8.5356e-03,  7.6234e-03, -2.8260e-03,
         2.4795e-03, -7.0586e-03,  4.5512e-03, -2.7812e-03,  4.2766e-03,
         7.4922e-04,  6.4884e-03, -2.8219e-03,  3.1827e-03, -4.4059e-04,
         6.2836e-03, -1.5435e-03, -6.2681e-04, -2.4651e-03, -2.6619e-03,
         5.6029e-04,  1.9431e-02,  5.5593e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0248, -0.7477,  0.0097,  0.0061, -0.0210, -0.0081, -0.0085, -0.0326,
        -0.0138, -0.0041,  0.0266,  0.0078,  0.0054,  0.0064, -0.0164,  0.0081,
        -0.0046,  0.0264,  0.0097, -0.0122,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0827, -0.7799, -0.0181, -0.0125, -0.0123, -0.0064, -0.0040,  0.0063,
         0.0034,  0.0038, -0.0343,  0.0054, -0.0185, -0.0125,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4080e-02,  7.3421e-01, -2.1352e-02, -1.2992e-02,  1.2271e-02,
        -1.5144e-03, -9.7941e-03, -2.1541e-03,  7.4129e-03, -7.5045e-03,
        -9.7655e-03, -1.1011e-03,  1.6814e-04, -1.2987e-03,  5.1973e-03,
         4.9695e-03,  2.3334e-02,  7.5463e-03,  6.6277e-03,  8.9331e-03,
         4.5335e-03,  1.4029e-02, -4.4029e-03, -1.3616e-02,  4.0569e-03,
        -1.6518e-02, -9.0859e-04,  4.9522e-03,  3.6209e-04,  1.7117e-02,
         1.1589e-02, -5.6830e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4338e-03, -7.6970e-01, -5.5052e-02, -4.9280e-02, -2.8522e-02,
        -6.5009e-03, -8.5141e-03, -4.7825e-03, -5.7625e-03,  6.8065e-03,
         2.3746e-04,  4.8622e-03, -3.4614e-03, -8.5075e-03,  3.5839e-03,
         2.6935e-03, -3.6532e-04,  8.9337e-03,  1.0250e-02,  1.9749e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3169e-02,  6.8479e-01, -1.7276e-02, -3.1777e-02, -2.5206e-04,
         1.1071e-02, -6.0263e-04, -1.1565e-02, -1.6776e-02, -1.9639e-03,
        -9.7965e-03,  4.0507e-03, -8.2418e-04,  4.5797e-03,  1.0470e-03,
         3.0931e-04, -1.3304e-03, -5.2866e-03, -5.0603e-03,  7.0992e-03,
        -3.2301e-03,  2.1266e-02, -3.6249e-03, -5.1617e-03,  2.7231e-03,
        -9.3041e-03, -4.1060e-03, -8.2919e-03, -5.2592e-03, -6.9975e-03,
        -1.3190e-04, -1.4031e-03,  1.6225e-02,  7.3807e-03,  6.8287e-03,
        -8.4845e-03,  3.8320e-06,  2.1920e-03,  2.7933e-03, -7.5268e-04,
         8.6741e-04,  5.7490e-03,  1.2105e-02,  2.6356e-03,  3.6403e-03,
         1.9685e-02, -1.0533e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.4910e-03,  8.1990e-01,  1.5273e-02, -1.9124e-03,  4.1525e-02,
        -1.1643e-02,  9.9080e-03, -8.9272e-04, -1.4634e-02,  1.1422e-02,
        -1.8025e-02, -1.2009e-04,  4.2573e-03,  8.1382e-03, -3.2863e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0554e-02, -7.0907e-01, -1.5806e-02, -7.3186e-03, -2.3821e-02,
         1.4797e-02,  1.0195e-02, -3.2328e-03,  5.6600e-03, -7.3543e-03,
         2.7297e-03, -1.7092e-03,  1.8633e-03, -1.2054e-02,  1.8503e-03,
        -4.2043e-04, -1.2105e-02, -5.6869e-03, -4.6694e-03,  1.4638e-02,
         2.5277e-03,  9.7106e-03,  3.5019e-03, -1.8205e-03,  8.7093e-05,
        -8.4993e-04, -7.5578e-03, -3.4924e-02,  1.2579e-03, -5.2516e-03,
        -4.3827e-03,  3.3108e-03, -3.2626e-04, -9.0386e-04, -2.4162e-03,
        -4.7017e-03,  2.3559e-03,  1.9112e-03,  6.4687e-03, -1.0369e-03,
        -2.6472e-03,  5.8758e-04,  2.8157e-03,  1.0142e-03, -2.8712e-04,
         9.5220e-03, -8.8468e-04, -2.1407e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 5.4712e-02,  7.6259e-01,  3.0472e-02,  8.8670e-04,  1.3452e-02,
         8.0332e-04, -5.3351e-03,  2.1488e-02,  1.9858e-03, -1.5342e-03,
        -8.0733e-03,  1.3425e-02, -5.1327e-03, -4.9660e-04, -4.3008e-03,
        -4.0369e-03,  2.0572e-03, -6.4201e-03, -7.6344e-03, -9.9780e-04,
         9.5270e-03,  2.1545e-03,  1.4055e-03,  7.7340e-03,  2.3323e-02,
         1.0025e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9189e-02, -7.4645e-01, -5.8331e-03,  5.1326e-03, -1.3301e-04,
        -1.3622e-02,  3.3566e-02,  4.8406e-03, -2.7242e-02, -9.5254e-03,
        -2.3195e-02, -2.7523e-03, -3.3437e-03, -1.5059e-03,  1.9177e-03,
        -4.6229e-03, -3.8889e-04,  1.1812e-02, -1.3183e-02,  1.9360e-02,
        -2.2387e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1707e-02,  7.2986e-01,  4.0085e-02,  2.9561e-02, -9.2804e-03,
        -1.5245e-02, -1.0700e-02,  1.7179e-03, -7.9433e-03, -4.3691e-03,
        -7.2286e-03,  3.3582e-03, -3.0016e-03, -7.1452e-03, -5.7944e-03,
        -2.4170e-03, -1.1509e-03, -3.7411e-03, -3.6115e-03, -1.4717e-02,
        -7.6187e-03,  6.4359e-04, -8.8044e-04, -3.1311e-04,  3.3044e-03,
         1.8130e-03, -5.0048e-03,  3.3531e-03, -1.2207e-03,  1.5649e-03,
        -6.6446e-04,  6.3727e-04,  5.0293e-03,  2.3020e-03, -1.3198e-03,
        -1.6925e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.3193e-03, -8.0593e-01, -9.4417e-03, -6.4698e-03, -9.1746e-03,
         4.6435e-03,  3.2937e-02, -7.6654e-03, -1.3467e-02, -1.4616e-02,
        -2.8737e-02, -5.9342e-03,  9.1588e-03,  7.0918e-03, -3.8838e-04,
        -8.8319e-03, -3.9774e-03,  2.8343e-04,  1.5317e-02,  9.6151e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0877,  0.7321,  0.0126, -0.0114,  0.0272, -0.0022, -0.0041, -0.0054,
        -0.0022, -0.0135,  0.0108, -0.0048, -0.0090,  0.0028,  0.0065, -0.0008,
        -0.0097,  0.0028, -0.0156, -0.0235, -0.0155,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5977e-02, -8.2375e-01,  3.8045e-03,  1.1553e-02,  1.2702e-02,
        -1.2059e-02, -9.1153e-03,  4.8890e-03, -1.8361e-02, -7.2027e-04,
         3.9361e-03, -2.7306e-02, -3.5832e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2237e-02,  7.2303e-01,  1.3480e-02, -1.5864e-02,  2.5712e-03,
        -5.5871e-03, -3.1002e-02, -4.5418e-04, -9.4521e-03, -1.2683e-02,
         4.4812e-03, -5.9042e-03, -3.0587e-04,  1.4012e-03,  1.0786e-02,
         7.6909e-03, -6.6748e-03, -7.2057e-03, -7.0022e-04,  1.4204e-03,
         5.3241e-03,  1.0066e-02,  5.4096e-05, -5.1182e-03, -1.6237e-03,
         7.1047e-04, -1.7442e-02,  1.2630e-02,  2.7644e-04, -2.2285e-03,
        -3.7324e-03, -9.6115e-05,  3.5750e-03, -6.9005e-03, -4.0780e-03,
        -2.0654e-02,  2.5621e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0879e-02, -8.3272e-01, -1.2687e-02, -1.1623e-02,  7.1772e-03,
        -7.5362e-04, -1.1525e-02, -3.3572e-03,  6.5588e-03,  1.6515e-02,
        -2.1941e-03, -2.4045e-03, -1.3784e-02, -4.3121e-03, -9.0453e-04,
        -9.5457e-03, -5.8294e-03, -4.6459e-03,  6.7014e-03, -9.8222e-03,
        -4.7807e-03, -1.0005e-02,  1.5025e-03,  9.7680e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0162e-01, -6.6530e-01,  3.4968e-03,  1.8692e-02, -2.4494e-02,
         5.4420e-03,  4.0762e-03, -1.0972e-02, -9.3783e-03, -5.8380e-03,
         2.3289e-02,  4.6939e-03, -1.2441e-02, -1.8683e-03,  8.8661e-03,
         2.5165e-02, -9.6415e-03, -8.0141e-03, -2.6718e-03, -3.8238e-03,
        -1.8468e-03,  5.8790e-03,  1.0359e-03, -3.4573e-03,  6.1559e-04,
         5.6109e-03,  3.2699e-04,  6.3329e-03,  9.6819e-04, -3.3875e-03,
         1.6495e-02,  4.2678e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2891e-02, -7.6264e-01, -5.3554e-02,  1.7083e-02, -1.9246e-03,
         2.1010e-02,  2.3189e-03,  1.5038e-02, -1.7742e-02, -3.8286e-03,
        -6.1666e-04,  9.8479e-05,  1.9576e-03, -1.3548e-03, -1.6979e-04,
        -4.9563e-03, -4.8026e-03, -9.9223e-04, -3.0714e-03, -7.4689e-04,
         4.4068e-03,  1.7770e-03, -4.9322e-05,  7.3025e-03, -1.8465e-03,
        -1.3125e-02,  4.7736e-03,  3.1742e-03, -1.9391e-03,  1.0235e-02,
         4.5780e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.1571e-02,  7.3632e-01,  4.6007e-03, -2.5735e-02, -5.0796e-04,
        -7.8158e-03,  6.2078e-03, -1.6612e-02, -6.0221e-03, -4.4388e-03,
         3.2976e-02,  1.8272e-05, -8.0440e-03, -8.8033e-03,  4.7240e-03,
         4.9941e-03,  5.6503e-03, -5.1835e-03,  2.7604e-03,  3.5785e-03,
         2.6370e-02,  1.0387e-02,  6.6776e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6571e-02,  7.7908e-01,  1.5533e-02,  4.4573e-03,  2.7607e-02,
         1.4069e-02,  3.3257e-03,  4.6494e-03,  3.1736e-03,  3.3838e-03,
        -3.7225e-04, -2.8653e-02, -1.0359e-02, -8.1913e-03, -2.6627e-03,
        -3.7539e-04, -1.0337e-02, -2.5067e-04, -2.0916e-03,  1.9352e-03,
         8.7564e-04,  4.7892e-03,  3.8817e-03,  5.4497e-03,  4.7455e-03,
         1.0060e-02,  1.3969e-02,  9.1492e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 4.7665e-02, -7.6542e-01,  1.3674e-02, -1.3169e-02, -1.1674e-02,
        -4.2442e-03, -1.5293e-02,  8.0028e-03,  6.6964e-04,  5.8350e-03,
        -4.2015e-03, -1.5836e-03, -7.2585e-03, -1.1227e-03, -5.5482e-03,
        -2.3678e-03, -9.7541e-04, -9.8288e-04,  2.5204e-03,  1.2513e-02,
        -8.1536e-03, -4.5957e-03, -4.1536e-03, -8.3356e-04, -3.1037e-03,
        -2.1483e-03, -1.5172e-02, -1.2609e-02, -2.4510e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0192e-02,  8.6720e-01, -1.0326e-02,  7.6140e-03,  7.0989e-03,
         1.4107e-03, -4.6151e-03, -2.0405e-03,  7.9156e-03,  6.2534e-03,
        -5.4285e-03, -1.2984e-03,  2.1783e-03, -9.3291e-03,  1.0616e-02,
        -1.9414e-03,  7.1745e-04,  1.4812e-02, -2.5542e-03, -1.0823e-02,
        -7.0322e-03, -5.2132e-03,  4.1893e-04,  2.9707e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0254, -0.7028,  0.0224,  0.0546, -0.0200,  0.0092, -0.0011,  0.0129,
        -0.0028,  0.0103,  0.0185, -0.0089,  0.0140,  0.0081,  0.0063,  0.0045,
         0.0064, -0.0009,  0.0129,  0.0045, -0.0025, -0.0051,  0.0029, -0.0022,
        -0.0016,  0.0026,  0.0028,  0.0011, -0.0036, -0.0010,  0.0084,  0.0110,
         0.0033,  0.0052], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2388e-02,  8.3039e-01, -2.5995e-02, -7.5931e-03, -1.4108e-03,
         1.4632e-04,  1.0035e-02, -6.7124e-04,  8.7189e-03, -4.1597e-04,
         1.2708e-03, -8.8377e-03,  1.0180e-02, -7.9994e-03,  2.6510e-03,
        -4.3140e-03, -2.4278e-03,  2.5509e-03, -4.3029e-03,  9.0871e-03,
        -1.0443e-02, -1.8170e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6384e-02, -7.5881e-01,  1.1854e-02,  6.5849e-03, -4.5786e-02,
        -1.5110e-02, -2.3263e-03,  6.8524e-03,  1.3021e-02, -1.3895e-03,
        -3.8928e-02,  2.6882e-02,  1.4183e-03,  2.8890e-04, -3.6371e-03,
        -1.7892e-03,  1.1674e-02,  2.7261e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1841e-02,  7.9727e-01,  2.8157e-02,  5.2531e-04, -1.4188e-02,
         1.6286e-02,  2.3862e-03, -6.1139e-03,  6.0615e-03,  5.6295e-03,
        -4.7418e-04,  4.9229e-03, -1.1555e-02, -3.4928e-03, -5.7896e-03,
         2.8836e-04,  1.0928e-04, -1.9265e-02, -1.3054e-02, -2.6946e-03,
        -1.3556e-02, -1.5340e-02, -1.0051e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1213e-02,  8.6286e-01,  1.4600e-02, -1.1103e-02, -1.8996e-02,
        -2.7294e-03, -4.6505e-03, -1.5394e-02, -1.0438e-03,  1.4901e-03,
         1.1398e-05,  3.5904e-03, -6.4310e-03,  1.2521e-02,  7.6645e-03,
        -8.5469e-03,  7.1500e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.0748e-02,  7.5275e-01,  2.9617e-02,  1.6585e-02,  2.3604e-02,
         3.1630e-03,  5.7085e-03, -1.0108e-02, -5.3287e-03, -7.4259e-03,
         4.7914e-03,  1.1145e-03, -1.2228e-03,  7.6291e-03,  2.7774e-03,
         1.5408e-02, -5.2891e-04,  3.1492e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0304e-02,  7.2149e-01, -4.3901e-02, -2.0815e-02, -1.0081e-02,
         1.8030e-04, -9.5467e-03,  1.4312e-02, -3.9232e-03,  7.7150e-03,
        -6.7208e-03, -2.2883e-03, -7.0337e-03, -7.7847e-03, -8.3914e-04,
        -2.0631e-02, -8.7745e-03,  1.7771e-02,  5.3610e-03, -1.1642e-02,
        -1.2823e-02, -5.7091e-03, -7.1646e-03, -8.1800e-03, -3.6679e-03,
         1.1340e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2737e-02,  7.8595e-01,  1.0761e-02,  1.0723e-02,  1.0348e-02,
        -8.9846e-04, -8.4800e-03,  9.3915e-03,  9.9533e-03, -1.5460e-02,
         6.1536e-03, -4.0237e-03,  2.5482e-03,  6.9681e-03, -6.4822e-03,
         2.3310e-03, -8.9026e-03,  3.1518e-03, -1.8872e-03, -5.5423e-03,
         1.1545e-03,  7.9763e-03,  5.5320e-03, -1.9721e-04,  2.8118e-03,
         4.0114e-03, -2.1641e-03,  2.8575e-03,  6.7436e-03,  1.2806e-02,
         1.1058e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0266,  0.7958,  0.0132, -0.0204, -0.0035,  0.0174,  0.0119,  0.0213,
        -0.0172, -0.0041, -0.0029, -0.0027, -0.0172, -0.0035, -0.0095, -0.0095,
         0.0093, -0.0142,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5662e-02, -8.8770e-01,  2.9681e-02,  7.8211e-03,  1.9797e-03,
         2.6338e-05,  6.7270e-03,  1.0402e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 7.1727e-02,  7.0186e-01, -4.2360e-04, -2.6012e-02, -1.8148e-02,
         1.9157e-02, -1.0177e-02,  3.4072e-04, -1.2981e-02, -4.2104e-03,
        -3.7964e-03, -1.3452e-02,  7.3685e-03, -4.6716e-03, -3.3748e-03,
        -2.5903e-04,  1.2159e-04,  1.7330e-03,  4.3701e-03,  1.5120e-02,
         1.6991e-03,  9.6041e-03, -2.5319e-03,  1.6213e-03,  1.0386e-03,
         1.7299e-03, -2.5133e-03,  3.3298e-03, -6.1192e-03,  2.4345e-03,
         8.5979e-03, -2.0453e-03, -6.1848e-04,  2.7084e-03,  3.6305e-03,
         6.2156e-03,  3.7926e-03,  1.1730e-02,  8.7397e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3401e-02, -7.5311e-01,  2.2813e-02,  1.9580e-03, -2.0593e-02,
        -1.5665e-02, -7.8965e-04,  1.4984e-03, -8.0468e-03,  2.1885e-02,
        -4.5409e-03,  8.1442e-04,  2.8621e-03, -7.4161e-05, -8.4358e-03,
        -1.6232e-03,  3.7979e-03,  6.7338e-04, -2.6901e-03, -3.2738e-03,
        -2.1850e-02, -1.5534e-02, -2.2668e-03,  1.2967e-03, -1.9726e-03,
         5.6086e-03, -2.1212e-04,  2.2712e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0402, -0.7635,  0.0330, -0.0094,  0.0120, -0.0048,  0.0209,  0.0142,
        -0.0150, -0.0119, -0.0523,  0.0166, -0.0064,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.5915e-04,  9.1300e-01, -5.8737e-03,  4.7051e-03,  1.2285e-02,
         1.3828e-03, -1.3672e-02,  9.2670e-03, -3.6477e-03,  5.0361e-04,
        -7.1612e-03, -1.2344e-02,  1.1459e-03, -3.8031e-03, -1.0252e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2474e-02,  6.2138e-01, -7.6150e-02, -1.4305e-02,  1.7503e-02,
        -1.8051e-02, -1.2021e-02, -1.5183e-02,  7.0310e-03, -4.3698e-03,
         4.7244e-03, -7.8470e-04,  1.5019e-02,  5.6755e-03,  1.9035e-02,
        -4.5046e-04,  2.9789e-04, -9.0684e-04, -2.4294e-04, -7.1106e-03,
        -3.0840e-03, -3.7179e-03,  3.0845e-03,  5.4531e-03,  8.2349e-03,
         2.2872e-03, -1.3822e-02, -5.9736e-03,  2.3612e-03,  1.0929e-03,
         2.6911e-03, -4.5930e-03,  4.2516e-03,  1.4623e-02, -3.7743e-03,
         6.7348e-03, -3.1176e-03, -2.6906e-03, -1.5107e-03, -1.9069e-03,
         3.0975e-04, -5.8117e-03, -9.5427e-04,  4.8631e-03,  9.2251e-03,
         1.0475e-02, -2.0640e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4160e-02,  8.0990e-01,  5.7299e-03, -2.4827e-03, -9.6634e-03,
         3.8961e-03, -2.1903e-03, -8.1729e-03, -2.1214e-02, -1.5557e-03,
         4.0450e-03,  5.4676e-03,  2.0766e-03, -5.3196e-03, -1.0370e-03,
        -2.8298e-03,  2.0353e-03, -9.2739e-03,  1.3607e-03, -1.2171e-03,
         8.9580e-04,  1.1663e-03,  7.4622e-03, -1.0545e-03,  4.2231e-03,
         1.2247e-03,  1.2148e-02,  2.9298e-03, -3.5907e-03,  4.2367e-03,
         1.0879e-02,  1.6330e-04, -1.8456e-03, -1.6395e-04, -2.0686e-04,
         2.5704e-05, -5.3382e-03,  3.3450e-05,  2.3113e-03,  9.7275e-03,
         8.4364e-03, -8.3136e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1386e-03,  7.5180e-01,  1.6850e-02, -2.9644e-02,  3.6161e-03,
         2.0400e-03,  1.8211e-02, -1.2681e-03,  1.2162e-03, -2.4268e-02,
         1.7929e-03,  2.3017e-02, -6.7885e-03, -3.8643e-03,  1.4198e-02,
        -3.3747e-03,  3.8885e-02, -6.1229e-03,  1.0454e-03,  1.1549e-03,
        -4.0025e-04, -6.6215e-03, -3.3958e-03, -1.3002e-02, -2.4282e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2198e-02,  8.0783e-01, -1.8462e-02, -9.2852e-03, -8.0802e-03,
        -2.7104e-03,  1.3831e-02,  1.6068e-03, -4.4252e-03, -5.2161e-04,
        -2.4571e-03, -6.7767e-03,  3.3530e-04, -6.0011e-04, -7.8984e-03,
         1.5727e-02,  3.2071e-03,  1.6632e-02,  7.1234e-03, -1.7807e-02,
         9.0346e-03,  2.6343e-02,  7.1126e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9699e-02,  7.7898e-01,  2.2067e-02,  1.8263e-02, -1.2201e-02,
         1.7861e-02,  6.0529e-03, -6.3369e-04, -1.0037e-04,  3.1060e-02,
        -7.7418e-03, -1.3725e-03,  9.7144e-03,  3.0452e-03, -1.4572e-03,
        -1.6594e-03, -1.1605e-02, -5.2517e-03, -1.0008e-03, -4.6835e-03,
        -1.9184e-04, -7.6633e-03,  1.7700e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1005,  0.7537, -0.0066,  0.0107, -0.0038,  0.0103,  0.0078,  0.0199,
         0.0238, -0.0088,  0.0130, -0.0254, -0.0158,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0762,  0.7247,  0.0344, -0.0377, -0.0242, -0.0121, -0.0262, -0.0131,
         0.0037,  0.0108, -0.0055, -0.0016, -0.0276, -0.0024,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.5908e-02, -7.6722e-01, -4.2226e-02,  7.6505e-03, -1.2042e-02,
        -1.5851e-02, -5.0589e-03,  4.5104e-03,  5.8523e-03,  8.5327e-03,
         5.9213e-03,  8.2220e-03,  4.5080e-03, -3.4242e-03,  1.5433e-03,
        -3.3721e-03, -9.8671e-03, -7.6425e-03,  6.4318e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 4.5442e-02,  7.6071e-01, -3.4402e-02, -1.1660e-03, -5.3058e-03,
         1.0674e-02,  2.1776e-03, -1.6268e-03, -4.7360e-03,  6.3695e-03,
         3.3295e-03,  1.3178e-03,  2.1653e-03,  8.2401e-03,  1.1679e-03,
        -3.4180e-02,  3.4342e-03, -5.1106e-04, -3.8826e-05,  1.4567e-02,
        -9.7594e-03, -3.7379e-03,  2.1329e-03,  1.2570e-03, -1.2743e-02,
        -2.8808e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.2197e-02, -7.2488e-01,  1.7298e-02,  2.1048e-02, -6.5313e-03,
        -5.5030e-03, -3.7543e-03,  2.3592e-03, -5.2679e-03,  6.2942e-02,
        -6.1649e-04, -6.7158e-03, -2.1603e-02, -2.8921e-02, -1.3912e-02,
         1.6456e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0600,  0.7356, -0.0037,  0.0197, -0.0212,  0.0305,  0.0360, -0.0019,
        -0.0072,  0.0120,  0.0059, -0.0154,  0.0075,  0.0060, -0.0023, -0.0088,
         0.0263,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0786e-01, -8.0612e-01, -2.0200e-02,  4.5189e-03, -6.8158e-03,
        -4.3880e-04,  3.1179e-04, -2.2966e-02, -3.0770e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0021e-01,  6.5104e-01,  2.9788e-02, -3.9695e-03, -8.7430e-04,
        -1.7611e-02, -1.4100e-02, -2.0087e-02, -1.1714e-02, -1.1346e-02,
         8.8634e-03, -3.6552e-03,  2.4622e-03, -2.3605e-03, -7.1017e-03,
        -1.5669e-03, -4.5188e-03,  1.0880e-03,  3.7416e-03, -7.7109e-04,
         5.4665e-03,  2.3964e-03, -9.6608e-04, -3.9634e-04,  2.6518e-03,
        -4.0184e-03,  1.3025e-02,  1.6395e-03, -2.4445e-03, -5.0565e-03,
        -8.9144e-04,  2.9382e-03, -5.3801e-04, -3.0495e-03,  2.1254e-03,
        -3.5533e-03, -4.1983e-03,  1.9775e-03, -1.1686e-02, -3.0631e-03,
        -2.0726e-03, -1.8312e-03, -1.5651e-02, -2.2683e-03, -1.2718e-03,
         7.9513e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6757e-03,  7.8686e-01,  2.8189e-02, -3.4301e-02,  6.6431e-03,
        -1.0357e-02,  1.8695e-02,  1.2480e-02, -2.3410e-05,  8.9312e-03,
        -3.8361e-03,  4.9055e-04,  7.2579e-03,  5.7238e-04, -3.7658e-03,
         6.9277e-03,  1.2169e-04,  8.8520e-03, -3.2862e-03,  1.2147e-03,
        -7.4227e-03, -6.8860e-03, -3.5566e-03,  2.9628e-04,  9.3196e-04,
        -8.9459e-03,  3.6356e-04, -4.2617e-03, -1.0957e-02,  1.0895e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0074, -0.8885,  0.0323, -0.0031, -0.0110, -0.0401,  0.0176,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.2568e-03, -8.8376e-01, -1.9370e-02,  5.0254e-03,  1.4855e-02,
        -2.9208e-03, -1.3132e-02, -1.6327e-02, -1.3179e-02,  1.9613e-04,
         2.4978e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1097e-02,  8.5838e-01,  1.7020e-02, -4.7302e-03, -1.1919e-02,
         2.1389e-02, -1.0151e-04,  1.8805e-03,  5.9432e-03, -1.0757e-03,
        -5.6561e-03,  1.5289e-02,  1.3974e-02, -1.5443e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0184e-02, -8.0683e-01,  2.4724e-02,  8.8571e-03, -4.0627e-03,
        -3.9297e-03, -1.8492e-03, -6.1044e-03,  8.1005e-03,  8.6096e-03,
        -7.4053e-04,  7.7360e-04,  5.8753e-03,  2.0399e-03,  7.3607e-04,
        -3.8533e-03, -2.5951e-03,  2.9174e-03,  1.7027e-03, -9.2712e-03,
        -7.3671e-03, -5.3528e-03, -2.6293e-03,  6.2873e-04,  5.9829e-03,
         9.2445e-03,  1.0211e-02, -4.5164e-03, -1.9521e-03,  3.2156e-03,
         1.1152e-02, -9.3170e-03,  1.1166e-02, -3.5073e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.9396e-04,  8.3422e-01,  7.3785e-03,  3.4388e-03,  8.5266e-03,
        -4.3462e-03,  3.6424e-03, -4.0524e-03, -8.8127e-03,  3.1659e-03,
         7.2895e-03,  3.6781e-05, -2.0947e-03,  8.2132e-04,  6.2097e-03,
        -5.0308e-03, -4.5026e-02, -1.2739e-02,  2.8449e-03, -4.2390e-04,
        -1.9550e-03,  4.4319e-03, -2.5668e-03,  1.3619e-03, -4.2129e-03,
         1.8340e-03, -1.2772e-02,  1.0076e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2268e-02,  7.4638e-01,  2.1892e-02, -3.7957e-02, -1.2109e-02,
        -3.4268e-03, -4.6526e-03,  3.6724e-03, -1.1928e-03,  1.8122e-03,
        -4.2163e-03,  4.4954e-04, -4.9879e-03,  3.4811e-03,  1.7138e-02,
         4.0389e-03, -1.5597e-02,  2.7433e-02,  1.6493e-02,  1.4257e-03,
         9.3727e-04,  1.5419e-03,  7.3695e-03,  1.9531e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 2.2168e-02,  7.5757e-01,  6.5517e-03,  2.6564e-02,  3.0833e-03,
        -1.5914e-02,  1.6579e-02,  8.2602e-03,  1.2238e-03, -4.5027e-03,
         5.7788e-03,  1.0529e-02,  9.4036e-03,  2.6674e-03,  2.6149e-03,
        -5.5963e-03, -1.3376e-02, -4.2164e-04,  4.2381e-03,  1.5682e-02,
        -1.5058e-03, -1.9389e-03, -1.5911e-03,  5.4960e-03, -3.9814e-04,
         2.6799e-02,  1.0878e-02,  1.3465e-03, -5.3982e-03,  8.3847e-03,
         3.5348e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0168, -0.7984,  0.0619, -0.0295, -0.0133,  0.0381, -0.0023,  0.0015,
        -0.0019,  0.0038,  0.0125,  0.0199,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1320e-02,  8.1328e-01,  1.4338e-02,  4.2814e-03,  1.6550e-02,
        -2.3641e-03, -1.6833e-02, -1.1960e-02, -2.8004e-04, -2.1739e-02,
        -9.2778e-04, -1.6987e-02,  1.7082e-02, -7.0905e-03,  1.4964e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7702e-02,  7.4565e-01,  5.5036e-02, -2.4543e-02, -4.2951e-04,
         5.8627e-03, -1.5322e-03,  1.0555e-02,  1.3189e-02,  4.0383e-03,
        -2.5143e-03,  3.1943e-03, -5.0431e-03, -1.7991e-03,  1.6715e-03,
         2.2171e-03, -1.6987e-03,  1.5352e-03,  4.4255e-04,  8.2747e-03,
         4.1173e-03,  7.3240e-04, -2.9512e-03,  9.5638e-04,  4.5520e-03,
         4.5969e-03, -2.5086e-03,  2.3267e-03, -4.4866e-03, -5.5203e-03,
         3.1658e-03, -4.1121e-03,  2.3043e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0257,  0.7643,  0.0654, -0.0323,  0.0044, -0.0064,  0.0116, -0.0027,
        -0.0126,  0.0101,  0.0014, -0.0076, -0.0107, -0.0067, -0.0070, -0.0311,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0350,  0.7617, -0.0726, -0.0152,  0.0339,  0.0245,  0.0145,  0.0105,
         0.0322,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0492,  0.7166,  0.0259,  0.0760, -0.1323,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3088e-02,  8.3592e-01, -2.9550e-03, -6.4313e-03, -4.8258e-03,
         4.7147e-03, -4.2801e-03,  1.7165e-03,  9.0250e-03,  5.3874e-03,
        -2.5984e-02, -4.8081e-03, -7.8094e-03, -8.5231e-03,  4.3133e-04,
        -4.3452e-03,  3.4934e-03,  9.1202e-04,  5.6829e-03,  1.5523e-04,
        -7.1692e-03,  2.1164e-03,  2.4321e-03, -3.9130e-03,  4.0283e-03,
         2.3842e-04,  9.6090e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0608,  0.8132,  0.0438,  0.0239,  0.0376,  0.0197,  0.0011,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0607e-02,  7.7999e-01,  2.5925e-02,  1.4723e-02, -1.2147e-03,
         5.0135e-03, -2.2963e-03, -6.8634e-03, -5.8652e-03, -1.1497e-02,
         3.8875e-03,  2.5994e-03, -2.8444e-03,  2.3317e-04, -5.6449e-03,
         1.2132e-02, -5.3943e-03,  2.6871e-02,  8.0158e-03, -3.2111e-02,
         1.4260e-03,  4.8481e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5662e-02, -8.3472e-01,  2.5879e-02, -8.5366e-04, -7.6132e-03,
         9.2042e-03,  1.9778e-02, -8.9147e-03,  1.1399e-02,  1.8446e-02,
        -1.1046e-02,  5.1881e-04,  5.5014e-03,  5.2122e-03, -2.1604e-03,
         7.0091e-03, -4.8774e-03,  3.2745e-04,  8.7899e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6790e-02,  7.9413e-01,  4.1389e-02, -3.8863e-04,  8.8106e-03,
         8.7400e-03, -1.8125e-03,  5.0847e-03,  8.2343e-03,  1.7052e-03,
         1.4887e-03,  2.0683e-03, -3.5767e-03,  3.4167e-03, -5.1896e-03,
         1.3285e-03,  7.5314e-03,  3.7373e-03,  3.4697e-03, -1.9876e-03,
         9.6447e-03,  1.2305e-02, -2.0780e-03,  1.2026e-04,  2.6429e-03,
        -3.6377e-03, -2.7000e-03,  3.0244e-03,  1.1604e-02, -5.5311e-03,
         1.5837e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 8.6100e-02, -6.1860e-01, -5.2507e-02,  1.0875e-02, -1.3602e-02,
         1.8821e-03,  8.0300e-03,  1.1086e-03,  8.4337e-04, -7.4562e-03,
        -6.9817e-03, -3.0683e-03,  1.2160e-02, -4.5092e-03, -2.2822e-02,
        -2.0445e-03,  8.2703e-03,  7.5977e-03,  1.7141e-02,  1.9663e-03,
        -2.8397e-03,  6.0133e-04, -3.9560e-03,  8.4803e-03, -1.9849e-02,
        -1.4237e-03,  1.3368e-03,  2.4078e-02,  1.7244e-03, -1.9875e-02,
        -1.4004e-02,  1.4270e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.0172e-03, -8.3223e-01, -6.8572e-05,  1.2514e-02, -2.2111e-02,
         5.5323e-03,  2.5046e-03, -1.4275e-02,  8.9060e-03, -4.6170e-03,
         1.1023e-02, -1.0082e-02,  3.0196e-03, -1.3833e-02,  4.1281e-03,
        -3.0270e-03, -5.1826e-03,  3.7774e-03, -3.7843e-03, -1.0084e-02,
        -9.0580e-03, -1.3220e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.5319e-04, -8.3284e-01, -4.0652e-03,  1.1923e-02, -2.0234e-04,
        -5.8650e-03,  1.1493e-02, -4.1485e-03,  7.3232e-03,  4.6370e-03,
        -8.8552e-03, -3.3550e-03, -1.6754e-03, -4.0893e-03, -7.1363e-03,
         1.2388e-03, -2.7228e-03, -5.1817e-03,  4.0912e-04, -9.0719e-03,
        -8.7593e-03,  1.0131e-02, -1.1426e-03,  4.5633e-03,  2.5819e-03,
         2.8443e-03, -2.9769e-03,  9.8599e-03, -8.2012e-03,  2.2054e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.6564e-03,  8.5783e-01, -2.6960e-02, -2.1169e-02, -5.2314e-03,
        -2.4314e-03,  5.3143e-04, -2.6832e-03, -1.8869e-03, -6.0196e-04,
        -2.4257e-02,  7.2216e-03, -1.3775e-02,  2.9952e-04, -5.4941e-03,
        -7.6739e-03, -7.7385e-03, -4.5633e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.7228e-02, -7.0788e-01, -9.6446e-02,  6.8253e-03,  2.3392e-03,
         2.1958e-02,  5.9507e-04, -1.3190e-02,  8.1506e-03,  9.7394e-04,
        -3.1382e-03,  5.3072e-04,  3.8076e-03, -6.0380e-03, -4.3742e-03,
         4.7887e-03, -2.8379e-03, -1.7890e-03,  3.0300e-04,  5.4746e-03,
         1.3422e-03, -2.9290e-03,  3.3394e-03,  1.5399e-03, -1.2803e-03,
         3.0573e-04, -4.0498e-03, -8.0901e-04,  2.4604e-03,  1.4149e-03,
        -8.2313e-04, -3.2798e-03, -2.1589e-03, -6.4060e-04, -1.6308e-03,
         7.1317e-03, -6.1965e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0787,  0.5600, -0.1351, -0.0122,  0.0047,  0.0225,  0.0071,  0.0256,
         0.0031,  0.0095, -0.0030, -0.0016,  0.0037,  0.0344,  0.0030,  0.0075,
         0.0133,  0.0207, -0.0026, -0.0032,  0.0041, -0.0009, -0.0302,  0.0133,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.5981e-03,  7.6088e-01, -6.2277e-03,  9.9225e-03,  7.3160e-03,
         1.4727e-03, -2.1636e-02,  5.7393e-03,  9.4754e-03, -1.0522e-03,
         3.5150e-03, -1.1489e-02, -7.8392e-03, -2.3609e-03, -2.5515e-03,
        -1.8708e-03, -1.0275e-02, -1.9297e-03, -5.7087e-04, -5.2765e-03,
        -2.1876e-03,  2.7047e-03, -2.9755e-03,  1.1679e-03,  3.6727e-03,
         2.2275e-02, -4.4360e-03,  3.6884e-03,  3.3535e-03, -2.2596e-03,
         1.2650e-02, -3.3384e-03, -1.5575e-03,  6.6924e-03,  6.6240e-03,
        -2.7896e-04, -2.0982e-03, -2.9542e-03,  1.9289e-04, -3.1796e-04,
        -4.4880e-03, -5.0044e-03, -2.1203e-03,  1.2538e-02, -4.2396e-03,
         7.1883e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0251,  0.7483,  0.0239,  0.0078,  0.0157,  0.0125, -0.0145, -0.0018,
        -0.0023, -0.0105,  0.0096, -0.0155,  0.0073, -0.0096,  0.0076,  0.0059,
        -0.0090, -0.0167,  0.0089, -0.0016,  0.0120, -0.0024, -0.0319,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0016,  0.8113, -0.0342,  0.0190, -0.0261,  0.0074,  0.0032,  0.0016,
        -0.0090, -0.0389, -0.0479,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.1745e-02, -7.8429e-01,  1.0092e-02,  3.0432e-02, -1.6093e-02,
         1.6996e-03,  1.0696e-02, -9.6080e-03, -1.8502e-02, -5.8058e-05,
        -4.0657e-03,  7.3796e-03,  9.5438e-03,  5.7924e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0915, -0.8127, -0.0026, -0.0203,  0.0164,  0.0077,  0.0141,  0.0060,
         0.0040, -0.0112, -0.0053,  0.0083,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1582e-02,  7.4383e-01,  6.8881e-02,  3.1239e-02,  3.7757e-03,
        -3.6674e-03, -1.4591e-02,  3.8833e-05,  3.6692e-03,  2.6450e-03,
        -3.1170e-03,  2.2025e-03,  8.2050e-03,  8.5168e-04, -1.2430e-03,
        -4.1397e-04,  8.9447e-04,  9.7142e-04,  1.9509e-03, -4.8138e-03,
        -1.1478e-03,  9.5268e-04,  2.5249e-03, -3.1793e-03,  4.3327e-03,
        -5.1007e-03,  1.4163e-03, -1.0190e-03,  5.4693e-04,  1.6485e-03,
         4.1286e-04,  4.5140e-03,  4.8661e-03,  3.2655e-03,  8.4307e-03,
         1.7994e-03,  3.9064e-03, -4.4276e-03, -8.0711e-04, -6.3900e-04,
         4.1811e-03,  2.2599e-03, -1.3008e-03, -3.6936e-04, -2.1179e-04,
         3.1170e-03, -2.1406e-04, -1.0586e-03,  9.4780e-04, -1.4775e-03,
         2.7960e-04, -5.5405e-05, -2.5627e-03,  1.1956e-02,  6.4922e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-3.3293e-02,  8.2224e-01, -8.0509e-03, -4.6856e-03, -7.5636e-03,
        -6.5861e-03,  7.2516e-04,  2.1744e-03,  1.9232e-03,  4.4640e-03,
         3.5639e-03, -3.7900e-03,  2.0823e-02, -4.5535e-03, -2.5207e-04,
         1.2570e-03, -8.6325e-04,  3.3511e-03,  1.7677e-04,  1.0346e-02,
        -3.6801e-03, -6.4078e-03, -4.9225e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.5157e-02,  7.8967e-01, -1.1088e-02,  1.6138e-02, -4.0671e-04,
        -1.4133e-02,  2.7141e-04, -4.5316e-03,  1.1368e-02,  2.5581e-02,
        -3.4135e-03,  7.9712e-03, -1.2075e-02, -9.2562e-03, -1.1354e-02,
         2.0476e-03,  4.2327e-03,  2.1309e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.1505e-03, -8.4621e-01,  1.9432e-02,  8.2751e-03, -2.0887e-02,
         1.0703e-02,  6.3555e-03,  2.6193e-03,  1.1392e-02,  1.1510e-02,
         2.5835e-03, -9.2301e-03, -4.7838e-04,  1.3647e-02,  5.9192e-03,
         2.2481e-03,  9.6546e-03,  1.0705e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2899e-02, -7.9726e-01,  3.9966e-02,  2.3571e-02, -1.2930e-02,
         1.3869e-02,  1.5941e-02, -8.4171e-03,  8.9094e-03, -9.8373e-05,
         4.9112e-03, -6.8782e-03, -1.6412e-02,  2.6050e-03, -3.5335e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7425e-02,  7.2149e-01, -1.3473e-02,  2.7250e-02, -3.3944e-03,
         6.2648e-03,  6.1782e-03,  1.3016e-02,  1.7045e-02,  1.5788e-02,
         1.3555e-02,  1.1665e-02,  6.3361e-03, -1.4517e-03, -2.7668e-03,
         4.4115e-03, -1.7403e-03,  7.1474e-03,  3.0402e-03,  6.2018e-03,
        -9.2327e-04,  7.3725e-03,  2.5517e-03, -2.0458e-04, -7.9965e-04,
         5.6137e-03,  1.7782e-02,  3.8635e-07,  8.2765e-03,  2.1474e-02,
        -4.2068e-03, -1.1560e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3062e-03, -7.8983e-01, -4.3293e-02, -8.1317e-03, -7.4678e-03,
        -2.7379e-03,  1.3419e-02,  3.8135e-04, -8.4002e-03, -4.6848e-03,
         8.6816e-04,  2.0692e-03, -3.8604e-03, -4.7423e-03, -1.6726e-03,
         3.0021e-04,  1.1542e-03,  7.1272e-03,  5.5162e-03,  3.0545e-03,
         1.1159e-02, -4.1389e-03, -7.5082e-03,  2.7238e-03,  7.0876e-04,
         2.8385e-03,  1.9926e-03, -4.8320e-03,  3.1477e-03,  6.0940e-03,
        -1.1832e-02, -1.7759e-02, -1.3253e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1199e-02,  8.1044e-01,  3.3546e-02,  3.1609e-03,  2.7583e-04,
         2.4212e-03, -1.1231e-03, -9.6959e-03,  3.2240e-03,  1.5011e-03,
        -5.5936e-03,  3.7244e-03, -1.1161e-04,  2.5762e-03,  5.4629e-03,
        -5.6039e-04,  3.1389e-03, -8.5399e-04, -2.2399e-03,  4.9429e-04,
        -1.6699e-03, -3.7441e-03, -4.4950e-04, -1.4072e-03,  2.2196e-03,
        -1.0598e-03, -5.9536e-04, -7.3850e-03,  2.9451e-03,  2.7105e-04,
         1.1984e-03,  5.2653e-03,  6.6080e-04,  4.3658e-03, -4.5041e-04,
        -1.1781e-02, -1.8195e-03, -2.8907e-03,  9.8982e-04,  2.2607e-03,
         4.3554e-04,  1.5121e-03,  1.5426e-03,  4.9530e-03, -3.8207e-04,
         2.1234e-03,  6.6181e-04,  1.7719e-03,  3.7931e-03,  1.6130e-03,
        -3.1696e-03,  2.4841e-03,  2.9124e-03,  3.2984e-03,  4.6525e-03,
         1.1225e-04,  8.0228e-03, -1.7838e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6515e-02,  7.2855e-01,  2.0261e-02, -2.3401e-02, -7.6286e-03,
        -3.1087e-03, -2.8566e-03, -8.3728e-03, -5.6610e-03,  1.2090e-02,
        -1.3900e-02,  3.4098e-03, -1.4291e-02,  9.8017e-03,  2.9073e-03,
        -3.5679e-03,  4.9943e-03,  2.0257e-04, -1.0341e-03,  4.3393e-03,
         7.8074e-04, -1.2598e-03, -9.5891e-03, -1.0952e-02, -2.5405e-02,
        -3.5125e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4119e-02,  7.6431e-01, -1.9574e-02, -1.8398e-02, -1.6357e-03,
         3.4012e-02, -7.7418e-03, -5.2619e-03, -1.1100e-02, -5.2548e-03,
         1.1127e-02,  2.8485e-03,  3.7051e-03, -2.7121e-03,  6.4715e-04,
        -2.1726e-03, -6.5924e-03, -2.8519e-03,  5.0359e-03,  4.8563e-03,
        -3.5731e-03, -8.7990e-03, -3.1605e-03, -5.3585e-03, -4.5150e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2846e-02,  7.5552e-01, -2.6732e-02,  1.7365e-02,  1.6962e-02,
         4.8960e-04, -5.0053e-03,  6.0983e-04, -8.1448e-03,  9.2631e-03,
        -6.6325e-03, -5.0116e-03,  1.2887e-02,  2.9316e-03,  5.9237e-05,
        -2.2952e-03,  3.4919e-03,  5.7864e-03,  3.4254e-04, -2.7106e-03,
        -2.1249e-03,  3.0483e-04,  3.3028e-04, -3.9107e-03,  2.1253e-03,
        -2.8210e-03, -1.0047e-02,  3.5422e-04,  2.3805e-03,  1.5895e-04,
         1.3280e-03, -2.6896e-03,  1.0534e-02,  3.1498e-03,  1.3966e-02,
        -2.7173e-03,  2.3165e-03,  3.9667e-04, -1.5969e-03,  2.7604e-03,
        -5.3165e-03,  2.9560e-03,  6.6211e-03,  9.1414e-03, -4.8640e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.1862e-02,  7.2594e-01, -6.1827e-02,  2.2377e-02, -1.3646e-02,
        -1.9600e-02, -1.3866e-02,  1.9070e-03, -1.6091e-02,  5.8880e-03,
        -7.6570e-03, -8.2962e-03, -5.1701e-03,  8.1047e-04, -1.3945e-03,
        -2.5829e-04, -7.8641e-03, -5.0050e-03,  2.2136e-03,  1.4793e-03,
        -2.3825e-03, -2.7455e-03,  1.7136e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0789,  0.7587, -0.0421, -0.0182, -0.0113,  0.0008,  0.0107, -0.0122,
         0.0033, -0.0232,  0.0186,  0.0222,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 1.3924e-04, -7.6905e-01, -4.4007e-02,  1.2388e-02, -4.0057e-04,
        -3.5702e-02,  7.6009e-03,  5.2816e-03, -1.4670e-02, -9.0744e-04,
         3.5469e-02,  2.8811e-03,  4.2640e-03,  4.5012e-04, -5.6371e-03,
         1.6446e-03, -2.4885e-03, -6.2588e-03,  2.1808e-03,  8.3805e-03,
         3.0237e-03, -4.1406e-03,  3.3035e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8515e-02, -7.3470e-01,  2.8496e-02,  1.8248e-02,  1.1422e-02,
         1.6231e-02,  1.3679e-02,  1.0800e-02,  1.2879e-02, -1.3228e-03,
         9.2204e-03,  5.5754e-03,  6.0337e-04, -1.2807e-03,  4.7514e-03,
         3.8447e-03,  9.4389e-03,  4.6839e-03, -2.2598e-03,  6.2240e-04,
        -5.4183e-03, -8.2900e-03, -5.2490e-03,  4.4208e-03,  7.5314e-03,
        -4.8926e-03,  9.0736e-03, -2.7453e-04,  5.4051e-03, -5.5462e-04,
         3.9036e-03,  7.2912e-03, -1.9119e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7850e-02,  8.6085e-01,  3.0855e-02, -1.4562e-02,  7.4031e-03,
         1.1872e-03, -8.6326e-03, -2.3763e-03,  5.8759e-03, -3.4884e-03,
         3.9422e-03, -2.0951e-04,  8.7468e-03,  3.3965e-03,  5.3439e-03,
        -1.2772e-02, -7.6476e-04,  9.4563e-04, -7.9637e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0328,  0.8070, -0.0214,  0.0213, -0.0168,  0.0127,  0.0033, -0.0010,
         0.0032,  0.0244,  0.0139,  0.0170,  0.0237, -0.0013,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1498e-02, -8.7257e-01, -1.0216e-02,  3.3904e-03,  2.5018e-02,
        -4.0698e-03,  5.3635e-03,  1.0439e-02, -9.7435e-03,  1.4842e-03,
         2.6780e-03,  7.8932e-03, -7.4303e-04,  3.1540e-03, -1.3245e-03,
         8.5181e-03,  5.6963e-05,  3.7258e-03, -2.0204e-03,  4.6867e-03,
         2.2104e-03,  9.1955e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.8377e-02,  8.2921e-01,  5.6017e-03,  1.3768e-02,  1.0564e-02,
        -5.9742e-03, -8.5904e-04, -1.5421e-02,  2.9092e-03, -1.9497e-03,
        -2.1583e-03, -7.7314e-03, -5.9649e-03,  2.1539e-04, -1.1807e-03,
         5.2227e-03,  7.7886e-03,  3.5684e-03,  2.3237e-03,  4.7394e-03,
         1.8383e-03,  4.3748e-03, -2.4410e-03,  8.0714e-04, -3.5256e-03,
        -5.6943e-05,  2.5150e-03,  4.6413e-03,  4.0362e-03,  2.3544e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6412e-02,  7.2604e-01, -6.3139e-02,  8.1430e-03,  3.7871e-04,
         1.2306e-02, -2.5690e-03,  1.9188e-02, -7.9110e-03,  6.2985e-03,
        -2.5314e-02, -6.2864e-03, -9.8959e-03,  3.1448e-02,  1.1339e-02,
        -6.4896e-03, -4.6845e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.5742e-04,  8.9908e-01, -1.8524e-02,  8.1522e-04, -5.2903e-03,
         1.1833e-02, -3.5067e-03,  1.0711e-02,  4.7058e-03,  2.6698e-02,
        -1.7976e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0335e-02, -7.7703e-01, -4.7711e-02,  9.3928e-03,  1.4978e-02,
         2.9132e-02, -9.0745e-03,  5.8633e-03, -3.1858e-03, -6.5982e-03,
         4.2963e-03, -1.6278e-03, -6.5055e-03, -2.3179e-02, -2.5562e-02,
         5.8237e-04,  2.4945e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8626e-02,  8.2720e-01, -1.1601e-02,  1.1089e-02, -5.4111e-03,
        -7.3950e-03, -1.7897e-03, -1.9248e-03, -2.8955e-03, -1.1744e-02,
        -8.0934e-04,  1.5066e-03, -5.7881e-03, -8.8803e-03, -1.0418e-02,
         6.2262e-04, -1.4761e-03, -5.1198e-03, -2.3687e-02,  2.8934e-03,
        -1.3780e-03, -2.9064e-03,  1.1348e-02, -7.4359e-05,  8.9677e-03,
        -1.4448e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0317e-02,  7.9004e-01, -3.0419e-03, -1.3703e-02, -1.8048e-02,
         1.8956e-02,  2.3641e-03,  1.1343e-02,  1.9405e-02, -4.5831e-03,
         1.1522e-03, -1.3832e-04, -3.3634e-03,  4.1995e-03, -1.8994e-03,
         3.7885e-04, -5.3662e-04, -3.4540e-04, -4.3589e-03,  5.9486e-04,
        -1.7280e-03, -1.1067e-02, -7.7430e-03, -2.8669e-03, -1.0184e-02,
        -2.5890e-03, -5.8227e-03, -4.9655e-03, -1.2129e-02,  1.2132e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4969e-02,  8.3415e-01,  1.8454e-02, -2.9091e-02,  1.7800e-02,
        -1.3555e-02,  3.2997e-03,  6.2785e-04,  3.8928e-04, -1.2676e-02,
        -9.0541e-03, -9.9284e-03,  7.6703e-03,  6.2181e-03, -5.5735e-03,
         5.3743e-04,  5.4164e-03,  3.4213e-03, -7.1724e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 9.5229e-03,  7.9165e-01,  7.1953e-03,  1.3677e-02,  1.0674e-03,
         2.9097e-03,  2.6242e-02, -3.7063e-03,  1.0624e-02,  7.8623e-03,
         9.8683e-03, -6.7634e-03, -1.0651e-03,  5.4455e-03, -6.4349e-03,
         1.1153e-03,  4.6269e-03,  6.5400e-03, -3.4302e-03,  3.6792e-03,
        -5.5713e-04, -5.6468e-03,  5.3746e-03,  1.0185e-02, -1.0191e-02,
        -3.1473e-02,  1.3151e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.4350e-02,  8.5082e-01, -3.8137e-03, -2.2516e-02, -5.0572e-03,
         1.5015e-02, -2.6528e-03,  7.3219e-04, -5.9953e-03, -5.8936e-03,
        -1.3118e-02,  6.6963e-03, -4.1481e-03, -9.1942e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6978e-02, -8.0165e-01, -5.7612e-03, -2.3631e-03, -2.0013e-03,
        -1.5226e-02,  6.8812e-05,  1.1354e-02, -2.5207e-02, -5.0284e-03,
        -5.6372e-03, -7.1530e-03, -6.1452e-03,  2.5488e-03, -1.5446e-02,
         2.2546e-03, -4.8698e-05,  2.9932e-03, -2.6912e-03, -7.6546e-03,
         2.1994e-03,  7.3763e-03,  1.7002e-03,  1.7113e-03, -5.5908e-03,
         4.3215e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0098,  0.7884,  0.0095, -0.0227,  0.0089, -0.0515,  0.0106, -0.0080,
         0.0033,  0.0069, -0.0029,  0.0465,  0.0082, -0.0230,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2473e-02,  8.5533e-01,  2.1965e-02,  4.6427e-03, -5.9822e-03,
        -3.5257e-03, -2.8000e-03, -8.4946e-03,  1.1027e-03, -1.8402e-03,
        -2.1806e-03, -1.0677e-03, -1.2349e-03,  2.3245e-04,  2.2100e-03,
         1.9316e-03,  4.0670e-03, -6.8715e-05,  4.8726e-04,  2.9029e-03,
         1.5990e-03, -4.2929e-04, -1.0035e-03,  1.6055e-03,  3.0819e-03,
        -5.7844e-03, -3.5339e-03,  8.0702e-03, -1.2993e-03, -3.4820e-03,
        -1.5612e-03,  1.4009e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2823e-02, -8.1191e-01,  2.7081e-03, -6.4924e-03, -3.2960e-02,
        -3.5876e-02,  3.7256e-03, -1.7923e-03, -6.6791e-03,  4.5007e-03,
        -6.7025e-04,  2.6769e-03,  5.9439e-03, -3.6102e-03,  3.6293e-03,
        -8.0681e-03,  4.0321e-03,  2.8350e-03, -2.7333e-03, -3.5655e-03,
         3.2585e-03,  3.3443e-03,  8.3279e-03, -2.9679e-03, -4.8663e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0587,  0.7583, -0.0335,  0.0516,  0.0292,  0.0686,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.1268e-02,  8.0463e-01,  4.7722e-02, -5.1406e-03,  1.2103e-02,
        -1.3050e-02,  4.3517e-03, -3.5737e-03, -1.9471e-03, -8.3363e-03,
        -1.2532e-03,  2.1288e-03, -4.6389e-03,  2.3043e-03, -9.2281e-04,
        -5.2373e-04, -3.0574e-03, -1.8533e-03,  4.8923e-04, -3.1026e-03,
        -1.1582e-03, -2.4811e-03, -4.3558e-03,  2.1229e-03,  1.5307e-04,
        -1.5683e-03, -1.9569e-03, -2.6920e-03, -2.0078e-04,  6.7342e-03,
         1.4830e-03, -2.0439e-03,  4.5342e-04, -1.1620e-04, -3.4642e-03,
        -4.5607e-03,  2.0592e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0112, -0.7757, -0.0325, -0.0089, -0.0157, -0.0084, -0.0128, -0.0023,
        -0.0021,  0.0075,  0.0031, -0.0012,  0.0054,  0.0040, -0.0043,  0.0065,
         0.0097, -0.0151, -0.0053, -0.0008,  0.0031, -0.0035,  0.0024,  0.0033,
        -0.0097, -0.0010, -0.0028, -0.0047, -0.0024, -0.0020, -0.0015, -0.0032,
         0.0016, -0.0141,  0.0122,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2399e-02,  7.1682e-01, -1.4292e-01, -8.8494e-03, -5.7939e-05,
         4.2650e-03,  3.8311e-03,  1.0748e-02, -1.6418e-02,  1.8458e-02,
         6.3929e-04, -2.3037e-02, -1.9787e-02, -1.6491e-02, -5.2727e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9039e-02, -7.8932e-01,  8.2908e-03,  2.9242e-02, -8.6592e-04,
        -5.1522e-03, -4.6815e-03, -1.4882e-02, -9.3086e-03,  6.7473e-05,
        -4.6752e-04,  9.2138e-03,  5.7859e-03,  5.8499e-03, -1.4937e-03,
        -2.9238e-04, -3.0143e-03,  7.0455e-03, -1.4324e-03,  3.0007e-03,
         5.5686e-03,  2.2860e-03,  3.4030e-03,  8.1538e-03, -5.0267e-03,
         6.1637e-03, -5.7521e-04, -1.4999e-03,  5.5354e-03,  5.9949e-04,
         1.4383e-03,  9.5759e-03, -2.1727e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4598e-02,  8.0629e-01, -1.4539e-02, -3.8928e-03, -5.9507e-03,
         1.8247e-02, -2.3842e-03,  5.6086e-03, -1.4201e-02,  1.0720e-03,
         7.6370e-03,  4.3306e-03, -4.7068e-03, -1.5445e-02,  4.0113e-05,
        -3.6195e-03, -3.0561e-04,  4.7032e-03,  4.2817e-03, -2.3835e-03,
        -7.3046e-03, -3.1209e-03,  5.3964e-04, -7.6593e-04,  3.4332e-03,
        -2.6300e-03, -1.2981e-02, -8.9877e-03,  6.0021e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 1.8525e-02,  8.5238e-01,  2.9727e-02, -2.8277e-02,  8.3256e-04,
        -1.9734e-02, -5.3487e-03,  3.4902e-03,  3.6985e-03, -8.6441e-03,
        -3.2599e-03, -1.6442e-03,  6.0868e-03, -1.8016e-03, -1.6549e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2319e-02,  7.9946e-01, -1.5670e-02, -2.3358e-02, -1.2915e-03,
         4.0191e-03, -9.5187e-03, -3.1241e-03,  4.5127e-03,  1.1192e-03,
        -1.9845e-03, -5.8489e-03, -2.2121e-04, -5.1342e-03,  2.3021e-04,
        -3.9082e-03, -6.2679e-03, -1.2321e-03, -7.5183e-04, -5.3389e-03,
        -6.3364e-04, -2.0523e-03, -4.4818e-04,  5.3263e-04,  5.5483e-04,
         4.6203e-03, -5.1613e-04,  5.4961e-03, -1.0056e-03, -1.6134e-03,
         5.3782e-04, -6.4681e-04,  2.5824e-03,  2.0065e-03, -6.0320e-03,
         3.6849e-03, -2.7181e-03,  5.7136e-04,  7.5486e-04, -5.9254e-03,
        -7.5385e-04, -1.3876e-03,  7.5959e-04, -2.1563e-03,  1.7079e-03,
        -4.1397e-03,  1.8212e-03, -8.3762e-03, -2.0654e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8712e-02,  8.2449e-01, -1.6402e-02, -1.3959e-02, -8.2429e-03,
        -4.6452e-03, -7.6595e-04, -2.1791e-04,  2.7457e-03, -1.6914e-04,
        -3.4487e-04, -4.3132e-03, -7.2644e-03, -1.4083e-03, -4.4207e-04,
        -1.5235e-03, -3.5191e-03,  2.1155e-03,  3.4280e-03, -1.0213e-02,
        -7.0007e-03, -2.9604e-03, -4.4678e-03, -1.9445e-03, -3.3833e-03,
         5.0719e-04, -2.3411e-03,  3.2870e-03,  5.4194e-03, -1.0673e-03,
         1.9519e-03,  7.3495e-05, -4.2650e-03,  1.0985e-02,  1.5428e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.1512e-02, -7.9818e-01, -2.2623e-02, -3.3360e-03, -1.6783e-02,
        -8.3507e-03,  3.6070e-03, -6.0149e-03, -9.0702e-03, -7.7473e-03,
        -4.8348e-03, -9.8674e-03, -5.0962e-04, -2.4168e-03, -2.9322e-03,
        -1.0157e-04, -6.1616e-03, -1.6301e-02, -9.6465e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9978e-02, -8.2304e-01,  4.2205e-02, -2.0958e-02, -7.9540e-03,
        -3.0281e-03, -2.3019e-02, -9.5121e-04, -3.1921e-03,  6.0937e-04,
         1.1146e-02, -6.0745e-04,  1.0507e-02,  2.1255e-02, -1.1552e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0358,  0.7569,  0.0388, -0.0079,  0.0113, -0.0117, -0.0137,  0.0054,
         0.0108,  0.0029,  0.0035, -0.0031, -0.0088,  0.0049,  0.0096, -0.0048,
        -0.0013,  0.0008, -0.0018,  0.0181, -0.0098, -0.0383,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0164,  0.8118, -0.0101, -0.0150,  0.0019, -0.0015, -0.0105, -0.0025,
        -0.0165, -0.0023, -0.0163, -0.0062,  0.0033,  0.0009, -0.0042,  0.0008,
        -0.0033, -0.0020, -0.0032, -0.0041,  0.0083, -0.0012,  0.0010,  0.0016,
         0.0024,  0.0021,  0.0057, -0.0069,  0.0066, -0.0313,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8634e-02,  7.9162e-01, -5.0935e-02,  9.4947e-03,  3.4723e-02,
         3.6068e-04, -2.1402e-03, -5.9080e-03,  1.8745e-02, -1.2426e-02,
        -2.5015e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.2915e-02, -8.6741e-01, -1.1449e-02, -6.0676e-03,  1.0342e-02,
         2.3529e-03, -1.6351e-03,  2.6408e-03,  2.3983e-03, -7.0313e-03,
         7.1361e-03,  2.9530e-04,  1.2330e-03, -3.0593e-04,  3.2594e-03,
        -1.3622e-04, -4.5404e-04,  3.8733e-03,  3.0129e-03,  1.1097e-02,
        -2.3733e-04, -4.7226e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1228e-02,  7.5117e-01,  5.2191e-02,  1.3581e-02,  7.3850e-03,
         3.8027e-03,  1.1972e-02,  1.9242e-03, -7.0505e-03, -3.0287e-03,
        -1.6548e-03, -7.0517e-03,  1.9693e-03, -1.3633e-03, -4.7927e-03,
         8.4184e-04, -2.7141e-03,  8.0192e-03,  9.1620e-03,  1.4390e-03,
        -1.5919e-03, -2.8891e-03, -8.3295e-03, -4.9841e-03, -6.9182e-03,
        -8.4376e-03,  6.6832e-03,  6.9614e-03,  4.7282e-04, -6.2888e-03,
         2.7815e-03,  1.0152e-02,  1.1165e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7424e-02, -7.1194e-01, -2.9012e-02,  4.8760e-03, -1.1341e-02,
        -8.6660e-03, -1.0694e-02, -4.0548e-03, -7.1863e-03,  5.9491e-03,
         2.3356e-03,  4.9431e-03,  2.0373e-03,  2.4231e-03,  1.0803e-02,
        -1.5731e-02, -8.9136e-03,  7.7905e-04,  2.7644e-03, -4.3498e-04,
        -7.2609e-03,  3.2217e-03,  5.0465e-03,  5.1125e-03,  2.0523e-02,
        -1.4010e-02, -3.9754e-02,  2.2768e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.7071e-03, -8.2047e-01, -4.5301e-02,  4.7691e-03,  1.2907e-02,
        -4.8587e-03, -2.5229e-03, -2.8645e-03,  2.1667e-02, -1.8181e-03,
         1.8046e-02, -2.1324e-03,  4.7091e-03, -2.9791e-03,  3.0253e-03,
         3.0291e-03,  4.7955e-03,  1.2888e-03,  1.5754e-04, -1.1875e-03,
         8.9958e-03, -1.0941e-03,  1.3144e-02,  6.4979e-03,  4.0260e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-1.3766e-02,  7.6075e-01,  4.8734e-03, -1.6095e-02,  5.3040e-03,
         2.9717e-02, -4.4142e-04, -8.6558e-03, -8.8905e-04, -1.5403e-03,
        -5.6564e-03, -3.5150e-03, -9.5342e-03, -1.2821e-02, -2.2449e-03,
        -2.8481e-02, -3.5994e-03, -4.7833e-03,  2.3142e-03, -7.8295e-03,
         1.8329e-03,  1.5604e-04, -1.9814e-03,  4.9919e-03, -4.3021e-03,
        -8.2989e-03,  8.0688e-03, -1.7651e-03,  4.3434e-04, -2.0375e-03,
         5.0398e-03, -2.5952e-03, -7.6359e-03,  3.0783e-03, -3.8497e-03,
         2.6437e-03, -6.5489e-04, -1.0862e-03, -8.8698e-03, -3.4786e-03,
         4.3852e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8555e-02,  7.5580e-01,  7.3575e-02, -5.5513e-03,  4.2891e-02,
         1.3962e-03,  3.4511e-04,  1.1675e-02, -5.2771e-03, -2.2059e-04,
         5.4713e-03, -1.8353e-02,  5.2068e-03,  7.6407e-04,  2.8122e-03,
        -1.0784e-02, -1.8992e-03, -3.8865e-03,  7.0149e-04, -5.0058e-03,
        -2.9359e-03, -1.0007e-02, -1.7769e-03,  9.3674e-03, -5.7439e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.6959e-02, -8.0810e-01, -3.6698e-02,  1.7540e-02,  8.8799e-04,
         1.0018e-03,  9.5364e-03, -1.3233e-02,  8.8763e-03,  1.3688e-02,
        -2.6578e-03, -4.0141e-04,  6.1275e-03,  1.0530e-03,  4.3340e-03,
        -2.0080e-03,  3.1348e-04, -2.7821e-03,  8.5127e-03,  3.1060e-03,
         1.2180e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0639,  0.7595,  0.0452, -0.0154, -0.0202, -0.0139, -0.0158, -0.0097,
        -0.0026,  0.0057, -0.0018, -0.0095, -0.0016, -0.0087, -0.0099, -0.0016,
        -0.0123,  0.0026,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2830e-03,  7.8948e-01,  1.4588e-02,  2.5880e-02,  1.1981e-02,
         1.5449e-02, -9.2734e-03,  8.0502e-05, -3.3431e-03,  3.2256e-03,
         3.8544e-03, -1.0653e-02, -6.2885e-03, -3.1759e-03,  2.4896e-03,
        -2.4359e-03, -1.4009e-02, -5.6866e-03,  2.3828e-03,  3.2136e-03,
         6.4867e-04,  2.8697e-03,  1.8121e-03, -7.8181e-03,  3.7726e-03,
         1.9616e-03, -6.4013e-03,  4.3111e-03,  1.8142e-03, -2.3697e-04,
         1.7336e-03, -1.1029e-03, -3.8582e-03,  1.3297e-04, -1.0786e-02,
         4.6519e-03, -2.2788e-04,  1.6087e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.5024e-03,  8.1307e-01, -6.4212e-02, -1.6306e-02,  1.2504e-02,
        -2.1633e-04, -2.0312e-03,  3.0054e-03,  2.4617e-03,  1.5011e-02,
         6.1052e-03, -1.2348e-02,  1.4085e-03, -2.0589e-03, -1.8539e-03,
         4.1485e-03, -4.2951e-04, -5.8977e-03,  1.1352e-03, -2.7296e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3330e-02, -8.2722e-01, -3.5156e-02, -1.7982e-03, -1.3635e-02,
        -2.7051e-02, -5.2750e-03, -7.7668e-03,  3.4159e-03, -7.4778e-04,
        -1.4997e-02,  7.4256e-03,  2.7046e-02, -5.1339e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0931e-02,  7.1838e-01, -5.0967e-02, -4.7358e-03,  1.8200e-02,
        -5.2252e-03, -3.7082e-02, -1.0377e-02,  7.6097e-03, -6.2784e-03,
        -3.0159e-03, -2.1485e-02, -5.8259e-03, -9.6583e-03, -5.3828e-03,
         2.1152e-04,  4.1350e-03,  4.6574e-03, -9.1794e-04, -3.5410e-03,
        -8.3062e-03, -3.8743e-03,  6.4289e-04,  7.3183e-03,  6.1444e-04,
        -3.4675e-03,  5.9977e-04,  3.3473e-03, -8.6493e-03,  7.4160e-03,
        -5.7537e-03,  1.1394e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.4795e-02, -7.3591e-01, -2.6084e-02, -3.3512e-02, -2.3182e-02,
         1.3267e-02, -1.8509e-02,  1.2432e-02, -2.7412e-03, -4.9860e-03,
         2.2758e-03,  6.3717e-03, -6.6093e-04, -2.3353e-03, -5.4891e-03,
        -8.1306e-03,  1.6570e-03, -6.7431e-03,  1.5303e-02,  3.5619e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9687e-02,  7.0556e-01, -1.3610e-02, -3.2115e-02, -2.1911e-02,
        -2.6017e-03, -5.7027e-03, -6.9344e-03, -1.5961e-02,  8.0936e-03,
        -5.8302e-03,  3.2649e-04, -4.8518e-03, -2.7696e-04, -1.5208e-03,
         1.3054e-03, -5.1647e-03, -2.1088e-02, -3.4612e-03, -1.6688e-03,
        -7.6568e-03,  2.1543e-03, -4.4244e-03,  1.8315e-03, -2.7224e-03,
        -2.1006e-03, -9.8224e-03, -7.7874e-03, -2.7189e-03, -1.0773e-03,
         6.8409e-03,  7.2437e-04,  9.5150e-03,  2.6626e-03,  1.1352e-03,
         1.0163e-02,  2.4422e-03, -3.1263e-03, -2.9275e-03,  1.2559e-03,
        -1.7648e-03,  3.8145e-03,  1.0739e-02,  1.2445e-03,  4.0440e-03,
         5.4311e-03,  1.2197e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0134,  0.7919,  0.0220, -0.0182,  0.0227,  0.0122,  0.0036,  0.0031,
         0.0248, -0.0162, -0.0221,  0.0080,  0.0021, -0.0043, -0.0355,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.6234e-03,  7.1676e-01,  1.0178e-02, -2.9797e-02,  1.3121e-02,
        -1.7922e-02,  5.3668e-03,  6.4785e-03,  7.2878e-03,  1.1933e-03,
         8.4847e-03, -1.3260e-02,  6.7294e-04, -2.2664e-03, -2.8682e-03,
        -5.0027e-03,  2.0657e-03,  4.1462e-03,  2.6434e-03, -1.1029e-03,
         2.9167e-03, -5.0845e-03, -2.5141e-03,  1.4007e-02, -2.3377e-03,
         8.9987e-04,  8.7743e-03, -1.6771e-03, -5.5723e-03,  5.5563e-03,
         2.1058e-03, -2.0858e-03,  7.2682e-04,  9.8657e-04,  6.9382e-03,
         1.2643e-03,  3.4523e-04, -3.5383e-04, -1.4897e-02, -1.6160e-03,
         4.5400e-03, -6.8836e-04, -3.9865e-03,  1.3296e-04, -3.9046e-03,
        -3.6579e-03,  3.5010e-03,  3.8688e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 6.2872e-03, -8.3234e-01, -3.0285e-03,  4.6497e-03,  5.9532e-03,
         1.9986e-02,  3.4713e-03,  4.4841e-03,  3.9459e-03,  2.2100e-03,
        -3.0351e-03,  1.9738e-02,  8.4972e-03,  2.7941e-03,  1.2460e-02,
        -3.9564e-04,  2.0461e-02,  5.2810e-05,  3.1262e-03,  4.9481e-03,
         1.7654e-03, -1.5061e-03,  2.4768e-04,  2.5585e-03, -1.0838e-02,
         2.1216e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7823e-02, -8.3319e-01,  1.7802e-04, -1.4785e-02, -3.0181e-03,
        -1.7376e-03,  1.8162e-02,  1.9024e-03, -1.0877e-02, -1.4731e-03,
        -2.5638e-02,  8.0046e-03,  5.9765e-03, -3.9620e-03, -1.1203e-03,
        -1.6825e-04,  1.6443e-02,  6.0052e-03, -6.4346e-03, -2.2426e-02,
        -6.7520e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8269e-02,  7.9832e-01, -1.7449e-02,  4.9597e-03,  5.4278e-03,
        -6.0727e-03, -3.8925e-03,  2.7819e-03, -2.6077e-03,  1.9124e-03,
         1.9489e-03,  4.0618e-04, -7.6612e-05, -9.3119e-04, -5.3650e-03,
        -4.5788e-03,  1.4840e-03, -1.0283e-02,  7.9899e-03, -2.2018e-02,
        -5.2341e-03,  3.4152e-04, -1.8026e-04,  6.9330e-03,  8.7454e-03,
         1.0993e-03, -3.2572e-03, -7.3748e-04, -8.6206e-04,  1.2589e-04,
        -2.6093e-03, -1.3596e-03,  2.1569e-03, -3.9473e-03, -3.6180e-03,
         2.2016e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3450e-02,  7.8980e-01, -9.1555e-03, -5.1045e-02, -3.3237e-03,
         5.2439e-03, -9.8109e-03, -1.2493e-02,  6.6960e-03, -4.9662e-04,
         1.7166e-03, -5.0792e-03, -1.0872e-02, -7.7134e-03, -1.0267e-03,
        -1.5238e-03, -1.2262e-03, -1.2264e-02,  2.6993e-02,  2.0068e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3989e-02, -6.9931e-01, -4.4352e-02, -1.0405e-02, -2.1780e-02,
         4.7248e-03, -1.4439e-02,  6.9107e-03,  2.3946e-03, -4.4913e-04,
        -4.1461e-02, -8.4683e-03,  8.0332e-03, -7.2756e-03, -1.5756e-02,
        -3.3607e-03,  5.3589e-03,  9.7572e-03, -6.4778e-03, -2.7800e-02,
         7.4911e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2035e-02,  8.8785e-01,  1.3870e-02,  4.2604e-03, -9.7983e-03,
        -1.9432e-02, -4.6851e-03, -2.5327e-04,  1.0475e-02,  1.1550e-02,
        -2.7764e-03,  5.3371e-03, -7.6778e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.1033e-03, -7.9261e-01,  2.0255e-02,  8.8866e-04, -9.5257e-03,
         7.6358e-03,  1.6871e-02,  5.9841e-04,  2.4897e-02, -2.0512e-04,
         8.9282e-03,  1.1497e-02,  3.3255e-03,  1.9610e-05, -1.6875e-02,
        -4.4871e-03, -1.6055e-03, -9.3795e-03, -2.3641e-03, -2.5945e-03,
         3.6500e-03, -2.1202e-03,  9.6090e-03,  3.8768e-03, -2.0169e-04,
         1.3821e-03,  2.1530e-03, -8.1302e-03, -4.0258e-04, -3.9881e-04,
        -2.5519e-03, -8.4115e-05, -5.8645e-03,  1.8244e-03, -3.9103e-03,
        -1.9428e-04, -1.1982e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1056e-02,  8.3871e-01,  8.5138e-04, -8.3612e-03, -1.1983e-02,
        -4.5137e-03,  3.9563e-03,  6.0794e-03, -7.8615e-03, -1.0676e-02,
        -1.9308e-02,  2.0572e-03, -2.8918e-03,  6.1320e-04, -1.1711e-02,
         2.0686e-03,  4.6832e-03, -4.0836e-03, -4.8282e-03, -5.2144e-03,
         2.7663e-03, -5.5269e-03,  7.0848e-04,  9.4949e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0192,  0.7441,  0.0371,  0.0117, -0.0061,  0.0084, -0.0041,  0.0136,
         0.0035, -0.0230, -0.0175, -0.0031,  0.0079,  0.0023, -0.0053, -0.0096,
        -0.0103,  0.0023,  0.0014,  0.0026,  0.0049, -0.0019, -0.0096, -0.0022,
        -0.0028, -0.0034, -0.0010, -0.0073,  0.0025, -0.0015, -0.0214, -0.0084,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2762e-02, -7.8117e-01,  1.0681e-03,  3.9271e-03,  1.4041e-02,
         1.2796e-02,  4.1024e-03,  1.6129e-02,  2.9441e-03,  1.0972e-02,
         4.2679e-03,  3.2064e-03, -1.2465e-03,  4.2558e-03,  5.4396e-03,
        -6.6553e-03,  1.4756e-03,  6.7865e-04,  4.3943e-03, -2.9953e-03,
         4.0163e-03, -4.0389e-03,  2.4286e-03, -8.8022e-04,  8.8668e-04,
        -5.8849e-03,  1.7987e-02,  6.5177e-04, -6.8056e-03,  2.4453e-02,
        -2.7444e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8913e-02,  7.9415e-01,  1.7024e-02,  6.7347e-03, -8.5698e-03,
        -4.9563e-03,  5.4347e-03, -1.7106e-03,  2.4279e-02, -7.5566e-03,
         1.5478e-02, -7.0328e-03, -1.3607e-02, -5.4570e-03, -5.3350e-04,
        -6.8591e-03, -2.9357e-03, -6.2206e-03, -4.0697e-03, -6.1471e-03,
         8.5677e-03, -2.2532e-03,  1.1505e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7170e-02, -8.0292e-01, -1.3375e-02,  2.5175e-03, -2.6526e-02,
        -1.1673e-02, -3.7843e-03,  8.2630e-03,  3.8459e-03,  7.2252e-03,
         5.5226e-03,  2.1329e-02,  3.5184e-03,  9.1976e-03,  5.2096e-04,
        -3.1069e-05,  1.5985e-03,  2.8401e-03, -4.6522e-04, -3.0205e-03,
         9.3273e-03,  4.9771e-03, -4.8798e-03, -3.7096e-03, -1.6083e-04,
        -2.2349e-03, -1.8044e-02, -1.1325e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 7.7809e-03, -8.3284e-01, -7.6793e-03,  4.3507e-04, -2.5907e-03,
        -1.8580e-02, -3.6418e-03,  1.1070e-02, -8.4510e-04, -3.6102e-03,
        -2.6611e-03, -4.5541e-04,  6.7597e-04,  3.7932e-03,  1.8501e-03,
        -3.3867e-03, -5.0359e-04, -9.7207e-03,  2.0972e-03,  2.4074e-03,
        -1.6129e-02, -4.5996e-03,  7.0996e-03,  1.2307e-03,  4.4366e-03,
         1.8203e-03, -1.5938e-02, -1.8577e-02,  1.3547e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8945e-02, -8.1797e-01,  6.3323e-03, -7.7991e-03, -2.7705e-02,
         1.1342e-02,  7.6958e-03, -3.8582e-03, -4.4215e-03, -2.9284e-03,
        -6.2609e-03,  7.3538e-03,  1.3037e-02, -5.7671e-04, -1.4150e-03,
         5.4941e-03,  2.0937e-03,  1.5100e-03,  4.0190e-03,  2.5142e-03,
         9.8520e-03,  5.9501e-03,  9.6684e-03, -1.2604e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.4843e-03,  7.9452e-01,  2.3337e-03,  1.1790e-02,  1.1465e-02,
         1.2823e-02, -6.1201e-03,  1.3693e-02,  1.2449e-02, -4.6279e-03,
         2.2525e-03,  1.5560e-02,  7.1586e-03, -6.1531e-03, -4.3267e-03,
         2.4744e-03, -1.7950e-03,  4.0864e-03, -5.8080e-03,  5.0759e-04,
         1.4341e-03,  1.4901e-03, -2.9556e-03, -3.3720e-03,  1.8170e-03,
        -4.1074e-03, -4.6308e-03,  3.8116e-05,  2.4464e-03,  1.6282e-03,
        -7.1650e-03, -3.4241e-03,  1.5329e-02, -2.1732e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.3922e-03,  8.2951e-01, -3.9396e-02, -1.4546e-02,  8.0286e-03,
         5.2758e-03,  2.9146e-03,  1.1982e-02, -2.3310e-03, -1.7862e-02,
         8.5733e-04,  1.1064e-03,  1.3808e-03, -1.0729e-02,  7.7469e-04,
         7.2845e-03, -9.4117e-03, -2.3639e-03, -1.0408e-03,  1.7077e-03,
        -1.5252e-02, -6.8508e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0521, -0.7560, -0.0113,  0.0041, -0.0402,  0.0156,  0.0291,  0.0010,
         0.0043, -0.0017, -0.0211,  0.0117,  0.0080, -0.0156,  0.0072, -0.0032,
         0.0096, -0.0082,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0089, -0.8338,  0.0114,  0.0129, -0.0055, -0.0056, -0.0094,  0.0104,
        -0.0060, -0.0018,  0.0048, -0.0074,  0.0010,  0.0074, -0.0071,  0.0037,
         0.0064,  0.0096,  0.0148,  0.0040,  0.0026,  0.0190,  0.0062,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.0320e-02, -8.2311e-01, -3.2203e-02, -2.3001e-02, -2.7964e-03,
        -6.5277e-04,  9.6395e-03,  8.2234e-03,  1.2512e-02,  4.3136e-03,
         1.5664e-04,  1.2103e-02,  6.8081e-03, -4.0476e-03,  2.9142e-03,
         1.4623e-03,  5.7362e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0355, -0.7440, -0.0562, -0.0177, -0.0269,  0.0023, -0.0249,  0.0092,
        -0.0045,  0.0243, -0.0061,  0.0027,  0.0168,  0.0041,  0.0048,  0.0031,
        -0.0104,  0.0065,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3626e-02,  7.7985e-01, -3.4593e-02,  1.2333e-02, -7.1947e-03,
        -6.7614e-03,  4.7726e-03, -2.0170e-03, -4.9796e-03, -1.2798e-02,
        -1.2469e-02, -2.1703e-03, -1.0691e-02, -6.8386e-03, -9.6271e-03,
         1.1058e-02,  3.5483e-03,  1.6833e-03,  6.5180e-03, -1.9874e-03,
        -2.0526e-02, -5.3133e-04,  7.3408e-03, -2.2007e-03, -1.7626e-02,
        -6.2608e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1545e-02,  8.5356e-01,  1.3639e-02, -1.8447e-02, -5.1846e-03,
        -1.1373e-02, -5.9997e-04, -2.8520e-03, -8.9911e-03, -1.0442e-03,
         9.6061e-03, -3.5353e-03, -3.9394e-03,  5.2237e-03, -3.2683e-03,
        -1.8756e-03, -3.8336e-03,  1.1043e-03, -8.1268e-05, -2.3816e-03,
         1.3900e-03, -1.3485e-02,  7.2188e-04,  2.9921e-03,  1.0646e-03,
         4.2749e-03,  3.7362e-04, -2.4114e-03,  5.4369e-04,  9.0157e-03,
         1.6437e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0424e-02, -8.4917e-01, -2.8667e-02,  9.7990e-03, -1.1166e-03,
        -4.8592e-03, -1.1061e-02, -2.3229e-02, -8.1995e-03, -1.5844e-03,
         9.1774e-03,  4.9045e-03,  8.7293e-03,  1.3557e-04,  1.1446e-02,
        -5.3977e-03,  6.9574e-04, -1.1407e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0380,  0.8412, -0.0334, -0.0263,  0.0278,  0.0221,  0.0069,  0.0042,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 6.3855e-02,  7.3310e-01,  8.2852e-03, -1.8720e-02, -1.7922e-02,
         8.4796e-03, -1.1968e-02, -3.5349e-03, -1.2810e-02,  2.4444e-02,
        -2.8442e-03, -4.2564e-03, -6.4130e-04,  5.2984e-04,  2.0712e-03,
         3.2353e-03,  1.9876e-03, -3.5676e-03, -6.9889e-03, -6.3878e-03,
        -4.7956e-03,  8.3692e-04, -6.0545e-03,  2.6769e-03, -9.5282e-03,
        -2.9877e-04, -3.1520e-03,  5.1242e-03,  2.1014e-04,  1.3158e-02,
        -2.5202e-03, -7.9560e-04,  6.0432e-04, -1.5797e-03,  4.1661e-03,
        -1.3288e-03, -6.2848e-03, -6.4013e-05, -1.1921e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0331e-03,  8.0809e-01, -9.6803e-03, -5.0774e-03,  1.2699e-02,
        -1.5862e-03, -1.9101e-02, -2.1127e-03,  5.8267e-03, -1.1504e-02,
         5.4373e-03,  4.6004e-03,  1.4233e-02, -6.0327e-03, -8.2135e-03,
        -6.4325e-04,  7.4437e-03, -1.2505e-03,  1.0133e-03,  8.3168e-03,
         1.7111e-02, -4.9377e-03, -6.0598e-03, -1.9940e-03, -3.6485e-03,
         5.7802e-04, -4.0677e-03, -2.2712e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3462e-02, -7.9483e-01,  4.1361e-02, -2.0509e-02,  6.8893e-03,
        -5.6473e-03,  2.0411e-02,  1.6752e-03, -5.7183e-04,  1.0840e-02,
        -4.7211e-03,  1.5546e-02, -2.3534e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0202, -0.8124,  0.0457,  0.0132, -0.0026,  0.0013, -0.0024, -0.0094,
        -0.0073, -0.0104,  0.0108,  0.0120, -0.0027,  0.0208,  0.0288,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4005e-02,  7.0919e-01,  1.5710e-02,  1.5145e-02,  1.4572e-02,
         8.2326e-03,  8.7564e-03, -1.4709e-02, -1.5569e-02,  6.2662e-03,
        -1.7512e-02,  5.0494e-03,  1.1692e-02, -1.2597e-03,  4.4951e-03,
         1.3090e-03,  9.9994e-04, -6.7097e-03,  5.0553e-03, -9.9363e-04,
        -2.6830e-03, -1.5213e-03, -5.4798e-03, -2.1214e-03,  7.1538e-03,
         3.1773e-03, -1.6434e-03,  4.6135e-03, -6.2141e-03, -5.3047e-04,
        -1.4376e-03,  2.0516e-03,  1.1559e-03, -3.1598e-03,  8.2607e-03,
         3.9031e-03, -2.1264e-03, -9.5015e-04, -2.9337e-03, -1.9581e-03,
        -8.3312e-03, -3.5848e-03,  2.8813e-04,  9.1859e-05,  4.8414e-03,
         4.5504e-03, -8.0037e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2512e-03,  8.3632e-01,  3.6599e-02, -5.3248e-03, -2.1653e-03,
        -8.7308e-03,  1.9284e-03,  2.5918e-03, -6.4635e-03,  1.2606e-03,
        -2.7452e-03, -1.1111e-02, -2.8933e-03,  4.6711e-04,  8.3578e-04,
        -2.5841e-04, -2.7750e-03,  1.2477e-03,  3.4871e-03,  1.0861e-03,
        -2.5806e-04, -5.9616e-03, -1.9582e-04, -3.4524e-03,  2.3214e-03,
         3.7237e-03,  7.8196e-03, -3.1814e-03, -3.7900e-03,  3.0870e-03,
         2.5852e-03,  6.8248e-04,  1.0879e-03,  4.1103e-03,  4.6514e-03,
         8.2541e-04, -5.2170e-03,  5.3393e-04,  2.0617e-03, -3.2228e-03,
         3.6151e-03,  1.0773e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4555e-02,  7.9806e-01,  2.9003e-02, -1.3974e-02, -5.1510e-04,
         2.6534e-03,  6.6726e-03,  1.4415e-03, -4.1428e-03, -6.9043e-03,
         9.4978e-03,  3.0941e-03,  8.5388e-04,  3.4002e-03,  1.8434e-02,
        -9.1380e-04,  1.1937e-02,  4.5340e-03,  3.5272e-03,  1.2772e-03,
        -3.2692e-03, -1.8743e-03,  3.9119e-03,  1.3431e-02, -1.2120e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7097e-02,  8.1122e-01, -1.9337e-03, -7.1319e-03, -3.2810e-03,
        -9.6834e-03, -1.8217e-02, -6.7153e-03, -8.9802e-03, -6.9141e-03,
         1.4950e-03,  9.7262e-03,  4.8573e-04,  1.1679e-02, -9.1580e-03,
        -4.0084e-03,  4.2337e-03, -4.1904e-03, -1.8992e-03,  1.4196e-02,
         1.2964e-02,  9.1543e-03,  5.6332e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0446e-02,  7.9663e-01,  1.3867e-02,  1.4711e-02, -2.8189e-03,
         2.1352e-03,  8.4693e-03, -4.0663e-04,  9.5470e-03,  3.6588e-02,
         1.9657e-02, -9.8103e-03, -4.9503e-03, -5.6379e-04, -1.2259e-03,
        -2.6389e-03,  1.8592e-04,  1.6396e-03, -2.7089e-03,  1.7278e-03,
         6.5450e-03,  8.7557e-04, -3.1847e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1021,  0.7442,  0.0226,  0.0056, -0.0094,  0.0129,  0.0123,  0.0035,
        -0.0110,  0.0080,  0.0045,  0.0227, -0.0411,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0682, -0.7627,  0.0360,  0.0228,  0.0020, -0.0015,  0.0078,  0.0109,
        -0.0085, -0.0125, -0.0111, -0.0392, -0.0152,  0.0016,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.0104e-02, -7.8071e-01, -6.3803e-02,  1.7112e-03, -2.8243e-03,
         5.8628e-03, -5.0770e-03, -1.5519e-03, -3.7082e-03,  1.9632e-02,
         9.5810e-03,  2.0552e-03, -5.4826e-03, -1.7070e-02, -2.7500e-03,
        -2.7146e-04,  1.0617e-02, -1.4709e-02, -2.4795e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 2.6940e-02,  7.6814e-01, -5.6637e-03, -2.1618e-02, -8.8363e-03,
         1.2595e-02, -1.6342e-04,  4.0579e-03,  2.5999e-03, -3.3590e-04,
         2.6627e-04, -6.2667e-03,  1.0476e-03,  2.7228e-03, -8.2193e-04,
        -3.5717e-02,  2.9053e-03, -7.6050e-03,  1.4494e-03,  1.5980e-02,
         1.9464e-03,  7.5045e-03, -1.8275e-03,  5.3260e-03,  5.9188e-03,
        -5.1744e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0056,  0.8513, -0.0164, -0.0156, -0.0117,  0.0049, -0.0039, -0.0105,
         0.0067, -0.0141, -0.0026,  0.0056, -0.0141, -0.0091, -0.0127, -0.0152,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0684,  0.7542, -0.0026,  0.0181, -0.0162,  0.0489,  0.0146, -0.0056,
        -0.0009, -0.0076, -0.0161, -0.0048, -0.0056,  0.0045,  0.0043,  0.0015,
         0.0263,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0650,  0.7417,  0.0007, -0.0305,  0.0429,  0.0253,  0.0164,  0.0433,
        -0.0342,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9435e-03,  7.1046e-01, -2.1698e-02, -8.0108e-03, -2.4367e-02,
         2.4416e-03,  3.8974e-03,  5.1996e-03, -1.3203e-03,  2.4478e-03,
         1.7736e-02,  2.9597e-03,  3.8020e-03,  1.7666e-03,  3.3418e-03,
         5.9562e-03,  8.9592e-03, -4.1949e-03,  1.1224e-02, -3.1630e-03,
        -3.8366e-04, -1.4955e-03,  9.5464e-03, -1.6303e-03,  7.7469e-03,
        -1.3770e-02,  4.9416e-03,  1.2358e-03,  8.0447e-04, -8.2582e-03,
        -3.1044e-03,  5.0705e-03, -1.9479e-03, -6.3848e-03, -1.0973e-03,
        -7.6289e-03,  2.9078e-03,  4.9695e-03, -3.5206e-03,  9.6276e-04,
        -7.3055e-03,  1.5631e-02,  1.7695e-03, -1.0154e-03, -7.7532e-03,
        -3.3225e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9291e-03,  8.2284e-01, -1.7618e-02, -1.7291e-02, -7.7386e-03,
        -2.6633e-03, -3.1026e-04, -3.7096e-03,  2.5843e-03,  9.8069e-03,
        -8.0496e-03, -8.2748e-04,  4.4790e-03, -1.8267e-03, -1.4433e-03,
         3.6215e-03,  2.8824e-03,  3.9924e-04,  7.9510e-03,  2.2070e-02,
         3.2201e-03,  4.2547e-03, -7.7613e-03, -3.9228e-03, -3.3783e-03,
        -1.2840e-02, -8.9006e-03, -6.7228e-03, -1.7511e-03,  7.2115e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.1828e-02,  8.8014e-01,  3.7855e-03, -6.7703e-05,  1.7018e-02,
         3.1191e-03, -2.4040e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0265, -0.7761, -0.0235, -0.0022,  0.0251,  0.0169,  0.0084, -0.0106,
         0.0295,  0.0324,  0.0487,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0455,  0.7457,  0.0321,  0.0412,  0.0087,  0.0178, -0.0179, -0.0010,
         0.0260,  0.0131, -0.0076,  0.0141, -0.0146, -0.0146,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0508e-03,  8.3029e-01,  3.1043e-02,  7.7553e-03, -2.1107e-03,
         4.4288e-03, -6.0781e-03,  3.6135e-04, -1.1370e-02, -3.8071e-03,
        -4.5994e-03,  7.0600e-03, -1.3705e-03,  3.0264e-03, -8.7089e-04,
         9.9129e-04,  2.0210e-03, -3.0856e-03, -2.2226e-03,  2.2016e-03,
         1.0255e-02,  1.2433e-02, -4.0046e-03,  4.7383e-03, -2.8824e-03,
         5.9562e-03,  4.4891e-03,  3.1670e-03, -2.1403e-03, -2.6781e-03,
         9.4609e-03,  3.0799e-03,  3.2493e-03,  5.7167e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2651e-02,  7.7309e-01,  5.3480e-02,  8.8368e-03,  1.3977e-03,
         1.0052e-02,  5.3234e-03, -1.4241e-02, -7.8062e-03,  2.8222e-03,
         1.2558e-03, -2.8138e-03,  3.7905e-03,  8.1355e-03, -2.1050e-02,
        -7.0739e-04,  5.4088e-04,  5.0864e-03, -1.3880e-03,  7.1539e-04,
         1.7799e-03,  4.5126e-05, -1.8330e-02, -5.2090e-04,  3.4740e-03,
         1.2649e-04,  1.9179e-02,  1.1363e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.6617e-03,  8.7802e-01,  1.9810e-02,  3.7903e-03,  5.8679e-03,
        -5.6664e-03, -4.1634e-03,  9.9798e-04, -5.9984e-03,  2.2482e-03,
        -2.1520e-04,  2.2720e-03,  1.0383e-02, -1.0305e-02, -1.2030e-03,
         1.8376e-03,  1.0608e-04, -1.0629e-02, -2.2910e-03,  2.8897e-03,
         4.9147e-03,  1.7822e-03,  3.0543e-03, -1.1896e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 1.3470e-02,  7.4161e-01,  4.6861e-02, -4.2968e-03,  1.3373e-02,
         6.9341e-03, -1.2634e-02, -9.5904e-03,  1.5629e-02, -7.3420e-03,
        -1.3460e-02,  6.1533e-03,  7.1535e-03, -1.2497e-04,  1.9251e-03,
         3.0939e-04, -4.7241e-03, -3.3792e-03, -9.3954e-03,  2.1359e-02,
         2.2646e-03, -4.3728e-03, -6.7629e-03,  4.3612e-03,  1.1348e-02,
        -8.1127e-04,  2.0024e-04,  1.2420e-03,  4.4002e-03,  1.1170e-02,
        -1.3344e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.9839e-02,  8.6667e-01, -9.0916e-04,  1.3570e-02,  3.4050e-03,
         5.3928e-03, -1.0196e-02,  7.0182e-04,  2.0278e-03,  1.1466e-02,
        -6.3811e-03,  9.4400e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.0033e-02, -7.3953e-01, -5.7354e-03,  1.6689e-02, -1.4574e-02,
         3.7861e-03,  2.4659e-03,  3.7607e-03,  2.3577e-04, -3.0983e-03,
        -1.0788e-03,  5.4222e-02, -2.1525e-02,  3.5442e-02, -1.7821e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4799e-02, -8.3646e-01, -1.0215e-02, -8.7380e-04,  2.8934e-03,
        -4.9226e-03, -1.5711e-02, -6.2648e-03, -1.6079e-03, -1.4123e-02,
         1.4540e-02, -3.3213e-03,  9.1553e-03,  3.4718e-03,  2.0579e-04,
         3.3644e-03,  2.4108e-03,  3.8090e-03, -1.6677e-03, -2.5358e-03,
        -1.4707e-03,  1.7275e-03, -2.3266e-03, -6.1013e-04, -7.3355e-03,
         1.4061e-03, -4.8276e-04,  6.9110e-03,  1.6786e-03, -3.3898e-03,
        -6.0036e-03, -1.3961e-03,  2.9104e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0370, -0.8012, -0.0370,  0.0343,  0.0058,  0.0024,  0.0011,  0.0062,
         0.0155,  0.0059,  0.0094, -0.0009,  0.0103,  0.0059, -0.0141, -0.0128,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0096, -0.8844,  0.0202, -0.0168, -0.0050, -0.0050, -0.0221,  0.0121,
        -0.0249,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0638,  0.7958, -0.0795, -0.0037, -0.0571,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2548e-02, -7.9069e-01, -1.5407e-02,  1.0422e-02,  5.3363e-04,
        -2.1855e-03,  5.4022e-03, -9.6569e-03, -2.1860e-03, -1.1098e-02,
        -2.6209e-02, -7.2699e-03, -8.8969e-03,  1.7221e-02, -3.5759e-03,
         8.3127e-04, -1.0898e-02, -4.2646e-03,  2.1455e-03,  1.7106e-03,
         6.3711e-03, -9.3917e-05,  2.7245e-03, -4.4722e-03, -5.2616e-04,
         1.1212e-02, -1.4494e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0044,  0.8427,  0.0086,  0.0075,  0.0098,  0.0776, -0.0494,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1059e-02, -8.6372e-01,  3.3333e-03,  7.0488e-03,  1.1715e-02,
        -3.9520e-03, -1.3650e-02,  9.6922e-03,  2.6683e-03,  9.5032e-03,
        -3.6591e-05, -2.1635e-03,  7.5962e-03,  1.4581e-03, -1.3938e-03,
        -1.7160e-05,  2.4493e-03, -3.4406e-04,  8.7712e-03,  1.7562e-03,
        -2.6615e-03, -5.0129e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6945e-02, -8.6514e-01,  2.3600e-03, -6.2127e-03,  2.4028e-02,
         6.1196e-03,  1.2595e-02, -4.2318e-03,  5.6367e-03, -1.5238e-03,
        -3.2574e-03,  1.0011e-02,  1.4635e-03, -7.9565e-03, -1.3876e-03,
        -1.8108e-03,  2.6354e-04, -5.4910e-04, -8.5115e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6885e-02, -8.2092e-01, -2.8967e-02,  6.5476e-03, -1.7406e-03,
        -8.2597e-04,  1.4206e-02, -5.3671e-03, -4.4008e-03, -3.9714e-03,
        -2.3259e-03, -3.2928e-03, -1.3907e-03, -5.7137e-03, -4.4401e-03,
         1.2602e-03, -1.4540e-03, -3.0083e-03,  2.2158e-03, -6.5323e-03,
        -3.4496e-03,  1.7532e-03,  2.3200e-03,  1.0409e-02, -1.8100e-03,
        -2.5647e-04,  1.9726e-04,  5.3597e-03, -1.6221e-02, -2.0080e-02,
        -2.6748e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 4.9915e-02, -6.8866e-01, -7.2611e-02,  1.9530e-02,  9.5608e-04,
         1.3834e-02,  2.4499e-04,  1.9361e-03,  5.0628e-03, -2.6645e-03,
        -2.3385e-03,  8.5170e-03, -9.4645e-03, -3.1877e-03, -6.0344e-04,
        -6.2264e-03,  2.5036e-03, -1.6507e-02,  4.8387e-03,  2.5941e-03,
         5.1939e-03,  2.1742e-04,  1.3157e-03,  2.8526e-03, -8.4656e-04,
         2.4848e-04,  7.4306e-03,  1.6505e-02,  2.8557e-04,  5.3941e-03,
         6.6201e-03,  4.0893e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0027, -0.8188,  0.0011, -0.0160,  0.0118, -0.0099,  0.0070, -0.0078,
        -0.0235,  0.0065,  0.0176,  0.0042,  0.0092,  0.0091,  0.0014, -0.0138,
         0.0010,  0.0053,  0.0049,  0.0081,  0.0119, -0.0085,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2976e-02,  8.1282e-01,  3.4988e-03, -9.1163e-03, -3.8327e-03,
         2.5490e-03, -1.2347e-02, -1.1392e-02, -1.0789e-02,  2.0647e-03,
         8.9280e-03,  3.6840e-03,  1.3211e-02,  5.6085e-03,  6.3518e-04,
         4.9860e-03, -3.1529e-04,  1.3861e-02, -2.1524e-03,  1.6816e-02,
         9.8832e-03, -5.0043e-03,  1.3315e-03, -1.6072e-03, -1.9065e-03,
        -4.9085e-03, -7.5666e-04,  7.3444e-03,  9.2912e-03,  6.3861e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1458e-02,  8.6810e-01,  2.0382e-02, -1.5540e-02,  2.0266e-03,
        -1.0753e-02,  8.0807e-04, -6.7174e-03, -6.1669e-03, -4.7916e-03,
        -6.4874e-03,  7.1769e-04,  1.3604e-04,  1.4479e-02, -2.1873e-03,
        -3.4427e-03, -3.0187e-03, -2.7860e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5325e-02,  8.2849e-01,  9.3373e-03,  1.2014e-03, -2.1347e-03,
        -1.2396e-03, -1.2232e-02,  6.1081e-03,  3.7220e-03,  2.4102e-03,
        -3.0201e-03, -2.1016e-03, -2.0227e-03,  2.1512e-03, -5.0759e-03,
        -1.1375e-02,  9.0005e-03, -3.4452e-03,  5.2988e-05,  4.2890e-03,
         2.3898e-03, -4.8572e-03, -4.5747e-03,  5.3779e-03,  5.6647e-03,
        -1.1642e-04,  3.3836e-03,  7.2893e-04, -4.7732e-04,  3.0446e-03,
         1.2329e-03, -1.8185e-03, -2.5330e-03,  4.3096e-03,  2.7728e-03,
        -6.8756e-03, -1.5108e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7804e-02,  7.8240e-01, -7.5484e-02,  3.0356e-03,  1.1313e-02,
        -3.6949e-03, -4.8006e-03,  9.5467e-03,  2.6579e-03,  2.0083e-03,
         2.6768e-03,  3.4253e-03, -2.9826e-04, -3.2731e-03,  1.5896e-03,
         4.5960e-03,  8.5626e-03, -3.2872e-03,  7.7396e-04, -1.2975e-03,
         1.5033e-02, -2.0845e-03,  3.5404e-03, -2.6816e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1745e-02,  7.7212e-01, -1.4344e-02, -1.1822e-02,  7.3541e-03,
         3.1586e-03, -8.1953e-03, -3.1394e-03,  3.9317e-03, -9.6674e-03,
        -3.7373e-03, -9.9304e-03, -3.0715e-03, -1.0084e-03, -4.5637e-03,
        -1.9955e-03, -2.9910e-03, -1.1604e-03,  8.6965e-04,  5.6446e-04,
         4.6256e-03, -1.1410e-03,  2.1739e-03,  2.5028e-03,  9.4392e-03,
         6.0907e-03, -9.6771e-03, -2.4301e-03,  4.9341e-03, -3.0477e-03,
         9.5613e-03, -2.3847e-03, -1.2471e-03,  4.3808e-03,  6.4696e-03,
        -3.1853e-03, -6.3863e-03, -2.6410e-03, -3.4655e-04, -3.1436e-03,
        -1.0168e-03, -3.9498e-03,  2.0634e-03,  4.4302e-03,  3.7634e-04,
         6.9837e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7122e-02,  8.1360e-01, -1.4108e-02, -1.0472e-02, -8.6508e-03,
        -4.4527e-03,  4.7208e-03, -2.5135e-03,  1.6058e-04, -1.4570e-02,
         2.8737e-03, -9.0942e-03,  6.1117e-03, -8.5365e-03,  3.3019e-03,
         6.0492e-03,  6.0558e-03, -5.4550e-03, -1.8139e-03,  5.8111e-03,
         1.1743e-02,  5.0043e-03,  2.7776e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.9825e-02,  8.3161e-01, -3.6934e-02, -7.2836e-03,  5.4510e-03,
         1.3330e-02,  2.3666e-05, -1.6982e-02, -8.9315e-03,  9.6310e-04,
         8.6629e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0266,  0.8355,  0.0240,  0.0071, -0.0128, -0.0160,  0.0057,  0.0116,
         0.0181, -0.0037,  0.0120,  0.0085,  0.0154, -0.0031,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0122,  0.8838,  0.0210, -0.0067, -0.0090,  0.0060,  0.0018, -0.0012,
         0.0087,  0.0223,  0.0262, -0.0010,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9000e-02, -7.0763e-01, -5.2623e-02, -1.6782e-02, -1.1764e-03,
         1.0400e-02,  1.4436e-02, -2.2421e-03, -7.1281e-03, -7.1001e-03,
         8.3750e-04, -3.7808e-03, -1.3909e-03, -1.4157e-03, -7.4934e-04,
        -4.2598e-04,  2.2313e-03,  2.6652e-03,  7.5219e-04,  5.7746e-03,
         4.0196e-04, -1.2825e-03, -4.6790e-03,  8.5957e-03, -3.1459e-03,
         1.8658e-02, -2.8144e-02, -4.7279e-03,  4.9096e-04,  1.9389e-03,
        -6.9078e-04, -2.2110e-05, -2.8970e-03,  7.6292e-04, -6.1818e-04,
        -8.4501e-04, -2.5647e-04, -1.6135e-03,  4.3661e-03,  2.0038e-03,
        -2.0764e-03, -8.2739e-04,  1.1338e-03, -8.1578e-04,  2.6724e-04,
         4.1415e-03, -2.8225e-03,  8.6701e-04, -3.6462e-04, -4.7282e-04,
        -2.0169e-03,  2.4835e-03,  6.2362e-03, -5.4127e-03, -1.5385e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 3.7579e-03, -8.1242e-01,  3.3873e-02,  1.4851e-03, -1.2252e-02,
        -1.3116e-02,  5.0435e-03,  1.3673e-02, -1.3558e-02, -6.8612e-03,
         1.8459e-02, -6.8646e-04, -1.0205e-02, -5.9863e-03, -2.0234e-03,
        -8.7469e-03, -4.6901e-03,  2.7310e-03, -6.1263e-04, -1.1203e-02,
         1.0266e-02,  2.2716e-04,  8.1271e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9164e-02,  7.9883e-01, -6.5410e-03,  1.8689e-02,  5.7012e-03,
         1.3105e-02,  5.5326e-04,  7.7800e-03, -1.4273e-03,  8.2767e-03,
        -1.3666e-02, -7.5870e-03, -5.2197e-03,  9.3954e-04,  8.5635e-04,
        -6.0158e-03, -1.0669e-02,  3.4977e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5803e-02,  8.2711e-01,  8.7284e-03,  1.9896e-02,  8.8585e-03,
         1.7384e-02, -8.0100e-05,  1.9039e-02,  3.6235e-04, -4.4211e-03,
        -6.0672e-03,  1.3072e-02,  2.2951e-02, -5.4497e-03, -2.2487e-03,
         4.8965e-03, -2.5523e-04, -3.3777e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1949e-02,  8.4536e-01, -4.4695e-03,  1.1310e-02,  9.1949e-04,
         1.9274e-02, -1.1123e-02,  4.4437e-03, -1.3318e-02,  4.5530e-04,
        -2.4341e-03,  5.1556e-03, -9.3506e-03,  7.6461e-04, -3.9673e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3413e-02,  7.9900e-01, -4.7263e-03,  6.8183e-03,  9.8876e-03,
        -8.3462e-04, -2.9595e-03,  1.1148e-03,  6.7383e-03,  1.1233e-02,
         1.4059e-02, -9.8044e-04,  4.1707e-03,  5.8532e-03,  3.1595e-03,
        -1.9472e-03, -2.6791e-03, -1.9003e-03, -2.5596e-03,  1.0547e-02,
         1.1222e-02,  1.2568e-03,  6.7806e-04, -2.1521e-03, -1.2813e-02,
         3.2597e-03,  5.7335e-03, -3.0692e-03,  2.1062e-03,  5.7023e-03,
        -7.4091e-03,  2.0012e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.2998e-02,  7.9144e-01,  8.4459e-03,  5.7307e-03, -1.0232e-02,
        -5.1250e-04,  1.0589e-03,  2.9574e-03,  1.8235e-03,  1.7094e-03,
        -8.5967e-04,  6.1967e-04, -4.6135e-03, -8.0695e-03, -2.5549e-04,
        -5.7430e-03, -3.8923e-03, -7.0528e-03, -1.1434e-02, -5.0774e-03,
         4.0746e-03,  8.1609e-03, -1.1898e-02,  3.4244e-04, -6.5279e-03,
        -4.6883e-03, -7.0319e-03,  5.0654e-03, -3.7925e-03, -1.1326e-03,
        -6.6047e-03, -4.2226e-03,  1.9293e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.8244e-02,  6.6899e-01,  1.5715e-02,  1.5874e-02,  2.0096e-02,
         2.7403e-03,  1.5168e-03, -5.2400e-03,  1.5947e-03,  9.2719e-03,
         7.7564e-04, -6.1857e-03,  1.1098e-02,  2.0604e-03,  1.9955e-02,
        -3.4032e-03, -2.2366e-03,  1.1212e-04, -6.7611e-03, -5.1377e-03,
        -1.1110e-03, -2.3645e-03, -6.0424e-04, -1.2221e-02,  7.4964e-03,
        -2.2981e-03,  2.0783e-03,  3.1766e-03,  4.9768e-03,  3.8946e-03,
         2.3180e-03,  4.3210e-03,  7.9097e-03,  8.9654e-04, -1.6868e-03,
        -3.2664e-04, -1.7833e-03, -2.0639e-03,  8.0558e-05,  1.6083e-03,
        -1.3066e-03, -1.0426e-03, -6.9521e-04,  4.1561e-03,  3.2382e-03,
         1.2767e-02, -1.9453e-03, -2.0698e-03, -1.0651e-02,  1.0500e-03,
        -4.5872e-03,  7.6339e-03,  1.2099e-03,  2.1586e-03,  3.9119e-03,
         4.1691e-03, -1.9695e-02, -7.4935e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0698e-02,  7.6696e-01, -2.9701e-02,  2.0361e-02,  2.1227e-03,
         1.0194e-03,  1.4911e-03, -3.0345e-03, -4.7301e-03,  4.3271e-03,
         1.2854e-05,  3.6236e-03,  8.4922e-03, -1.0188e-02,  4.3962e-03,
         4.3308e-03,  1.0078e-02, -2.0447e-03, -3.5424e-03,  2.7492e-05,
        -3.2974e-03, -2.1969e-03, -9.2592e-03, -5.9494e-03, -5.6033e-03,
        -5.2509e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0599,  0.6814, -0.0777,  0.0039,  0.0038,  0.0163, -0.0133, -0.0046,
        -0.0053,  0.0056,  0.0132, -0.0060,  0.0083,  0.0102, -0.0071,  0.0020,
         0.0029,  0.0034, -0.0050, -0.0023, -0.0033, -0.0151, -0.0219, -0.0044,
        -0.0231,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7317e-02, -7.4079e-01, -4.7152e-02, -1.3579e-02, -2.5912e-03,
         6.8920e-03, -5.4941e-03,  6.8411e-03,  8.5623e-03,  9.6994e-03,
         7.5017e-03,  1.4937e-03, -5.1728e-03, -7.0885e-03,  6.8615e-05,
         6.1611e-04, -2.6266e-03, -5.9394e-03,  9.1399e-04,  4.1027e-03,
        -8.7184e-03,  2.7104e-03,  3.6990e-03, -1.4335e-03, -1.1043e-03,
        -1.5145e-03, -4.1022e-03, -8.4378e-04, -4.7840e-04, -3.9975e-03,
        -4.4595e-04,  4.6494e-03,  2.8850e-03, -1.5732e-03, -8.1958e-03,
         2.4075e-03,  2.2803e-03,  6.6918e-04,  2.2327e-03,  8.0681e-04,
         3.5622e-03,  4.0730e-03,  4.3844e-05, -5.4768e-03,  3.7657e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4191e-02,  7.3743e-01, -2.9965e-02,  6.6288e-03,  3.9040e-03,
        -8.5312e-03, -8.7564e-03, -1.0717e-02, -3.7340e-03, -2.0334e-02,
        -6.8343e-03,  1.1644e-02,  4.0454e-03,  5.8946e-03, -1.1395e-02,
         9.0511e-03, -8.7572e-03, -6.8155e-05,  9.6759e-04,  3.7810e-03,
         1.1149e-02,  5.0604e-02,  1.6176e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5289e-02,  8.1007e-01, -5.7745e-02, -2.1060e-02, -8.1458e-03,
         5.8127e-03, -6.0849e-03, -5.3947e-03, -1.6292e-04, -7.1871e-03,
        -2.9553e-02,  1.3497e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 0.0193, -0.8075, -0.0076,  0.0152, -0.0074, -0.0150,  0.0051, -0.0044,
        -0.0110,  0.0135,  0.0203,  0.0067,  0.0110, -0.0111, -0.0063,  0.0028,
        -0.0064, -0.0034, -0.0041,  0.0045,  0.0037,  0.0043, -0.0092,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3636e-02,  7.9443e-01,  1.8034e-02, -1.3670e-02, -8.2568e-03,
        -1.4480e-02, -9.5885e-03,  2.9834e-03, -5.7576e-03,  1.2930e-02,
         3.2596e-03, -9.5633e-03,  1.4368e-03,  9.2234e-04, -1.1052e-02,
        -4.5128e-03, -3.6969e-03, -7.9816e-03, -5.6748e-04,  3.8348e-03,
        -6.3194e-03,  1.6513e-03, -7.9255e-03, -1.3155e-03, -5.2099e-03,
         1.1876e-03, -5.9016e-03, -6.0954e-03, -3.9388e-03,  2.2431e-03,
        -6.7370e-04, -1.6639e-03,  5.2782e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2273e-02,  8.2367e-01,  8.0682e-03,  2.9869e-03,  3.6612e-03,
        -3.6586e-03, -1.6538e-02, -1.4097e-02,  4.6604e-03,  9.2552e-03,
        -2.1954e-03,  5.2191e-03,  7.5861e-03,  2.0209e-02,  8.3515e-03,
        -6.2366e-04, -4.5098e-03,  2.1829e-02, -1.0604e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1090, -0.7405,  0.0180, -0.0198,  0.0286, -0.0151, -0.0133,  0.0080,
         0.0071, -0.0107,  0.0060,  0.0148,  0.0054,  0.0037,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0284,  0.8022,  0.0339, -0.0190, -0.0134,  0.0109,  0.0103, -0.0086,
         0.0082,  0.0051,  0.0022, -0.0012,  0.0028,  0.0090, -0.0037,  0.0058,
        -0.0041, -0.0044,  0.0032,  0.0049, -0.0015,  0.0172,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0344e-02, -7.8549e-01,  8.5608e-03, -9.7156e-03, -4.3635e-02,
        -5.4644e-03,  5.8646e-03,  6.5233e-03, -3.3259e-03, -1.1944e-02,
         3.0783e-03, -8.7517e-03,  1.4501e-02,  3.1778e-03, -5.7534e-03,
         1.4117e-04, -3.9784e-03, -8.6664e-04,  6.7993e-03,  1.1860e-03,
         1.3607e-03, -1.3207e-03, -1.3317e-03,  8.6227e-03,  6.0155e-03,
        -6.0608e-03, -1.4340e-04, -3.5915e-03, -8.4064e-03, -2.4051e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0084,  0.7852, -0.0375,  0.0048,  0.0077, -0.0110,  0.0366, -0.0138,
        -0.0130,  0.0040, -0.0085, -0.0033, -0.0023,  0.0126,  0.0092, -0.0080,
         0.0343,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0825,  0.7477,  0.0407,  0.0240, -0.0091,  0.0232, -0.0114,  0.0249,
         0.0325,  0.0013, -0.0028,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6890e-02,  7.3007e-01,  6.7618e-02, -3.4078e-03, -4.3919e-03,
         1.8772e-02, -4.2433e-03, -1.1096e-02, -5.9405e-04, -1.2424e-03,
         4.2011e-03, -4.1979e-03, -7.4492e-03,  4.5653e-02,  2.5230e-02,
        -1.5285e-02, -1.9661e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.0787e-03,  8.0470e-01, -2.2586e-02,  1.6999e-02,  1.5734e-02,
        -5.8357e-03, -3.1054e-03,  3.9151e-03, -2.1818e-04, -4.5494e-03,
        -8.9887e-03,  1.2483e-02, -6.2221e-03,  1.0920e-03, -3.5072e-03,
         2.3089e-03, -2.7195e-03,  1.3126e-04, -1.1799e-02,  1.2918e-04,
        -1.2481e-03, -7.5339e-03,  1.8859e-02, -9.4711e-05,  5.2378e-03,
        -3.2921e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3267e-03,  8.0872e-01,  3.9172e-02, -1.3400e-02, -3.2245e-03,
         3.5860e-02, -1.6205e-02, -4.3880e-04,  1.9273e-03, -3.9482e-03,
        -5.9987e-03, -4.9108e-03, -4.7031e-03, -4.6008e-03, -3.2744e-04,
         1.2059e-03, -2.9737e-03,  8.9298e-04,  1.4518e-04,  8.0838e-04,
        -7.9399e-03,  3.4340e-03, -4.0630e-03, -3.3605e-03, -3.9579e-03,
         6.5268e-04, -3.7954e-04,  5.3311e-03,  4.1388e-04, -1.4676e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3342e-03,  8.3114e-01, -1.0509e-02, -2.1605e-02,  1.4763e-02,
        -7.1036e-03,  7.1142e-03, -5.9654e-05,  1.4540e-02,  5.5970e-03,
        -5.6749e-03, -1.4438e-03, -3.4568e-03,  2.6050e-03, -7.7051e-03,
         3.6954e-03, -1.2368e-02, -8.8260e-04,  4.7402e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 0.0633, -0.6891,  0.0161, -0.0278, -0.0019,  0.0054, -0.0192,  0.0044,
         0.0009,  0.0008,  0.0013,  0.0039, -0.0017, -0.0095,  0.0088, -0.0096,
        -0.0151, -0.0116, -0.0083, -0.0032,  0.0024,  0.0080,  0.0102, -0.0018,
         0.0059, -0.0577,  0.0121,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.6018e-02, -8.2832e-01,  1.0486e-02,  6.4286e-03, -6.0082e-03,
        -9.0906e-03, -9.6728e-03, -1.6044e-03, -4.7034e-04, -3.8749e-03,
         1.2412e-04, -3.4101e-02, -1.2479e-02, -1.1321e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4806e-02,  8.2881e-01, -5.6205e-03,  3.3492e-03, -1.8699e-03,
         9.5570e-03, -1.3732e-03,  7.8578e-03,  1.4716e-02,  4.0331e-03,
         1.1969e-02,  3.1079e-04, -1.1662e-03,  5.5859e-03, -7.0633e-04,
        -4.1090e-03, -4.2727e-03, -3.9121e-03, -9.4709e-03,  4.8794e-03,
        -5.8781e-03, -8.8626e-04,  3.3876e-04, -7.7209e-04, -2.0857e-02,
        -3.2892e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.7182e-03,  8.2629e-01,  2.3533e-02,  3.0957e-02,  1.2689e-02,
        -1.1335e-02,  9.9455e-03,  4.6621e-04, -6.2044e-03,  1.2193e-02,
         7.3211e-03,  3.6903e-03,  3.5730e-02, -9.9296e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8394e-02,  8.1654e-01,  1.4863e-02, -1.1019e-02, -1.1705e-02,
        -1.1120e-03,  2.6868e-03, -1.0309e-02,  1.0593e-03,  3.4914e-03,
         2.3015e-03, -5.3027e-03, -1.5979e-03,  3.2345e-03, -1.8633e-03,
         3.5800e-03,  2.2480e-03, -3.3865e-05,  2.8049e-03, -3.9168e-04,
        -1.2258e-03,  2.3156e-04, -1.0946e-02, -5.7709e-03, -5.4721e-03,
        -3.0790e-03, -4.9897e-03,  5.5875e-04,  6.8241e-03,  6.4208e-03,
        -5.6969e-03,  2.4246e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7059e-02, -8.3672e-01, -2.5962e-02,  8.7481e-03,  2.0749e-02,
        -1.2117e-02,  7.0227e-03,  1.3629e-02, -9.3814e-05,  5.5428e-04,
         8.1074e-03, -1.8328e-03,  2.1731e-03,  3.5351e-04, -1.2097e-04,
        -1.3318e-04,  1.1025e-03,  3.9354e-03,  5.1837e-03,  1.6503e-03,
         6.6182e-03, -2.3642e-03, -1.4717e-03,  1.2080e-02,  2.1430e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5518e-02,  8.2236e-01, -3.4937e-04, -9.3162e-02,  2.6650e-02,
         2.1962e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2412e-02,  8.1579e-01,  1.7656e-02, -1.8271e-02,  5.8057e-03,
        -7.8054e-03, -2.0076e-03, -1.1005e-03,  3.3645e-03, -5.6163e-03,
        -1.4981e-03,  8.9004e-03,  6.0533e-04,  7.3163e-03, -9.7424e-03,
        -2.2899e-03,  3.9388e-04, -3.9629e-04, -2.7634e-03,  7.0420e-04,
        -4.5179e-04,  1.8016e-03, -5.9757e-03,  3.7613e-04,  2.1158e-03,
         6.6177e-03,  2.5934e-04, -3.1387e-04, -7.6037e-04,  4.2174e-03,
         5.9004e-04, -7.5987e-04,  6.1539e-03, -6.3458e-04,  3.7686e-03,
         6.6785e-03,  1.4086e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6581e-02, -7.4960e-01, -1.7866e-02, -3.8000e-02, -4.9419e-03,
        -1.1268e-02, -5.7748e-03,  1.5377e-03,  2.3129e-03, -3.1095e-03,
        -8.6091e-04, -6.2749e-03, -1.4902e-02, -5.9663e-03, -1.7264e-02,
         2.5569e-02,  1.9532e-03, -7.4508e-03, -3.9224e-03, -5.2463e-03,
         8.3810e-03,  4.1438e-03,  5.0723e-03, -8.0321e-04, -1.9402e-03,
         1.2350e-03,  5.5334e-03, -3.4075e-03,  3.8952e-03,  3.6126e-04,
        -5.8108e-04, -7.2786e-04,  5.1203e-04, -4.6612e-05, -2.9611e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0502, -0.6999,  0.1334,  0.0105, -0.0100, -0.0087,  0.0036,  0.0130,
         0.0073, -0.0059,  0.0120,  0.0157,  0.0049,  0.0067,  0.0181,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7909e-03,  8.6630e-01,  1.1887e-02,  1.1369e-02,  4.4658e-04,
         9.7666e-04, -3.2807e-03,  1.0312e-02,  1.2024e-02, -5.3215e-04,
        -6.5593e-03, -8.4280e-03, -1.5175e-03, -6.0229e-04, -7.7514e-03,
         3.0491e-03, -3.4329e-03, -5.1175e-03,  3.6918e-03, -1.3459e-03,
         1.1998e-05, -4.4577e-03,  3.5098e-03, -2.7959e-03,  2.3600e-03,
         3.8756e-04, -2.9253e-03, -5.6809e-04, -3.2478e-03,  2.2273e-03,
         1.7061e-03,  1.8459e-03,  1.3546e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.5358e-03, -8.2107e-01, -1.2060e-03, -2.3634e-02,  6.2631e-03,
        -7.4218e-03,  5.3789e-03, -5.8147e-03,  9.4598e-03,  5.8864e-03,
        -2.8506e-03, -8.6955e-03, -5.3370e-03,  1.2419e-02,  7.6511e-03,
         3.5552e-03,  4.5911e-03, -3.0083e-03, -2.5075e-04, -2.2349e-03,
         7.1791e-03,  4.2826e-03,  5.8623e-04,  5.2596e-04, -1.4170e-03,
        -3.9309e-05,  1.4784e-02,  9.5321e-03,  1.8390e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 6.7639e-02,  8.0608e-01, -1.6340e-02, -8.9033e-03,  1.1109e-02,
         2.2665e-03, -2.3149e-03, -1.0056e-02,  6.8728e-04, -2.0145e-02,
        -6.3207e-03, -1.1563e-02, -2.5362e-03, -2.1061e-02, -1.2982e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0877e-03,  8.5391e-01,  1.1339e-02,  3.2745e-03, -1.3556e-03,
         6.6405e-03,  5.0814e-04, -2.0732e-03,  4.3184e-03, -3.9937e-03,
        -7.1642e-03, -2.5738e-03, -6.9865e-04,  7.8362e-05,  8.8923e-04,
         8.7349e-05, -1.8706e-03,  1.7778e-03, -1.8515e-04, -2.2177e-03,
        -2.2396e-03, -1.4422e-03, -4.9788e-03,  2.9241e-03,  5.4118e-03,
         3.2646e-03, -6.9664e-03,  2.3181e-03, -4.5288e-04, -2.0449e-03,
         1.0624e-03, -1.9941e-03, -9.4531e-04, -2.5124e-03, -5.2814e-03,
         2.5547e-03, -3.1883e-03, -7.5781e-04,  2.5077e-03, -5.6449e-04,
        -6.7922e-03, -2.3958e-04,  6.6055e-04, -2.8780e-03, -6.0760e-04,
        -1.2401e-02,  5.1490e-04,  4.1108e-03, -7.3390e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.0372e-02,  7.6854e-01, -2.7175e-02,  1.5748e-03, -1.0120e-02,
         6.2872e-03, -1.4324e-05,  9.3304e-04, -1.9395e-02,  1.8397e-03,
        -3.6596e-03, -1.8556e-03, -4.0363e-03,  2.6132e-03, -6.7432e-03,
        -2.1983e-03,  2.3275e-02, -9.9438e-04,  1.8789e-03, -5.5619e-03,
        -2.5646e-03, -2.4584e-03, -9.5525e-03,  4.5685e-05, -3.6120e-03,
         6.8303e-04, -2.6948e-03,  2.6044e-03,  4.4668e-03, -1.0334e-04,
         3.7153e-03, -6.6063e-03, -3.6836e-03,  4.4407e-03,  2.3704e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4236e-02,  8.7137e-01, -1.2382e-02,  3.6615e-02, -4.5270e-03,
         1.0192e-02,  1.1324e-02,  2.6149e-03,  2.2714e-03, -3.4901e-03,
         1.7324e-03, -4.0972e-04, -4.6010e-03, -1.4267e-03, -2.0711e-03,
        -6.1422e-03, -3.9217e-04, -4.4340e-03,  9.7650e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0171, -0.8246,  0.0204, -0.0376, -0.0101,  0.0028,  0.0031, -0.0061,
        -0.0242,  0.0064, -0.0022,  0.0064, -0.0065,  0.0040,  0.0285,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8515e-03,  8.6225e-01, -6.9210e-03, -4.7582e-03,  1.9114e-02,
         1.5584e-02, -3.7281e-03,  1.6655e-03,  4.5576e-03,  8.5203e-03,
        -7.6036e-04, -3.5423e-03, -1.4698e-02,  3.8845e-03, -9.6770e-03,
         2.2874e-03, -5.5505e-03,  3.9794e-04, -1.0829e-02,  1.9576e-06,
         1.5578e-04,  1.7268e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9792e-02,  8.1749e-01,  2.5611e-02, -1.0053e-02, -1.1019e-02,
        -1.9307e-03, -4.1976e-03,  5.3139e-03,  2.7473e-03, -9.6320e-03,
        -1.0213e-02,  4.8283e-04,  2.2216e-03,  5.8348e-03, -6.1105e-03,
        -2.2904e-04,  6.0460e-03,  1.0432e-02, -5.3168e-03,  1.0311e-03,
         2.2071e-04,  3.3830e-03, -4.3012e-04,  6.4404e-03, -1.3226e-03,
         2.5531e-03, -8.6213e-04,  3.6505e-03,  5.8170e-04, -4.8539e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1146,  0.7185, -0.0229, -0.0217, -0.0380, -0.0140,  0.0044, -0.0071,
        -0.0145,  0.0201, -0.0242,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7704e-02,  8.4510e-01, -5.0266e-03,  8.1984e-03,  1.8201e-02,
         3.4974e-03,  2.3877e-03, -2.7172e-03,  3.8797e-03,  9.6484e-03,
        -1.9276e-03, -1.3650e-04,  3.6226e-03,  8.2050e-05, -3.8870e-03,
         9.0066e-04, -1.8162e-03, -4.0692e-03,  3.2773e-03, -2.8780e-03,
         1.2510e-02, -8.5337e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.5974e-03,  8.3044e-01,  1.9596e-02, -6.2499e-04,  1.7604e-03,
         1.0402e-02, -7.4232e-03,  7.4891e-03, -3.4233e-04, -8.0654e-03,
        -1.7234e-02, -5.2248e-03,  3.7576e-04,  4.1795e-03, -8.3144e-03,
         1.1655e-03, -1.1500e-04, -6.5007e-04,  7.2736e-03, -8.3878e-04,
        -5.7317e-04, -2.9326e-03, -1.0386e-02, -1.6233e-03, -8.4770e-04,
        -1.6436e-03, -5.9690e-04, -4.3633e-03,  1.5874e-03, -1.1825e-04,
        -1.1665e-02, -1.2586e-02,  1.0961e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6322e-02,  8.0284e-01,  3.9129e-02, -2.2584e-02, -3.3878e-03,
         4.2029e-03, -1.3446e-03,  4.1205e-03,  1.3440e-03, -3.9762e-03,
        -3.5528e-03,  1.1781e-03, -7.6514e-03, -7.3237e-03,  1.1237e-04,
        -6.5169e-03,  1.1020e-03, -9.1530e-03, -6.8000e-03,  1.8357e-03,
         1.2833e-02, -6.0798e-03,  1.2903e-03, -1.5646e-03,  8.8099e-03,
         5.9050e-03, -1.4697e-02,  4.3417e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1134e-03,  8.1672e-01,  1.6993e-02,  4.4322e-03,  7.9836e-03,
        -1.7296e-03,  3.8568e-03,  1.0253e-02, -1.5299e-02, -5.2275e-03,
        -1.8493e-02,  1.3624e-04,  3.9122e-03, -3.8433e-03, -2.4189e-03,
        -7.1854e-03,  1.1554e-04, -7.1442e-03, -4.2236e-03, -3.1639e-03,
        -1.3769e-02,  6.8336e-03, -1.6007e-02, -1.4033e-02, -1.3109e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 6.9769e-03,  8.0030e-01,  3.7996e-03,  2.4900e-02,  3.2631e-03,
        -7.3279e-03, -8.2836e-04, -3.1927e-04,  2.6913e-03, -2.6504e-03,
        -2.1216e-03, -5.2078e-03,  2.6495e-04,  2.1450e-03, -7.0053e-04,
        -3.9736e-02, -5.4192e-03,  1.5121e-03, -6.0327e-04, -2.1095e-07,
        -2.8652e-03, -4.3211e-03, -6.0501e-04, -1.5875e-03,  4.2227e-03,
        -1.3292e-03,  5.7545e-03, -3.9782e-03,  1.2846e-04,  3.1913e-03,
        -7.7554e-03, -4.3969e-03, -6.5121e-03,  5.2343e-04, -4.6130e-06,
         9.8883e-04, -2.0560e-03, -2.9309e-03,  4.8827e-03, -8.7580e-03,
         2.2439e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9019e-03,  8.5099e-01,  1.7933e-02, -2.5013e-02,  1.2132e-02,
        -2.5219e-03, -6.2423e-03,  1.3501e-02,  9.2046e-04,  5.6179e-03,
        -4.4561e-03,  8.0802e-03,  4.8588e-03, -4.6876e-04, -6.5226e-04,
        -1.6212e-05, -1.9955e-03,  5.6977e-03,  3.9158e-04, -5.5170e-03,
        -1.4959e-03, -3.6478e-03,  1.2141e-03, -2.2788e-02,  1.9507e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.4476e-03,  8.3023e-01,  5.1468e-02,  2.7818e-03,  1.3979e-03,
         1.2732e-02,  2.2164e-02,  8.4730e-03,  1.9262e-03,  9.9970e-03,
         7.4091e-03,  2.8284e-03,  8.8248e-04,  1.0243e-02,  5.9205e-03,
        -1.4266e-03,  1.5901e-03,  9.7190e-04, -2.5620e-03, -3.9648e-04,
        -2.0155e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0322,  0.7622,  0.0309, -0.0160, -0.0132,  0.0049, -0.0059,  0.0019,
        -0.0048, -0.0045, -0.0363, -0.0271, -0.0075, -0.0078, -0.0044, -0.0059,
        -0.0224,  0.0121,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1969e-02,  8.0092e-01,  2.8743e-02,  3.3501e-02,  3.7055e-03,
        -2.7054e-03,  9.1348e-04, -1.4535e-02, -5.8562e-03, -2.8960e-05,
         6.2517e-03, -3.2985e-03, -4.8902e-03,  5.7104e-03, -9.4293e-04,
         1.6906e-03, -7.1222e-03,  5.5653e-03, -2.2738e-03, -1.2217e-04,
        -9.9455e-04, -3.2551e-03,  9.8089e-04, -9.3644e-03,  7.8037e-03,
        -7.5166e-04, -9.1886e-03,  2.2453e-03, -3.6035e-03, -2.5485e-03,
         1.3386e-03, -5.0741e-04,  3.5373e-03,  9.0023e-04,  9.0456e-04,
         4.7252e-03, -4.4721e-03,  2.1325e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0152,  0.8340, -0.0198,  0.0200, -0.0076, -0.0012, -0.0050,  0.0175,
         0.0122, -0.0010, -0.0120, -0.0051, -0.0039,  0.0110,  0.0026,  0.0018,
         0.0012, -0.0051, -0.0056, -0.0183,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.0027e-04,  8.7791e-01, -8.6534e-04, -7.9412e-03,  9.4237e-03,
         2.1378e-02,  4.1093e-03,  1.9529e-02,  4.3404e-03, -8.4837e-03,
         6.6790e-03, -4.4633e-03, -5.3948e-03, -2.8781e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9998e-02,  7.9066e-01, -2.8692e-02,  1.1876e-02, -1.9199e-03,
        -3.5422e-03, -6.7995e-03, -2.6225e-03, -7.6392e-03,  4.5748e-03,
        -3.8331e-03,  1.7112e-03, -3.1597e-03, -1.0303e-03,  4.7815e-04,
        -7.2927e-03,  3.9924e-04,  5.0357e-03,  3.2677e-03, -2.3098e-03,
        -1.5678e-04,  9.7788e-03, -3.1983e-03,  1.4287e-03,  1.0250e-04,
        -1.4308e-02,  6.8140e-03,  5.3545e-03, -1.1281e-03, -1.8824e-02,
         1.4401e-03, -1.0621e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.6928e-02,  8.2992e-01,  6.4185e-03, -7.3409e-03,  2.2281e-02,
        -1.2191e-02,  9.9176e-03, -3.8771e-03,  3.7060e-03, -2.6288e-03,
         5.4287e-04, -6.8313e-03, -1.4076e-03,  8.7816e-03,  2.8952e-04,
         3.0998e-04, -9.8096e-04,  9.5569e-03, -5.6268e-03, -4.6036e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4779e-02,  7.0725e-01, -2.4010e-02, -1.2476e-02,  1.7725e-03,
         8.1456e-03,  2.2758e-03, -1.5611e-04, -5.8101e-03, -1.4546e-03,
         1.7382e-03, -1.7237e-03, -5.0491e-03,  1.0159e-03, -9.1332e-04,
        -7.0199e-03, -2.6504e-03, -6.2507e-03,  1.2763e-03,  2.4771e-03,
         1.1664e-03,  1.7077e-03, -7.6746e-03,  1.1140e-02, -2.8262e-03,
        -7.5344e-03, -9.7845e-03, -1.5738e-02,  3.9450e-03, -5.7110e-03,
        -2.5581e-05,  5.5680e-03,  9.6408e-03,  8.5392e-03,  1.8456e-03,
         9.1359e-03,  1.1686e-03, -1.6835e-03, -2.6356e-03, -3.7317e-05,
        -7.3702e-03,  7.0918e-03, -2.4193e-03, -1.8690e-04, -2.8749e-03,
         2.9722e-02, -2.4585e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1043,  0.7524, -0.0181, -0.0134,  0.0174, -0.0017,  0.0017, -0.0077,
         0.0041, -0.0068, -0.0067,  0.0041,  0.0085,  0.0115, -0.0418,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2672e-04,  8.0820e-01,  4.3071e-03, -2.3992e-03,  1.7114e-03,
        -1.3362e-03,  6.5139e-03,  1.5628e-02, -2.7277e-04, -3.4243e-03,
         9.2687e-04, -9.8997e-04, -1.6801e-04,  1.2659e-03, -6.9075e-03,
        -3.4149e-03, -3.6789e-03,  4.2760e-03,  1.9424e-03, -9.6486e-03,
         5.7364e-03, -2.9726e-03, -5.3273e-03,  9.6081e-03,  3.2929e-03,
        -9.5971e-03,  3.1063e-03, -1.2876e-02,  7.2828e-04, -8.7421e-04,
         3.0059e-03, -9.7130e-03, -3.0427e-03, -4.4100e-03,  4.4649e-03,
         3.9863e-03,  2.3232e-03, -1.5392e-03, -5.7252e-03, -1.1427e-03,
         3.2209e-03, -6.4946e-04, -1.5894e-03, -9.2559e-04, -1.5496e-03,
        -1.2978e-03,  4.8274e-03,  1.5031e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-1.4139e-02,  8.3846e-01,  7.6750e-03, -7.8845e-03,  1.4488e-02,
         1.7428e-02, -5.2881e-03, -4.2533e-03,  6.4387e-05,  5.9221e-03,
        -1.7602e-03,  2.4332e-04, -1.0777e-02, -1.0226e-03,  1.9229e-03,
         1.9844e-03,  4.0811e-04,  1.0878e-03, -2.0172e-03, -4.2392e-03,
        -4.6905e-03,  5.3841e-03,  8.5501e-03, -5.4085e-03,  1.2741e-02,
        -2.2156e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1818e-02,  7.9991e-01,  3.6905e-03,  1.7762e-02,  4.5098e-03,
         2.1338e-02, -8.4544e-03,  8.8534e-03, -8.0426e-04, -5.2288e-03,
         2.3526e-02, -7.1196e-03,  4.0061e-04, -2.4179e-03, -9.5858e-03,
        -2.3292e-03, -4.8856e-03, -6.5133e-03,  1.5724e-02,  7.8620e-04,
         4.3475e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9740e-03,  7.8196e-01,  1.4625e-02, -1.1733e-03, -3.4739e-02,
        -2.0242e-02,  8.0527e-03,  6.5963e-03, -4.8179e-03, -4.9959e-05,
         1.9391e-03,  3.2468e-03, -3.9861e-03, -4.3512e-03, -8.0380e-03,
        -3.5237e-03, -1.2015e-03, -8.6394e-03,  8.9952e-03, -9.3674e-03,
        -1.0411e-02, -6.0898e-04,  4.8388e-03, -5.2997e-03,  1.1408e-02,
         1.1160e-03, -5.0912e-03,  5.7901e-04, -3.2409e-04,  9.5976e-03,
        -5.4319e-03,  7.3376e-04,  8.0330e-03,  2.1994e-04, -5.3430e-03,
        -2.4433e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0034e-02,  7.1958e-01, -7.3286e-02,  6.1190e-03,  9.8960e-04,
         6.7236e-04, -1.2871e-02, -6.4201e-03, -1.3134e-02, -7.2025e-03,
        -5.6141e-03, -3.5294e-03, -1.2297e-02, -2.0388e-02, -5.7983e-03,
        -5.5468e-03, -5.9111e-03, -3.3731e-03, -1.1768e-02, -3.5468e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.9747e-03, -8.3590e-01, -3.1366e-02, -2.6859e-03,  3.9042e-04,
        -7.3279e-03,  1.2240e-02,  8.4113e-03, -5.6972e-03,  1.7892e-03,
        -1.4116e-02,  5.8789e-03,  3.9399e-03,  7.8765e-03, -1.2615e-02,
         1.9221e-03,  3.7196e-03,  7.1562e-03,  7.3357e-03, -1.9167e-02,
        -4.4859e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2136e-02,  8.1138e-01, -2.9439e-02, -1.2547e-02, -2.9647e-02,
         3.4503e-03, -1.2168e-02, -1.7129e-02, -1.9919e-02, -1.2277e-02,
        -4.8575e-03,  3.9695e-04,  4.6499e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.7908e-03,  7.6926e-01,  2.7064e-03, -1.3627e-02, -3.0829e-02,
        -1.7706e-02, -2.5021e-02, -4.8211e-03, -8.2749e-03, -1.4623e-03,
        -4.5058e-03, -1.4467e-02, -2.5657e-03, -1.5248e-03,  1.0348e-02,
        -1.9756e-03, -4.9663e-04, -1.8487e-03, -3.9596e-03,  2.2282e-03,
        -5.1794e-03,  1.2294e-02, -1.7343e-03, -3.3061e-03,  1.7317e-03,
         4.9648e-03, -6.0359e-03,  3.4281e-03,  2.6665e-03,  1.3018e-04,
         1.9757e-03,  2.3149e-04,  3.4811e-03, -2.1440e-03,  5.9310e-03,
        -1.8197e-02,  1.5192e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3499e-02,  7.8964e-01,  7.0578e-02, -9.9356e-05, -2.1460e-02,
        -1.3280e-02, -2.8083e-03, -2.3640e-03, -6.8735e-03, -7.4836e-03,
        -1.1187e-02,  2.4404e-03,  8.7072e-03, -2.2490e-03, -5.4851e-03,
         4.1039e-03,  3.9328e-03, -2.7720e-03, -3.5310e-03,  2.9425e-03,
        -6.4332e-03, -3.6036e-03, -2.3731e-03,  2.1501e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.8279e-02,  7.5818e-01,  1.3527e-02,  1.4408e-02, -2.0514e-03,
         6.7726e-03, -9.5593e-04,  4.1855e-03,  9.0973e-03, -1.4324e-02,
         4.5494e-03, -8.8505e-03,  1.4017e-02, -3.4507e-03,  3.2407e-03,
        -5.7953e-03, -5.7039e-03, -3.7935e-03,  7.3847e-04,  1.3499e-03,
         3.5494e-03, -1.3198e-02, -9.6775e-03, -5.5012e-03, -9.0961e-03,
         1.2572e-03, -1.8667e-03, -3.2462e-03, -1.1709e-03, -5.4114e-04,
        -4.3779e-03, -1.3246e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1504e-02,  7.8282e-01,  1.7421e-02, -8.7130e-03,  4.1697e-04,
        -4.2829e-03, -1.3166e-02,  4.9166e-03,  4.2958e-03, -8.2834e-03,
         1.6475e-03, -2.1031e-03, -1.2188e-03,  1.5429e-03, -7.4209e-04,
         1.7479e-03, -6.0806e-03, -1.8245e-03, -3.8965e-03, -4.5297e-03,
        -1.0315e-02,  6.5966e-03, -3.8848e-03, -5.7392e-03, -1.5535e-03,
         1.2618e-02,  1.8546e-02, -7.0832e-03, -1.5578e-03, -1.2594e-02,
         8.3543e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0069e-02,  7.7643e-01, -8.0625e-02, -4.0787e-03,  2.3373e-03,
         5.6149e-03, -3.9336e-03, -1.1455e-02,  9.7797e-03,  7.1692e-04,
         1.0114e-02, -7.6259e-03, -1.4147e-02,  7.4013e-04,  4.8677e-03,
         9.9897e-03,  3.4021e-03, -2.3930e-03,  4.6214e-03, -6.8302e-03,
        -9.1887e-03, -1.5510e-02, -5.5285e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6776e-02,  7.5906e-01,  4.5891e-02,  4.5813e-03,  2.2294e-02,
         4.0634e-03, -4.1964e-03, -1.3254e-03, -2.3692e-03,  1.0232e-03,
         5.0640e-03, -2.8033e-03, -4.2984e-03, -1.0002e-02,  1.0475e-02,
         2.8649e-03,  3.3844e-03, -2.6557e-03,  1.1010e-03, -2.3588e-03,
        -1.6734e-02, -6.6001e-03,  2.0295e-03,  1.6437e-04, -4.6918e-03,
         4.9479e-03,  2.3391e-02,  1.4856e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-1.6172e-02,  8.6237e-01, -1.7410e-03, -8.6008e-03, -1.7981e-02,
         5.5755e-04,  1.9581e-03, -1.9542e-03, -6.9309e-03, -1.1196e-02,
        -1.5073e-04, -4.5649e-03, -3.3264e-03, -9.5932e-05, -5.1248e-03,
         1.8266e-04, -6.0711e-03,  4.3197e-03, -3.9147e-03,  1.4961e-03,
         7.5709e-03,  4.3122e-04, -6.2047e-03, -3.4632e-03, -6.8749e-04,
         6.8978e-04,  3.6788e-03, -1.0098e-02,  8.4695e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4406e-03,  8.3655e-01,  8.3300e-03, -1.2832e-02, -9.0584e-03,
         9.6727e-03,  3.1336e-03, -4.8625e-03,  2.0145e-03,  1.3611e-02,
        -3.9785e-03, -5.6240e-03, -1.1745e-02, -9.2902e-03, -4.7342e-03,
         3.2689e-04,  6.9173e-03,  2.6437e-03, -1.3746e-02, -8.5063e-03,
        -8.1338e-04,  7.5663e-03, -1.3654e-02,  3.9504e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2606e-02,  7.2560e-01,  2.4738e-02,  3.8639e-02,  2.4978e-02,
         1.5900e-02,  1.5067e-03, -4.0078e-04,  1.7414e-02,  1.0870e-03,
         8.7763e-03,  2.6570e-03,  6.0820e-03, -4.9394e-03, -1.3295e-02,
         5.7896e-03, -5.6493e-03,  1.8436e-03,  2.5868e-03,  1.7786e-03,
         4.8701e-05, -8.5660e-03, -6.4651e-05,  7.7078e-04, -7.3743e-04,
        -3.4748e-03, -1.1266e-02, -1.8040e-05, -2.0931e-04,  2.6998e-03,
        -7.3439e-03,  1.8626e-02, -8.1850e-03, -2.1723e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.2370e-02, -8.0041e-01,  1.2061e-02,  2.1913e-02, -1.7016e-03,
        -5.1263e-03, -1.8701e-02, -2.5397e-03, -8.5043e-03, -6.0802e-03,
         2.7443e-03,  7.2296e-03,  2.4729e-03,  6.0920e-03,  2.4702e-03,
        -2.3704e-03, -2.0860e-03, -4.2027e-04,  6.3531e-04, -1.0021e-02,
         9.4510e-03,  1.4603e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5484e-03, -8.9765e-01,  9.6099e-03, -1.6435e-02, -9.4067e-03,
         6.0378e-03,  2.1113e-03,  3.8838e-03,  3.0994e-03, -1.6783e-03,
        -3.0430e-02,  9.9837e-04,  3.0132e-03, -9.2602e-04, -5.6651e-03,
        -1.5179e-03, -8.4146e-04, -5.1470e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7435e-02,  8.0184e-01,  2.8993e-02,  5.2629e-03,  1.1021e-02,
         6.4710e-04,  6.1643e-03, -1.0263e-02, -6.8229e-03,  6.6014e-03,
         2.7557e-04, -5.0233e-04, -1.0658e-02, -6.0333e-03, -1.1729e-03,
        -6.1278e-03,  2.6264e-03, -9.3447e-03, -1.1197e-02, -5.9232e-03,
        -1.0607e-02,  2.6100e-02, -4.3824e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2307e-02, -8.4774e-01,  9.3527e-04, -7.8689e-03,  3.6287e-02,
        -4.5341e-03,  3.1192e-03,  4.5515e-03,  1.4829e-02,  1.1754e-02,
         7.4183e-03, -3.7762e-04,  1.9593e-03, -7.6379e-03, -2.2365e-03,
        -1.4768e-02,  2.1677e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5551e-02, -8.3495e-01,  5.9320e-03,  3.6224e-02,  1.8390e-03,
         2.3057e-03, -2.5746e-04, -1.1293e-02, -1.7910e-02, -4.8938e-04,
         6.0109e-03,  3.4501e-03,  5.3583e-03,  2.3729e-03,  8.2038e-03,
        -6.4178e-03,  1.9034e-02,  2.2403e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.0923e-02,  7.8577e-01,  1.8781e-02, -6.4807e-03,  7.2346e-03,
        -1.7131e-03, -9.8131e-03,  8.1170e-03, -5.3962e-03,  2.1661e-02,
        -1.6011e-02, -3.8340e-03, -1.3379e-03,  1.5030e-03, -4.1454e-03,
         1.9471e-03,  2.9078e-03,  8.8024e-03,  9.8882e-03,  2.1114e-03,
        -5.9407e-03,  3.2915e-03, -1.0116e-02,  2.4214e-04, -7.3374e-03,
        -1.4697e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.1716e-02,  8.3399e-01,  6.1917e-04, -5.1055e-03,  5.0909e-03,
        -6.4813e-03, -1.0681e-02, -1.0209e-03,  1.3901e-02, -1.4345e-02,
         5.5891e-03,  3.2413e-04, -5.5547e-03, -7.5619e-03, -6.5240e-03,
        -6.6777e-04, -5.2718e-03,  1.4353e-03,  3.7823e-03, -1.5286e-03,
         6.3580e-03, -1.0065e-03,  2.8392e-03,  3.3797e-03, -1.9385e-03,
         5.3849e-03, -1.7625e-04, -5.1469e-04, -1.1737e-03,  4.3329e-03,
        -1.7004e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0521,  0.7863, -0.0236, -0.0137,  0.0008, -0.0012,  0.0083,  0.0096,
         0.0076, -0.0315, -0.0058, -0.0160, -0.0028, -0.0051, -0.0066, -0.0033,
        -0.0076, -0.0183,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0390, -0.8713,  0.0093,  0.0075,  0.0028, -0.0361, -0.0284,  0.0055,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 3.1072e-02,  7.8977e-01,  3.1406e-02, -7.8215e-03,  5.1403e-03,
         1.3305e-02,  1.0320e-02,  7.1110e-04,  3.5158e-03, -1.5013e-03,
         7.9074e-05, -1.1776e-02,  5.8279e-03,  7.5465e-03, -7.8787e-03,
        -4.1219e-03,  1.1353e-03,  1.4575e-03,  3.6355e-03,  6.7681e-03,
         3.9415e-03,  1.1189e-03, -9.9107e-04, -1.4518e-03, -2.5376e-03,
         3.6207e-03,  1.9408e-03, -7.5932e-04,  2.6043e-03, -2.1518e-03,
         2.0373e-03,  7.9422e-04, -2.3982e-04, -1.6374e-03,  1.1416e-02,
         1.3887e-03,  4.6501e-03,  5.4451e-03,  6.4820e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1890e-02,  8.4610e-01,  2.7246e-02, -1.2900e-03,  7.0616e-03,
        -3.4651e-03, -9.7686e-03, -1.9652e-03, -6.6719e-03, -2.4846e-03,
        -1.4272e-03,  1.3105e-04,  5.4709e-04, -3.2662e-03, -3.1589e-03,
        -4.1179e-05, -6.4777e-04, -6.0735e-03,  1.0884e-03,  1.1758e-02,
        -8.3340e-03, -1.2339e-02, -3.3676e-03, -9.1474e-04,  6.2056e-04,
         2.0920e-03, -1.1799e-02, -4.4539e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1847e-02, -8.2087e-01,  3.2089e-02, -2.3571e-03,  1.9566e-03,
         8.4698e-03,  7.0381e-03,  8.4423e-03, -4.6614e-04,  9.3961e-03,
        -2.5799e-03,  2.6181e-02, -1.8302e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.7774e-02,  7.8384e-01,  6.9306e-03, -2.6072e-02,  5.8723e-03,
         8.1727e-03,  1.0325e-02,  8.9613e-03, -1.6548e-04,  6.4553e-03,
         5.4228e-05, -2.1210e-02,  3.5435e-03,  6.8841e-03, -2.3743e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5310e-02,  7.7836e-01,  1.8682e-02, -1.3061e-02,  7.5821e-04,
         1.2141e-02,  2.8737e-03,  3.9423e-03,  1.1015e-03,  5.3019e-03,
         6.7078e-03,  6.3595e-03, -5.9405e-03,  3.2419e-03,  4.2190e-03,
         6.0549e-03, -1.2024e-03, -6.3558e-03, -1.1275e-03, -6.2625e-03,
        -2.7880e-03,  1.3197e-03,  1.6961e-03,  2.5547e-03,  4.9313e-03,
         7.1292e-04,  6.3312e-04,  4.6886e-03, -7.2862e-03,  9.2596e-04,
        -4.2760e-03,  2.3268e-03, -5.5690e-03,  1.1053e-04,  8.6068e-03,
         3.2052e-03,  3.4274e-04,  1.6306e-03, -1.8031e-03,  9.0392e-04,
         1.8131e-03, -5.5545e-03,  1.6409e-03,  9.3007e-05,  3.3348e-04,
        -4.9858e-03, -2.0268e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5664e-02, -8.7049e-01,  1.7810e-02,  5.4905e-03, -2.5764e-03,
         5.1919e-04,  3.4186e-03, -1.3706e-03, -9.4899e-03,  1.9242e-04,
        -1.9684e-03, -6.3672e-04, -3.5139e-03,  5.7644e-03, -9.7342e-04,
        -2.2790e-03, -4.1627e-04, -5.2594e-04, -2.1613e-03,  1.4652e-03,
        -1.6499e-03,  4.7734e-03, -1.8497e-03,  1.5522e-03, -3.0153e-03,
        -7.1773e-04, -7.4132e-03, -7.9844e-04,  2.9326e-03,  7.8970e-04,
         2.8402e-03,  1.9666e-03, -1.4436e-03,  6.0771e-05, -1.2358e-03,
        -5.1378e-04, -1.6863e-03, -2.6782e-03, -7.6624e-04,  3.9062e-04,
         1.6157e-03,  2.5842e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0020,  0.7848,  0.0369,  0.0126,  0.0166,  0.0015,  0.0211,  0.0072,
         0.0054, -0.0153, -0.0013,  0.0119,  0.0014,  0.0056,  0.0088, -0.0032,
         0.0317,  0.0064,  0.0041,  0.0032, -0.0070,  0.0027, -0.0012, -0.0065,
         0.0017,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6305e-02,  8.1635e-01,  3.0700e-02, -1.7017e-03,  6.5138e-03,
        -3.5388e-03,  4.1468e-02,  2.2494e-03,  7.0017e-03, -1.2193e-04,
         7.7334e-04,  7.4948e-04, -2.0841e-03,  1.9448e-03, -6.4811e-03,
        -7.7750e-03,  1.4305e-02,  1.1357e-02, -2.1893e-03,  3.4588e-03,
        -6.1737e-03, -5.1614e-03, -1.5965e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0105, -0.8465, -0.0242, -0.0014,  0.0054, -0.0038, -0.0065, -0.0019,
        -0.0032, -0.0156, -0.0130, -0.0019, -0.0077,  0.0032,  0.0096,  0.0035,
         0.0092, -0.0012, -0.0090, -0.0044, -0.0037,  0.0036,  0.0112,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3811e-02,  8.4343e-01, -4.9104e-03, -2.0902e-02, -6.1184e-03,
         1.4618e-04,  2.8768e-02,  2.8037e-02,  3.5641e-03, -3.8132e-03,
        -6.8588e-05, -5.1044e-03, -1.3238e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0161, -0.8581,  0.0208,  0.0089,  0.0066,  0.0080,  0.0172, -0.0056,
        -0.0104,  0.0060, -0.0119, -0.0108,  0.0131, -0.0066,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7058e-02, -8.7554e-01, -9.8005e-03,  1.3383e-02, -1.3886e-02,
         1.4633e-03,  3.7363e-03, -4.9154e-03, -6.5469e-04,  4.4811e-03,
         1.0561e-02,  4.3854e-03,  1.4645e-03,  4.5479e-03, -1.7069e-03,
         8.0088e-04,  7.5998e-03,  1.0137e-02, -3.8810e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 1.3699e-02,  8.4444e-01,  8.4157e-04, -2.0783e-02,  8.5692e-03,
         2.0255e-02,  1.6631e-03,  1.4796e-02,  3.6549e-03,  1.7073e-02,
         1.3315e-03, -2.7805e-03, -8.4288e-04,  2.5559e-03,  9.7404e-04,
        -1.1237e-02,  5.6940e-03,  2.7850e-03,  2.2260e-03,  2.9662e-03,
        -3.0671e-03, -1.2873e-03, -3.2205e-03,  5.8897e-03, -1.9769e-03,
        -5.3923e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5753e-02,  7.7564e-01, -2.8124e-02,  7.4875e-04,  2.5973e-02,
        -8.3046e-03, -3.5048e-03, -7.0342e-03, -5.8319e-03, -7.3857e-02,
        -2.2241e-03,  6.8288e-04,  1.5433e-02,  1.3209e-02,  5.7337e-04,
        -3.1117e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7674e-02,  8.5367e-01,  2.0344e-02,  5.2828e-02,  1.5149e-03,
         6.9837e-03,  2.7729e-03,  3.2679e-03, -3.1289e-03,  3.5111e-04,
        -7.5772e-03,  3.4239e-03,  5.1525e-03,  8.2111e-03, -3.4411e-03,
         7.1629e-03,  2.4920e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0945,  0.8237,  0.0024,  0.0080, -0.0297,  0.0168, -0.0068,  0.0046,
         0.0135,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.7257e-02, -7.2171e-01,  2.9359e-03,  1.7298e-02,  3.2391e-02,
        -7.6553e-03, -8.7997e-03, -6.1122e-03, -1.1665e-02, -2.1931e-03,
        -2.4139e-02, -7.8646e-03, -1.4794e-03, -3.1582e-03, -8.8179e-04,
         4.5861e-03, -4.5896e-03,  4.1902e-03,  5.0454e-03,  4.3341e-03,
         1.0478e-03, -3.4128e-03,  8.1840e-04, -9.8757e-04,  2.8501e-03,
         6.5748e-04,  1.1571e-02, -1.7894e-03, -3.3837e-03, -9.6657e-04,
        -6.1865e-04, -2.5919e-03, -6.9841e-03, -2.4843e-03,  9.6278e-04,
        -1.3010e-03,  2.1425e-03, -9.2520e-03,  3.2327e-03,  1.1448e-03,
         5.8964e-04, -1.5394e-03,  1.5387e-03, -2.3030e-03, -5.7873e-03,
        -1.7618e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3395e-02,  8.4273e-01, -1.1987e-02, -2.2895e-02,  5.9720e-03,
        -1.0043e-02,  1.4422e-03, -4.9856e-03, -4.2247e-03, -1.1311e-03,
        -5.5516e-03, -1.1869e-03, -3.5425e-03,  1.3922e-03,  3.4650e-03,
         3.2853e-03,  2.3058e-03,  2.0481e-03, -4.5613e-03,  5.2970e-03,
        -7.6723e-03, -1.2310e-03, -8.7111e-03,  1.1027e-04, -1.8000e-03,
        -1.5998e-03,  6.7563e-03, -3.1576e-03, -2.6136e-03,  1.4901e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0171, -0.8816, -0.0141, -0.0449,  0.0166,  0.0027,  0.0229,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4003e-02, -9.1704e-01, -5.0964e-03,  1.6016e-02,  2.1931e-03,
        -3.7363e-03, -3.3121e-03, -1.3674e-02, -2.2855e-05,  2.7136e-03,
        -2.2192e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0375,  0.8499,  0.0356,  0.0129, -0.0040, -0.0117,  0.0101, -0.0032,
         0.0112,  0.0056,  0.0013,  0.0053,  0.0022, -0.0095,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4785e-02,  8.1013e-01, -1.6630e-02, -1.8855e-02, -2.7950e-04,
        -2.9962e-03, -1.3981e-02, -7.5032e-03, -2.6213e-03, -5.4286e-03,
         2.0474e-03,  2.7480e-03, -9.2512e-03, -1.1124e-02,  1.2265e-03,
         2.1083e-03, -1.3455e-04, -1.2680e-03, -2.1353e-03,  1.5463e-02,
         4.7414e-03,  8.1541e-03, -2.0713e-03,  3.1148e-03, -1.4027e-03,
         4.5030e-03,  4.6046e-03,  1.7332e-03, -9.0103e-04, -9.0408e-04,
        -1.0718e-02,  9.8008e-03,  4.3938e-03,  2.2444e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4111e-02,  7.8114e-01,  4.9544e-02,  6.0464e-03,  1.3431e-02,
        -4.0288e-03, -2.4272e-03, -7.8371e-03, -8.2679e-03,  3.9933e-03,
         7.7118e-03, -7.2234e-03, -2.5831e-03, -6.2801e-05, -1.2299e-02,
        -5.3433e-03, -1.1959e-03, -1.1971e-02,  3.9303e-04,  1.0474e-03,
        -2.9735e-03,  3.0020e-03, -5.0164e-03, -2.3205e-04,  4.0783e-03,
        -1.7205e-03, -9.0764e-03, -3.2455e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3296e-02,  8.2800e-01,  1.6826e-02, -9.9168e-03,  3.8268e-04,
        -4.9267e-03,  5.6412e-03,  3.0669e-03, -1.3677e-03,  1.5620e-03,
         5.2300e-03, -6.7061e-03,  1.5762e-03, -1.4083e-02,  9.1076e-03,
        -3.3887e-03, -1.3608e-02, -5.3001e-03,  1.1951e-02,  6.3553e-03,
        -5.0431e-03,  6.5942e-03, -1.4517e-02, -1.1558e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 8.0794e-02,  6.8492e-01, -1.7278e-02, -5.8663e-03,  9.1649e-04,
        -5.0163e-03,  2.6540e-03, -6.0889e-03,  1.2167e-02,  3.9825e-03,
        -1.4868e-02, -1.3409e-02,  5.4666e-03,  6.1195e-03,  5.7748e-04,
        -8.5084e-03, -1.4084e-03, -6.2439e-03, -9.8596e-04, -2.0062e-02,
        -3.7110e-04,  2.1159e-03, -9.3407e-03,  1.0148e-02, -2.3585e-04,
         7.3726e-03, -1.0091e-02, -6.1827e-04,  1.6685e-03, -3.8193e-02,
         2.2511e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.7996e-02,  8.1730e-01, -2.1268e-02, -6.1156e-03,  1.8811e-02,
        -1.1124e-02, -3.1705e-03,  1.5416e-03,  1.8661e-02, -1.7615e-03,
        -1.7644e-03,  4.9148e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3044e-02,  7.9651e-01,  1.8900e-02,  7.3713e-02,  6.8149e-03,
        -1.4027e-04, -3.6769e-03, -1.3420e-03,  3.8780e-03, -7.0395e-03,
        -1.0871e-02, -2.0870e-02,  6.3247e-03, -6.5334e-03,  3.4658e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3259e-02,  8.2329e-01, -8.5073e-03,  8.6124e-03,  1.6676e-02,
         2.3248e-03,  2.4563e-02,  1.1855e-02, -3.3853e-03,  7.2681e-03,
         5.3798e-04,  7.2286e-03,  2.0394e-03,  7.7501e-04, -1.2341e-03,
        -2.1365e-03, -5.1876e-03,  1.6622e-03,  1.0455e-02, -6.0811e-03,
        -8.4524e-05, -2.1659e-03,  1.2008e-03, -1.9396e-06, -5.6164e-04,
         8.2582e-04,  5.4801e-03,  1.6729e-03, -8.4619e-04,  2.5681e-03,
        -1.1090e-03, -2.9591e-03,  2.3445e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0099,  0.8523, -0.0128, -0.0048, -0.0041, -0.0109, -0.0075, -0.0133,
        -0.0028, -0.0009,  0.0015,  0.0104, -0.0150, -0.0321, -0.0148,  0.0069,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0052,  0.8247,  0.0222,  0.0143, -0.0077,  0.0074, -0.0153, -0.0385,
         0.0645,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1063,  0.7955, -0.0905, -0.0025,  0.0051,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6609e-02,  8.1517e-01, -2.1138e-02,  3.8121e-03,  2.7792e-03,
        -5.3249e-03, -2.3930e-03, -8.8143e-03,  1.3037e-02,  1.0740e-02,
         1.6386e-03, -5.3189e-03,  2.4847e-03, -1.7897e-02, -4.0272e-03,
         2.2162e-04,  6.5949e-03,  3.1013e-03, -1.4237e-02,  1.1844e-03,
        -1.8473e-03,  3.5431e-03, -2.2014e-03,  8.1706e-03,  1.5508e-03,
         4.1835e-03, -1.1980e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0264,  0.8580, -0.0024,  0.0018,  0.0362,  0.0485,  0.0266,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2257e-03,  8.0476e-01, -2.7365e-03, -2.0142e-02,  6.1070e-03,
         7.4940e-03,  3.4959e-02, -9.3388e-03, -1.2165e-02,  7.0025e-04,
         3.1624e-04,  1.2112e-02, -1.8459e-03,  5.5265e-03,  1.5372e-02,
         4.0218e-03, -2.7696e-03,  4.3645e-03, -1.4400e-02,  1.5133e-02,
         2.1317e-02,  1.9626e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0305,  0.8529,  0.0019,  0.0035,  0.0060,  0.0034, -0.0041,  0.0211,
        -0.0055, -0.0046,  0.0080,  0.0058, -0.0062,  0.0138, -0.0023,  0.0061,
        -0.0020, -0.0038, -0.0184,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9468e-03, -8.4007e-01, -1.3402e-02,  8.7352e-03,  1.1445e-02,
         1.3894e-02, -1.7684e-03, -1.0977e-02,  5.6078e-04,  2.9838e-03,
        -1.5500e-03,  8.7496e-04,  1.4262e-03,  6.1030e-04,  4.9328e-03,
        -6.2137e-04, -6.5235e-04, -1.7390e-03,  3.9417e-03, -4.1859e-04,
        -4.9412e-03, -5.1431e-03,  6.7316e-04,  1.7203e-03, -7.3134e-03,
         3.2974e-03,  4.8446e-03,  7.2068e-04, -1.1189e-02,  8.9577e-03,
         2.1649e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-2.2499e-02,  7.8077e-01, -4.2786e-02, -1.1936e-02, -1.4819e-02,
        -1.2002e-02,  8.1415e-04, -1.2146e-03,  2.0284e-03,  2.0442e-03,
         4.9340e-03,  2.6802e-03, -7.7051e-03,  7.6489e-03,  6.8409e-03,
         1.4598e-03,  1.8232e-02, -2.9001e-03, -4.8403e-04,  2.2345e-03,
        -1.8945e-03, -4.7824e-03,  2.1225e-03, -6.9223e-03, -6.7136e-03,
        -3.4058e-03,  1.0898e-03, -9.6806e-03, -5.8498e-04,  1.5352e-03,
         1.9629e-03, -1.3272e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0724e-03,  7.9471e-01, -1.2009e-02,  3.6386e-03, -9.5727e-03,
        -3.2972e-03, -1.6491e-02,  1.1297e-02,  5.1731e-02, -7.8187e-04,
        -4.1820e-02,  5.6113e-03, -4.6318e-03,  8.9815e-03,  2.3312e-03,
        -8.7226e-03, -3.5423e-03, -1.5786e-03,  2.6739e-03,  1.3415e-03,
         6.9808e-03,  6.1814e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6590e-03,  8.4916e-01,  2.3244e-02,  3.7223e-03, -2.2914e-03,
         3.3771e-03, -1.7421e-02,  4.0377e-03,  3.6161e-04,  6.6856e-04,
         8.2266e-03,  3.6791e-03,  1.1170e-02,  4.4351e-03,  7.3301e-04,
         9.5614e-04, -2.9980e-03,  5.0383e-03, -5.7939e-03, -1.3353e-04,
        -2.5880e-03, -4.0667e-03, -3.0869e-03, -4.3188e-03, -5.2138e-03,
        -6.3584e-03, -1.2919e-03,  2.1323e-03,  1.0100e-02,  1.0733e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.8683e-02,  7.6401e-01, -5.1911e-02, -2.6869e-02, -2.2359e-03,
        -8.1876e-03,  2.9538e-03,  9.3179e-04, -1.3080e-02,  4.9743e-03,
        -1.1533e-02,  6.0483e-03, -1.1791e-02,  1.6602e-04,  1.4912e-03,
        -2.0556e-03, -2.4855e-02, -1.8220e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.9625e-03,  8.0705e-01,  8.3592e-03,  2.0909e-02,  6.7335e-03,
        -2.0004e-03,  5.8859e-04,  1.0212e-02,  1.4102e-02,  3.6435e-03,
        -3.0431e-04, -1.1587e-03,  1.2635e-04,  2.9737e-03, -3.9388e-03,
        -7.4653e-04,  1.8483e-03,  2.0725e-02, -9.8092e-03,  1.0842e-03,
        -1.3214e-03,  1.8386e-03,  4.0424e-03,  1.4463e-02,  6.3290e-03,
        -4.9060e-04,  3.6352e-03, -9.1507e-04, -2.1616e-04,  1.2625e-03,
        -1.7189e-03,  2.5926e-03,  2.6194e-03,  6.8569e-03, -1.9602e-04,
        -6.4604e-03, -2.0763e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.1413e-02,  7.4548e-01, -2.8426e-02,  1.0261e-02,  1.1667e-02,
        -4.4530e-03, -2.8618e-03, -2.0892e-03,  2.5000e-04, -3.1947e-03,
        -3.2844e-03,  1.0124e-02,  1.1558e-03, -1.1281e-02, -1.0249e-02,
        -4.1777e-03,  4.6508e-03, -5.6059e-03, -2.1727e-03,  7.8095e-03,
         1.0331e-02, -2.9013e-03, -3.7377e-02, -8.7835e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2841e-02,  8.0391e-01,  2.7417e-03, -1.5410e-02,  1.3397e-02,
         3.5598e-03, -6.0900e-03,  6.7817e-03,  2.0079e-03, -6.0938e-03,
        -4.0140e-03, -4.7235e-03, -3.2788e-03,  7.5713e-04, -7.4657e-03,
         1.5274e-03, -3.5840e-03, -3.6491e-03, -2.0249e-04, -5.5565e-03,
        -6.9661e-04, -2.3564e-03, -1.3420e-03,  3.2400e-03,  6.0400e-04,
         1.5704e-02, -2.2466e-03,  5.7068e-04,  9.9372e-04, -1.2483e-03,
         8.2292e-03, -5.4417e-03, -8.8743e-04,  2.9455e-04,  3.5954e-03,
        -2.8747e-03, -7.0903e-04, -2.8973e-03, -3.4277e-04,  3.7061e-04,
        -1.0087e-03,  4.3354e-04, -2.1311e-03,  7.0362e-03, -2.9116e-03,
        -4.2394e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.6886e-03,  8.1125e-01, -1.4002e-02, -1.5599e-03, -4.2150e-03,
         8.8142e-03, -2.0336e-02, -1.2285e-03,  6.1576e-03, -1.2860e-02,
         1.4841e-02,  7.9034e-03,  1.3277e-02, -3.7770e-03,  3.7414e-03,
         2.0194e-03,  6.0422e-04,  1.6461e-02, -7.0548e-03,  3.5999e-03,
         2.4036e-02,  7.8265e-03,  5.7457e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3943e-02,  9.3969e-01, -9.2303e-04, -4.0325e-04,  3.5538e-03,
         1.0935e-02,  5.0263e-03, -2.3893e-03, -5.7305e-03,  2.8046e-03,
         4.6000e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5754e-02,  8.4972e-01,  4.1790e-04,  2.9011e-03, -1.1053e-02,
        -9.2859e-03,  3.7424e-03, -9.3935e-03,  2.2196e-02,  1.2341e-02,
         4.7092e-03, -1.9883e-03,  2.2004e-02, -2.4494e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0100,  0.8675, -0.0104, -0.0333, -0.0149,  0.0014, -0.0143, -0.0086,
         0.0023,  0.0071,  0.0211, -0.0090,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5575e-02, -7.8158e-01, -3.2271e-02, -3.2168e-02, -5.0162e-03,
         1.2910e-02, -9.1630e-03, -1.4438e-02, -2.5296e-03, -1.1396e-03,
         2.5600e-03,  2.6339e-03,  1.1950e-03, -2.3079e-03,  2.9455e-03,
        -2.8916e-04,  3.9821e-04,  1.7289e-03, -2.3388e-03,  2.2581e-03,
         1.1820e-03, -6.0345e-04, -2.1009e-03,  5.3491e-04, -1.3892e-03,
        -1.6325e-03, -1.2011e-02,  1.1581e-03,  3.5818e-03,  3.9697e-03,
         1.1069e-03, -7.9370e-04, -1.1513e-03, -4.0322e-03, -1.2405e-03,
         1.2150e-03,  2.6226e-03,  8.2168e-04,  2.0248e-03, -4.6160e-06,
        -1.1287e-03,  3.5358e-04, -9.5939e-04, -3.0034e-04,  2.7986e-03,
         4.7116e-04, -2.6576e-03,  3.1664e-03,  3.1686e-03, -1.0020e-04,
        -1.1998e-03,  1.8616e-03,  4.5905e-03,  7.1602e-04, -7.9010e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-3.9252e-02,  8.0615e-01,  2.3378e-02, -2.2330e-02,  1.3627e-02,
         6.4617e-03,  3.3660e-04, -2.6779e-03, -1.5661e-02, -9.0993e-04,
         2.9376e-03, -1.9443e-03, -9.3076e-04, -7.9548e-03, -5.0873e-04,
         1.9849e-04, -3.6355e-03,  2.4046e-02,  6.7942e-03,  1.1547e-02,
        -3.1461e-03, -2.6088e-03,  2.9592e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0643, -0.7817,  0.0233, -0.0125, -0.0025,  0.0095,  0.0022, -0.0019,
         0.0145,  0.0101,  0.0017, -0.0028,  0.0085, -0.0038, -0.0034,  0.0122,
         0.0409,  0.0043,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.0191e-02,  7.7452e-01, -9.3928e-03, -1.1894e-02,  2.1253e-02,
        -2.6611e-03,  5.8681e-03,  2.2166e-02, -6.7503e-03,  4.1051e-03,
         4.8790e-03, -6.0929e-03, -4.5702e-03,  5.0372e-03, -5.0877e-04,
         1.2667e-02, -2.6405e-03,  1.4799e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3516e-03, -8.8212e-01,  4.0342e-02,  8.6692e-03, -1.0396e-02,
         5.3486e-03,  1.8088e-04,  7.1708e-03,  7.6007e-03,  6.4058e-04,
        -3.5333e-03, -5.2619e-04, -8.8472e-03, -2.0925e-02,  2.3503e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5269e-02,  8.3669e-01, -1.2807e-02,  5.4648e-03,  1.4138e-03,
        -2.1234e-03,  2.1616e-03, -3.8640e-04,  3.1901e-03, -4.0203e-04,
         3.0820e-03,  3.8862e-04,  5.6489e-04,  3.5494e-04, -1.8522e-03,
         1.1296e-02, -8.0036e-03, -4.8208e-03,  1.6466e-05, -1.1121e-03,
         3.2535e-03, -3.6320e-03,  3.0335e-03, -1.8236e-03,  1.0065e-02,
         1.3530e-03,  1.2983e-02,  7.9204e-04, -1.3574e-03,  5.6156e-04,
         5.3863e-03, -4.3613e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7482e-02,  7.9635e-01,  1.8183e-02,  8.1492e-04,  4.7178e-03,
         1.2150e-03, -2.1052e-03,  7.1157e-03, -1.6014e-02, -5.9348e-03,
         2.3412e-05, -3.8517e-03, -6.0056e-03, -1.2203e-02, -9.9824e-04,
         2.1008e-03, -6.0116e-03, -6.6699e-03,  3.3051e-03, -6.0876e-04,
        -4.6770e-03,  2.9058e-03, -7.1681e-03,  5.6819e-03, -1.8739e-03,
         5.0445e-03, -1.9454e-03, -2.4073e-03,  9.1431e-04, -2.0798e-03,
         4.5073e-06,  7.4187e-03,  2.6172e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5868e-03,  8.3134e-01,  2.4545e-02,  1.6424e-02,  1.9045e-02,
         7.1635e-03,  2.7177e-03, -6.5044e-03,  9.9075e-03,  1.1650e-04,
        -2.5200e-03, -8.6685e-03,  3.2126e-03,  2.7740e-04,  3.6311e-03,
         2.0006e-03,  8.7066e-04,  2.8839e-04, -3.5063e-03,  5.1323e-04,
        -6.3427e-04, -3.1492e-03, -1.3581e-03, -1.3145e-03,  8.8074e-04,
         3.6644e-04, -8.0036e-04, -1.4340e-04, -4.7746e-04,  2.3754e-03,
         2.4364e-04,  1.9147e-04, -3.2221e-04,  2.3743e-03, -2.4056e-03,
        -2.4627e-03, -3.6577e-03, -2.7733e-03, -1.0366e-03,  5.9690e-04,
         2.6992e-04,  1.7473e-03, -1.0929e-03, -2.0503e-03,  2.3097e-03,
         7.3965e-04, -7.0559e-04, -2.1125e-03, -1.1650e-03, -3.3380e-03,
         2.9247e-03, -3.6129e-05, -7.0989e-04, -1.9854e-03,  1.0258e-03,
        -2.5859e-03,  1.2795e-03, -5.1973e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7147e-02,  8.2079e-01,  1.1968e-02,  1.1430e-02, -1.0757e-02,
         5.4354e-03, -2.9066e-03, -1.0097e-02, -4.5059e-03,  1.3080e-02,
        -1.0737e-02,  6.0232e-04,  1.8884e-02, -2.2693e-03, -1.5220e-03,
        -1.7588e-03,  5.1864e-03,  6.2923e-05, -4.0764e-03, -4.9524e-03,
         1.3189e-03,  1.2423e-03, -3.2639e-03, -1.9019e-04, -1.3701e-02,
        -1.2120e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0687, -0.7904,  0.0148,  0.0204,  0.0063,  0.0095, -0.0030,  0.0031,
         0.0032,  0.0048,  0.0012, -0.0024,  0.0046, -0.0028,  0.0050, -0.0011,
        -0.0014, -0.0009, -0.0023, -0.0021,  0.0035,  0.0111, -0.0039, -0.0016,
         0.0319,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9513e-02,  7.8461e-01,  1.5103e-02, -1.1468e-03, -2.2977e-02,
        -8.2971e-03, -8.2238e-04, -1.4938e-03,  4.5662e-03,  6.0916e-03,
         3.7948e-03, -7.0464e-03, -4.1714e-03,  8.7717e-03, -1.4586e-03,
         1.2110e-03, -4.1290e-03,  7.9270e-03, -9.1655e-04, -1.4206e-03,
         1.4932e-03,  5.3768e-03, -3.7212e-03, -5.7004e-03,  8.5786e-04,
        -1.2291e-02, -5.9361e-03, -2.2065e-04, -1.0891e-03, -2.8156e-03,
         5.2699e-04, -5.5854e-03, -3.5393e-03,  1.0400e-03,  7.1691e-04,
        -2.9243e-03, -2.8148e-03, -5.9205e-03, -2.0602e-03,  5.5254e-04,
        -3.8067e-03, -2.7195e-03,  4.1195e-03, -2.8089e-04,  8.4246e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0160e-02,  8.2106e-01,  1.4754e-02,  5.6181e-03,  9.1449e-03,
        -4.8997e-03,  6.2615e-03,  5.5009e-03, -2.2305e-02, -2.5356e-03,
        -1.3216e-03, -3.7922e-03,  3.9610e-03,  3.5830e-03,  1.8455e-03,
        -1.9496e-02,  7.9338e-04,  6.2038e-03,  7.1520e-03,  7.9402e-03,
         1.8675e-03,  1.2211e-02,  7.5947e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0453,  0.8440, -0.0029,  0.0138, -0.0034,  0.0040, -0.0371,  0.0054,
         0.0115,  0.0016, -0.0010,  0.0299,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 2.3648e-02, -8.0387e-01,  1.3541e-02,  1.3681e-02,  7.2772e-03,
        -6.0754e-03,  5.1745e-03,  1.1126e-02, -2.9720e-03,  1.4386e-02,
         2.4851e-02,  1.0482e-03,  2.0703e-03, -3.0949e-03,  6.3476e-03,
         1.0492e-02, -2.1384e-03, -1.1611e-02, -2.1730e-05, -4.4576e-03,
         4.2992e-03,  1.4001e-02,  1.3812e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.2320e-02,  7.8183e-01,  8.9992e-03, -1.1963e-02,  1.4538e-04,
        -9.4769e-03, -4.4719e-03,  1.2070e-02, -1.6644e-03, -9.7013e-04,
         4.2842e-04, -3.9529e-03, -2.2962e-03,  9.7710e-04, -2.1315e-03,
        -2.8578e-03, -6.3520e-03, -2.9922e-02, -2.0329e-03, -3.8408e-04,
         4.6650e-03, -4.2949e-03, -1.2248e-03, -4.9387e-03, -3.4577e-03,
        -1.8746e-03, -6.7927e-03, -4.0768e-04, -8.6327e-03, -6.9206e-04,
        -1.5371e-03,  5.1204e-03,  1.1183e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0037,  0.8953,  0.0051, -0.0159, -0.0117,  0.0014, -0.0093,  0.0014,
         0.0009,  0.0035, -0.0009, -0.0068,  0.0030, -0.0051,  0.0043, -0.0077,
        -0.0089,  0.0073, -0.0078,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0883e-02,  8.1174e-01, -4.2744e-02, -1.3035e-02, -4.0006e-03,
         3.1184e-02,  4.1814e-03, -1.1331e-02,  5.9401e-04, -5.4532e-03,
        -4.0885e-03, -6.4111e-03,  1.1904e-02,  1.2446e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1720e-02, -8.7009e-01, -7.6364e-03,  1.0209e-03, -5.6374e-04,
        -9.1191e-03, -7.3820e-03, -4.9970e-03, -8.0601e-03, -8.2358e-04,
        -1.9577e-03, -2.4114e-03, -1.9767e-03, -1.6826e-02, -4.7586e-03,
         3.6103e-03, -8.4737e-03, -2.1869e-03, -4.6803e-03,  5.2660e-04,
         4.9184e-03, -6.2623e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1383e-02, -8.3052e-01, -7.4686e-03,  3.8555e-03, -3.2335e-03,
        -1.2299e-02,  6.3102e-03,  2.6116e-02,  3.0427e-04,  2.2214e-03,
        -3.1230e-03,  2.2434e-03,  1.1348e-02,  4.6058e-03, -7.2966e-03,
        -3.9406e-03, -5.7826e-03, -5.8614e-03,  3.0409e-03, -1.2878e-02,
        -3.1888e-03, -1.2741e-03, -4.0886e-03,  6.5540e-03,  2.2087e-03,
        -3.4072e-04, -1.2825e-03, -1.1574e-03, -1.2876e-02, -3.1950e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.5137e-02,  7.3097e-01, -1.7213e-02, -3.0328e-02, -1.1854e-03,
         2.1975e-02,  1.3139e-02,  5.1964e-04,  4.7212e-03,  4.6460e-03,
        -8.3173e-03,  2.5110e-03, -8.4390e-03,  4.9424e-03,  1.4168e-02,
        -2.7481e-02, -1.4310e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0093,  0.8831,  0.0086, -0.0078, -0.0107, -0.0042,  0.0079,  0.0109,
         0.0431,  0.0045, -0.0099,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4293e-03, -8.2840e-01, -1.0305e-03,  6.8679e-04,  2.3850e-02,
         8.1142e-03, -4.3214e-03,  2.8318e-03,  4.0335e-03, -3.7698e-03,
         1.2706e-02, -5.8572e-06, -1.4099e-03, -9.8488e-03, -1.8503e-02,
        -4.5358e-02,  3.0705e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8706e-02, -8.4256e-01, -1.9492e-02, -2.5386e-02, -1.8399e-04,
         9.3199e-03, -1.3269e-03,  2.3312e-03,  7.0178e-03,  6.7506e-03,
         1.5537e-03,  2.4149e-03,  9.4975e-06, -9.4124e-03, -1.4963e-03,
        -2.7238e-03,  5.3566e-04,  5.4484e-03,  6.2542e-03,  1.6914e-03,
        -1.1948e-03,  2.4578e-04, -7.0441e-03, -4.1513e-04, -8.1434e-03,
         1.8345e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7082e-02,  8.0352e-01,  1.7323e-02, -1.0511e-02, -4.1189e-03,
         2.0535e-02, -1.1418e-03,  7.3311e-03,  1.1994e-02, -2.4123e-03,
        -2.1230e-03, -1.6392e-02, -2.3283e-03,  6.5162e-03, -5.8221e-03,
        -2.7051e-03, -1.0884e-04, -2.8783e-03, -6.9122e-04,  2.2865e-03,
        -3.1385e-03,  7.9175e-04, -4.7923e-03, -7.7391e-03,  9.2364e-03,
         1.9805e-03, -2.8570e-03, -5.3353e-04, -3.9797e-03,  2.7127e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.9704e-02, -7.6637e-01,  3.6955e-03, -2.5320e-03, -4.5863e-02,
         2.3395e-03,  1.1701e-02,  6.3430e-03, -2.8565e-03,  7.8659e-03,
         1.3140e-04,  9.8863e-04, -1.9088e-03,  8.0375e-03,  4.6542e-03,
         2.1459e-03,  1.0655e-02, -3.3489e-03, -6.8856e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-2.0308e-02, -8.1148e-01,  4.3754e-03, -5.2371e-02, -2.7490e-03,
         5.5923e-03, -1.4699e-02,  2.7440e-03, -3.4590e-03,  1.5717e-03,
        -5.8303e-03,  2.6470e-03, -2.5677e-03, -6.4138e-04,  1.1253e-02,
        -5.3302e-03, -3.4827e-03, -3.7345e-03, -3.5973e-03, -1.5607e-03,
         6.9112e-03,  1.6263e-03, -1.2224e-03,  5.9670e-03,  1.4120e-02,
        -6.9912e-04,  9.4618e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8774e-03, -8.1634e-01, -1.8017e-02, -6.9047e-04, -1.7749e-04,
        -1.3414e-03,  2.2240e-03, -8.0673e-03, -2.2283e-02, -7.1905e-03,
         4.7036e-03, -5.2730e-02,  3.9731e-02,  2.4623e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3926e-03, -8.5474e-01,  1.3380e-02, -1.3131e-02, -5.9822e-03,
        -4.5464e-03, -7.2691e-03,  7.8843e-03, -2.1942e-02,  1.8422e-03,
         7.9729e-04, -7.7231e-04, -5.6659e-03, -1.9045e-03, -5.2990e-03,
         8.4171e-04, -4.5682e-04,  1.5018e-02, -1.0924e-03, -3.1890e-03,
         1.1358e-04,  4.2927e-03, -2.7103e-03, -5.8591e-03,  1.3714e-04,
         1.7744e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0318,  0.8251,  0.0153,  0.0042,  0.0054, -0.0171,  0.0090,  0.0168,
        -0.0031, -0.0123, -0.0011, -0.0112,  0.0369, -0.0107,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1761e-02,  8.3266e-01,  1.2171e-02, -9.9478e-03, -1.4603e-02,
        -5.8354e-03, -2.9876e-03, -7.5646e-03, -2.1084e-03, -1.2148e-03,
        -3.3905e-03, -1.1123e-02, -2.9828e-03,  6.6122e-03,  8.9101e-03,
         1.2421e-03,  6.0359e-03,  4.0995e-03, -3.2101e-03,  1.3678e-03,
        -1.6191e-03,  3.3781e-04,  1.6874e-03,  2.8912e-03,  3.0391e-04,
         1.3318e-03, -2.5604e-03, -1.3605e-03, -5.0103e-04,  1.5754e-03,
        -1.5158e-02,  2.0850e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3849e-02,  7.9997e-01,  1.9615e-02, -2.0964e-02, -6.0598e-03,
         1.2807e-02, -1.9409e-02, -5.1036e-03,  5.5948e-03, -3.2796e-03,
         4.6893e-04, -1.7833e-05,  9.4017e-04,  2.8401e-03,  7.5677e-03,
        -1.4805e-03,  9.0180e-04,  9.3672e-03, -2.2097e-03,  7.5912e-03,
        -2.2415e-03,  2.3421e-03,  3.5705e-03,  1.1825e-03, -5.0630e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0026,  0.9064,  0.0330, -0.0228, -0.0103, -0.0249,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.0493e-03,  7.8868e-01,  4.0713e-02, -1.0670e-03,  1.9433e-02,
        -3.0604e-03,  6.9064e-03,  4.2851e-03,  6.7357e-04, -1.0613e-02,
        -4.4591e-04,  2.6110e-02, -6.5725e-04, -5.5164e-03,  1.8484e-03,
        -4.0803e-03, -4.6994e-03,  5.7682e-03, -1.8940e-03,  1.2697e-04,
        -6.4018e-03, -8.5628e-03, -1.7493e-03, -3.0801e-03, -1.4938e-03,
        -3.0243e-03,  1.7456e-04, -3.5500e-03, -2.0041e-03,  1.5403e-02,
         2.2759e-03,  5.5909e-03,  3.3733e-03,  9.5696e-04, -5.2122e-03,
         3.6425e-03,  1.8737e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4463e-02, -8.2184e-01,  3.4012e-03,  2.2302e-02,  2.0324e-03,
        -5.9210e-03, -7.8772e-03, -1.2789e-03, -3.5392e-03,  7.0321e-03,
        -7.3382e-03, -5.0215e-03,  7.7590e-03, -3.4960e-03,  3.7571e-03,
         8.7539e-03,  4.7403e-03, -2.8140e-03, -3.4757e-03, -9.8805e-04,
         8.8922e-03,  2.9805e-03,  2.5024e-04,  2.2891e-04,  1.8196e-03,
        -4.1443e-04, -1.2602e-04,  2.8126e-03,  4.0675e-03,  1.4037e-03,
         1.8289e-03,  1.3480e-03,  5.5541e-05, -7.1300e-03,  8.8095e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1493e-02,  7.2855e-01, -1.3726e-01,  1.6085e-02, -5.5838e-04,
         2.7886e-03,  2.8312e-03, -4.8275e-03,  7.4491e-03,  3.0099e-03,
        -1.9549e-02, -1.5743e-02, -1.5017e-02,  1.4630e-02, -1.0207e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0039e-02, -8.2492e-01, -4.3189e-03, -3.5269e-03,  7.7574e-03,
        -1.4574e-02,  1.3243e-03, -2.3606e-04, -3.0307e-03,  2.0933e-03,
         1.7944e-03,  3.5389e-03,  8.7449e-04,  1.3518e-04, -3.9488e-03,
         8.2750e-04, -2.8859e-03, -1.7790e-03, -1.9177e-02,  1.2567e-03,
        -3.7269e-04, -8.9488e-03,  6.0463e-03,  1.0620e-03,  6.4980e-03,
        -5.1784e-03, -3.1173e-05, -1.2832e-03,  1.8592e-03, -3.4773e-03,
        -2.4353e-03, -1.8575e-02,  6.1978e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3675e-02, -7.8891e-01,  1.0412e-02,  1.2440e-02, -1.3050e-02,
         1.1193e-02,  2.9292e-03, -2.3826e-03,  6.5803e-03,  2.7373e-03,
         8.2125e-03, -1.3862e-02, -6.7600e-03, -2.9498e-03,  9.4365e-03,
         3.4985e-03,  3.5769e-03,  7.4360e-04,  9.2223e-03,  2.0121e-03,
         4.1141e-03,  3.4939e-03, -8.1540e-03, -1.6186e-03, -3.1834e-03,
        -4.1378e-03,  1.1261e-02,  1.6956e-02,  2.4987e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 1.5710e-02,  8.7415e-01,  5.0674e-03, -1.7188e-02,  3.6171e-05,
         8.3791e-03,  1.0336e-02,  9.4812e-03,  1.0772e-02, -1.5396e-02,
         1.3602e-03,  4.4837e-03,  3.5719e-03, -1.9049e-02, -5.0146e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2401e-02,  8.2287e-01, -6.8681e-03, -5.5524e-03, -7.9258e-03,
        -7.7523e-04,  7.4513e-04,  1.0796e-03,  8.6075e-03, -1.9324e-03,
         7.3079e-03, -2.0125e-03,  4.2442e-04, -2.7128e-03,  1.4805e-03,
        -6.9116e-04,  1.2002e-03, -9.3940e-05,  1.9338e-03,  6.2970e-05,
        -2.8903e-03, -8.6953e-04, -1.8533e-03,  8.8503e-04,  3.7430e-04,
        -2.9445e-04,  5.3090e-04,  4.9632e-03,  1.0024e-03,  1.9290e-03,
         6.7096e-04, -1.4627e-03, -2.3344e-03, -6.8885e-04, -6.2799e-03,
         1.7741e-03,  3.5505e-03,  3.2932e-04,  2.5659e-03, -1.8984e-03,
        -4.6254e-03,  4.2386e-03, -5.1752e-03, -4.1581e-03, -2.4751e-03,
        -5.3203e-03, -2.5084e-03,  1.4461e-03,  1.6227e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.4979e-02,  6.9453e-01,  1.0311e-01, -6.9300e-03,  7.5683e-03,
         1.9976e-03, -1.5438e-03,  1.4248e-03, -1.4475e-02, -4.7105e-03,
        -1.4761e-03,  3.6502e-04,  3.4128e-03,  1.2458e-03,  1.3028e-03,
        -1.5389e-03,  1.9115e-03,  1.0470e-03,  1.0196e-03,  5.1791e-03,
        -1.4198e-03,  6.3359e-03, -4.7728e-03,  1.4254e-03, -2.4220e-03,
        -1.2879e-03, -5.8336e-03,  7.6994e-03,  4.8612e-03, -7.7595e-03,
        -3.0275e-04,  5.7444e-03,  7.2532e-04,  1.7373e-03, -1.7901e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0131, -0.8596, -0.0151,  0.0228,  0.0046, -0.0135, -0.0017, -0.0120,
        -0.0034,  0.0023,  0.0059, -0.0016,  0.0025, -0.0027, -0.0122, -0.0036,
        -0.0078, -0.0074, -0.0082,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3762e-02, -8.3892e-01,  1.0509e-02, -1.7254e-02, -5.3287e-03,
        -1.4590e-02,  8.5953e-03,  2.3049e-03,  5.6377e-03,  3.6568e-04,
         4.3148e-04, -1.0540e-02, -8.0200e-03,  3.5271e-02,  8.4683e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.8348e-04,  8.9646e-01, -1.9091e-03, -9.1487e-03, -7.7349e-03,
        -5.6652e-04, -6.6772e-03,  4.0431e-04,  9.9314e-03,  4.3290e-03,
         4.0850e-03,  3.2029e-03,  4.1511e-03,  5.1016e-04,  3.7219e-03,
         4.3521e-03, -1.5643e-03, -1.0901e-02, -1.6577e-03,  9.2011e-04,
         1.9529e-02, -7.3566e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4895e-02,  8.1683e-01, -5.0838e-02, -1.1878e-02, -2.0286e-03,
         1.6554e-03, -1.2182e-02, -4.8490e-06,  2.0284e-04,  6.2402e-03,
        -4.1191e-03,  4.8625e-03, -7.4476e-04,  3.5413e-03, -2.3582e-03,
        -3.8586e-04, -1.2253e-03,  3.8075e-03, -7.5650e-03,  2.6033e-03,
        -5.1317e-03,  2.6419e-03,  5.5123e-03,  5.6814e-03,  1.1545e-02,
        -2.4251e-04,  7.1387e-04,  9.0948e-03, -4.8801e-03,  6.5846e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.9526e-02,  7.3350e-01, -4.2369e-02, -2.1982e-02, -8.2537e-02,
        -2.0670e-02, -9.4696e-03, -7.9969e-04, -3.8710e-03, -5.1488e-04,
         4.7575e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.6229e-03, -8.7033e-01,  5.1570e-05,  1.5631e-02,  5.0807e-03,
         8.0234e-04,  2.7893e-03,  4.2215e-03,  7.9226e-03, -1.9956e-03,
        -3.5876e-03, -6.6712e-03,  4.4790e-03, -2.2664e-03, -3.4756e-03,
        -3.0663e-03,  8.3357e-03,  9.5202e-03,  8.3464e-04,  7.2979e-03,
         2.5295e-02,  9.7246e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5037e-02, -8.0447e-01, -1.6792e-02,  7.4605e-04, -1.2100e-02,
         1.1235e-02, -3.4400e-03,  5.8809e-03,  1.1209e-02,  6.2958e-03,
         4.4669e-04, -6.5213e-03, -2.0869e-03,  4.8266e-03,  3.8569e-04,
         9.0234e-03,  9.7177e-04, -1.3256e-03,  4.3649e-03,  2.9172e-03,
         2.2560e-03,  3.7641e-03,  5.8039e-03,  3.6002e-03, -7.9445e-04,
         1.6616e-03, -3.0850e-03,  9.1674e-04,  9.3466e-03,  3.7167e-03,
         2.6493e-03,  5.0048e-03, -1.7325e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5912e-02, -7.7975e-01, -4.1255e-02,  9.5620e-03, -6.2868e-04,
        -1.9692e-04,  2.9504e-03,  3.4977e-03,  2.1336e-03,  4.8677e-03,
         8.4829e-03, -5.0215e-03, -1.4564e-03,  7.2211e-03,  3.6764e-03,
        -8.6411e-03, -9.3978e-04,  1.1044e-02,  4.7604e-03,  5.9720e-03,
        -1.0742e-05,  1.4302e-02, -9.9721e-03,  1.0656e-03,  7.3051e-04,
        -5.7949e-03,  1.0734e-03,  4.9086e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2261e-02, -8.6486e-01, -1.5878e-02, -6.8715e-03, -1.2039e-02,
        -1.2921e-03, -8.7711e-03, -3.9847e-03,  1.2783e-03,  1.4378e-03,
         8.2414e-03, -8.0202e-03, -6.3251e-03, -4.3292e-03, -2.3518e-03,
        -8.3981e-04,  6.7484e-04,  5.3977e-03,  1.4336e-03, -5.8680e-03,
         4.5038e-03, -5.6335e-03,  1.1171e-02, -2.9753e-03, -3.5633e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-3.2386e-02, -7.7302e-01, -1.6843e-03, -1.8360e-02,  1.0069e-03,
        -1.4898e-02, -4.6469e-03,  2.7971e-03,  2.2204e-03, -4.8563e-04,
        -9.2838e-03, -7.4877e-03,  1.2831e-03,  3.6124e-03,  6.1596e-04,
         3.5918e-02,  3.6161e-03, -1.2914e-03,  5.6612e-04,  1.5901e-03,
        -4.1669e-03,  4.0592e-03,  4.0895e-03, -7.9932e-04, -2.6607e-03,
         9.0319e-03, -3.2284e-03,  4.1735e-04,  3.9259e-03,  6.9717e-03,
         3.8655e-03,  1.6410e-03,  4.1885e-03,  6.3668e-05,  1.2125e-03,
         5.7078e-03,  9.1823e-04, -1.7928e-03, -2.5339e-03, -2.0502e-03,
        -1.9900e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.8886e-02,  8.2275e-01, -6.4090e-03, -2.5983e-02,  7.1202e-03,
        -3.5389e-03, -7.0114e-04,  1.3086e-03, -2.4190e-03, -6.8976e-04,
        -3.8886e-03,  4.8358e-03,  7.2142e-03, -1.0830e-03,  6.7837e-03,
        -5.4356e-03, -3.3840e-04, -5.3787e-03,  3.8151e-03, -1.8643e-03,
        -3.5915e-03, -6.6401e-03,  3.7756e-04, -1.0649e-02, -1.8294e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1808e-02,  7.5519e-01,  3.5370e-02, -4.3727e-02,  1.3872e-02,
        -5.2255e-03, -6.5411e-03,  4.5147e-03, -1.8116e-03,  3.0041e-03,
         3.5713e-04,  2.6042e-03, -3.5218e-03, -3.4242e-03, -4.7065e-03,
        -3.0796e-03, -7.5197e-03, -1.0587e-03, -1.0221e-02,  4.3283e-02,
         1.9165e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0195,  0.8859, -0.0031, -0.0013, -0.0075,  0.0018, -0.0062, -0.0027,
        -0.0010,  0.0019, -0.0056, -0.0280,  0.0022, -0.0067, -0.0082, -0.0098,
        -0.0015, -0.0072,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2853e-02,  8.2710e-01, -6.3460e-03, -1.9544e-02, -1.5325e-02,
        -1.0577e-03, -4.6953e-03,  1.0569e-04,  1.9029e-03,  3.5836e-03,
         5.4680e-03, -3.7056e-03,  1.7168e-03,  1.4947e-03, -2.0655e-03,
         7.5789e-03, -3.8664e-03,  2.2318e-03,  8.6419e-03,  2.4200e-04,
        -4.9384e-03, -2.3476e-03,  2.7925e-04, -2.2357e-03,  4.9646e-03,
        -2.8305e-03, -5.0435e-03,  6.3565e-03, -3.8758e-03, -4.1358e-03,
         3.7320e-03, -1.4623e-03,  7.3371e-04,  1.2277e-03, -1.9390e-03,
         7.5139e-03,  7.1549e-03,  9.7091e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.7179e-04,  8.0513e-01, -2.3573e-02,  2.3666e-02,  1.5594e-02,
         5.7934e-03,  1.5788e-03,  1.6436e-02,  3.8598e-03, -3.6245e-04,
         1.9591e-03, -4.1652e-03, -3.1570e-03,  1.9182e-02, -1.8648e-03,
         1.4402e-02, -1.8614e-03, -4.1553e-02, -1.1979e-02,  3.0096e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4773e-02,  8.7649e-01, -6.8902e-03, -3.1927e-02,  1.2615e-02,
         1.2505e-03,  2.1972e-03,  3.1592e-03,  4.5070e-03, -2.8162e-03,
         5.1429e-04, -1.3217e-02,  4.7406e-03,  4.9059e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0883e-03, -8.2499e-01, -1.9942e-04,  4.8967e-03, -2.2151e-02,
         1.3391e-03,  1.7804e-02,  4.2431e-03, -3.9667e-03, -5.2454e-04,
         2.3972e-03, -8.8433e-03,  5.0778e-03,  5.9572e-05, -1.8736e-03,
        -1.8489e-03,  3.5846e-04, -9.0234e-03, -6.8630e-03, -1.5908e-03,
         1.4924e-03, -1.8306e-02, -9.7992e-03, -1.5987e-02, -5.9764e-03,
         2.7633e-03, -1.4049e-03,  9.4654e-04,  6.6391e-04, -1.4393e-02,
        -4.5240e-03,  2.6069e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4255e-02, -8.6598e-01,  2.0372e-02,  4.0736e-03,  3.5955e-03,
         1.4340e-02,  2.5993e-04, -6.6004e-04,  1.6457e-03,  9.2545e-04,
         3.0118e-03,  1.1245e-02,  1.9202e-04, -2.4330e-03, -1.5874e-02,
         1.2287e-03,  1.8264e-03, -9.8371e-03,  1.3570e-02,  4.6729e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0935e-02,  7.5483e-01,  1.7278e-02, -1.5489e-02,  9.1218e-03,
        -6.9164e-03, -5.3061e-04, -1.1336e-02, -4.5447e-03,  3.3642e-03,
         1.4945e-03,  6.9149e-03, -4.5459e-04,  3.9371e-03,  5.1631e-04,
        -5.4705e-03, -1.3088e-03, -1.5017e-04, -4.1798e-04, -5.7155e-03,
         4.4623e-03,  1.0496e-02,  4.2004e-03, -8.1904e-03,  1.4607e-03,
        -2.0271e-03,  4.3242e-03, -2.9616e-03,  8.7951e-04, -1.1257e-03,
         7.7076e-03,  4.7543e-03,  8.9866e-03,  6.1373e-03,  2.7984e-03,
         2.4587e-03, -1.4058e-03, -3.5675e-04, -2.3843e-03,  1.1931e-03,
        -6.6164e-03,  4.6898e-03,  9.0892e-04,  7.9471e-04,  1.9232e-03,
         2.0362e-03,  2.3991e-02,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.0110e-04,  8.7934e-01,  8.0305e-03,  7.5928e-03,  2.1776e-02,
         5.1203e-03,  6.8917e-03,  2.9807e-03, -5.2146e-03,  2.9304e-03,
        -1.8053e-02,  2.7811e-03, -2.8460e-03,  1.4823e-02, -2.0922e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8163e-02,  8.0007e-01,  2.9973e-04,  3.9403e-03,  7.3803e-03,
        -6.3903e-05, -6.3564e-03, -9.4906e-04, -6.2483e-03, -3.7113e-03,
        -2.7362e-03, -4.9078e-03,  1.2610e-03,  1.6934e-02, -1.0996e-03,
        -3.3010e-04, -4.1195e-04,  2.5428e-03,  2.7958e-03, -6.9782e-03,
         3.5595e-03, -2.8515e-03,  2.0745e-04,  1.5127e-03,  1.1397e-03,
        -4.2938e-03,  6.4530e-03,  1.5742e-02, -9.9456e-04,  2.9685e-03,
        -6.6500e-03, -2.7190e-03,  3.3415e-03,  3.0624e-03,  2.5232e-03,
         6.5830e-03,  2.2030e-03, -1.1222e-03, -4.9184e-03,  2.2730e-03,
         9.8770e-04,  1.2042e-03, -3.2391e-04,  2.0777e-03, -1.7048e-03,
        -7.0541e-04,  3.4620e-03,  2.7237e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 7.1104e-03,  8.6081e-01, -1.5821e-02, -4.1473e-03,  1.3130e-02,
         1.7510e-02,  3.6529e-03, -2.6285e-04, -1.4590e-02,  5.9766e-03,
         7.1147e-04, -1.6223e-03, -2.6702e-03, -3.9600e-04, -4.5366e-03,
         2.5417e-05, -1.0745e-02,  4.0654e-03, -2.6204e-03, -2.8374e-03,
        -2.2927e-03,  1.9266e-03, -4.0710e-03, -1.0696e-03, -9.3128e-03,
        -8.0890e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0612, -0.7919, -0.0073, -0.0181, -0.0024,  0.0042, -0.0020, -0.0095,
        -0.0085, -0.0037,  0.0199, -0.0027,  0.0061, -0.0015,  0.0063, -0.0034,
        -0.0056,  0.0147,  0.0041, -0.0043,  0.0226,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9742e-03,  8.0458e-01,  1.5442e-02, -4.4536e-02, -3.0722e-03,
        -1.3273e-02, -3.3392e-04, -9.2655e-04, -1.7935e-03, -1.9706e-03,
        -6.4809e-03, -3.5956e-03, -1.5317e-02, -1.4060e-02,  1.8671e-03,
        -2.6078e-03, -6.2281e-04, -1.2848e-02, -4.8873e-03, -1.9763e-03,
        -1.7236e-03,  3.8693e-03,  4.5927e-03,  3.4804e-03,  1.9455e-03,
         1.6008e-03,  3.9078e-04,  1.0341e-03, -1.8590e-03, -1.2302e-03,
        -1.7040e-03,  1.4625e-03,  5.0756e-03, -2.0704e-03, -7.6758e-03,
         4.1168e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9651e-02,  8.2290e-01, -3.3764e-02,  1.7587e-02, -8.5103e-03,
        -1.0508e-02, -1.6336e-02,  5.4455e-03, -1.4151e-04, -7.7080e-03,
         2.1416e-02,  2.5072e-03, -2.3676e-03, -1.3962e-02, -2.4655e-03,
         1.0460e-03, -7.2792e-04, -6.6895e-03, -2.8593e-03, -3.4066e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2871e-02, -8.5039e-01, -1.6976e-02, -5.0827e-03, -2.9655e-02,
        -6.4007e-03,  4.3062e-03,  5.9408e-03,  3.9799e-03, -3.8231e-04,
        -7.3753e-03,  2.7875e-03,  8.8873e-03,  4.5594e-03, -1.4188e-02,
        -4.2610e-03, -9.3943e-05,  2.5309e-03,  1.3610e-02,  5.0961e-03,
        -6.2973e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0312, -0.9021,  0.0015, -0.0123,  0.0066,  0.0036,  0.0010,  0.0038,
        -0.0104, -0.0014,  0.0039, -0.0060, -0.0160,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7326e-03,  8.5256e-01,  3.3637e-03, -1.2987e-02,  2.7229e-03,
        -6.5854e-03, -1.6303e-02, -1.0511e-02, -5.3404e-04,  1.6674e-03,
         5.0889e-03,  3.6215e-03, -7.6694e-03, -1.9421e-03,  5.8882e-03,
         6.8123e-03,  8.5801e-04,  1.6373e-03, -2.9213e-03, -3.8141e-04,
        -2.8150e-03,  2.6638e-03, -6.6049e-03, -1.5230e-03,  5.0732e-04,
        -2.0914e-03, -9.2391e-03,  4.0227e-03,  1.3841e-03,  1.6910e-03,
         3.0065e-03,  1.5077e-03,  5.5000e-03, -1.5344e-03, -7.9705e-04,
         2.1399e-03,  5.1839e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3716e-04,  9.0388e-01, -2.3417e-03, -8.6294e-04, -1.0682e-02,
         8.8825e-06,  3.5667e-03, -3.0498e-03, -8.5884e-03,  3.9744e-04,
        -6.5425e-03,  1.3616e-03,  5.5338e-03,  2.2975e-03,  2.3344e-03,
         6.6891e-03,  2.2039e-03,  6.5612e-03, -2.1936e-03, -9.3453e-03,
        -2.0870e-03,  7.6019e-03,  3.7078e-03,  7.6286e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0769e-03,  7.9685e-01,  1.1647e-02, -5.4773e-03,  3.0035e-03,
         6.3319e-03,  1.3343e-03,  1.8318e-02,  2.1380e-02, -5.3156e-03,
        -1.6586e-02,  1.3380e-02,  5.1349e-03,  2.0055e-05,  9.5743e-03,
        -8.2010e-03, -2.6557e-03,  2.2823e-03, -2.5288e-03,  4.6819e-03,
        -2.4961e-04, -5.3348e-03,  1.9761e-04,  1.4325e-03, -1.1937e-02,
         1.2547e-03,  9.9022e-04, -3.4999e-03,  4.6354e-03, -2.2154e-03,
        -2.2392e-02, -8.0861e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6857e-02,  8.1974e-01,  2.4623e-02, -1.7248e-02, -1.0046e-02,
        -4.6966e-04, -2.0994e-03, -4.3423e-03,  1.2014e-02, -7.1691e-03,
        -5.0878e-03, -1.9495e-03, -4.8368e-03, -2.6107e-03, -2.5664e-03,
         3.1002e-03, -3.6170e-03, -5.7946e-04,  1.2318e-03, -3.6351e-03,
        -3.4301e-03, -5.6813e-03,  5.6111e-03, -3.3456e-03,  3.1672e-03,
         4.2149e-03,  7.5661e-03,  1.2948e-03,  1.0560e-02, -7.3114e-03,
         3.9987e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8541e-02,  8.0254e-01,  3.9008e-02, -2.5863e-02,  2.5639e-03,
        -1.3425e-03,  1.1026e-03, -1.3862e-02,  4.7862e-03, -6.1692e-04,
         6.5919e-03,  1.7773e-03,  2.6586e-03,  2.8154e-03,  6.3134e-03,
         3.4357e-03,  7.7959e-03,  9.5060e-03,  5.8103e-03,  4.9066e-03,
         1.1723e-02, -2.5780e-04,  6.1782e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2026e-03,  8.5511e-01, -1.0193e-02, -1.7090e-02, -4.1811e-04,
         1.2077e-02, -7.8345e-03, -6.7567e-04,  1.1063e-02,  2.1681e-04,
         4.8321e-03,  8.6289e-03, -1.5945e-03, -1.4430e-02,  2.9192e-03,
         5.4249e-03, -2.1646e-03, -4.6398e-03,  1.6369e-03,  2.9786e-03,
        -4.4114e-03, -3.1241e-03,  5.0765e-03, -4.7491e-03, -2.5470e-03,
         1.4254e-03, -2.5384e-03,  1.0995e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 5.7327e-02, -8.1845e-01, -7.0089e-03, -7.3169e-03, -6.2239e-03,
        -1.4560e-02, -5.8477e-03,  2.9327e-03, -6.3990e-03,  3.2890e-03,
        -2.2030e-03,  6.5720e-03, -3.6486e-03, -2.3262e-03,  4.6874e-03,
        -3.7647e-03, -2.4929e-04, -7.6498e-03,  5.4230e-03, -7.9564e-04,
        -3.8624e-03, -1.2080e-03, -4.2086e-04,  1.0413e-03,  6.3065e-04,
         5.4641e-03,  1.8704e-03, -1.7157e-02, -1.6729e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.7100e-03,  8.8601e-01, -8.8634e-03,  5.7339e-03,  8.0016e-03,
        -4.1557e-03, -1.2135e-02, -2.1299e-04, -7.4544e-03,  1.3968e-02,
         3.0236e-03, -1.8676e-03, -2.7271e-03,  3.1683e-04, -1.6908e-03,
         9.4963e-04,  2.0082e-03, -4.4939e-04,  5.7338e-03, -3.8912e-03,
        -7.8220e-03,  1.4225e-03, -3.9662e-03,  1.1886e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3373e-02,  7.7695e-01,  2.6866e-02, -1.9154e-02,  2.8271e-03,
         1.0227e-02,  2.5685e-03,  4.9619e-03,  1.2352e-02,  1.3077e-03,
         2.1586e-03,  4.4444e-03,  8.9124e-03,  2.8217e-03,  9.6024e-03,
         1.9922e-03, -1.3152e-03,  1.0226e-03,  2.2602e-03,  5.1284e-03,
        -5.9489e-04,  3.6286e-03,  2.2022e-03,  5.9080e-03,  1.1493e-02,
         4.6979e-03,  5.2631e-04,  5.3707e-03,  1.8805e-03,  2.3011e-03,
         6.7330e-04, -1.4133e-02,  1.3259e-02, -3.0897e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1200e-02,  8.3900e-01,  1.2080e-02, -2.7624e-03,  3.2165e-03,
        -4.4509e-03, -6.1403e-04,  2.1138e-03,  3.9511e-03,  4.4082e-03,
        -7.7631e-03, -1.0169e-02, -1.1304e-02, -3.2854e-03, -1.3052e-03,
        -9.8179e-03,  7.0495e-04,  6.8261e-04, -7.5372e-03,  1.7570e-02,
        -4.1913e-03, -3.1870e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1915e-02, -8.3739e-01, -2.3090e-02, -9.9589e-04, -9.3745e-03,
         3.2538e-03, -2.6953e-03,  2.0616e-02,  5.5648e-03,  1.5702e-04,
        -1.4924e-02,  7.2580e-03,  1.9055e-03,  1.6195e-03,  2.1610e-03,
         6.8135e-03,  7.8676e-03,  2.2403e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7016e-02,  8.6233e-01,  4.1826e-03, -3.7050e-03,  2.1380e-04,
         1.8985e-03,  2.3369e-03,  1.5060e-02,  8.9812e-03,  1.0199e-02,
        -3.5063e-03,  5.7911e-03, -9.2047e-03, -1.5613e-02, -1.7263e-03,
         3.6505e-03,  6.1737e-04, -7.9963e-03, -1.3033e-02,  2.8617e-03,
        -3.6155e-04, -9.0840e-03, -6.2618e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.0997e-03, -9.0018e-01, -1.4235e-02, -1.2054e-02,  1.3979e-02,
        -3.4317e-03, -3.8022e-03,  6.0352e-03, -2.4346e-03,  4.5438e-03,
         4.6302e-04,  4.2794e-03,  4.5981e-03, -7.0754e-03, -3.0398e-03,
        -4.8372e-03, -6.9097e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2907e-01,  7.1343e-01,  1.3588e-02,  3.3092e-05,  5.0996e-03,
         4.5346e-03,  2.6189e-02, -7.0300e-03,  4.9013e-03,  4.2508e-03,
         4.1968e-03, -4.2296e-03,  9.2165e-03, -1.0746e-03, -5.2323e-03,
         1.7759e-02,  1.8181e-02,  3.1981e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5203e-02,  8.1332e-01, -1.5744e-02,  8.8167e-03, -1.6553e-02,
        -5.3697e-03, -3.5525e-03, -4.7474e-03, -8.0200e-03,  6.7512e-03,
        -1.9459e-03, -5.1655e-03,  1.2983e-02,  1.1080e-03,  9.1924e-04,
        -1.8024e-02,  7.2032e-04, -7.1364e-03,  4.1334e-03,  3.7866e-03,
        -3.1628e-03, -3.6211e-03,  2.7488e-03, -5.0160e-03, -1.7227e-02,
         1.4227e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6316e-03,  8.2466e-01,  6.6994e-03, -1.4246e-02,  9.1773e-06,
         1.1075e-02, -2.4015e-02,  3.4959e-03,  5.7773e-03, -1.0388e-02,
        -7.1112e-03, -7.8800e-03, -2.3654e-03,  6.4357e-03, -7.3760e-03,
         3.2361e-03, -9.2319e-03, -2.7789e-03, -1.1203e-03, -1.1888e-02,
         1.6177e-03,  1.2016e-03,  1.2411e-03, -6.0674e-03,  4.8444e-03,
        -6.6966e-04, -7.8450e-03,  5.9959e-04, -1.4543e-03, -7.6430e-03,
        -3.3904e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0366, -0.8322,  0.0042,  0.0078,  0.0010,  0.0073, -0.0024, -0.0222,
        -0.0059,  0.0173, -0.0031,  0.0017,  0.0053,  0.0062,  0.0207,  0.0045,
         0.0117,  0.0099,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0090,  0.8502, -0.0525, -0.0311,  0.0071, -0.0074, -0.0209,  0.0219,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 3.6063e-02,  8.3228e-01,  7.8563e-03, -2.0403e-02,  3.5546e-03,
         1.1923e-02, -5.5347e-04, -9.6634e-05, -5.7124e-03,  7.4109e-03,
         3.2487e-04, -9.9394e-03,  2.6895e-03,  1.8451e-03,  3.3069e-03,
         1.1843e-03,  1.2318e-04,  2.6938e-03,  1.0656e-03,  4.3628e-03,
         8.5021e-05,  1.9178e-03, -1.0853e-03, -4.8834e-04,  2.5620e-03,
         2.6718e-03, -2.3482e-03,  2.3859e-03, -3.4943e-03,  1.1752e-03,
        -1.2166e-03,  2.2822e-03,  1.4330e-03,  3.1746e-03, -2.4033e-03,
        -7.1701e-03, -2.5303e-03,  6.0909e-05,  8.1224e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5228e-02, -7.7721e-01,  7.4336e-02,  1.0961e-02, -8.2033e-04,
         2.8501e-03,  5.5273e-03,  5.1809e-03,  9.3360e-03,  1.0842e-02,
        -2.2482e-04, -2.3451e-03, -9.7394e-03,  4.7546e-03,  2.0978e-03,
         1.1568e-03,  1.5203e-03,  9.1166e-03,  4.0157e-03,  1.9649e-03,
        -2.1693e-03, -8.1147e-03,  3.9615e-03,  4.3266e-03, -2.1194e-03,
        -1.0856e-03, -2.7399e-03,  6.2527e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5669e-02, -8.3041e-01,  2.5041e-03,  4.7646e-02,  8.6915e-03,
        -6.1430e-04,  1.3469e-02,  7.1189e-03, -3.8312e-03,  3.3943e-03,
        -3.7796e-04, -7.9160e-03, -1.8358e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0226,  0.8452, -0.0139,  0.0100, -0.0017,  0.0108, -0.0148,  0.0068,
         0.0018, -0.0112,  0.0016, -0.0047, -0.0036,  0.0121, -0.0391,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.8975e-03, -7.8124e-01,  1.7943e-03,  1.5623e-03, -1.8067e-02,
         1.2547e-02, -5.0705e-03,  9.1207e-03, -6.8582e-03,  8.4630e-03,
         1.0466e-02, -2.4605e-03, -4.9038e-03, -4.6188e-04,  2.3436e-03,
        -2.5085e-03,  1.4514e-03,  1.4961e-02,  4.8095e-03,  3.5777e-03,
         1.1633e-03,  1.2570e-03,  4.0385e-03,  2.8216e-03,  5.0243e-03,
        -4.1161e-04,  1.3932e-03, -4.5668e-03,  3.3106e-03,  2.6682e-03,
         1.4706e-03, -1.0081e-04, -2.6548e-03, -5.9116e-03, -4.2097e-03,
        -3.5667e-03,  3.6922e-03,  1.8002e-03,  3.1679e-03,  1.1596e-03,
        -3.8549e-03,  2.9592e-03,  1.2278e-03,  2.6968e-03, -6.1433e-03,
        -4.7829e-03,  2.5377e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7104e-02, -8.4312e-01, -1.4669e-02, -9.9619e-03, -1.2467e-04,
         1.2526e-03,  5.3508e-03, -1.0722e-04, -4.6545e-04,  4.0882e-04,
         1.6823e-03, -2.0534e-03, -7.5449e-04,  2.2837e-04, -3.7734e-03,
        -1.2795e-03, -5.9417e-03,  8.4198e-04,  1.7573e-03, -9.3301e-04,
        -3.6271e-04, -2.0744e-04,  3.9563e-04, -2.9308e-04, -1.5840e-03,
         2.6570e-03, -1.2140e-02,  1.1797e-03,  1.9594e-03, -4.0510e-03,
         8.9390e-04,  4.2028e-03,  6.2904e-04,  6.2465e-04, -4.0502e-03,
         6.6363e-04, -1.4425e-03, -3.4261e-03, -3.0398e-03,  6.7875e-03,
        -2.2891e-03,  2.5310e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0926e-02,  7.7278e-01,  1.5268e-02,  2.6494e-03,  6.3186e-03,
         1.4718e-02, -5.2336e-03, -4.1051e-03,  4.1764e-03,  5.8048e-03,
         8.4298e-03,  9.4634e-03, -4.7262e-04, -1.1372e-04,  1.2168e-02,
        -6.0220e-03,  4.1607e-02,  9.0403e-03, -1.9025e-03,  1.8926e-03,
        -9.7536e-03, -6.1418e-03,  1.5574e-03, -4.3093e-03, -5.1443e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5697e-02,  8.7504e-01, -2.6450e-03,  2.8429e-03, -3.4498e-03,
        -7.7571e-03, -1.3255e-02, -2.9152e-03,  5.6885e-04, -2.1924e-03,
         2.0413e-03,  1.3175e-02, -7.0231e-04,  5.8710e-03, -4.1906e-03,
         2.6529e-03,  2.9464e-03, -4.2889e-03,  4.9571e-03,  9.5876e-03,
         1.3620e-03,  1.5867e-02, -5.9917e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5927e-02, -8.0263e-01, -7.4885e-02, -2.8360e-03,  1.0008e-02,
         7.8972e-03,  6.2731e-04, -6.6836e-04,  8.1114e-03,  2.5322e-03,
         6.4386e-03,  1.6295e-03,  9.8820e-03,  6.1538e-03,  6.3899e-04,
         5.1669e-03, -5.0305e-03, -1.8753e-03,  1.2949e-03,  4.6987e-03,
         3.3212e-03,  1.2855e-02,  1.4895e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0668,  0.7546, -0.0416, -0.0197, -0.0120, -0.0009,  0.0235, -0.0049,
         0.0053, -0.0035,  0.0156, -0.0412, -0.0103,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0233, -0.8173,  0.0325,  0.0075,  0.0142, -0.0024,  0.0134, -0.0206,
        -0.0154, -0.0047,  0.0012, -0.0129,  0.0123,  0.0224,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0186,  0.9071,  0.0142, -0.0112, -0.0050,  0.0027, -0.0011, -0.0023,
        -0.0039, -0.0012, -0.0033,  0.0035,  0.0024, -0.0041,  0.0026,  0.0022,
         0.0072, -0.0056,  0.0017,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 3.5044e-02,  8.3004e-01,  9.2999e-03, -7.4059e-03,  1.4747e-03,
        -1.5758e-03, -8.8488e-04, -5.4842e-03,  5.5547e-03,  4.1645e-03,
        -3.8924e-03, -5.5253e-03, -9.4013e-04,  2.1278e-03, -4.9357e-04,
        -3.0597e-02,  5.6406e-03,  1.5551e-03,  2.1086e-03,  1.2224e-02,
        -5.7112e-03, -6.0643e-04, -2.1877e-03,  2.1964e-03, -1.7127e-03,
        -2.1550e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2725e-02,  9.0080e-01, -1.9266e-02,  9.4424e-03, -1.5046e-02,
        -3.4124e-03,  7.4496e-04, -9.2199e-03, -1.2244e-03,  3.9837e-03,
         3.5385e-03, -1.3503e-03, -4.9787e-03, -6.6374e-03,  3.0011e-03,
         4.6326e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0129,  0.8466,  0.0283, -0.0107, -0.0093, -0.0026, -0.0056, -0.0016,
         0.0226, -0.0031, -0.0096, -0.0024,  0.0050,  0.0021, -0.0062, -0.0104,
         0.0209,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0756e-02,  7.9499e-01,  6.8047e-03, -6.9332e-03,  3.6223e-02,
        -1.0754e-03,  5.5881e-04,  4.8951e-02, -4.3710e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1617e-02,  7.0290e-01,  2.7681e-02, -7.2417e-03, -2.7144e-02,
         1.2584e-02,  2.2120e-03,  7.3602e-03,  1.9878e-03,  3.2259e-03,
         1.0008e-02, -4.8160e-03,  5.2491e-03, -4.4138e-03,  4.7457e-04,
         6.3903e-03, -8.3034e-03,  4.2181e-03,  7.5732e-03,  4.7353e-03,
         5.0652e-03, -4.3635e-04,  2.6005e-03, -2.2011e-03,  4.3980e-03,
        -2.5863e-03,  2.5521e-03,  2.8116e-03,  3.3058e-03,  6.8360e-04,
         4.3217e-03,  2.0325e-03,  1.7619e-03,  7.4515e-04,  2.2513e-03,
         1.4924e-03,  1.6091e-03,  4.3101e-03, -4.9178e-03, -7.4531e-04,
        -4.1589e-03, -1.0153e-03, -3.8016e-03, -1.5788e-03, -1.1035e-03,
        -2.7378e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.9635e-03,  8.2901e-01,  8.0803e-03, -3.5633e-02,  3.2612e-03,
        -4.8019e-03,  3.3707e-04, -4.8071e-03,  2.1285e-03,  7.2063e-03,
        -1.2470e-03, -6.3074e-04,  1.4382e-02, -5.7545e-04,  1.1528e-05,
         5.5107e-03,  6.0103e-03,  3.6882e-03,  1.6634e-03,  8.7047e-03,
        -1.9029e-02,  2.2958e-04,  2.9843e-03,  2.8866e-04,  2.0844e-03,
        -2.3328e-03, -6.0499e-03, -5.6572e-03, -3.0864e-03,  1.1601e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0224, -0.8730,  0.0321,  0.0249,  0.0281, -0.0176, -0.0019,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1569e-03, -9.0156e-01,  8.3363e-03,  8.4458e-03,  2.3618e-03,
         6.4083e-03,  1.3338e-04, -2.7509e-02,  5.0386e-03, -3.2740e-02,
        -6.3124e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1428e-02, -7.8653e-01, -2.2732e-02, -4.0733e-02, -3.9248e-03,
        -2.4091e-02, -1.3258e-02, -3.1922e-03, -9.9959e-03, -2.2313e-04,
        -9.8434e-04, -6.5714e-03, -4.2787e-03,  4.2054e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3907e-02, -8.3681e-01,  4.4699e-03, -2.0127e-04,  9.8734e-03,
         9.5080e-03,  5.7389e-03,  7.9600e-03,  1.0967e-02,  2.6017e-03,
        -1.7258e-03, -2.2527e-03,  8.0196e-03,  1.1317e-04,  1.5074e-03,
         2.3128e-03, -2.6581e-03,  2.3571e-03, -8.0069e-04, -5.3335e-03,
        -3.7615e-03, -8.3985e-03,  3.1094e-03,  1.1435e-03,  2.8925e-03,
         6.2164e-03, -3.6070e-03, -1.2768e-03,  7.8731e-04, -1.0361e-02,
        -1.8617e-03, -6.5696e-03,  1.9800e-03, -1.8916e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9955e-02, -7.9090e-01, -4.8477e-02,  1.1935e-02, -9.2710e-03,
         4.3448e-03, -1.1476e-02,  4.7419e-03,  3.9624e-03, -7.8398e-03,
         8.7466e-03,  5.5343e-03,  1.9607e-03,  5.8870e-03,  9.7456e-03,
        -6.7749e-04, -2.8588e-03, -4.4328e-03,  4.8992e-03,  7.5193e-03,
        -5.7261e-03, -1.8772e-03,  5.7731e-03, -2.3811e-03, -2.9017e-03,
         2.6257e-03, -5.4123e-03,  8.1347e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9095e-02,  8.5227e-01, -7.3097e-04, -5.9883e-03,  6.5224e-03,
        -5.5164e-03, -4.6474e-03,  4.3095e-03,  3.2473e-03,  1.0658e-03,
        -1.3428e-04, -4.9274e-03,  1.4117e-04,  5.3215e-03,  5.8951e-03,
        -6.8937e-03, -3.4659e-03,  1.5334e-02,  5.2456e-03, -3.9334e-04,
        -4.0214e-03, -2.2076e-03, -6.6353e-04, -1.1961e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-2.3867e-03,  8.3238e-01,  1.7798e-02,  4.8172e-03, -3.5719e-03,
         1.1233e-03,  5.2391e-03,  1.6853e-03,  3.8792e-03, -2.5339e-04,
        -2.3865e-02,  2.0317e-03,  4.4952e-03,  7.0964e-03,  3.2739e-03,
        -5.2400e-03, -6.9707e-04, -2.1224e-03,  2.5360e-03,  1.9602e-02,
         2.6533e-03,  6.5059e-03, -6.5445e-03,  2.3367e-03,  1.8525e-03,
         4.2020e-04,  1.5429e-03,  6.7321e-04,  3.8042e-03, -1.9941e-02,
        -9.6362e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0272,  0.8744, -0.0094,  0.0093,  0.0120, -0.0017,  0.0089, -0.0052,
         0.0057,  0.0060, -0.0125, -0.0278,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8587e-02,  8.9531e-01,  1.0436e-04, -2.3307e-02,  5.2054e-03,
        -1.1291e-03,  2.4556e-03, -3.9565e-03,  7.1616e-04, -1.0616e-02,
         4.7872e-03, -1.2799e-02,  4.8029e-03,  4.1201e-03, -2.0999e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5514e-02,  8.1735e-01,  5.2225e-03, -1.0550e-02,  1.2391e-02,
         2.9561e-03,  1.3966e-02, -1.6536e-02,  5.3988e-03,  9.6046e-03,
        -2.4985e-04, -2.3207e-03, -8.3467e-04, -5.9083e-03,  7.2223e-03,
         8.5291e-04, -4.5949e-03, -1.0046e-02,  3.4613e-03, -6.9067e-03,
        -8.3270e-04, -1.4090e-03, -1.8140e-03, -2.7776e-04,  2.1391e-03,
         8.3861e-03,  2.7268e-04, -2.8724e-03, -1.9831e-03,  1.3209e-03,
         4.9377e-04, -6.1734e-03, -1.3637e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8795e-02,  8.7616e-01,  6.6766e-03, -1.9430e-02,  1.8158e-03,
         2.5421e-04, -5.3414e-03, -1.5569e-02, -1.1226e-02, -2.7782e-03,
         1.4181e-04, -6.0018e-03, -1.5745e-04, -1.6105e-02,  6.8888e-04,
        -8.8564e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0036,  0.8724, -0.0203, -0.0189,  0.0131, -0.0112, -0.0248,  0.0061,
        -0.0295,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0289,  0.8565, -0.0752,  0.0115, -0.0279,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4057e-02, -8.3368e-01, -9.1671e-03, -1.3356e-02,  5.5333e-03,
        -2.7344e-03,  8.9239e-03,  1.2050e-02, -6.4729e-03,  6.0998e-04,
        -1.1301e-02,  7.6385e-04, -6.6598e-03,  5.1250e-03,  6.0408e-04,
        -2.6522e-03, -3.5140e-03, -1.6454e-03,  3.6205e-03, -3.2739e-03,
        -1.9495e-03, -9.6238e-04, -1.5095e-03,  1.5222e-03, -1.0888e-03,
         2.9669e-03,  4.2600e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0337,  0.9486, -0.0021,  0.0021, -0.0061, -0.0064,  0.0010,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8604e-02,  8.5535e-01,  1.0362e-02, -1.3363e-02, -1.4488e-02,
         6.7171e-04,  2.4717e-03,  8.7202e-04, -6.7588e-03, -4.5714e-03,
         6.8176e-03,  8.0849e-03, -9.4734e-03, -4.2919e-03, -2.8506e-03,
        -5.2529e-03, -5.0200e-03, -4.9891e-03, -6.6410e-03, -8.2071e-03,
         4.8966e-03,  5.9593e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0366, -0.8373,  0.0085, -0.0210, -0.0057, -0.0090,  0.0062, -0.0042,
         0.0052,  0.0047, -0.0080, -0.0024, -0.0011, -0.0147,  0.0073,  0.0011,
         0.0063, -0.0034, -0.0173,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6270e-02, -8.1268e-01, -2.8626e-02, -9.9332e-03,  1.2627e-03,
        -1.8002e-02, -6.7981e-03, -3.7047e-03, -5.3473e-03, -7.6469e-04,
        -5.4117e-03,  1.1409e-03, -8.5849e-04, -4.8331e-03,  2.4288e-03,
        -2.2155e-03, -4.9371e-03, -4.6361e-03, -4.4668e-03, -1.8705e-03,
        -2.1602e-03, -5.1109e-03, -3.0760e-03, -1.2764e-04,  8.1721e-04,
        -1.1970e-03, -1.3969e-03,  2.2751e-04, -6.5799e-03,  5.9396e-03,
         7.1764e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 1.5236e-02, -7.6394e-01, -7.6823e-02,  1.8286e-02,  2.3853e-03,
        -5.7410e-04,  9.8838e-03,  4.5752e-03,  6.8603e-03,  1.1542e-03,
         1.6580e-04,  5.1084e-03, -1.5789e-03,  3.2235e-03, -2.8970e-03,
         8.4326e-04,  4.4046e-03, -5.8106e-03, -8.4996e-03, -7.4684e-04,
         1.8839e-03,  4.2087e-03,  1.4673e-03,  1.6058e-02, -2.0985e-03,
         2.7791e-04, -1.8308e-03,  2.2645e-02,  4.7840e-03, -3.9164e-03,
        -1.7470e-03, -6.0853e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8910e-02,  8.4616e-01,  2.0774e-03,  2.3922e-03,  4.7922e-04,
         6.2353e-03, -2.7782e-03, -1.9592e-03,  5.3692e-02, -1.2705e-03,
        -2.7718e-03,  1.0308e-03, -2.4489e-03, -3.9848e-03, -1.2642e-03,
         3.6143e-03,  4.8279e-03, -5.2321e-03, -8.6212e-03,  1.3637e-03,
         2.1421e-03,  6.7494e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.7699e-03,  8.4362e-01, -1.4980e-02, -8.0679e-03, -7.2974e-03,
        -1.5592e-03, -5.1890e-03,  8.1885e-03, -1.3209e-03, -2.0889e-03,
         6.7941e-03,  4.2866e-05,  1.1602e-02,  5.6983e-04, -6.6921e-03,
         1.0520e-03,  1.5870e-03,  9.9545e-03,  1.5145e-03, -1.5966e-02,
         3.1379e-04, -6.3826e-03,  3.2059e-03, -4.2759e-03, -4.2442e-03,
        -3.9177e-03, -1.5462e-03,  1.2978e-02, -3.3412e-03,  4.9361e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.9334e-03,  9.0040e-01, -1.4778e-02, -1.1598e-02,  8.5620e-03,
         1.9054e-04, -1.2560e-03, -6.3526e-03, -3.9567e-03,  7.4890e-04,
         2.6178e-04,  8.1053e-03, -1.5540e-03,  1.3804e-03, -4.3412e-03,
        -4.6451e-03,  5.1986e-03,  1.9739e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4243e-02,  7.9798e-01,  1.4054e-02, -1.4625e-03, -1.8928e-02,
        -7.7843e-03, -7.0489e-03,  7.2811e-03,  2.6102e-03,  5.3429e-04,
         1.1318e-03,  6.0166e-04,  7.8858e-04,  4.9830e-03, -3.9580e-03,
        -5.6065e-03,  2.5440e-03,  7.7223e-04, -1.2077e-02,  6.7980e-04,
         2.2615e-03, -2.4977e-03,  1.6320e-03,  7.1772e-03, -1.7067e-03,
        -9.9776e-04,  2.3703e-03,  1.3963e-03,  5.8020e-04,  1.3044e-03,
        -1.1566e-03,  2.5681e-03,  3.7097e-03,  2.7099e-03,  5.6578e-03,
        -1.9314e-02, -7.8956e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0513,  0.7695, -0.0710, -0.0041,  0.0044,  0.0132, -0.0014,  0.0173,
         0.0012,  0.0009, -0.0010, -0.0011,  0.0037,  0.0053, -0.0021,  0.0018,
        -0.0047,  0.0040,  0.0024,  0.0020,  0.0134, -0.0008,  0.0113,  0.0123,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2659e-02, -8.3039e-01, -3.2294e-02, -1.7536e-03, -1.1249e-02,
        -6.3690e-04,  4.6318e-03, -4.8539e-03, -9.0654e-03,  1.8613e-03,
        -1.1016e-03,  4.7895e-03,  4.6711e-04, -1.5220e-03,  1.7898e-03,
        -1.9914e-03,  3.2922e-03,  1.1996e-03,  2.0227e-03, -2.2197e-03,
        -2.1030e-03,  1.6846e-03,  2.8999e-03, -7.6341e-04, -4.9562e-03,
        -3.6236e-03,  2.6992e-03, -3.5072e-03, -4.7796e-03, -1.5046e-04,
         3.4763e-03,  6.8067e-04,  8.4260e-04, -2.1287e-03,  4.1630e-04,
        -8.4408e-03,  6.9138e-04,  4.3350e-03,  7.4503e-04,  1.3542e-03,
         4.4598e-04, -2.2421e-03, -7.1233e-04, -7.4204e-03, -6.8893e-03,
        -2.2217e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.8961e-03,  8.4997e-01, -1.6708e-03,  5.8874e-03, -2.6530e-02,
        -1.0995e-02, -1.6370e-03, -7.5558e-03,  2.2909e-03, -9.4859e-03,
         4.6442e-03, -3.7944e-03,  4.7640e-03, -3.1599e-03,  1.4705e-03,
         4.4754e-03,  1.3690e-02, -3.7174e-03,  1.1480e-04, -6.3242e-05,
         1.0912e-02, -2.3718e-02, -3.5583e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0415, -0.8276,  0.0451, -0.0173, -0.0185,  0.0053,  0.0012, -0.0117,
        -0.0040,  0.0022,  0.0255,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0643,  0.8042, -0.0262, -0.0200,  0.0051, -0.0034,  0.0082, -0.0010,
         0.0138,  0.0086,  0.0166,  0.0048,  0.0195, -0.0041,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5697e-03,  8.9806e-01, -2.0803e-03,  2.7384e-03,  5.8507e-04,
        -2.9319e-03, -2.4094e-02, -9.2748e-04,  6.5773e-03, -3.1559e-03,
         2.5786e-02,  2.7489e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1127e-02, -8.2038e-01, -3.9641e-03, -2.6155e-02,  7.8589e-03,
         1.5196e-05,  2.0226e-03,  4.8873e-04, -1.1074e-04, -3.4725e-03,
         2.8510e-03, -3.3853e-03, -1.9463e-03, -1.9864e-03,  1.4493e-03,
        -1.0500e-03,  1.0849e-03,  1.1211e-03, -4.4027e-04,  5.4636e-03,
         1.4724e-03, -2.2456e-04, -4.5736e-03,  3.2051e-03,  2.8456e-03,
         4.6957e-03, -4.8288e-03, -1.0092e-04,  2.8829e-03, -2.6631e-03,
        -1.4171e-03, -2.1584e-03, -6.1951e-04, -1.0555e-03, -1.4945e-03,
        -2.5417e-04,  1.2212e-03,  1.4570e-03,  1.9664e-03,  3.3621e-04,
        -9.1660e-05,  1.8905e-04, -9.2099e-04, -6.4069e-04,  1.7135e-03,
         2.5097e-03, -3.5912e-03,  3.0648e-03,  1.5909e-03,  2.5661e-03,
         1.4177e-05,  9.3464e-04,  2.7132e-03, -3.3922e-03,  1.0217e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-5.1616e-04, -8.7892e-01, -1.8603e-02,  7.6220e-03, -3.1033e-03,
        -2.7459e-03,  4.7017e-03,  7.7495e-03,  8.2629e-03, -7.8103e-03,
        -2.7683e-03,  2.2460e-05, -1.1962e-02, -1.3841e-03, -1.7204e-03,
        -7.2289e-03,  2.0568e-03, -1.7729e-02, -3.3342e-03, -5.1244e-03,
         2.6033e-04, -5.5994e-04, -5.8115e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.7963e-02,  7.7723e-01, -4.3915e-02,  8.3873e-04, -5.3344e-03,
        -2.8204e-03, -1.0653e-03, -1.3661e-03,  2.2315e-02, -1.7460e-02,
        -5.8775e-03, -9.2762e-03, -4.3102e-03,  1.4618e-02, -7.9027e-03,
         2.9365e-04, -7.2149e-03,  1.0200e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5994e-02,  7.6050e-01,  7.0947e-02,  2.6757e-02,  1.3097e-02,
        -1.9065e-02, -1.1512e-03,  1.0103e-02, -1.7263e-03,  4.9256e-04,
        -4.5187e-03,  1.9577e-03,  9.1524e-03, -7.9843e-04, -6.7814e-03,
        -6.5548e-03, -8.6084e-03,  1.7926e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1437e-03,  9.0456e-01, -2.1363e-02, -1.0243e-02, -6.5180e-03,
         1.0271e-02, -2.0548e-03, -1.7366e-03,  3.8353e-03,  1.2826e-02,
         6.0359e-04, -1.0047e-02,  1.1949e-03,  8.2751e-03, -2.3312e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2345e-02,  8.4075e-01,  1.3493e-03, -1.5872e-02,  7.5577e-03,
        -5.2765e-03,  1.1566e-02, -1.0080e-03,  6.4971e-03,  1.0082e-02,
         2.9505e-04, -1.1480e-03,  4.9311e-03,  3.1197e-03,  2.6794e-03,
         1.0640e-02, -2.5495e-03,  2.2707e-03,  4.0423e-03,  1.3160e-02,
         1.6163e-03, -2.1222e-03, -2.1311e-03, -1.7365e-04,  1.9750e-03,
         1.5591e-03,  1.1271e-02, -1.3301e-03,  2.3086e-03,  1.0841e-02,
        -1.8065e-03,  5.7237e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2974e-02, -8.3103e-01,  4.5725e-03, -6.1271e-03, -1.7091e-02,
        -1.4799e-03,  1.5177e-03,  1.5597e-03,  1.6188e-02, -7.4332e-03,
         2.8973e-04,  2.8519e-03,  7.6362e-04,  3.0625e-03,  1.2218e-04,
         1.3372e-02,  8.5072e-04,  8.1606e-04, -3.2055e-03, -1.1206e-03,
         1.6889e-03,  1.1311e-02,  6.9223e-03,  1.0391e-03, -1.7752e-03,
        -1.5878e-03, -1.2769e-03, -8.6500e-03,  2.4278e-03,  8.4163e-04,
         9.5067e-03, -4.2418e-03, -2.3048e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0335e-02, -7.9783e-01,  6.5797e-03, -3.3257e-02, -1.4105e-02,
         4.2502e-03,  6.0000e-03,  7.3375e-03,  5.4738e-03, -9.4652e-03,
        -1.8421e-03, -3.7331e-03, -3.2068e-03,  4.0130e-03, -2.5837e-03,
         4.7599e-03,  1.1074e-03,  4.0487e-03, -1.4226e-03,  1.2145e-03,
         6.8616e-04,  2.0148e-03,  9.1818e-04,  1.4426e-03, -7.9701e-04,
         8.5833e-04, -2.2483e-03,  4.9307e-04,  3.9038e-04,  2.2228e-04,
        -3.2523e-03, -3.1452e-03,  7.9180e-04,  5.7196e-03,  1.7931e-03,
         4.3831e-03,  1.0660e-03,  2.7006e-03,  1.0266e-03, -6.5186e-03,
         9.3281e-04,  1.5136e-03,  1.4238e-03,  9.0830e-04, -3.2436e-04,
        -2.5665e-03,  5.8956e-04,  6.2100e-05, -1.1347e-03,  1.1675e-04,
         3.1739e-03, -2.5881e-03, -3.3577e-03, -5.6414e-04, -2.0223e-03,
         3.0107e-03, -4.3568e-03, -8.3232e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.2431e-02,  7.5904e-01, -5.4395e-03, -2.8680e-02, -7.9389e-03,
        -1.1162e-02, -2.8898e-03, -2.4525e-03, -7.5640e-03,  6.8807e-03,
        -2.1101e-04, -2.2592e-03, -7.3749e-03, -1.4890e-02,  4.1573e-04,
        -1.1812e-03, -2.3721e-03, -1.3994e-03, -4.6958e-03,  1.4704e-03,
         5.3928e-03, -5.6826e-03, -4.3883e-03, -7.2943e-03, -1.1252e-02,
        -2.5240e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8368e-04,  8.7354e-01, -2.1505e-02, -9.8420e-03, -2.2343e-03,
         1.6403e-02, -2.2314e-03, -1.7460e-03, -9.5523e-03, -5.6563e-03,
        -3.5087e-03,  1.5479e-03,  7.3740e-03,  7.0224e-03, -1.4795e-03,
         7.8510e-04,  3.2381e-03, -2.9754e-03,  1.3663e-03, -1.2004e-03,
        -8.2675e-04, -8.7396e-03, -1.2801e-02, -8.3729e-04,  3.4020e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0319e-02, -7.9738e-01, -3.2816e-02, -6.2264e-03, -1.8144e-03,
        -3.2792e-03, -4.3718e-03, -7.9749e-03, -3.7728e-03, -7.8650e-03,
         1.3542e-03,  1.5043e-03, -3.7782e-03, -1.6297e-02,  1.0928e-02,
        -8.4662e-03,  6.5929e-03,  6.3052e-04, -4.1726e-03, -3.0401e-03,
         4.8744e-04,  5.1707e-03, -3.0543e-03,  2.4643e-03,  3.1105e-03,
         4.3698e-03, -9.1623e-03, -4.6184e-04,  2.4956e-03,  2.6368e-03,
        -4.4670e-03,  1.9410e-03, -2.3817e-03, -4.2778e-03,  3.8940e-03,
        -1.6676e-04,  1.6907e-03,  2.1865e-03,  3.4190e-03,  6.3617e-04,
         5.7369e-03,  1.6405e-03,  2.0335e-04,  7.7115e-04,  5.9160e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7402e-02,  7.9365e-01,  2.1505e-02,  1.6605e-02,  6.3834e-03,
         9.3327e-04,  4.3105e-03,  6.9464e-03, -8.3161e-03,  6.6297e-04,
         4.9276e-03, -8.6902e-03, -8.3759e-03,  3.9932e-03,  1.9776e-03,
         6.1797e-03, -1.3137e-02,  1.9997e-03,  6.0430e-03, -3.9826e-04,
         6.8133e-03, -2.2856e-02, -1.7894e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0189,  0.8658,  0.0042, -0.0392,  0.0077,  0.0176,  0.0161,  0.0023,
        -0.0024, -0.0161,  0.0066,  0.0032,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-1.5926e-02,  8.5362e-01, -2.6539e-02, -1.4840e-02, -1.0217e-03,
         1.0486e-02, -5.2083e-03, -6.3189e-03,  9.1048e-03, -2.9131e-03,
        -6.8728e-03, -4.4462e-03, -5.8158e-03,  9.5564e-04, -5.5062e-03,
        -2.7966e-03,  3.6795e-03,  5.4804e-04, -1.4170e-03,  7.6719e-03,
         5.7031e-03, -2.6774e-03,  5.9324e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.0729e-02,  8.2123e-01,  2.4786e-02, -4.7733e-03,  3.6166e-03,
        -7.7084e-03, -1.0923e-02,  2.8366e-03, -2.4075e-03,  9.7170e-03,
        -3.2729e-03, -1.8110e-03, -3.3511e-03, -1.6272e-03, -1.6947e-04,
        -4.6062e-03, -2.0300e-03, -1.0016e-02, -1.5654e-03,  4.5498e-04,
         2.2457e-03, -9.8541e-04, -2.5321e-03, -2.0761e-03, -1.0814e-03,
         3.3997e-04, -6.0699e-03, -3.3021e-03, -5.5637e-03,  3.0802e-04,
        -1.9370e-03,  1.3046e-02,  2.8784e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0583,  0.7903, -0.0267, -0.0215, -0.0165, -0.0036,  0.0086, -0.0047,
        -0.0020, -0.0023,  0.0041,  0.0022, -0.0176, -0.0039,  0.0054, -0.0046,
         0.0036,  0.0167,  0.0074,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2941e-02,  7.8050e-01, -2.7428e-02,  4.5175e-02, -4.0471e-03,
         6.7738e-03,  9.1014e-03, -7.5118e-04,  6.2450e-03, -5.1742e-03,
         2.6753e-02, -2.6864e-02,  1.3613e-02, -4.6333e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2471e-02,  8.8115e-01,  3.1529e-02, -2.5302e-03,  3.1186e-03,
         2.4684e-03,  1.2666e-03, -5.4839e-03,  2.1712e-03,  2.9286e-03,
        -7.0358e-04,  1.9527e-03,  5.3313e-03, -1.9445e-02,  1.6827e-03,
         2.4745e-03,  1.4685e-03,  3.0422e-03,  3.1243e-03,  9.9662e-04,
         7.2009e-04,  1.3940e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9181e-03, -8.8140e-01, -1.0560e-03, -9.9415e-03,  3.3191e-04,
         3.8903e-03,  1.0838e-02, -3.7227e-03,  4.2576e-03,  1.6976e-03,
         1.3999e-03,  3.2948e-03, -2.0934e-03,  3.4064e-03, -6.0831e-03,
         1.3937e-03,  7.4393e-03,  1.9310e-03, -6.7563e-04, -1.3375e-03,
         1.1286e-03, -1.1509e-03, -1.1458e-04,  2.1031e-03,  3.7397e-03,
        -3.3098e-03,  4.2342e-03,  2.7795e-03, -3.1692e-03,  2.8161e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5105e-02,  8.4213e-01, -1.1588e-02, -9.9155e-03, -5.1950e-04,
         7.7461e-03,  2.3619e-02,  7.7609e-03, -6.8546e-03, -8.6399e-03,
        -1.8245e-02, -7.1842e-03,  6.5289e-03,  7.5708e-03,  5.1848e-03,
        -9.5867e-03, -1.1821e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0745,  0.7430,  0.0469,  0.0358, -0.0094,  0.0046, -0.0165, -0.0034,
         0.0406, -0.0196, -0.0055,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0336e-03, -8.4362e-01,  1.0564e-02,  1.3032e-02,  6.5018e-04,
         3.7223e-02,  2.4204e-02,  2.7528e-03,  7.0430e-04, -4.3320e-03,
        -3.4139e-03,  7.2608e-03,  2.2411e-03, -2.6347e-02, -1.0523e-02,
         4.3310e-03,  7.6901e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3902e-02,  8.2904e-01, -4.1025e-02,  2.2728e-02, -4.5449e-03,
         8.4654e-03, -5.7706e-04, -1.4867e-03,  5.0467e-03, -5.7424e-04,
        -5.5104e-03,  6.2465e-03,  2.3075e-03,  2.1299e-03,  1.1272e-02,
         2.8728e-03, -2.2038e-04, -5.9746e-03, -1.4897e-02,  1.4876e-03,
        -2.0053e-03,  1.6365e-04,  2.0537e-03, -1.9047e-03, -2.0944e-03,
         1.4703e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.6018e-02,  7.9385e-01,  2.3426e-02, -1.3090e-02,  9.5188e-04,
         1.3473e-02,  8.8762e-03,  8.5675e-03,  3.1122e-03, -2.9130e-03,
        -3.3074e-03, -6.9851e-04, -5.6740e-04,  3.6965e-03, -1.1054e-03,
        -7.2675e-03, -5.6689e-03, -4.7022e-03,  1.6708e-03, -3.7914e-03,
         7.7901e-03, -8.8581e-05,  2.1848e-03, -7.7043e-04,  5.6176e-03,
        -1.7510e-04,  2.9580e-03, -1.0348e-03, -4.9672e-03,  1.7662e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.4649e-03,  8.7726e-01, -2.3257e-02,  5.8327e-04,  2.2307e-02,
        -7.8999e-03,  5.3445e-03,  6.9831e-03,  1.1447e-03,  1.9491e-03,
        -5.4894e-03,  2.6254e-03, -6.1710e-03,  3.0684e-03,  6.7612e-04,
         1.0059e-03, -1.4149e-02, -6.1502e-03,  9.4720e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-8.1245e-02,  7.7854e-01, -1.6188e-02,  7.4523e-03,  3.5364e-04,
        -1.9105e-03,  1.4586e-02,  1.1965e-03,  8.6857e-03,  2.4235e-03,
         1.0261e-03,  4.3034e-04,  2.6267e-03,  3.3233e-03, -7.9483e-03,
         3.4565e-03,  8.3349e-03,  1.8428e-02,  1.3272e-03,  5.1611e-03,
         3.6643e-03, -3.4421e-03, -3.9617e-04, -1.8803e-03, -1.0137e-02,
        -6.8381e-05,  1.5768e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0023,  0.8759,  0.0212,  0.0032,  0.0088,  0.0091,  0.0021,  0.0016,
         0.0063,  0.0015, -0.0026,  0.0093, -0.0141, -0.0422,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0837e-03,  8.4727e-01, -1.8766e-02, -6.6643e-03, -5.4059e-03,
         1.3919e-03,  6.7289e-03,  8.8735e-03,  2.1248e-02,  9.9231e-03,
        -2.1585e-03,  4.8458e-03, -1.4233e-03,  5.0747e-03, -1.0899e-03,
        -7.7608e-03,  4.2754e-03, -5.7642e-03,  4.7547e-03,  3.4415e-03,
        -5.5234e-04,  1.4718e-03, -3.1310e-03, -1.2672e-02, -2.4658e-03,
         6.7668e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5300e-02,  8.7660e-01,  1.1382e-02,  4.8356e-04, -2.5876e-03,
        -1.7075e-02,  5.5509e-03, -5.6247e-03, -1.1162e-03, -2.1648e-03,
         5.6479e-04, -3.8276e-03, -5.8754e-03, -2.1844e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.4524e-03,  8.4183e-01, -2.9773e-04, -2.1343e-02, -2.8820e-03,
         3.4351e-03,  2.5624e-03,  2.4253e-04,  2.6202e-03,  3.6890e-03,
        -1.6719e-03, -4.6126e-06,  3.1825e-03, -2.1669e-03,  9.1681e-03,
         4.7111e-03, -4.7743e-03,  7.8866e-03, -9.4994e-04, -7.0758e-03,
         1.1992e-03,  4.6904e-03,  9.2949e-03,  4.2796e-03, -9.3996e-04,
         2.2992e-03, -7.5230e-03,  8.5615e-04,  3.7288e-03,  7.4216e-03,
         2.1632e-03,  2.6661e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2069e-02,  8.0545e-01, -6.3574e-03, -2.6428e-02, -1.5794e-03,
         1.5171e-02, -1.8549e-02, -3.1602e-03,  3.8623e-03,  9.3184e-04,
         3.5607e-03, -2.1494e-03, -4.0621e-03,  2.5318e-03,  1.0083e-02,
         2.2439e-03,  5.0199e-03, -4.2530e-03,  2.8815e-04,  3.6378e-03,
        -2.3246e-03,  3.2803e-03, -1.6334e-03, -3.0650e-03, -2.8309e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0295,  0.8943,  0.0273, -0.0064, -0.0191,  0.0234,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6733e-02,  8.0180e-01,  5.7336e-03,  1.2098e-02,  1.6611e-02,
         1.8239e-03,  5.1364e-03, -1.9692e-03,  9.1495e-03, -5.4901e-03,
        -2.3344e-03,  7.1124e-03,  1.2452e-03,  1.5309e-02, -3.9162e-03,
         4.6874e-03,  2.3122e-03,  3.4712e-03,  1.7417e-03,  6.0579e-03,
        -4.3756e-03,  6.9667e-03, -5.2019e-03, -1.0805e-04, -3.1830e-04,
        -2.7863e-03, -2.2782e-03,  8.1621e-03,  6.1788e-04,  1.2354e-02,
         2.9731e-03,  1.5686e-03,  1.6069e-03, -2.5135e-03,  7.3619e-04,
        -1.3390e-03,  1.3604e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0374e-04, -8.3350e-01, -2.4098e-02, -2.4049e-02, -1.2888e-03,
        -1.2132e-02, -3.4807e-03, -5.3061e-03,  1.6563e-03,  1.2240e-03,
         1.0266e-03, -1.9143e-03, -1.4523e-03, -5.7431e-03,  3.3223e-03,
         5.6673e-03,  7.1755e-03, -4.8549e-03,  5.1383e-04, -9.0348e-03,
        -1.8370e-03, -8.8176e-04,  6.3569e-03,  2.6819e-03, -2.1541e-03,
         1.4534e-03,  2.8440e-03, -1.1645e-02,  3.7244e-03, -4.0997e-03,
         1.7720e-03, -1.6984e-03, -4.2552e-03,  2.9423e-03,  3.9151e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3907e-02,  7.2337e-01,  2.5294e-02,  4.2809e-02,  6.1086e-03,
         2.1508e-02, -1.2866e-02,  1.0441e-02, -6.6662e-04,  3.0419e-02,
         1.1922e-02, -1.3473e-02,  5.6894e-03, -3.2029e-02, -2.9498e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7965e-02,  8.3627e-01, -1.0337e-03,  6.7949e-03, -7.9978e-03,
        -4.7309e-03,  2.0075e-03,  2.0443e-02,  4.9051e-03,  1.6376e-03,
         2.0011e-03, -4.1129e-03,  8.1903e-03,  1.0471e-03,  4.5962e-04,
         1.7852e-03,  1.4196e-04, -1.4076e-02, -3.0811e-03, -5.1982e-03,
         2.6070e-03,  2.2488e-04, -5.0532e-03,  1.1901e-03, -2.5305e-03,
         5.4996e-03, -1.4938e-03, -1.1390e-03, -2.7779e-03,  1.9390e-03,
         3.2012e-03,  8.9103e-03, -9.5597e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.2337e-02, -7.8693e-01,  1.3287e-02,  1.4590e-03, -4.4630e-05,
         1.2406e-02,  8.8311e-03,  9.3157e-04,  1.1226e-02,  1.3814e-03,
         5.2534e-03, -1.3287e-02,  4.9630e-03,  2.4244e-03, -1.3207e-02,
         3.2001e-03,  3.0295e-03,  1.5946e-03, -3.0879e-04, -2.1543e-03,
         1.3748e-02, -1.0457e-03,  1.9504e-03,  5.0517e-04, -1.7602e-03,
        -7.8080e-03,  7.5346e-06,  6.7095e-03,  8.2107e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([-2.1681e-02,  8.9150e-01, -1.0233e-02, -9.1035e-03,  9.5029e-03,
        -7.5897e-03, -2.7812e-03, -8.1830e-03,  3.0046e-03, -9.4746e-03,
         1.0805e-03, -7.6534e-04,  5.8940e-03, -2.7646e-03, -1.6444e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.8247e-04,  8.5833e-01, -1.0206e-02,  7.5214e-03, -1.1644e-02,
         8.4073e-03,  1.0914e-03, -2.9108e-03,  3.8971e-03, -2.3845e-03,
        -2.5597e-03, -4.9419e-03, -2.8156e-03,  1.0489e-03,  8.8248e-04,
        -6.2628e-03, -1.0222e-03,  1.2791e-03, -8.7592e-04, -3.1584e-03,
        -1.5186e-03, -3.7023e-03, -6.0104e-03, -1.2791e-04, -5.1167e-04,
        -8.4917e-06,  4.1156e-04, -1.1293e-04, -4.6571e-03,  1.9579e-04,
         2.2802e-04,  4.2598e-04, -5.1964e-04, -1.2850e-03, -2.8000e-03,
         2.8682e-03, -2.8869e-03,  2.7440e-03, -2.5574e-03, -1.6656e-03,
        -3.3413e-03,  1.7918e-03, -6.6011e-04,  1.4461e-04,  1.6724e-03,
        -1.0119e-02, -4.8118e-03,  1.9042e-04,  9.9087e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.4926e-03,  8.5516e-01,  1.2504e-02, -4.9242e-03,  5.8914e-04,
         1.9798e-03,  4.2554e-03,  2.0999e-03, -1.4692e-02,  4.3666e-03,
         4.0424e-03,  1.7200e-03, -7.6377e-03,  2.3061e-03, -1.0518e-03,
         1.7059e-04,  3.8578e-03,  1.5767e-03,  4.5511e-03,  2.8667e-03,
         3.1821e-03,  3.7884e-03,  4.2388e-03,  5.0104e-03,  4.9621e-03,
        -7.4982e-03, -1.6632e-03,  2.1492e-03,  1.8249e-03, -1.6862e-03,
         3.3584e-03,  3.8713e-03, -4.5830e-03,  7.6324e-03, -1.0709e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5476e-02,  9.0504e-01, -3.0250e-03, -7.0673e-03, -9.7794e-03,
         6.1897e-03, -5.4489e-04,  1.2424e-03, -4.9380e-03, -2.3155e-03,
        -7.8373e-04,  2.4253e-03, -9.0028e-03, -1.3693e-03,  4.1733e-03,
         5.5366e-04,  3.2247e-03, -1.1989e-02,  1.0856e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.7422e-04, -8.6315e-01, -7.4963e-04, -4.1056e-03, -6.0510e-03,
        -1.6679e-02, -3.7273e-03, -3.6010e-03, -9.7251e-03, -3.5243e-04,
         3.1168e-03, -1.9983e-03, -3.9611e-02,  2.6739e-02,  1.9415e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0265,  0.8665, -0.0197,  0.0051,  0.0042, -0.0062,  0.0104,  0.0056,
        -0.0016, -0.0031, -0.0041,  0.0017, -0.0074, -0.0035, -0.0022, -0.0046,
        -0.0039, -0.0048, -0.0078,  0.0040,  0.0013, -0.0055,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2239e-02,  8.3268e-01, -2.0589e-02, -1.4814e-02, -1.2414e-02,
        -1.5315e-02, -8.3675e-04,  9.6201e-03,  4.8237e-04,  7.2999e-04,
        -5.7035e-03,  4.3415e-03, -5.2695e-03, -8.8296e-04, -3.1774e-03,
        -6.4434e-03,  5.3609e-03,  8.3966e-03, -2.4686e-03, -3.6545e-03,
         7.5991e-04,  5.7524e-04,  1.8792e-03,  5.4195e-03, -9.1885e-03,
        -1.3603e-03,  5.3276e-03, -4.5286e-04, -4.7816e-03,  4.8334e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.5429e-02,  7.8592e-01, -8.0774e-03, -2.3711e-02, -2.6788e-02,
        -6.1310e-03,  2.7651e-04,  8.8340e-04, -1.6337e-02, -1.6528e-02,
         1.9917e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7793e-02, -8.6992e-01, -5.0301e-03,  1.0329e-02,  6.5623e-03,
         5.8050e-03,  4.4856e-03,  2.8585e-03, -3.5051e-03, -8.5929e-04,
         4.8894e-03,  9.2970e-03,  4.9151e-03, -9.6602e-04, -6.3107e-04,
         2.6114e-03, -7.9197e-04,  2.7380e-03,  1.4314e-03,  8.1964e-03,
        -2.3265e-02,  3.1163e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2499e-02,  8.1395e-01, -3.2523e-02,  1.7800e-03, -4.5200e-03,
         4.1184e-03, -3.1175e-03,  4.6505e-03, -1.1598e-02, -3.3643e-03,
        -8.9070e-03, -8.6697e-03, -3.2939e-03, -7.8082e-03,  1.8180e-03,
        -3.7865e-03, -7.0100e-03, -1.8190e-03,  3.7424e-03, -1.9603e-03,
         1.3531e-03, -2.6968e-03, -2.3177e-03, -1.1606e-03, -7.4688e-03,
        -4.7266e-03,  1.6179e-03, -3.4384e-05, -5.3790e-03, -1.7905e-03,
         1.3326e-04,  2.2813e-03,  1.8105e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0036, -0.8605, -0.0213,  0.0042,  0.0035, -0.0200, -0.0009,  0.0164,
         0.0010,  0.0010, -0.0009,  0.0030,  0.0038,  0.0038,  0.0066, -0.0102,
        -0.0021,  0.0011,  0.0109, -0.0035, -0.0033, -0.0010,  0.0050,  0.0054,
         0.0026, -0.0014,  0.0016,  0.0013,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3503e-03, -8.7350e-01, -1.1455e-02,  1.6527e-02, -1.2671e-02,
         2.3554e-03,  1.0635e-02,  4.3148e-03,  1.8106e-02,  1.3775e-03,
         1.1127e-02, -3.9402e-03, -4.5322e-05,  1.1586e-03, -5.5645e-03,
         1.3138e-03, -2.9506e-03,  2.6155e-03, -2.0384e-03, -4.9185e-04,
        -2.3349e-03,  1.8789e-04,  7.9262e-04, -4.7389e-04,  8.6705e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 2.7741e-02,  8.5574e-01,  8.3915e-04,  2.8721e-03, -3.0538e-03,
         1.0279e-03, -2.2081e-03, -5.2187e-04,  1.4752e-02,  9.9441e-04,
         8.4348e-03,  2.4970e-03,  4.1793e-04,  9.5042e-04, -3.6568e-03,
        -2.2122e-02, -2.5697e-03,  6.4519e-05, -4.5972e-04,  3.6986e-03,
         1.3944e-03, -2.5164e-03, -1.0692e-03,  9.3578e-04,  4.4214e-03,
        -3.2674e-03, -8.3019e-04,  6.8803e-04, -3.0493e-03, -2.0955e-03,
        -6.2644e-04, -6.2787e-03, -4.5635e-03, -1.1350e-03, -1.4011e-03,
         2.0096e-03,  1.8982e-04,  9.5451e-04,  1.3774e-03, -4.5998e-03,
        -1.9718e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3245e-02,  8.5343e-01,  3.6440e-03, -1.6273e-02,  1.4203e-02,
         5.9498e-03, -6.6047e-03,  3.1050e-03, -3.2690e-03,  3.4938e-03,
         3.0420e-03, -2.2181e-02,  5.5770e-03,  3.2579e-05,  9.0737e-04,
        -9.5274e-04,  1.2590e-03, -3.8839e-03,  2.0380e-03,  5.3135e-03,
        -1.9215e-03,  4.9951e-03,  3.6361e-03, -1.0837e-02,  2.0348e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0064,  0.8807, -0.0066, -0.0100, -0.0087, -0.0035, -0.0023, -0.0019,
        -0.0030,  0.0125,  0.0026,  0.0070, -0.0022,  0.0060, -0.0068, -0.0045,
         0.0030,  0.0037, -0.0086, -0.0056,  0.0144,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0863e-02,  8.3105e-01,  3.2879e-02, -1.9389e-02,  2.8441e-03,
         3.1060e-03, -7.8655e-03, -1.6798e-03,  1.9077e-03,  4.3446e-05,
        -7.5426e-03, -1.2015e-02,  5.0360e-03, -9.0970e-03, -7.3975e-03,
         5.3653e-03, -1.4739e-03,  1.0441e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0937e-02,  8.0547e-01, -1.9945e-02,  5.1425e-03, -8.5459e-03,
         7.1740e-04, -2.0668e-03, -6.2380e-03, -4.9460e-03, -3.2748e-03,
        -3.7018e-03, -3.1636e-03,  3.4508e-03,  2.5255e-03, -1.7433e-03,
         1.1833e-02,  1.7712e-03,  8.2494e-03, -2.0884e-04, -5.8133e-03,
         9.2164e-04,  3.1797e-04,  6.3918e-04,  1.5920e-03,  1.0481e-02,
        -2.7410e-03, -5.0971e-03, -1.7894e-03, -8.1031e-03, -5.1886e-03,
         5.6507e-03,  1.7229e-03,  5.8533e-03,  1.4448e-03,  2.8426e-03,
         4.0157e-03, -2.7530e-03,  2.9104e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2616e-02, -8.6877e-01,  3.8113e-02, -8.2917e-03,  4.6365e-03,
        -1.1313e-03, -3.3839e-04,  5.2247e-03,  2.7629e-03,  4.4184e-03,
        -6.9973e-03,  2.6362e-03, -2.4310e-03, -4.2400e-03,  2.5658e-03,
        -1.9736e-03,  2.1242e-03,  1.9409e-02,  2.7800e-03, -8.5349e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.4814e-03,  8.2749e-01,  2.1734e-02,  1.1376e-02,  3.9682e-02,
         1.2358e-02,  3.1301e-03,  2.1403e-02,  8.7970e-03,  8.4915e-03,
        -1.8801e-02, -6.3460e-03, -1.0274e-02, -6.3143e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1303e-02,  8.3039e-01,  1.8011e-02,  2.5954e-02, -3.2135e-03,
        -7.6315e-04, -2.9270e-03,  3.4746e-04, -9.0978e-04, -6.0153e-03,
        -5.6436e-03,  7.4446e-04,  2.1209e-03,  5.1929e-03,  7.7302e-03,
         5.6103e-04, -6.0967e-03, -3.1809e-03,  1.0506e-03,  2.0110e-03,
        -9.3827e-04,  2.0494e-02,  2.4691e-03,  1.2264e-02, -2.7163e-03,
        -1.4384e-04,  4.8698e-03,  1.1345e-03, -9.9595e-04,  4.5648e-03,
        -5.7090e-04, -4.6746e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.2363e-02, -8.7409e-01,  7.6387e-04, -1.8252e-03,  3.4795e-03,
        -2.2301e-03, -4.7967e-03,  1.8310e-03,  7.3293e-03,  3.2741e-03,
         2.4942e-03,  4.3287e-03, -3.1312e-03, -3.9235e-04, -4.6104e-03,
         9.2596e-04,  3.6010e-03,  3.2146e-03,  1.5206e-02, -1.0109e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.8590e-02,  7.4975e-01, -3.3960e-02, -2.2962e-03, -1.2619e-03,
        -4.2780e-03, -4.8830e-03, -1.0030e-02, -1.4092e-03,  3.0723e-03,
        -4.3461e-03, -1.1903e-04, -6.8572e-04,  1.2402e-03,  3.1894e-04,
        -5.7356e-04, -2.7276e-03,  8.3623e-03, -3.9039e-03,  1.2582e-03,
         4.0668e-03, -5.4763e-03, -5.4284e-04,  6.0832e-03, -7.2770e-03,
        -9.3610e-03, -1.4263e-02, -4.6990e-03,  2.9852e-03, -3.1232e-03,
         7.8870e-04,  1.1866e-03,  6.7075e-03, -1.4290e-03,  7.7870e-03,
         7.5398e-06, -5.2792e-03,  8.8262e-05, -4.0512e-04,  1.5006e-03,
        -5.3228e-04,  5.6489e-03,  4.3122e-03, -1.2189e-03,  7.4181e-04,
        -1.5412e-02,  6.0093e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.3978e-03,  8.4671e-01, -1.2930e-03,  6.7463e-03,  3.4129e-02,
         2.1165e-03,  7.4444e-03, -2.6033e-04,  7.1899e-03, -9.1176e-03,
        -2.2378e-02,  1.1234e-02,  1.3829e-02,  1.9356e-02, -8.7973e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.7305e-03,  7.8831e-01,  1.4310e-02,  3.4578e-03,  3.3868e-03,
         1.6912e-03, -5.2668e-03,  1.6598e-02,  1.8709e-04, -2.2513e-03,
        -7.3977e-04,  7.8821e-04, -5.1004e-03,  1.0889e-02, -8.0567e-03,
         7.0378e-03, -4.5327e-03,  1.8574e-03, -1.0297e-02, -5.7116e-03,
         1.0874e-04, -3.4555e-03,  1.2036e-03,  1.8351e-03, -2.5614e-03,
         7.2173e-04,  1.1107e-02, -3.5882e-02, -3.4205e-03, -2.5276e-03,
        -2.3924e-03,  1.6624e-03,  1.4974e-03,  1.3006e-04,  5.2970e-03,
         4.0016e-03,  1.9024e-03,  1.8653e-03, -5.0709e-03,  8.8625e-04,
        -1.4112e-03,  2.3633e-03, -1.8291e-03, -2.1863e-04, -9.0580e-04,
        -2.0270e-03,  4.6882e-03,  1.8320e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 4.1689e-02,  8.1519e-01,  9.2996e-03, -8.7818e-03,  8.3087e-03,
        -1.6335e-02,  2.0119e-03,  1.4636e-03, -7.8506e-03, -1.0563e-02,
         1.6365e-03, -1.5582e-03, -7.7347e-03,  5.9174e-04, -2.4901e-03,
        -2.2334e-03, -1.8017e-02,  1.4142e-03,  1.3134e-03, -2.7134e-04,
         2.0358e-03,  2.1222e-03,  1.2032e-03, -3.4584e-03,  2.1554e-02,
         1.0877e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3517e-02, -8.3565e-01,  1.0447e-02, -2.1291e-03, -2.1799e-03,
        -1.2783e-03,  7.0781e-03, -7.4693e-03, -6.2785e-03, -3.6050e-03,
        -1.5463e-02,  4.8846e-03, -2.1985e-04, -2.6517e-03,  3.9496e-03,
         4.9524e-03, -8.4312e-03,  2.6438e-03, -1.0859e-02,  7.6567e-03,
        -8.6581e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0675e-02, -8.3525e-01,  5.0223e-04, -1.3500e-02,  1.1753e-02,
         5.9457e-03,  3.2166e-03, -4.6043e-04,  5.5533e-03,  6.0936e-03,
         4.5191e-05, -3.9945e-04,  8.7454e-03,  1.3059e-03, -8.1039e-03,
        -6.8051e-03,  5.5424e-03, -2.8877e-03, -6.1043e-03, -3.1650e-03,
         1.1201e-03,  1.8759e-03, -6.4424e-03,  2.8484e-03, -9.1847e-03,
         1.5441e-03,  4.2666e-04,  1.5948e-03, -8.4747e-04, -1.7348e-03,
         2.0275e-03, -1.4524e-03, -1.0662e-02,  3.6591e-03,  3.7888e-03,
         4.7359e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0022,  0.7932, -0.0220, -0.0192,  0.0042,  0.0267, -0.0263,  0.0043,
        -0.0106,  0.0113,  0.0123,  0.0048, -0.0040, -0.0127, -0.0038, -0.0049,
         0.0053, -0.0059,  0.0034, -0.0232,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6982e-02, -8.2044e-01,  4.0927e-02, -1.2309e-02,  2.9846e-03,
         4.4629e-03,  3.8793e-03, -8.2921e-03,  2.2989e-03,  5.9363e-03,
        -2.2955e-02, -1.0615e-02,  9.8400e-03,  7.6578e-03, -1.2301e-02,
        -5.5831e-04,  3.5748e-04, -4.9054e-03,  9.6014e-03, -2.6001e-04,
        -2.4330e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0513, -0.8138,  0.0208,  0.0040,  0.0106,  0.0069,  0.0129,  0.0099,
        -0.0030,  0.0038,  0.0035, -0.0147, -0.0449,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2827e-02, -8.5510e-01, -2.4593e-02, -4.4356e-03, -2.0047e-03,
         5.0732e-03,  5.6341e-03,  3.8686e-03,  1.4410e-02,  1.9881e-03,
         1.8172e-03,  2.7106e-03,  1.5223e-03,  2.3838e-04, -3.9155e-03,
         4.7874e-03, -5.9232e-04,  1.1194e-03,  3.4633e-03,  2.1945e-03,
        -5.9521e-03, -3.1378e-03,  1.5102e-03,  1.8358e-03, -3.4444e-03,
         3.2613e-04, -6.3963e-04, -2.7053e-03,  2.2688e-03,  3.0608e-03,
        -6.1921e-05, -1.3214e-03, -6.3196e-03,  2.3189e-03, -7.6958e-03,
         4.7625e-03,  3.4351e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1422e-03, -8.7370e-01,  1.7744e-02,  2.7488e-04,  1.7068e-02,
         5.2375e-03,  4.1415e-03,  3.9805e-03,  4.0555e-03,  4.1855e-03,
         5.6752e-03,  9.4412e-05, -3.1594e-03, -2.6101e-03,  1.2125e-03,
         6.2505e-03,  2.0405e-03, -4.5625e-03,  1.2148e-03, -1.7831e-03,
         4.6320e-03,  6.9753e-03,  2.8112e-03, -2.2445e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9712e-03,  8.4565e-01,  3.6460e-02, -5.4850e-03,  7.1439e-03,
        -8.3796e-05, -3.2597e-03,  1.2855e-02,  2.2228e-03,  8.0366e-05,
        -2.7498e-03, -5.2409e-03,  2.6995e-03, -4.7051e-03,  7.4086e-03,
        -2.4642e-03, -4.7528e-04,  1.8380e-03, -6.6411e-03,  2.7363e-03,
         4.6877e-05, -3.8577e-03,  1.2637e-03, -1.3051e-03, -3.4352e-04,
        -9.1161e-04, -2.3296e-03, -1.2648e-03,  3.5353e-03,  5.7547e-03,
        -1.0975e-02, -1.5237e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6703e-02, -7.9515e-01, -1.4135e-02,  9.7878e-03,  1.2110e-02,
         1.6227e-02,  2.5642e-04,  3.4431e-03, -8.3984e-03, -2.7802e-03,
         1.5937e-03,  2.8573e-03,  1.3178e-03, -1.8281e-03,  2.7234e-03,
        -6.1892e-03,  6.2596e-03,  3.2362e-03,  1.3991e-04, -1.2234e-03,
        -4.7014e-03, -2.7784e-03, -8.7418e-05,  1.6866e-04, -5.0329e-04,
         3.9228e-03, -1.8409e-02,  1.6947e-03, -7.0208e-03, -1.5673e-02,
         1.8678e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1240e-03,  8.2694e-01,  3.6256e-02,  1.1512e-02, -2.4904e-03,
        -1.1859e-02,  3.3201e-03, -8.7501e-03,  1.2676e-02,  7.3380e-04,
         2.1176e-02, -5.4181e-03, -4.7409e-03, -1.6255e-02, -4.7279e-03,
         6.1268e-03, -6.1291e-04, -4.3735e-03, -9.5436e-04,  6.1799e-04,
        -1.1953e-02,  5.0099e-03,  1.3703e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4783e-02, -8.1885e-01, -7.3342e-03,  1.5544e-02, -7.9393e-03,
         7.0949e-04, -1.5918e-03,  1.3763e-02, -2.9002e-04,  1.2446e-03,
         1.6745e-03,  1.7491e-02,  5.6636e-03,  8.4581e-03, -6.0720e-04,
        -1.3085e-03,  5.4558e-03,  5.8764e-03, -6.4235e-04, -4.2577e-03,
         2.2752e-03,  5.2305e-03, -1.2564e-02, -3.3185e-03, -8.3079e-04,
         2.4956e-03, -3.8190e-02, -1.6061e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 1.0750e-02, -8.5494e-01,  1.0016e-02, -7.8101e-03, -7.5621e-03,
        -4.6190e-03, -5.0768e-03,  2.4790e-03,  2.3446e-03,  5.5498e-03,
         5.7991e-03,  2.7162e-03, -2.7306e-03, -1.4732e-03, -1.9826e-03,
        -5.5814e-03,  2.4871e-03, -2.1931e-02,  3.7433e-03, -6.8133e-04,
        -8.7282e-03, -1.0669e-03,  9.6654e-03, -6.1784e-04,  4.5284e-03,
        -3.4602e-03,  5.4202e-04,  8.4586e-03, -2.6549e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.7865e-02, -7.9196e-01, -9.6587e-04, -9.4043e-03, -1.6517e-02,
        -7.9467e-03,  1.1501e-03,  6.2972e-03, -1.3486e-02, -1.6116e-02,
        -9.3826e-03,  2.1679e-03,  1.5101e-03,  5.3227e-03, -1.3346e-03,
        -3.7806e-03, -7.4535e-03,  2.2599e-03, -1.8822e-03, -1.9189e-03,
         9.3300e-04,  7.2873e-03,  2.3039e-02,  1.5614e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5360e-02, -7.8606e-01,  5.6417e-02,  2.0760e-02, -7.4238e-03,
        -7.2539e-04, -3.7364e-03,  2.7736e-03, -7.9033e-03,  2.2755e-03,
         5.7397e-04, -6.2127e-03, -1.4875e-02, -1.2575e-03,  3.4958e-03,
        -2.3863e-03, -4.0792e-03, -1.6540e-04, -7.4387e-03, -8.7237e-04,
        -3.7697e-03,  6.7408e-03,  1.9628e-03, -4.6000e-03, -6.7260e-03,
         2.7323e-03,  1.5089e-03, -2.4410e-05,  2.1408e-03, -2.2280e-03,
        -5.4372e-03,  4.6432e-03, -1.1539e-02, -1.1505e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.1925e-03,  8.3922e-01, -5.0606e-02, -1.4215e-02, -8.1770e-04,
         8.7399e-03,  1.5098e-02,  7.9202e-03,  4.2939e-03,  1.7692e-03,
         1.2718e-03, -6.2080e-03, -2.3913e-03, -2.9668e-03, -6.1026e-04,
         2.1761e-03, -3.4727e-03, -3.5762e-03, -6.1276e-03,  2.5265e-03,
        -8.0238e-03,  1.3777e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3598e-02,  8.7708e-01,  1.1425e-02,  3.8425e-03,  2.2169e-02,
        -1.4788e-02, -1.0153e-03,  5.3993e-04, -6.6361e-03, -1.0982e-03,
         1.1886e-02, -7.6637e-03, -3.2723e-03,  3.1943e-03, -2.7215e-03,
        -8.3702e-03, -8.2771e-04,  9.8742e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4314e-02, -8.4037e-01,  3.1268e-05, -4.4295e-03, -1.4124e-03,
         3.6382e-03, -4.8470e-03,  5.3187e-03, -1.1056e-02,  1.4660e-03,
        -1.9737e-03, -2.6415e-03,  1.4779e-02,  1.8722e-04,  3.2416e-03,
         3.9279e-03, -7.6613e-04,  1.4460e-02, -1.5407e-03,  4.9634e-03,
         7.5016e-03, -1.3762e-02,  2.3375e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.1253e-02, -8.5538e-01, -1.2472e-02, -1.0621e-02,  1.6472e-02,
         2.8339e-03,  9.8799e-04,  2.7191e-03, -2.7812e-03,  1.0314e-03,
         1.0040e-02,  2.3910e-04,  1.5286e-03, -7.5484e-03,  9.2960e-03,
         8.8023e-03,  5.9909e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1069, -0.8071, -0.0079, -0.0024, -0.0041, -0.0039, -0.0041,  0.0086,
        -0.0032,  0.0022,  0.0025,  0.0035,  0.0048, -0.0012,  0.0049, -0.0116,
         0.0114, -0.0098,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3353e-02, -7.4507e-01,  4.8064e-02,  1.5033e-02, -2.3707e-03,
         4.7466e-04, -3.3927e-03,  4.9768e-03,  4.2301e-03, -2.5600e-02,
        -3.1441e-03,  7.7333e-03, -8.6446e-03, -2.8706e-05,  1.2775e-03,
         5.4176e-03,  2.2801e-02, -1.0037e-02, -5.8749e-03, -4.2278e-03,
         6.0454e-03,  2.5374e-03,  2.6121e-03,  3.1964e-03,  2.9557e-03,
         3.0897e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5131e-02,  8.5148e-01,  3.2639e-04,  5.5676e-03, -3.0238e-03,
        -8.8224e-03,  1.9621e-03,  3.0774e-03,  8.5150e-03, -8.5293e-03,
        -2.1950e-03,  2.0155e-04,  3.5919e-04, -4.4609e-03, -4.8608e-03,
         1.7908e-03,  2.6005e-03,  4.1370e-04,  4.6555e-03, -3.2747e-04,
         2.9038e-03, -4.8984e-03,  4.0717e-03, -2.3276e-03, -9.3730e-04,
         1.6780e-03, -3.2232e-04, -2.1805e-03,  3.1970e-03,  1.0879e-02,
        -3.8303e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0197, -0.9009,  0.0026,  0.0040,  0.0030, -0.0022,  0.0024, -0.0032,
        -0.0063, -0.0018,  0.0020,  0.0101,  0.0058,  0.0056, -0.0018, -0.0029,
         0.0019,  0.0238,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0138,  0.9013, -0.0130, -0.0219, -0.0021, -0.0051, -0.0058,  0.0370,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([-1.2420e-02,  8.4921e-01,  1.7795e-02, -4.2721e-03, -1.3026e-02,
         5.1574e-03, -5.1258e-03, -3.3616e-06,  1.1277e-04,  5.8772e-03,
         5.6482e-04, -9.6913e-03,  1.7380e-03, -2.3119e-03,  1.3216e-03,
         1.5410e-03,  1.5970e-03, -2.3853e-03,  3.2980e-03,  3.7500e-03,
         3.7823e-03,  3.9005e-03,  1.1421e-03, -4.9648e-03, -6.3507e-04,
         2.3736e-03, -3.6781e-03,  3.3291e-03,  5.2776e-04,  4.9145e-04,
         4.8397e-04,  2.0414e-03,  2.0375e-03,  4.5829e-03,  1.5925e-03,
         4.8313e-03,  1.8669e-03,  9.6511e-03,  6.8915e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3536e-02, -8.6982e-01,  9.9680e-03, -1.4711e-03, -7.0171e-04,
         1.7669e-03,  8.1964e-03,  5.9255e-03,  2.3597e-03, -3.5756e-03,
         1.1982e-03, -5.1368e-03, -4.4446e-03,  9.0697e-03, -1.4151e-03,
         1.1188e-04, -2.5925e-04,  8.8764e-03, -4.1410e-03, -8.4890e-03,
        -5.0897e-03, -5.1123e-03,  2.0453e-03,  4.7857e-04,  5.9709e-04,
        -2.7884e-03,  6.9561e-03,  6.4707e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.8207e-02,  8.2772e-01, -2.8653e-02,  7.3718e-03, -1.3926e-02,
         1.0831e-02, -1.2054e-02, -1.1911e-04,  3.5777e-03,  3.2664e-03,
         5.4388e-04, -1.2413e-02, -1.3207e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1733e-03, -9.0882e-01,  1.5594e-02,  1.1330e-02,  9.1580e-04,
        -6.6926e-03,  3.6300e-03, -6.2648e-03, -8.1762e-03, -6.4739e-03,
        -7.2753e-04,  1.3121e-02,  4.9744e-04,  2.2409e-04,  1.1363e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4893e-02, -8.1004e-01, -2.6330e-02,  6.8682e-03, -1.0818e-02,
         4.0479e-03, -8.2583e-03, -2.1529e-03,  6.5574e-03,  4.9247e-03,
         1.2263e-03,  2.2600e-04,  1.1219e-02, -1.1511e-02,  1.1359e-03,
         1.8993e-03, -2.6191e-03, -7.3674e-04,  4.6288e-05,  1.9299e-03,
         1.3068e-03,  2.6505e-03, -1.1662e-03,  2.8955e-03, -3.7035e-03,
        -8.1819e-03, -2.5262e-03, -5.2860e-03,  4.1129e-03, -6.8993e-04,
         1.2898e-03,  2.0520e-03,  2.1914e-03,  7.6554e-04, -8.0901e-03,
         7.9526e-04,  1.0748e-03, -8.6651e-04,  2.9127e-03,  2.9485e-04,
        -2.5459e-04,  3.7479e-03, -1.2593e-03,  2.4306e-03, -2.1505e-03,
        -1.8146e-05,  9.8429e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9001e-03, -8.8295e-01, -1.0894e-03,  8.1616e-03,  1.2378e-02,
         1.1410e-02,  1.8610e-03,  4.5340e-03,  4.7840e-03,  4.7062e-04,
        -1.7718e-03, -2.9769e-03, -9.8255e-04, -1.2365e-03,  1.0439e-05,
         8.0123e-04, -1.5409e-03, -2.4794e-03, -1.8969e-03,  9.4774e-04,
        -1.7748e-03, -3.4921e-03, -1.8139e-03, -1.7271e-03, -2.2992e-03,
         4.7051e-03, -2.6681e-03, -1.8181e-03,  1.5967e-04, -4.4059e-03,
         4.7468e-07, -2.5918e-03, -4.1736e-04, -4.8845e-03, -1.1155e-03,
        -3.1562e-03,  8.9660e-05, -3.6190e-03, -2.0279e-04,  2.3729e-03,
        -5.4967e-03,  6.0047e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7263e-04,  8.5182e-01,  2.1757e-02, -1.0150e-02, -6.0243e-03,
         1.8433e-03,  3.1123e-02,  8.1307e-04,  7.4499e-04,  6.9693e-03,
        -2.2018e-04,  7.9421e-03,  3.9183e-03,  3.3342e-03,  7.4336e-03,
        -8.1597e-04,  1.6646e-02, -9.4322e-05,  2.6232e-03,  2.1024e-03,
        -6.3047e-03, -2.2667e-03, -9.8003e-04, -2.6327e-03, -1.1173e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1705e-02,  8.4965e-01, -2.3905e-02, -3.6398e-03,  2.4263e-03,
        -1.3085e-02,  1.6693e-02, -3.7572e-03, -8.7762e-03, -2.9125e-03,
        -1.1566e-03, -2.7651e-04,  5.2690e-03, -1.3332e-03, -2.8892e-03,
        -3.1039e-03,  1.8927e-03,  1.1788e-02, -5.6423e-03, -2.6698e-03,
        -4.4063e-04,  1.2698e-02,  4.2879e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2779e-02, -8.1128e-01, -7.1799e-02, -9.8897e-03,  1.5222e-02,
        -3.0738e-03, -9.1949e-03,  1.7931e-03,  6.4452e-04, -2.3442e-02,
        -2.9876e-03,  6.7969e-04,  5.3066e-04, -3.2122e-03, -2.8598e-03,
         2.0782e-03, -4.5104e-03,  6.6257e-03,  1.0083e-03, -5.5175e-04,
         8.0262e-03, -6.1470e-03,  1.6659e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.0870e-02,  7.9004e-01,  3.4874e-03, -2.5201e-02,  4.7209e-04,
         1.5275e-03,  2.4612e-02,  2.6528e-02,  1.4584e-02,  5.1391e-03,
        -7.8637e-03, -2.7549e-02,  1.2123e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0028,  0.8992, -0.0041, -0.0154,  0.0058, -0.0010, -0.0023,  0.0130,
         0.0263,  0.0055,  0.0047,  0.0017, -0.0040, -0.0142,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.9878e-03,  8.9134e-01, -2.2931e-02, -3.2006e-03,  8.3347e-03,
         2.6024e-03,  4.3854e-03,  1.8368e-03,  1.0062e-03,  1.0423e-03,
        -1.2798e-03,  4.0334e-03,  7.9650e-03,  6.7100e-04, -1.3605e-02,
         1.3480e-03,  1.5623e-03,  1.8041e-02,  7.8288e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 3.6797e-02,  8.3569e-01, -9.8777e-04, -1.4985e-02, -1.0198e-02,
         8.0890e-03,  2.9112e-03, -4.0620e-03, -3.5324e-03,  4.1488e-03,
         1.7934e-03, -6.2287e-03, -4.7804e-04,  6.3054e-03, -1.8008e-03,
        -2.1071e-03,  1.9678e-03,  6.0046e-03,  7.7306e-04,  1.0902e-02,
        -5.3106e-03,  4.8988e-03,  5.0759e-04,  2.1281e-03, -1.5838e-02,
        -1.1555e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2877e-02,  8.5156e-01, -1.0415e-03, -1.6954e-02, -1.1574e-02,
        -4.5860e-03, -8.6139e-03, -1.2765e-02,  6.1306e-04,  1.7620e-02,
         5.7411e-03,  1.8341e-03, -1.0222e-02,  6.0228e-03,  7.4512e-04,
        -7.2295e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2376e-02,  8.3994e-01,  2.0068e-02,  1.7067e-03, -8.9434e-03,
         4.1230e-03,  1.1023e-02,  4.0257e-03,  1.6197e-02,  1.2769e-03,
         5.9307e-03, -2.0107e-02,  3.6689e-03, -1.2749e-04,  9.4345e-03,
         6.2615e-03, -4.7887e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0615,  0.8796,  0.0333, -0.0053,  0.0038, -0.0049,  0.0028,  0.0051,
        -0.0037,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1174e-02,  7.5685e-01,  1.3633e-02, -5.1387e-03, -1.4465e-02,
         7.9254e-03,  5.0324e-03,  9.0328e-04, -4.0628e-04, -3.5148e-03,
         4.0689e-03, -3.3447e-03, -1.7414e-03, -3.1215e-03,  1.9459e-03,
         6.0435e-03, -3.6083e-03, -1.3639e-03,  2.8368e-03,  2.2637e-03,
        -2.4243e-04,  2.4879e-04, -7.3563e-03, -2.0697e-03,  2.8962e-03,
        -9.8608e-03,  6.5286e-04, -3.2751e-03,  3.1527e-03, -3.0184e-03,
         4.4209e-03, -2.2916e-03,  4.7485e-03, -5.7813e-04,  3.3968e-03,
         9.3334e-04,  2.6654e-03, -3.0844e-03, -1.2157e-02, -1.3583e-03,
        -9.0158e-03, -1.7370e-02, -1.1219e-02, -6.6537e-03,  5.3116e-03,
         2.6370e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1361e-02, -8.2462e-01,  1.8016e-02,  1.4900e-02,  7.2597e-03,
        -4.6625e-03,  2.0876e-03, -7.4980e-04,  5.4939e-03, -4.5354e-03,
         1.0222e-03, -1.4869e-03, -2.1667e-02, -3.0320e-03,  4.2150e-04,
        -4.7991e-03, -1.9093e-03, -1.0260e-02,  4.1357e-03, -9.6099e-03,
         1.0502e-02, -2.0867e-03, -3.0583e-03,  3.7105e-03,  3.9949e-03,
         6.7171e-03,  7.1363e-03,  3.2013e-03, -4.2677e-03, -3.2959e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0329, -0.9298,  0.0110, -0.0048,  0.0047, -0.0076, -0.0091,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0120,  0.9048, -0.0064, -0.0081,  0.0097, -0.0166,  0.0041,  0.0040,
        -0.0258,  0.0025, -0.0061,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.8595e-03,  9.1407e-01,  3.0325e-02,  7.3784e-03, -3.5230e-03,
        -2.6768e-03,  3.1994e-03,  1.9559e-03,  1.0132e-02, -4.6200e-05,
        -2.2197e-03,  4.8181e-03, -1.0919e-03, -1.4709e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0725e-02, -8.0485e-01, -1.5483e-02,  1.7211e-03, -7.0120e-03,
         4.3086e-03, -5.5980e-03,  1.6006e-02,  1.1878e-02,  1.3194e-03,
        -1.3972e-03,  3.2219e-04,  2.3949e-03, -1.8972e-03,  1.3646e-03,
        -5.2288e-03, -1.5965e-03, -7.2309e-03,  9.1167e-03, -8.3958e-03,
        -2.6458e-02, -1.9375e-02, -4.1559e-05,  2.1046e-03,  1.5775e-03,
        -1.7615e-03,  1.7361e-03, -4.7248e-03, -1.9582e-03, -3.0752e-04,
        -5.3496e-04,  1.1009e-03,  6.6577e-03, -1.3812e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.4046e-03,  8.1772e-01,  5.2265e-02, -1.7606e-04, -1.7691e-03,
        -5.2998e-04,  2.0251e-03, -6.3636e-03, -3.5259e-03, -1.7871e-03,
         8.4582e-03, -4.9301e-03,  8.9938e-03, -3.0494e-03,  6.8670e-03,
        -4.0274e-03, -3.5216e-03,  1.8120e-02, -1.6517e-03,  2.0344e-03,
        -3.9607e-03, -5.7332e-03, -6.9623e-03, -1.0948e-03, -2.1244e-03,
        -3.0840e-03,  1.5068e-02,  5.7531e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2930e-02,  8.4406e-01,  9.8893e-03, -2.3606e-02, -7.1187e-04,
         1.5400e-02, -1.6708e-03,  4.4736e-03,  7.0824e-04, -2.6177e-03,
         1.4862e-03,  2.9530e-04, -1.1198e-02,  8.4488e-03, -5.0853e-03,
        -7.4665e-03,  6.8677e-03,  1.3501e-02,  1.0490e-02,  2.2081e-03,
        -2.5824e-03,  1.8466e-03,  7.2214e-04, -1.1734e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 2.3340e-02,  7.9622e-01,  4.2352e-02, -6.6215e-03,  1.0657e-02,
        -6.1735e-03,  1.7184e-03,  1.5897e-03, -5.5186e-03, -2.3205e-03,
        -6.3784e-03, -3.5915e-04, -3.9346e-03,  3.7638e-03,  2.4020e-03,
        -2.1666e-03, -1.3531e-03, -7.8582e-04,  5.4923e-03,  1.8505e-02,
         3.1847e-03,  2.2273e-04, -1.2784e-03, -2.8228e-03,  1.9045e-04,
         1.0205e-02, -8.3823e-03, -5.2088e-03,  1.1002e-04, -1.0505e-02,
        -1.6237e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0615,  0.8075, -0.0230, -0.0247, -0.0201,  0.0015,  0.0083,  0.0054,
         0.0155,  0.0142,  0.0070,  0.0114,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5589e-03,  8.7696e-01, -1.1502e-03,  6.0563e-03,  1.7074e-02,
         8.9418e-03,  4.2500e-03, -8.1015e-04, -6.0136e-03, -1.1300e-02,
         3.3194e-03, -1.9708e-02,  2.2987e-02, -5.8678e-03, -1.1999e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.9646e-03,  8.7098e-01,  1.6283e-02, -4.0285e-03,  1.8507e-02,
         1.9090e-03,  2.3272e-03,  1.3699e-03,  5.0893e-04,  4.0391e-03,
        -2.5610e-03, -5.6623e-04, -6.1127e-03,  3.3521e-04,  1.5823e-03,
        -3.0870e-04,  2.7891e-03, -1.3315e-03,  4.5097e-03, -4.4403e-03,
         2.4404e-03,  1.9232e-03, -2.9075e-03,  1.6640e-03, -8.0370e-04,
        -2.0411e-03,  2.7991e-03, -5.8664e-04, -4.4007e-03,  5.3097e-03,
        -1.7562e-04, -1.5411e-02,  8.0834e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.3094e-03, -8.7925e-01, -2.0813e-03,  1.7163e-02, -1.6857e-03,
         4.0759e-03,  1.2643e-02,  1.3812e-02,  1.0388e-02,  1.7953e-03,
         1.3650e-02,  3.4685e-03,  8.6552e-03, -2.1917e-04, -1.4888e-02,
        -9.9178e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9912e-03,  9.2735e-01, -1.2416e-02, -9.5459e-04, -2.3685e-02,
        -7.0651e-03, -6.7678e-03, -1.8989e-02, -7.8425e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0789,  0.8798,  0.0067, -0.0103, -0.0243,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.2007e-02, -8.3357e-01, -4.4910e-03,  6.7726e-03, -1.1874e-04,
         4.7766e-03,  1.6002e-03, -1.0371e-04, -1.0277e-02, -1.7521e-02,
         2.5587e-03,  9.3132e-03,  6.3736e-03,  3.6122e-03, -9.5107e-04,
         5.3174e-03, -2.7858e-03,  2.4029e-03,  7.6843e-04, -2.8775e-03,
         1.4323e-03, -3.4340e-04, -4.5765e-03,  5.9166e-03,  1.2199e-03,
         8.2414e-03, -6.9074e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1023,  0.8359,  0.0110,  0.0120,  0.0220,  0.0128,  0.0041,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2965e-02,  8.5385e-01, -2.3151e-02, -5.3714e-03, -4.9742e-03,
        -7.3319e-03, -4.1227e-03, -2.9486e-03, -1.4008e-02, -1.1430e-02,
        -2.7459e-04, -2.8742e-03,  2.2026e-03,  8.4037e-03, -8.2399e-04,
        -1.1100e-03, -3.9424e-05, -1.8130e-03,  7.1633e-03, -1.2166e-02,
        -6.9199e-03,  1.6054e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.7657e-03, -8.1005e-01, -8.2265e-02, -1.1523e-02, -1.1917e-02,
        -5.9125e-03,  3.6258e-03, -1.8099e-02, -1.0592e-03,  1.3109e-02,
        -3.2099e-03, -3.6251e-03,  7.2687e-03, -1.2268e-02,  1.4190e-03,
        -1.5171e-03,  2.7543e-04, -4.9737e-03, -1.1611e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.2965e-02,  8.2129e-01,  1.5122e-02, -1.2267e-02, -5.8160e-03,
         1.1062e-03, -1.0417e-02,  5.0345e-03,  6.2581e-03, -3.7052e-03,
        -3.5032e-03,  2.2646e-03,  3.0605e-04,  1.8357e-03, -1.4710e-03,
        -1.6488e-03,  6.6423e-04, -1.2082e-03, -2.1658e-04, -1.8772e-03,
         7.5395e-03,  7.4462e-03, -9.3883e-05, -3.6743e-04,  1.1377e-03,
         4.8153e-04, -3.9816e-03,  9.9694e-04,  4.2830e-03, -1.6143e-03,
        -1.3083e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 1.4082e-02, -8.1425e-01, -3.5588e-02,  5.8513e-04,  2.1039e-03,
         6.4322e-03,  5.6366e-03,  5.5365e-03,  6.8997e-03,  3.5196e-03,
         2.1660e-03,  2.0866e-03, -7.7957e-03, -1.3907e-02, -4.5213e-03,
         1.1419e-03, -9.1423e-03, -2.5084e-03, -7.4120e-03, -4.4453e-04,
         6.5587e-03,  5.0808e-03, -3.0075e-03,  3.7859e-03, -8.0232e-03,
        -2.7331e-04,  3.0753e-03, -6.1806e-03,  3.0029e-03, -6.7799e-03,
        -3.0181e-04, -8.1710e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.0137e-04,  8.7100e-01, -1.9006e-02,  4.7551e-04, -1.3294e-02,
         4.9208e-03, -2.7294e-03,  1.8904e-03,  2.8901e-02,  9.7615e-04,
        -2.4153e-03,  9.0669e-03, -4.2443e-04, -9.1565e-03,  4.2052e-03,
         1.0925e-02,  2.5507e-03,  4.1485e-03, -2.8585e-03,  2.8994e-03,
         5.9471e-03, -1.5100e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3036e-03,  8.3059e-01, -1.6258e-02, -2.0196e-02, -4.0697e-04,
        -4.1745e-03, -1.3017e-02, -6.5124e-03, -2.1108e-03, -6.1994e-03,
         5.6314e-03,  4.7320e-03,  1.5248e-02, -1.7709e-04, -1.6576e-03,
         6.5092e-04, -2.9132e-03,  6.4166e-03, -1.4298e-03, -1.0502e-03,
         1.3224e-03, -2.0968e-03, -4.3829e-04, -2.6386e-03, -3.2850e-03,
         3.9695e-03, -2.0350e-03,  7.7653e-03, -1.6644e-02,  1.7127e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0658,  0.7974,  0.0079, -0.0074, -0.0116, -0.0116,  0.0046, -0.0130,
        -0.0132, -0.0046, -0.0134,  0.0073, -0.0095,  0.0098,  0.0028, -0.0039,
         0.0145, -0.0016,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5381e-02,  7.9262e-01,  8.0516e-02,  4.2331e-03, -1.1043e-02,
        -1.3194e-02, -4.7481e-03,  3.4024e-03, -6.9398e-04, -5.8658e-04,
        -1.1584e-03,  2.2909e-03, -5.2502e-05,  4.8373e-03,  8.1689e-04,
        -6.8557e-03, -1.6051e-04,  2.8219e-03, -2.6736e-03, -2.2396e-03,
        -1.0398e-03, -2.7796e-03,  1.4015e-03,  2.5422e-03, -2.2023e-03,
         1.0209e-03, -2.7529e-03, -9.8875e-04,  4.9162e-04, -6.7922e-04,
        -1.3932e-03,  3.9066e-03,  3.0513e-03, -4.6877e-03,  5.0220e-04,
         5.6031e-03,  4.6299e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5682e-02,  8.5197e-01, -2.2038e-02, -1.1042e-02,  5.5099e-03,
         4.8424e-03,  5.9106e-03, -5.2068e-03, -4.8563e-03, -3.3173e-03,
         1.1837e-03,  1.6350e-04,  2.8128e-03,  8.5163e-03,  6.6836e-03,
        -3.1427e-03,  2.8126e-03,  1.3458e-03, -8.0308e-04,  1.8922e-03,
        -8.0081e-03, -6.7096e-04,  1.0425e-03,  2.0548e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.0716e-02,  7.8435e-01,  1.6202e-02,  3.8201e-03,  4.3847e-03,
        -2.3867e-03, -3.5802e-03, -5.0700e-04,  2.7974e-03, -4.6527e-03,
        -3.8184e-04, -2.7120e-03, -8.5657e-03,  8.6772e-04, -9.7139e-04,
         1.2253e-04, -1.3187e-03, -8.1058e-04,  7.3921e-04,  2.4624e-03,
         1.3339e-03, -7.0545e-03, -1.3314e-03,  9.6804e-04,  3.5554e-03,
         1.0214e-02, -1.3739e-03,  2.3588e-03,  8.5332e-04, -9.1987e-04,
        -6.9357e-03,  1.7346e-03,  5.8410e-04,  3.3751e-03,  2.9939e-03,
         1.1618e-03, -4.1412e-03, -8.8887e-05,  1.3209e-03, -1.7378e-03,
         3.4603e-03, -1.0110e-03, -2.2757e-03,  4.4314e-03,  2.0332e-03,
         1.0406e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3236e-03,  8.3462e-01, -3.2045e-02, -8.9525e-03, -6.1475e-04,
         1.8455e-03, -1.9965e-02,  3.1528e-03, -1.8268e-03, -1.3220e-02,
        -1.1712e-03, -8.1090e-03,  7.1270e-03, -5.4543e-03, -9.6967e-06,
        -4.5536e-03, -3.4596e-03, -1.6734e-02, -1.0890e-02, -1.4019e-03,
         1.1328e-02, -2.4674e-03,  8.7334e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.5117e-03, -8.7339e-01,  2.3087e-02, -3.8372e-03, -1.0030e-02,
         3.5306e-04, -1.1407e-02, -1.4231e-02,  2.0553e-02, -2.3741e-03,
         3.1226e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0302,  0.8715,  0.0329,  0.0017, -0.0076, -0.0067,  0.0022, -0.0051,
         0.0077, -0.0028,  0.0060,  0.0016, -0.0128, -0.0112,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8957e-02,  8.8100e-01,  3.0011e-03, -1.2879e-02, -7.5321e-04,
        -1.9641e-03, -9.9273e-04,  8.2629e-03,  8.6843e-03,  3.2614e-02,
         1.4816e-02, -1.6079e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.7180e-03,  8.3967e-01,  1.3007e-02,  4.4314e-03,  3.2498e-03,
        -5.5943e-03,  9.3330e-06,  7.9657e-03,  2.8703e-03,  4.7366e-03,
        -1.2002e-03, -6.6405e-03,  3.3971e-03,  1.6260e-03, -3.1721e-03,
         1.4945e-03, -1.3627e-03, -4.6492e-04, -9.6144e-05, -7.0021e-03,
         1.4358e-03, -6.6967e-04, -1.1633e-03, -6.8297e-04,  4.6712e-05,
         1.8058e-03,  1.8416e-02, -2.2052e-03, -4.3153e-03, -1.9625e-03,
        -1.4753e-04,  5.5926e-03, -5.8315e-04, -3.4236e-03,  1.6408e-03,
        -3.0557e-03,  2.1974e-03,  2.6750e-05,  1.7445e-04, -1.1168e-03,
         1.6874e-03,  1.0336e-03,  5.9847e-04,  8.1669e-04, -8.5986e-04,
        -3.6744e-03,  1.7845e-03, -6.3203e-04,  4.0886e-04, -4.6119e-03,
        -3.6498e-03, -4.1288e-04, -2.4401e-03,  4.0460e-03,  4.9689e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-1.7980e-02,  8.3370e-01, -2.3196e-02, -1.7852e-02, -2.5192e-03,
         1.3497e-02, -3.6182e-03, -1.4552e-03, -2.2393e-02,  9.2790e-04,
         5.0825e-03,  2.4888e-03,  6.4736e-03, -8.3817e-03,  1.0252e-03,
         4.3913e-03, -9.6085e-03,  6.8732e-03,  2.0266e-03,  5.6254e-03,
         6.9049e-03,  6.1379e-04, -3.3682e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0322,  0.8580, -0.0204,  0.0088,  0.0082,  0.0050,  0.0052,  0.0051,
        -0.0067, -0.0117, -0.0018,  0.0045, -0.0068,  0.0058,  0.0051, -0.0033,
        -0.0050,  0.0065,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2262e-02,  7.7921e-01,  3.9114e-02,  1.7721e-02,  2.2312e-02,
         1.1243e-02,  1.5133e-02,  5.8855e-03,  3.3250e-03, -1.5254e-02,
        -2.8252e-03, -7.5750e-04,  1.5319e-02, -9.5754e-03,  6.9170e-03,
         4.2994e-03, -1.6426e-03,  2.7209e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0401e-02,  8.6213e-01, -1.5008e-02,  2.7178e-03,  9.3463e-03,
         1.8209e-02,  1.7669e-03,  3.0887e-03, -6.0022e-04,  8.1607e-05,
        -3.6481e-03,  2.2853e-03,  6.4054e-03,  2.3639e-02, -4.0671e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.8103e-04,  8.2539e-01, -1.3562e-02,  9.8174e-03, -3.5884e-03,
        -8.1244e-03, -5.0482e-04, -1.2141e-03, -7.2195e-03,  6.2753e-03,
         2.7526e-03, -6.2018e-04, -7.2303e-04, -1.4261e-03, -5.1949e-03,
         9.2156e-03, -3.1932e-03,  6.3968e-03,  2.3897e-03, -6.5966e-03,
         4.7901e-03,  6.3902e-03,  3.5912e-03, -2.6884e-04,  1.0154e-02,
         9.9497e-05,  1.5500e-02, -9.1981e-04, -2.6850e-03,  1.2575e-02,
         7.5090e-03,  2.0637e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9123e-02,  8.4999e-01,  1.9736e-03,  1.1821e-02,  1.7875e-02,
         3.0116e-03, -7.6510e-03,  7.3693e-04, -1.2923e-02, -1.4498e-04,
         2.6486e-03, -1.0813e-02, -1.0567e-03,  2.1787e-03, -1.5566e-03,
         5.5920e-03,  1.0051e-04, -2.0022e-03, -3.5091e-03, -5.4659e-04,
        -2.2133e-03,  7.6719e-04,  5.9650e-04, -2.7794e-03,  7.4191e-04,
         1.3798e-04, -3.5352e-03,  1.4734e-03, -4.5266e-03, -4.5448e-03,
        -1.5066e-03, -1.0131e-03,  1.0907e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4711e-02,  8.4567e-01, -3.1864e-03, -3.2167e-03, -3.7292e-03,
        -6.0383e-03,  5.7793e-04, -1.0977e-02,  2.1290e-04,  4.4461e-03,
         4.2708e-03,  5.6846e-03, -3.8228e-03, -7.9822e-04, -1.8327e-03,
        -3.1214e-03,  6.3784e-03, -8.4983e-04, -3.8830e-03,  1.2834e-03,
         2.2854e-03, -3.7035e-04,  2.6322e-03, -1.7991e-03,  1.8066e-03,
        -7.3774e-04, -6.2957e-04,  2.5024e-03,  1.9425e-03, -2.9344e-03,
         1.3564e-03,  3.4915e-03,  3.4070e-03,  2.8067e-03, -1.2677e-03,
        -2.7843e-03,  5.4956e-04,  2.0660e-04, -3.0472e-04,  1.8088e-03,
         6.1066e-04,  7.4994e-04,  1.3084e-03, -1.0576e-03,  1.3739e-03,
         1.4076e-03, -8.5903e-04, -1.3221e-03,  2.7992e-04, -1.8144e-03,
         6.5752e-04,  3.1711e-03, -3.2477e-05,  2.9967e-04,  4.0282e-04,
         3.8162e-03,  1.7729e-04, -1.0350e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4728e-02,  8.3527e-01,  2.3210e-02, -1.1416e-02, -2.3085e-03,
         1.4161e-03, -3.5244e-03,  8.3868e-04, -5.8650e-03,  3.8096e-03,
         2.3981e-04,  1.2182e-03, -2.3672e-03, -3.7429e-03, -9.4196e-04,
        -6.9600e-03,  3.0310e-03, -2.0046e-03, -2.9701e-03, -4.6033e-03,
        -2.0705e-03, -3.4307e-03, -6.2930e-03, -2.5210e-03,  3.1345e-04,
        -4.9017e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1278e-01,  7.7020e-01,  6.5888e-04, -1.0916e-02, -5.5704e-03,
         1.1989e-02, -4.5950e-03,  6.9046e-03,  5.7710e-03,  9.7285e-03,
        -1.6181e-04,  6.3364e-03,  4.0653e-03,  6.2480e-03,  8.7438e-04,
         1.7889e-03,  9.1263e-03,  3.3519e-03,  3.9103e-03,  1.0338e-03,
        -2.2077e-03, -7.1040e-03, -4.2068e-03, -6.0305e-03,  4.4438e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4792e-03, -8.1909e-01, -2.6469e-02,  6.2696e-03,  6.8763e-03,
         4.5905e-03, -6.2186e-03,  1.2209e-03,  1.1058e-03, -3.6924e-03,
         3.1152e-03, -1.5531e-03, -7.0865e-03, -1.3175e-02, -2.3605e-05,
         2.7935e-03,  5.1331e-03, -1.8271e-03, -1.1609e-03, -2.4423e-03,
         3.9824e-03,  1.3267e-02,  5.3553e-04,  1.8372e-03,  4.0136e-04,
         5.7204e-03, -3.2948e-03, -1.8613e-04, -4.1445e-04, -2.9317e-03,
        -3.2001e-03, -8.5161e-04, -2.4017e-03, -6.1753e-03, -6.7450e-03,
         6.6766e-03, -2.8441e-03, -1.3326e-03, -1.2271e-03, -2.2272e-03,
        -1.6055e-03,  1.9216e-03,  1.8076e-03, -5.5690e-03,  3.5165e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1870e-02,  8.1260e-01, -4.2978e-02,  1.4195e-02, -2.6056e-03,
        -1.5932e-02, -7.3830e-03,  8.8139e-04,  7.3578e-03,  7.8202e-03,
        -9.5315e-03, -3.9538e-03, -2.9447e-03, -2.0004e-03, -1.1139e-03,
         2.1126e-02,  1.0127e-03, -1.3784e-03, -1.2470e-03,  2.3222e-04,
        -1.3991e-03, -6.8131e-03, -3.6239e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0032,  0.8085, -0.0158, -0.0315, -0.0090,  0.0526,  0.0357, -0.0015,
         0.0019, -0.0018, -0.0334, -0.0051,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 2.2019e-02, -8.1834e-01, -1.2657e-02,  1.1793e-02,  1.9706e-02,
        -9.1164e-03, -6.2942e-04,  3.8688e-03, -2.6692e-03,  6.5860e-03,
         1.4501e-02, -3.0048e-05,  5.8768e-03,  5.4592e-03,  1.1381e-02,
         2.1411e-03, -3.4585e-03,  1.0984e-02,  3.5425e-03,  1.0151e-02,
         5.0602e-03,  3.7990e-03,  1.6226e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2801e-02,  8.6861e-01,  5.2682e-03, -8.7378e-03, -3.3132e-03,
        -5.7268e-04, -3.1768e-03,  8.2260e-03, -4.9402e-03,  8.5980e-03,
         1.4337e-03, -5.1147e-04,  2.3702e-03,  1.2314e-03,  2.7901e-03,
        -2.2507e-03, -4.5901e-03, -1.1844e-03,  5.6971e-04,  4.5671e-03,
         5.4121e-03, -3.5585e-03, -3.4914e-03, -5.5995e-03, -1.0671e-03,
        -3.2232e-04, -4.6366e-03,  2.0970e-03, -3.0877e-03,  2.8533e-03,
         1.8468e-03,  3.3919e-03,  6.8902e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0084e-02,  8.9305e-01,  1.0150e-02, -7.4936e-04, -9.5183e-03,
        -1.3159e-04, -2.6645e-03, -9.1136e-04,  2.4150e-03, -7.4930e-05,
         7.4034e-03, -9.6916e-03, -9.4245e-03,  5.7839e-03,  3.5667e-03,
        -1.7156e-02, -6.6606e-03, -1.6866e-03, -8.8824e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3532e-04,  8.8455e-01, -1.3914e-02,  3.5100e-02,  1.2134e-02,
         1.9660e-02,  4.3814e-03,  4.5988e-03,  6.6270e-03,  1.2282e-03,
         7.7452e-03, -3.1499e-03,  2.2664e-03,  4.4061e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0636e-02,  8.6235e-01,  3.8168e-02, -8.1584e-03, -1.4136e-03,
         1.6284e-03,  1.4918e-03, -1.6162e-02,  4.7729e-03,  1.2901e-03,
         7.1931e-04, -3.6324e-04,  1.9058e-03,  5.3906e-03,  1.5181e-03,
        -2.6738e-04,  3.5405e-03,  3.1299e-03,  3.9956e-03,  2.7450e-03,
        -6.0138e-03, -4.3372e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2857e-02,  8.4353e-01,  1.3564e-03, -5.1316e-03,  5.2883e-03,
        -8.6705e-03, -4.7084e-03, -5.6588e-03, -5.4667e-03,  2.3066e-03,
        -2.6143e-03, -5.3710e-04,  7.0865e-03, -4.0668e-03,  1.8362e-03,
        -6.0018e-04, -4.0889e-03,  1.2517e-03, -2.5869e-03, -4.0226e-03,
        -1.5664e-03, -3.9012e-03,  8.3055e-04, -4.8565e-03, -1.0102e-02,
         2.8662e-03, -3.1440e-03, -1.7025e-04, -1.1653e-02, -7.2471e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.5200e-03,  8.5351e-01, -1.2613e-02, -7.7387e-03, -3.2978e-03,
        -6.6613e-03, -6.1019e-03,  3.5788e-03, -2.7331e-03, -9.3447e-03,
        -2.3485e-02, -1.3696e-02, -1.4226e-02,  1.3516e-02,  7.4070e-03,
        -7.3935e-05,  1.3494e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4632e-02,  8.3286e-01,  1.8465e-02, -1.4587e-02, -2.4757e-02,
        -3.6997e-03, -3.6867e-03,  2.3592e-02,  1.5504e-02, -1.7892e-02,
        -3.2519e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3627e-02,  7.7757e-01,  1.8395e-02, -2.5569e-03, -7.7048e-03,
        -2.4687e-02, -2.5413e-02, -9.4937e-03,  5.8954e-05,  9.0477e-03,
         1.6994e-03, -8.1363e-04, -1.3484e-02,  5.9055e-02,  6.7779e-03,
        -3.3929e-03,  6.2178e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3832e-02,  8.5820e-01, -7.4586e-04,  3.6794e-04, -8.7262e-03,
         4.2983e-03, -9.1978e-03, -4.7516e-03,  7.0978e-03, -5.5695e-03,
         1.0418e-02,  1.2410e-02, -5.7912e-03, -1.5401e-06, -9.8379e-03,
        -2.4819e-03, -5.5151e-03, -5.5427e-03, -1.1070e-02, -3.9211e-03,
         2.9611e-04, -2.2634e-03, -1.4168e-04, -5.2299e-03, -3.3556e-03,
        -8.9352e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0817e-01,  7.6629e-01,  4.2318e-02, -9.1842e-03, -5.5954e-03,
         6.4049e-03, -6.2433e-04, -7.1911e-04, -7.1515e-03, -6.8956e-03,
        -1.5821e-03, -5.0393e-03, -6.9557e-04,  2.5840e-03, -9.7999e-04,
         2.4121e-03,  1.6043e-03, -2.0522e-03,  1.9072e-04,  3.5352e-03,
        -3.2220e-03,  2.5498e-03, -3.9617e-03, -2.6653e-03, -4.1401e-03,
         1.5418e-03,  1.3784e-03, -4.3910e-03, -5.2470e-04,  1.6050e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7035e-02,  8.6387e-01,  8.8741e-03,  3.2408e-03,  1.2429e-02,
         6.9029e-03, -1.7425e-02, -2.8413e-03, -2.7033e-03, -2.3123e-03,
         4.8258e-03, -3.8634e-03, -4.6992e-04, -6.7700e-03,  3.3985e-03,
        -3.8139e-03, -2.6311e-03, -5.0402e-03,  1.5493e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-3.1594e-02,  7.6135e-01,  6.1690e-02,  4.0678e-02,  1.0748e-02,
         1.3965e-03,  6.2192e-03, -3.5396e-03,  4.4930e-03,  3.8638e-03,
         2.5692e-03, -2.0007e-04,  3.2902e-03,  6.4707e-03, -1.4956e-02,
        -9.9819e-03,  3.3290e-03,  6.0686e-03, -7.0095e-04,  5.4863e-03,
         2.2605e-03, -1.0791e-03,  4.5141e-03,  1.8219e-03, -2.9972e-03,
        -2.4970e-04,  8.4527e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6983e-02,  8.4954e-01, -5.4666e-03, -2.0513e-02, -4.4315e-03,
         1.2726e-02,  5.4437e-04,  8.3911e-03,  4.9691e-03, -9.5278e-03,
        -4.5170e-04,  3.1911e-02,  5.7731e-04, -1.3966e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4359e-02,  8.4729e-01,  1.1165e-02,  7.6836e-04, -1.9498e-02,
         1.7973e-02,  6.6441e-03,  2.2683e-04, -1.3437e-02, -2.8041e-03,
        -3.2766e-03, -3.1774e-03, -2.2339e-03, -1.8432e-03,  4.0968e-03,
        -4.2405e-05, -1.2725e-04,  1.3581e-03,  2.0358e-03, -2.1061e-03,
        -4.6561e-03,  3.3811e-03,  3.1290e-03, -6.2780e-03,  5.9528e-03,
        -2.1441e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0101,  0.8546,  0.0069,  0.0026,  0.0058, -0.0319,  0.0095,  0.0129,
         0.0022, -0.0087,  0.0028,  0.0081, -0.0042,  0.0395,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1393e-02,  8.5777e-01,  2.4177e-03, -8.7679e-03, -8.5146e-03,
        -2.8866e-03, -1.5938e-03, -1.0585e-02, -5.6259e-03,  1.0477e-03,
        -5.0203e-03, -2.9072e-03,  4.6843e-04,  3.1892e-03, -1.3200e-03,
         1.0175e-03,  4.2188e-03, -1.1841e-03, -1.3097e-03, -1.6795e-03,
        -2.5898e-04, -3.0360e-03,  6.3217e-03,  1.8699e-03,  9.7446e-03,
        -5.5030e-03,  4.9220e-05, -8.1611e-05,  2.0036e-03,  9.5396e-04,
         7.0010e-03,  2.0258e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3626e-02,  8.4343e-01,  1.4196e-02, -7.3183e-03, -9.2931e-03,
         1.2658e-02, -2.2063e-02,  5.7554e-04,  4.5884e-03, -2.8687e-03,
        -7.6772e-03,  4.4850e-03,  3.5756e-03,  3.4049e-03,  2.6039e-03,
         4.0174e-03,  3.2703e-03,  1.0317e-02,  4.6937e-03, -2.3413e-03,
        -1.9663e-03,  1.9373e-03, -5.3448e-03,  2.6506e-03, -1.1098e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0513,  0.8886, -0.0049,  0.0079, -0.0269,  0.0203,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3408e-02,  8.0857e-01,  1.4290e-02, -7.5736e-03,  6.2809e-03,
        -1.3042e-03,  1.1798e-03, -7.4242e-04,  2.1525e-03, -3.7842e-03,
         9.6641e-04,  1.5271e-02,  2.5869e-04,  6.8873e-03, -3.9473e-03,
        -3.2874e-03, -7.2189e-04,  7.0332e-03,  3.5930e-03,  2.7399e-03,
        -3.2848e-03,  2.4840e-03, -4.9890e-03,  1.7220e-03,  1.4670e-03,
         1.4659e-03, -9.0791e-04, -6.0706e-04, -1.4175e-03,  1.4312e-02,
         3.9805e-03, -1.9997e-05, -8.2412e-04, -3.3735e-03,  2.5389e-03,
        -1.8150e-03,  7.9836e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0355e-02, -8.8993e-01,  2.0338e-02, -1.6139e-03, -1.6631e-06,
         3.2068e-04,  2.1561e-03, -4.0930e-04,  3.6169e-05, -6.8698e-04,
        -3.7605e-03, -2.3617e-03,  3.4443e-03, -1.4646e-03, -2.5530e-03,
         6.4246e-03,  2.5627e-03, -9.3122e-03,  1.3864e-04, -5.5152e-03,
         4.9596e-03,  6.8237e-04,  4.3656e-03, -3.1654e-03, -2.8804e-03,
        -1.7924e-03,  3.1336e-03,  3.5483e-03,  3.0746e-03, -3.1248e-03,
         8.8244e-04, -1.5086e-03,  2.0869e-04, -1.3891e-03,  1.8944e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0383,  0.7746, -0.0721,  0.0064, -0.0034,  0.0096,  0.0190, -0.0031,
         0.0018,  0.0161,  0.0089, -0.0034, -0.0095, -0.0267, -0.0069,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3643e-03, -9.0605e-01, -7.3893e-04, -4.2037e-03,  1.2472e-03,
        -4.0342e-04, -1.5694e-03, -9.1505e-03, -1.4557e-03,  4.9995e-03,
         1.2656e-03, -1.5704e-03,  3.2447e-03,  9.0883e-04, -1.4341e-03,
        -8.7813e-03, -2.6737e-04,  5.2103e-03, -4.3874e-03,  1.2211e-03,
         7.0633e-03,  5.9205e-03,  2.6438e-03, -1.8922e-03, -1.5002e-03,
        -1.9032e-03, -1.4273e-03, -4.7193e-03,  3.9490e-03,  1.7827e-03,
         2.7387e-03,  1.8054e-03, -2.1848e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1242e-02,  8.1799e-01,  7.0448e-03, -3.9249e-03,  3.1029e-03,
        -1.9556e-02, -1.2065e-02,  4.1210e-03,  1.3231e-03,  4.8785e-04,
         3.3450e-03,  1.0886e-03, -5.7977e-03, -1.3548e-02, -1.7108e-03,
        -1.7540e-03, -6.7283e-04, -3.9609e-03,  4.2650e-03, -5.4130e-03,
        -8.8643e-03, -6.1309e-03,  3.9313e-03,  7.7438e-05,  1.2518e-03,
        -1.7424e-04, -1.3987e-03,  1.6192e-02,  9.5705e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([-0.0097,  0.8537,  0.0056, -0.0146, -0.0014, -0.0120,  0.0036,  0.0245,
        -0.0060, -0.0207,  0.0023, -0.0036,  0.0039,  0.0057, -0.0327,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6079e-02,  8.5704e-01, -6.5735e-03, -1.4277e-02, -2.4018e-03,
         1.6038e-03, -2.9961e-03, -1.6458e-03,  3.6585e-03,  2.1889e-04,
         3.7988e-03, -3.0032e-03, -1.7072e-03, -9.4479e-04,  2.8792e-03,
        -3.9121e-03, -3.4504e-04,  1.7012e-03,  7.5254e-05, -1.8398e-04,
         2.0775e-03,  1.3223e-03, -1.0354e-03,  8.0994e-04, -7.2520e-04,
         1.4720e-03, -5.5408e-03,  3.8222e-04,  7.5415e-04, -2.4195e-03,
         1.0554e-03, -5.9026e-04, -2.6447e-03, -1.6881e-03,  6.5311e-04,
         3.5462e-03,  2.1110e-03,  4.8603e-04,  1.1229e-03, -2.9488e-03,
         1.9199e-06,  3.6383e-03,  1.0208e-03, -1.1800e-04,  1.2695e-03,
        -1.2489e-02, -1.0342e-03, -3.3695e-03,  8.6311e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9577e-03,  8.3620e-01, -2.6659e-02, -1.5178e-03,  9.0954e-03,
        -5.8612e-03,  6.2355e-03, -2.3779e-03,  8.8439e-03,  5.8077e-03,
        -3.1072e-03, -1.4299e-03, -1.8183e-03,  2.0596e-03, -2.6969e-03,
        -4.4573e-04,  2.3656e-03,  2.4801e-03,  1.9181e-03, -2.8607e-03,
        -2.6100e-04,  1.2315e-03, -2.3979e-03,  6.7866e-04, -2.0365e-03,
         1.7759e-03, -4.1125e-03,  4.5265e-03, -2.0501e-04, -3.4935e-03,
         3.4093e-03,  8.0435e-04, -2.9281e-03,  2.6948e-02, -1.9458e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.6964e-03,  8.9233e-01, -1.4836e-02,  1.3717e-02, -2.9911e-03,
         1.3326e-02, -2.9412e-03, -4.1036e-03,  2.4117e-03,  2.1064e-03,
         2.2968e-03,  8.1458e-04,  5.4667e-04,  1.2600e-03,  3.3457e-03,
        -2.0742e-03, -4.9992e-04,  1.4383e-02,  1.7314e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0238, -0.8276,  0.0153, -0.0260,  0.0069, -0.0039, -0.0050,  0.0026,
         0.0021,  0.0026,  0.0098, -0.0059, -0.0027,  0.0097,  0.0560,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8272e-02,  8.8826e-01, -1.5559e-02, -6.8373e-03, -4.3441e-03,
         2.9554e-03, -1.2150e-04, -2.0129e-03,  1.5340e-02, -1.2807e-03,
        -1.1509e-03, -1.4510e-03, -5.5450e-03,  2.7877e-04,  3.8081e-03,
        -2.3354e-03, -5.7369e-03, -1.6298e-03, -4.6177e-05,  3.4781e-03,
        -3.3072e-03, -1.6253e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7100e-02,  8.1650e-01, -2.8611e-02, -1.4459e-02, -7.2035e-03,
        -1.2944e-02, -2.2112e-03,  5.2378e-03, -6.1853e-03,  1.8165e-03,
        -4.7976e-03,  5.7676e-03, -2.8535e-03, -3.9581e-03, -1.8974e-03,
        -1.8467e-03, -6.5340e-03, -1.9036e-03, -3.8887e-03, -1.8483e-03,
         1.6661e-04,  1.0033e-02, -5.6647e-03, -1.4478e-03,  1.2451e-02,
         1.1661e-03, -2.8195e-03,  8.6911e-04, -9.1638e-03, -8.6546e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0095,  0.8129, -0.0342, -0.0094, -0.0553, -0.0128, -0.0064,  0.0215,
         0.0032, -0.0082,  0.0266,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5375e-02, -8.8948e-01,  8.0872e-03,  4.6202e-03, -8.0246e-03,
        -7.9097e-04, -4.1055e-03,  4.4914e-03,  9.9428e-04, -6.9723e-03,
         4.4617e-03, -1.3053e-03,  6.3349e-03, -1.4681e-03, -3.1994e-03,
        -4.2186e-03, -1.1026e-03,  3.0160e-03, -1.1651e-03,  1.4991e-03,
         2.5679e-02,  3.6055e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6418e-02,  8.4583e-01, -5.3479e-03, -2.1639e-02,  1.7777e-03,
         3.8654e-03, -3.2773e-03, -1.1105e-03, -5.0265e-03, -5.5383e-03,
        -2.0459e-02,  1.2524e-03, -3.7763e-03, -3.2008e-03, -3.1997e-03,
         6.6220e-04,  2.1112e-03, -4.8815e-03, -5.0892e-03, -8.4373e-04,
        -2.3827e-03, -1.5543e-04, -7.9237e-03, -2.9246e-03,  1.3210e-03,
        -2.6234e-03, -4.6080e-04,  5.6331e-05,  2.0712e-03,  1.3354e-04,
         9.9328e-03, -2.5398e-03,  1.2171e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.8154e-04,  8.7043e-01,  1.4344e-02,  1.5032e-02, -6.1364e-04,
         4.9706e-04,  1.6017e-05, -1.2128e-02,  4.3200e-03, -7.1796e-03,
         2.2606e-03, -8.2793e-04, -5.7403e-04, -5.7321e-03, -8.7108e-03,
         8.4738e-03,  1.5121e-04,  4.9092e-03, -1.5334e-03, -4.6380e-03,
         4.9173e-03,  3.4645e-03,  2.5908e-03, -2.6362e-03, -3.4578e-03,
         4.4821e-03,  5.8345e-03,  9.2684e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7446e-02,  7.7650e-01,  1.3186e-01,  1.5820e-03,  5.4045e-03,
        -1.9853e-03, -9.0532e-04,  4.1174e-03, -7.0318e-03,  6.2993e-03,
         5.6897e-03,  1.1094e-02, -2.1398e-04,  1.5985e-03,  3.4668e-03,
         2.4228e-03,  6.9851e-03, -7.4308e-04,  2.6161e-03,  5.2106e-03,
         2.6915e-03,  1.1864e-04,  2.2901e-03,  1.0452e-03,  6.7827e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 1.1149e-02,  8.7340e-01, -3.4514e-03, -5.4658e-03, -7.5369e-03,
        -1.4621e-03,  2.2321e-04,  1.6776e-03, -1.7503e-03, -2.5074e-03,
         6.2996e-04,  3.1412e-03, -1.5861e-03, -2.2295e-03, -4.7093e-04,
        -2.6022e-02, -3.3479e-03,  1.9244e-03, -1.2921e-03,  2.4234e-03,
         3.9928e-03, -4.5674e-03,  3.8222e-03,  2.9532e-03,  1.4435e-03,
        -5.6176e-04, -4.0088e-03,  5.8330e-04, -3.5320e-03,  2.0749e-03,
        -8.9319e-04, -2.9627e-03,  2.0391e-03, -1.7097e-04,  2.0549e-03,
         2.7420e-03, -3.8024e-03, -7.0559e-04,  7.9702e-05,  3.0292e-03,
        -2.2912e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7807e-02, -8.4604e-01, -6.5006e-03,  1.3800e-02, -1.1804e-02,
         2.1507e-03,  3.2574e-03, -1.0567e-02,  2.1433e-03, -8.2455e-06,
         2.3689e-03,  5.7715e-03, -2.9003e-03,  3.2044e-03, -1.1251e-02,
         8.1286e-03,  4.0850e-04, -3.2567e-03,  2.8204e-03,  5.8799e-03,
         2.1365e-03,  1.2473e-02,  5.8880e-04,  9.1657e-03,  1.5568e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9834e-02,  8.5777e-01,  2.2975e-02, -1.6684e-02,  2.4866e-03,
         3.9261e-03, -7.7337e-03,  1.0461e-02, -7.1568e-03, -4.7785e-04,
        -8.9452e-04,  3.8324e-03, -5.4089e-03,  2.1544e-03, -2.4568e-04,
        -6.9119e-03, -6.4495e-03,  3.6580e-03, -9.8826e-03,  2.9099e-03,
         8.1498e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4470e-03,  8.7176e-01, -7.0359e-03, -1.0495e-02, -4.4724e-03,
         1.1161e-03, -6.7181e-03, -1.7004e-02, -1.5947e-03, -1.2008e-03,
        -8.8313e-03, -1.0928e-02, -4.9550e-04, -7.6227e-03, -9.6409e-03,
        -9.4451e-03,  2.0051e-02,  6.1371e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4589e-02,  8.9564e-01,  1.4111e-03, -3.7001e-03, -5.5582e-03,
         5.5293e-03,  2.2896e-03,  4.7548e-04, -2.2546e-03,  4.6890e-04,
         6.1464e-05,  1.2436e-03, -7.0015e-03,  3.2916e-03, -3.3955e-03,
         3.9110e-03,  1.5018e-03,  1.7208e-03, -7.1463e-05, -1.8394e-03,
        -3.1677e-03, -9.8818e-04, -1.2940e-03, -2.1747e-03, -8.1971e-03,
         1.5526e-03, -3.0037e-03, -5.3618e-03, -3.1824e-03, -8.9250e-04,
         8.3774e-04, -1.5027e-04, -1.6250e-03, -5.6601e-04,  1.4468e-03,
         4.0238e-03,  3.0965e-03, -2.4862e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.4172e-03, -8.9504e-01,  1.7161e-02, -1.5148e-02, -2.7911e-03,
        -2.9638e-03,  1.2223e-05, -1.1753e-03, -3.2260e-03, -5.6286e-03,
         4.3198e-03, -2.1892e-03, -4.8937e-04, -5.3266e-03, -3.8056e-03,
        -1.1924e-03, -1.4907e-03,  3.1787e-03,  6.5830e-03,  1.8862e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3718e-02, -8.5174e-01,  3.6156e-02,  4.3960e-04, -7.3815e-03,
        -2.4767e-02, -1.5383e-03, -1.6494e-02, -7.3534e-03, -9.6468e-03,
        -1.3854e-02, -2.7938e-03,  1.0083e-03, -1.3112e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7147e-02,  8.6899e-01, -4.4638e-04,  6.8666e-04,  6.2833e-03,
         1.8219e-04, -9.6411e-04, -1.2953e-03,  2.0669e-04, -1.1378e-03,
        -2.1227e-03, -2.6599e-04, -3.0128e-03, -1.8012e-03,  3.0639e-03,
         3.3343e-03,  1.5276e-03,  6.4648e-03, -5.4707e-04, -3.4931e-05,
        -1.4675e-03,  1.7220e-02, -1.5343e-03,  5.2208e-03, -3.3103e-03,
         4.6350e-04,  9.7335e-04,  6.6694e-03, -5.0421e-03,  3.1926e-03,
         2.4084e-04, -5.1502e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1602e-02,  8.6587e-01,  1.1862e-02,  1.3213e-03,  8.3704e-03,
        -1.2679e-02,  8.2038e-03, -3.9985e-03, -6.6510e-03, -9.2197e-03,
        -5.8872e-04, -7.0458e-03, -2.5280e-03, -5.5685e-04,  1.1283e-03,
         1.2195e-03,  2.6356e-05, -5.2225e-03, -8.0716e-03, -2.3833e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8482e-02,  8.2516e-01,  7.5603e-03,  8.7290e-03, -2.1098e-03,
         8.0575e-04, -1.5504e-03, -1.8474e-03, -9.7581e-03, -3.5944e-03,
        -5.6139e-03, -4.2038e-03, -3.8086e-03, -3.1286e-03, -3.0546e-03,
        -5.3660e-04, -2.6988e-03, -7.3471e-03, -3.0530e-03, -3.6577e-03,
        -2.7028e-04,  1.7284e-03,  3.4035e-03,  2.1919e-03, -3.2855e-03,
        -5.1295e-03,  2.4505e-03, -8.9869e-03, -3.5848e-03,  6.3376e-03,
         4.1697e-03,  2.0675e-03,  2.9187e-03, -1.1228e-03,  4.8194e-04,
         7.9385e-03, -5.5075e-03, -1.7863e-03, -2.9341e-04, -2.0745e-03,
         1.1745e-03, -9.0029e-04,  6.7814e-04,  3.1615e-03, -1.5325e-04,
         6.5488e-03,  4.9513e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9062e-02,  8.7649e-01, -1.1962e-02,  2.4216e-03,  3.9653e-02,
         2.4732e-03, -2.6672e-03,  8.9407e-05,  7.9822e-04, -1.5304e-02,
        -6.9658e-03, -2.0818e-03,  2.0637e-03,  1.2465e-02, -5.5027e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.6589e-03,  8.4557e-01, -1.3120e-02, -8.1708e-05,  7.2250e-03,
        -6.7184e-03, -4.1190e-03, -7.9078e-03, -2.2742e-03,  3.5752e-04,
        -1.1097e-03,  1.5436e-03,  1.1966e-03,  1.0811e-02,  1.8426e-03,
        -9.1628e-04,  4.0411e-04, -1.1450e-03,  6.0876e-04, -3.4956e-03,
         1.9657e-03, -9.3610e-04, -9.0410e-04,  2.8913e-03,  2.9266e-03,
        -4.6961e-03,  5.0253e-03, -6.6401e-03,  1.2926e-04,  5.0662e-03,
        -8.7059e-04, -3.8600e-03,  2.0540e-03,  1.1658e-03,  3.3299e-03,
         4.6067e-03, -2.4436e-03,  1.8641e-03, -2.5320e-03, -1.7576e-03,
         2.5467e-03, -1.3173e-03, -3.9490e-03, -1.0874e-03, -2.2628e-03,
         3.6053e-04,  6.2132e-03,  8.4939e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-5.9782e-02,  7.9629e-01,  1.4930e-02, -1.3311e-02,  1.5501e-02,
         2.2450e-03,  1.1750e-02, -4.6055e-03, -6.5907e-03,  4.8627e-05,
        -5.5339e-03, -1.3508e-02, -5.5299e-03, -7.6964e-04,  1.7278e-03,
         1.1517e-03, -8.6077e-03,  5.5932e-03,  6.4189e-04, -2.8141e-03,
         2.6625e-04,  6.5405e-03, -5.4417e-03,  1.0290e-02,  6.0398e-03,
         4.9615e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.7291e-03,  8.6866e-01, -9.8157e-03, -4.9972e-04, -1.7424e-03,
         2.2245e-03,  3.6557e-03,  1.1624e-02,  6.1236e-04,  2.7989e-04,
         2.2245e-02, -8.8829e-03,  2.4088e-03,  1.3269e-03, -4.3177e-03,
        -3.0401e-03,  1.3816e-02, -1.2948e-02,  1.0626e-02, -2.8577e-03,
        -1.2691e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7328e-02,  8.1545e-01,  1.9160e-02,  7.4244e-03, -1.9098e-02,
         1.9978e-03,  2.5878e-04, -5.3334e-03,  3.1446e-03,  1.4547e-04,
         4.3624e-03,  1.7187e-03, -4.7441e-03, -2.3695e-04, -5.2384e-03,
        -2.8403e-03,  3.6643e-03, -6.5189e-03,  8.3059e-04,  3.3867e-04,
        -3.1914e-03, -6.2286e-04,  1.4103e-03,  5.0051e-04,  1.0274e-02,
        -9.9036e-04, -1.3294e-04,  3.4397e-03,  1.4613e-03,  1.1570e-03,
        -7.0575e-03,  3.5003e-03,  8.0843e-03,  1.6536e-03, -4.7458e-03,
         1.1947e-02,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0192,  0.8387,  0.0057,  0.0075,  0.0096,  0.0062,  0.0302, -0.0106,
         0.0044, -0.0052,  0.0026, -0.0079, -0.0080, -0.0100, -0.0025,  0.0032,
         0.0017, -0.0126, -0.0067,  0.0075,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3211e-02, -9.0640e-01,  2.6066e-03, -7.1133e-03, -8.2672e-03,
         1.1157e-03,  2.7435e-05, -1.7049e-03, -2.2169e-03,  3.6686e-03,
        -5.7662e-04,  4.2645e-03,  6.8974e-03,  3.6253e-03, -6.5161e-03,
        -1.3123e-03,  1.0528e-02,  1.0096e-02,  1.6383e-03,  2.9950e-03,
        -5.2184e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1532e-04,  9.3482e-01,  2.5178e-03,  4.5357e-03,  4.5836e-04,
        -8.7730e-04, -2.4911e-03,  1.9285e-04,  4.0987e-03,  1.0856e-02,
        -9.3616e-03,  2.7378e-02,  1.7975e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.0058e-03,  8.7169e-01,  9.4315e-03, -5.9727e-03, -3.0866e-03,
         4.4261e-04,  1.5373e-03, -2.4425e-03,  6.8211e-04, -3.3613e-03,
         4.3481e-03, -4.1762e-04,  1.3511e-03, -2.8317e-03,  1.0763e-02,
        -2.3758e-03, -5.4417e-03,  6.8114e-03, -2.4625e-03, -2.2950e-03,
        -2.7976e-03, -4.2001e-03, -1.4399e-03, -3.9844e-03,  1.6881e-03,
         4.9850e-04, -5.3840e-03, -1.2653e-03,  1.4653e-03,  2.1139e-03,
         1.5801e-03, -1.8903e-03,  2.3146e-03, -3.8685e-03,  5.3192e-03,
         9.6548e-03,  8.7886e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6681e-02,  8.8040e-01, -1.5197e-03, -1.3048e-02,  8.7005e-03,
        -3.3326e-03, -2.8039e-03, -1.1532e-02,  1.0685e-03,  5.9230e-03,
        -2.4854e-04,  1.1210e-03,  7.6172e-03, -3.0525e-04, -3.5079e-03,
         3.3916e-03,  4.2668e-03,  2.1963e-03, -3.3474e-03, -7.6005e-03,
         3.5594e-03,  4.9285e-03, -7.0405e-03, -5.8550e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.0712e-02, -7.7053e-01,  9.8863e-03,  8.4973e-03,  1.4120e-02,
        -6.3651e-03, -1.0914e-02, -2.8556e-03, -9.7128e-03,  2.5718e-03,
         6.0370e-03, -3.1876e-03,  1.6784e-02,  2.0357e-03, -1.1866e-02,
         9.2565e-04,  5.9984e-03, -5.7157e-04, -1.5083e-03, -2.7404e-03,
         1.3135e-03,  2.8201e-04,  1.2720e-04, -8.0004e-04, -1.9346e-03,
         9.4987e-04, -2.8980e-04,  2.5333e-03,  1.4532e-03,  3.6861e-05,
         2.8258e-03, -9.6353e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.2534e-03,  8.5106e-01,  1.1344e-02, -1.9944e-03, -7.9487e-03,
         2.4296e-03,  7.2210e-04, -6.7344e-03,  9.0886e-03, -1.6565e-03,
        -3.1844e-03, -5.8236e-04, -2.9868e-03, -1.1075e-03, -1.5301e-03,
         1.3118e-02, -2.7837e-03, -9.6240e-04,  6.1413e-03,  5.7238e-03,
         1.0637e-02, -9.6888e-04,  3.1553e-03, -6.0302e-04, -4.3132e-03,
        -6.3843e-03,  3.8374e-03, -2.6827e-03,  1.5907e-02, -1.0545e-02,
         2.6130e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2210e-02,  8.3294e-01,  4.4198e-02,  2.6207e-03,  8.0388e-03,
        -3.8295e-03, -1.6488e-03, -1.0278e-02,  7.7039e-03,  3.6858e-05,
        -2.0513e-03, -4.0831e-05,  4.9578e-03, -7.9152e-03,  1.0545e-03,
        -5.4728e-04,  4.7239e-03,  1.2928e-03,  1.6208e-03,  1.4613e-03,
         8.0444e-03,  1.5262e-02,  7.5180e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3645e-03,  8.6533e-01,  1.7026e-02, -7.1746e-03,  7.8318e-03,
         4.3622e-03, -3.3809e-03, -2.3532e-03, -7.1538e-03, -3.8876e-03,
        -1.7088e-04,  5.7062e-04, -3.2189e-03, -2.0323e-02, -3.8764e-03,
         3.8888e-03, -2.4833e-03, -1.3559e-05,  4.4915e-04,  3.2289e-03,
        -6.7157e-03, -2.6396e-03,  4.3994e-03, -2.3232e-03,  7.3237e-04,
         2.0695e-03,  1.5707e-02, -4.3304e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 1.9861e-02, -8.8511e-01,  4.4515e-03, -1.3965e-02, -2.5184e-03,
        -2.0562e-03,  2.0701e-04,  1.3448e-03,  9.5761e-03,  6.2153e-03,
        -4.1594e-03,  5.3148e-03,  1.9182e-03, -8.3664e-05,  2.2864e-03,
        -3.5878e-03,  9.7942e-04, -3.8157e-04,  7.9582e-03,  6.4376e-03,
        -5.0769e-04, -1.1996e-03,  4.7943e-03,  1.0202e-03, -9.7661e-04,
        -2.7148e-03, -2.7729e-03,  2.7710e-03,  4.8260e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0655e-02, -8.7109e-01,  2.0863e-02, -5.6292e-03, -3.6441e-03,
         8.5642e-03, -1.6332e-03,  7.2454e-03,  4.5721e-03, -6.4186e-04,
        -4.5069e-03,  9.2919e-03,  5.2215e-03,  1.4437e-04,  1.1937e-03,
         6.0971e-04, -2.5007e-03, -1.5706e-03,  1.4035e-03,  3.2760e-03,
         1.9622e-03,  9.8004e-04,  1.0526e-02, -2.2769e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6421e-02, -8.2607e-01, -2.7419e-02, -4.8265e-03, -7.2457e-03,
        -8.7322e-03,  7.6899e-04, -2.8390e-03, -7.3840e-03,  3.6456e-04,
        -4.1204e-03, -1.8615e-02, -4.5484e-03,  9.8469e-05, -5.7197e-04,
         3.5105e-03,  8.4506e-04, -1.2493e-04,  5.3130e-03, -3.0563e-03,
        -8.0528e-03,  4.7847e-03,  1.0926e-03, -5.1301e-03, -1.6770e-03,
         3.3378e-03,  1.0099e-02, -9.2552e-04,  4.1718e-03, -2.7727e-03,
         1.3725e-03,  9.6889e-04,  1.2011e-04,  2.6190e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0208e-02, -8.7487e-01,  4.4718e-03,  2.4620e-03, -3.0685e-03,
        -4.4955e-04, -1.4274e-02, -2.4755e-03, -5.1243e-04,  9.6630e-04,
        -2.9720e-03,  8.7643e-03, -1.0108e-03,  4.0282e-03,  3.0283e-03,
        -7.8530e-05, -4.3163e-04, -8.4296e-04,  4.7738e-03,  1.9172e-03,
        -1.6111e-03, -6.7864e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.2644e-03, -8.2908e-01,  1.4453e-02, -3.1104e-02, -3.7135e-02,
        -4.1831e-03, -4.4288e-04,  6.6228e-03,  4.5426e-03, -3.2509e-03,
        -6.9382e-03,  8.6707e-03, -4.0449e-03, -7.7265e-03,  4.2432e-03,
         1.1988e-02, -2.6484e-03, -1.4661e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3359e-02, -8.4789e-01,  1.7972e-02, -2.3311e-02,  4.8652e-03,
         8.2299e-03, -1.0220e-03,  4.4812e-03,  6.7479e-03, -7.9448e-03,
         1.5576e-04, -3.0641e-03,  4.6552e-03, -8.7512e-04, -3.2181e-03,
        -6.1337e-04, -2.3525e-03,  7.4502e-03,  2.5633e-03,  1.8423e-03,
         4.7700e-03, -5.5453e-03,  7.0709e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7964e-03,  9.1667e-01,  1.4089e-02,  5.4540e-03, -1.1960e-02,
         1.4041e-03,  6.8742e-04, -3.5265e-03,  3.4880e-05,  4.3712e-03,
        -1.5209e-02, -1.2074e-03, -2.5095e-03,  9.3925e-03, -8.8657e-05,
        -6.8111e-04, -2.9155e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3623e-02, -8.3917e-01, -4.3683e-02, -3.0999e-03, -9.4328e-03,
         2.9539e-03, -2.5531e-04, -3.0066e-03, -3.0481e-03, -1.1486e-02,
        -7.2269e-03, -8.1951e-04, -5.2157e-04, -1.2780e-03, -2.5734e-03,
        -1.2108e-03, -2.0336e-02,  1.6270e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5381e-02, -8.8401e-01,  1.5057e-02,  6.0145e-03, -2.3234e-03,
        -3.0865e-03, -2.1317e-03, -2.5007e-03,  9.0314e-04,  6.1964e-03,
         4.6967e-04, -1.8191e-03,  6.4184e-04,  3.6580e-03,  2.7829e-03,
        -8.7123e-03,  8.6936e-03,  5.8955e-03, -3.3327e-03,  1.6523e-05,
         2.9447e-03, -4.0585e-04,  2.9317e-03,  5.6512e-03, -3.8743e-03,
         1.0563e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0235e-02,  8.2199e-01,  1.6382e-02, -7.2754e-03,  1.3881e-02,
         2.3598e-02, -9.3182e-03,  1.1849e-02,  6.2507e-03, -1.3113e-02,
        -2.0676e-03, -1.5768e-03, -4.3687e-03,  2.7668e-04, -7.1620e-03,
         8.6938e-04, -3.9753e-03,  3.8984e-04,  4.6536e-03,  8.8416e-04,
        -1.3329e-03,  1.2599e-03, -1.9300e-03,  5.0488e-03,  4.6302e-03,
        -5.1188e-03,  4.1356e-03, -2.1566e-03, -3.8304e-04,  7.4559e-03,
        -6.4317e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.0243e-02, -8.2617e-01, -2.5436e-02, -5.6830e-04, -2.7164e-03,
         1.0410e-02, -7.2723e-03, -1.4785e-02, -7.7723e-03,  2.2116e-02,
        -6.7749e-03,  2.1282e-03, -7.6930e-03,  8.8088e-04,  3.6470e-03,
         6.6784e-03, -3.6983e-03,  1.0094e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0350,  0.8235, -0.0406, -0.0098,  0.0011, -0.0074, -0.0161,  0.0664,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([-2.8330e-02, -8.6409e-01, -8.9553e-03,  7.4894e-03, -7.2859e-04,
         2.0592e-03,  4.6815e-03,  4.1969e-03, -5.6999e-03,  2.7008e-03,
        -2.5631e-03,  7.1995e-03,  8.2296e-04, -4.4477e-04,  4.4276e-04,
        -3.8985e-03, -2.3487e-03, -1.7833e-03, -9.3972e-04, -2.0884e-03,
        -3.3771e-03, -1.1865e-03,  6.3157e-04, -7.3702e-04,  3.6577e-03,
         1.4888e-03,  3.0044e-03, -2.7029e-03, -2.0962e-03, -1.8002e-03,
         6.1471e-04, -8.8280e-04, -2.0653e-04,  1.3024e-03,  6.1763e-04,
        -1.3874e-03, -2.7203e-03, -2.1731e-03,  1.7950e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1807e-02, -8.4692e-01,  1.0873e-02,  7.8557e-03, -5.2301e-03,
         4.0613e-03,  9.2186e-03,  1.2880e-03,  1.9086e-03,  4.3311e-03,
         1.2702e-03, -1.4704e-03, -5.1374e-03,  1.2243e-03, -3.1870e-03,
         6.1502e-03,  7.1571e-03,  1.5270e-02,  2.7658e-03, -6.9662e-03,
         1.6212e-02,  6.4178e-03,  1.1379e-04,  1.3913e-03, -1.2597e-03,
         5.9870e-03, -7.6346e-03, -6.8937e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8856e-02, -8.8050e-01,  7.3195e-03,  4.2664e-03,  2.2136e-04,
         2.7886e-03, -5.3547e-04,  1.1220e-02, -5.2913e-03,  1.0677e-02,
        -1.3574e-02,  1.5704e-02,  9.0489e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5656e-02, -8.7351e-01,  2.1137e-02,  3.3643e-02,  7.7143e-03,
         2.0507e-03,  1.0513e-02, -4.6020e-04,  4.5021e-03,  4.0166e-03,
         5.1536e-03,  5.1344e-03, -1.0575e-03,  6.0839e-03,  9.3668e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1470e-02, -8.5519e-01, -1.8109e-04, -5.6240e-03, -7.1196e-03,
        -1.1401e-02, -2.0085e-03,  5.7020e-04, -6.4783e-03, -1.0811e-03,
         4.6788e-04, -4.1521e-04,  4.0030e-03, -6.0109e-03, -3.1413e-03,
         2.6095e-03,  4.1532e-04, -9.8023e-04, -3.0231e-03,  5.5428e-03,
        -1.8751e-03, -3.5810e-04, -8.7947e-04, -1.2827e-03, -1.1354e-02,
        -7.3402e-03,  1.6496e-03, -4.2381e-03,  2.9774e-03,  9.2099e-05,
         3.4870e-03, -4.6788e-04, -8.9727e-04, -3.7404e-03, -9.7051e-03,
         1.2470e-03,  1.7110e-03,  2.2500e-03,  6.6477e-04,  4.3589e-04,
        -4.8034e-03,  2.1144e-03, -1.3582e-03,  1.2221e-03, -3.6808e-03,
         1.8054e-03, -6.2669e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.4592e-03, -8.9747e-01,  1.0494e-02,  8.1636e-03, -2.2052e-03,
         3.1989e-03,  1.0298e-03, -1.6483e-03, -3.5753e-03, -2.1903e-03,
        -3.6084e-03,  1.0167e-03,  7.9981e-04,  1.9785e-03,  6.7625e-04,
        -3.6173e-03,  9.7858e-04, -3.6131e-04,  3.3690e-03, -1.0361e-03,
        -1.2478e-03, -5.5034e-04, -8.0613e-04, -3.3447e-03, -2.7252e-03,
        -1.6340e-03, -4.8850e-03, -2.6135e-03,  1.6038e-03, -3.2539e-03,
        -2.0659e-03,  2.5309e-04, -1.7348e-03, -3.0015e-03, -3.4690e-03,
         1.9457e-03,  3.5584e-04, -1.0759e-03,  3.7516e-04, -1.8845e-03,
         4.2917e-04,  4.8691e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1226e-02,  8.4626e-01, -1.4330e-03, -1.4249e-02,  1.3597e-02,
         5.9190e-03,  2.3618e-02, -4.7958e-03, -7.3084e-04, -6.6673e-03,
        -8.4117e-04, -4.2072e-03, -5.0353e-03,  1.9143e-03,  9.7254e-03,
         4.3903e-03,  2.4882e-02, -1.3643e-03,  1.2125e-03, -1.6210e-03,
        -5.7292e-03,  2.0688e-04, -6.8947e-03, -2.1856e-03,  1.2983e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4486e-02,  8.4658e-01,  1.6432e-02,  6.4242e-03,  4.3855e-04,
         3.4489e-04,  1.0287e-02, -2.9779e-03,  2.4281e-03, -2.1381e-03,
         1.2150e-04,  6.5000e-03,  5.4197e-03, -4.2160e-03, -3.7942e-03,
         4.4162e-04, -3.2449e-03,  1.7518e-03,  3.1872e-03,  2.8453e-03,
         1.4993e-02,  5.5431e-03, -5.4017e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1650e-02, -8.6575e-01, -2.2982e-02, -1.3943e-03,  1.0603e-02,
        -7.3983e-04, -2.5914e-03, -4.5247e-03,  9.2639e-03, -1.7593e-02,
         5.1180e-04,  2.4666e-04, -7.1575e-03,  5.9328e-03,  7.0417e-03,
         5.7944e-04, -5.7548e-04, -3.4188e-03, -2.6941e-04,  1.8461e-03,
         8.8209e-04,  2.7003e-03,  1.7415e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0993,  0.7881,  0.0140, -0.0114, -0.0069, -0.0131,  0.0151,  0.0197,
        -0.0022,  0.0029,  0.0151,  0.0042, -0.0079,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3606e-02, -8.8566e-01, -2.6047e-02,  1.5177e-02, -1.6833e-03,
         3.2231e-03,  7.2612e-03, -1.9022e-03,  3.9171e-03,  6.5108e-03,
        -2.5344e-03, -6.2532e-04,  1.3377e-02,  1.8474e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.8038e-02,  8.6852e-01, -1.3269e-03, -1.4299e-02,  1.6148e-03,
        -5.5909e-04, -4.7061e-03, -3.9297e-04, -3.5810e-03, -6.4371e-03,
         4.8189e-03,  4.8690e-03, -1.9712e-03,  4.2993e-03, -2.6381e-04,
        -6.4087e-04, -1.0161e-02, -1.2186e-03,  2.2779e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 3.1676e-02, -8.5778e-01, -4.3210e-03,  9.0969e-03, -8.1795e-03,
         4.4462e-03, -7.4520e-04,  1.0681e-02, -2.4151e-04, -2.1467e-03,
        -2.6820e-03, -6.3953e-03,  6.5145e-04, -9.7644e-03, -1.5239e-03,
         2.2506e-02, -3.7928e-03, -1.6157e-03, -3.9003e-03, -3.8109e-03,
        -3.2197e-05,  8.2922e-04,  4.0263e-03,  1.2503e-05, -1.6935e-03,
         7.4505e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.6274e-02,  8.0153e-01,  1.1903e-02,  1.0052e-02, -6.5199e-03,
        -1.1453e-02, -1.5064e-02,  2.1312e-03, -1.5682e-02, -2.4871e-02,
         7.4874e-04, -2.4713e-04,  3.6416e-03,  9.1442e-03, -6.8436e-03,
         1.3892e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2486e-02, -8.8597e-01, -1.5380e-02, -3.4692e-02,  2.8395e-03,
        -8.2718e-03, -1.3591e-03,  1.3000e-04, -4.1760e-03, -8.9391e-03,
         6.2439e-04, -1.2773e-04, -4.7611e-03,  1.9102e-04,  1.6834e-03,
        -8.2692e-03,  1.0055e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0181,  0.9216, -0.0050,  0.0048,  0.0015,  0.0018, -0.0128,  0.0208,
         0.0136,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2376e-02, -8.0090e-01, -2.6897e-02,  9.2109e-03, -1.8242e-03,
        -2.0135e-02, -1.0451e-03,  2.7815e-03, -4.4671e-03,  5.1935e-03,
        -8.2835e-03,  1.8227e-03,  1.7913e-04,  1.1118e-03,  1.8709e-03,
        -2.4348e-03,  1.2506e-03,  1.0767e-03, -2.5331e-03, -4.3420e-03,
        -1.1271e-04,  4.0734e-04,  1.7864e-03,  5.0469e-03,  3.2011e-03,
         3.4661e-03,  1.1090e-02,  2.2567e-03, -1.5474e-03,  4.2716e-04,
        -4.4018e-03, -1.6387e-03, -6.9255e-04,  2.8418e-03, -2.5694e-03,
         8.4792e-03,  3.7355e-03,  5.9215e-04,  6.9085e-03,  2.3381e-03,
         1.2121e-03, -1.1641e-03,  1.3002e-02,  6.1406e-03,  3.5655e-03,
        -1.6351e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1095e-02,  8.7619e-01,  2.5386e-02, -8.4095e-04,  1.8702e-04,
        -4.7332e-04, -1.2694e-03,  1.1665e-03, -6.5357e-03,  6.4762e-03,
         1.2987e-03,  2.4935e-04, -1.8548e-03, -1.1727e-03,  6.4064e-03,
         3.3760e-03,  1.1341e-03,  2.2950e-03, -3.0096e-03,  2.8916e-03,
        -1.6054e-03,  1.3242e-04, -1.8150e-04,  1.4661e-03, -4.3971e-03,
        -6.2685e-04,  7.8960e-04, -4.6574e-04, -9.5254e-03, -7.4972e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0347, -0.8701,  0.0216,  0.0213,  0.0095,  0.0365,  0.0063,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0543,  0.8785,  0.0020, -0.0044,  0.0084,  0.0054,  0.0127,  0.0051,
         0.0047, -0.0017, -0.0229,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7647e-02,  8.7502e-01,  1.1733e-02,  3.9988e-02,  1.1544e-02,
        -3.6853e-03,  9.3829e-03, -6.2209e-03,  6.5916e-03,  8.6750e-04,
        -6.0777e-03,  1.7094e-03, -6.5086e-03, -3.0246e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.0165e-03,  8.7331e-01, -1.5279e-02, -5.9041e-03,  5.8201e-04,
        -5.8234e-03, -2.7686e-03, -1.4101e-03,  4.1359e-03, -6.8951e-04,
         1.5404e-03, -5.5797e-04, -2.2792e-03, -6.2868e-03,  2.3263e-03,
         6.9998e-04,  6.5415e-04, -5.7361e-04, -1.0988e-03,  8.3734e-03,
         1.3418e-02,  8.0178e-03,  1.8713e-03, -7.9659e-05, -1.7885e-03,
        -4.5780e-03,  6.3096e-05,  1.2761e-03, -2.6197e-03,  1.8811e-03,
        -2.3085e-03,  2.1739e-05, -5.2502e-03,  1.5518e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.7129e-03,  8.3394e-01,  8.7190e-03,  3.4393e-03, -6.2952e-03,
        -2.8240e-03,  1.0669e-02, -2.3078e-03,  4.6177e-03, -6.0383e-04,
         6.5437e-03,  1.3325e-03, -3.2982e-03, -4.8262e-03,  1.6792e-02,
        -6.3785e-03, -4.5209e-03, -3.0686e-02, -1.5549e-03, -2.2322e-03,
         2.0712e-03, -2.0202e-03, -2.7497e-04, -1.0932e-03, -1.6617e-03,
        -5.9099e-03, -5.2632e-03, -2.2417e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5926e-03,  9.1349e-01, -1.3359e-03, -1.2569e-04,  3.2166e-04,
        -2.4919e-03, -7.0071e-03,  3.5823e-03, -2.8329e-04,  2.9027e-03,
        -2.4707e-03, -9.7912e-03,  2.7407e-03,  4.8480e-03, -3.3100e-03,
         3.2050e-03,  3.5172e-03,  4.1337e-03,  3.9201e-03,  4.5395e-04,
         3.0003e-03,  2.1532e-03, -3.7720e-03, -1.7555e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-4.4929e-03,  8.6088e-01,  9.9974e-03,  7.2766e-06, -3.7661e-04,
        -8.8417e-03,  5.2637e-03, -2.1867e-03,  5.3242e-03, -5.9577e-04,
        -1.9150e-02, -6.6951e-03, -1.3746e-03,  2.9754e-03,  5.7457e-04,
         5.5851e-04, -1.1166e-03, -3.0916e-04,  8.1863e-03,  1.6138e-02,
         2.5965e-03,  6.5220e-04, -4.2448e-03,  2.1159e-03,  3.0037e-03,
         7.1019e-03,  3.2592e-03,  3.3239e-03,  3.9809e-03, -1.1883e-02,
        -2.7935e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0183,  0.8914, -0.0160,  0.0409, -0.0031, -0.0027,  0.0031,  0.0023,
         0.0070,  0.0047, -0.0059, -0.0045,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1632e-02,  8.9749e-01,  6.4045e-03,  3.7226e-02, -1.6470e-03,
        -2.6883e-04, -5.5005e-03, -2.4182e-03, -4.4014e-03,  1.2786e-03,
        -7.4860e-04, -2.0970e-02,  2.7111e-03, -3.3825e-03, -3.9198e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.4954e-03, -8.8632e-01, -1.7065e-02,  1.3673e-02,  8.4108e-04,
        -4.9182e-04,  2.3471e-03,  4.7328e-04, -1.9115e-03,  2.7725e-03,
         8.2705e-04, -2.3376e-03,  5.2028e-04,  4.5168e-04,  4.4356e-03,
         3.6373e-03,  2.5148e-03,  2.1566e-03, -7.3508e-03,  3.6711e-03,
         3.1855e-03,  1.6081e-03, -2.9801e-03, -1.1767e-03, -5.1819e-04,
        -1.4554e-03, -3.3051e-03,  2.7255e-03,  3.1793e-03,  1.3900e-03,
         5.7661e-03,  8.3446e-03, -3.0683e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5475e-02, -8.9928e-01,  1.8962e-02,  9.0242e-03, -7.5399e-03,
         8.6543e-06,  6.7281e-03,  6.6033e-03,  1.8896e-03,  1.9189e-03,
        -4.8270e-03, -5.0054e-03, -2.1513e-03,  2.0725e-03,  4.6082e-03,
        -1.3905e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0126,  0.9094, -0.0196, -0.0055,  0.0074,  0.0216,  0.0104,  0.0027,
         0.0107,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0412, -0.8663,  0.0129, -0.0397,  0.0399,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.6384e-03,  8.8236e-01,  7.6400e-03,  9.8455e-04,  4.9627e-03,
        -2.1862e-03, -6.7209e-04,  6.4601e-03,  1.3040e-02,  4.9794e-03,
         1.2657e-02, -4.3991e-03,  2.8329e-03, -6.4411e-03, -8.3149e-03,
        -2.0041e-04,  4.1853e-03, -1.5405e-03,  1.7290e-03,  4.5034e-03,
        -2.2691e-03, -7.7572e-04,  1.9133e-03, -2.9705e-03,  1.3659e-03,
        -1.1185e-02,  7.9658e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1085, -0.8272, -0.0193, -0.0148, -0.0149, -0.0060,  0.0093,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0541e-03,  8.6271e-01, -2.6874e-04, -1.2167e-02, -3.9988e-03,
         3.5346e-04,  7.3392e-03,  6.3261e-03, -8.9684e-03,  4.6090e-03,
        -2.1579e-03,  3.1332e-03, -1.3293e-03,  3.4195e-03,  7.5476e-03,
         3.2029e-03,  8.7646e-04, -6.8622e-03,  6.3371e-03, -3.7823e-02,
        -8.3363e-03,  1.1181e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4073e-02, -8.8854e-01,  6.2318e-03, -6.8374e-03,  1.2753e-02,
         1.6972e-04,  3.0123e-03, -5.1567e-03, -5.3020e-04,  6.7754e-04,
         9.6910e-03, -9.1083e-03,  1.4454e-02, -2.5657e-03,  4.3308e-03,
        -3.3642e-03, -8.4739e-04, -7.5934e-04, -6.8983e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0104e-02, -8.3715e-01, -9.4970e-03,  5.1675e-03,  1.0805e-02,
         1.0858e-02,  1.0946e-02, -2.4274e-03, -1.0265e-02,  8.1836e-04,
        -5.6219e-05, -3.1100e-03, -2.3493e-03, -4.2238e-03,  2.9054e-03,
        -3.3889e-03,  5.2363e-03, -2.1876e-03,  1.0082e-03, -2.4114e-03,
        -3.6280e-03, -4.7814e-03, -2.9894e-03,  4.1794e-04,  8.8786e-05,
        -3.1963e-03,  4.8260e-05,  1.0733e-03, -3.5503e-03,  5.0789e-03,
        -1.0235e-02,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 3.1906e-02, -8.0518e-01, -5.7045e-03,  1.2946e-02, -1.0822e-02,
         9.3299e-03,  1.2572e-02, -1.5794e-03,  1.8715e-04, -2.7381e-04,
        -2.1009e-04, -5.5863e-03,  7.4670e-03, -1.5395e-02, -2.0514e-03,
        -7.0275e-03,  9.4348e-03,  2.5453e-03,  1.1898e-02, -3.0607e-03,
        -9.0261e-04,  1.4350e-03, -1.6723e-03,  4.5033e-03,  5.8775e-04,
         1.3868e-03,  3.2192e-03,  9.7310e-03, -2.6506e-03, -4.5071e-03,
        -5.5581e-03,  8.6661e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5625e-02, -8.9020e-01, -1.1319e-02, -5.8987e-03,  9.5487e-04,
         1.4052e-03,  1.6978e-03, -9.5224e-03,  5.3757e-04, -2.6827e-03,
         5.8575e-03, -7.8049e-03,  5.4960e-03, -9.4454e-03, -8.0700e-05,
        -2.8555e-03, -2.2489e-03, -8.7696e-04,  2.9905e-03, -3.7506e-04,
        -9.2300e-03, -2.8995e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4191e-02,  8.3814e-01,  1.2041e-02,  6.2245e-03, -6.3627e-04,
         2.1417e-03, -6.0687e-03, -6.0684e-03, -3.0287e-04, -1.1914e-03,
         9.7770e-03,  8.7028e-04,  6.9038e-03,  3.7096e-03, -5.0776e-03,
        -3.5334e-04, -5.1676e-03,  1.3799e-02, -7.5115e-03, -1.6600e-02,
         1.5204e-04, -4.4710e-03,  7.8156e-04, -6.3036e-03, -2.0851e-03,
        -5.7924e-03,  2.4859e-03, -4.0956e-04,  8.7557e-03,  1.9886e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0122,  0.8220,  0.0291, -0.0035, -0.0018, -0.0306,  0.0110, -0.0057,
        -0.0056,  0.0040,  0.0097,  0.0165, -0.0049,  0.0017,  0.0015,  0.0043,
         0.0344,  0.0016,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.0966e-02,  7.9162e-01, -9.1738e-03, -6.8179e-03, -1.0483e-02,
        -2.4117e-02, -4.7912e-03,  1.3019e-03,  6.5481e-04,  1.2538e-03,
         3.2513e-03, -1.9333e-03, -9.8395e-05,  1.1067e-03,  1.3068e-03,
        -2.9183e-03,  6.9450e-03,  1.2060e-02, -1.3883e-02,  1.6349e-03,
        -2.0915e-03, -5.5959e-03,  7.2790e-03,  1.3884e-03,  4.5866e-03,
         3.0256e-04,  3.3067e-03,  2.2809e-03,  4.2574e-03,  3.9301e-03,
         2.7908e-03,  3.4570e-03,  4.0766e-03,  2.7031e-03,  2.6290e-04,
        -4.6918e-03,  6.7944e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1217e-02,  8.1187e-01, -5.4101e-02,  1.2660e-02,  2.9184e-03,
         2.7476e-03,  2.7525e-03, -2.9985e-03,  4.1613e-03,  2.6934e-03,
        -3.0368e-03, -6.2906e-03,  1.2600e-03,  4.9929e-04,  4.6504e-03,
        -2.7173e-03, -5.4298e-03, -2.2118e-04, -3.4267e-03,  9.8265e-04,
        -6.7389e-03, -1.6882e-04,  5.1068e-03, -1.3501e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4258e-02,  8.4949e-01, -1.3712e-02,  9.8492e-03,  1.8195e-02,
         9.6750e-03, -1.8330e-03,  1.4080e-04,  6.3415e-03, -4.4735e-03,
         2.1356e-03,  1.0547e-04, -3.3037e-04, -1.4290e-03,  5.3604e-03,
         2.7469e-03, -1.1488e-03,  1.7291e-04,  2.0487e-03,  5.0269e-04,
         9.3619e-04, -1.3743e-03, -1.3504e-03,  2.7635e-03,  3.4968e-03,
         9.4723e-03,  2.0221e-03,  1.2154e-03,  2.4420e-03, -5.6912e-04,
         1.0593e-04, -8.9666e-04, -6.9627e-04,  1.0041e-03,  1.3667e-03,
        -4.6844e-03,  1.3801e-03, -1.6607e-04, -9.0368e-04,  2.4125e-03,
        -3.1008e-03,  1.3915e-03, -1.5810e-03,  8.5120e-03, -1.9373e-03,
        -2.7267e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1464e-03, -8.4488e-01,  1.5809e-03,  1.0728e-02,  8.0096e-03,
        -6.1576e-03,  1.8577e-02,  3.6984e-03,  3.8630e-03,  2.3187e-02,
         2.4215e-03,  2.5206e-03,  4.7676e-03,  8.2187e-04,  1.9538e-03,
        -7.4436e-04, -5.5621e-03, -1.7982e-02,  3.7507e-03, -3.5825e-03,
        -1.7079e-02,  5.9619e-03, -7.0276e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5139e-02, -9.1510e-01, -5.1595e-03, -1.4231e-02, -5.8366e-03,
        -3.2091e-03, -1.0794e-03, -2.6481e-03,  2.7709e-04,  2.3949e-02,
         3.3677e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5162e-02, -8.6694e-01, -3.2471e-02, -7.7511e-03,  6.4581e-03,
        -1.0664e-02, -7.2645e-03, -2.9928e-05, -4.2335e-03,  1.3323e-04,
        -1.3521e-02, -1.1832e-02, -1.1284e-02, -2.2514e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4196e-02,  8.7774e-01,  3.4058e-02, -2.2905e-02, -5.9773e-03,
         3.8400e-03, -8.2534e-03,  8.7351e-05,  7.2363e-03,  8.0283e-03,
        -1.3260e-02, -4.4168e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0881e-02, -8.1000e-01, -2.4971e-02, -2.8317e-02,  1.8401e-03,
         3.8215e-04,  2.3181e-03, -5.4642e-03, -4.5537e-03, -3.7871e-03,
        -1.3950e-03, -4.7106e-03, -3.2899e-04, -2.4887e-03,  2.9026e-03,
         1.4471e-03,  6.9406e-04, -2.0886e-03,  1.4801e-03,  3.8330e-03,
         8.7780e-04, -1.0695e-03, -3.8974e-03,  2.7221e-03, -1.4006e-03,
         4.3121e-03, -4.7053e-03,  3.1313e-03, -5.2866e-04,  2.0883e-03,
        -2.9524e-03, -5.3683e-03, -2.1130e-03, -3.9809e-03, -1.0572e-03,
         2.2755e-03, -9.3732e-04, -2.5828e-03,  1.8178e-03,  1.7412e-03,
        -1.7083e-03, -1.2554e-03,  2.8014e-04, -1.5074e-03, -1.0597e-03,
        -3.5112e-03, -1.4767e-03, -8.8974e-04,  1.6940e-04,  4.8679e-03,
         7.2196e-04,  1.0455e-03,  1.8761e-03,  4.3616e-06, -6.1887e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 9.4556e-03, -8.6275e-01, -2.1962e-02, -8.0727e-03, -1.2620e-03,
         3.6107e-05,  4.9861e-03,  1.0467e-02,  4.3142e-03,  5.9925e-03,
         6.1390e-03,  2.3312e-03, -4.5596e-03,  3.0383e-03,  5.2505e-04,
         2.2045e-03, -9.5520e-03, -2.0176e-02, -2.7827e-03,  2.6640e-04,
         2.6696e-03,  1.2978e-02, -3.4778e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0283,  0.8439,  0.0081, -0.0168,  0.0106,  0.0092,  0.0054,  0.0088,
        -0.0076,  0.0024, -0.0031,  0.0022, -0.0137, -0.0048, -0.0016, -0.0019,
        -0.0187,  0.0129,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0109, -0.8745, -0.0085, -0.0121, -0.0155, -0.0080,  0.0106, -0.0040,
        -0.0101, -0.0021, -0.0030, -0.0089,  0.0027,  0.0054, -0.0011, -0.0062,
         0.0085, -0.0079,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1106e-02,  8.8065e-01, -2.6976e-02, -1.2041e-02,  1.3923e-02,
         5.6945e-03,  6.0420e-04,  2.2838e-03,  1.5216e-03,  6.0027e-03,
        -5.0529e-03, -7.1246e-03,  7.5919e-03,  2.8345e-03,  6.5965e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4032e-02,  8.5322e-01,  4.9783e-03,  3.5225e-03, -4.3033e-03,
         1.1155e-03, -4.0412e-04,  1.3122e-02,  5.7955e-03, -8.1324e-04,
         2.5751e-03, -4.6374e-03,  2.3707e-03, -2.5269e-03, -1.8848e-03,
         4.8034e-03,  3.7681e-03, -2.4178e-03, -1.6903e-03,  3.3868e-03,
        -3.6350e-03, -1.8687e-03, -1.1715e-04,  6.7767e-04,  3.1446e-03,
        -1.9757e-03,  7.3531e-03, -2.8091e-03, -1.3853e-02,  2.1456e-02,
        -6.7161e-04, -1.0668e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.7742e-03,  8.6669e-01, -2.1072e-02,  4.5903e-03,  4.6148e-03,
         3.6293e-03, -4.7571e-03, -5.8749e-04,  3.8022e-03,  2.2814e-03,
         2.5432e-03,  7.8719e-04,  1.9809e-03, -3.0380e-03,  1.6789e-03,
         3.5499e-03, -4.4962e-04, -8.4174e-03, -3.0984e-03, -3.3177e-03,
        -1.1239e-03, -1.0355e-02, -5.7171e-03, -4.2846e-03,  1.8042e-03,
        -3.3971e-04,  9.0741e-04,  1.4249e-02, -3.5416e-03,  2.3900e-03,
         3.4932e-03, -2.3633e-04,  4.8982e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3926e-02,  8.3509e-01,  4.3052e-03,  1.4250e-02,  7.2630e-05,
         1.4801e-03, -1.2109e-03, -7.2141e-03, -5.7058e-04,  4.1260e-03,
        -4.5007e-03, -1.1467e-02,  4.8398e-04, -1.0088e-03, -3.9698e-03,
         1.8293e-03,  4.3431e-03,  6.9313e-04,  3.1395e-03, -2.7363e-04,
        -1.0975e-03,  6.8158e-04,  4.2668e-04,  2.4633e-03,  2.8798e-03,
         1.0715e-03,  4.6015e-03,  5.1019e-03,  7.9454e-04, -1.5767e-03,
         5.8234e-03,  1.7424e-03,  4.7563e-03,  3.5148e-03,  7.0900e-04,
        -3.7485e-03,  1.2906e-04,  2.4973e-03,  2.5596e-03,  6.6845e-03,
        -4.8775e-04,  1.4259e-04, -9.7982e-05,  2.1356e-03,  4.7644e-03,
         1.5899e-03,  1.1035e-03, -1.2305e-03,  2.1484e-03, -1.5397e-03,
         3.7399e-03,  1.0597e-03,  2.0347e-03,  1.1191e-03,  4.9854e-04,
         4.9195e-03, -2.7887e-03,  1.7795e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2778e-02,  8.6458e-01, -1.0230e-02, -2.3333e-03, -7.5643e-04,
         4.1240e-03, -5.5115e-03, -4.3706e-03, -1.5010e-03, -2.1569e-04,
         1.9931e-03,  4.4343e-03, -5.7383e-03, -1.5566e-02,  2.6391e-03,
        -5.4354e-03,  8.0688e-03,  2.9487e-03, -5.0883e-05,  6.0428e-03,
        -1.6935e-04,  5.7502e-03,  4.1153e-03,  1.3125e-02, -3.5388e-03,
         3.9873e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7111e-02,  8.7249e-01, -1.4605e-02,  4.3497e-03,  4.7547e-03,
         1.7508e-02, -2.5340e-03, -2.1550e-03,  1.9622e-03,  4.3168e-03,
         1.9290e-04,  2.1496e-03,  2.1821e-03,  5.6068e-03,  2.8784e-05,
         2.0494e-03, -2.9831e-03, -9.1779e-04,  4.8854e-03, -7.8917e-03,
        -9.1110e-04, -1.7391e-03, -1.2117e-02, -9.5359e-03, -5.0194e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1262e-03,  8.4170e-01,  2.7546e-02, -4.3972e-03,  3.8738e-03,
        -4.5054e-03, -1.9551e-03, -2.2757e-03, -2.7252e-03, -3.9356e-04,
         6.2441e-04, -4.4981e-03, -2.7707e-03,  1.2070e-02,  1.8270e-04,
        -5.5601e-03, -4.6427e-03,  2.9032e-03,  3.8664e-03, -1.0999e-03,
         1.4233e-05, -2.1522e-03,  4.2199e-04,  9.6897e-04, -6.5752e-03,
        -5.8973e-03,  1.9012e-03,  8.4352e-04,  2.3782e-03, -4.4140e-03,
         2.8592e-03, -5.0787e-03, -5.3652e-03,  1.3437e-03,  8.7890e-04,
        -5.4780e-03,  2.6656e-03, -2.4257e-03, -1.7256e-03,  4.6329e-03,
        -3.0320e-04, -6.4749e-04,  5.0434e-03,  2.3672e-03, -1.9060e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0098,  0.8777, -0.0132, -0.0029,  0.0011, -0.0115, -0.0088, -0.0027,
        -0.0048, -0.0020,  0.0019, -0.0024, -0.0035,  0.0022,  0.0031,  0.0104,
        -0.0091, -0.0032,  0.0026,  0.0048,  0.0033, -0.0063,  0.0127,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0912,  0.7351, -0.0357, -0.0063,  0.0124,  0.0535,  0.0212,  0.0114,
        -0.0060, -0.0043,  0.0054,  0.0176,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-6.6510e-03, -8.5672e-01, -5.8117e-03,  1.2814e-02, -2.1118e-03,
        -7.3920e-03,  1.2381e-02,  5.6078e-03, -5.9034e-03, -4.3102e-04,
         1.9017e-02, -5.2930e-03,  8.7485e-03,  3.0316e-03,  1.3064e-03,
         6.2313e-03, -6.7547e-03,  5.2870e-03,  4.4974e-03,  6.9652e-03,
         1.3482e-04,  1.1585e-02, -5.3275e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6072e-02,  8.4677e-01,  1.7699e-02, -8.8394e-03,  4.3469e-04,
        -5.9046e-03, -4.2083e-04,  6.4508e-04, -3.0617e-03, -6.8460e-03,
        -4.8991e-04, -3.8018e-03,  4.5605e-03,  4.3818e-03,  1.2860e-03,
        -7.4508e-04, -3.7408e-03, -9.3390e-03, -1.3546e-03,  3.2617e-03,
        -6.1735e-03,  3.3217e-03, -5.0940e-03, -8.1205e-03, -4.0041e-04,
         1.9567e-04, -2.2192e-03, -2.2266e-03, -3.1423e-03, -4.2920e-04,
         4.7429e-03,  3.4169e-03,  1.0860e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9910e-02,  8.7387e-01,  8.2509e-03,  4.8525e-03,  1.4837e-02,
         1.9130e-03,  3.4039e-03, -1.1893e-02, -3.1749e-03, -1.3518e-02,
         3.7813e-03,  1.4123e-04, -6.1109e-03, -5.8985e-03,  4.7149e-03,
        -8.1301e-03, -8.7629e-03, -6.8272e-03, -1.4136e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0162,  0.9138,  0.0021, -0.0083, -0.0102,  0.0105,  0.0047, -0.0049,
        -0.0027,  0.0035, -0.0081, -0.0046, -0.0094,  0.0010,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.9337e-03,  8.9819e-01,  3.8559e-03, -1.1591e-02, -1.8505e-03,
         6.3913e-03, -1.8323e-03,  6.5444e-03, -4.9331e-03,  1.2600e-03,
        -4.4961e-03, -3.5503e-03,  2.0521e-03, -1.2025e-02,  4.7231e-04,
        -6.5260e-03,  5.2653e-03,  6.8925e-06, -4.6368e-04, -6.2817e-03,
         1.0975e-02,  3.5062e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.2581e-03,  8.7066e-01,  1.1828e-02,  1.4331e-02,  3.9103e-03,
        -1.4875e-03,  2.6784e-04, -7.9339e-03,  3.5012e-03, -1.2147e-02,
        -8.0511e-03,  6.2566e-03,  5.5649e-03,  3.1669e-03,  3.2257e-03,
        -6.2062e-04,  3.8242e-03,  2.3220e-03,  2.4600e-05, -6.0184e-04,
        -1.1143e-04,  1.1540e-05,  4.2487e-03, -7.9869e-03, -4.6689e-03,
        -6.0795e-04,  3.2695e-03,  3.4079e-03,  5.7828e-03, -9.1795e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0170,  0.8475, -0.0087,  0.0048, -0.0015, -0.0131,  0.0190, -0.0096,
         0.0099, -0.0015, -0.0049, -0.0015, -0.0082,  0.0132, -0.0097, -0.0245,
         0.0054,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6383e-02,  9.1952e-01,  1.0707e-02,  1.3968e-02, -3.8603e-03,
        -1.6134e-03, -4.5529e-03,  1.7661e-02,  3.1428e-03, -7.8869e-03,
         7.0585e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0230, -0.8186, -0.0139, -0.0056, -0.0044, -0.0051,  0.0027,  0.0038,
        -0.0033, -0.0058,  0.0083, -0.0024, -0.0032, -0.0486, -0.0054,  0.0104,
         0.0356,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1052e-02, -8.5616e-01, -5.0182e-03, -2.8314e-02, -1.0833e-03,
        -6.7235e-03, -1.5356e-03, -1.4275e-03, -7.0617e-03, -3.4663e-03,
        -4.0878e-03, -5.5758e-03, -5.1837e-03, -7.3131e-03, -9.9116e-03,
        -3.3295e-03, -1.6021e-03,  5.6605e-03,  2.2556e-03, -3.2008e-03,
        -1.9108e-03,  3.6110e-04, -2.0069e-03, -5.8421e-03,  7.2447e-03,
         2.6705e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7197e-02,  8.3564e-01,  1.6366e-02, -1.2790e-02, -4.4764e-03,
         1.1470e-02, -1.2949e-03,  5.9992e-03, -6.7600e-03, -1.3016e-03,
         1.1375e-03, -1.4489e-02, -4.6562e-04, -3.0017e-03, -3.2565e-03,
        -1.3301e-03,  2.8157e-04,  2.6321e-03, -2.0456e-03, -3.1457e-03,
         1.8008e-03, -8.3888e-04, -1.8936e-03, -3.1377e-03,  5.5400e-03,
        -2.5559e-03, -8.6360e-04, -1.6017e-03,  1.0974e-03,  5.5857e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9219e-03,  8.7614e-01,  2.5260e-02, -1.6170e-02,  1.4793e-02,
        -7.9958e-03, -7.5164e-03, -1.0557e-02, -1.3301e-03, -4.9069e-04,
        -4.3724e-03, -1.2870e-02, -9.2258e-05, -5.5718e-03,  3.4791e-03,
         2.6733e-03,  1.5006e-03, -1.2563e-03,  4.0074e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 4.1464e-02, -7.8927e-01, -1.1419e-02, -1.5699e-02, -5.2337e-03,
        -2.4075e-03,  3.4061e-03, -6.9495e-03,  4.7201e-04, -6.3401e-03,
        -7.6202e-03,  5.0541e-04, -1.5757e-03, -1.0193e-02,  3.3976e-03,
         6.3219e-03, -1.4159e-02, -5.0066e-03, -6.3014e-03, -8.4725e-03,
        -4.7824e-03, -3.9050e-03, -9.3315e-04,  3.8237e-03,  3.4222e-03,
        -2.4371e-02,  1.2545e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0322,  0.8590, -0.0046, -0.0062,  0.0081, -0.0125, -0.0068,  0.0038,
         0.0177,  0.0055,  0.0057,  0.0083, -0.0143, -0.0154,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6070e-02,  8.7246e-01,  1.3912e-02, -1.6414e-02, -5.8284e-03,
        -7.5010e-03, -3.8457e-03, -2.7453e-03, -6.3128e-03,  3.5324e-03,
         4.7607e-03, -5.0544e-04, -2.0026e-03, -1.6148e-03, -6.6903e-04,
        -8.4278e-03, -1.3466e-03, -6.1058e-03, -1.3422e-03, -3.7877e-03,
        -1.6864e-03, -3.7058e-03,  1.1304e-03, -4.6956e-03, -8.0434e-03,
         1.5551e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5241e-02,  8.6321e-01, -2.2590e-02, -1.2125e-02, -5.3208e-03,
        -1.5411e-02,  8.7155e-03,  2.1833e-03,  3.3913e-05, -3.6059e-03,
         5.4937e-03, -4.6400e-03, -7.3272e-03, -1.4102e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8849e-02,  8.8930e-01, -3.7594e-03, -8.9581e-04, -5.3267e-03,
        -1.7955e-03, -1.8927e-03, -6.7308e-03,  3.5404e-03,  3.5030e-03,
         9.9497e-04, -6.7497e-03,  2.4022e-03,  7.9673e-04,  5.6461e-03,
         1.5755e-03, -1.6293e-03, -1.0922e-05, -3.3957e-03,  6.0183e-03,
        -1.5663e-03,  2.2041e-03,  1.2453e-03, -2.9880e-04,  5.5104e-03,
        -4.0339e-03,  6.1298e-04,  3.4701e-04, -5.6048e-03,  4.5975e-03,
        -1.2183e-03,  7.9430e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3897e-02,  8.7106e-01,  1.9791e-02,  4.4480e-03, -1.9448e-02,
         8.2311e-03, -3.0742e-03, -3.2217e-03,  2.0567e-04, -4.4517e-03,
        -5.7786e-03,  8.2371e-03, -5.2507e-04,  3.3468e-03, -5.8755e-04,
         2.8045e-03,  8.6409e-04, -3.4371e-03,  1.4299e-03,  1.0283e-03,
         1.4097e-03,  5.7810e-03,  2.4861e-04, -5.0161e-03, -1.1674e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0090,  0.8836,  0.0640, -0.0055,  0.0012,  0.0367,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.1527e-03,  8.3870e-01,  3.3383e-02, -1.0265e-02, -6.9116e-04,
        -1.0991e-02,  4.9810e-03, -6.0217e-03,  1.7696e-03, -3.4347e-03,
        -5.8346e-03, -1.8589e-03,  5.8802e-05,  1.1604e-02, -4.9117e-03,
        -1.7691e-03, -4.3595e-03,  3.4080e-03, -1.1691e-03, -1.5324e-03,
        -5.0771e-03, -9.2613e-05, -5.8046e-03, -6.2295e-04, -1.9560e-03,
         9.2635e-03, -1.7989e-03, -1.3981e-03, -2.8590e-03, -5.7861e-03,
         8.6152e-04, -7.7639e-04,  1.5318e-03, -6.8591e-04, -1.3241e-03,
        -2.7680e-03,  2.4971e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3545e-02, -8.8115e-01, -2.6950e-03,  1.7133e-03, -5.2806e-03,
        -8.2720e-03, -3.6346e-03, -1.3914e-04,  2.7079e-03, -1.1232e-03,
         3.3436e-03,  9.9327e-04, -2.4890e-03,  3.1994e-03, -1.4711e-03,
         3.6985e-03,  8.9231e-03, -4.3993e-04, -1.4043e-03, -1.5265e-03,
         1.0060e-02,  3.7609e-03,  5.9683e-03,  5.4075e-03,  1.8848e-03,
         6.7077e-03,  5.4159e-03, -3.2511e-03,  2.2132e-04, -1.6978e-03,
         1.3554e-04, -2.0207e-04,  2.1030e-03,  3.4912e-03,  1.9418e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7013e-02,  7.9569e-01, -9.9788e-02, -8.4262e-03, -2.4759e-03,
        -2.8785e-05, -5.4990e-03, -3.8651e-04, -4.9438e-03,  1.3476e-03,
        -9.3078e-03, -9.9293e-03,  6.2937e-03, -5.5317e-03,  1.3339e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9013e-02, -8.6423e-01,  2.5132e-02, -1.3165e-03, -4.4263e-03,
        -6.1233e-03, -8.1303e-03, -6.6585e-03, -2.7491e-03, -3.5160e-03,
        -3.1259e-03, -8.1263e-05,  1.1903e-03,  1.4287e-03, -5.9889e-04,
        -7.5613e-04, -1.4913e-03,  3.2318e-03, -6.6108e-03,  4.3161e-04,
        -1.7859e-03, -8.2325e-04,  4.0300e-03, -2.7224e-03, -1.4031e-03,
        -1.9079e-03, -2.0366e-03,  8.5655e-04, -1.8769e-04,  1.9228e-03,
        -1.8993e-03, -6.6420e-03, -3.5369e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4954e-02, -8.7922e-01,  9.5991e-03,  3.2528e-03,  3.3007e-03,
         1.5085e-02,  1.0852e-03,  2.9708e-04,  1.0866e-03,  7.8176e-04,
        -1.4791e-03, -1.2275e-03, -2.4213e-03,  6.0266e-03,  4.6694e-03,
        -2.1310e-03,  2.9362e-03,  4.6376e-03, -4.8362e-04, -3.3975e-03,
         4.6566e-04,  1.5595e-03, -7.6227e-04, -9.0635e-04, -3.2099e-03,
         3.0925e-04,  6.6256e-03,  3.2162e-03, -4.8769e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 5.7224e-02,  8.7562e-01, -5.1048e-03, -6.2055e-03, -5.7823e-03,
        -2.8356e-03, -3.5466e-03,  3.1923e-03,  4.8246e-05, -1.4607e-02,
        -4.4571e-03,  2.7238e-03,  3.1394e-03, -5.5196e-03, -9.9937e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3244e-03,  8.7222e-01, -8.7983e-03, -1.5439e-02, -7.3136e-03,
         3.3836e-03, -5.8095e-03,  9.6590e-04,  6.5401e-03,  1.3099e-03,
        -4.3619e-03, -1.3820e-03,  5.0668e-04,  8.6057e-04, -1.5573e-03,
        -2.1093e-03,  1.5621e-03, -6.2585e-05,  2.1354e-03,  1.4466e-04,
        -1.5457e-03, -4.2539e-03, -3.4068e-03,  9.2460e-04, -1.1990e-03,
         5.4379e-04,  9.6870e-05,  4.4367e-03, -3.5411e-03, -9.7582e-07,
         1.8701e-03, -2.9817e-03,  2.8103e-04, -1.1788e-03, -8.9749e-04,
        -8.9320e-04, -5.1337e-04,  7.7261e-04, -8.5378e-04,  2.6782e-04,
        -1.2634e-03, -2.0578e-03, -1.8690e-03, -2.1345e-03, -1.5190e-04,
        -3.0456e-03, -2.7870e-03, -4.9786e-03,  9.4660e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1049e-02,  8.2898e-01,  2.1895e-02, -7.2683e-03,  2.9764e-04,
        -4.2717e-03, -9.8551e-03, -2.5743e-06,  2.7637e-03,  1.1564e-03,
        -3.7637e-03, -1.7760e-04,  1.1280e-03, -2.0618e-03, -1.5015e-03,
        -8.6517e-03,  2.2008e-03,  1.2026e-03, -2.7274e-04,  7.9005e-04,
        -2.6045e-03,  3.7173e-03, -4.7295e-03,  3.5989e-04, -9.8843e-04,
         8.2701e-04,  4.6304e-04, -2.7643e-03,  3.3335e-03, -4.5362e-03,
         2.7373e-03,  3.0107e-03,  1.0015e-03,  1.4003e-03,  8.2348e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4905e-02,  9.0443e-01, -2.0258e-02,  1.6531e-03, -5.7970e-03,
         5.7757e-03, -1.3010e-03,  1.9805e-03,  4.7238e-03, -5.4033e-03,
        -1.7451e-03, -4.2225e-03, -3.4881e-03,  1.8434e-03,  1.1784e-03,
         9.4142e-03,  1.0419e-02,  7.6948e-05, -1.3862e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8337e-02,  8.6606e-01, -6.8909e-03,  2.9433e-02,  9.2080e-03,
         8.1989e-04,  3.0675e-03, -3.3497e-03, -7.0902e-03,  2.6395e-04,
         7.0810e-04,  2.3421e-03,  2.9900e-02, -1.6298e-02, -6.2282e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6157e-03,  9.0352e-01,  7.0601e-03,  4.0125e-03,  2.7927e-03,
        -2.9181e-03, -7.5893e-03, -4.4260e-03, -5.2085e-03,  8.9417e-04,
        -7.9208e-03,  8.4240e-03, -5.4269e-03,  2.3215e-03, -5.2572e-03,
        -1.5717e-03, -6.6957e-04, -9.2280e-03, -9.2280e-03,  1.1752e-03,
        -5.5577e-03, -1.1833e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0041e-02,  8.6849e-01,  2.0897e-03, -8.6741e-03, -5.0270e-03,
         1.5506e-03,  3.1756e-04, -2.7416e-03, -1.8812e-03,  1.7776e-03,
        -1.3458e-02, -1.5019e-03,  5.1019e-03,  2.4133e-04, -6.9819e-03,
         5.4625e-04, -5.3565e-03,  1.9304e-03,  6.2628e-03,  1.6318e-03,
        -2.4326e-03,  1.7940e-03,  2.9329e-03,  2.9125e-03,  5.6477e-03,
         8.2291e-04, -2.4237e-03,  1.1798e-03,  3.0226e-03,  1.1227e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0403,  0.8419, -0.0438, -0.0180,  0.0048, -0.0070, -0.0102,  0.0073,
         0.0114, -0.0134,  0.0019,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3545e-03, -9.0914e-01,  1.5612e-02,  5.4835e-03, -5.6234e-03,
        -3.3330e-03, -4.9025e-03,  1.3867e-03, -2.9083e-04, -1.2294e-03,
         4.8923e-03, -8.6189e-03, -6.5358e-03, -9.7651e-03,  8.0682e-04,
        -4.0848e-03, -2.3070e-03, -4.3564e-04,  5.0446e-04,  4.5442e-04,
         7.8128e-03,  4.4232e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5538e-02,  8.7166e-01, -1.5203e-02,  3.2131e-03,  3.5749e-03,
        -1.1467e-02, -2.1563e-03,  2.0133e-03, -3.5068e-03, -2.4486e-04,
        -3.5081e-03, -8.1857e-04,  4.6415e-03,  6.0201e-05,  2.6072e-03,
        -7.4507e-04,  7.0788e-03,  5.1461e-03, -2.8874e-04, -1.0856e-03,
        -1.8118e-03, -6.0826e-03, -8.4430e-03, -8.9891e-04, -7.3891e-04,
         1.0789e-03,  4.1111e-03,  4.5882e-03, -4.2270e-03,  3.3920e-03,
         2.1552e-03,  5.1794e-03,  2.7325e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1919e-02,  8.3826e-01,  2.7974e-02,  2.1812e-02, -2.5227e-04,
         8.5655e-03,  2.3712e-03,  5.0204e-03,  1.0156e-03,  7.0017e-04,
         1.7454e-03,  8.4900e-04,  1.8724e-03,  6.0070e-04,  1.8977e-03,
         9.7468e-04,  2.3888e-03, -4.2936e-03, -6.5264e-03,  2.6060e-03,
         1.2805e-03, -1.6766e-02,  3.5067e-03,  4.4077e-03, -1.2024e-02,
         1.0507e-02,  5.9191e-03, -3.9495e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2736e-02, -9.0294e-01, -1.3849e-02,  9.5878e-04, -4.9271e-04,
        -3.0613e-03,  1.1965e-02, -7.7575e-03, -2.3375e-04, -2.3660e-03,
        -9.5906e-04, -3.9127e-03, -3.1987e-03, -4.2124e-03,  6.0382e-04,
        -1.1861e-03, -5.8916e-03,  1.7308e-03,  1.8621e-03, -1.2058e-05,
         8.3733e-03, -1.4567e-04, -3.5580e-03, -5.6839e-03,  2.3143e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 2.0668e-02,  8.0809e-01,  1.3162e-02, -7.5068e-03, -5.0768e-03,
         2.7930e-04, -5.5105e-03, -5.5458e-03,  3.5026e-03, -9.6660e-04,
         4.2469e-03,  1.3726e-03, -3.5482e-03, -3.5428e-03, -3.0055e-03,
        -1.2845e-02, -5.3515e-03,  3.6581e-03, -3.3245e-03,  2.0348e-03,
         9.6454e-04, -2.7194e-03,  2.1271e-03,  1.4905e-03,  2.1577e-03,
        -4.5570e-03,  3.1606e-03, -1.6015e-03, -4.2982e-03, -4.9071e-03,
        -3.9755e-03, -2.5876e-03, -9.6973e-03, -2.2317e-03, -1.5322e-03,
        -8.2985e-04, -1.5309e-03, -6.0780e-04,  2.8328e-03,  2.3594e-02,
        -9.3546e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5161e-02, -8.6930e-01, -1.4393e-02,  8.8327e-03,  3.7222e-03,
        -1.0818e-03,  3.5513e-03, -9.5673e-03,  2.5265e-03,  1.0759e-03,
        -5.7484e-04,  5.0548e-03, -2.4408e-03, -9.7573e-04,  2.8576e-03,
         7.1009e-03, -1.7677e-03, -1.5323e-04,  1.9088e-03,  8.6330e-03,
         3.2933e-03,  9.8670e-03, -4.1369e-04,  5.2500e-03, -1.0498e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.9053e-03,  8.3252e-01,  3.7730e-02, -2.9593e-02,  3.5123e-03,
         6.7428e-03,  8.2509e-03,  1.0763e-02, -2.0102e-03, -3.7615e-03,
         4.3504e-03,  9.3531e-03, -3.9523e-03,  2.2758e-03, -4.0066e-03,
        -3.0732e-03,  8.5187e-04,  6.5440e-04, -2.2906e-02,  2.5708e-03,
        -6.2209e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0134,  0.8763,  0.0066,  0.0012, -0.0019, -0.0113, -0.0025, -0.0139,
         0.0029,  0.0064, -0.0146,  0.0030, -0.0013, -0.0108, -0.0029, -0.0069,
        -0.0156, -0.0085,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1429e-02,  8.5505e-01,  6.7554e-03, -1.0213e-02, -6.6056e-03,
         4.6428e-03,  1.0323e-03, -8.5878e-03, -5.0211e-03, -3.4097e-03,
        -3.0901e-03, -9.7705e-04,  5.3188e-04,  1.8014e-03, -2.7765e-03,
         4.3264e-03, -3.5389e-03,  1.6550e-03, -2.4530e-03, -4.2231e-03,
        -2.2360e-03, -2.2570e-03, -1.6432e-03, -9.9195e-04, -2.0154e-05,
        -2.1006e-03, -1.5533e-03,  7.3388e-05, -3.9723e-04, -2.3362e-03,
        -8.7583e-04,  2.0623e-04,  1.8128e-03, -8.6696e-04,  5.4596e-03,
         7.1252e-03,  2.1376e-02,  5.4962e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0537, -0.8204,  0.0029, -0.0129, -0.0021, -0.0056,  0.0075,  0.0048,
        -0.0100, -0.0034,  0.0085,  0.0022, -0.0029, -0.0015,  0.0113, -0.0056,
        -0.0136, -0.0105,  0.0076,  0.0130,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9984e-03, -8.8844e-01, -9.0504e-03, -2.3651e-02, -1.9967e-02,
        -3.8266e-03, -5.3474e-03, -1.7443e-02, -4.9307e-03, -9.3311e-04,
        -1.1104e-02,  3.0716e-04, -4.6419e-03, -5.3571e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4243e-02,  8.4038e-01,  3.7267e-02, -5.0921e-03, -7.2137e-04,
        -1.4307e-03, -7.8803e-03, -2.7961e-03, -1.4166e-03, -2.0586e-03,
        -4.0785e-03, -5.6291e-03,  3.0484e-03, -4.8604e-03,  2.3296e-03,
         1.7890e-03,  2.8963e-04,  3.7531e-04, -2.5943e-03,  2.6628e-05,
        -4.3602e-03,  2.6105e-02,  2.5766e-03,  1.0756e-03, -6.4841e-04,
        -3.4797e-03,  5.5818e-03, -1.7585e-03,  2.4196e-03,  1.2178e-02,
         1.2342e-03, -2.7495e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8650e-02,  8.4297e-01,  1.5655e-02,  2.4088e-02, -1.0168e-02,
         6.0679e-03, -2.8152e-03, -2.2451e-02,  6.8115e-03, -1.1761e-02,
        -1.2278e-03, -2.6696e-03,  1.9287e-03, -1.6524e-03,  3.7252e-03,
        -4.7227e-04, -2.1978e-03, -4.1562e-03, -7.3421e-03, -3.1926e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8336e-03,  8.1786e-01,  4.1546e-03,  3.0074e-03,  1.6633e-02,
         1.0656e-02,  2.1624e-04, -4.9710e-03, -2.4851e-03,  6.7912e-04,
        -2.9738e-03,  6.2601e-03, -6.7270e-04, -1.4007e-03,  4.6219e-04,
         1.1130e-03, -2.5837e-03, -1.3866e-03, -4.1368e-03,  3.9737e-04,
         3.4068e-03, -1.0722e-02, -8.7540e-03, -3.6180e-03, -4.9449e-03,
        -3.9723e-03,  4.7491e-04, -3.1095e-03, -1.0466e-02,  1.0260e-03,
         2.6259e-03,  4.4220e-03,  1.0745e-03,  5.4102e-03,  7.1979e-03,
        -3.3941e-04,  1.9696e-03, -3.3351e-03, -1.6609e-03,  2.3758e-03,
        -5.6283e-03,  5.2005e-03,  2.7359e-03,  1.6321e-03,  5.6801e-03,
         5.4710e-03, -7.8579e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.1345e-03,  8.8049e-01, -5.0057e-03, -1.2540e-02,  2.5302e-02,
         6.5048e-04,  1.2993e-02, -5.1140e-03,  1.7751e-02, -2.6146e-04,
        -1.4889e-02, -1.6880e-03, -1.2709e-03,  1.0442e-02, -4.4700e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5874e-02,  8.4247e-01, -4.6968e-04, -1.3306e-03,  4.0672e-03,
        -7.8401e-03, -3.4190e-03,  6.8944e-05,  1.3895e-04, -2.1410e-03,
        -3.4431e-03, -6.1747e-04,  5.5724e-04,  8.4356e-03, -4.2714e-03,
        -1.8519e-04,  6.1663e-03, -5.7281e-04, -3.6031e-03, -4.6344e-03,
        -9.5895e-05, -2.5843e-03, -1.9755e-03,  1.5582e-03, -1.1025e-03,
        -3.1698e-04,  1.6475e-03,  3.4181e-02, -1.5337e-03,  3.3386e-03,
        -1.5499e-03, -3.0638e-03,  1.0286e-03,  6.6583e-05,  2.5099e-03,
        -1.9377e-03,  2.4800e-04,  2.3241e-03, -4.2214e-03, -7.8282e-04,
         1.2172e-03,  3.6178e-04, -3.0807e-03,  3.2039e-04, -1.3757e-03,
         4.3076e-04,  5.3658e-03,  1.4753e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-1.7342e-02,  8.3014e-01, -3.2224e-02, -9.3187e-04,  1.6508e-02,
         2.8246e-03, -1.4741e-03,  6.2010e-03,  1.7434e-02,  6.6991e-03,
         4.8316e-03, -1.1857e-02, -9.9070e-04,  1.0203e-03,  1.2165e-03,
        -3.2436e-03, -1.2360e-02, -1.0538e-03,  4.0264e-04,  4.3405e-05,
         2.4378e-03,  7.1748e-03,  1.4969e-03,  6.2264e-03,  3.2225e-03,
        -1.0642e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7963e-02,  8.4253e-01, -1.1819e-02,  1.6547e-02,  2.6698e-04,
         3.8435e-03, -7.2018e-03,  4.2250e-03,  7.7428e-03,  2.1099e-03,
         2.5139e-02, -6.2635e-04, -1.5726e-03,  3.2535e-03, -2.5829e-03,
         7.8904e-04,  4.8387e-04, -8.3785e-03,  1.8968e-03,  5.6240e-03,
         5.4066e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8093e-02,  8.6032e-01,  9.5693e-03,  2.9911e-03, -6.8881e-03,
        -3.2938e-03, -3.7475e-04,  5.3691e-05,  1.3888e-03,  1.2854e-03,
        -1.5637e-03,  2.9952e-03, -8.5500e-03, -1.9996e-03,  2.3480e-03,
        -1.4350e-03,  7.7494e-03, -5.8385e-03,  9.1527e-03, -6.6694e-04,
        -2.1488e-03, -5.4828e-05,  6.3178e-03,  1.8712e-03,  1.2776e-02,
         8.9357e-05,  7.2827e-04,  1.2466e-03,  6.1447e-04,  6.2124e-05,
         8.8610e-04, -4.4963e-04, -1.6369e-03,  2.8221e-03, -3.7835e-03,
         7.9575e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3214e-02, -8.2137e-01,  1.3538e-02, -3.0446e-02,  2.5652e-03,
        -4.7162e-03,  1.0926e-02,  3.2389e-03, -5.3528e-04, -2.4788e-03,
        -1.7393e-02, -6.5762e-03,  2.5751e-03, -2.8292e-03,  7.5156e-04,
        -3.7351e-03,  6.7569e-05,  6.3832e-03, -6.3980e-03,  2.0265e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2913e-02,  8.7163e-01, -1.9100e-02,  1.2535e-02,  7.3433e-03,
        -4.9934e-03, -2.6093e-03, -9.1097e-03,  3.4006e-03, -6.3305e-03,
        -1.6210e-03, -4.1081e-04, -4.6157e-03,  3.2472e-03,  4.8328e-03,
         6.0406e-03,  2.2357e-03, -2.4168e-03, -5.2650e-03, -6.2552e-03,
        -1.3093e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2119e-02, -9.3807e-01,  1.1561e-03, -1.0727e-03, -6.6905e-04,
         2.2118e-03,  7.2998e-03,  1.6165e-02,  8.9937e-03, -4.6028e-03,
         4.7589e-03,  2.5071e-03,  3.6921e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2371e-02,  8.3066e-01,  1.5173e-02, -3.9762e-03,  1.8111e-04,
        -4.6739e-03, -1.1837e-02, -9.1456e-03, -6.7384e-03, -5.3542e-03,
        -1.1782e-02, -1.1321e-02, -4.2645e-03, -2.4752e-03, -1.1332e-03,
         1.0180e-03, -6.8902e-03,  1.3925e-05, -2.4052e-03, -2.1514e-03,
        -1.5087e-03,  3.5907e-03, -1.0627e-03, -3.1948e-03, -1.6464e-03,
        -4.0760e-04, -1.2056e-02,  6.3494e-04,  1.6796e-03, -3.3493e-03,
         7.2156e-03,  2.2350e-03,  3.9602e-03, -5.1493e-03,  1.1563e-03,
         6.7789e-03, -8.0780e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.5185e-03,  8.6954e-01,  4.3554e-02, -6.9606e-03,  1.4944e-03,
         4.3938e-03,  7.4407e-04,  1.4283e-02, -4.7503e-03,  3.6092e-03,
         3.8055e-03, -2.8634e-03, -1.7201e-03,  1.9587e-03,  7.3826e-03,
        -1.2350e-03,  3.0372e-04, -5.1862e-04, -3.6661e-03, -5.6732e-03,
        -3.4306e-03,  7.1567e-04,  7.7393e-03, -1.3882e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0715e-02, -7.9595e-01, -1.0981e-02, -1.0837e-02, -6.7284e-03,
        -1.4833e-03, -7.8711e-03, -1.3489e-02, -1.4447e-02,  3.6749e-03,
         1.2476e-02, -6.7641e-03, -1.1821e-02,  3.4732e-03, -1.1612e-02,
         7.3183e-03, -6.5199e-03, -8.3505e-03, -2.0257e-03,  1.1766e-03,
        -1.6418e-03,  1.8947e-03,  1.2363e-04,  6.0269e-03,  8.8090e-04,
         9.7921e-04, -2.4883e-03,  1.8636e-03, -1.1507e-03,  9.8021e-04,
         1.6685e-02,  7.5745e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5125e-02,  8.0529e-01,  1.2325e-02, -1.6546e-02,  4.4097e-03,
        -1.5466e-02,  2.5139e-03,  4.1801e-03,  1.8422e-03,  2.8926e-03,
        -1.3922e-03,  2.8572e-03,  2.5277e-03,  5.9969e-04,  8.8829e-04,
         2.6521e-03,  3.1842e-03, -1.2119e-03,  3.8173e-03,  4.4054e-03,
         3.9530e-03, -2.0372e-03,  3.5941e-03, -1.0921e-04, -2.1156e-03,
         9.0006e-03,  5.4576e-03, -3.7833e-03,  1.2796e-02, -9.6632e-03,
         3.3600e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1248e-02,  8.0715e-01, -2.6718e-02, -5.3093e-03,  9.3484e-03,
        -9.7370e-03, -8.5844e-03, -2.1819e-02,  3.5335e-03, -1.4090e-03,
         9.8067e-03,  2.2792e-03,  7.8576e-04, -2.3189e-02, -1.4551e-04,
        -6.0117e-03,  2.5220e-03,  1.1231e-03,  3.4023e-04,  4.3085e-03,
        -8.6448e-03, -1.0632e-02, -5.3544e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3265e-03,  8.9814e-01, -1.3906e-03, -1.0403e-02, -3.8803e-03,
         7.5494e-03, -2.9913e-03, -3.7434e-03,  3.0028e-03,  6.3661e-04,
         2.9653e-03,  2.7277e-03, -4.4997e-03, -1.2261e-02,  7.0009e-04,
         6.5577e-03,  6.2308e-04, -2.5985e-03,  1.0624e-03, -8.4596e-04,
        -9.2302e-03, -1.8739e-03,  2.7675e-03,  8.2619e-04, -6.0651e-03,
         9.1703e-04, -2.1058e-03,  8.3134e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-3.6245e-02,  8.7142e-01, -1.0277e-02, -5.8079e-04, -3.6561e-04,
         1.9787e-03,  2.0563e-03, -5.4748e-03, -5.1125e-03, -4.6791e-03,
         1.5147e-03, -3.8003e-03, -5.1833e-03,  4.0827e-03, -2.1512e-03,
         6.6166e-03, -1.9858e-03,  1.7915e-03,  6.0473e-03, -6.0734e-03,
        -9.4048e-04,  3.0422e-04,  4.0484e-05, -3.7152e-03,  3.9018e-03,
        -3.2471e-04,  7.6704e-03,  5.2992e-03,  3.6730e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.9634e-02,  8.5637e-01, -3.1921e-03,  7.9818e-03,  3.1772e-02,
        -1.3468e-03, -8.2994e-04,  4.0693e-04, -4.5787e-03,  1.5046e-03,
         9.8699e-05, -5.6429e-03, -5.9805e-03, -1.7569e-03,  4.8295e-03,
        -3.1477e-04, -6.0731e-03,  2.2733e-03, -4.2644e-03, -4.2175e-03,
        -1.5948e-03, -1.1280e-02, -8.5395e-03,  5.5231e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5179e-02, -8.2034e-01,  2.8496e-03, -1.7746e-02, -8.3293e-03,
        -1.9405e-02,  3.3154e-03,  8.5805e-03, -1.2217e-02,  2.6670e-03,
        -1.3313e-03, -1.0921e-02, -3.9393e-03,  2.2862e-03,  4.7089e-03,
        -1.6447e-03,  2.4759e-03, -2.3082e-04, -3.2488e-03, -3.6939e-03,
        -1.2708e-02,  2.9277e-03,  8.1944e-04,  1.5866e-03, -2.8268e-03,
         2.5403e-03,  5.9407e-03,  6.0721e-04,  2.9860e-03,  3.7369e-03,
         4.5935e-03,  2.1954e-03,  6.6630e-03,  4.7544e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2731e-02, -8.7767e-01, -1.1486e-02, -1.0728e-02,  3.0004e-03,
         1.2969e-03, -1.2550e-03,  1.7744e-03, -1.8581e-03, -7.4550e-03,
         3.9694e-03,  6.7600e-03, -1.7849e-03,  2.8702e-03, -2.8854e-03,
        -2.1276e-03,  2.3063e-03, -2.3438e-04, -3.8364e-04, -9.6890e-03,
         7.2647e-03,  4.6698e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4641e-02,  8.7934e-01, -2.8206e-02,  1.9271e-03, -2.4010e-04,
        -3.2743e-03,  1.0447e-03,  8.5044e-03,  1.8853e-03,  5.0216e-03,
         2.2134e-02, -3.7190e-03, -1.5821e-03, -8.3400e-04, -2.4962e-04,
        -3.4088e-03, -3.8712e-03,  1.0118e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.2728e-03, -8.8348e-01,  1.3593e-02,  9.0373e-03, -1.7344e-03,
         5.2905e-03, -3.3274e-03,  1.3784e-04, -2.5176e-03, -4.2158e-03,
         2.6869e-03, -1.1746e-04,  4.7528e-03,  1.6498e-03,  3.7075e-03,
         3.9818e-04, -6.5199e-04,  1.1805e-02,  1.1941e-02,  5.9022e-03,
         1.1881e-03,  1.6871e-02, -6.7229e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.9220e-03, -8.7691e-01,  2.3521e-02,  4.9159e-03,  3.6745e-02,
        -5.0878e-03, -6.4216e-03,  6.9589e-03,  8.9074e-03,  4.6876e-03,
        -3.3563e-03,  1.1916e-03,  5.6022e-03, -5.3908e-03, -3.4862e-03,
         1.4163e-03,  4.7903e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0073,  0.8484,  0.0357,  0.0036, -0.0046,  0.0066,  0.0152,  0.0034,
        -0.0028,  0.0137,  0.0095, -0.0018, -0.0017,  0.0043,  0.0037, -0.0034,
        -0.0233,  0.0109,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3966e-02, -8.2733e-01,  3.8360e-02,  2.3008e-02,  5.3949e-03,
         1.5380e-03,  1.5650e-03,  2.2067e-04,  7.1800e-04, -2.5490e-02,
         5.8340e-04,  6.7063e-03,  2.5792e-03, -5.7544e-03, -5.1474e-03,
        -7.5588e-03,  6.2792e-03,  1.6567e-03, -1.4268e-03, -3.5437e-03,
        -1.7551e-04,  6.8569e-04, -6.5913e-03,  3.2199e-04,  4.3393e-03,
         9.0623e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3714e-02,  8.4961e-01,  1.9153e-02, -9.4857e-03, -2.5212e-03,
        -1.7160e-02, -2.1145e-03,  3.4292e-03,  4.1529e-03, -9.1237e-03,
        -7.3819e-03,  1.0223e-04, -7.6995e-03, -3.1240e-03, -3.4406e-03,
         5.5372e-05,  1.2612e-04,  1.0435e-03, -2.0609e-03, -5.5181e-03,
         5.5129e-04, -7.0485e-04,  1.1278e-03,  1.7290e-03, -3.2451e-03,
         2.1855e-03, -4.6903e-03,  4.5479e-03,  3.6601e-03, -2.3469e-04,
        -6.3031e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.0161e-02,  8.9078e-01, -8.3648e-03,  3.4403e-03,  2.1497e-03,
         1.0966e-03, -7.6066e-04, -1.4297e-03,  1.9078e-03, -1.3430e-02,
        -2.9597e-03, -4.8482e-03,  1.2715e-03, -2.4016e-03, -9.2446e-03,
        -6.9943e-03,  4.8716e-04, -8.2723e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0038,  0.9352, -0.0185,  0.0021, -0.0071, -0.0036, -0.0140,  0.0157,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 4.0350e-02,  8.1959e-01,  1.6651e-02, -1.5863e-03, -3.3336e-03,
         2.5791e-03, -3.8432e-03,  5.0988e-04, -3.0231e-03,  6.5542e-03,
         3.4022e-03, -2.5011e-03,  6.8894e-04, -1.8262e-03, -7.2745e-04,
         6.6440e-03,  1.2939e-03, -1.7907e-03,  3.7141e-03,  1.5840e-02,
         1.0800e-02, -1.2608e-02,  5.8404e-04,  1.2445e-03,  4.8845e-04,
        -1.2494e-03,  5.9299e-04,  3.4625e-03,  6.0823e-03,  2.3603e-03,
         3.1228e-04,  1.2991e-03,  2.7929e-03, -4.3683e-04,  7.1118e-04,
         3.6570e-03,  7.2790e-05,  1.2048e-02,  2.7449e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5245e-03,  9.1365e-01, -1.4030e-03, -2.2953e-03,  3.2924e-03,
         4.5842e-03, -1.0045e-02, -7.1757e-04, -4.3659e-03,  2.1135e-03,
        -9.4069e-04,  7.8327e-04,  2.6659e-03, -6.3722e-03,  4.8921e-03,
         3.1238e-03, -2.4282e-03, -9.2673e-03, -1.8129e-03, -2.7485e-03,
        -2.3529e-03, -2.1640e-03, -4.1151e-03, -3.8699e-04, -3.6269e-04,
        -3.0193e-03, -2.4512e-03, -2.1183e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.5702e-02, -8.8626e-01, -3.8361e-03,  6.1213e-03, -3.8737e-03,
         6.4837e-03,  1.3411e-03, -2.8743e-03, -8.8916e-03, -4.6889e-03,
        -9.2543e-03,  5.9922e-04,  7.0011e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5350e-02,  8.8842e-01,  1.1331e-02,  2.9469e-03, -6.5478e-03,
        -5.4095e-03, -3.6481e-03, -1.2612e-04,  3.1394e-03, -5.5535e-03,
        -9.9022e-03, -1.1008e-02, -5.7986e-03, -1.0651e-02,  1.6917e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.1638e-03, -8.5586e-01, -1.4289e-02, -1.6920e-03, -9.4409e-03,
         1.0166e-02, -7.9854e-03,  8.6304e-03,  2.4418e-03, -3.7721e-03,
        -1.3866e-03, -2.4938e-03, -2.9774e-03,  1.3099e-03,  2.4382e-03,
         2.5298e-03,  4.0542e-05,  4.2324e-03,  2.0893e-03, -4.0430e-04,
        -1.1118e-04,  2.2581e-04, -7.8160e-04,  5.5013e-03, -9.5918e-03,
        -7.1910e-03, -2.4550e-03, -9.8046e-04, -2.2498e-03, -4.5009e-04,
        -6.2109e-04,  2.7405e-04,  2.4530e-04, -3.9265e-04, -2.0692e-04,
        -1.4328e-04,  5.3817e-03,  8.2735e-04,  2.4051e-03,  1.2313e-03,
        -1.5041e-03,  6.8268e-03, -2.5986e-04, -1.0204e-05, -1.4811e-03,
        -1.5595e-03,  4.7490e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7412e-03, -9.0087e-01,  7.8717e-03, -4.2687e-03, -8.5864e-04,
         1.0278e-02,  1.6433e-03,  5.0044e-03, -7.7402e-03, -2.6584e-03,
         1.3781e-05, -1.8720e-03,  1.5989e-03, -1.0881e-03,  1.3746e-03,
         3.0073e-03, -3.2772e-03,  2.3305e-03,  3.9048e-04, -4.1496e-03,
         1.1245e-03,  1.2727e-03, -4.3748e-03, -7.5181e-04,  8.8576e-04,
         7.6406e-04, -6.5839e-03,  3.3248e-04,  5.6813e-04, -1.4466e-03,
         8.9838e-04, -1.3406e-04,  9.4690e-04, -1.6082e-03,  2.3204e-03,
         5.8007e-04, -4.0704e-03, -4.5209e-04, -1.1650e-03,  3.8902e-03,
         1.2467e-03,  1.5480e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9042e-02,  8.7858e-01, -9.0779e-03, -1.0162e-02,  1.1170e-02,
         8.1724e-03,  9.6876e-03,  2.7794e-03,  1.6944e-03,  2.1601e-03,
        -2.4300e-03,  1.2352e-02,  3.0774e-03,  6.8001e-04,  4.4633e-03,
         8.5896e-03,  6.7335e-03, -1.4054e-03,  3.7057e-03, -4.4644e-04,
        -9.6056e-04, -5.3417e-04,  3.3937e-04,  1.2672e-03, -4.9533e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0845e-02,  9.1100e-01,  1.3581e-02,  3.0644e-04, -2.9896e-03,
         3.1998e-03,  1.2478e-02, -1.8175e-03, -2.3658e-03, -2.9444e-03,
        -3.1165e-03, -3.4551e-03, -4.2346e-04, -5.9274e-03, -5.5160e-03,
         5.0033e-03, -5.2819e-03, -4.4133e-04, -4.1426e-04, -1.7072e-03,
        -8.9816e-04, -5.3966e-03,  8.8800e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0589e-03, -8.3957e-01, -2.6819e-02, -1.1174e-02, -1.9834e-02,
         3.9382e-04,  1.6456e-04,  3.3416e-04, -1.7514e-03, -5.1422e-02,
        -9.0459e-03, -3.8381e-03,  6.2651e-03, -6.3203e-03, -1.6916e-03,
        -1.2847e-03, -1.4633e-04, -4.8279e-03,  2.5747e-03,  3.0193e-03,
         1.9738e-03, -3.8882e-04,  6.0986e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0766,  0.8082, -0.0134, -0.0110, -0.0077,  0.0164,  0.0097,  0.0040,
         0.0167,  0.0071, -0.0098,  0.0174, -0.0020,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0038, -0.9401,  0.0094, -0.0019, -0.0032,  0.0017,  0.0019,  0.0055,
        -0.0033, -0.0018,  0.0026, -0.0020,  0.0096,  0.0132,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.8840e-02,  8.7353e-01,  8.6373e-03, -5.7322e-03, -1.5912e-03,
        -1.5096e-03,  5.8322e-03,  2.2511e-03,  2.3866e-03,  3.4069e-03,
         8.3878e-03,  2.6985e-03,  2.1359e-03,  1.8631e-04, -1.7325e-03,
         2.4081e-03,  2.2372e-03,  1.3588e-02,  2.9042e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 3.9652e-02,  8.3391e-01, -8.1302e-04, -7.1709e-03,  2.0021e-03,
         2.4160e-04,  5.8861e-03,  4.6802e-03, -6.7002e-03, -8.7344e-03,
        -1.9772e-03, -7.0102e-03,  1.6711e-03,  1.1817e-02, -4.7099e-03,
        -9.8722e-03, -1.9681e-03, -1.1854e-04,  4.5291e-04,  7.9161e-03,
        -2.0757e-03,  8.8274e-04, -2.7093e-03,  3.0763e-03, -1.9646e-02,
        -1.4310e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3251e-03,  8.7898e-01,  1.7178e-03, -4.0905e-03, -1.3613e-02,
        -1.7110e-02, -6.1222e-03, -1.6651e-03,  1.6338e-03, -2.8723e-02,
         1.1470e-02,  5.3068e-04, -1.3909e-02, -2.2964e-03, -6.3284e-03,
         8.4855e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0083,  0.8296,  0.0014,  0.0564, -0.0013,  0.0102,  0.0100, -0.0046,
        -0.0104,  0.0167,  0.0105, -0.0014,  0.0052, -0.0022,  0.0093, -0.0079,
         0.0145,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0404, -0.8736,  0.0275,  0.0093,  0.0068,  0.0056,  0.0082, -0.0266,
        -0.0019,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4879e-03,  8.0071e-01,  2.6518e-02, -3.6212e-02, -8.5478e-03,
         5.3216e-03,  5.8539e-04, -9.0037e-06,  2.7213e-03, -3.5388e-03,
         4.2382e-03,  7.9237e-03, -1.4645e-03, -5.3012e-03,  6.1201e-03,
         2.3549e-03,  1.1971e-03,  3.4408e-03,  3.1855e-03,  3.9892e-03,
         1.2768e-04,  6.7423e-03, -1.0427e-04, -2.6876e-04,  5.7133e-03,
         2.3097e-03, -3.0036e-03,  3.6790e-03,  3.1526e-03,  8.6928e-04,
         3.2632e-03,  1.8749e-03,  5.2731e-03, -1.2544e-03,  5.2099e-04,
        -1.6430e-03, -5.9938e-06,  7.8201e-03, -5.5811e-05, -8.3971e-04,
        -1.6927e-03,  2.3213e-03, -1.6458e-03,  5.0281e-04,  3.1115e-03,
        -1.4332e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8309e-02,  8.3619e-01, -1.0927e-02, -1.8933e-02, -5.0969e-03,
        -4.7138e-03,  4.8318e-03, -7.1599e-04, -7.7650e-03,  5.1612e-03,
        -4.4014e-03, -2.7673e-03,  1.4395e-02, -6.2089e-03, -9.4088e-04,
        -6.9803e-04,  4.1207e-03, -2.0041e-04, -8.3510e-04,  4.0399e-03,
        -3.4101e-03, -4.8455e-03,  1.4327e-03, -4.0901e-03,  1.4507e-03,
        -1.1670e-03,  5.5646e-03,  1.0362e-03, -5.5933e-03,  1.0160e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0439, -0.9164,  0.0128,  0.0160,  0.0028,  0.0028,  0.0053,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2283e-03, -9.3463e-01,  4.4193e-03,  5.9485e-03, -1.9453e-03,
         9.3251e-04, -2.5452e-03, -6.4471e-03,  7.0617e-03, -4.0261e-03,
         2.8812e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7062e-02,  8.8125e-01,  2.5275e-02, -8.3853e-03,  1.0677e-03,
         9.8701e-03, -5.6147e-03, -3.4018e-03,  6.8669e-04,  4.7450e-03,
        -2.1959e-03, -3.1455e-03, -1.8221e-03, -1.5475e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5322e-02, -8.7003e-01,  5.1795e-03,  1.2263e-03,  1.3718e-02,
        -1.3289e-03, -2.6494e-03,  1.4897e-03,  8.6424e-04,  3.9584e-03,
        -2.0750e-03, -2.8330e-03,  9.2560e-03,  5.4569e-04, -3.4099e-04,
        -2.9039e-03, -2.4158e-03,  7.9415e-04, -5.1626e-03,  2.1319e-03,
        -1.5414e-02, -7.0520e-04,  2.5129e-03,  3.2465e-03,  1.7755e-03,
        -3.1721e-03, -1.5331e-03, -1.7283e-03, -1.6706e-03, -3.3179e-03,
        -1.4074e-03, -6.2632e-04,  3.9962e-03,  4.6673e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2192e-02,  8.2291e-01,  2.7898e-02, -1.8052e-02, -6.8902e-03,
         2.8377e-03, -3.3376e-03, -1.0260e-02, -5.6893e-04, -1.0577e-02,
        -2.8757e-03, -3.5307e-03, -2.3287e-03, -3.2547e-03, -1.9045e-02,
        -7.5014e-05,  2.3517e-03, -5.6594e-03, -1.2122e-03, -5.0564e-03,
         6.5613e-03,  1.6360e-03, -5.3057e-03,  3.4164e-03, -3.4300e-03,
         2.6445e-03, -1.1362e-03, -4.9619e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6250e-02,  8.9104e-01,  1.2654e-02, -1.1842e-02, -1.4815e-03,
        -5.5863e-04, -7.1663e-03,  2.0488e-04, -4.0985e-03, -1.1430e-03,
         5.6979e-04, -7.0194e-03, -1.3229e-03, -6.6934e-04, -3.2196e-03,
        -1.6587e-04, -4.3871e-03,  1.0817e-02,  3.5795e-03,  3.8723e-04,
        -7.4307e-04,  1.5721e-03, -2.3998e-03, -6.7081e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 1.2996e-02,  8.9466e-01,  1.1427e-03, -1.2560e-04,  1.7818e-03,
         5.6783e-03, -4.6555e-03, -2.7136e-03, -3.0667e-03, -1.5828e-03,
        -1.9547e-02,  1.5789e-03,  9.3071e-04, -5.4654e-05,  1.7797e-03,
         1.1157e-03, -2.8647e-03, -5.3154e-03, -3.7903e-03,  3.3662e-03,
         2.5977e-03, -3.3168e-03, -5.4016e-03,  4.6573e-06,  1.0321e-04,
         1.9566e-03,  1.7847e-04,  2.4700e-04,  3.5027e-03,  6.4853e-03,
        -7.4637e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2780e-02, -9.1277e-01, -5.0274e-03, -1.2483e-02, -2.8613e-04,
        -7.5095e-03, -8.6954e-03,  2.6254e-04,  1.9134e-03, -2.8235e-03,
         3.8418e-03,  1.1609e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0383,  0.8833,  0.0096, -0.0072,  0.0082,  0.0011, -0.0029, -0.0042,
         0.0039,  0.0060,  0.0018,  0.0026,  0.0170, -0.0049,  0.0090,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0750e-02, -8.9260e-01,  6.5728e-05, -1.2703e-03, -2.4836e-03,
        -3.4224e-03, -4.2678e-03,  5.0856e-03,  5.3258e-03, -8.2785e-04,
        -1.4825e-03, -1.1467e-03, -3.0164e-03, -1.4568e-03,  1.1595e-03,
         4.7047e-04,  1.2426e-03,  1.2254e-03,  6.9808e-03,  3.2183e-03,
        -1.2348e-03, -2.2449e-04,  8.6551e-04,  8.7227e-04, -1.5854e-03,
        -4.2789e-03,  1.1236e-03,  6.0562e-04,  1.4170e-03, -1.7089e-03,
        -2.0454e-03,  2.1603e-03,  1.4375e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0271,  0.8665,  0.0147, -0.0159,  0.0020,  0.0034, -0.0057,  0.0051,
        -0.0074, -0.0018, -0.0041, -0.0051, -0.0028,  0.0149, -0.0079, -0.0158,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0160,  0.9177,  0.0018, -0.0026,  0.0066, -0.0068, -0.0122, -0.0232,
         0.0130,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0295,  0.8544, -0.0550,  0.0109, -0.0502,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4309e-02,  8.5482e-01, -1.6830e-02,  6.5272e-03, -2.6088e-03,
        -8.3858e-03, -4.1221e-03,  1.3756e-02, -1.2148e-06, -5.1061e-03,
        -1.3617e-03, -4.0014e-03, -5.5622e-03, -4.6332e-04, -5.0105e-03,
        -3.2334e-03,  2.7800e-03, -2.3385e-03,  3.3626e-03,  6.0282e-03,
        -4.4594e-04, -1.0666e-03,  1.5112e-03,  1.8808e-03,  6.6633e-04,
        -8.8347e-03,  4.9876e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0470, -0.8588, -0.0214, -0.0161, -0.0030,  0.0413,  0.0125,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6465e-02,  8.5103e-01,  8.1156e-03, -4.9204e-03,  8.6353e-03,
         4.8510e-03, -6.9682e-03, -2.0759e-03, -8.4551e-04,  2.5052e-03,
         9.3847e-04, -8.1157e-03,  5.4721e-04, -2.0801e-03,  8.0487e-04,
         2.9253e-03,  7.3228e-03, -6.5533e-03,  6.9191e-03,  1.7675e-02,
        -3.0582e-03,  1.6647e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1280e-02, -9.2575e-01, -3.8696e-03, -9.9118e-03, -1.9409e-03,
        -2.4066e-03, -5.0780e-04, -2.1993e-03,  1.2293e-03,  2.8723e-03,
         7.8623e-04, -1.9537e-03,  4.0043e-03, -4.6945e-03,  5.4450e-03,
         8.6218e-04,  2.0852e-03,  5.0369e-03,  3.1678e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0710e-02, -8.9771e-01, -1.8023e-03, -7.9525e-04,  4.1379e-03,
         2.1098e-03, -7.3385e-04, -1.0550e-03, -2.6360e-03, -1.9146e-03,
         5.6404e-04, -2.0299e-03,  1.2214e-04, -2.8285e-03,  4.5870e-04,
        -8.4132e-04,  2.3539e-03, -3.0317e-03, -1.4822e-03, -3.3360e-05,
        -5.8496e-04, -8.8772e-03, -2.0022e-03, -4.3586e-03,  2.9774e-03,
        -1.3297e-04,  1.1169e-03,  1.1981e-03, -1.2930e-02, -1.6928e-02,
        -1.5440e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-2.6377e-02,  8.1809e-01,  5.0234e-02, -6.9030e-04, -5.8364e-03,
        -3.6778e-03,  3.2789e-03, -1.8847e-05, -8.2635e-03,  1.0181e-03,
         5.0934e-03,  1.8328e-03, -3.6585e-03,  2.6019e-03, -7.7714e-04,
         1.9398e-03, -7.0272e-03,  6.7696e-03, -2.2512e-03, -2.0154e-03,
        -2.9774e-03,  6.2101e-04,  3.3119e-03,  7.1527e-04, -3.0434e-03,
        -2.1908e-03, -3.2945e-03, -1.5128e-02, -8.0102e-04, -4.1124e-03,
         6.4136e-03,  5.9434e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1137e-02, -8.5819e-01,  9.5991e-03, -5.5592e-04, -8.7821e-04,
        -2.1168e-03, -9.6368e-03, -4.5000e-03, -7.9328e-03,  2.9482e-03,
         1.2834e-02,  2.1350e-03, -2.6479e-03,  2.6857e-03,  3.1140e-04,
         3.0479e-03, -2.6149e-03, -6.8456e-03,  7.3436e-03, -1.0292e-03,
        -5.4205e-03, -5.5909e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0094e-02,  8.4490e-01,  2.0309e-04, -8.5626e-03, -9.5715e-04,
        -2.8025e-03, -9.3622e-04, -1.2310e-02,  7.1518e-05, -8.1811e-04,
         3.4804e-03, -1.6460e-03, -3.5444e-03, -4.8326e-04, -1.1599e-02,
         1.0571e-03, -3.8182e-04,  1.1925e-02, -1.9692e-03, -1.7654e-03,
        -5.2328e-03, -4.6968e-03,  1.0017e-03, -3.7606e-03, -2.7811e-03,
         4.3153e-03, -3.6365e-03, -1.3855e-02, -4.4859e-04,  2.0762e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2017e-02,  8.8447e-01, -1.4044e-02,  3.9311e-03, -4.7045e-03,
        -5.1627e-03, -4.1956e-03, -5.3797e-03, -3.1191e-03, -5.8883e-03,
        -2.6455e-03,  6.3228e-03, -2.6363e-03,  7.1878e-04, -4.8854e-03,
        -7.1263e-04, -2.8980e-02, -1.8699e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8586e-02, -8.5726e-01, -2.4321e-02,  5.6284e-03,  1.2711e-02,
         1.4571e-02, -5.9063e-04, -2.1695e-03, -2.4973e-03, -2.3177e-03,
        -3.3925e-03, -1.2686e-03, -2.4822e-03, -5.9094e-03, -5.7908e-04,
        -1.4538e-03,  6.2184e-04,  3.3130e-03,  4.3471e-03, -3.8992e-05,
         7.3671e-04,  7.7655e-04, -4.3563e-05,  5.5453e-04, -2.0768e-03,
        -5.0193e-04, -1.3228e-03, -1.4902e-03, -1.9351e-03, -1.7974e-04,
        -5.4735e-04,  6.9023e-04, -4.8151e-04, -7.8360e-03, -2.1163e-03,
         7.2512e-03, -7.4026e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2248e-03, -8.8099e-01,  2.0592e-02,  2.2504e-03,  2.2290e-02,
         3.0993e-03,  3.6398e-03,  8.7308e-03,  2.8383e-03,  3.2902e-03,
        -4.2385e-03,  1.3883e-04,  2.2596e-03, -3.6578e-03, -7.7762e-03,
        -4.4955e-04,  7.3398e-03, -5.7459e-03,  1.3196e-03, -2.7283e-04,
         7.0043e-03,  3.2640e-03, -6.9236e-04,  6.8918e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6933e-03,  8.7547e-01, -1.0732e-03,  2.9881e-03,  1.0844e-02,
         6.3116e-03,  3.3996e-03,  1.6434e-03,  2.6899e-03,  3.8026e-03,
         1.8007e-03, -6.8496e-03, -3.8322e-03, -9.0159e-04, -3.3694e-04,
         1.8857e-03,  1.4658e-03, -1.0868e-03, -2.9231e-03, -1.5704e-03,
         9.3586e-04, -4.3940e-03, -5.2303e-04,  2.0621e-03, -2.6948e-03,
         1.7701e-03,  1.4476e-03, -1.0469e-03,  2.2798e-03,  3.1566e-03,
        -6.3612e-03,  1.3925e-03, -8.6413e-03,  2.4420e-03,  6.1190e-03,
         4.2526e-03, -7.3625e-04, -7.1692e-04,  9.0398e-04,  3.2491e-03,
        -1.9068e-03,  6.2321e-04, -1.8307e-03,  4.0591e-03,  1.4363e-03,
         2.4536e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1011e-02,  8.4741e-01, -3.3180e-02, -4.7866e-03, -9.5779e-03,
        -5.9879e-03, -7.7360e-03, -3.3265e-03,  1.1586e-03,  3.0723e-03,
         1.5533e-04,  1.0538e-04,  4.7462e-03, -1.5641e-03,  3.4776e-03,
        -2.5690e-03,  5.9194e-03,  5.7707e-03, -1.9628e-03,  5.3758e-04,
         1.2545e-02,  2.2071e-03,  1.1193e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0680e-02,  9.1071e-01, -1.5174e-02,  4.0792e-03, -5.7429e-03,
         8.8135e-03,  2.2944e-04, -1.6919e-03,  6.0927e-03,  2.4658e-02,
         2.1329e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3333e-02,  9.0550e-01,  1.6233e-02, -7.9879e-03, -6.6129e-03,
        -1.0070e-02, -6.3643e-04, -4.7593e-04,  5.7856e-03, -1.6724e-03,
         7.3848e-03,  3.0394e-03, -7.6194e-04,  5.0336e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0746e-02,  8.7634e-01, -1.5260e-02, -1.8822e-02, -4.1956e-03,
         6.4887e-04, -5.7771e-05, -2.8918e-03,  4.2439e-04,  3.3886e-03,
        -1.1467e-02, -5.7587e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1022e-02,  8.3364e-01,  8.7699e-03,  1.8141e-02,  7.9324e-04,
         1.0185e-03, -1.0437e-02, -5.0218e-05, -1.4855e-03,  3.3993e-04,
        -1.9252e-03, -1.7942e-04, -1.1341e-03,  4.5342e-04,  2.3754e-03,
         5.9658e-05,  1.2614e-03,  5.0007e-04,  3.0075e-03, -1.0080e-03,
        -1.2519e-03,  9.1749e-04,  5.9565e-03, -2.5393e-03,  7.3888e-04,
        -1.5486e-02,  1.1835e-02, -3.3339e-04, -2.8584e-03,  5.4049e-03,
        -6.8418e-04,  5.2794e-03, -3.4705e-04,  7.1228e-04,  3.4772e-03,
        -4.0833e-03,  2.3238e-03,  3.5968e-04, -1.6672e-03, -1.1510e-04,
         1.9103e-03,  1.7790e-03, -1.2945e-03,  1.4400e-04,  2.1154e-03,
        -2.5923e-04,  9.8928e-04, -4.8788e-04, -7.6418e-04, -2.1333e-03,
        -2.1431e-03, -4.0155e-04, -1.5062e-05, -1.4639e-03,  1.2607e-04],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-3.0157e-02, -8.9231e-01,  2.8254e-03, -4.3992e-03,  1.5064e-03,
        -4.0330e-03, -8.7401e-04,  1.2009e-03,  6.1098e-03, -2.3977e-03,
         7.8487e-03, -1.5718e-03,  1.9489e-03,  5.0386e-03, -1.5995e-04,
        -2.7899e-03, -3.8912e-03, -1.1447e-02, -1.6044e-03, -6.4089e-03,
         2.8462e-03, -2.2084e-03, -6.4190e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.5240e-02, -8.8067e-01, -1.3474e-02, -3.6769e-03, -7.7827e-04,
        -8.6188e-04, -1.2046e-02,  3.2147e-03, -1.0004e-02,  1.8824e-03,
         3.3397e-04,  3.4387e-03,  3.5073e-03, -3.3850e-03, -5.4972e-04,
         3.4157e-03,  9.7538e-03, -3.7633e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.5476e-02,  8.3006e-01,  1.1157e-02, -6.7943e-03,  2.2346e-02,
         1.3319e-02,  8.5703e-03, -1.5199e-03, -2.6002e-03, -5.4108e-03,
         1.1096e-02, -1.7610e-06,  1.9795e-03, -7.6540e-03, -9.3006e-03,
         9.3152e-03, -5.3635e-04, -1.2867e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.8770e-03,  8.8588e-01,  1.0471e-02, -5.7686e-03,  9.6612e-03,
        -4.4550e-03,  1.3883e-02,  7.9491e-03, -5.6731e-04,  4.8935e-03,
         4.9888e-03,  1.0969e-02,  5.5763e-03, -6.2860e-03, -2.3774e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.6332e-05,  8.7904e-01,  1.3488e-03,  1.5199e-02, -3.4083e-03,
        -2.8616e-03, -1.8572e-03,  5.3112e-03, -3.8500e-03,  5.8801e-03,
         9.1856e-03, -3.2169e-04,  2.4215e-03, -3.6529e-05, -1.3232e-03,
         6.9029e-03,  1.0742e-03,  1.0695e-03,  7.3071e-04, -6.3854e-03,
         3.3152e-03, -4.8249e-03,  1.6148e-03,  8.2519e-04,  1.3810e-02,
         4.2186e-03,  1.1451e-03,  9.2103e-04, -1.2212e-03,  1.1931e-02,
         9.9872e-04, -6.8850e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5932e-02, -8.6120e-01,  7.6719e-03, -1.3060e-03,  3.1218e-04,
        -1.4970e-03,  9.7492e-04, -1.1229e-03,  7.2941e-03,  6.6509e-04,
         1.1883e-03,  6.1056e-03,  3.7573e-03,  1.8657e-03,  2.7728e-03,
         6.4206e-03,  1.6037e-03,  3.6609e-03,  5.1934e-03,  1.1345e-03,
         7.0105e-03,  1.7382e-02,  5.3394e-03,  7.4025e-03,  2.7583e-03,
        -7.3300e-05, -9.0933e-04, -9.1504e-03,  2.7522e-03, -5.8245e-04,
        -1.0808e-03,  3.3656e-03, -5.1003e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0535e-03, -8.7944e-01, -2.2300e-02,  3.5010e-03, -4.3949e-03,
         1.6816e-03,  9.2194e-04,  2.6895e-03,  3.3505e-03, -2.1643e-03,
         1.7726e-04,  1.6995e-04,  3.5994e-03,  1.7082e-03, -4.0400e-03,
        -1.4429e-03, -2.3123e-03,  1.3290e-03, -7.8589e-04,  2.1169e-04,
         7.0518e-04,  8.5439e-04, -2.0973e-04,  1.6122e-03, -2.6428e-03,
        -2.2939e-03, -1.8943e-03,  2.6621e-04, -1.6194e-03,  6.0738e-03,
        -2.9823e-03,  1.2713e-03, -3.1614e-03,  1.4415e-03, -2.5494e-03,
        -4.0648e-03,  1.2726e-03, -1.9508e-03, -1.9155e-03, -7.2863e-04,
         2.8361e-04,  8.4331e-05, -1.2206e-03, -1.0679e-04, -6.7004e-04,
        -1.8476e-03, -2.2589e-03, -8.4006e-04, -5.1112e-04, -2.7604e-04,
        -1.9574e-03,  1.1561e-04, -4.2540e-04, -1.4970e-03, -1.4010e-03,
        -1.2426e-03,  7.4066e-04, -6.7386e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6538e-02,  8.6027e-01, -5.6295e-03,  2.5106e-03, -6.6782e-03,
        -6.5148e-04,  8.9337e-04, -3.0236e-03,  6.8423e-04,  1.6877e-03,
         2.8251e-03,  5.0390e-03, -8.2728e-03, -8.3798e-03,  4.3063e-04,
        -1.9454e-03, -1.0731e-03, -1.8180e-03, -3.9738e-03,  1.8717e-03,
        -4.5426e-03, -3.7329e-04,  8.0455e-04,  5.9432e-04, -2.1878e-03,
        -2.7304e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0880e-03,  8.1972e-01, -1.6046e-02, -1.3037e-02, -1.4878e-02,
         3.0421e-02, -4.8600e-03,  2.4313e-03, -4.6129e-03,  1.3465e-02,
         8.3882e-03, -4.8837e-05,  4.7756e-03,  3.2792e-03, -1.2831e-04,
        -2.2711e-03,  4.6315e-03, -1.7030e-03,  1.3808e-02, -2.8643e-03,
         4.0992e-03, -5.9491e-03, -6.8299e-03, -9.3823e-03, -1.1281e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4654e-03, -8.5967e-01, -2.3999e-02,  5.8603e-03,  1.0397e-02,
        -1.1137e-03, -7.1894e-04,  1.2680e-03, -4.2771e-03, -2.5972e-03,
         1.3180e-03, -1.0794e-03,  4.0503e-03, -9.2956e-03,  3.6132e-03,
         8.6214e-04, -4.2959e-04, -3.6345e-03,  5.4394e-04,  4.5250e-03,
         1.6329e-03,  1.2254e-03,  8.1722e-04,  2.3737e-03, -7.2530e-04,
         8.5578e-04, -3.4058e-03, -1.8651e-03,  2.1729e-03,  1.1017e-03,
        -3.1031e-03,  3.9791e-03,  3.5656e-03, -5.6144e-04, -1.6580e-04,
         9.7156e-04, -3.2149e-03,  6.1465e-04,  1.4194e-03, -3.7408e-03,
        -6.3360e-04,  3.9542e-03, -2.2202e-03, -4.8580e-03, -9.1024e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9791e-02,  8.3669e-01, -6.4176e-03,  1.3664e-02,  3.5354e-03,
        -1.0292e-02, -7.4490e-03, -1.4980e-03, -4.1107e-03, -1.0325e-02,
        -1.1964e-02, -3.7696e-03,  1.3083e-03, -2.6883e-03,  2.8197e-03,
         2.3772e-02, -6.5081e-04, -2.8010e-03,  2.5946e-03,  2.0561e-03,
        -4.2584e-03,  4.2108e-03, -3.3348e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0134,  0.8526,  0.0264, -0.0199, -0.0037,  0.0156, -0.0095,  0.0201,
         0.0056, -0.0102, -0.0195, -0.0037,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-4.7770e-03, -8.5279e-01,  2.0368e-02,  2.2030e-02, -2.3593e-04,
         2.8085e-03,  2.1602e-02,  6.7725e-03, -2.1898e-03,  8.1485e-03,
         4.7351e-03,  6.8336e-03,  4.3789e-03, -2.8719e-03, -1.8081e-03,
        -3.7699e-03, -2.7717e-03, -1.8153e-03,  3.9544e-04,  6.2309e-03,
         4.6033e-03,  1.5731e-02, -2.3362e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5373e-02,  8.5519e-01,  2.6479e-03, -4.1195e-03, -2.5190e-03,
        -8.1209e-03, -2.3144e-04,  4.5215e-03,  3.9680e-03,  8.4539e-03,
         2.4968e-03, -3.0255e-03, -3.5274e-03,  7.9512e-04,  2.6445e-03,
        -6.2599e-03, -2.2458e-03,  5.8238e-04, -2.7523e-03,  4.5385e-03,
        -1.8037e-04, -1.2413e-03,  8.6940e-04, -3.3256e-03, -9.2770e-04,
        -2.3228e-04, -6.0190e-03, -7.6335e-04, -5.8408e-03, -9.1083e-05,
         4.0593e-04,  9.6873e-03,  1.6407e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0711e-02,  8.9590e-01,  2.6636e-02,  3.7555e-03,  2.4785e-03,
         4.4625e-03, -3.3746e-03,  8.6377e-03,  4.2813e-05, -5.7059e-04,
        -5.3132e-03, -6.5568e-05,  4.6531e-03, -2.7461e-03,  8.3466e-04,
        -6.4941e-04, -3.7816e-03, -6.8331e-03, -8.5573e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3342e-02,  9.0515e-01, -1.2625e-02,  5.4471e-03, -2.6723e-03,
         1.1607e-02,  4.8183e-03, -1.4978e-03, -2.1350e-03, -6.4964e-04,
        -4.5625e-03,  2.5790e-03,  4.9716e-03,  7.9386e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0960e-02,  8.6997e-01,  4.8205e-02, -3.6548e-03, -1.1629e-02,
         8.8614e-03,  4.5098e-03,  3.8218e-03, -1.2771e-03,  3.5412e-04,
         6.0483e-03, -2.8176e-03,  3.5643e-03, -6.5990e-04, -5.0187e-03,
        -1.1357e-04, -6.0703e-03,  1.6984e-03,  1.3094e-03, -2.5683e-03,
        -6.4471e-03,  4.4064e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3503e-03,  9.0338e-01,  7.9922e-03,  1.2336e-03, -2.1860e-03,
        -3.2836e-04, -2.7135e-03, -2.9670e-02, -1.3803e-03,  3.3588e-03,
        -1.5735e-03, -3.8235e-03, -9.5140e-03,  1.6456e-03,  1.9750e-03,
         2.2320e-04,  1.3495e-03,  7.8669e-04, -2.0801e-03,  3.6706e-03,
        -1.1775e-03,  2.2628e-03,  1.5891e-03, -2.1145e-03, -6.4829e-04,
         1.3730e-03,  3.6052e-03,  2.1881e-03, -2.0913e-03,  2.7183e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0088,  0.8806, -0.0147, -0.0103, -0.0048,  0.0103,  0.0124, -0.0126,
        -0.0030,  0.0054,  0.0009, -0.0049,  0.0025,  0.0047, -0.0034, -0.0034,
        -0.0171,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0021, -0.8513, -0.0325,  0.0123,  0.0097, -0.0091, -0.0020, -0.0302,
         0.0114, -0.0099,  0.0295,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3217e-04, -8.4704e-01, -1.2423e-02,  1.0099e-02,  1.6614e-02,
        -3.1671e-02,  2.0947e-02,  6.5946e-03, -8.5352e-04,  3.0797e-03,
        -6.7373e-03,  2.0473e-03,  5.9384e-03, -1.5425e-02, -4.9482e-03,
         4.4996e-03, -1.0545e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1688e-02,  8.5183e-01, -5.8006e-03,  1.1225e-02, -2.8506e-03,
        -2.3450e-04,  2.1097e-03, -4.5354e-03,  4.4167e-03, -6.6059e-03,
        -1.7971e-02,  3.8962e-03,  4.0804e-03,  9.3533e-03,  1.3427e-02,
         6.0227e-03,  1.0503e-04,  1.2432e-03, -1.2607e-02, -8.1209e-03,
        -3.3244e-03, -5.2559e-04,  9.7199e-03,  1.1517e-03, -4.1814e-03,
         2.9750e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4311e-02,  8.6840e-01,  7.6080e-03, -1.2916e-03,  5.2324e-04,
         2.0603e-02,  2.8172e-03, -3.5624e-04,  1.1548e-02, -4.6959e-04,
        -8.0211e-03, -5.5884e-03, -2.7968e-03, -1.6076e-03,  1.0899e-03,
        -2.2339e-03,  6.3498e-04, -1.0439e-03, -9.2754e-04, -3.7149e-03,
        -8.2539e-04, -7.0765e-04, -3.5665e-03, -2.0295e-04,  2.2195e-03,
        -7.2709e-04, -6.6347e-04, -1.1605e-05, -3.0168e-03,  2.4753e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0443e-02,  8.9920e-01,  2.7181e-03, -9.6133e-04, -7.7898e-03,
        -1.1789e-02, -4.8330e-03, -5.4417e-03, -1.2735e-04, -8.6621e-03,
        -4.4644e-03, -1.4397e-02, -5.2828e-03,  2.1599e-03,  1.7195e-03,
        -1.4635e-04, -2.5087e-03, -8.4440e-04, -6.5068e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-6.6693e-02,  8.1961e-01, -2.4752e-02,  3.7316e-04,  6.4888e-03,
        -8.0767e-04, -2.1507e-03, -4.9131e-03,  2.8165e-03,  1.4875e-03,
         1.2327e-04, -1.4017e-03,  1.8009e-03,  6.2056e-03, -1.0884e-02,
         2.2029e-03,  5.5211e-03, -4.1096e-03,  3.6006e-03,  3.6916e-03,
        -7.4153e-04, -6.8895e-04, -3.6916e-03, -5.0919e-03, -5.4306e-03,
         1.1654e-02, -3.0632e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8341e-02,  9.1219e-01,  4.3752e-03, -8.2418e-03, -6.5419e-03,
         5.8249e-03,  1.0117e-02,  2.5474e-03,  5.6888e-03,  6.2728e-05,
        -4.5207e-05, -1.0885e-03, -3.7302e-04,  4.5635e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.7071e-03,  9.1175e-01,  1.1293e-02,  1.3640e-03,  1.3000e-03,
         1.1611e-02,  1.1107e-03, -8.2318e-05,  2.2718e-03, -1.6280e-03,
         2.6066e-03, -1.4040e-04, -2.0718e-03,  2.9745e-03,  1.1689e-03,
        -3.6218e-04, -2.5511e-03,  1.5233e-03,  1.7482e-03,  4.4095e-03,
        -1.4706e-03, -3.4220e-03, -3.3453e-03,  1.2725e-03, -3.1543e-03,
        -1.7656e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3576e-02,  8.8385e-01,  1.4852e-02, -1.2410e-02, -3.2987e-03,
        -2.3145e-02,  3.4324e-03, -7.1875e-04, -3.0064e-03, -6.8595e-04,
        -3.3582e-03,  2.5170e-03, -7.2671e-03, -2.7883e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2406e-02,  8.9320e-01,  3.5169e-03,  4.2468e-03, -7.8869e-03,
         7.1343e-03, -1.7261e-03, -3.2517e-03,  1.1509e-03,  3.7093e-03,
         1.4054e-03, -2.4558e-03, -6.3593e-05, -2.0377e-03,  3.3381e-03,
        -6.4890e-05,  1.1557e-03, -4.7264e-03, -2.5146e-03, -1.8389e-03,
        -1.4379e-03,  1.3651e-03, -8.0519e-04,  8.6037e-04, -9.0108e-04,
        -6.6844e-04,  1.6386e-03, -3.1310e-04,  3.2906e-04,  5.0758e-03,
         9.2831e-03,  1.9492e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6005e-02,  8.9250e-01,  4.5148e-03, -1.2194e-02, -1.2062e-02,
         3.9108e-03, -2.1957e-03, -1.0754e-02,  2.8837e-03, -8.8904e-04,
         4.6050e-04, -1.1366e-03,  2.1025e-03,  3.2010e-03,  1.4024e-03,
         1.9620e-03, -1.1923e-03, -4.2593e-03,  8.9193e-04, -2.2706e-03,
        -1.5088e-03, -1.9599e-03, -7.5326e-04, -3.9726e-03, -1.5015e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0068,  0.9270,  0.0375,  0.0230,  0.0015, -0.0043,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8049e-02,  8.6078e-01,  1.7485e-02, -4.2205e-03,  5.5829e-03,
        -7.2584e-03, -2.0930e-03, -4.3915e-03, -1.0922e-03, -1.4090e-03,
        -2.3225e-03, -2.1818e-03, -2.0001e-03,  1.0729e-02, -4.6674e-03,
         5.4551e-04,  4.5326e-03,  1.1877e-03,  9.3128e-04,  3.2418e-04,
        -5.9860e-03, -2.7287e-03, -5.9568e-04,  7.4507e-04, -2.1939e-03,
         1.0638e-03, -1.0643e-03,  1.5993e-03,  1.7569e-03, -1.2053e-03,
         3.1563e-03, -8.5372e-04,  2.8675e-03, -1.7453e-03,  2.0156e-03,
        -5.4857e-03, -3.1581e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5392e-02, -8.6275e-01,  1.1569e-02, -5.1038e-03,  4.2789e-03,
         2.9492e-04,  2.4262e-03,  1.6701e-03,  1.5961e-03,  1.8356e-03,
         9.3490e-04, -8.4946e-04,  6.9618e-03, -2.8178e-04,  5.0641e-04,
         6.7602e-03,  4.5824e-03, -3.9886e-03, -2.6254e-03, -1.3934e-03,
         2.8219e-02, -3.0078e-03,  6.0212e-03,  3.1581e-04,  4.8944e-04,
         4.0554e-03,  4.4750e-04,  2.3638e-03,  1.5834e-03,  2.3093e-03,
         5.1663e-05, -8.2224e-04, -3.2158e-06, -9.2828e-04,  3.5782e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9827e-02,  8.2299e-01, -2.6793e-02,  3.0494e-02, -1.1982e-03,
         9.3933e-03, -1.4326e-02, -2.3077e-03, -4.3671e-03,  9.2814e-03,
         2.9557e-03, -2.6631e-02, -7.6599e-04, -5.5937e-03,  3.0759e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.2420e-03, -9.3041e-01,  8.2294e-04, -8.2146e-03, -5.7426e-04,
         5.9160e-04, -3.6727e-03,  2.7802e-03,  1.0396e-03,  1.6626e-03,
         1.9409e-03,  1.8093e-03, -2.5972e-05,  4.6244e-04,  2.5008e-04,
        -8.2353e-04, -1.3677e-03, -3.6512e-03, -1.4553e-03, -5.6842e-04,
         1.6208e-03,  2.0247e-03,  1.2952e-03, -4.8386e-04, -3.0181e-03,
        -7.8685e-04, -1.6395e-03, -2.0018e-04, -2.9013e-03, -1.9576e-03,
        -3.0517e-03,  2.7794e-03, -6.8713e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3937e-02, -8.5901e-01,  3.7083e-03,  1.3731e-04,  2.7493e-03,
         2.4128e-02,  4.7998e-04, -7.3233e-03,  4.3577e-03, -1.3891e-03,
        -8.5646e-03, -8.4785e-03, -2.3596e-03,  7.1969e-03,  1.8802e-03,
        -1.1029e-04,  4.9722e-04,  1.5752e-03,  2.2819e-03, -4.7669e-03,
         4.9470e-03,  1.7369e-03, -3.3762e-03, -1.5991e-03, -2.5935e-03,
         5.8376e-04, -1.9215e-03,  1.7695e-03,  6.5411e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([-2.2320e-02, -8.8224e-01,  1.6299e-04,  7.6746e-03,  3.2997e-03,
        -2.2276e-03,  3.7973e-03, -7.3799e-03, -5.3957e-03,  2.9326e-02,
         2.1376e-03, -1.2538e-03, -6.2957e-04,  2.3185e-03,  2.9835e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.9868e-03,  8.7137e-01, -1.2809e-02,  8.8707e-03, -1.5099e-04,
         1.8426e-03,  1.9823e-03, -2.9874e-04,  5.5451e-03,  2.5742e-03,
        -3.8002e-03,  1.7915e-04, -8.4196e-04,  1.4643e-03,  1.5586e-03,
         1.8374e-03, -1.6682e-03,  5.1147e-03,  1.0770e-03,  3.5771e-03,
        -2.4676e-03, -2.0501e-03, -3.2754e-03,  1.0778e-03, -3.8108e-04,
         1.6837e-03,  1.2105e-03,  2.2911e-03,  9.6853e-05,  1.7115e-03,
         3.3644e-03, -2.7700e-03, -8.1218e-04,  1.5426e-03, -5.0735e-03,
         3.4501e-03, -4.7978e-04, -3.4116e-04,  6.9368e-04, -1.7355e-03,
        -1.3810e-03,  3.2869e-03,  1.8974e-03, -1.1148e-03, -2.5585e-03,
        -4.9261e-03, -1.2702e-03, -5.6218e-03, -5.8830e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2878e-02,  8.0095e-01, -3.6597e-02,  3.3636e-03,  1.5088e-02,
        -5.4694e-04,  2.1175e-03,  4.3205e-03,  4.2368e-03,  1.7105e-03,
         6.2043e-05,  7.8998e-04,  2.7302e-03,  1.6176e-03, -4.5279e-05,
        -7.6073e-04,  3.9733e-03, -4.7212e-04,  3.7928e-03, -4.8604e-03,
         1.4488e-03,  2.6396e-03, -1.5675e-03,  6.5485e-04, -5.5775e-04,
         3.6535e-03, -4.8644e-03,  4.8039e-03,  3.4707e-03, -3.1746e-03,
         2.3302e-03, -3.9801e-03, -5.2148e-03,  2.0640e-03, -2.8660e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6984e-02, -8.8468e-01,  2.4778e-03,  4.4181e-03, -1.5380e-03,
         5.5713e-04,  6.7378e-03, -4.3575e-03, -9.1218e-03,  3.0054e-03,
         2.2470e-03, -1.7495e-03,  1.1876e-03, -1.6162e-03, -5.3051e-03,
         3.1977e-03,  9.8626e-03,  5.6913e-03, -2.5263e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8572e-03, -9.0821e-01,  7.8936e-03,  5.9478e-04,  2.5241e-03,
        -1.1425e-02, -5.1767e-03,  8.7517e-04,  1.6698e-03,  4.7808e-03,
         8.6923e-03,  2.2026e-03, -2.3521e-02,  1.4954e-02,  4.6190e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8767e-02,  8.9362e-01, -4.3080e-04,  6.2720e-03, -1.0346e-02,
        -3.6159e-03,  9.6084e-04, -3.7791e-03, -7.2848e-03, -2.8228e-03,
        -2.7018e-04, -4.8890e-03, -8.9881e-03, -2.4511e-03, -2.4839e-03,
         5.1132e-04,  1.1189e-03, -1.6487e-03, -5.4031e-04,  5.5616e-03,
        -1.0606e-02,  3.0359e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9991e-02,  8.4923e-01, -1.8970e-02, -9.1962e-03,  6.5696e-04,
         1.5232e-03, -5.8595e-03,  3.2631e-03, -1.2678e-03,  1.2310e-03,
        -4.1719e-03,  1.9952e-03,  5.2568e-03,  4.1071e-03, -2.0429e-03,
        -1.2804e-04, -2.3082e-03, -4.2668e-05, -2.9510e-03,  6.2566e-04,
        -1.1580e-03,  4.0192e-03, -9.2337e-04,  5.3816e-03, -3.6769e-03,
         6.5105e-04, -6.7849e-03, -3.3686e-03, -2.2812e-03,  1.6941e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8944e-02,  9.0881e-01, -1.2386e-02, -1.7151e-03,  1.3473e-02,
        -8.7607e-03,  9.2470e-04, -6.5944e-03, -7.4366e-04,  9.1422e-03,
        -8.5011e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3348e-02, -8.9062e-01, -1.0308e-02, -1.4078e-02, -1.0475e-02,
        -6.4238e-03, -3.4237e-03, -3.4632e-04,  2.1822e-04, -1.8522e-03,
         1.2568e-03, -7.0270e-03, -6.1738e-04, -4.9294e-03, -2.9123e-03,
        -7.1688e-03,  2.0821e-03,  2.0969e-03, -1.5424e-03,  1.9371e-03,
        -9.2912e-03,  8.0472e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2873e-03, -8.5752e-01,  7.7050e-03,  7.3114e-03, -3.5466e-03,
         4.3295e-03, -6.1298e-03, -1.4507e-03, -9.5059e-04,  1.2483e-03,
         6.7879e-03, -5.4794e-03,  1.5561e-05,  5.1260e-03,  3.7578e-03,
         8.0163e-03,  5.9250e-04,  7.4386e-03,  9.9200e-03,  4.4558e-03,
        -1.2018e-03, -8.8136e-04,  2.7809e-03,  5.5665e-04,  5.8748e-03,
         2.8516e-03,  9.4340e-04,  1.5794e-03, -2.9910e-03,  5.5938e-03,
         3.0458e-03, -1.1325e-03, -2.5499e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6696e-02,  8.0053e-01,  3.6348e-02, -9.0177e-03,  4.9650e-03,
         5.7932e-03,  9.6135e-04, -8.0677e-03,  1.3944e-03, -6.3830e-04,
        -5.8445e-04,  1.5789e-03,  2.3876e-03, -4.7374e-03,  1.4893e-04,
         1.0613e-02,  3.5968e-03, -3.4025e-04, -4.0947e-03,  3.8266e-03,
         4.9671e-03, -9.5096e-03,  1.3997e-04, -1.0458e-03, -9.2027e-03,
         3.1374e-03,  2.1297e-02, -1.4381e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5602e-02, -8.8388e-01, -2.8371e-02,  1.7782e-03,  6.2320e-05,
        -6.1372e-03, -1.4985e-03,  1.1369e-03,  1.0062e-02, -2.9721e-03,
         7.7033e-03,  7.5175e-04,  7.4238e-04, -2.4086e-03,  1.8932e-03,
         6.9347e-03,  1.4079e-03,  2.4078e-03,  3.3487e-03, -1.0579e-03,
         3.9568e-03,  1.7364e-03,  1.7508e-04,  8.8446e-03, -5.1317e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 1.7074e-02,  8.4566e-01, -3.0978e-03, -1.7807e-03, -1.0617e-02,
         8.9130e-03,  3.9758e-03, -5.9168e-03,  1.8149e-04, -1.9658e-03,
         4.3577e-05, -1.0073e-03, -1.1180e-03,  2.4512e-03, -1.7515e-03,
        -9.8362e-03, -3.1331e-03,  1.3718e-02,  2.1251e-04,  2.1409e-03,
         2.7032e-03,  1.5986e-04, -2.5446e-03,  2.1471e-04, -1.0735e-02,
        -4.0539e-03,  3.7933e-03, -1.0870e-04, -1.6965e-03, -2.2537e-03,
        -2.5112e-03, -3.3107e-03,  4.7820e-04, -2.5919e-03, -3.8089e-05,
         3.6599e-03,  1.7995e-03,  1.3875e-03,  4.9902e-03, -7.0714e-03,
         9.2994e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2018e-02, -8.7510e-01,  1.0416e-03,  2.0060e-02, -7.9074e-03,
         4.4748e-03,  3.8842e-04, -4.2553e-03,  2.3663e-03,  3.9082e-03,
         3.2525e-03,  3.9745e-03, -3.3448e-03, -1.8813e-03, -4.4373e-03,
         4.5843e-03, -1.7070e-03,  1.8657e-03,  7.3500e-04,  9.5056e-03,
         5.5110e-03,  5.4363e-03,  1.6419e-03,  3.0222e-03, -7.5820e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.5020e-02,  8.3684e-01,  1.5170e-02, -1.8256e-02, -2.7904e-03,
         2.8836e-03,  5.4236e-03,  4.7830e-03, -5.2083e-03, -1.2954e-03,
         6.8988e-03, -2.7441e-04, -5.4794e-03,  6.8846e-04, -5.1432e-03,
        -4.7859e-03, -7.2337e-03, -1.1977e-05, -7.3877e-03, -3.4237e-03,
        -1.0050e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6608e-02,  8.4786e-01,  1.5418e-02, -1.2675e-02, -6.6345e-04,
         6.8168e-03, -4.4897e-03, -8.8776e-03, -4.2148e-03, -6.6475e-04,
        -5.2346e-03, -2.9491e-02,  3.8260e-03, -2.1927e-03, -3.3835e-03,
        -4.8628e-03,  1.4890e-02, -7.8288e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4748e-02,  8.4597e-01, -1.7575e-02, -3.6412e-03, -2.1360e-03,
         6.7445e-03, -4.3345e-03, -9.8384e-03, -5.1587e-03,  1.0179e-03,
        -2.4333e-03, -1.0458e-03,  1.3726e-03, -2.8461e-04,  1.2923e-03,
         4.4996e-03, -3.1423e-03, -3.9133e-03,  2.4850e-03, -3.4873e-03,
        -1.6165e-03, -9.0261e-04,  1.4627e-03, -8.4669e-04,  1.9566e-03,
        -1.4705e-03,  7.6513e-04,  5.9360e-04, -2.1688e-03, -1.3280e-04,
         3.1112e-03,  4.9945e-04,  4.8660e-04,  7.7419e-03,  8.5613e-04,
         2.4335e-03,  2.2023e-02,  5.8089e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0669e-02, -8.7793e-01,  2.3000e-02, -4.0265e-03, -2.5867e-04,
        -2.8504e-04,  3.6745e-03, -1.1649e-04,  2.2589e-03, -2.6577e-03,
         8.3166e-04,  5.3214e-03,  5.5097e-04, -2.2196e-03, -2.3911e-03,
        -1.5856e-03, -5.2025e-03,  1.8852e-02, -2.3565e-02, -4.6084e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8746e-02,  8.7599e-01, -9.5788e-03,  8.0537e-03,  1.3433e-02,
         1.4428e-02,  8.4022e-05,  1.9914e-02,  3.4955e-03, -3.8721e-03,
        -7.2520e-03, -1.4079e-02,  2.7681e-03,  8.3098e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.3541e-04,  8.7342e-01,  4.3658e-03, -1.4123e-03,  5.7132e-03,
         5.6504e-03,  6.3866e-03, -3.8435e-03,  3.6943e-05,  3.7927e-03,
        -9.3964e-03, -7.4556e-05,  3.6977e-04,  3.3948e-03,  2.7212e-03,
        -3.6276e-03, -1.6731e-03,  5.4513e-03,  3.4702e-03,  1.5266e-03,
        -1.9697e-03,  1.8402e-02,  1.4514e-03,  1.3690e-02, -5.1258e-04,
        -2.8217e-03,  2.5512e-03,  1.5916e-03, -1.5945e-03,  1.1369e-02,
         2.7713e-03,  4.0161e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3977e-03, -9.0738e-01, -9.6049e-03, -1.6648e-02, -8.7888e-03,
        -2.3941e-05, -3.4044e-03,  1.6557e-04, -5.5038e-04,  1.3995e-03,
        -8.2245e-04,  9.8273e-03,  4.0784e-03, -1.0161e-03, -1.7858e-03,
         1.8941e-03, -3.0891e-04,  1.0433e-02, -7.2087e-05,  2.0393e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1007e-02,  8.4309e-01, -1.5756e-02, -5.6306e-03, -5.6979e-03,
         7.2822e-03, -2.6311e-03, -2.5700e-03,  1.8215e-03, -2.3177e-03,
        -3.6770e-04, -5.2203e-03, -1.7891e-03,  1.1244e-03, -7.8709e-04,
        -1.5826e-04, -1.9119e-03, -4.4298e-03, -7.6487e-03, -1.9997e-03,
         3.0964e-03,  2.0931e-03,  9.5478e-04,  1.1097e-02, -6.1682e-03,
         1.4738e-03, -8.4458e-03, -1.5537e-03,  5.1761e-03, -1.6604e-04,
         2.5367e-03, -9.9180e-04,  6.1630e-03,  5.1650e-03,  4.3314e-04,
         1.3380e-03,  7.9117e-04, -1.1109e-03, -1.0547e-03, -1.9021e-04,
         3.1620e-03, -1.6093e-03,  1.1703e-04,  9.6619e-06,  3.3774e-03,
        -2.2539e-03,  6.2297e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0330, -0.8748, -0.0021,  0.0042, -0.0105, -0.0055,  0.0017, -0.0082,
        -0.0164,  0.0012, -0.0046, -0.0072, -0.0015,  0.0074, -0.0217,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2657e-03,  7.9139e-01,  3.6993e-03,  6.5473e-03,  7.3771e-03,
        -3.7705e-03, -7.2907e-03,  9.3649e-03, -1.6485e-03, -1.7162e-03,
        -4.5934e-04, -9.0341e-04,  5.2716e-04,  1.0663e-03, -3.5113e-03,
        -5.4827e-03,  2.1152e-03,  2.1674e-03,  1.2041e-02, -2.7891e-03,
         1.8671e-03, -4.0580e-03, -9.4816e-04,  3.3129e-03, -1.4456e-03,
        -7.6513e-03,  1.8703e-03,  3.0709e-02, -7.1035e-03,  1.0025e-03,
        -3.5235e-03, -2.5014e-03, -6.6442e-04, -1.1306e-03,  2.8802e-03,
         1.6282e-03, -2.5113e-04,  7.7664e-04, -6.5009e-03, -7.6659e-04,
        -1.4830e-03, -3.7928e-03, -3.7960e-03, -1.5143e-03, -2.1109e-04,
        -6.0713e-03,  6.5532e-03,  2.8856e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-1.8151e-02,  8.4610e-01, -1.0647e-02, -1.4502e-03,  1.9920e-02,
         2.1207e-02,  7.0160e-03,  4.3304e-03, -1.0011e-02, -9.9489e-03,
        -3.4711e-03,  6.0247e-03, -5.0296e-03, -1.6928e-03, -9.1705e-04,
        -1.3300e-03, -2.6658e-03, -2.5909e-03, -1.8431e-04,  1.8339e-04,
        -5.8378e-04, -6.8079e-03,  3.7408e-03,  2.9721e-03,  9.9265e-04,
        -1.2036e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2012e-02, -8.9966e-01,  1.7796e-02, -4.2200e-03,  1.1377e-03,
        -2.8116e-03,  6.4014e-03,  1.6105e-03,  2.0917e-04,  3.4780e-03,
        -2.3363e-03,  8.4741e-03,  3.0154e-03, -1.7231e-03,  5.4422e-03,
        -6.4958e-04, -3.1793e-03,  5.2326e-03,  3.1233e-03, -1.4546e-02,
        -2.9389e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4133e-02,  8.4419e-01,  1.6953e-02, -4.3215e-03, -2.4172e-02,
        -4.4268e-03, -2.4636e-03, -8.3203e-04, -1.1501e-03, -1.3929e-05,
        -4.5894e-03, -3.0038e-03, -1.0984e-03, -5.6221e-03,  1.5317e-03,
        -3.3699e-03, -9.8237e-04, -3.3906e-03,  5.0880e-03, -7.2785e-03,
        -2.4139e-03, -1.7070e-03, -6.8022e-03, -7.4018e-03,  1.7143e-02,
         4.4719e-04, -1.4770e-03, -2.6612e-03, -8.9897e-04,  7.3463e-04,
        -1.1253e-03,  1.5156e-03,  3.8895e-03, -5.7573e-04, -2.0700e-03,
        -5.2541e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.4599e-02,  8.4922e-01, -1.3520e-02,  2.2373e-03, -4.7373e-03,
         5.5720e-03, -4.8275e-03, -1.3935e-03,  4.4226e-03, -2.4227e-03,
         8.7149e-03, -6.8285e-04,  4.0423e-03, -1.2231e-03, -3.8623e-03,
         6.8295e-03,  3.3356e-03, -7.3874e-03,  1.0680e-02, -1.0286e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6680e-02, -8.0716e-01, -2.4746e-02, -4.7071e-03,  2.9170e-03,
         8.7308e-03, -3.6683e-03,  1.1387e-02, -1.4330e-02, -1.8683e-03,
        -9.6901e-03,  9.6584e-05,  7.5375e-03,  2.1263e-04, -2.0301e-02,
         3.0693e-03,  5.0898e-03, -1.0641e-03,  1.1376e-03, -1.3781e-02,
         1.8236e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0103e-02,  9.0303e-01, -1.4799e-02,  2.9367e-03, -7.1689e-04,
         1.4516e-03,  1.0219e-03, -3.5747e-03, -2.5912e-03,  8.3777e-04,
        -9.5563e-04,  3.2883e-02, -5.1026e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7487e-02,  8.7804e-01,  1.0957e-02, -5.7645e-03,  1.6445e-04,
        -6.2871e-03, -5.2708e-03, -4.0795e-03, -2.8413e-03, -8.2549e-04,
        -3.5507e-03, -5.3815e-03, -2.2054e-03,  2.5203e-03, -3.1565e-03,
        -1.6853e-03,  9.0265e-04, -2.6893e-03, -1.6954e-03, -1.4234e-03,
        -1.0797e-03,  3.7305e-03, -3.7377e-03, -2.4887e-03, -1.9282e-03,
         3.6978e-03, -3.1352e-03, -3.0810e-04,  1.6357e-03, -2.8481e-04,
        -1.7102e-03,  4.6091e-03,  1.3877e-03, -1.8887e-03,  2.1990e-03,
         6.4274e-03,  2.8281e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4199e-02,  9.0897e-01, -4.9929e-03,  5.5061e-03,  4.4560e-03,
        -5.4993e-03, -1.6200e-03,  1.9534e-03,  5.1258e-03, -1.8861e-03,
        -6.4693e-03,  7.1011e-03,  5.5845e-03, -1.7643e-03,  1.5178e-03,
         1.5907e-03,  3.9348e-03,  9.1134e-04, -2.3581e-03,  1.4782e-03,
        -1.2034e-03,  9.0591e-04,  6.5399e-03,  4.4297e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4801e-02,  8.5540e-01,  1.6170e-02, -8.2322e-05,  3.8474e-03,
         3.1954e-03, -3.0805e-03,  4.9067e-03,  1.5996e-03, -8.5388e-03,
        -7.5182e-03, -1.4205e-03,  8.8783e-03, -3.1047e-04,  9.6300e-04,
        -1.1503e-02, -6.8506e-03,  9.9906e-05,  3.7115e-03,  2.6699e-03,
        -2.2093e-03, -5.1710e-03,  4.7799e-05, -5.8238e-03, -2.0198e-03,
        -1.6642e-03, -7.1419e-04, -6.3953e-03, -2.8686e-04, -1.8961e-03,
        -5.5345e-03, -1.2693e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8332e-02,  8.8573e-01, -1.4975e-03, -3.3027e-03, -7.0282e-03,
        -1.2527e-02,  8.8224e-03,  1.9874e-03, -5.8499e-04, -4.0542e-03,
         3.6470e-03, -1.4199e-03, -7.9839e-04,  4.5585e-04, -1.7270e-03,
         2.0787e-03, -5.2774e-04, -1.0665e-03, -1.9630e-03,  3.4085e-03,
         6.5632e-04,  2.9052e-03, -1.1070e-03,  5.1738e-04,  2.5686e-04,
        -8.6305e-03, -6.1626e-03, -1.7184e-03,  1.4592e-03,  5.4341e-03,
         1.9061e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4395e-02, -7.9652e-01, -3.2388e-02,  2.1835e-03, -6.1922e-03,
        -2.1458e-04,  2.0301e-02,  3.9582e-02, -2.4188e-03, -2.0581e-03,
         1.0681e-03, -1.1281e-02, -5.2512e-03,  1.2461e-02, -3.1502e-03,
        -1.1006e-03,  4.7043e-03,  2.3839e-03, -2.4031e-03, -2.7730e-05,
        -6.4840e-03,  2.1027e-04,  2.3217e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6514e-02, -8.8960e-01, -9.4177e-03,  4.5643e-03, -1.2733e-03,
        -1.1191e-02,  2.1812e-03,  1.0191e-02,  4.1270e-03, -1.0184e-03,
        -4.8060e-03,  2.3477e-03,  3.6440e-03,  1.5610e-02,  3.3578e-04,
        -1.9688e-03, -5.0250e-04,  2.7450e-03,  2.0347e-03, -1.3820e-03,
        -1.7995e-03,  4.3396e-05, -3.7832e-03,  3.1241e-04,  1.6317e-03,
        -2.5891e-03,  3.4587e-03,  9.2346e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 4.3833e-04,  8.8760e-01, -2.9021e-02,  2.1893e-03, -6.3861e-03,
        -2.0185e-04,  2.0647e-03, -9.1562e-04,  2.0392e-03, -7.2752e-03,
         1.7167e-03, -3.0912e-03,  2.6977e-03,  2.9119e-03, -2.9523e-03,
         2.1103e-03, -5.2682e-04,  6.7189e-03, -4.5040e-03, -2.5062e-03,
         4.6723e-03,  2.2560e-03, -6.0388e-03, -2.7423e-04,  4.9799e-03,
         3.4939e-03,  5.6853e-03, -2.0283e-03,  2.6996e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0911e-02,  8.4707e-01, -5.5288e-03,  1.3317e-02,  1.6102e-02,
         2.8252e-03,  4.9821e-03, -5.9109e-03, -1.0603e-03,  5.7687e-03,
         6.0845e-03, -5.6533e-03, -2.2397e-05, -1.0314e-03, -1.0842e-03,
        -5.1137e-03,  5.4313e-04, -7.6723e-03,  1.8535e-03,  8.2298e-04,
        -1.1285e-03, -1.2256e-02, -1.6521e-02,  6.7368e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9598e-02,  8.6639e-01, -1.9832e-03,  5.0954e-03,  7.3797e-03,
         8.3007e-03, -2.7999e-04,  5.2849e-03,  9.9226e-03,  1.7215e-03,
         1.2093e-03,  4.8962e-03,  9.5201e-03,  1.5673e-03,  1.6839e-03,
         4.3847e-03, -4.3047e-03,  3.1713e-03,  2.5548e-03, -6.1937e-05,
        -2.5808e-03,  1.6143e-04,  1.1329e-03,  5.6124e-03,  2.7736e-03,
        -1.4071e-03, -6.9041e-03,  1.9252e-03, -3.1919e-03,  3.8462e-03,
        -2.7168e-03, -1.1320e-03, -4.8864e-03,  2.4180e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4889e-02,  8.9080e-01, -2.0070e-02, -1.7149e-02, -3.9880e-04,
        -5.4480e-03,  1.9954e-03,  4.7591e-03,  1.6482e-03, -2.6322e-03,
        -3.7305e-03, -2.1392e-03, -3.7434e-03, -8.0893e-04,  2.1875e-03,
        -3.4833e-03, -7.6898e-03,  6.0558e-03, -1.0258e-03,  1.4865e-03,
         2.5486e-03, -5.3103e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4532e-02,  8.7261e-01, -1.0816e-02,  1.4401e-02, -2.4237e-03,
        -4.5232e-03, -4.3614e-03,  1.0790e-02,  2.1858e-03, -3.5615e-03,
        -1.4211e-03, -2.8744e-03, -4.0363e-03, -5.4221e-03, -5.5043e-04,
         7.6292e-03,  5.3645e-03,  2.2499e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6782e-02,  8.7156e-01, -2.5525e-02, -3.9920e-03,  3.9832e-03,
        -5.5365e-03, -3.3832e-03,  4.8272e-03, -3.7031e-03,  5.8581e-03,
         3.5129e-04, -3.5674e-04, -4.3436e-03, -1.5493e-03,  3.5710e-03,
         1.2015e-03, -5.5889e-03, -3.2752e-03, -7.4222e-03, -3.2316e-03,
         1.8293e-03, -9.7448e-03, -2.3887e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0174,  0.8833,  0.0225,  0.0018, -0.0164,  0.0016,  0.0015, -0.0059,
        -0.0094, -0.0020, -0.0019,  0.0011, -0.0023,  0.0092, -0.0015, -0.0061,
         0.0160,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0334, -0.8687,  0.0012, -0.0110,  0.0014, -0.0052,  0.0024,  0.0069,
         0.0085,  0.0109, -0.0098,  0.0035, -0.0010, -0.0022,  0.0030, -0.0101,
         0.0178, -0.0030,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2024e-02,  8.3694e-01, -2.9107e-02,  3.6229e-03, -4.7843e-04,
         2.8894e-03, -4.6507e-03,  4.4533e-03, -2.8992e-03,  1.9865e-02,
         3.4337e-03,  9.3898e-03,  6.5876e-03, -4.2281e-03, -6.5157e-03,
         7.0958e-03, -5.8406e-03,  4.4556e-03,  2.3053e-03, -3.5804e-03,
        -2.8711e-03, -3.3829e-03,  3.2360e-03,  4.1263e-04,  2.4067e-03,
        -1.7328e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.3043e-04,  8.9666e-01,  8.8705e-03,  5.8869e-04, -1.6505e-03,
        -6.9822e-03, -3.8323e-03, -2.9240e-03, -6.3154e-04,  2.8517e-03,
         3.1275e-03, -4.4941e-03,  2.6225e-03, -6.0003e-03, -6.4548e-05,
         4.7298e-03,  3.9112e-03,  3.7220e-04,  2.8200e-03, -1.5626e-03,
         4.1503e-03, -4.5511e-04,  2.0957e-03,  4.7162e-03,  4.7392e-03,
         8.1999e-03, -3.0539e-03,  2.2951e-03,  1.0854e-03,  8.9787e-03,
         4.6006e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3453e-02,  8.7982e-01, -9.9709e-03, -6.0152e-03, -7.8539e-03,
         9.2128e-03,  2.7198e-03,  1.2432e-02, -1.1854e-03, -9.2396e-03,
        -1.4310e-03, -2.7036e-03,  1.6480e-03, -2.0244e-03,  3.2222e-04,
        -5.9194e-04, -6.6834e-03,  2.6911e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0152,  0.8991,  0.0052,  0.0130,  0.0175, -0.0235,  0.0114,  0.0152,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 1.9447e-02,  8.9830e-01, -4.3318e-03, -6.2045e-03, -1.0659e-02,
         2.9420e-04, -2.9618e-03, -4.1201e-05, -2.7268e-03, -9.0324e-04,
        -2.8410e-04,  3.0313e-03,  2.0468e-03,  8.1010e-04,  8.2774e-04,
        -1.1934e-03,  1.6776e-03, -1.1224e-03,  2.3885e-03,  3.9000e-03,
         1.6085e-03,  4.0240e-04,  2.0533e-03, -1.0783e-03, -1.3226e-03,
        -4.8453e-05, -2.6036e-03,  1.0324e-03,  3.1485e-03, -8.2103e-04,
         4.0764e-03,  2.1288e-03,  9.4598e-04,  1.5296e-03, -1.1809e-03,
        -8.7691e-04, -1.5806e-03,  4.0934e-03,  6.3198e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1281e-03, -8.8873e-01,  2.6004e-02,  2.1855e-03,  1.0381e-03,
         2.0127e-04,  3.5674e-04,  7.7311e-04,  6.2226e-03,  3.2898e-03,
         1.7891e-03,  2.2576e-03, -2.3864e-04,  3.6916e-03, -3.0515e-03,
         4.7002e-03, -5.6031e-03,  1.1601e-02,  2.3306e-04,  4.8833e-04,
         8.2908e-03,  4.4875e-03, -1.7838e-03,  2.3242e-05, -6.5963e-04,
        -4.4632e-04,  1.5811e-02, -9.1500e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0497, -0.8613, -0.0020, -0.0319, -0.0030, -0.0035,  0.0104, -0.0036,
        -0.0019,  0.0102, -0.0067, -0.0012, -0.0146,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3424e-02,  9.0674e-01, -2.0139e-02, -7.4533e-03, -1.8159e-03,
        -1.6805e-03, -2.8062e-03,  1.8281e-03, -5.7972e-04,  2.5830e-03,
        -4.4304e-03, -1.9315e-04, -2.8490e-03,  8.7504e-04, -1.2598e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4821e-03, -8.6320e-01,  1.1229e-02, -1.9486e-03, -6.4404e-03,
         6.5260e-03,  3.1280e-03,  5.4240e-04, -9.1034e-03, -2.5589e-03,
        -4.6506e-04,  1.1655e-04,  4.0282e-03, -1.0303e-02, -3.3423e-03,
         2.1555e-03, -9.5868e-04,  3.7137e-03,  2.7903e-03,  2.6082e-03,
         1.2886e-03,  4.9216e-04, -2.7136e-03,  1.5034e-03, -3.1795e-03,
        -4.4291e-03, -2.1455e-03, -4.1381e-03,  2.1220e-03, -2.4519e-03,
         7.3135e-04, -1.8146e-04,  3.0792e-03, -3.0325e-03, -3.4736e-03,
        -6.4480e-04,  2.1183e-03,  2.6135e-04,  2.1781e-03, -1.0999e-03,
        -3.0719e-03,  3.9946e-03, -4.2747e-03,  2.0777e-04, -2.1354e-03,
         6.3217e-04,  6.7787e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2234e-03,  8.9238e-01, -5.9435e-03,  8.3950e-03, -1.1581e-02,
        -6.0379e-04, -3.8964e-03, -5.2484e-03, -4.5700e-03,  2.6903e-03,
         2.7619e-03,  7.2292e-04,  2.3135e-03,  6.2462e-04,  4.5616e-04,
         8.7804e-04,  3.3873e-03,  4.9937e-03, -8.4687e-04, -2.0251e-03,
         3.7142e-04, -1.1530e-03,  1.3269e-03, -4.8246e-04,  2.4779e-04,
         1.0119e-03,  1.0152e-02,  1.3374e-03, -8.5998e-04,  4.2525e-03,
        -9.3814e-04, -1.0109e-03,  1.2002e-03,  5.5485e-03, -5.9730e-04,
        -3.2574e-04,  1.6009e-03, -8.2102e-04,  2.5879e-03,  2.2415e-03,
         1.4883e-03, -1.9073e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.0025e-03, -8.9253e-01,  3.2236e-03, -8.9856e-04,  4.5562e-03,
        -1.1627e-03, -2.7479e-02,  4.6034e-03, -4.5891e-03, -3.8480e-03,
         4.6068e-03,  1.3838e-02,  4.1672e-03,  2.1967e-04, -3.0303e-03,
        -6.5418e-05, -1.1456e-02,  2.5800e-03, -7.8305e-04, -3.7407e-04,
         6.3208e-05, -9.2074e-04,  7.2593e-04,  4.6282e-03, -4.6437e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6042e-02,  9.2724e-01, -4.8952e-03, -4.4170e-04, -1.4573e-03,
         2.7589e-03,  2.8463e-04, -9.9151e-04,  1.4900e-03,  7.7855e-04,
         1.5532e-03, -3.3613e-03, -2.4554e-03, -1.0906e-02, -2.3268e-03,
         2.7297e-03,  2.1067e-03,  1.7828e-03, -8.1798e-04,  5.6825e-05,
         2.9432e-04,  3.5041e-03,  1.7288e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5127e-03, -9.1296e-01,  5.5124e-03, -3.0785e-03,  6.9935e-03,
         1.3157e-03, -6.1259e-03, -9.7031e-03,  4.4643e-03, -9.8200e-03,
         5.4012e-03,  4.8294e-04,  2.0034e-04,  6.2579e-05,  3.2706e-03,
         1.1951e-03, -1.5568e-03,  5.7136e-03,  2.5825e-03,  5.5254e-03,
         2.9175e-03,  2.7247e-03,  6.8790e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0822,  0.8380,  0.0091, -0.0187, -0.0085,  0.0035,  0.0034,  0.0028,
         0.0177,  0.0010,  0.0053, -0.0050,  0.0049,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1857e-02,  8.6502e-01,  1.1510e-02,  2.3852e-03, -1.9766e-03,
        -5.9718e-03, -5.4627e-03, -3.3307e-03, -5.8593e-04, -1.6593e-03,
        -7.5568e-03, -2.1266e-02, -4.9797e-03,  6.4350e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3831e-02,  9.0377e-01, -1.7529e-03, -5.4825e-03, -1.0197e-03,
        -2.5901e-03,  2.5375e-03, -6.2698e-03,  4.6696e-03,  7.5681e-04,
        -1.2663e-02, -2.4483e-03, -6.6446e-04,  7.6389e-04,  3.9312e-04,
        -4.3268e-04, -4.3810e-03, -1.4072e-02,  1.1497e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-3.5009e-03,  8.9604e-01,  3.8222e-03,  4.7932e-03,  8.7469e-04,
         9.0844e-04,  6.6974e-03, -7.6233e-04,  3.0303e-03,  6.2897e-03,
         2.4157e-03, -1.3552e-02,  1.7690e-03,  9.6061e-03,  1.9855e-03,
         2.8132e-03,  6.7557e-03,  5.0766e-05,  6.2418e-03,  5.7303e-03,
         2.3433e-03,  5.6559e-03, -2.1810e-03,  8.0055e-04,  6.4199e-04,
         1.0734e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1665e-02,  8.8411e-01,  1.5219e-04,  1.0561e-02,  2.6162e-03,
         4.3194e-03, -2.1513e-03, -1.2537e-02, -9.6511e-03,  8.6274e-03,
         6.6654e-03, -6.8938e-03,  3.6443e-04, -3.1104e-03,  2.0087e-02,
        -6.4829e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2854e-03,  8.5145e-01, -1.1063e-02,  6.1038e-02,  6.9085e-03,
         2.2503e-02, -7.4392e-04,  2.0620e-03, -2.6541e-03,  4.3320e-03,
         5.4162e-03,  7.8823e-04,  8.7871e-04, -6.2306e-04,  1.0662e-02,
         1.1716e-02, -1.8762e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0198,  0.9085, -0.0251, -0.0029, -0.0154,  0.0079, -0.0128,  0.0048,
         0.0029,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.5296e-03,  8.5237e-01, -4.9957e-03, -1.9947e-03, -1.7699e-02,
         3.0003e-03, -9.8863e-04,  9.5239e-03, -4.4648e-03, -2.3643e-03,
         2.4398e-03,  6.5198e-03,  1.4030e-03, -3.3003e-03,  8.3553e-03,
         3.9700e-03,  3.6480e-03,  6.2843e-03,  8.6267e-04,  5.5767e-03,
         2.9589e-03,  2.5885e-03, -2.5590e-03, -4.3021e-03,  2.9123e-03,
        -1.7375e-03, -3.4666e-03, -1.5549e-03,  3.2314e-03, -2.2235e-04,
         3.7036e-03,  9.9079e-04,  2.4079e-03, -2.8357e-03,  1.8347e-03,
        -2.0359e-03, -1.0431e-03,  1.1597e-03, -3.1805e-04, -1.8539e-03,
         1.6124e-03,  1.0442e-03, -2.3962e-03, -5.8069e-04,  2.3360e-03,
        -4.0255e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.3995e-03,  9.1307e-01, -1.0858e-02, -7.5474e-03, -3.3899e-03,
        -9.5779e-04,  1.6313e-05,  4.5531e-03, -6.9506e-04, -3.5900e-03,
        -1.3047e-03, -2.9105e-03,  5.3071e-03,  1.6432e-04,  2.4999e-03,
         4.3368e-03,  4.5310e-03, -2.8733e-03,  9.4427e-04,  1.2766e-03,
         1.5073e-03,  4.0922e-05, -5.2254e-03,  2.0937e-03,  2.6762e-03,
        -2.7061e-03,  4.4544e-03,  5.4494e-04,  4.4354e-03, -1.0934e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0256,  0.9318, -0.0092, -0.0088, -0.0016,  0.0072, -0.0158,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0191, -0.9161,  0.0165,  0.0018,  0.0019,  0.0019, -0.0063, -0.0113,
        -0.0104,  0.0126,  0.0020,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8937e-03,  9.1413e-01, -1.0590e-04,  1.7873e-02,  5.2915e-03,
         8.0221e-03,  5.8226e-03,  7.6850e-03,  8.4043e-03,  2.3595e-03,
         2.6853e-04,  1.4261e-02, -3.3614e-03, -8.5261e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.4392e-03,  8.9921e-01,  9.5954e-03, -1.7715e-03,  6.6173e-03,
        -8.5722e-04,  6.6447e-03, -3.6939e-03,  1.1905e-03, -5.9865e-03,
         1.5279e-03, -1.3780e-04, -1.6953e-03,  1.7420e-03, -1.2983e-03,
         1.1185e-03,  2.8021e-03, -5.6569e-05, -1.4571e-03,  3.6226e-03,
         5.2445e-03,  4.5054e-03,  1.9067e-03, -6.1027e-04, -1.5419e-03,
        -2.0597e-03, -3.8820e-03,  2.8842e-03,  7.8732e-04,  1.4516e-03,
         4.2746e-03, -7.4890e-04, -2.6160e-03,  1.0023e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7430e-02,  8.7446e-01,  1.1304e-02, -1.3917e-03, -1.5331e-03,
        -1.1631e-03,  2.5067e-04, -7.2443e-03,  1.4811e-03, -1.9309e-03,
         3.3075e-03, -2.0173e-03, -9.3777e-04, -3.5905e-03, -2.0433e-02,
        -1.2281e-03, -1.5964e-02,  4.8196e-05, -1.7625e-03, -5.3890e-04,
         1.1119e-03,  1.9630e-03,  1.5794e-02,  1.5638e-03, -1.2953e-03,
        -3.3352e-03,  1.4768e-03, -5.4460e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5207e-02,  8.2423e-01,  1.9187e-02,  7.9728e-03,  6.6777e-03,
        -8.3279e-04, -4.6398e-03,  2.8980e-03, -1.9903e-03, -4.1508e-04,
         2.4071e-03, -6.0984e-03,  6.2664e-03,  8.6792e-03, -7.4236e-03,
        -5.2778e-03,  1.2414e-02, -1.0726e-02,  1.3971e-03,  1.9156e-03,
         9.2378e-04,  1.6074e-03,  1.1796e-02, -9.0132e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 8.6095e-03,  8.3052e-01,  1.7176e-02,  1.7572e-02, -1.8919e-04,
         1.7552e-03,  6.2422e-03,  2.5000e-03,  1.1170e-02, -2.9003e-03,
         2.9136e-03,  3.8941e-04,  4.9707e-03,  1.3512e-03,  2.8811e-03,
        -2.3898e-03, -3.6281e-03,  3.8694e-04,  4.0785e-03,  2.2297e-02,
         2.1381e-03, -6.9153e-03, -4.9561e-03,  1.3018e-03,  4.2793e-03,
         1.0279e-02, -9.9486e-04, -4.3991e-05,  5.9273e-03, -1.0373e-02,
        -8.8665e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0360,  0.8606, -0.0109,  0.0030,  0.0050,  0.0340,  0.0107, -0.0012,
         0.0023, -0.0026, -0.0310, -0.0028,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0535,  0.8583, -0.0080, -0.0117,  0.0019, -0.0039, -0.0084, -0.0041,
        -0.0011, -0.0043, -0.0172,  0.0013,  0.0033,  0.0099,  0.0130,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4318e-02,  8.6995e-01,  1.4013e-02, -1.4454e-03, -5.2771e-03,
         1.0625e-04, -2.1983e-03, -2.4069e-03, -3.5429e-03,  3.1999e-03,
         1.8018e-03,  3.3848e-03, -5.5085e-04, -2.6440e-03, -2.1462e-03,
        -1.7219e-03, -1.7512e-04, -6.5420e-03,  1.1455e-03,  2.7240e-03,
         1.2960e-03,  5.1513e-04, -1.6247e-03,  7.4118e-04, -2.5394e-03,
        -1.4639e-04,  1.1867e-03, -1.1818e-03, -1.5824e-03, -1.0855e-04,
         6.7225e-05,  9.8969e-04,  8.7222e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.2993e-02,  8.8262e-01,  3.7421e-04, -1.5011e-02,  1.1034e-03,
        -9.4862e-04, -1.3847e-02, -6.3999e-03, -5.1892e-03,  6.4751e-04,
        -8.8749e-03,  1.4017e-04, -1.3509e-03,  5.6675e-03, -3.9683e-03,
        -8.6118e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0668,  0.8834,  0.0065,  0.0117, -0.0028, -0.0127,  0.0033,  0.0020,
         0.0107,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0156,  0.8446, -0.0696,  0.0092, -0.0610,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.8536e-03,  8.8542e-01, -1.4343e-02,  4.3522e-03,  2.2227e-03,
        -1.2191e-02, -3.0805e-03,  2.7664e-03,  4.9068e-03, -3.8326e-03,
         6.6025e-03, -1.4555e-03, -5.9543e-03, -1.1619e-03, -2.2629e-04,
        -4.0319e-03,  3.7233e-03,  2.2874e-03, -3.6615e-03, -4.0060e-03,
        -2.8382e-04, -1.8603e-03,  3.9641e-03, -4.7568e-03,  1.5869e-03,
        -4.8440e-03, -8.6293e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0906, -0.8179,  0.0024,  0.0018, -0.0219, -0.0313, -0.0343,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8041e-02,  8.9099e-01,  6.5578e-03,  4.2332e-03,  3.6732e-04,
         2.4055e-03,  1.4288e-02,  3.3336e-03,  2.3380e-03, -1.2904e-02,
        -3.5365e-03, -2.8440e-03,  5.0711e-03,  1.1144e-02, -1.5527e-03,
         1.3672e-03,  4.7795e-03, -5.8062e-04, -4.8831e-03,  3.1159e-04,
         6.1431e-03, -2.3304e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4232e-03, -9.0581e-01,  4.7369e-03, -8.2471e-03,  2.0829e-02,
        -2.0122e-03,  3.0796e-04,  1.6951e-03,  4.0085e-03,  2.0884e-03,
         2.8604e-03, -6.0781e-03,  2.3342e-02, -5.0546e-03,  5.8398e-04,
        -2.2005e-03,  5.7692e-04,  4.4171e-03, -2.7230e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1867e-03,  9.0838e-01,  2.5432e-02,  3.8148e-04, -1.1230e-03,
         7.3704e-03, -2.1531e-03, -3.6154e-03,  5.2256e-03, -2.8423e-03,
         2.1553e-03,  1.1584e-03, -6.4243e-04,  4.0329e-03,  1.6865e-03,
        -4.4298e-04,  2.3907e-03,  3.0817e-03, -3.1657e-04,  1.0373e-03,
         5.2431e-03,  1.7412e-03,  6.3561e-05,  6.3176e-04,  1.2711e-03,
         6.3908e-04, -3.0310e-03, -1.2585e-03, -2.9639e-03, -4.7615e-03,
        -1.7423e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-3.7170e-02, -8.2728e-01, -3.7681e-02, -4.6987e-03, -9.8432e-04,
         4.0180e-03, -3.5117e-03,  7.3640e-04,  3.8944e-03,  9.8299e-05,
         2.4793e-03,  2.4828e-03,  1.9619e-02, -6.6784e-03, -2.9821e-03,
         4.8867e-03,  1.9781e-03, -3.6023e-03,  1.4985e-03,  2.1412e-03,
         1.1307e-03,  2.9849e-03,  1.3299e-03, -6.3828e-03, -1.3951e-03,
         2.3797e-03,  6.0401e-04,  1.1363e-03, -2.8415e-03, -5.4943e-03,
        -1.5958e-04, -5.7375e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3508e-02, -9.0835e-01,  1.3402e-02, -4.3948e-05,  9.5983e-04,
        -1.6745e-03, -6.2493e-04, -2.7420e-03, -1.5743e-02,  3.0712e-04,
         3.9448e-03, -6.3581e-03,  2.3869e-03, -1.6957e-03, -5.7703e-03,
        -7.0416e-03, -3.7848e-03,  2.0390e-03,  3.5987e-03, -2.4326e-04,
         6.9475e-04, -5.0850e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1689e-02,  9.1220e-01, -1.0624e-03, -8.0228e-03,  7.2820e-04,
         4.0657e-03,  3.6668e-04, -1.8394e-03,  1.2756e-03,  1.1286e-03,
         1.0878e-04,  1.9252e-04,  5.5160e-04,  8.9167e-04, -1.1980e-02,
         2.1942e-03, -5.2628e-03,  1.0667e-02,  4.7771e-04,  4.0369e-03,
        -1.1301e-03,  9.6410e-04,  6.5817e-04, -9.7299e-04,  1.0093e-03,
         5.9465e-03,  8.6903e-04, -3.4313e-03, -5.8423e-04,  5.6898e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4375e-02,  9.3310e-01, -1.0576e-02, -5.0710e-03, -1.2116e-03,
         1.4803e-04, -1.3513e-03,  1.4940e-03, -4.1631e-03,  1.5270e-03,
        -1.3564e-03,  8.1843e-03, -3.0765e-03,  5.7079e-03, -2.3318e-03,
         1.5413e-03,  1.5694e-03,  3.2123e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.9678e-03,  8.9821e-01, -4.3561e-03,  4.1688e-03, -8.0592e-03,
        -4.9667e-03, -7.9661e-04,  5.6998e-04,  1.2876e-04,  5.0613e-04,
         2.8088e-04,  6.3910e-04,  2.2522e-04, -1.8445e-03, -1.0918e-03,
        -9.1724e-03,  2.9276e-03, -8.5640e-03,  7.0228e-04, -1.6388e-03,
        -1.5784e-03,  5.5192e-03,  3.7409e-03,  2.0828e-03, -1.1888e-03,
        -3.0795e-04, -5.4237e-03,  1.6196e-04, -6.8973e-05, -5.4870e-04,
        -9.9181e-04, -3.3065e-03, -1.6480e-03, -1.4485e-03,  1.6444e-03,
         4.1950e-03,  7.3308e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5435e-02,  8.9842e-01,  4.6220e-03,  1.2198e-02, -6.3177e-03,
         4.1908e-03,  3.7865e-04, -6.4752e-04, -1.3897e-04,  4.4196e-04,
         8.4276e-04,  6.8551e-03,  4.4699e-03,  1.4933e-03,  6.6068e-03,
         2.8141e-04, -1.8621e-04,  1.5409e-03, -6.7314e-04,  2.6084e-03,
         4.5686e-03, -2.4692e-04,  2.0210e-03,  4.8119e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7721e-02,  8.8229e-01,  4.2760e-04, -1.1431e-02, -1.3783e-03,
         2.6119e-03, -4.5548e-03,  2.9854e-03, -3.7465e-05, -2.1058e-03,
         1.3429e-03, -2.8588e-03, -4.4728e-03, -4.3697e-04,  1.4945e-03,
         2.6521e-04, -3.5629e-03,  2.0016e-04,  8.4986e-04,  1.9365e-03,
         4.2115e-04, -6.9024e-03,  4.6485e-04,  7.5447e-05,  3.8505e-03,
         8.3210e-03,  1.2639e-03, -2.9188e-04, -2.9125e-04,  1.7476e-03,
         2.3992e-03,  1.2571e-03,  2.1451e-03,  3.0963e-03,  2.8238e-03,
         3.2254e-03, -2.1119e-03, -2.0749e-03,  1.5421e-03,  2.2494e-03,
         2.8357e-03, -2.1369e-03, -4.1517e-04,  5.3108e-04,  1.2687e-03,
         3.2954e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1691e-02,  8.7595e-01,  3.4552e-04, -5.4772e-03, -8.1998e-03,
         7.4945e-03, -1.4773e-02,  2.8709e-04,  3.2358e-03, -2.9113e-03,
         1.7886e-03,  5.1335e-03, -4.0033e-03,  1.3300e-03,  2.2986e-03,
         5.0877e-03,  9.1470e-03,  1.8162e-02,  1.5856e-05,  5.4534e-03,
         7.1690e-03,  5.4989e-03,  4.5440e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6106e-02, -9.1048e-01,  1.7614e-02,  1.3063e-03,  2.0551e-03,
        -2.6816e-03, -1.1804e-03,  8.5888e-04, -1.7478e-02,  6.4634e-03,
        -2.3772e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1102e-03,  9.5019e-01,  7.7256e-03, -5.2066e-03,  5.3503e-04,
         9.9981e-04, -4.6908e-04, -5.0603e-03,  4.2150e-03,  9.2529e-03,
        -1.5310e-03, -1.8681e-03,  6.9883e-03,  8.4363e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2630e-04,  9.4716e-01, -1.0092e-03, -1.5664e-02, -1.0914e-02,
         4.1807e-03, -3.3320e-03, -7.8107e-04,  1.9853e-03, -6.1540e-03,
        -2.8401e-03,  5.7578e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7480e-02, -8.6399e-01, -3.4995e-05, -2.8953e-03,  2.8290e-03,
         4.1518e-03,  3.5590e-03, -6.6826e-03, -5.3846e-03, -5.3261e-03,
         4.2419e-04,  4.5864e-04,  1.7752e-03, -3.1016e-03, -5.0086e-05,
         5.0271e-05, -1.2343e-03, -1.7404e-04, -1.6501e-03, -1.8093e-04,
         1.1600e-03, -1.1242e-03, -4.5117e-03, -8.3586e-04,  1.7234e-03,
         4.4819e-03, -4.7183e-03, -1.1063e-04,  1.6963e-03, -3.4646e-03,
        -6.6318e-04, -1.0622e-02,  3.8470e-04, -7.3786e-04,  8.6511e-05,
         4.0190e-03, -9.2102e-04,  4.6199e-04,  1.6140e-03,  1.1653e-03,
        -1.5199e-03, -2.2305e-03,  1.3569e-04, -5.7095e-04, -9.5843e-04,
         3.7677e-04, -9.3380e-04,  9.7377e-05,  1.1744e-03, -6.1604e-05,
        -5.2259e-04, -1.2658e-04,  1.4088e-03, -1.5260e-03, -1.2424e-02],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-1.9485e-02,  8.7621e-01, -2.8968e-02,  7.6565e-04,  4.4100e-03,
        -8.1923e-05,  1.9316e-03,  3.5709e-04,  2.2245e-03,  8.7795e-04,
        -1.4750e-02, -4.4543e-04,  6.6184e-03, -3.8469e-03,  5.4026e-03,
         2.5421e-03, -2.8536e-04,  1.4678e-02, -6.4479e-04,  2.1482e-03,
        -5.9114e-03, -1.9218e-03, -5.4974e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0259,  0.8641,  0.0086, -0.0064,  0.0096,  0.0060, -0.0095, -0.0022,
        -0.0027, -0.0027, -0.0115,  0.0009, -0.0096,  0.0025,  0.0023, -0.0059,
        -0.0020,  0.0275,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0502,  0.8486,  0.0078,  0.0058,  0.0140,  0.0048,  0.0118,  0.0033,
         0.0060, -0.0077,  0.0022,  0.0058,  0.0104, -0.0083, -0.0024, -0.0022,
         0.0063,  0.0025,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.1839e-03,  9.0387e-01,  1.1221e-03,  6.9278e-04, -7.7882e-03,
         3.3288e-03,  2.4027e-03, -6.0235e-03,  6.8224e-03,  3.3599e-03,
        -5.0706e-04,  1.0540e-02,  9.1686e-03,  1.3044e-02, -2.3142e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5881e-03,  8.7577e-01,  5.7940e-03,  1.5868e-03,  5.3253e-05,
        -2.5038e-03,  5.3554e-04,  1.2072e-03,  9.9062e-03,  1.3009e-03,
        -2.7749e-03, -1.5154e-03,  2.1013e-03,  1.8842e-03, -1.4111e-03,
         3.1292e-03,  1.1653e-03,  2.8186e-03, -1.7628e-03,  1.0171e-02,
         1.1349e-02,  2.2469e-03, -1.2601e-04,  5.3839e-04,  1.3616e-02,
        -2.1369e-03,  1.6156e-02, -1.6277e-04, -7.7177e-04,  8.9746e-03,
        -6.9319e-03,  4.0126e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3782e-02,  8.5984e-01, -1.6535e-03, -6.5606e-04,  3.7002e-03,
         1.6317e-03, -2.9137e-03, -2.4846e-03, -1.5293e-02,  3.6204e-03,
         1.3849e-03, -1.7925e-03, -3.5626e-03, -5.1224e-03,  1.6859e-04,
         4.9579e-04, -1.7131e-03, -1.1205e-02, -2.8541e-03, -7.5285e-04,
        -2.9111e-03, -3.1326e-03, -1.1993e-02,  1.9573e-03,  1.5629e-03,
         3.6906e-03, -3.3861e-04,  1.2861e-03,  7.6979e-04,  4.3569e-03,
         8.9530e-03, -7.3148e-04,  3.6952e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4795e-03,  8.9560e-01,  1.0570e-02, -3.7949e-03,  1.6440e-03,
         3.6560e-03,  5.4361e-04, -1.8240e-03, -3.9883e-04,  5.8796e-04,
        -8.1193e-04, -3.6379e-03, -4.0658e-03, -2.6051e-03, -5.7312e-03,
         1.1063e-03,  1.5194e-03, -2.9578e-03, -1.8115e-03, -5.5951e-04,
        -7.6165e-04,  1.3181e-03, -1.3449e-03,  1.7900e-03,  6.2857e-04,
        -6.5199e-04, -1.1831e-03,  1.9174e-03,  4.7511e-04, -3.2909e-03,
         9.9947e-04,  4.9324e-04,  8.5207e-04, -2.3178e-03, -7.1063e-04,
         1.0775e-03, -7.1119e-04, -1.5909e-03,  8.9043e-04,  8.8053e-04,
         2.0316e-03, -1.2801e-04,  2.1373e-04,  1.8135e-04,  1.8219e-03,
         2.7495e-03, -4.3182e-04, -7.4342e-05,  1.6044e-03,  2.1153e-03,
         1.9838e-03,  1.9901e-03, -2.6584e-04, -6.1922e-04,  1.8609e-03,
         7.1741e-04, -2.4909e-03,  4.9257e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4948e-02,  8.5570e-01, -2.2787e-02, -1.1199e-02,  4.3313e-04,
        -2.5150e-03,  7.2595e-04,  1.3803e-03, -6.6019e-03,  3.0820e-03,
         3.6597e-03, -2.6277e-03, -6.4217e-03, -8.9996e-03, -1.8135e-04,
        -1.3236e-03,  6.5154e-03,  2.8637e-04, -1.9241e-03,  2.1118e-03,
         1.2550e-03, -1.3700e-04, -4.1393e-03,  5.3794e-03, -1.0084e-03,
        -1.4656e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5133e-02,  8.8257e-01, -8.4435e-03, -3.9695e-03, -7.5822e-03,
         3.7725e-03,  2.9955e-03,  1.0494e-04,  1.4933e-03, -2.0992e-03,
         1.2496e-02, -1.9755e-04,  6.2674e-03,  3.9203e-03,  1.4128e-03,
        -2.1082e-03,  1.6059e-03,  7.7160e-04,  6.7348e-03, -1.1826e-03,
        -1.7928e-03, -5.4090e-03, -3.6200e-03, -3.8170e-04, -3.9360e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2466e-02,  8.7649e-01,  3.7167e-03, -9.6357e-03,  4.5388e-03,
        -1.6034e-03,  5.3366e-03, -2.7453e-04,  1.6639e-04,  6.6465e-03,
        -4.9009e-05,  7.0480e-04, -3.8557e-04,  3.2580e-03, -1.0468e-03,
         7.8497e-04, -1.2874e-03,  4.9518e-04,  9.4130e-04,  5.0307e-03,
        -1.6494e-03,  5.9812e-03, -3.4921e-03, -1.9136e-03, -1.9171e-03,
        -5.5135e-03,  7.0939e-03,  4.1426e-04, -2.9849e-03,  8.8450e-04,
        -2.3070e-03, -4.1396e-03, -1.0740e-03,  8.3592e-04, -2.0343e-03,
        -2.4294e-03,  8.5145e-04, -4.4065e-03, -3.5229e-03, -1.2347e-03,
        -2.9656e-03,  8.8140e-05,  3.2701e-03,  2.5979e-03,  1.5432e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3432e-02,  9.0450e-01,  1.0074e-02, -3.7064e-03, -4.8162e-03,
        -5.5397e-03, -1.4016e-03, -6.6006e-03, -4.6475e-03, -1.3242e-02,
        -2.4676e-03,  1.5005e-03,  1.8497e-03,  4.2389e-04, -1.5479e-04,
        -6.8524e-03,  4.7440e-03,  1.4635e-03,  1.5653e-03,  3.9095e-03,
         4.2940e-03, -1.8349e-03, -9.8024e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0860,  0.8128,  0.0190, -0.0132,  0.0058,  0.0185,  0.0054, -0.0030,
        -0.0033,  0.0036,  0.0203, -0.0091,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-7.6145e-04,  8.9818e-01, -2.6001e-02, -1.7202e-02,  3.0857e-04,
         4.9723e-04, -2.6674e-03,  9.5944e-04,  8.9381e-03, -3.7941e-03,
        -6.1424e-03, -1.9133e-03,  7.1713e-04, -2.4641e-04,  4.1085e-03,
        -4.2432e-03,  4.8218e-03, -5.3732e-04,  1.9034e-03, -4.9838e-03,
        -2.5466e-03,  7.5604e-03, -9.6732e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7767e-03,  8.8582e-01, -1.4187e-03, -1.1502e-03,  2.7451e-04,
        -7.5896e-03,  3.1794e-03, -1.6060e-03,  1.7605e-03,  6.4046e-03,
        -3.2389e-04, -6.2804e-03, -9.1147e-04,  1.3726e-03, -3.0095e-03,
        -1.8966e-03, -3.9020e-03, -4.5783e-04, -1.1489e-03,  4.6263e-03,
        -1.0361e-04, -2.4018e-03, -1.1721e-02, -3.0673e-03,  6.6678e-04,
        -2.8207e-03, -8.1030e-03,  7.2498e-04, -5.2671e-03, -6.7394e-04,
         1.6384e-03,  1.5929e-02,  8.9698e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3452e-02,  9.3126e-01, -1.0380e-02, -2.6069e-03,  3.1828e-03,
        -1.6348e-03, -4.5829e-03, -1.0017e-03,  2.2191e-03, -2.0080e-03,
         3.7081e-03, -5.9202e-03,  8.4374e-04,  3.1022e-03,  4.6784e-03,
        -2.2286e-03,  3.3468e-03,  1.8539e-03,  1.9930e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1856e-03,  9.3507e-01, -1.5856e-02,  9.9888e-03,  5.5535e-03,
         9.6352e-03,  7.8613e-03,  2.1770e-03, -5.0287e-03, -1.8309e-03,
         1.2325e-03, -1.7561e-04,  3.2898e-03, -1.1116e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4632e-02,  9.2122e-01,  1.2588e-02, -2.8890e-03, -1.1815e-02,
        -3.3674e-03, -8.0183e-04, -8.9346e-03,  2.4097e-03,  2.1073e-03,
         1.0306e-03, -6.4041e-05,  2.5448e-03, -3.7551e-03, -2.4454e-03,
        -2.3292e-04, -2.1868e-03,  1.0529e-03, -1.8746e-04, -1.6761e-03,
        -8.0983e-04,  3.2458e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8405e-02,  9.0168e-01,  5.0086e-03, -2.0836e-03,  4.4769e-03,
        -3.5503e-03, -1.8792e-03, -1.5346e-02, -2.3956e-03, -1.4390e-03,
        -8.2993e-04, -1.8538e-03, -2.1087e-03, -9.9006e-04, -1.2970e-03,
         5.2166e-04, -3.7182e-03,  1.1534e-03,  7.7186e-04, -1.7673e-04,
        -1.0909e-03,  2.1432e-04, -2.5983e-03, -3.9522e-03, -3.5762e-03,
         1.4420e-03,  8.3096e-04, -3.1392e-03,  1.1771e-02,  1.7018e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3098e-02,  8.5842e-01, -2.0128e-02,  4.1266e-05,  5.8400e-03,
         9.1827e-03,  1.7581e-03, -4.9096e-03,  4.8573e-03,  2.3031e-03,
        -4.1603e-03,  1.0054e-03, -6.1062e-04,  1.2576e-02,  6.9837e-03,
         1.4644e-02, -1.9482e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0512,  0.8569,  0.0350, -0.0041,  0.0009,  0.0010, -0.0018,  0.0128,
         0.0153, -0.0042, -0.0167,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.2794e-05, -8.8114e-01, -2.0312e-02,  2.5746e-03,  7.5352e-03,
         7.7575e-04,  2.7979e-03, -4.0624e-04,  3.3486e-03, -4.3153e-03,
         3.6912e-06, -5.3714e-03,  3.4143e-03, -2.2820e-02, -2.1050e-02,
         8.4965e-03,  1.5548e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8095e-02,  8.4043e-01, -1.1429e-02,  3.8985e-02,  6.7472e-03,
         1.0583e-03,  2.7274e-03, -8.9916e-04,  4.4166e-03,  1.1555e-03,
        -5.7663e-03,  1.3384e-03,  1.0196e-03,  5.6897e-03, -1.7827e-03,
         1.7996e-03, -2.0251e-03, -3.1200e-03, -1.2650e-02, -3.4736e-03,
        -3.9355e-03, -3.3035e-03,  3.6526e-03, -7.5120e-04, -3.6188e-03,
        -2.0130e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1375e-02,  8.7377e-01,  3.5059e-02, -5.4915e-04, -3.9780e-03,
         8.8568e-03,  1.7663e-03,  3.2580e-03, -5.9899e-03,  3.3221e-03,
         4.3185e-04, -2.4039e-03, -4.4372e-03, -1.9072e-04,  2.8585e-04,
         1.5383e-03, -1.2018e-03,  3.0842e-04, -1.0533e-03, -4.1634e-04,
         2.5717e-03,  3.8087e-03, -3.2537e-03, -8.2318e-04,  8.1661e-04,
        -1.5822e-03,  2.2614e-03, -2.6134e-03, -9.2992e-03, -2.7779e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4033e-03,  8.8715e-01,  1.9774e-03,  8.9985e-03, -1.0608e-02,
        -7.8510e-03,  1.5425e-02, -1.2900e-02, -2.9175e-03, -1.9769e-03,
         2.4260e-04, -4.3736e-03, -1.0344e-02, -3.9988e-03,  5.6228e-03,
         1.3244e-03,  4.9606e-03,  1.0235e-02, -3.6945e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 1.5966e-02,  8.5634e-01, -7.3043e-03,  8.7792e-03,  8.4994e-03,
        -5.5754e-03,  3.3655e-02,  3.4186e-03,  3.0573e-03,  1.3880e-03,
         1.1889e-03,  8.4337e-04,  2.0166e-03,  1.7353e-03, -5.3588e-03,
        -3.9454e-03,  1.9878e-03,  1.6504e-02,  1.4947e-03,  4.3544e-04,
         4.0708e-03, -2.5493e-04, -1.9883e-03, -3.3279e-03, -4.3438e-03,
        -6.0862e-03,  4.3775e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0246,  0.9104,  0.0043,  0.0052, -0.0019,  0.0076, -0.0013, -0.0047,
         0.0044,  0.0056, -0.0082,  0.0137, -0.0058, -0.0024,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1877e-02,  8.8598e-01,  5.7897e-03, -9.3465e-04,  4.0713e-04,
        -4.9385e-04,  1.0792e-02,  1.8932e-03,  4.9516e-03,  9.9778e-03,
         7.0716e-03,  3.9625e-03,  5.7915e-04,  2.3132e-03,  4.5992e-03,
         9.2795e-03,  8.2524e-03, -1.2024e-03,  7.8231e-03,  7.9071e-03,
        -2.5114e-03, -1.5517e-03, -1.7523e-03, -1.4449e-03, -3.4751e-03,
         3.1788e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.3760e-03,  9.3334e-01,  1.0575e-02,  5.8937e-04, -4.3154e-03,
        -2.1088e-02,  3.2282e-03,  2.8650e-03,  2.1303e-03,  8.7323e-04,
         4.5028e-03,  7.7268e-04,  2.0979e-03, -5.2457e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0058e-03,  8.9654e-01,  2.8026e-03, -9.5746e-03, -5.4186e-03,
        -3.2574e-03,  8.8100e-05, -3.6586e-03,  6.9057e-04,  2.4944e-03,
        -6.3067e-04,  2.9152e-04,  3.6139e-03, -3.4111e-03,  2.3081e-03,
         4.4724e-03, -4.7421e-04,  3.7089e-03, -2.4297e-03,  1.2535e-03,
        -2.8937e-03, -3.6663e-05, -3.2340e-03, -4.3959e-03,  4.6968e-03,
         2.3898e-03, -6.1357e-03,  5.5686e-04, -5.3041e-04, -1.1250e-03,
         1.5037e-03,  2.4380e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.7136e-03, -8.9342e-01, -2.4060e-02, -7.1495e-03,  7.5533e-03,
        -2.7882e-04,  1.0655e-02,  1.0056e-02, -9.1747e-04,  3.6508e-03,
         6.4923e-03, -1.7155e-03, -1.7185e-03,  2.6913e-04, -3.2731e-03,
        -2.2445e-03, -1.9539e-03, -6.8366e-04, -1.9599e-03, -2.8632e-03,
         7.1063e-04, -4.7724e-03,  3.1778e-03, -3.5064e-03,  2.0170e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8561e-02,  9.0249e-01, -8.7220e-04, -3.2490e-02, -2.8873e-02,
         6.7146e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2465e-02,  8.6369e-01, -5.8583e-03, -4.9548e-03,  6.8815e-03,
        -6.5607e-03,  2.5819e-03, -1.6275e-03,  1.1695e-03,  2.5834e-03,
         6.6193e-04, -1.8382e-03, -8.4499e-04,  2.6234e-03, -3.4245e-05,
        -6.5872e-04, -3.0004e-03, -1.7675e-04,  9.1374e-04,  3.7466e-03,
        -8.4453e-04,  4.2895e-03, -2.6399e-03,  6.8368e-04,  7.4040e-04,
         6.2921e-03, -3.8630e-04,  1.8745e-03, -1.4405e-03,  7.6471e-03,
         1.7946e-03, -8.8458e-04, -8.4536e-04, -1.1485e-03,  2.3126e-03,
        -2.8279e-03, -4.7578e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0539e-02, -8.7933e-01,  5.9270e-03, -3.0959e-03, -1.4053e-03,
        -4.0571e-04, -1.6072e-04,  1.0466e-03,  4.8865e-03, -3.2572e-03,
         1.2958e-03, -2.7998e-03, -2.1499e-03, -4.2924e-03, -9.6352e-03,
         4.0779e-03,  4.8845e-03, -5.4690e-03, -1.3423e-03, -4.4188e-03,
         6.4205e-03,  1.0822e-03,  1.7340e-03, -1.0533e-03, -1.9900e-03,
         5.2673e-04,  4.2294e-03, -2.4491e-03,  4.1216e-03, -4.5303e-03,
         2.3694e-03, -3.1358e-03,  5.3398e-03, -2.0711e-03,  8.5245e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9149e-02,  8.1864e-01, -7.0779e-02,  9.0863e-03,  6.2183e-03,
         2.1781e-03,  3.5656e-03, -2.1956e-04,  1.2714e-03, -6.7523e-03,
         2.5495e-03,  8.4471e-03,  1.1547e-03,  2.8417e-02,  1.1568e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.2452e-03,  9.0581e-01, -8.2617e-03,  3.5165e-03,  9.6830e-03,
         2.5436e-03,  2.0543e-03, -1.0836e-04,  2.1424e-03,  3.0193e-03,
        -2.2632e-03, -2.8180e-04, -1.7399e-03, -3.3035e-03, -6.2690e-04,
        -1.2547e-03,  1.0746e-04,  4.8880e-03,  3.7613e-03,  5.5492e-04,
         6.9410e-04,  1.8303e-03,  2.4678e-03, -1.9119e-03,  2.8797e-03,
        -2.7233e-03, -1.4152e-03, -6.9482e-04, -1.0892e-03,  8.3675e-04,
        -2.5433e-03, -1.5303e-03,  1.7221e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4266e-02,  8.8542e-01, -7.4069e-03, -1.2793e-03, -3.6045e-03,
        -1.2718e-02, -1.4914e-03,  2.7490e-03,  1.7092e-03,  5.3544e-04,
         2.0352e-03,  6.2511e-03, -6.7132e-03, -1.1039e-02, -6.3655e-03,
         1.5838e-03,  1.4106e-03, -3.4491e-04, -1.9921e-03, -1.0534e-03,
        -4.2967e-03, -7.9992e-04, -7.1733e-04,  1.2221e-03,  2.5744e-03,
         2.0563e-03, -1.2609e-02,  2.5045e-03, -3.2466e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 7.8286e-03,  9.1879e-01,  1.6961e-02,  1.3481e-03, -4.7154e-03,
        -1.2454e-03, -5.5250e-04, -3.1268e-03,  5.0120e-03, -7.4109e-03,
         4.5458e-05,  1.8562e-04,  2.8244e-03, -3.8092e-03, -2.6143e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6159e-02, -8.7673e-01,  9.8839e-03,  6.0376e-03,  5.6204e-03,
        -1.0167e-03,  7.2523e-04,  9.3157e-05, -2.4834e-03, -3.4405e-04,
         1.1228e-03,  2.1539e-03,  1.9664e-05,  9.0726e-04,  1.3720e-03,
         1.5937e-03, -2.0859e-03,  4.0662e-04, -5.5455e-04, -1.0268e-03,
         6.5234e-04,  1.1767e-03,  4.3384e-04,  7.0350e-04,  1.3701e-03,
        -1.8707e-03,  2.7220e-03,  1.2045e-03,  3.0269e-03, -6.0458e-04,
        -9.0374e-04, -9.8742e-05, -3.9921e-04,  3.7120e-04,  1.0401e-03,
        -1.6189e-03, -5.0897e-04,  1.1431e-03, -7.1440e-04,  1.5622e-03,
         6.2751e-04, -7.4062e-05, -2.5035e-04, -4.6774e-04,  1.1772e-03,
         2.9038e-03,  2.3794e-03,  2.3970e-04, -9.4173e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7809e-02,  8.3675e-01, -1.9003e-02, -5.5292e-03,  4.1900e-03,
         5.2061e-03,  2.7158e-03, -1.2832e-03,  6.5450e-03,  5.1274e-03,
        -3.6424e-04,  1.7048e-03, -8.2407e-04,  4.5867e-04, -1.9248e-03,
        -3.3039e-03,  6.1544e-03,  1.5572e-03, -4.8039e-04,  2.8384e-04,
        -2.0235e-03,  2.2627e-03, -4.1771e-03,  2.5440e-03,  3.4795e-04,
         1.4701e-03,  1.5686e-03,  9.9532e-03,  3.7351e-03, -1.8700e-03,
         3.9525e-05, -9.4734e-04, -1.8236e-03,  1.2611e-02, -1.3407e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.9007e-03, -9.3437e-01,  1.1312e-02, -5.7011e-03, -5.9652e-03,
         2.1043e-03,  2.6829e-03,  1.4124e-03,  1.3445e-03, -1.6545e-03,
        -5.2816e-04, -2.7025e-03,  8.6268e-04, -1.9562e-03, -4.3717e-03,
         4.4888e-03, -1.3250e-03,  2.5125e-03, -9.8040e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0160, -0.9009,  0.0042,  0.0060, -0.0047, -0.0056, -0.0036,  0.0026,
        -0.0051,  0.0070,  0.0080,  0.0047,  0.0010,  0.0258, -0.0047,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8783e-02,  8.5225e-01,  1.4065e-02,  6.8308e-03, -1.1270e-03,
        -5.7437e-04,  2.2620e-04,  4.9314e-03, -1.5122e-03, -3.5069e-03,
         8.6208e-03,  4.8118e-03,  2.6945e-03, -1.7498e-03, -2.6136e-03,
         1.7183e-03, -6.1148e-04,  1.9434e-03,  8.1527e-04,  4.1170e-03,
         4.2500e-02, -4.0011e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1472e-02,  8.9959e-01, -9.2377e-04, -1.3027e-03,  2.7849e-03,
        -2.5516e-03,  4.0264e-03, -1.2795e-03, -1.9808e-05,  4.5998e-03,
        -9.6990e-04,  6.2293e-03,  5.2265e-03,  6.3949e-04, -7.2499e-03,
         1.0638e-03, -8.2626e-04,  4.8656e-03,  3.6874e-03, -2.2539e-03,
         8.0469e-04,  5.1005e-03, -1.6788e-03,  7.1805e-03, -2.3944e-03,
        -9.5183e-04, -1.2521e-04,  2.3981e-03,  3.1752e-03,  4.6279e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0268,  0.8839, -0.0161, -0.0189,  0.0143, -0.0116, -0.0050,  0.0032,
        -0.0038, -0.0059, -0.0104,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2770e-03,  9.2006e-01, -1.1465e-02,  1.8774e-03, -1.8733e-03,
         3.2959e-03,  4.1923e-03, -4.9351e-03, -1.1160e-03,  2.5117e-03,
        -4.5514e-03,  1.4891e-03,  3.5878e-03,  1.4904e-03, -2.3389e-03,
         8.7234e-04, -2.6584e-03, -3.4296e-04, -4.6181e-04, -8.5566e-03,
         1.5895e-02,  3.1536e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3257e-02,  8.7116e-01, -1.4794e-02, -1.4329e-03, -2.7102e-03,
        -4.1050e-04,  2.6953e-03,  1.0361e-02, -1.6061e-03, -6.7493e-04,
        -2.3034e-03,  5.6731e-03,  7.5484e-04, -8.5385e-05, -3.2754e-04,
        -4.2644e-03, -4.4871e-04,  4.3250e-03,  1.0138e-02, -5.9943e-03,
        -2.0454e-04, -9.0622e-04, -2.7468e-03, -1.0914e-03, -2.1827e-03,
        -5.7584e-04,  2.4843e-04,  3.6636e-03, -6.6313e-03, -1.4293e-04,
         3.6197e-03,  2.2118e-04, -4.3494e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8646e-03,  8.9166e-01,  3.3768e-02,  3.8197e-03, -7.6415e-03,
        -9.4851e-04,  1.6584e-03, -1.1198e-02, -2.4705e-03, -9.6989e-03,
        -4.3328e-03,  1.5941e-05,  1.5815e-03, -2.1557e-03, -3.7319e-03,
        -5.6547e-03,  6.2728e-04, -1.7572e-03, -2.6963e-04, -2.3286e-04,
         1.3536e-03,  2.5229e-03, -6.4519e-04,  1.3502e-03, -5.0432e-04,
         2.2409e-03,  2.7667e-03,  3.5279e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.5888e-03,  9.2753e-01, -3.9413e-04, -5.0135e-03, -5.8604e-03,
         3.2778e-04,  3.7739e-03, -3.5102e-04, -8.8114e-03, -1.1498e-05,
        -2.8201e-03, -1.8196e-03,  3.8858e-04,  2.3568e-03,  1.1815e-03,
        -6.2552e-04,  5.2136e-04, -9.3371e-03, -5.0987e-04,  1.0457e-02,
         1.5335e-03,  1.4881e-04, -7.0256e-03,  9.3242e-04,  6.7468e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-1.3744e-03,  8.5814e-01,  9.3792e-03, -1.2315e-02,  2.1774e-03,
         8.4618e-04,  8.3345e-03,  3.2265e-04,  2.8016e-04,  1.9961e-03,
        -1.5841e-03, -4.8883e-04, -4.6281e-04,  4.0692e-04,  3.0495e-03,
        -1.3266e-02, -5.6323e-03, -9.3506e-04, -2.5444e-03, -8.2675e-03,
         3.3371e-03,  6.9592e-04, -6.6912e-03, -1.4819e-03,  8.5371e-04,
        -4.0008e-03,  4.6675e-04,  4.9267e-04, -9.0020e-04,  1.5613e-03,
         1.0849e-03, -8.9801e-04, -3.6158e-03, -1.8573e-04,  1.8453e-03,
         1.9469e-03, -1.0170e-03,  3.4449e-03,  3.9548e-03,  1.3763e-02,
        -1.5955e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2945e-03,  9.1404e-01,  1.0351e-02, -3.8625e-03,  2.8753e-03,
        -5.2409e-03, -1.1722e-03,  3.5995e-03, -2.6656e-04, -3.3129e-03,
        -9.6817e-04, -9.1329e-03,  6.0630e-03,  7.1281e-04,  4.8239e-03,
        -4.5731e-03,  1.1971e-03,  2.3089e-04,  4.6024e-04, -3.0671e-03,
        -1.1194e-03, -1.0549e-02,  6.8041e-04,  1.1869e-03,  9.2210e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0482e-02,  8.9354e-01,  2.0391e-02, -1.8254e-02,  3.0988e-03,
         5.5282e-03,  6.8615e-03,  8.5111e-05,  2.0512e-03,  8.1022e-03,
         6.8057e-03,  4.5219e-03,  2.6004e-05,  3.0514e-03,  2.5390e-03,
        -4.4447e-04, -5.0434e-03,  1.4774e-03, -3.4558e-03,  2.8618e-04,
        -3.9593e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2364e-02,  8.7054e-01, -8.4648e-03, -7.3050e-03,  4.9177e-04,
        -1.4356e-03,  3.0944e-03, -7.2305e-03, -2.5441e-03,  1.1951e-02,
         9.9539e-04, -1.6474e-02, -5.5263e-04, -5.5430e-03, -8.9799e-03,
        -1.0915e-03, -8.9572e-03, -1.1990e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4732e-02,  8.7613e-01,  1.3160e-02, -5.1182e-04,  1.5409e-04,
         5.1456e-03,  3.7773e-03, -6.3310e-03, -5.4916e-03, -3.2900e-03,
         3.4612e-03, -3.5407e-03, -3.4450e-03,  2.1871e-03,  1.5050e-03,
         8.7825e-03, -5.8298e-03,  2.1867e-03,  2.9257e-03, -2.6974e-04,
        -3.5903e-04,  9.0817e-04,  8.9107e-04,  7.2598e-04, -1.8334e-03,
         4.4784e-03,  1.8546e-03, -2.6565e-03, -4.4091e-03,  4.6442e-04,
         1.2439e-03, -2.6271e-04,  1.7967e-03,  6.1655e-04,  7.2269e-03,
         4.5079e-03, -9.7380e-04,  1.9387e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8408e-02,  8.2874e-01, -3.1922e-02, -4.6332e-03, -6.3742e-04,
         1.3087e-03, -9.4578e-03, -1.0800e-02, -1.2246e-03, -1.9823e-03,
        -3.2233e-03, -3.7341e-03, -3.8855e-03, -1.2879e-02, -4.2797e-03,
        -2.6758e-03,  6.9748e-04, -3.1674e-02, -6.6768e-03,  1.1162e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0163,  0.8880, -0.0072,  0.0248,  0.0035, -0.0045,  0.0050,  0.0074,
         0.0063,  0.0039,  0.0107, -0.0040, -0.0071,  0.0112,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.4288e-02,  7.9867e-01, -1.3450e-02, -4.4059e-03,  3.9017e-04,
         2.8975e-04, -3.0195e-03, -3.1478e-03, -3.8103e-03,  4.2203e-03,
        -5.6613e-03,  1.1820e-03, -1.4307e-03, -1.4897e-03, -1.7390e-03,
        -6.0287e-03,  1.5536e-03,  8.6534e-04,  5.3389e-03,  5.3405e-03,
        -1.9009e-03,  1.4685e-02,  2.0207e-03,  5.6779e-03, -5.4975e-04,
        -6.1060e-03,  1.0488e-03,  9.5576e-04, -3.1065e-03, -1.1199e-02,
         1.8854e-03, -4.5429e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1764e-02, -8.9398e-01,  6.6038e-03, -3.6174e-04,  1.8864e-02,
        -2.4468e-03, -8.7710e-04,  1.8055e-02,  1.2227e-03,  4.1224e-03,
         8.6487e-03,  9.5896e-03,  3.4397e-03,  4.4048e-03, -5.2590e-03,
         7.6954e-04, -1.1069e-03, -5.4602e-03, -1.6324e-04,  2.8629e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1152e-02,  8.4997e-01,  2.0412e-02, -7.5671e-03,  3.8427e-03,
         7.7385e-03, -5.8457e-04, -3.4429e-03,  1.4321e-03,  2.2873e-03,
        -4.0252e-03, -8.2059e-04, -2.8972e-03, -2.4422e-03, -1.2214e-03,
        -3.1078e-03, -3.5048e-03, -2.2113e-03, -2.9669e-03, -1.9114e-04,
         1.3079e-03, -7.8647e-03, -4.9374e-03,  4.5151e-03,  4.5200e-04,
        -2.7913e-03, -2.9181e-03,  7.0073e-04, -1.1122e-03,  2.3260e-04,
        -1.8708e-03,  6.9976e-04,  1.3751e-03,  4.8008e-03,  1.6622e-03,
         1.8947e-03,  3.4158e-04, -5.2077e-03,  7.1093e-04,  3.8158e-04,
         3.3281e-03, -1.0274e-03, -1.8057e-03, -4.6064e-04,  2.9519e-03,
        -5.9238e-04,  2.2388e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2713e-03,  8.9001e-01, -1.0814e-02, -1.1996e-02,  3.5352e-02,
        -1.6101e-03, -4.1160e-03, -1.9116e-04, -6.9292e-03, -3.7396e-03,
        -9.4265e-03, -3.2606e-03,  4.2400e-03,  1.4074e-02,  1.9701e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0115e-02,  9.0752e-01,  5.6748e-03,  6.7026e-03,  4.1485e-04,
        -2.2660e-04,  1.8704e-03, -4.3788e-03,  2.3460e-03, -1.8891e-03,
        -3.1740e-03, -1.2711e-03, -3.2277e-03,  2.5552e-03,  7.2004e-04,
         4.5741e-04, -1.8410e-03, -1.7325e-03,  1.6009e-03, -5.3390e-04,
        -2.1423e-04, -3.1230e-03, -1.5556e-04,  3.0366e-03,  1.3183e-03,
        -1.0339e-03,  3.5757e-03,  1.7790e-03, -6.9135e-04,  1.8191e-03,
        -6.1045e-04, -1.1045e-03, -9.1469e-04,  1.1547e-03,  2.6785e-03,
         4.0269e-03, -9.5688e-05,  5.3741e-04, -4.4662e-03,  1.5937e-03,
         1.9122e-03,  8.4674e-04, -1.2360e-03, -7.1660e-04, -4.5051e-04,
        -2.8041e-04, -1.3744e-04,  2.2348e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-2.8044e-03,  8.7307e-01,  1.0612e-03, -1.8071e-03,  9.1574e-03,
        -1.1404e-02,  1.6263e-03, -8.2974e-03, -7.8519e-03, -1.0909e-02,
        -9.9619e-04, -1.3002e-02, -4.4374e-03,  3.3950e-04, -2.1566e-03,
        -4.1998e-03, -9.7755e-03, -7.4762e-03, -1.3293e-03, -9.8764e-04,
         5.8471e-04, -2.6950e-03, -1.2265e-04, -4.2831e-03,  1.0462e-02,
         9.1636e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1480e-02,  9.3479e-01,  1.2222e-02,  4.4218e-04, -1.8222e-03,
        -8.7581e-04, -1.1061e-02, -1.3241e-04,  1.6642e-03,  9.4034e-04,
         2.5566e-03, -5.4450e-03, -1.4564e-03, -8.9788e-05, -1.1919e-03,
         5.9464e-04, -1.4527e-03, -6.8430e-03,  3.3823e-04,  1.4422e-03,
         3.1565e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6316e-03,  8.7066e-01, -7.6258e-03, -8.6638e-03, -1.2811e-02,
        -4.6584e-03, -2.1862e-03, -3.1011e-03, -2.2007e-03,  4.6986e-04,
        -8.7576e-04, -3.8736e-04, -2.6095e-03, -2.4185e-03,  5.7371e-03,
        -6.1670e-04, -1.2194e-03, -1.2696e-03,  6.7217e-04, -4.5705e-03,
        -7.6808e-04, -5.4646e-03, -7.7166e-03,  2.7423e-03,  6.2390e-03,
        -8.6636e-03, -2.3401e-03, -3.4719e-03, -3.6378e-04,  2.9944e-04,
        -4.8594e-03, -1.8111e-03,  5.8088e-03, -1.6406e-03, -4.6666e-03,
        -5.7571e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0120,  0.8732, -0.0165, -0.0015, -0.0046,  0.0013, -0.0011,  0.0028,
        -0.0060,  0.0074,  0.0073,  0.0024, -0.0029, -0.0101, -0.0067,  0.0050,
         0.0041, -0.0075, -0.0081,  0.0195,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9863e-02,  8.9581e-01, -6.6999e-05,  6.0163e-03, -8.4470e-03,
        -3.9243e-03, -2.0065e-03, -5.5698e-03,  2.0976e-02,  2.2844e-03,
        -8.4091e-04, -4.0373e-04, -4.7054e-03,  1.5970e-03,  1.6512e-02,
        -7.6838e-04, -1.1721e-03, -4.8545e-03, -5.0452e-04, -2.5304e-03,
        -1.1413e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3891e-02, -9.3127e-01, -8.1959e-03,  2.8471e-04, -9.9020e-04,
        -2.0337e-03, -1.8488e-04,  1.9759e-03,  5.4461e-03, -1.2220e-03,
        -7.5175e-03, -2.7475e-03,  4.2385e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9368e-05,  9.2365e-01,  1.6904e-02,  1.1088e-03, -1.8010e-03,
        -3.9153e-04,  4.4968e-03, -6.2374e-04,  4.7427e-03, -2.4068e-03,
        -2.2095e-03, -5.5604e-03, -1.2207e-03, -2.8840e-03,  1.6533e-04,
         4.5140e-04, -4.4391e-04, -1.3529e-03, -1.5656e-03, -1.7919e-03,
         8.7194e-04, -6.0732e-04, -1.0297e-03, -9.7584e-05,  1.5879e-03,
        -1.9721e-03, -5.3124e-03,  1.6035e-04,  9.0005e-04,  5.4076e-04,
         1.5254e-03,  1.7714e-03,  3.8504e-04,  1.3473e-03,  2.1017e-03,
         2.6463e-03,  3.3378e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5476e-02,  8.7116e-01,  2.0103e-02,  4.2458e-03, -1.3553e-03,
        -5.4584e-03, -7.4065e-03,  1.2877e-03, -1.4647e-03, -1.9206e-04,
        -1.3216e-03,  6.2069e-03, -2.9223e-04,  4.7018e-03, -1.7249e-03,
         3.1989e-03, -4.3913e-03,  1.0500e-04, -3.2260e-04,  2.6353e-03,
         2.9588e-03,  8.4750e-04, -1.1249e-03, -3.2018e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8607e-02,  8.5513e-01,  1.8752e-02,  7.9825e-03,  1.4000e-03,
         1.0734e-03,  1.4451e-03,  4.2109e-03, -2.8528e-04, -2.8003e-03,
        -4.1399e-03,  1.6697e-02,  5.3909e-03, -1.5590e-03,  4.6477e-03,
        -1.7046e-03,  2.7739e-03,  1.9184e-03,  1.9833e-04,  1.7124e-03,
        -1.0497e-03,  3.2039e-03, -2.6785e-04, -1.3717e-04,  5.5263e-04,
         1.2273e-03, -2.3993e-03,  6.2175e-06, -1.8324e-03, -3.4000e-03,
        -2.7487e-03, -1.0749e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5399e-02,  8.6491e-01,  1.0684e-02, -3.1960e-03,  1.0326e-02,
        -1.1341e-02,  6.0332e-04,  4.5140e-03,  5.6971e-03,  8.5616e-03,
        -6.6398e-04, -1.9418e-03,  1.1798e-04, -2.7909e-03, -8.7077e-04,
         2.3107e-03, -1.1329e-03,  4.9220e-04,  3.7979e-03,  4.4148e-03,
        -2.0629e-03, -1.8550e-03,  2.0052e-03,  5.3872e-03,  1.2686e-03,
         6.6674e-03,  8.4709e-03, -1.2544e-03,  4.0171e-03,  2.9839e-03,
        -2.6217e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6603e-02, -8.5545e-01, -2.6620e-02, -4.0968e-03, -4.6721e-03,
        -1.1452e-03, -2.0949e-03,  1.8883e-02,  1.9981e-03,  2.5432e-03,
        -4.4275e-03, -3.6468e-03,  5.9409e-03,  1.1728e-02, -3.3330e-03,
         3.8637e-03, -2.5981e-03, -6.3566e-03, -1.9463e-04,  4.0940e-03,
        -7.5213e-04,  3.0804e-03,  5.8751e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3840e-02,  8.6021e-01, -9.3085e-03, -8.5597e-03, -9.0078e-03,
        -5.3832e-03, -3.7244e-04, -7.3460e-03,  2.4718e-03,  3.1672e-03,
        -3.0068e-04, -9.5593e-03, -2.9598e-03, -4.1291e-03,  8.0525e-03,
         2.0691e-03,  4.0834e-04, -3.0092e-03,  2.3164e-03,  1.3082e-03,
        -1.8197e-03,  2.6988e-03,  8.8957e-03, -4.8808e-04, -3.5061e-03,
        -2.6182e-03,  1.7051e-02,  9.1406e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-2.2218e-02,  8.8979e-01,  1.4018e-02, -3.7148e-03, -5.4606e-03,
        -9.3477e-04,  3.5976e-03, -4.5993e-03, -3.5752e-03,  1.3125e-03,
         5.3084e-03,  7.2222e-04, -1.1285e-03,  1.5209e-03,  1.3630e-03,
         3.7666e-03, -7.3847e-04,  7.4643e-03, -2.0773e-03, -1.2862e-03,
         1.7423e-03,  6.6718e-04, -1.5750e-03,  3.0322e-03,  3.3883e-03,
         8.4367e-04,  5.2040e-03,  3.8301e-03, -5.1188e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.7921e-03,  9.0058e-01,  2.2393e-03, -6.1934e-03,  5.1021e-03,
         1.2722e-02,  6.8336e-03,  7.8719e-05, -2.7161e-04,  4.4055e-03,
         1.8192e-03,  1.3613e-03, -7.2413e-03,  5.0825e-04,  1.8344e-03,
         1.0802e-03,  4.1966e-03, -6.6096e-03,  8.8483e-04,  3.7666e-03,
         1.2833e-03,  9.0681e-03, -1.2296e-02, -8.3209e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3008e-02, -8.2432e-01,  9.4719e-03, -1.3471e-02, -3.1325e-02,
        -5.1231e-03, -1.4847e-03, -5.1942e-03, -3.7655e-03, -4.8810e-03,
        -3.9920e-03, -8.3404e-03, -1.4007e-02,  8.3802e-04, -1.3261e-03,
        -8.9316e-04, -5.3690e-04, -1.0077e-03, -9.0063e-03, -1.8599e-03,
        -9.6161e-04,  1.5881e-03, -9.1739e-04, -1.9914e-03,  3.2268e-04,
         1.6776e-03,  7.2747e-03, -2.7258e-07, -1.0137e-03, -1.2788e-03,
        -1.2481e-03, -5.0057e-03,  7.9078e-03, -4.9602e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.5536e-03, -9.3780e-01,  8.4781e-03,  1.6199e-03, -1.1194e-03,
         3.5764e-03,  5.7727e-04,  1.4988e-03,  1.3007e-03,  3.9168e-03,
        -2.9586e-03, -1.8448e-04,  2.8663e-04,  2.8989e-03,  1.7589e-03,
        -6.6532e-03, -1.5141e-03, -9.5972e-04,  1.1534e-05, -7.8929e-03,
         4.7829e-04,  7.9603e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8352e-02,  9.0892e-01,  1.6593e-02,  1.2224e-03,  7.6436e-03,
        -4.2346e-03,  7.1129e-04,  4.1569e-04, -6.0251e-04,  8.1591e-03,
         7.4288e-04, -9.1059e-04, -7.9255e-04,  1.8818e-03,  5.5252e-03,
        -2.6095e-03,  8.2371e-03,  2.4417e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4893e-02,  8.8980e-01, -1.3362e-02, -3.6794e-03,  3.1062e-03,
        -8.9276e-04,  1.5242e-03,  1.2978e-03,  2.8954e-04,  4.9023e-03,
        -1.6282e-03, -3.4162e-03, -7.9564e-03,  5.2831e-03,  9.2320e-04,
         4.4653e-03,  4.4503e-03, -1.2123e-02, -6.1174e-03, -4.0586e-03,
        -5.5019e-04,  4.1864e-03,  1.0932e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.2171e-03, -9.3704e-01,  2.9474e-04,  5.3368e-03,  1.2774e-02,
         1.6693e-03, -1.9585e-03,  6.1994e-03,  6.2314e-04,  2.1501e-03,
         3.6122e-03, -1.0234e-03,  7.9495e-04, -2.1125e-04,  5.4191e-03,
         1.1364e-02,  2.3127e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1735e-02,  8.7852e-01,  3.7773e-02,  4.3830e-03,  2.1319e-03,
         2.9651e-03, -5.9448e-03, -1.9511e-03, -7.0242e-04, -3.3938e-03,
         4.5851e-03, -1.9189e-03, -3.7627e-03,  1.0476e-03,  3.4390e-03,
         6.3078e-03,  2.4554e-03, -1.6985e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7198e-02,  8.7846e-01, -1.7901e-03, -8.1440e-03,  1.6297e-04,
         3.5935e-03, -4.4055e-03, -1.3438e-03, -4.1601e-03,  1.5199e-02,
         8.0085e-04,  1.1160e-03,  6.9298e-03, -1.3657e-03, -2.4006e-03,
        -2.0695e-03,  6.1855e-03, -3.8513e-03,  4.5173e-03, -4.9231e-03,
        -3.2841e-03, -6.3127e-04, -2.5826e-03, -9.9178e-05,  2.0958e-03,
        -1.2694e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2482e-02,  8.6210e-01,  3.1596e-02, -3.9469e-03,  1.1549e-03,
        -1.7496e-04, -4.5650e-03,  1.2423e-02,  1.4535e-03, -1.2375e-02,
        -5.8108e-03, -6.6552e-03, -4.6083e-03, -2.4489e-03, -4.3513e-03,
         7.4299e-03,  5.4204e-04,  1.7312e-03,  1.6668e-03, -1.9402e-03,
        -2.3466e-03,  7.9068e-04, -2.3396e-03, -1.3414e-03, -2.2004e-03,
        -2.1847e-03, -1.5437e-03,  1.1234e-03, -2.6744e-03,  2.2670e-03,
        -1.7283e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.0803e-03, -9.1686e-01, -8.6942e-04, -3.9418e-03,  7.6342e-03,
        -1.9702e-03,  1.4018e-03, -8.5492e-03, -1.6429e-03, -1.2006e-02,
        -1.3191e-03,  2.1030e-03, -6.3610e-03,  4.9923e-03, -1.5216e-03,
        -2.3223e-04, -1.6872e-02, -2.6461e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7020e-02,  9.0811e-01, -2.2621e-02,  3.7853e-03, -2.1084e-03,
        -1.4184e-02, -2.9257e-04,  1.1874e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([-6.8266e-03,  8.8852e-01, -6.1522e-03, -7.7447e-03, -9.0058e-03,
        -1.4090e-03, -5.7433e-03, -2.4290e-03,  3.5628e-04, -1.7318e-03,
         1.3789e-03, -4.6150e-03,  1.0428e-03, -7.1714e-03, -3.4053e-04,
         3.2640e-03,  3.8951e-03, -4.4627e-03, -5.5055e-04,  4.0801e-03,
         6.1421e-03,  5.8805e-03, -2.7325e-03, -4.6589e-04, -6.8081e-04,
        -6.9595e-04, -1.0847e-03, -7.3994e-04,  3.9267e-03, -9.6428e-04,
         6.1031e-04, -5.1683e-04,  2.2662e-04, -1.6327e-03,  1.7365e-03,
         2.1827e-04,  1.2721e-03,  6.6343e-03, -3.1198e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.1089e-03,  8.9957e-01, -2.1017e-02,  5.7113e-03, -3.9898e-03,
        -2.6187e-03, -1.8724e-03, -3.0097e-03,  3.5540e-03,  1.4505e-04,
        -7.9680e-04,  4.0417e-03,  3.2201e-03, -3.0156e-03,  2.6498e-03,
         5.0108e-04,  2.0051e-03, -1.6278e-03,  5.0465e-04,  2.1088e-03,
        -3.6674e-03, -4.6234e-03,  3.1959e-03, -4.0124e-04,  1.5522e-03,
         1.2957e-03,  6.3359e-03, -7.8609e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9050e-02, -9.0042e-01,  1.3044e-02, -2.1774e-02, -8.8991e-04,
        -2.6962e-03,  1.0682e-02,  5.4796e-03, -1.2490e-02, -1.3277e-03,
         4.4194e-04,  4.9010e-03,  6.8042e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0695e-02, -9.3288e-01, -1.5652e-04,  2.3920e-03,  9.3826e-03,
         6.0955e-03, -5.5905e-03,  3.3872e-03,  1.7288e-03, -1.1555e-03,
         1.1211e-03,  9.5301e-03, -5.2869e-04,  7.8986e-03,  7.4616e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0770e-02, -8.9323e-01, -1.2095e-02,  2.2690e-03, -1.4211e-03,
         3.4441e-03, -5.4769e-03, -2.6047e-03, -2.0423e-03,  2.9487e-03,
        -4.2097e-03, -2.1435e-03,  7.9385e-03,  1.7833e-03,  2.3820e-03,
         3.5824e-03, -9.4814e-06, -1.0315e-03,  2.2737e-03,  2.6023e-04,
        -8.5781e-04, -2.6037e-04,  3.3451e-05,  3.8600e-03, -9.8949e-04,
        -1.2129e-03,  5.6369e-04, -3.2324e-03, -1.4263e-03,  1.1251e-03,
         3.3439e-03, -8.5229e-04,  2.1483e-03,  3.5748e-04, -5.1296e-03,
        -4.2690e-04, -2.4400e-04, -6.5149e-04,  2.0046e-03,  1.4129e-03,
         4.0200e-04,  1.8527e-03, -1.3038e-03, -4.8132e-04, -1.4965e-03,
         1.4683e-03,  9.4657e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.4452e-03, -9.0676e-01,  5.0925e-03, -1.3696e-02,  6.0929e-03,
        -2.7532e-04, -3.3321e-03,  4.9402e-03,  6.9057e-03, -1.2895e-03,
         4.4302e-03,  7.0479e-04, -3.1002e-04,  6.2954e-04,  6.2375e-04,
         2.1342e-04, -3.3475e-03, -2.4510e-03, -9.2854e-04, -9.8523e-05,
        -1.1896e-03, -4.7958e-04, -2.3765e-03, -5.7451e-04, -1.1181e-03,
        -1.5528e-03, -1.0298e-02,  5.5091e-04,  9.5224e-05,  1.4436e-03,
        -9.2846e-04,  7.4047e-04, -1.4963e-03, -2.0823e-03, -2.0314e-03,
         1.6119e-04,  3.1383e-04,  3.5947e-04, -2.8157e-04,  1.1222e-04,
         1.7275e-04,  7.5718e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2833e-02,  8.5853e-01, -3.4784e-03, -2.0725e-02, -7.9954e-04,
         3.7759e-03,  1.5480e-02, -1.8483e-03,  6.8258e-04, -1.4294e-03,
         4.2109e-05,  7.9015e-03, -5.1646e-03,  6.3581e-03,  5.7518e-03,
        -6.5951e-03,  8.8226e-03, -2.5534e-04, -3.6736e-03,  3.9181e-03,
         9.1972e-04, -1.8148e-03, -7.1299e-03, -3.0329e-03, -9.0344e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7677e-02,  8.9911e-01,  1.5215e-02, -4.5894e-03,  2.9678e-03,
        -2.0173e-03,  6.9585e-03, -1.8716e-03,  4.9257e-03, -7.3800e-04,
         3.0484e-03,  2.2128e-03,  4.7211e-03, -3.6561e-03,  1.0570e-03,
         2.0818e-03,  4.0465e-03, -6.9596e-04,  1.9708e-03,  5.2816e-03,
         5.6914e-03,  1.0390e-03, -8.4233e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5132e-03,  9.1148e-01, -2.6872e-03, -2.6295e-03,  3.5517e-04,
        -1.0762e-03, -1.5482e-03, -2.4931e-03,  8.3751e-04,  1.2664e-02,
         2.0889e-03,  1.6607e-03,  1.2365e-02,  6.5436e-03, -6.0676e-03,
        -2.5397e-03,  4.7519e-03,  2.5885e-03, -1.1182e-03, -1.3581e-03,
         2.9781e-03, -1.1765e-02,  5.8853e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4204e-01,  7.9982e-01,  1.6767e-05, -1.5795e-02,  3.4427e-04,
        -9.3298e-04,  1.6168e-02, -6.8520e-03,  6.6803e-03,  2.0649e-04,
        -2.6110e-03,  8.6486e-04,  7.6714e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0822e-03,  9.3111e-01,  3.9148e-03, -7.4399e-03, -7.7488e-05,
         1.0440e-02, -2.2280e-03,  3.7600e-03, -1.4965e-03, -9.7493e-03,
         2.0202e-03, -7.6434e-03,  1.1627e-02, -4.4081e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0533e-02, -9.1467e-01, -1.2524e-02,  9.7119e-05, -3.3287e-03,
         6.0014e-04, -8.6753e-04, -1.5323e-03,  3.9744e-04,  2.6169e-03,
         4.1718e-03, -2.8497e-03, -3.9357e-03, -2.9825e-03, -3.3912e-04,
        -1.5942e-03,  6.6439e-03, -2.2850e-03, -1.8030e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 4.8072e-03,  9.1467e-01,  8.8127e-03, -8.6921e-03, -1.3092e-03,
         5.4052e-03,  4.2933e-03,  1.4215e-03, -6.4053e-05,  5.0551e-03,
         8.5806e-05,  3.4899e-04,  5.0035e-04,  6.3634e-03,  3.4236e-03,
        -1.1110e-02,  5.7778e-03,  2.2989e-03,  5.8189e-03,  6.8067e-04,
         8.7341e-04, -2.1986e-04, -2.9492e-04,  2.9200e-03, -1.3944e-04,
        -4.6150e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3764e-04,  9.3207e-01, -7.2347e-03, -1.3489e-02,  2.5642e-03,
        -7.7395e-04, -3.1922e-03, -1.9798e-02, -1.8127e-03,  4.0125e-04,
        -3.1493e-04, -1.0435e-03,  1.0578e-05,  5.9303e-03, -9.7552e-03,
        -6.7291e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9904e-03,  8.6736e-01,  4.0830e-03, -2.1525e-03, -1.4332e-03,
        -9.1772e-03, -2.9831e-03, -1.1804e-02,  1.4299e-02, -3.2083e-03,
         4.9443e-03, -8.2309e-03, -4.1177e-04,  9.5516e-03, -9.6802e-03,
        -1.5449e-02,  3.0242e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6914e-02, -9.0282e-01, -1.4406e-02, -5.1136e-04, -2.7203e-02,
        -4.2521e-03,  2.4582e-03, -3.2595e-03, -8.1799e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0344e-02, -7.9057e-01,  9.7814e-03,  8.9791e-03,  1.4985e-02,
        -4.2949e-03, -3.6105e-03, -3.2146e-03,  2.4453e-05,  2.1924e-03,
        -3.9229e-03, -1.0345e-03,  1.8432e-05,  1.5887e-03, -2.8248e-03,
         2.6188e-03, -4.3308e-03, -7.0768e-03,  9.6843e-04, -1.1981e-03,
        -2.1379e-03, -4.1308e-03, -2.5887e-03,  2.2162e-03, -4.3122e-03,
        -2.1424e-03,  6.2021e-03, -1.6503e-04,  3.8728e-03,  1.3586e-03,
        -1.9635e-03,  9.5153e-04, -2.9024e-03,  1.3446e-03,  5.8722e-04,
         5.3249e-03,  6.5403e-03, -5.8290e-03,  5.1961e-03,  2.5970e-03,
         2.0019e-03,  7.9103e-03,  6.4055e-03, -3.0375e-03, -2.3654e-03,
         1.2334e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2132e-02,  8.8731e-01,  1.9250e-02, -3.0147e-03,  8.1271e-03,
         7.4115e-04, -1.0170e-03,  3.2542e-03, -3.3265e-04,  3.6654e-03,
        -7.0745e-04, -1.8535e-03, -1.5600e-02, -1.0431e-04,  1.2547e-03,
        -5.2063e-03,  8.5180e-04, -4.5964e-03, -2.4915e-03,  1.2794e-03,
         1.3759e-03, -1.5280e-03, -1.6910e-03, -3.4771e-04,  6.9357e-04,
        -2.1340e-03, -4.9561e-03, -7.7818e-04,  1.8868e-03, -1.1822e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0020,  0.9383,  0.0146,  0.0055,  0.0109, -0.0195, -0.0091,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0114,  0.9512, -0.0027, -0.0027,  0.0097, -0.0016,  0.0039,  0.0034,
        -0.0014, -0.0071, -0.0048,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0076,  0.9243,  0.0075,  0.0012, -0.0018,  0.0076, -0.0029,  0.0060,
         0.0058,  0.0034, -0.0036,  0.0019,  0.0032, -0.0231,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.1126e-03, -8.9317e-01, -2.1655e-03,  5.4064e-03, -2.4935e-03,
        -2.0309e-03, -1.1593e-03,  2.6729e-03,  2.6655e-03, -1.3333e-03,
        -1.0335e-03, -1.2177e-03,  2.1816e-03,  3.7253e-03,  9.5703e-04,
        -1.7542e-03, -1.7117e-04, -1.2130e-03, -3.2121e-03, -1.6178e-02,
        -2.9241e-03, -7.5427e-03,  2.5923e-03,  1.2500e-03,  2.5748e-03,
        -1.6584e-03, -5.3444e-03, -7.1163e-04, -6.3918e-04,  5.0552e-03,
         1.6896e-03,  4.9533e-03, -7.0305e-03,  4.1795e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9999e-02, -8.6558e-01, -9.4737e-03, -1.3652e-02, -2.1566e-03,
         3.0662e-03, -5.2683e-03,  5.1763e-03, -1.2359e-03, -1.3755e-03,
        -3.0677e-04,  2.3071e-03, -1.0885e-03, -5.8152e-04,  1.2398e-02,
         4.5138e-03, -4.3498e-03, -2.7285e-03,  2.9755e-03,  2.4599e-03,
         7.3548e-05, -2.8639e-03,  2.8909e-04,  2.1389e-03,  1.4554e-03,
         1.4300e-03, -7.1174e-03,  3.9389e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9655e-02,  9.1935e-01,  7.2323e-03, -1.4383e-03,  1.0135e-02,
        -3.4534e-04, -3.2128e-03,  1.4466e-03,  1.6320e-03,  2.2903e-03,
         5.8219e-04,  5.2054e-04,  4.6567e-03, -2.1724e-05, -2.8685e-03,
         4.4613e-04,  5.2671e-03, -2.7802e-03, -2.3553e-03,  2.1085e-03,
         3.8936e-04, -2.8931e-03, -2.7055e-03,  5.6687e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 3.5204e-02,  8.4607e-01,  3.3278e-02,  2.2823e-02,  1.6632e-03,
        -4.3838e-03,  5.8987e-04, -2.3067e-03,  2.2450e-03, -1.1042e-03,
        -8.8137e-03,  6.9275e-04,  2.8765e-03,  2.4100e-03,  3.8657e-03,
        -1.8197e-03, -2.7564e-03,  1.1215e-03, -1.9977e-03,  4.3467e-03,
         2.0544e-03,  3.7090e-03,  2.3487e-03,  3.9465e-04,  2.3612e-03,
         4.2731e-03,  1.5295e-03,  1.0179e-03, -8.2324e-04,  6.4132e-04,
         4.7893e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0371, -0.8665, -0.0176, -0.0181,  0.0037, -0.0078,  0.0010,  0.0055,
        -0.0227, -0.0023,  0.0095, -0.0083,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0290,  0.8843, -0.0097,  0.0067,  0.0088, -0.0041, -0.0088, -0.0045,
         0.0018, -0.0019, -0.0050, -0.0095,  0.0169, -0.0035,  0.0053,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.8474e-03,  9.1943e-01,  2.6608e-02,  7.5488e-03,  4.7423e-04,
         5.3549e-03, -7.0868e-04,  2.2688e-03, -8.7072e-05,  1.9422e-03,
         1.7310e-03, -6.7140e-04,  9.7470e-04, -7.9154e-04,  2.3867e-03,
        -1.5362e-03,  1.0531e-03, -1.7478e-03, -1.9160e-04, -3.3112e-03,
         6.3737e-04, -2.2991e-04,  1.0625e-03,  1.5898e-03,  1.3167e-03,
         8.4366e-04,  8.2582e-04, -3.6385e-03, -1.6940e-04, -6.8286e-04,
         8.7217e-04,  2.0734e-03,  3.8956e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2113e-02,  9.3969e-01, -1.5120e-03, -5.5135e-03, -1.3352e-03,
        -5.8496e-04, -3.8547e-03, -2.4148e-04,  1.4124e-03, -4.1907e-03,
        -2.0746e-03, -9.7196e-04, -7.0378e-03,  6.0414e-03, -1.9741e-03,
         1.4565e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0256,  0.9589,  0.0012, -0.0058,  0.0030,  0.0011, -0.0019,  0.0013,
        -0.0012,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0250e-04, -9.0683e-01,  5.5728e-02,  2.2019e-02,  1.5024e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.3568e-02, -8.9196e-01, -6.1221e-03,  7.7763e-06,  1.3023e-03,
         1.9203e-03,  5.4045e-03,  2.9863e-04,  6.2839e-03,  3.0941e-03,
         2.8872e-03,  4.2408e-03,  5.4720e-03,  1.1373e-03,  1.5714e-03,
         1.5858e-03,  2.9105e-03,  5.6514e-04, -2.2215e-03,  5.3957e-05,
        -5.9032e-03,  6.5948e-04,  5.2802e-05,  1.6521e-03,  9.5619e-04,
         1.9321e-04,  7.9810e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0438, -0.9004, -0.0194,  0.0022, -0.0185, -0.0039, -0.0118,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3147e-02,  8.7887e-01,  2.1387e-02,  3.3071e-03,  4.8434e-04,
        -2.3962e-03, -8.4470e-03, -5.2276e-03, -5.6532e-03, -3.1491e-03,
        -8.4057e-04,  1.1276e-03, -1.2873e-03,  4.1662e-03,  8.0332e-04,
         7.6714e-04,  1.8973e-03,  4.7362e-04, -1.9355e-03, -1.6762e-02,
        -3.9931e-03,  1.3879e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9858e-02, -9.1572e-01, -3.6670e-03,  7.0040e-03, -5.5872e-03,
        -5.7166e-03, -1.2872e-03, -6.1239e-03,  1.0422e-03,  5.3002e-03,
         5.1981e-03,  1.3719e-03, -1.2770e-03, -8.7540e-03,  2.1183e-03,
         4.1597e-03,  3.3054e-04, -1.0324e-03, -4.4563e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.2159e-03, -9.0041e-01,  7.0212e-03,  1.9579e-03,  5.9954e-03,
        -3.6748e-03, -3.4968e-03,  2.6070e-03, -1.2532e-02,  5.6830e-03,
         8.1324e-04, -1.9093e-03,  1.5644e-03, -2.7749e-03,  3.6689e-03,
         2.6578e-03, -8.7849e-05, -2.0611e-04, -7.4909e-04,  3.6805e-03,
        -2.9416e-03, -5.3891e-03, -1.0363e-03, -1.7151e-03, -2.2461e-03,
         5.7249e-04,  2.1514e-03,  9.5919e-04, -1.2320e-03,  7.5810e-03,
        -3.4656e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 4.4704e-02, -8.3373e-01,  1.1134e-02,  9.7943e-03, -2.5381e-04,
        -1.7819e-04,  1.7249e-03,  2.6548e-03,  1.9335e-03, -1.4488e-03,
         6.6138e-04, -7.8227e-04, -3.8040e-03, -9.9616e-03, -2.6240e-03,
         3.0154e-03, -2.5322e-02, -6.4634e-04,  8.8487e-03, -2.5192e-03,
         1.7135e-03,  2.3706e-03, -2.4828e-03, -6.1296e-04, -4.2864e-03,
         6.0094e-04,  2.5579e-03, -1.4510e-03,  2.3384e-03, -4.4381e-03,
        -5.6476e-03, -5.7566e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8024e-03, -9.3104e-01,  6.5816e-03,  2.3109e-03,  1.2736e-03,
         3.5792e-03,  1.3121e-03,  4.0979e-03, -1.4361e-02,  2.7292e-03,
         2.9550e-03, -8.5899e-04, -6.7921e-04,  1.8942e-03, -1.8538e-03,
         2.2071e-03,  2.5702e-04, -2.7543e-03,  3.4090e-03, -1.5783e-03,
         1.6564e-03, -7.8126e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5765e-02,  9.1482e-01, -7.4644e-03, -6.9118e-03, -2.9412e-03,
        -1.0795e-03, -2.7137e-03, -3.4926e-03,  1.9053e-04,  1.6234e-03,
        -1.0085e-03, -3.8918e-04,  1.2055e-03, -2.4682e-04,  5.1238e-04,
        -2.8586e-04, -4.4096e-03,  1.0862e-02, -1.9862e-03,  4.5320e-04,
        -1.7558e-03, -5.4914e-04,  7.4419e-04, -1.2768e-03, -1.1130e-03,
        -2.6986e-04, -1.3903e-03, -7.6377e-03, -1.3315e-03,  5.5661e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5808e-02,  8.8074e-01, -3.3347e-02, -6.7113e-03, -3.7316e-03,
         2.2262e-04,  9.7998e-03, -2.0593e-03, -5.0242e-03,  5.3261e-03,
        -5.0358e-03,  5.7223e-03, -3.2424e-03, -3.4372e-03, -1.1567e-03,
        -7.2018e-03,  5.5126e-03,  5.9190e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.0314e-03,  8.7275e-01,  3.3299e-02,  1.3924e-03, -3.2256e-03,
        -9.6496e-03, -2.3444e-03, -2.1474e-03, -5.1374e-03,  1.7303e-03,
         1.9855e-04,  9.1304e-04,  1.1242e-03, -1.4731e-03, -2.7897e-03,
        -5.6582e-03,  1.1410e-03, -2.1793e-03, -2.7986e-03, -1.6426e-03,
        -1.4422e-03, -5.2868e-03,  3.8659e-03,  6.0084e-04,  8.7272e-04,
         6.8488e-04,  2.2542e-03,  7.4496e-04, -3.1243e-03,  7.9303e-04,
         7.8065e-04, -2.3743e-03,  1.8953e-03,  1.3341e-03, -1.0739e-03,
        -9.1821e-03,  3.0645e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9009e-02,  9.0450e-01,  3.6902e-04, -3.8005e-04, -2.9049e-03,
         4.9192e-03, -1.0099e-04,  6.4800e-03,  1.3660e-03,  5.5761e-04,
         2.6764e-03,  1.0424e-02, -1.0699e-02,  2.6630e-03,  9.8956e-03,
         5.4575e-04,  7.6506e-05,  3.0224e-03, -2.8756e-04,  1.2252e-04,
         2.0831e-03, -9.6457e-04,  7.4488e-04,  5.2107e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8991e-02, -8.7855e-01, -2.5306e-03,  2.8600e-03,  1.4079e-03,
         2.5488e-03,  3.7572e-03, -1.0413e-03,  1.5093e-03,  1.9713e-03,
         3.9822e-03,  1.9884e-03,  4.8423e-03, -6.9753e-04,  1.1058e-03,
        -2.2566e-04,  1.9816e-03,  6.1808e-04,  1.5884e-03,  1.3800e-03,
         1.5294e-03,  4.2507e-03,  2.2723e-03, -1.8398e-03,  4.0151e-04,
        -6.0920e-03, -2.0693e-05,  2.4012e-04,  5.5635e-04,  2.4666e-03,
         3.0718e-05,  5.7030e-03,  2.1257e-03, -7.4327e-04, -7.5117e-05,
        -9.0766e-04,  1.2818e-03,  7.2058e-04, -6.6558e-04, -2.0275e-03,
         2.2695e-03,  7.1792e-04,  1.1098e-05, -2.5960e-03,  1.4142e-03,
         5.4670e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1113e-03,  8.6491e-01, -1.0674e-02, -1.5770e-02, -2.8435e-02,
         2.4904e-03, -9.4147e-03,  2.3593e-04,  4.3892e-04, -3.8590e-03,
        -3.9847e-03,  4.4189e-04,  5.8424e-03,  6.3968e-04,  2.1302e-04,
        -2.4095e-03, -4.3872e-03,  1.1508e-02,  1.0869e-03,  2.9664e-03,
         9.6311e-03,  1.0804e-02,  8.7466e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0234, -0.9350,  0.0093, -0.0032, -0.0042,  0.0035, -0.0018,  0.0020,
        -0.0035, -0.0107,  0.0035,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3956e-02,  9.0780e-01,  1.2787e-02, -1.9826e-03,  4.7891e-03,
         2.9215e-03,  5.4781e-03, -2.7006e-03,  5.2064e-03, -1.1791e-04,
        -2.8739e-03,  6.3817e-05,  2.5546e-03,  6.7693e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0258,  0.8825, -0.0034, -0.0091, -0.0056, -0.0028, -0.0086, -0.0010,
        -0.0029, -0.0047, -0.0218,  0.0319,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6743e-02,  8.9047e-01,  9.2602e-03,  3.7949e-03, -2.5370e-03,
         6.5049e-04,  3.1482e-03, -5.4757e-05,  2.2808e-03,  2.5313e-03,
        -1.7689e-03, -1.7565e-03,  1.1076e-03, -7.5306e-04, -2.6621e-03,
        -2.3549e-03, -2.1381e-04, -2.9514e-03, -6.0399e-04, -1.7588e-03,
        -1.3029e-03,  6.2993e-05,  3.5962e-03,  1.0849e-03,  3.0289e-04,
         1.2401e-03,  8.9446e-03, -1.0822e-03, -1.0513e-03, -2.2598e-03,
        -9.3198e-04,  5.1077e-03,  4.8465e-04,  1.4509e-03,  3.2205e-04,
         3.6407e-04, -2.7111e-04,  1.6055e-03,  1.0763e-03, -1.4514e-03,
         1.2195e-03, -5.5492e-04,  9.9798e-04,  2.5803e-04, -6.6493e-04,
         2.9082e-04, -5.7944e-04, -7.6773e-05,  5.6195e-04, -1.6708e-03,
        -1.5770e-04, -4.7234e-04, -1.2630e-03,  3.6245e-03,  6.2087e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 8.4798e-04, -9.3802e-01, -1.7776e-03,  3.5281e-03,  2.3855e-03,
        -1.0205e-03,  1.7763e-03,  7.2904e-03,  1.7211e-03,  1.1602e-04,
         5.9863e-03, -3.1556e-03, -5.7039e-03, -2.1946e-03, -7.2709e-04,
        -2.6271e-03,  2.1388e-03, -2.6640e-03, -1.1802e-03, -1.4827e-04,
         1.1953e-02, -2.3520e-03, -6.8406e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6868e-02,  8.8444e-01,  1.3860e-02, -1.3550e-02,  6.9079e-03,
        -1.7598e-03, -5.9441e-03, -3.5532e-03, -6.3924e-03, -7.2646e-03,
        -7.6137e-03, -3.7534e-03, -8.0482e-03, -2.0104e-03,  2.6230e-04,
        -5.9130e-03, -7.9038e-03,  3.9570e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1654e-04,  8.8759e-01, -8.0989e-03, -5.9161e-03,  7.9945e-03,
         4.7756e-03, -2.3998e-02, -3.3984e-03,  6.6051e-03,  6.1933e-03,
         6.4410e-03,  1.7404e-03,  1.6204e-02, -2.4467e-03,  4.4173e-03,
         4.0887e-03,  4.6666e-03,  5.2139e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8340e-02,  8.7310e-01,  2.0279e-03,  3.0115e-04, -2.8329e-04,
        -6.0221e-03,  2.0195e-03, -1.6659e-02,  9.2031e-03,  3.8368e-03,
        -2.3261e-03,  1.3792e-02,  7.7229e-03,  9.5967e-03, -3.4772e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5140e-02,  8.7761e-01,  1.2207e-02,  3.9221e-03, -4.4470e-03,
        -1.8577e-03,  8.6165e-05,  2.8394e-03,  4.6405e-03,  3.2665e-03,
        -5.9528e-03,  9.3569e-04,  3.6877e-03,  1.2971e-03, -1.0172e-03,
         5.1177e-03,  3.2269e-03,  2.6594e-03, -2.2673e-03,  9.2648e-03,
         3.5500e-03, -1.6408e-03, -1.4921e-03, -1.0086e-03, -3.3665e-03,
        -3.3069e-03,  1.6215e-03,  9.5375e-04,  2.8823e-03,  6.0638e-03,
         6.6189e-03,  6.0476e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2256e-02, -8.9860e-01, -5.7795e-03, -5.6819e-05, -7.8552e-04,
        -2.7623e-03,  5.7888e-03,  1.2878e-04,  5.7932e-03,  3.3831e-03,
         5.1333e-03,  2.9563e-03, -1.8461e-03,  1.3307e-03,  4.9771e-04,
         8.3068e-04,  2.5000e-03,  9.1073e-04,  4.3866e-03,  1.2247e-03,
         4.5839e-04,  4.6671e-04,  2.9656e-03, -1.0615e-03,  3.0073e-04,
         6.8391e-04, -1.9089e-04, -1.0197e-02,  3.6667e-03, -5.2669e-04,
        -1.2096e-03, -9.5077e-04, -1.0366e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.8000e-03,  9.0505e-01, -1.8330e-03,  6.0858e-03,  1.5321e-03,
         7.9353e-04, -6.7013e-04, -7.0431e-03, -6.8632e-04, -6.3838e-04,
        -1.5574e-03, -6.9939e-03, -6.0881e-03, -1.2547e-03, -2.3068e-03,
         4.9009e-05, -3.2416e-03, -1.6530e-03, -3.2413e-03, -1.0239e-03,
         1.5979e-04,  2.3986e-03, -2.4177e-04, -2.5384e-03,  2.1296e-03,
        -6.1386e-04,  1.0250e-03, -1.9632e-03,  1.1644e-03,  7.7346e-04,
         1.2591e-03, -3.2276e-04, -3.9950e-04, -5.5150e-04,  5.3050e-04,
        -2.5571e-03, -2.4756e-03, -1.5257e-03, -1.5832e-04,  2.6807e-04,
        -3.7057e-04, -1.9169e-03,  5.4976e-04,  3.2411e-03,  1.0184e-04,
         1.5149e-03, -1.1477e-03, -7.4970e-04,  8.3178e-05, -4.3067e-04,
         6.6120e-04,  5.6228e-04, -1.6144e-03, -1.3806e-04,  1.5818e-03,
         1.7194e-03, -1.0777e-03,  1.9433e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1820e-02,  9.0629e-01,  3.1687e-04,  1.8971e-04, -7.1976e-03,
        -3.3157e-03, -3.9873e-03, -2.6114e-03, -1.1678e-03,  1.1176e-03,
         3.1085e-03, -1.8214e-03, -8.1866e-03, -9.2665e-03, -2.3639e-03,
        -3.8636e-03, -1.8344e-03,  9.4306e-04, -3.9727e-03,  4.1202e-03,
         1.1472e-03, -7.3524e-05, -2.6637e-03, -5.0523e-04, -7.7197e-03,
        -3.9225e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2189e-02,  8.7809e-01, -2.8480e-02, -9.0680e-03, -9.3424e-03,
         5.8878e-03, -2.0771e-03, -2.1539e-03, -4.5556e-03, -7.5235e-03,
        -3.6731e-03, -1.6329e-03, -7.0105e-04,  3.0263e-03, -2.4169e-04,
        -1.1161e-03, -5.1320e-03, -3.6238e-04, -1.7767e-03,  6.2877e-04,
         1.0126e-03, -1.9454e-03, -5.2402e-04,  5.2230e-03, -1.3638e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.1247e-03, -8.9318e-01, -9.4048e-03, -3.6962e-03,  6.8241e-03,
         3.1020e-03,  3.2441e-03,  6.7724e-04, -2.8054e-03,  5.4756e-04,
        -3.2458e-04, -3.4720e-03, -7.5695e-03, -7.1242e-03,  2.2308e-03,
         2.9873e-03,  1.2506e-04,  6.0948e-05, -6.8523e-04, -9.7047e-04,
         2.1085e-03, -6.5743e-04, -8.8737e-05,  3.2209e-03,  2.6488e-04,
         2.5556e-03,  3.7799e-04, -9.1739e-04,  6.2506e-04,  2.1907e-03,
        -8.8741e-05, -1.5140e-04, -5.1132e-03,  1.8233e-03,  5.5555e-03,
         1.4820e-03,  7.1334e-04,  2.2118e-03, -3.8633e-04, -9.5656e-04,
        -2.6660e-04,  2.0195e-03,  1.6160e-03,  2.2022e-03,  7.2498e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3998e-03,  9.2045e-01, -8.5591e-03, -3.0185e-03, -2.6357e-03,
        -1.3312e-03, -5.3629e-04, -4.9883e-03, -9.0676e-03, -2.3336e-03,
         2.3882e-03,  5.0176e-04, -1.5805e-03,  3.1447e-05,  1.9171e-03,
        -1.6177e-03,  1.9745e-03,  4.4646e-03,  6.3710e-03,  1.3914e-03,
        -4.5546e-03,  7.0638e-03,  6.8197e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0140,  0.8870, -0.0111, -0.0046,  0.0036,  0.0181,  0.0165,  0.0064,
        -0.0154, -0.0112,  0.0036,  0.0086,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 8.9561e-03,  9.0502e-01,  6.2989e-03, -1.2360e-02, -7.5611e-03,
        -2.7890e-04, -5.6827e-03, -4.2543e-03,  2.7718e-03, -5.3759e-03,
        -4.7050e-03, -2.1230e-03, -1.6607e-03, -3.3961e-03, -1.6921e-03,
        -4.1238e-03,  3.3722e-03, -6.8937e-04,  8.0228e-05, -6.7766e-03,
        -5.5796e-03, -1.2824e-03, -5.9627e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3586e-02,  8.7758e-01,  3.8432e-02,  5.1965e-04, -5.0892e-03,
        -3.4864e-03,  1.7334e-03,  2.3721e-03, -3.8822e-03,  4.7752e-03,
         1.9143e-03,  6.3480e-05,  1.2595e-05,  1.7521e-03,  2.9051e-03,
        -1.6639e-03, -3.5443e-03, -3.5740e-03,  9.4919e-05,  2.3244e-03,
         1.6694e-03, -5.3774e-04, -2.5370e-03, -1.2116e-03,  3.9319e-04,
        -2.6205e-03, -3.2360e-03,  3.6709e-04, -1.9024e-03, -1.4385e-03,
         1.5965e-03,  5.6763e-03, -7.5061e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.0865e-02,  8.6918e-01,  2.3882e-02, -1.3979e-02, -4.5761e-03,
        -3.5873e-05, -6.6913e-03, -1.9499e-03,  2.5204e-03, -1.0113e-02,
         4.8205e-03, -6.9802e-03,  4.6800e-03, -2.8628e-03, -1.5995e-04,
        -1.0200e-04, -2.4960e-03,  4.0719e-03, -3.6630e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.5428e-03,  9.3587e-01, -1.8639e-03,  9.2918e-03,  2.0644e-03,
        -3.6858e-03,  1.2048e-02, -8.0240e-04, -3.5218e-03,  1.7041e-03,
        -5.6809e-03, -3.6081e-03,  3.8346e-03,  8.4815e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.9251e-03,  9.1150e-01,  1.6398e-02, -3.4744e-03,  1.0971e-02,
         3.7585e-03,  5.4136e-03,  3.8765e-03,  7.1792e-03,  8.7446e-04,
        -8.5884e-04, -6.2976e-04,  5.2726e-03, -5.4914e-04,  6.7538e-04,
         1.6333e-03,  3.4049e-03,  2.2365e-03,  1.8597e-03,  2.7757e-04,
        -8.5235e-04,  8.3800e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3482e-02, -8.9214e-01,  1.4701e-02,  2.5676e-04, -7.9424e-03,
         4.4304e-03, -1.1475e-03,  2.4247e-03, -1.6757e-03, -4.2593e-03,
        -4.5738e-04, -4.7428e-03, -6.2722e-03, -6.3591e-04, -7.4242e-04,
         1.4215e-03, -1.5466e-04, -1.8854e-03,  1.5093e-03,  2.0907e-03,
         1.7943e-03, -1.5031e-03, -1.7317e-03,  1.2421e-03,  3.6992e-03,
        -4.6003e-04, -9.1993e-04, -1.2675e-03, -7.5693e-03, -7.4452e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2847e-02,  8.9442e-01,  3.2573e-03, -9.5492e-03, -7.1607e-03,
         1.2626e-02, -6.1148e-03, -4.5871e-03,  3.6001e-03,  2.5331e-04,
        -3.1048e-03, -3.6282e-03, -7.1265e-04, -1.3028e-03,  1.8789e-03,
        -1.2897e-02, -2.2059e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3787e-02,  8.9359e-01,  1.5677e-02, -7.7166e-03, -1.9474e-04,
         7.9733e-03, -1.8145e-03,  1.5609e-02, -2.0071e-02, -5.6729e-03,
         7.8917e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5770e-03, -9.1071e-01, -1.1841e-02,  1.0697e-02, -1.6507e-04,
        -8.5422e-03,  4.5670e-03, -1.1820e-03, -7.7651e-06,  1.2268e-03,
         4.5431e-03,  1.7558e-03,  3.5022e-03, -1.3471e-02, -1.2448e-02,
         6.2413e-03,  7.5190e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.8095e-03,  8.9720e-01,  1.0258e-02,  2.1584e-02,  5.2690e-03,
         3.5516e-03,  4.7835e-03, -1.4596e-03,  3.1139e-03,  2.1519e-04,
        -1.2772e-03, -3.6152e-04,  6.1835e-04,  3.8328e-03,  1.4463e-02,
         2.4160e-03,  9.7354e-04, -1.8713e-03, -3.1134e-03,  5.8834e-04,
        -3.9672e-04,  3.1345e-03,  6.7652e-03,  3.2173e-03,  4.8797e-04,
         4.2353e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7721e-02,  8.9800e-01,  1.2186e-03, -6.9467e-03, -2.5813e-03,
         7.9649e-03,  8.0295e-04,  2.3922e-04, -1.5306e-03,  1.9262e-04,
         6.3681e-06, -8.8003e-04, -6.4070e-04, -3.5402e-04, -2.2563e-04,
        -2.9104e-04, -1.8146e-03,  5.8117e-04, -6.9708e-04, -2.1774e-03,
        -1.6564e-03, -1.6935e-03, -2.2660e-03,  4.9018e-03,  2.1312e-04,
        -3.5740e-03,  4.4345e-04, -2.1418e-03, -5.8546e-03, -2.3877e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1292e-04,  9.2942e-01,  1.2106e-02, -4.3577e-03,  8.8011e-03,
        -6.5192e-03, -2.3487e-03, -4.6209e-03,  4.1813e-03,  9.4885e-04,
         2.9953e-03,  1.3241e-03,  2.2210e-04, -2.5692e-03, -1.4399e-03,
         9.9484e-04, -2.1070e-03, -1.8199e-03,  1.3109e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-1.0075e-04, -8.7447e-01,  1.4329e-02, -7.8549e-03,  7.6889e-03,
         5.8280e-03, -7.9599e-03,  6.7964e-03,  2.9460e-03, -5.3520e-03,
         1.0041e-03, -7.4385e-04,  3.5710e-03, -2.0446e-03,  5.4072e-03,
         2.4516e-03, -5.3556e-03, -8.9883e-03,  2.9539e-03, -2.2029e-03,
         7.8871e-03, -1.5947e-03,  2.6306e-03,  5.8717e-03, -3.5589e-03,
        -7.0441e-03, -3.3654e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2741e-02,  9.1487e-01,  6.0249e-03, -1.2021e-02,  1.9485e-03,
         5.8018e-03,  3.4204e-03,  3.9958e-03,  9.1328e-03, -7.4901e-04,
        -1.4141e-03,  5.0200e-03,  6.9538e-04,  2.1611e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8694e-02,  8.8201e-01,  8.2858e-03, -1.6469e-03,  5.7338e-04,
         4.7280e-03,  9.6332e-04,  1.0050e-03,  5.9082e-03,  2.0622e-03,
         6.6093e-04,  3.1888e-03, -1.6281e-03,  5.7935e-03, -1.3636e-03,
        -6.9250e-03,  1.5228e-03, -6.6823e-03,  9.0583e-03,  3.5128e-03,
        -3.9479e-03, -3.5834e-03, -3.9733e-03, -2.2754e-03,  1.6700e-03,
        -8.3417e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9099e-02, -8.8581e-01, -5.2978e-03,  5.3514e-03,  7.6442e-03,
         1.7522e-02, -3.2033e-03,  6.3539e-03,  2.6187e-03, -1.9335e-02,
        -2.3409e-04, -3.5130e-03,  1.4440e-03,  1.2579e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9833e-02,  8.8665e-01,  9.7719e-03,  1.2001e-03, -9.0277e-04,
        -1.6318e-03,  7.9713e-03, -1.8602e-03,  3.2914e-03,  3.4982e-04,
         1.8859e-03,  2.6502e-03,  2.0775e-03,  1.6100e-03,  5.5138e-03,
        -8.5273e-04, -2.9364e-03,  3.4908e-04, -1.2085e-03, -2.7769e-03,
        -2.2977e-03, -1.0559e-03, -2.6135e-03, -6.1075e-03,  6.4868e-04,
        -8.2528e-04, -1.0977e-02,  4.8531e-04,  9.9323e-04, -9.4435e-04,
         4.7043e-03, -3.0280e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3532e-02, -8.8864e-01,  1.1966e-02, -7.9384e-03,  3.8102e-03,
         2.5778e-03,  6.8291e-03,  1.0165e-02,  4.8291e-03,  2.5611e-03,
         8.2165e-04, -3.4341e-03,  3.8216e-04, -3.0017e-06, -1.8280e-03,
         4.3633e-03,  6.2454e-04,  3.5079e-03, -3.3286e-03, -1.5599e-03,
         4.5340e-04, -2.1067e-03,  9.7679e-06,  2.5518e-03,  2.1762e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4806e-02,  9.5236e-01,  1.7699e-02,  6.7969e-04,  3.3380e-03,
        -1.1118e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6171e-02,  8.9599e-01,  5.9504e-03, -9.9249e-04,  7.6163e-03,
         3.2818e-04,  3.9329e-03, -1.4563e-03, -6.4931e-04,  2.3445e-03,
        -2.2666e-03,  2.9664e-03,  1.1193e-03,  3.5442e-03,  7.0710e-05,
        -2.3292e-03,  2.5194e-03,  2.8620e-03,  1.3898e-03, -3.3714e-04,
        -3.6627e-03,  1.2434e-03, -3.0332e-03,  7.4633e-04,  1.5937e-03,
         1.2756e-03, -6.4050e-04, -9.3392e-04, -1.5602e-03,  6.6651e-04,
         1.0129e-03,  4.8867e-04, -3.6962e-04, -2.6100e-03, -1.3905e-03,
         4.3290e-04, -3.5055e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1207e-02, -8.6260e-01, -4.0615e-03, -2.3067e-02, -4.2863e-04,
        -8.5578e-04, -1.6009e-03, -5.8971e-03,  1.2605e-03, -1.0977e-03,
        -3.7657e-05, -3.6853e-03, -3.0838e-03, -2.6334e-03, -9.0279e-04,
         6.4241e-03,  2.8240e-03, -7.3271e-03, -5.2981e-04, -3.1443e-03,
         6.6393e-04,  6.8685e-04,  6.8761e-04,  6.8952e-04,  1.0448e-03,
         2.2548e-03,  5.2914e-03,  8.7116e-03,  1.7783e-03, -4.9893e-03,
         3.0302e-04, -1.1977e-03, -1.7007e-03,  2.1603e-03,  5.1691e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.7038e-03,  9.1013e-01, -3.3137e-02, -3.9522e-05, -1.9380e-03,
         6.6413e-04, -2.2964e-03,  5.0408e-04,  5.6953e-03,  1.3860e-02,
         9.5208e-03, -1.0873e-02, -1.0166e-03, -1.3261e-03,  2.9057e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5517e-02, -9.1546e-01, -6.0457e-03, -3.1060e-04, -2.8732e-03,
        -1.7522e-03,  1.9961e-03,  2.9984e-03, -6.6166e-04, -1.6365e-03,
         1.0187e-03,  2.9478e-04, -4.4118e-03, -7.7122e-04,  5.6477e-04,
        -4.9316e-03, -2.5734e-03, -1.6343e-04, -4.6395e-05, -1.2331e-03,
        -9.1929e-04, -2.6760e-03, -1.7880e-03, -2.4810e-03,  1.1290e-04,
         4.7826e-04, -1.8583e-03,  5.2800e-04, -3.3392e-03, -9.1812e-04,
        -1.2186e-03, -6.5388e-04, -7.7647e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1824e-03, -9.0067e-01, -1.4025e-02, -7.9055e-04,  1.8290e-03,
         8.6547e-03,  4.4218e-03,  1.2838e-03,  1.1848e-02,  2.2923e-03,
         2.8433e-03,  5.4776e-03,  3.9268e-03, -5.9570e-03,  1.0566e-03,
         1.9815e-03, -1.2159e-03,  8.9446e-04,  1.0517e-03,  3.4149e-04,
         2.0189e-03,  3.6099e-03,  1.7588e-03,  2.1126e-03, -3.0813e-03,
        -2.6923e-03, -1.4487e-03, -2.3680e-03,  5.1662e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 6.2321e-03,  9.3696e-01,  5.6171e-03, -4.9483e-03,  1.6860e-03,
        -6.3628e-03, -3.1192e-03,  1.8134e-03, -1.2847e-03, -5.0341e-03,
         1.4474e-04,  6.4094e-03,  3.6091e-03,  3.2926e-03, -1.3487e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4656e-03,  9.2822e-01,  2.5480e-03, -2.1078e-03, -2.5543e-03,
        -5.8064e-04,  2.9596e-03,  4.1680e-03,  5.8862e-03,  1.5624e-03,
        -1.5600e-03, -2.6342e-03, -1.0721e-03,  2.0132e-03,  1.8253e-03,
        -4.3351e-03, -1.3410e-04,  4.0239e-04,  5.2343e-06,  5.3170e-04,
         1.4520e-04, -1.1898e-03, -1.8466e-03,  3.1944e-04,  3.2843e-03,
         1.0721e-03, -1.4183e-03,  4.0514e-04, -1.2704e-03, -5.8048e-04,
         4.9127e-04, -2.2181e-03, -3.2151e-04,  4.4984e-04,  6.3562e-04,
         1.0060e-03, -4.0332e-06,  4.0749e-04,  1.6997e-03, -2.1819e-03,
        -1.8597e-04,  2.0442e-03, -9.8337e-04, -8.4267e-04,  1.3876e-04,
         1.3417e-04, -1.0470e-03, -1.1419e-03,  1.9670e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8466e-02, -8.3071e-01,  3.7348e-02, -8.1862e-03, -4.6416e-03,
        -7.6100e-03, -3.6401e-03,  5.8066e-03, -1.1162e-03, -4.2686e-03,
         2.5844e-03, -2.1136e-03, -6.6113e-04, -1.3851e-03, -2.9245e-04,
         1.3659e-03, -4.2091e-03, -1.3672e-03,  2.3658e-03, -4.0280e-03,
         1.8671e-03, -2.5626e-03,  2.7829e-05, -3.2151e-03, -1.5492e-03,
        -5.4432e-04,  1.3033e-03,  1.1038e-03, -1.7151e-03,  5.6855e-03,
        -1.4191e-03,  1.4491e-04, -2.1414e-03, -7.6372e-03,  6.9202e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2557e-02,  9.0646e-01, -9.3100e-03,  1.4621e-03,  2.4101e-03,
         7.2049e-03, -9.8688e-03,  1.8786e-03, -2.1421e-03, -2.8350e-03,
         1.5500e-03,  1.2142e-03, -3.0862e-04, -3.7730e-03,  1.3356e-03,
        -3.5804e-03,  2.9460e-03, -2.6701e-02,  2.4662e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0061,  0.9108,  0.0055,  0.0028,  0.0037,  0.0038, -0.0011, -0.0026,
         0.0033, -0.0030, -0.0127,  0.0010, -0.0137, -0.0185, -0.0112,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7006e-03,  9.4196e-01, -4.3906e-03,  3.2443e-03,  1.2534e-03,
         2.5147e-03,  8.6991e-04,  3.3258e-03,  4.9388e-04, -5.0254e-03,
        -3.3840e-04,  2.2575e-03, -4.7765e-03, -7.4068e-03, -2.4400e-03,
        -3.1712e-03,  1.2116e-03, -1.1722e-03,  2.7437e-05, -3.5861e-03,
        -8.8239e-03, -1.1903e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3272e-02, -8.8583e-01,  2.0042e-02,  3.7791e-03,  2.0802e-03,
         4.5668e-03,  3.7251e-03, -1.2970e-03,  6.9518e-03,  6.1763e-03,
         4.8270e-03,  3.8187e-05, -2.1788e-03,  5.0600e-03, -2.1963e-04,
        -2.6511e-04,  6.3948e-03, -1.3959e-03, -8.1188e-04, -1.3932e-03,
         3.2389e-03,  5.2440e-04, -5.9162e-04, -1.3211e-03, -5.0047e-03,
         4.1815e-04,  2.2444e-03,  1.1820e-03, -4.6060e-03,  5.6211e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1682e-03,  9.1274e-01, -2.5132e-02, -9.0769e-03, -6.0016e-03,
        -3.7438e-03,  3.3227e-05, -5.9910e-03,  9.4825e-03,  1.2287e-02,
        -1.1339e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1871e-02, -9.2999e-01, -1.0327e-02,  1.0391e-03, -5.1981e-04,
        -4.2789e-03, -3.8027e-04,  4.0548e-04, -5.6128e-03, -5.5632e-04,
         2.5866e-03, -3.0688e-03, -4.7658e-04, -3.0980e-03,  1.3941e-03,
        -3.2685e-03,  1.9972e-03,  1.1758e-03, -1.4020e-03,  2.1414e-03,
        -1.1524e-02,  2.8878e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9442e-03, -9.1118e-01,  3.7701e-03,  2.6792e-03,  1.6574e-03,
        -8.0480e-03, -3.7916e-03, -4.5179e-03,  6.8259e-03, -1.6659e-03,
         7.0223e-05, -3.0236e-05, -3.1680e-03,  4.4693e-03,  1.1038e-03,
         1.0914e-03, -4.3696e-03, -1.9718e-03,  7.7347e-03, -1.7326e-03,
         7.1509e-04,  2.1952e-03,  9.8727e-05,  3.8283e-04, -1.5218e-03,
         1.1487e-04,  8.5074e-04, -2.9255e-03,  2.3018e-03,  4.5517e-03,
        -4.5938e-03, -5.8032e-03,  2.1226e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7460e-02,  8.7249e-01,  2.8119e-02, -2.4976e-03,  3.3774e-03,
         6.0462e-03,  1.3578e-03, -1.2730e-02,  2.1910e-03, -5.3783e-03,
         6.8465e-04, -2.0393e-03,  2.1242e-04,  3.0753e-04,  1.8569e-03,
         1.3551e-03,  4.9869e-03, -1.0754e-03, -9.5266e-05, -2.1376e-03,
         4.7736e-03, -1.1579e-02,  6.8518e-03,  1.1614e-03,  8.7430e-04,
         6.0680e-04,  3.2237e-03,  4.5343e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2222e-03, -8.6199e-01, -4.5095e-02, -1.7818e-03, -8.8514e-04,
         4.3493e-03,  5.5368e-03, -9.9033e-03,  4.2582e-03,  9.1622e-03,
         2.4416e-03, -3.4843e-03, -7.9928e-04, -1.2640e-02, -2.6024e-03,
        -7.7232e-03, -1.6913e-03,  2.1947e-03,  1.6075e-04,  1.5720e-03,
         1.3708e-03,  1.4414e-03,  2.9429e-03,  9.8583e-03,  2.8939e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 1.6501e-02,  9.0240e-01,  1.8190e-02,  1.4670e-03,  1.5274e-04,
        -5.1224e-03, -9.2607e-06, -4.8353e-04,  1.9317e-03,  7.6793e-04,
         3.8519e-03,  6.5011e-05, -2.2778e-03,  3.7186e-03,  7.0097e-04,
        -3.1981e-03, -2.1014e-03,  9.7643e-04,  7.4567e-04, -1.7224e-05,
        -4.0030e-05, -1.0880e-03, -2.4999e-03, -1.4508e-03,  1.4092e-03,
         1.8647e-05, -1.4325e-03, -3.1276e-03, -1.8128e-04,  4.6588e-03,
        -4.3785e-03, -3.0782e-03,  1.1109e-04, -1.3761e-03,  9.3337e-04,
        -2.0115e-03, -2.5177e-03, -1.2185e-03,  7.5624e-04, -1.1292e-03,
         1.9035e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3612e-02,  9.1445e-01, -5.5802e-04, -5.4570e-04,  7.9571e-03,
         2.8748e-03, -2.7476e-03,  1.1927e-03, -1.7785e-03,  2.3186e-03,
         2.5662e-03,  5.3610e-04,  6.1204e-03,  8.1529e-06,  5.9754e-03,
        -5.0931e-04,  3.9648e-03, -2.2251e-05, -1.4178e-04, -1.1896e-02,
        -4.4628e-03,  2.0494e-03,  2.4029e-04, -1.7444e-03,  1.1732e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8361e-02, -8.8461e-01, -1.3131e-02,  1.0405e-02, -1.1388e-04,
        -7.2133e-03, -1.8380e-03, -1.0093e-02,  3.9245e-03, -1.5863e-04,
        -5.0969e-03, -3.1637e-03, -2.0257e-03, -3.0671e-03,  7.8182e-04,
        -5.2797e-04, -1.7677e-03, -8.3588e-04,  4.6959e-03, -6.6626e-03,
         1.5229e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1027e-02,  8.7401e-01,  7.2148e-03, -7.8231e-03, -5.0287e-03,
        -3.1106e-03, -5.0850e-03, -8.7866e-03, -4.0602e-03, -8.6359e-03,
        -8.9226e-03, -4.4106e-03, -4.2615e-04, -1.0149e-02, -2.7170e-03,
        -8.6425e-03, -8.0237e-03, -1.9272e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4402e-02,  9.0237e-01,  4.0880e-03,  1.8120e-03, -8.8283e-04,
         7.6095e-03,  5.2260e-03,  3.3946e-03, -4.1937e-03,  3.5632e-03,
         2.9067e-03, -1.8428e-03,  2.9889e-03,  5.9357e-03,  1.6130e-03,
         3.5543e-03, -1.1667e-03, -2.1714e-03,  4.0748e-03, -1.8674e-03,
         7.4417e-05, -6.6313e-04,  6.9902e-04,  5.1511e-04, -2.5517e-04,
        -4.1327e-04, -9.4845e-04, -1.2267e-03,  1.4921e-03,  1.8220e-04,
         1.6376e-03,  2.6583e-03, -1.4422e-03, -5.1549e-04,  5.7491e-03,
         5.2855e-03,  6.8402e-05,  5.1296e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6455e-03, -9.1175e-01,  9.4273e-03,  5.8588e-03, -8.2763e-03,
         4.3049e-04,  1.5319e-03, -6.3819e-03, -1.0456e-03, -3.5005e-03,
         2.2450e-03,  4.2998e-03, -3.7197e-04,  1.6477e-03, -6.3624e-03,
        -1.6662e-03, -4.0859e-03,  1.9694e-02,  1.4091e-03, -8.3672e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0132,  0.9060, -0.0069,  0.0080,  0.0037, -0.0028,  0.0037, -0.0015,
         0.0069,  0.0111,  0.0188,  0.0036,  0.0051, -0.0089,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0056e-02,  8.6652e-01,  3.5364e-03,  6.5537e-03,  3.4431e-03,
        -3.9839e-04,  2.9577e-03,  2.0370e-03,  3.9504e-03,  5.0597e-03,
        -1.1065e-03,  3.7347e-03,  6.7927e-04,  1.6919e-03,  4.6253e-03,
         2.3676e-03,  3.6374e-03,  2.6849e-03,  3.7368e-03,  1.3375e-03,
        -3.2178e-03,  4.1084e-03,  3.9233e-03,  2.6269e-03,  3.2492e-03,
        -3.1932e-03,  1.4891e-03,  3.5679e-03, -2.2803e-04,  1.8262e-04,
         1.6114e-03, -2.4854e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5157e-02, -8.7704e-01, -1.1403e-02, -1.3849e-02, -7.1039e-03,
        -7.3298e-03, -5.5685e-03,  2.1464e-03, -1.9449e-03,  3.8240e-03,
        -8.4208e-04,  1.2789e-02, -1.5608e-03, -6.1962e-04, -1.3536e-03,
        -8.5732e-04, -8.6179e-04,  3.8888e-03,  7.2125e-03, -1.4649e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.3707e-02,  8.4539e-01,  2.7590e-02, -9.0514e-03, -1.0941e-03,
         4.2500e-03, -1.9160e-03, -2.7766e-03,  1.0667e-03,  6.4813e-04,
         4.4650e-04,  2.0803e-03, -4.2639e-04, -1.2752e-03, -3.4353e-04,
        -4.1624e-03,  1.2044e-03, -2.6996e-03,  1.9009e-03, -1.8617e-03,
         3.1240e-03,  3.9620e-03, -6.5715e-04,  1.1198e-03,  1.8008e-03,
        -1.0798e-03, -6.3193e-03, -3.5002e-03,  7.7302e-04, -1.3053e-04,
         2.6657e-03,  5.2226e-04,  2.0962e-04,  5.6434e-03,  1.2161e-03,
         1.4343e-03, -7.6326e-04,  4.0589e-04,  1.8598e-03,  9.9997e-04,
        -2.7312e-03, -1.3296e-03, -7.9147e-04,  4.9869e-04,  6.2350e-05,
        -9.3085e-04,  1.5792e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4064e-03,  9.3010e-01,  8.9807e-04, -8.3647e-03,  2.0407e-02,
         3.7033e-03, -6.7252e-04,  2.0445e-03, -4.5374e-03, -3.1020e-03,
         1.2375e-03, -4.1234e-03,  7.9192e-04,  5.0946e-03, -1.0516e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.9817e-03,  8.9448e-01, -1.1277e-03, -2.5162e-03, -7.2304e-03,
        -2.2185e-03, -8.7796e-04, -2.8018e-03, -6.5125e-03,  3.2527e-03,
         6.7827e-04, -3.0629e-03,  2.6880e-03,  1.2216e-03, -1.4351e-03,
        -4.4416e-03,  5.6151e-03,  1.5492e-03, -1.4988e-03, -1.3852e-03,
        -1.3173e-03, -3.2405e-03, -2.1028e-03,  4.7614e-03,  8.3717e-04,
        -1.3781e-03,  3.2393e-03, -4.0792e-03,  2.9267e-03,  1.9699e-03,
        -3.2251e-04, -1.2419e-03,  3.1983e-03,  2.0415e-04,  1.3053e-03,
         2.9359e-03, -3.0985e-04, -1.0033e-03, -3.0305e-03,  3.9073e-04,
        -1.1558e-03, -9.0802e-04,  3.3892e-04, -9.7817e-04, -1.6461e-03,
        -3.4299e-03, -2.4618e-05,  1.4499e-04], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-2.9306e-03,  9.1734e-01, -1.2305e-02, -7.4973e-03,  7.5180e-03,
        -5.5690e-03,  6.8314e-03, -2.0767e-04, -1.2451e-02, -2.3649e-03,
        -1.2379e-03,  2.5884e-04, -1.1611e-03,  6.6890e-04, -9.9052e-04,
        -1.0992e-03, -7.0852e-03, -3.2059e-03, -7.6649e-04, -1.2283e-03,
         1.4106e-04,  5.5507e-04, -6.8651e-04, -3.0170e-03, -1.1650e-03,
        -1.7170e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5908e-02, -9.1186e-01,  8.9079e-04, -1.1855e-03, -1.6728e-03,
         4.2441e-04, -5.8445e-03, -6.3550e-03,  2.6460e-03,  1.6518e-03,
        -4.7389e-03,  2.4583e-03, -3.3413e-03,  3.8619e-04,  3.6549e-03,
         2.9846e-03, -1.4423e-02,  4.2021e-03, -9.7132e-03, -2.1324e-03,
        -3.5284e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2541e-03,  9.0043e-01, -4.2411e-03, -5.1888e-03, -6.3461e-03,
         8.9455e-04,  1.9762e-04, -5.8081e-03, -3.1038e-03,  3.3507e-04,
        -1.0777e-03,  2.0966e-03, -1.4724e-04, -6.8237e-03,  2.9919e-03,
        -2.6275e-03,  1.3405e-03, -9.1288e-03, -1.8454e-03, -3.5033e-03,
         1.3406e-03, -7.8306e-04, -6.2541e-03,  2.2036e-03,  1.7320e-04,
        -6.1482e-03,  8.2446e-05,  9.3638e-04,  2.9606e-03,  6.4597e-04,
        -4.3802e-03,  2.8124e-03,  2.6845e-03, -8.1312e-04, -2.8093e-04,
         5.1189e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4009e-02,  9.2344e-01, -4.6550e-03,  6.5145e-03, -1.0995e-03,
        -4.1167e-03,  3.2682e-03,  4.7865e-03,  2.9783e-04,  3.8912e-03,
         2.3861e-03,  2.7582e-03, -2.8640e-03, -1.2211e-03, -1.9854e-03,
        -2.3482e-03, -2.0032e-03, -2.1551e-03,  1.4823e-02, -1.3747e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7078e-02, -9.1396e-01, -9.6014e-03, -5.7599e-03, -2.0093e-03,
         1.2803e-03, -1.9113e-03,  6.1095e-03, -4.4822e-03,  8.7179e-04,
         3.8435e-03,  5.1328e-03, -3.1649e-03, -2.4399e-05,  1.9272e-03,
        -1.0203e-03, -5.4929e-03,  7.4972e-03,  1.4371e-03,  1.1730e-04,
        -7.2790e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5364e-02, -8.9456e-01,  1.5766e-02, -5.6516e-03,  4.9178e-03,
        -1.4745e-03, -1.0509e-03,  1.4169e-02,  1.4265e-02,  2.5606e-04,
        -1.1222e-03,  1.2355e-02, -9.0429e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8217e-02,  8.3986e-01,  3.7012e-02,  9.0718e-03, -4.8859e-03,
         2.6196e-04,  7.0156e-03, -2.7661e-03,  4.6535e-04,  2.2847e-03,
         1.4133e-03, -1.5019e-03,  1.8820e-03, -1.8755e-03,  3.5932e-03,
        -3.5128e-04, -1.8054e-03,  9.1995e-04, -1.1302e-03, -4.4897e-03,
         1.9647e-03,  2.6686e-03,  6.8968e-04, -2.1689e-04,  2.7244e-03,
         1.7783e-03, -2.7539e-03,  8.3717e-04,  1.4670e-03,  2.0332e-03,
         2.5692e-03, -2.3540e-03,  2.2112e-03, -1.0786e-03,  3.1524e-03,
         7.0834e-03, -3.6162e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.6454e-03, -9.0899e-01, -8.6507e-03,  9.1366e-03,  2.5650e-04,
         2.0058e-03, -3.8603e-03,  1.1059e-02,  2.3029e-03, -4.8198e-05,
         1.3077e-03, -1.6425e-03, -8.8996e-04,  3.7743e-03,  2.5485e-03,
        -4.6143e-03,  2.4556e-03,  1.4439e-03,  4.8531e-03,  4.0917e-03,
        -6.6644e-04,  2.5861e-03, -1.1454e-02, -3.7117e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5704e-03, -8.8757e-01, -2.2397e-02, -8.8093e-03, -1.0905e-03,
        -1.6661e-03, -1.7613e-03, -1.0465e-02, -3.2855e-03,  4.4026e-03,
         9.1963e-03, -4.3291e-03, -5.9356e-03, -2.6129e-03, -7.7403e-03,
        -1.0545e-03,  4.5254e-04,  1.2924e-03, -2.5875e-03,  7.5384e-04,
        -3.9902e-04,  2.9978e-03, -1.1450e-03,  3.6046e-03,  6.6145e-04,
         5.8837e-04,  1.1342e-03,  3.5327e-03,  4.7109e-04, -3.0264e-04,
        -2.0081e-03,  4.1804e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2369e-02,  8.8013e-01, -4.7341e-03, -7.9713e-03, -4.9661e-04,
        -1.1429e-02,  3.2453e-03, -5.6742e-03,  4.7704e-03,  3.4797e-03,
         2.5881e-03, -2.0394e-03, -8.9113e-04, -8.7328e-06,  1.9857e-03,
         5.8790e-05, -1.1866e-03, -1.6942e-03,  6.2028e-05,  5.0320e-03,
         8.8980e-03,  3.0350e-04, -1.9199e-03,  1.1448e-03, -1.0297e-03,
         6.8802e-03, -6.1997e-03, -2.5336e-03,  3.1839e-03,  4.7528e-03,
         3.3028e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2698e-02, -8.8574e-01, -5.9855e-03, -1.6387e-03, -3.0676e-04,
         2.2410e-03,  1.9474e-03,  2.1202e-02,  1.0730e-02,  1.9269e-03,
        -8.3572e-03, -1.8392e-03, -8.8520e-03,  9.5492e-03, -2.9984e-03,
         1.0121e-03, -1.4972e-03, -3.8552e-03, -6.5792e-03, -5.2877e-04,
         7.2461e-05, -9.2227e-04,  9.5219e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6279e-03,  9.4344e-01, -1.1659e-03,  4.0090e-04, -8.8221e-03,
         3.1233e-03, -1.7342e-03,  1.3294e-03, -2.0891e-04,  9.1437e-04,
        -9.2032e-05, -7.6073e-04, -6.0944e-04, -1.7053e-03, -4.4158e-03,
         2.6555e-04, -3.0286e-03, -1.1870e-03, -1.5595e-03, -5.3560e-04,
        -3.4914e-03, -4.8216e-03, -2.2799e-03, -6.1620e-04, -5.9487e-05,
         3.6617e-03,  1.8686e-03,  6.2751e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 2.4888e-02, -9.1132e-01, -2.8899e-03, -1.7497e-03, -2.4450e-03,
         1.0106e-03, -2.2974e-03,  3.3211e-03, -1.8364e-03,  2.0226e-03,
        -2.0255e-03,  2.3965e-04, -2.3977e-03,  5.6023e-04,  1.3588e-03,
        -3.4749e-03,  1.3428e-03, -3.8351e-03,  2.0220e-03,  7.3809e-04,
        -1.9495e-03,  1.6054e-03, -3.4237e-03,  1.1272e-03, -1.3803e-03,
        -2.6922e-03, -4.8638e-03,  6.0239e-03, -5.1581e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0040e-02, -9.0059e-01, -1.4609e-02,  3.2518e-03, -1.6803e-03,
        -1.6013e-02,  1.5287e-03, -2.8502e-03,  8.1642e-04,  8.0366e-04,
         1.5821e-05,  1.6592e-03, -1.0600e-03, -1.2515e-03, -2.7007e-03,
        -2.0532e-05,  2.5956e-03, -1.8839e-03, -9.9214e-03, -3.0851e-03,
        -1.2446e-03,  1.7002e-02,  4.7230e-03,  6.5024e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1877e-02, -8.8225e-01, -5.9910e-03, -3.6691e-03, -3.5483e-03,
        -1.6566e-02, -9.3637e-04, -6.8456e-04,  3.0035e-03, -4.0354e-03,
        -1.0185e-03,  8.2811e-03, -6.9360e-03, -4.3952e-03,  3.8680e-03,
        -7.4063e-04,  2.8583e-05, -1.1828e-05,  6.9451e-04, -8.5724e-04,
         3.2582e-03,  4.5143e-03, -3.6706e-04, -4.9304e-03, -5.9973e-04,
         2.1104e-03,  4.0697e-03,  2.5214e-03,  4.4761e-03, -5.8286e-03,
         2.0528e-03, -2.4316e-03,  2.7632e-03,  6.8296e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7620e-03,  9.2416e-01, -1.5589e-02,  1.4408e-03, -2.2256e-03,
        -4.5406e-04,  2.3883e-03,  4.1048e-04,  7.7161e-04, -1.3854e-02,
         2.8553e-03, -4.4966e-04, -2.2385e-03, -1.5438e-03,  8.9205e-04,
        -3.0551e-03, -7.4954e-03,  7.8899e-04, -1.4353e-03,  3.0347e-03,
         5.1287e-03,  8.0267e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.8058e-03,  9.1225e-01,  8.5031e-03, -6.4150e-03,  1.1122e-02,
         5.7124e-04,  4.5180e-03,  5.8291e-03, -7.8351e-04, -2.7446e-03,
         1.3329e-02, -1.8820e-03,  2.7717e-03, -2.8183e-03, -2.9773e-03,
        -2.6291e-03,  7.8607e-03, -3.1893e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.3225e-03,  9.3238e-01,  7.6424e-03,  1.8367e-03,  3.3851e-03,
         8.3202e-04, -1.5879e-03, -7.1155e-04, -3.8958e-03,  2.0913e-04,
        -1.5933e-03, -3.0490e-03, -4.5145e-03, -9.9222e-04, -1.3956e-03,
        -4.5459e-04, -1.3033e-03, -2.3877e-03, -4.9487e-03, -2.7070e-03,
        -3.0393e-03,  1.2729e-02,  2.0872e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.2797e-03, -9.2023e-01, -2.8120e-03,  6.9596e-03,  9.9111e-03,
         4.1830e-03, -2.5626e-03,  6.5114e-03,  7.4430e-03,  4.2774e-03,
         6.9035e-03,  1.9803e-03,  3.0064e-03, -4.5733e-03,  3.3334e-03,
        -7.0906e-04, -7.3199e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4239e-02, -8.9271e-01, -2.2441e-02,  2.4235e-03,  4.8679e-03,
         3.2009e-03,  3.5688e-03,  5.8438e-04, -9.4744e-03, -1.1527e-02,
         2.8359e-03,  1.0897e-03,  5.2516e-04, -3.6483e-03, -5.1703e-03,
        -6.3915e-03, -1.0159e-02,  5.1445e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7076e-02,  8.9668e-01, -1.0253e-02,  2.2248e-03,  4.7977e-03,
         2.4467e-03, -4.1609e-03, -8.2721e-03, -7.1940e-04, -4.7099e-03,
         9.3496e-04, -3.4949e-03, -6.8239e-04, -6.0444e-04,  3.2017e-05,
        -1.9727e-03,  5.9381e-03, -1.0081e-02,  1.9145e-04, -3.4389e-03,
        -4.7138e-03, -4.0320e-03,  2.4964e-03,  7.8735e-04,  4.8180e-03,
        -4.4433e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4993e-02, -8.8819e-01,  1.6521e-02, -2.5199e-04, -6.5361e-03,
        -3.4525e-03, -5.2838e-05, -3.9137e-03, -1.3320e-02,  8.0021e-03,
         2.2882e-03,  1.9132e-04,  1.8951e-03,  9.6821e-04,  4.1931e-03,
         9.9933e-04, -2.8901e-03,  4.7373e-04,  1.5233e-03, -1.2878e-03,
         7.4401e-04, -5.4192e-03, -1.2607e-03, -3.4097e-03, -1.8640e-03,
        -1.1868e-03,  1.9772e-04, -7.1880e-04, -1.1199e-02,  5.7545e-05,
        -2.0027e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3359e-02,  8.7285e-01,  2.1378e-02, -7.8289e-04, -1.6298e-02,
        -2.3616e-03,  3.9391e-04,  5.4795e-03,  1.5360e-02, -4.5278e-03,
        -1.6720e-03, -2.5697e-03, -3.6034e-03, -3.5811e-03,  1.3988e-02,
         1.4849e-04,  9.7999e-03,  1.8432e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6821e-03,  9.1629e-01, -4.9644e-02, -9.7069e-04, -1.1975e-04,
         7.9931e-03,  1.5851e-02,  4.4501e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 1.9234e-02,  9.1559e-01,  3.2495e-03, -8.2807e-03, -9.6872e-04,
        -4.1248e-04, -3.8384e-03, -1.7841e-03,  1.6029e-03, -9.3725e-04,
        -1.0648e-03,  3.5251e-03,  5.3431e-03,  1.1656e-03,  1.2908e-03,
        -7.4091e-04, -1.5400e-03, -3.5163e-04,  3.4030e-03,  1.2909e-03,
         8.7106e-04,  1.5547e-03, -5.4611e-04, -6.1091e-04,  1.7176e-04,
        -1.5243e-03, -1.2908e-03,  1.0002e-03, -3.8006e-05, -1.6533e-05,
         1.5286e-03,  2.3656e-04, -5.6882e-04,  3.3314e-03,  9.6246e-04,
        -1.1266e-03,  1.2565e-03,  1.4608e-03, -6.2896e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.9027e-03, -8.5468e-01,  3.7621e-02,  5.8893e-03, -2.8523e-03,
        -5.3075e-04, -2.9283e-03,  2.6710e-03,  5.8634e-03,  1.9571e-03,
         2.8535e-03, -9.1985e-04, -3.2519e-03,  3.9170e-05,  8.6535e-04,
         3.2473e-03, -4.4839e-03,  2.0817e-03, -6.8569e-04, -6.7554e-03,
        -1.2871e-02, -4.3206e-03, -4.5302e-03,  8.1043e-04, -5.9969e-04,
         6.4752e-04,  2.1525e-02, -4.6172e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0514e-04, -9.4479e-01, -1.7042e-02, -9.5049e-03,  2.5767e-03,
        -1.4545e-04, -2.2772e-03,  7.6019e-04,  2.9267e-03,  1.4840e-03,
        -7.3338e-03,  6.7501e-03, -4.3072e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3928e-02, -9.2495e-01, -6.3643e-03, -5.3870e-03,  3.4687e-03,
        -8.9135e-03, -1.0674e-03, -1.6936e-02, -1.9189e-03, -9.6252e-04,
         4.6097e-03,  1.7194e-03, -4.1676e-04,  3.9858e-03,  5.3766e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3166e-02, -8.4687e-01,  2.5765e-02,  8.3934e-03, -5.3364e-03,
        -1.7475e-03, -5.4260e-04,  5.1923e-04, -3.2284e-03, -6.3027e-03,
        -1.1360e-03, -4.2603e-05,  2.4022e-03, -3.3468e-04, -2.4473e-03,
         3.0857e-03, -8.0619e-03,  4.1266e-03,  3.1108e-03,  4.1820e-04,
        -2.9211e-03,  4.3467e-04,  4.1313e-04,  1.2151e-03, -5.6245e-03,
        -1.0267e-03, -2.3996e-03, -4.4879e-03,  2.4393e-03, -1.4304e-03,
        -3.2092e-04,  6.9606e-05,  3.7948e-03,  5.7696e-04, -2.4389e-03,
         8.8573e-04,  8.8095e-04, -1.5684e-03, -1.2444e-03, -1.1965e-03,
        -2.2227e-03,  1.6056e-03, -2.9339e-03, -3.7858e-04, -8.7208e-04,
         3.3284e-03,  6.2470e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.2071e-04, -9.1653e-01, -8.7181e-03, -3.5172e-03,  9.0675e-03,
        -1.6541e-03,  1.2236e-03, -3.4642e-03, -5.2881e-03, -2.4178e-03,
        -1.8418e-03, -4.2736e-03, -1.0744e-03, -4.5954e-05, -9.0471e-04,
        -2.2858e-03, -7.7105e-03, -1.4960e-03,  2.3156e-03, -8.6734e-04,
        -8.5087e-05,  3.1786e-04, -3.0513e-03,  1.3384e-03,  1.9123e-04,
        -2.1777e-04, -6.5981e-03, -2.8220e-04, -4.5298e-04, -3.9063e-04,
        -1.0516e-04, -2.4749e-05, -1.8770e-03, -7.7807e-04,  1.6532e-04,
         1.3089e-03,  2.9327e-04,  1.4976e-04, -1.4738e-03, -1.3693e-03,
        -2.6502e-03, -1.4611e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8599e-02, -8.9968e-01,  1.9041e-03,  7.7248e-03, -7.0335e-03,
        -3.0423e-03, -9.6916e-03, -1.5815e-03,  5.6237e-05, -7.9133e-03,
         1.6346e-03, -1.2683e-03,  1.3219e-03, -8.2445e-04, -3.3669e-03,
         1.4595e-03, -4.9877e-03, -1.5525e-03,  1.5651e-03, -3.7825e-03,
         3.4226e-03,  7.9729e-04,  9.2514e-03,  2.1667e-03,  5.3708e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.9595e-02,  8.6596e-01,  1.0491e-02, -2.9983e-03, -4.9765e-04,
        -3.6420e-04,  6.5519e-03, -4.5560e-04,  2.8513e-03, -6.4035e-03,
         2.0951e-03, -4.1879e-04, -2.2969e-03, -1.1334e-02,  2.1387e-04,
         5.7489e-03, -2.0573e-03,  1.3200e-03, -3.1034e-03, -1.0260e-03,
        -3.2608e-03,  8.3287e-03, -2.6275e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4139e-02, -9.3490e-01, -5.1225e-03, -2.3471e-03,  2.4501e-03,
        -2.5318e-03, -4.7691e-04, -3.0205e-03, -6.0795e-05, -5.0447e-03,
        -6.8376e-04,  5.1199e-03, -6.2185e-04,  3.8177e-04,  1.4988e-03,
         1.5336e-03, -9.3796e-04,  2.7270e-03,  4.8232e-03,  4.4655e-03,
         2.5815e-03,  2.8964e-03,  1.6365e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.2689e-02,  8.6943e-01,  6.8004e-03, -4.6196e-03, -1.0192e-02,
         1.0103e-03,  4.3522e-03,  9.2646e-03,  1.1892e-02,  2.2368e-03,
        -6.2218e-03,  1.0596e-02, -6.9542e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3568e-02,  9.4892e-01,  3.0072e-03,  4.8121e-03,  3.6181e-03,
        -1.8140e-03, -3.4833e-03,  3.6871e-03, -1.8622e-03,  4.2492e-03,
        -1.7020e-03, -4.4769e-04, -7.6209e-03, -1.2107e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6658e-02, -9.0596e-01, -2.2229e-02,  4.3702e-03,  2.1841e-03,
        -1.5119e-03, -3.3315e-03,  2.5650e-03, -8.6643e-04,  4.4538e-04,
         3.9202e-04, -1.0053e-03, -4.0484e-04, -5.1930e-03,  7.6632e-04,
        -1.7533e-03,  2.4760e-03, -2.6189e-03, -5.2678e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 8.9188e-03,  9.3107e-01,  7.7889e-04,  5.4678e-04,  1.9901e-03,
         3.7056e-03,  7.3598e-04, -3.9930e-03, -3.6163e-03,  3.4284e-03,
         1.8802e-03, -2.1433e-03,  1.3736e-03,  2.5757e-03,  3.8668e-03,
        -4.4219e-03,  3.8222e-03,  5.6594e-03,  2.4012e-03,  2.4189e-03,
        -5.2882e-04,  1.0097e-04, -7.5288e-04,  3.3082e-03,  2.6328e-03,
         3.3259e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.4805e-03,  9.0458e-01,  1.0516e-02, -3.9042e-03,  1.1597e-03,
        -9.9190e-03, -3.7101e-03, -1.1772e-02, -6.3770e-03,  2.1459e-02,
         8.1611e-05, -2.6648e-03, -2.9197e-03,  6.9536e-03,  1.7141e-03,
         2.7897e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0184,  0.8699,  0.0245,  0.0206, -0.0029,  0.0053,  0.0081, -0.0035,
         0.0018,  0.0069,  0.0074,  0.0011,  0.0021, -0.0021,  0.0155, -0.0020,
         0.0078,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7636e-02,  9.2360e-01,  1.3444e-02, -1.1302e-03, -4.8932e-03,
         7.8227e-03, -2.6678e-04,  9.0425e-04, -1.0302e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9532e-02, -8.4146e-01, -3.4776e-03, -1.3618e-02,  9.5530e-03,
        -4.2959e-03,  5.7707e-04,  2.2303e-03,  1.5448e-03,  1.0714e-03,
        -4.5729e-05, -6.8537e-04, -2.4521e-03, -1.3444e-03, -3.4543e-03,
         1.2977e-03,  2.2416e-04, -3.7549e-03, -2.9409e-03, -2.7603e-03,
        -2.5285e-03, -4.4904e-04, -2.4896e-03, -4.8298e-04,  4.1894e-03,
         7.2316e-04,  8.7298e-03, -5.0541e-04, -7.6162e-04, -2.9109e-04,
        -5.8132e-04,  6.6327e-04, -1.0327e-03,  2.0225e-03, -5.9198e-04,
         2.3597e-03, -3.6656e-04,  5.5897e-05,  5.2310e-03,  2.2998e-03,
         1.4937e-03,  4.3796e-03,  4.3019e-03,  4.4043e-03, -5.3141e-03,
         1.7430e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.8308e-03,  9.0518e-01,  1.5228e-02, -9.8884e-03,  1.2380e-03,
         7.9208e-04,  4.8312e-04, -5.2943e-03, -2.7451e-03, -1.7627e-03,
        -1.5038e-03, -6.5065e-05, -8.5224e-03, -3.5307e-03,  3.1911e-03,
        -1.6376e-04,  1.5959e-03, -2.7367e-03,  2.8666e-03,  2.6706e-03,
        -1.6394e-03,  2.1503e-03,  3.0094e-03,  8.9355e-04,  1.8683e-03,
        -6.0296e-03,  1.1373e-03, -9.7436e-05, -1.2164e-03, -5.6664e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0048,  0.9535, -0.0037,  0.0140, -0.0013,  0.0084, -0.0142,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0041, -0.9513,  0.0012, -0.0123, -0.0086, -0.0038, -0.0039, -0.0041,
         0.0061, -0.0026,  0.0019,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3705e-02, -8.4564e-01, -9.9565e-03, -2.0166e-02, -1.2560e-02,
        -6.1252e-03, -1.8781e-03,  1.9483e-04, -4.9393e-03, -1.1996e-03,
        -2.9136e-03, -9.8473e-03, -3.4499e-02,  6.3768e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2837e-03, -9.1432e-01,  3.9215e-03, -8.6045e-05,  1.1740e-03,
         2.2673e-03,  3.2075e-03,  5.1134e-04, -1.5847e-03,  6.7759e-04,
         1.6326e-03, -4.8431e-03, -9.0878e-04,  6.7598e-03, -1.1833e-03,
        -1.2818e-03, -1.6613e-03,  6.5960e-04,  1.0172e-03, -8.1033e-03,
        -1.1207e-02, -2.4704e-03, -1.6769e-03,  1.0420e-03,  2.4535e-03,
        -7.5830e-04, -2.7870e-03, -2.3182e-04,  9.9873e-04,  1.3747e-03,
        -3.6012e-03,  2.2548e-03,  2.3869e-03, -7.6698e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4853e-04,  8.7080e-01,  1.8263e-02,  1.6395e-02, -5.1453e-04,
         2.5884e-03,  1.4339e-03, -1.1187e-02,  4.0966e-03, -7.5321e-03,
        -3.1345e-03,  1.0306e-03,  5.2544e-04,  1.7226e-03,  1.1829e-02,
        -7.2404e-03, -1.0266e-02, -6.1740e-03, -2.8744e-03, -3.8717e-03,
         4.1096e-03, -2.4940e-04, -4.6809e-03, -2.3615e-04,  8.6266e-05,
         1.9355e-03, -5.8129e-03,  1.2568e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3222e-02, -8.7702e-01, -2.3467e-02,  1.2274e-02, -7.4407e-03,
        -9.1062e-04,  2.3467e-03,  3.6255e-04, -8.5459e-04, -1.0541e-03,
        -1.6469e-04,  4.2840e-03, -3.0938e-03, -4.5602e-03, -1.8836e-03,
         1.4892e-03, -1.2918e-02,  1.0320e-02, -3.9874e-03, -2.0128e-03,
         4.5418e-04, -2.1444e-03,  1.0371e-03, -2.6978e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-1.8104e-03,  9.1732e-01, -8.1585e-03,  1.6619e-04, -2.5895e-03,
        -4.2227e-03,  7.8681e-03, -3.0997e-03,  3.1853e-03,  9.9097e-04,
        -5.6463e-03,  1.8568e-03,  1.5904e-03,  2.9267e-03, -3.4859e-04,
        -5.5617e-04,  3.4862e-03, -1.6945e-03,  1.8493e-03, -2.1044e-03,
        -1.9535e-04, -1.7628e-03, -2.5658e-03,  2.1114e-03,  1.7998e-03,
         8.7590e-03, -3.4797e-03, -2.0705e-03, -7.5698e-04, -2.5822e-03,
         2.4498e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0073, -0.9013,  0.0296,  0.0129, -0.0027,  0.0091,  0.0040,  0.0024,
        -0.0187,  0.0012,  0.0093, -0.0016,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4751e-02, -8.8793e-01,  4.3591e-03, -1.5958e-02, -6.3646e-04,
        -6.9984e-03,  1.2582e-03,  3.5571e-03,  5.0601e-03,  6.6463e-04,
         9.0527e-03,  1.2785e-02, -8.3438e-03,  4.0794e-03,  4.5685e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.8553e-03, -9.0982e-01,  7.1362e-03, -9.5348e-03,  8.0075e-03,
         3.9679e-03, -8.6095e-03,  1.8714e-03, -1.6644e-03, -2.2444e-03,
        -2.0957e-03,  1.0858e-03, -8.7107e-04,  1.0313e-03, -1.1531e-03,
        -6.9982e-04,  3.0449e-03,  1.2485e-03,  1.5126e-03,  4.1458e-03,
         2.0942e-05, -1.8535e-03,  1.7188e-03, -2.2521e-05, -2.8799e-03,
        -2.1582e-03, -3.0600e-04,  1.0569e-03,  2.8116e-04, -5.2065e-04,
        -1.8606e-03, -1.8419e-03,  8.8840e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1360e-02, -9.3311e-01, -1.1629e-02,  6.4410e-03, -1.8126e-03,
        -6.0712e-04,  2.2281e-03,  9.7369e-04,  9.0579e-03,  4.9502e-04,
         7.2160e-03, -2.0228e-03,  1.6259e-03,  3.4092e-03,  2.4243e-03,
         5.5902e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2900e-02,  9.3768e-01, -1.1880e-03, -4.4916e-03,  4.8310e-03,
         8.6423e-04,  6.8683e-04, -1.4663e-03, -1.5894e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0678,  0.9094, -0.0074,  0.0103,  0.0051,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3144e-02,  9.2669e-01, -1.1140e-02,  2.7949e-04, -4.3076e-03,
        -3.2147e-03, -7.8976e-04, -4.1371e-03, -1.1338e-03,  3.6628e-04,
        -2.6628e-03, -1.6938e-04, -8.8813e-03, -2.1421e-03, -1.6415e-03,
        -3.9333e-04, -2.5714e-04,  2.8381e-03, -3.9624e-04, -2.6310e-04,
        -2.6212e-03,  6.4412e-04,  8.9288e-04, -3.8199e-03,  1.1914e-03,
         2.7544e-03,  3.2273e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0260, -0.9263, -0.0198,  0.0023, -0.0080, -0.0154, -0.0021,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3422e-02,  8.9696e-01,  6.7119e-03,  7.8266e-03,  4.6830e-03,
        -1.3394e-04, -6.9475e-03,  1.5646e-03, -1.3537e-03,  3.6333e-03,
        -8.6339e-04,  1.9155e-03,  2.2892e-03, -3.0309e-03, -1.1273e-03,
         1.4755e-03,  4.1009e-03,  3.6418e-03, -2.1968e-03,  9.3175e-03,
         2.5682e-03, -4.2360e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3484e-04,  9.0463e-01,  1.6505e-02,  3.2229e-03,  8.7438e-03,
         9.6364e-04, -1.9568e-04,  8.0798e-03, -1.3338e-03, -4.5793e-03,
        -6.8006e-03, -1.2925e-03, -4.8450e-03,  7.0862e-03, -5.2283e-03,
        -3.6558e-03, -3.1914e-03,  7.9985e-03, -1.0817e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.3037e-03, -9.4618e-01, -2.3103e-03,  3.0833e-03, -3.7116e-03,
         1.9951e-03, -9.4149e-05,  2.6355e-03, -3.1614e-03, -7.0925e-04,
        -3.6748e-03, -2.7642e-04,  9.4309e-04, -9.9191e-04, -2.1652e-03,
        -2.7305e-04,  9.3237e-04, -7.0089e-04, -6.2922e-04, -1.5264e-03,
        -1.1016e-03,  5.7975e-04, -1.8363e-04,  6.9528e-04,  1.0739e-03,
        -1.0121e-03, -4.7581e-04,  1.1283e-03, -1.7854e-03,  3.0127e-03,
         4.6581e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 8.0975e-03,  8.3491e-01, -3.3348e-02, -3.5250e-03, -6.0291e-04,
        -2.9580e-03,  2.8812e-03,  2.2034e-04, -4.6205e-03, -1.2479e-03,
         1.4425e-05,  2.6090e-03, -2.7961e-03,  1.3399e-02, -8.6959e-06,
         3.2157e-03,  8.1442e-03,  4.4791e-05,  6.2828e-03, -1.7895e-03,
         5.5477e-03,  1.8733e-03,  6.6252e-03,  8.3094e-03, -7.5720e-03,
         1.8407e-03, -4.8017e-06, -2.1211e-02,  5.7320e-03,  4.0509e-03,
         3.9168e-03,  2.5976e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0571e-03, -9.5144e-01, -4.9169e-03,  3.1979e-03,  1.4884e-04,
         2.3033e-04,  6.7275e-04,  2.4862e-03, -7.8062e-03, -1.2529e-03,
        -1.8173e-03, -1.6148e-03, -4.2351e-03,  5.0664e-04, -2.3807e-03,
        -1.9040e-03, -2.6413e-03,  1.0527e-04,  2.3247e-03, -5.4074e-04,
         1.2190e-03, -5.4977e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1749e-02, -8.5985e-01, -2.1822e-02,  6.5039e-03,  2.3098e-03,
        -6.7652e-03, -3.4364e-03, -3.2486e-04,  4.5346e-03,  2.9332e-03,
        -2.9081e-04,  1.6633e-03, -1.9093e-03, -1.3466e-03,  7.4908e-03,
         1.0493e-03, -3.9716e-04, -1.3057e-02,  6.2347e-04,  2.7549e-03,
         8.5286e-04,  4.1111e-03,  1.5525e-03,  2.8466e-03,  2.6935e-03,
        -4.5637e-03,  3.0305e-03,  8.5244e-03, -3.1928e-04, -2.0695e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1973e-02,  9.2700e-01, -1.6174e-02, -8.1544e-03,  2.2297e-04,
        -6.8706e-03,  4.7670e-03,  2.4656e-03, -4.4133e-03,  1.4574e-03,
        -2.7134e-03,  2.9381e-03, -8.6247e-04, -4.7868e-04, -1.8535e-04,
         3.4751e-03, -5.6542e-03, -1.9600e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7584e-03, -8.8947e-01, -3.0117e-02, -2.6222e-03,  2.8349e-03,
         1.0482e-02,  8.9362e-04, -1.4541e-03,  3.9886e-03, -6.7083e-04,
        -4.8710e-05,  3.0594e-04, -6.9760e-04, -3.9817e-03, -2.9858e-04,
         3.1576e-03, -5.3379e-03, -1.0077e-03,  5.1870e-03,  1.2958e-03,
         1.6859e-04,  8.0715e-04, -2.2859e-03, -6.1752e-03, -2.2287e-04,
        -1.2392e-04,  3.2742e-03, -6.2155e-04, -1.8602e-03, -9.8643e-04,
        -6.5665e-04, -1.0772e-03, -4.8559e-03,  2.4914e-04,  3.4469e-04,
         3.0631e-03, -6.6218e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2776e-03,  8.9304e-01, -1.9942e-02,  1.2100e-02, -6.9799e-03,
         2.7690e-03,  2.1740e-03, -3.2928e-03,  1.6014e-03,  2.7205e-03,
        -4.9229e-05,  2.5613e-03, -3.9090e-03,  5.0489e-03,  4.5111e-03,
         1.1130e-03,  5.9490e-03,  7.8265e-04,  5.5509e-04,  5.1094e-04,
         6.3010e-03,  1.0430e-03,  1.2058e-02,  7.7144e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.2158e-03,  9.2548e-01, -8.5588e-03, -2.4816e-03, -2.6062e-04,
         4.2494e-03, -1.9461e-03, -2.4133e-03,  3.7640e-03, -5.9212e-04,
        -5.7307e-04, -3.9471e-04,  2.2773e-03, -1.3178e-03,  1.8727e-03,
         7.0831e-04,  3.7616e-04,  4.1250e-04, -1.5611e-04,  5.8561e-04,
         6.1767e-04, -5.7963e-03, -4.2951e-04, -3.7490e-04, -1.0506e-03,
         5.1976e-03, -1.0518e-03, -1.5419e-04, -1.2792e-03, -1.3818e-03,
         2.0848e-03, -2.5951e-03,  3.7322e-04,  1.3120e-04,  3.7895e-03,
        -2.8044e-03,  6.5723e-04,  9.2555e-04,  5.2976e-04, -5.4184e-04,
        -3.3315e-04, -1.6295e-04,  7.6386e-04,  4.6081e-04,  1.1523e-03,
         7.2600e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.4370e-03,  8.7131e-01, -3.8891e-02, -8.0060e-03,  5.0228e-03,
         1.5023e-03, -9.7509e-03, -1.8704e-03,  5.4675e-04, -6.6015e-03,
         3.2096e-03,  4.2738e-03, -1.8690e-03, -6.9213e-04, -6.2487e-04,
         2.0211e-03, -3.2940e-03, -1.4643e-02,  1.4693e-03,  2.0005e-03,
         4.8588e-03, -1.0070e-03,  9.1022e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1906e-02,  9.5551e-01,  7.4705e-03, -4.8711e-03, -2.7493e-03,
        -2.7445e-03, -1.8170e-04,  1.7005e-03,  3.0955e-03, -1.1717e-03,
         8.5952e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0409,  0.9092,  0.0088, -0.0031, -0.0011, -0.0059,  0.0034, -0.0080,
        -0.0010,  0.0024,  0.0040, -0.0024,  0.0034, -0.0064,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0356,  0.8915,  0.0029, -0.0017, -0.0049, -0.0011, -0.0010, -0.0068,
        -0.0059, -0.0318,  0.0115, -0.0053,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3836e-02,  8.5782e-01, -9.7503e-03,  9.1138e-03, -1.4427e-03,
         3.5468e-03, -6.0238e-03,  8.5032e-04,  3.6598e-03,  3.8971e-03,
        -3.1619e-03,  1.4897e-03, -1.0509e-04,  7.5619e-04, -1.7207e-03,
        -8.6565e-04, -8.7492e-04, -2.5134e-03,  2.1743e-03, -1.6017e-03,
        -1.0433e-03,  4.3314e-03,  2.4256e-03, -2.4188e-03, -4.8708e-04,
        -5.0370e-03,  1.9859e-02, -1.2122e-03, -7.7702e-04,  6.0448e-03,
        -4.8451e-04,  6.1505e-03,  6.5683e-04,  4.2897e-03,  2.0541e-03,
        -1.2727e-03,  5.9834e-04,  6.1434e-04, -3.0547e-03, -2.0383e-04,
         5.9834e-04,  1.1206e-03, -5.8263e-04,  9.1895e-04,  9.8346e-04,
         2.5061e-03,  8.7857e-04, -1.4238e-04, -1.3593e-03, -6.0816e-04,
        -2.0995e-04,  4.9283e-04, -2.4941e-04,  2.7416e-04,  8.5579e-04],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 8.6911e-03,  9.1138e-01,  7.3057e-03,  1.4007e-02,  1.5878e-03,
         1.1265e-03, -5.3796e-03, -1.6952e-03,  1.9361e-03,  6.4206e-03,
        -7.6570e-03, -1.5215e-03,  6.9226e-04,  3.0472e-03,  9.3562e-04,
         2.4075e-03,  1.1427e-03, -7.7339e-03,  3.0075e-03,  3.0195e-03,
         2.4915e-03, -2.3929e-03, -4.4267e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8270e-02, -9.1964e-01,  7.1385e-03, -1.4527e-03, -1.0630e-02,
        -1.8704e-03,  1.2693e-03, -2.8096e-03, -4.1502e-03, -6.5656e-04,
         3.4838e-03,  5.6222e-03,  8.3968e-04, -9.8388e-03, -2.2678e-03,
         1.2304e-03,  5.9370e-03,  2.8941e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9026e-02, -8.6729e-01, -2.9741e-03, -4.6965e-03, -1.6279e-02,
         7.1292e-04, -2.2400e-02, -5.3806e-03,  5.9749e-03,  6.4265e-03,
        -1.1690e-02, -3.1017e-03, -3.0025e-04,  6.3680e-05, -8.2663e-03,
         2.2997e-03,  1.9889e-03, -1.1300e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2934e-02, -8.9199e-01,  2.2240e-02,  8.8060e-03, -2.3949e-03,
        -1.0584e-02, -5.3763e-03, -1.6399e-04, -6.5678e-05,  3.2004e-03,
        -2.7361e-04, -1.2127e-02, -4.7303e-03, -1.0817e-02,  4.2990e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0243e-02, -8.6992e-01,  2.4281e-02, -1.3900e-03,  2.2866e-03,
         4.0109e-03,  2.5949e-03, -1.2052e-04, -1.2021e-03, -4.4287e-03,
        -2.7179e-04,  2.8031e-03, -4.7221e-03, -2.2280e-03,  1.5623e-03,
        -4.7996e-03,  3.1361e-03,  7.8018e-04, -1.2422e-03, -1.7207e-03,
        -9.1451e-03,  1.1484e-03,  1.7336e-03,  3.9831e-04, -1.5717e-03,
         1.7606e-03, -3.4409e-03,  2.0456e-04, -2.0370e-03, -8.5135e-03,
         3.9066e-03, -2.4006e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7145e-02,  8.9128e-01,  4.6669e-03,  3.6907e-03,  4.8807e-03,
        -1.4982e-03, -6.0280e-03,  1.8990e-05, -5.7525e-03,  3.0588e-03,
         5.9202e-03, -2.5506e-03,  1.8420e-03,  3.5473e-03, -3.2566e-04,
        -4.5178e-04,  2.5953e-04, -5.2100e-03, -7.4446e-04,  8.6830e-04,
        -5.2948e-04,  1.8739e-03, -1.0716e-03,  9.4890e-04,  1.8927e-04,
         1.5694e-03, -1.3981e-03, -1.4985e-03, -1.5123e-03, -4.1370e-04,
        -2.1650e-03,  2.9564e-03, -4.1357e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9305e-03, -9.2339e-01,  2.0559e-03,  6.1486e-03,  8.2623e-04,
         1.9160e-03, -3.2745e-03,  1.0772e-03, -1.9157e-03, -3.2401e-03,
         2.5796e-04, -2.6291e-03,  6.4740e-04,  1.8498e-03, -9.2304e-04,
        -1.6188e-03, -9.7707e-04,  1.4589e-03, -6.2546e-04, -2.9118e-04,
         8.0598e-04, -7.1181e-04,  1.0708e-03,  1.6115e-03, -1.4021e-03,
        -8.4354e-04,  4.7689e-06, -1.3834e-03,  2.1814e-04,  1.6982e-03,
        -3.8888e-04, -2.2424e-03, -1.9762e-03, -3.6315e-04, -1.1324e-03,
        -3.9949e-03,  1.1033e-03, -1.7663e-03, -5.2587e-04, -3.0379e-04,
        -5.4425e-04,  7.0035e-04, -7.3799e-04,  1.1357e-03, -2.5499e-04,
        -6.5011e-04, -7.4086e-05, -7.5987e-04, -2.0005e-03,  5.6659e-04,
        -3.5457e-03, -9.8696e-04,  1.4037e-03, -4.0475e-04, -2.2226e-03,
        -3.8411e-04,  4.7097e-04,  2.5589e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6255e-02,  9.2905e-01, -2.9830e-03, -9.6289e-03, -3.4574e-03,
         1.7011e-03,  2.8557e-03,  2.4739e-04,  5.4059e-03,  1.9476e-04,
        -1.5492e-03,  4.9759e-03, -1.6217e-03, -2.7358e-03,  2.7961e-03,
        -3.5860e-04,  3.1082e-03,  9.9022e-04, -8.9115e-04,  7.9922e-04,
         8.9965e-04, -2.0979e-04,  1.1789e-03, -1.3706e-03, -4.0271e-03,
         7.1179e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.6631e-03,  9.1949e-01, -1.7148e-02, -3.3827e-03, -5.3112e-04,
         3.3628e-03, -1.5288e-03, -2.9930e-03,  2.4918e-03, -2.3481e-03,
         2.3864e-03, -1.3520e-03, -2.5881e-03, -3.9238e-03, -2.0595e-03,
        -6.1944e-04, -2.5691e-04,  2.7563e-03, -6.4263e-04, -2.9283e-03,
        -5.4151e-04, -1.4209e-03,  1.2127e-03, -3.2822e-03, -1.1093e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4070e-02, -8.8107e-01, -4.3040e-03, -2.5155e-03, -5.7988e-04,
        -1.0351e-03,  2.3447e-03,  8.2763e-04, -1.1096e-03,  1.5489e-03,
        -3.5620e-03, -1.3275e-03,  2.9941e-03, -5.7358e-03, -3.0898e-03,
        -1.9978e-03,  2.9160e-04, -4.4860e-03, -2.7421e-03,  5.2296e-04,
         1.2692e-03, -1.6223e-03, -1.6001e-03, -1.8657e-03, -1.6424e-03,
         1.0532e-03,  3.1410e-04, -3.1219e-03, -4.9563e-04,  1.6352e-03,
         1.1772e-03,  3.9245e-03,  3.7614e-03, -6.8703e-04,  4.9680e-03,
         3.4069e-03, -1.1977e-03, -1.1383e-03,  6.7026e-04, -9.2835e-04,
         1.8692e-03, -2.1281e-03, -2.9920e-03, -1.0999e-03,  9.2720e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4717e-02,  8.9836e-01, -1.5797e-02,  6.3394e-03, -3.6377e-03,
         1.5777e-03, -2.6272e-03,  2.1331e-04,  1.5824e-03,  1.7284e-04,
         9.0768e-04,  5.9249e-03, -1.2799e-04,  4.8094e-03,  3.5687e-03,
        -5.9838e-03,  9.8773e-03,  2.8923e-03,  3.4764e-04,  3.3339e-03,
         1.3269e-03,  2.1684e-03,  1.3708e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5376e-02,  9.1453e-01,  5.1507e-03,  2.1357e-05,  1.5585e-03,
         1.0934e-02, -1.8727e-03, -1.7280e-03, -1.7734e-03, -1.0234e-02,
        -1.8594e-02,  8.2279e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([-2.4669e-02, -9.0519e-01, -2.2119e-03,  8.1220e-03, -7.7142e-03,
         1.0425e-03,  8.1105e-03,  2.3064e-03, -8.4771e-04,  4.1102e-03,
         8.0780e-04,  3.9752e-04,  2.3893e-03, -4.6352e-03,  1.2060e-03,
         2.7198e-03, -2.1289e-04, -2.6301e-03, -1.1199e-03,  1.0747e-02,
         1.6710e-03,  3.2953e-03,  3.8455e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7014e-02,  8.9656e-01, -5.7132e-03,  1.7448e-03, -5.2078e-03,
        -1.9025e-03,  8.6995e-04,  1.5666e-03,  1.8928e-03,  1.6766e-03,
         1.3783e-03, -9.2053e-04, -5.9988e-04,  2.2176e-03,  1.9003e-03,
        -6.2523e-04, -2.3506e-03,  3.3577e-03,  2.3480e-04,  1.7447e-03,
        -1.5691e-03, -1.3168e-03, -3.5836e-03, -1.9753e-03, -6.0097e-04,
        -1.7641e-03, -1.3899e-03, -8.9537e-04, -3.7243e-03,  1.6208e-03,
        -3.2843e-04, -7.7661e-03,  3.9888e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0540e-02, -8.8832e-01,  6.1626e-04,  1.0665e-02, -5.9119e-03,
        -1.3220e-03, -1.6441e-02,  5.5165e-03, -3.9160e-04, -8.1118e-03,
         1.5468e-03,  2.7228e-03, -1.5991e-02,  1.8867e-03, -6.6372e-03,
         4.8119e-03, -1.2162e-03,  1.6379e-02,  9.7410e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2779e-02,  9.2906e-01,  4.7675e-04,  2.7326e-03, -1.3443e-03,
        -8.0890e-03,  1.3082e-02, -2.6315e-03, -8.4460e-04,  9.0390e-03,
         5.1073e-04,  6.2431e-04,  7.5633e-04,  1.8033e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.4841e-03, -9.2287e-01, -1.0009e-02, -4.7952e-03,  1.8424e-03,
        -5.4981e-03,  3.0433e-03, -2.9195e-04,  2.8012e-03,  9.3884e-04,
        -1.5769e-03,  1.7406e-03,  1.0174e-03, -1.0567e-02, -1.9124e-03,
        -2.1917e-04, -3.6994e-03, -1.3231e-03, -1.8105e-03, -1.6219e-03,
         9.9381e-04, -1.2941e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.1072e-03,  9.0834e-01,  1.1724e-02, -5.7985e-03, -2.5425e-03,
        -2.4026e-03, -4.1204e-03, -1.3144e-02,  1.7144e-04,  1.9794e-03,
        -3.7713e-03, -2.2013e-03, -4.0455e-04,  8.2499e-04,  2.4414e-03,
        -4.7906e-04,  6.7435e-04, -7.7733e-04, -2.8024e-03, -9.8247e-04,
        -2.1580e-03, -1.7493e-03, -2.4385e-03, -1.0684e-03, -4.4702e-03,
         1.0546e-03, -5.4832e-04, -1.9224e-03, -3.8569e-03,  7.0412e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.8816e-03,  9.1726e-01, -8.6501e-03, -1.7255e-02, -5.2047e-03,
         4.0359e-03, -1.5726e-03, -4.5756e-04, -4.1710e-03,  2.1261e-03,
        -9.6152e-03,  3.2907e-03,  3.0625e-04, -5.9031e-04, -5.4593e-03,
        -1.0085e-02, -4.0421e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0246e-02,  9.0745e-01,  1.1178e-02,  2.4280e-02, -3.6842e-03,
         7.1566e-03,  8.7081e-04,  9.8756e-03,  4.5615e-03, -1.0596e-02,
        -9.6659e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0067, -0.9262, -0.0081, -0.0022, -0.0054,  0.0012, -0.0023, -0.0010,
         0.0022, -0.0010,  0.0061, -0.0019, -0.0013, -0.0171,  0.0049,  0.0025,
         0.0098,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9993e-04,  9.1842e-01,  2.0465e-03,  1.0796e-02,  8.4268e-05,
         3.6727e-03,  2.4745e-03,  9.4966e-05,  3.9174e-03,  1.0502e-03,
        -3.7337e-03,  2.4677e-03,  2.1835e-03, -8.1975e-04,  9.6234e-03,
         2.1904e-03,  3.2417e-03,  7.9509e-04, -6.0491e-03,  1.6625e-03,
         6.5770e-04,  2.0519e-03,  9.3277e-03,  1.5354e-03, -2.8253e-03,
         7.7757e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6871e-02,  8.9929e-01,  1.1307e-02,  1.9882e-03, -2.7677e-03,
        -2.4030e-03, -2.1149e-03,  1.7941e-03, -2.9581e-03, -1.5065e-03,
        -5.6584e-03, -1.2153e-03, -2.3229e-03,  3.5915e-04, -2.0927e-03,
         4.5420e-04,  1.7832e-03,  2.1982e-03, -2.5242e-04, -2.9412e-03,
        -3.9052e-03,  1.7590e-04, -2.1649e-03, -3.3923e-04, -9.4133e-05,
        -1.0135e-04,  6.8783e-04, -1.8428e-03,  2.2467e-03,  6.1600e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8064e-02,  9.2520e-01, -8.7456e-04, -1.1513e-02,  1.2931e-03,
        -2.6061e-03, -1.1490e-02, -2.7173e-03, -1.6290e-03, -4.3580e-04,
         4.5235e-04, -3.2483e-03, -5.0451e-03, -5.0419e-03,  1.1063e-03,
         7.3470e-04,  4.9652e-04,  4.6190e-03,  3.4352e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 9.2557e-03,  8.8006e-01,  1.0775e-02,  3.7216e-03,  2.5510e-03,
         3.1138e-03,  2.5388e-02,  1.0160e-03,  3.2893e-03,  6.6673e-03,
        -3.9162e-04,  5.4671e-04, -8.1492e-04,  4.6515e-03,  2.8765e-03,
         1.2915e-03,  7.7936e-03,  6.9349e-03,  4.5864e-03,  7.3040e-03,
         1.3718e-03, -1.2673e-03, -1.9442e-03, -9.3523e-04, -4.8182e-03,
         3.7797e-03,  2.8559e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0134,  0.9415,  0.0033,  0.0047, -0.0037,  0.0030, -0.0022, -0.0048,
        -0.0010,  0.0030,  0.0053,  0.0055, -0.0034,  0.0052,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.1189e-02,  8.8767e-01,  7.8264e-03,  3.9257e-03, -2.1669e-03,
        -4.4548e-03,  3.8414e-03, -4.4112e-04, -9.0333e-03,  4.7588e-04,
        -4.7792e-03, -1.3926e-03, -4.0649e-03,  2.4126e-03, -3.7292e-04,
        -6.8092e-03,  7.3517e-04, -4.5394e-03, -5.0414e-04, -4.7780e-04,
        -2.1872e-03,  1.3313e-04,  4.1058e-03,  1.6427e-03, -2.1758e-03,
        -2.6387e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.5940e-03,  9.0865e-01, -1.4749e-02, -8.0411e-03, -8.9069e-03,
        -8.3384e-03,  2.6062e-04, -2.6153e-03, -7.4538e-03, -7.6088e-03,
        -4.9861e-03, -8.2981e-03, -1.2877e-02, -6.2583e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3857e-03,  9.1241e-01, -1.6948e-03,  1.1898e-02, -3.4553e-03,
         3.1818e-03,  6.3652e-04,  1.0272e-03,  8.8043e-04,  6.5691e-04,
         2.4296e-03, -2.6016e-03,  3.6990e-03,  1.1223e-03,  3.7068e-03,
         1.7805e-03,  1.9778e-03, -3.1771e-06,  1.4784e-03, -1.9574e-03,
        -2.6291e-03,  6.6279e-03,  3.4545e-03, -9.3560e-04, -6.6360e-04,
         1.5670e-03, -3.7418e-03, -1.8725e-03, -4.5988e-03,  2.1768e-03,
         1.7310e-03, -1.0018e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9593e-02,  8.7490e-01,  1.9576e-02,  1.0621e-03, -9.0854e-03,
        -1.6874e-04,  3.7415e-03, -1.7300e-02,  3.1575e-03, -7.5119e-04,
        -1.9543e-03,  2.0639e-03,  9.0575e-04, -4.5711e-04,  7.6418e-04,
        -5.9828e-04,  3.1857e-03, -6.0240e-03, -3.6457e-03,  2.7326e-04,
         2.9563e-04,  3.4555e-03,  2.6132e-03, -8.0695e-03, -6.3561e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0233,  0.9235, -0.0152, -0.0119, -0.0145, -0.0117,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9983e-02,  8.7517e-01,  6.2693e-03,  5.8191e-03,  1.0030e-02,
         3.1242e-03,  2.9795e-03, -1.0647e-02, -2.0345e-03,  1.9062e-04,
         7.5917e-04,  3.5827e-03, -1.1108e-03, -2.2008e-03,  2.4015e-03,
         1.9716e-03,  3.2060e-04,  3.7153e-04,  4.3295e-04,  2.2120e-03,
        -2.8673e-03,  3.3948e-04,  5.3405e-03, -2.3252e-03, -7.1076e-04,
        -3.0630e-03, -1.0536e-03,  4.6139e-03,  5.0535e-04,  4.0826e-03,
         4.0691e-03,  7.1643e-04,  1.6399e-04,  5.6209e-04,  1.3497e-03,
         1.5399e-02, -1.2317e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7115e-02,  8.5041e-01, -1.3742e-02,  1.8799e-02, -8.7400e-04,
         1.2290e-02, -1.3740e-03,  4.2874e-03, -1.6957e-03, -1.8352e-03,
        -9.8710e-04,  3.0870e-03,  3.8772e-04,  1.0980e-03,  2.5547e-03,
        -2.6402e-03, -1.8992e-03,  4.2262e-03,  1.7743e-03,  2.8636e-03,
        -1.5981e-02, -1.6245e-03,  4.5274e-05, -1.6962e-03, -1.8410e-03,
        -2.3671e-03, -9.4504e-04,  5.4953e-03, -7.1566e-04,  2.7587e-04,
        -6.8453e-04,  7.4843e-04,  6.8642e-04,  4.9876e-04,  2.4529e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0108,  0.8209, -0.0954,  0.0081, -0.0021,  0.0016, -0.0126, -0.0095,
         0.0067,  0.0068,  0.0013, -0.0013, -0.0045, -0.0124,  0.0061,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6803e-02,  9.0658e-01, -4.9506e-03,  1.8670e-03,  8.9180e-04,
         3.8633e-03, -6.6697e-04,  3.4394e-03,  2.4521e-03,  3.6313e-03,
        -3.4057e-04,  7.6893e-04,  1.4041e-03,  1.4561e-03,  2.5318e-05,
        -5.4330e-04,  3.9132e-04,  5.5198e-03,  6.4730e-03,  5.7377e-03,
         2.1295e-03,  4.0929e-03,  1.6537e-03,  9.8001e-04,  1.2021e-03,
         7.9003e-04, -8.4472e-04,  1.3424e-04, -1.0506e-03, -8.1072e-04,
         1.2330e-03, -3.9724e-04,  6.8768e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4576e-02,  8.8795e-01, -1.9306e-02, -7.1382e-04, -1.2736e-02,
        -6.4003e-04, -3.6599e-04,  1.7473e-03, -1.0617e-02,  5.4980e-03,
         2.6837e-03, -3.5990e-03,  4.4134e-03,  3.2902e-03,  2.4340e-03,
        -1.4737e-03,  3.1993e-03,  1.1118e-03,  2.5378e-03, -1.9247e-03,
        -1.8773e-03, -5.2314e-04, -4.5026e-03,  8.4008e-04,  3.1621e-03,
        -4.9326e-04, -2.2353e-03, -3.6870e-03, -1.8565e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([-1.9328e-02,  9.1819e-01,  2.2103e-02, -9.6045e-03, -9.8445e-04,
        -7.2628e-03, -7.2089e-04,  6.7189e-03, -3.0559e-03, -2.8159e-03,
         6.4714e-04, -3.9280e-04,  1.6513e-03, -3.9153e-03,  2.6101e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4640e-03,  9.0626e-01, -1.3753e-02, -1.0035e-03, -2.3000e-03,
         1.2419e-03,  3.1490e-03,  6.4744e-04,  4.2973e-03, -1.1131e-03,
        -1.7320e-03,  1.6390e-04, -6.0395e-04, -4.5676e-04, -2.7949e-03,
         3.1991e-03,  1.1992e-03,  8.3247e-03,  2.5941e-03,  1.0219e-03,
        -5.8944e-04, -1.2162e-03, -3.1560e-03,  1.2665e-03,  1.7149e-04,
         2.7152e-03, -8.1610e-04,  2.9402e-03, -2.1655e-03,  1.0044e-05,
         1.7226e-03,  1.4769e-03,  2.9144e-05, -2.2060e-03,  7.4637e-05,
         2.6689e-03, -1.2323e-03, -1.0576e-03,  2.8095e-03, -1.4104e-03,
        -2.2356e-04, -8.2049e-04, -2.4537e-03, -2.2934e-03, -6.5836e-04,
        -1.0884e-03, -2.7387e-03,  4.5451e-04, -1.2158e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5362e-02,  8.5729e-01,  3.5689e-02, -7.1918e-03, -1.3935e-02,
         1.0099e-02, -3.6592e-03,  1.4988e-03, -6.8169e-03,  2.8345e-03,
        -9.2267e-04,  1.3039e-04, -2.7462e-03,  4.1097e-04, -3.4982e-04,
         3.4519e-05,  1.0969e-02,  9.4757e-04,  3.5274e-04,  3.1503e-03,
         1.1033e-03,  3.4473e-04, -1.2437e-03,  4.5321e-04, -1.4051e-03,
        -5.9317e-04, -2.3752e-03, -2.0650e-03,  1.1853e-03, -1.6497e-03,
         2.0533e-03, -2.6301e-03,  4.7132e-04,  5.6885e-03,  2.3458e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.0355e-03,  9.5161e-01, -7.3050e-03,  1.8754e-03,  1.7956e-03,
        -3.2720e-03,  1.3007e-03,  2.7862e-03, -4.3265e-03, -7.5845e-04,
        -4.6548e-03,  1.4457e-03, -1.1399e-03,  4.2331e-04,  5.0144e-03,
        -3.4937e-03,  2.4466e-03,  8.2141e-04, -4.9221e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5956e-02,  9.2479e-01, -9.5606e-03, -4.7881e-03,  4.2870e-04,
         7.4034e-03,  9.6004e-04, -1.5757e-03,  2.9651e-03, -6.9513e-03,
        -5.5138e-03,  3.4909e-03,  2.4670e-03, -1.8119e-03,  1.3430e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.2019e-03,  9.1264e-01, -1.7524e-02,  3.1810e-04, -7.0422e-03,
        -4.6068e-03, -3.7823e-04,  4.4529e-03, -1.1920e-03, -1.5342e-03,
        -3.1508e-03,  7.8354e-04,  2.1540e-03, -4.2063e-03,  3.0852e-03,
         1.7652e-03, -3.5708e-03, -2.7784e-03,  2.8754e-03,  7.2224e-03,
        -1.5332e-03, -7.9890e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0231e-02,  9.1791e-01,  7.0116e-03, -1.0269e-04, -3.1614e-03,
        -1.4377e-03, -1.4233e-03,  7.3086e-04, -1.0885e-02,  4.8389e-04,
        -4.2248e-03,  2.7708e-03, -9.5500e-04, -2.0522e-03,  5.0778e-03,
         1.3252e-03,  2.6890e-03,  6.6841e-04, -4.2409e-04, -9.6965e-04,
         4.2079e-03,  1.6479e-03,  1.2486e-03,  4.9710e-04,  6.2344e-03,
        -1.8501e-03,  6.2196e-04, -3.0629e-03,  1.0824e-03, -5.0073e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0110,  0.8681, -0.0355, -0.0066, -0.0031, -0.0012, -0.0032, -0.0074,
         0.0022, -0.0139, -0.0478,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4276e-02,  9.1919e-01, -8.9123e-04, -2.9671e-03,  1.2306e-02,
        -5.3251e-03,  3.4873e-04, -3.7295e-03,  5.5758e-03,  2.0076e-04,
         6.1455e-03,  1.7999e-03,  2.2922e-03,  3.2502e-03,  3.1131e-03,
         2.4671e-03, -2.6448e-03,  1.8389e-03,  3.5943e-04, -8.6938e-04,
         3.0979e-03, -7.3119e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6135e-02,  8.7175e-01,  1.7314e-02, -4.9208e-03,  2.2039e-03,
         8.3825e-03,  4.7468e-04,  4.4313e-03, -1.0595e-03,  1.0249e-03,
        -3.6815e-03,  4.5869e-03,  1.1958e-03, -3.9090e-03,  1.7857e-03,
        -9.1751e-04, -9.6756e-04,  2.9146e-03, -1.3932e-04, -6.7603e-04,
        -2.8370e-03, -2.3415e-03, -8.5395e-03, -1.4088e-03, -1.0045e-04,
        -1.2182e-03,  1.5556e-03,  5.3354e-03, -2.1394e-03, -6.2434e-04,
         6.4182e-03, -5.7130e-03, -3.2944e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2228e-03,  9.0436e-01,  1.7714e-02,  1.5491e-02,  4.0407e-03,
         5.9922e-04,  4.4648e-03, -6.5824e-03,  1.4767e-03,  4.2657e-04,
        -1.2088e-03, -3.0039e-03,  1.4027e-03,  3.8692e-03, -3.7189e-03,
        -7.9752e-04,  4.4043e-03, -3.2521e-03,  1.7222e-04, -5.6483e-04,
        -2.6506e-04, -2.1499e-03,  3.7568e-03,  2.5831e-03, -2.0335e-03,
        -5.7688e-04,  3.1697e-03,  3.6904e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.7068e-03,  9.2090e-01,  2.6037e-03,  8.9894e-04, -4.6792e-04,
         5.4776e-04, -1.3560e-03,  5.2466e-03, -2.1020e-02, -1.2819e-03,
        -3.0831e-03,  2.2028e-03, -5.1177e-03, -6.9449e-04, -1.1503e-03,
        -2.9154e-04, -8.8969e-04, -6.3519e-03, -5.3652e-05, -1.1381e-03,
         4.7944e-04, -3.5435e-03, -4.5941e-03,  2.1622e-03, -6.2146e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 4.1489e-04,  8.8485e-01,  1.1573e-02,  1.1171e-02,  8.7485e-04,
        -3.5908e-03,  4.5442e-03,  6.5917e-04,  1.5490e-03, -2.8396e-03,
         7.0391e-03,  2.6127e-03,  4.4491e-04,  5.5714e-04, -5.6023e-04,
        -1.3069e-02, -1.1432e-03,  1.2930e-03,  5.7668e-04, -5.8369e-03,
        -2.4760e-04, -1.1149e-03,  7.3556e-03,  1.2480e-03,  1.0501e-03,
         8.0009e-05, -1.3397e-03,  1.6797e-04,  2.8541e-04,  2.6036e-03,
        -4.3074e-03, -1.0551e-03, -3.8283e-03, -1.3528e-03, -2.1092e-03,
         1.1621e-03, -4.7643e-04,  1.3746e-03,  8.8671e-04, -2.9320e-03,
         9.8187e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1092e-02,  9.1200e-01,  1.0517e-02, -1.1756e-02,  2.6725e-04,
        -6.7057e-03, -9.3987e-04, -1.0869e-03, -2.7603e-03, -5.2296e-04,
         1.9378e-03, -2.2824e-03,  1.5395e-03, -8.5947e-04,  7.8610e-03,
        -2.1921e-03, -7.8094e-04, -2.5555e-03,  1.0651e-03,  3.2413e-04,
        -4.9355e-03, -4.8173e-03,  5.6885e-04, -6.5848e-03,  4.0448e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8695e-02,  8.8174e-01, -1.0621e-02, -2.7637e-03,  6.4786e-03,
         1.0234e-02, -8.2703e-03,  5.0535e-03,  3.6234e-03, -1.3530e-03,
         3.3275e-03,  7.4875e-03,  1.0999e-03, -2.2932e-03,  3.7845e-03,
         1.8190e-03,  7.6598e-03,  2.6442e-04, -4.4394e-03,  3.7417e-03,
         5.2545e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4356e-02,  8.8953e-01,  2.0493e-02, -8.5092e-03, -5.0298e-03,
        -2.8288e-03, -5.6415e-04, -2.6988e-03, -7.3104e-03, -8.8461e-04,
         1.1318e-02, -1.2051e-02,  9.6858e-03, -4.3142e-03, -4.0691e-03,
        -1.6101e-03, -2.9329e-03,  1.8185e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4234e-03,  9.1738e-01,  5.7293e-03,  4.8483e-03, -3.3033e-04,
         3.1855e-04,  1.7670e-03,  2.3953e-04, -2.4541e-03,  1.4865e-03,
         3.4848e-03, -2.7362e-03,  8.2848e-04,  7.1463e-04,  1.3462e-03,
         6.8653e-03, -1.4332e-03,  2.9274e-05,  2.6561e-03, -3.7071e-03,
         2.7010e-03, -1.2466e-04,  2.0402e-03, -1.2423e-03,  1.6856e-03,
        -5.2548e-03, -2.2942e-03, -1.8137e-03, -1.1855e-03, -3.9716e-03,
         7.0620e-04,  7.5950e-04, -2.9898e-03, -1.1529e-03,  6.1446e-03,
         1.7292e-03, -2.2533e-03, -1.1749e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1594e-02,  9.2985e-01, -1.5993e-02,  4.0364e-03,  4.4979e-03,
        -3.0371e-03, -6.5853e-03,  7.1370e-04,  1.9865e-03,  3.9618e-05,
        -1.5602e-03, -1.2821e-03, -4.5538e-04,  5.4973e-03,  4.5560e-03,
         6.3369e-04, -3.1535e-04, -5.7917e-03,  5.0827e-04, -1.0643e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.7223e-03,  9.3399e-01, -1.9451e-03,  1.1008e-02,  2.4217e-03,
        -8.1155e-03,  6.0039e-03, -6.2419e-04,  2.3044e-04,  1.1146e-02,
         1.0375e-02, -3.7721e-03,  1.9018e-03,  7.4530e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1077e-02,  9.1911e-01, -7.3154e-03, -4.3251e-04,  5.1681e-03,
         5.2616e-04,  6.3299e-04, -7.7056e-04,  2.6995e-03,  1.9123e-03,
         2.2184e-03, -4.6207e-03,  3.7177e-03,  9.7811e-05,  2.8917e-04,
        -3.5990e-03,  3.8946e-03,  2.3381e-03,  3.1854e-03, -1.4545e-03,
        -4.9468e-03,  4.2252e-03,  6.3370e-04,  2.8023e-03,  6.1440e-04,
        -4.5405e-03, -1.0233e-03,  5.5463e-05,  5.6180e-04,  7.9477e-04,
        -1.4158e-03, -3.3294e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9543e-02,  8.7900e-01, -4.6514e-03,  2.6290e-02, -3.9425e-03,
        -2.8351e-03,  4.4085e-03, -1.2875e-02,  3.3802e-03,  2.3017e-03,
        -9.7249e-04, -2.8237e-03,  7.0882e-04,  3.7543e-03,  1.0940e-02,
        -9.1068e-04,  5.9145e-04,  2.6039e-03,  5.2802e-04, -1.6936e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.7961e-03,  8.7198e-01,  1.8956e-02,  6.8078e-03,  1.1863e-03,
         1.3391e-03,  1.5459e-03, -3.9493e-03,  2.9209e-03, -6.0513e-03,
         1.3740e-03,  2.4960e-03, -7.8576e-05, -1.0845e-03, -1.4823e-03,
        -5.0332e-03, -1.0506e-03,  5.8946e-04, -4.0667e-03, -9.8713e-04,
         7.0674e-04, -2.7105e-03, -3.4482e-03,  2.7783e-03,  1.2281e-03,
        -3.3586e-03, -6.0857e-03, -5.6660e-04,  7.5214e-04,  3.3889e-03,
        -1.8343e-03,  1.6016e-03, -1.6755e-03,  3.8985e-03, -3.6251e-04,
         1.6093e-03,  2.3953e-03, -2.8755e-03,  1.0787e-03,  2.2429e-03,
         1.3501e-03,  3.6489e-04,  4.5684e-03, -4.9986e-04, -1.0741e-03,
        -2.9163e-03,  2.8571e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.5653e-03,  9.3073e-01, -1.5181e-02,  5.0611e-03,  8.7200e-03,
         1.1934e-03, -5.1500e-04, -3.4709e-03, -2.1122e-03,  5.1466e-03,
        -7.7895e-03, -4.9181e-03, -3.7470e-04,  3.6810e-03, -2.5384e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6248e-02,  8.6872e-01,  6.6756e-03,  3.2904e-03, -1.7899e-03,
        -2.9089e-03, -4.8884e-04, -1.6396e-03,  4.1875e-03,  4.1316e-03,
         3.0988e-03,  1.5016e-03,  1.5664e-05,  6.2615e-03,  6.8850e-04,
         1.4268e-03,  2.9298e-03,  1.0577e-03,  7.8701e-03, -8.5506e-04,
         8.8173e-04, -5.4566e-03,  7.3255e-04,  5.0693e-03,  4.9933e-04,
        -2.4920e-03,  3.7006e-03,  5.2430e-03,  5.4113e-03, -2.2094e-04,
        -1.1563e-03, -1.2399e-03,  3.6248e-04,  8.9376e-04,  5.9822e-04,
        -1.3235e-03, -1.0917e-03, -9.2818e-04, -3.4098e-03,  1.0028e-03,
        -1.1209e-03, -5.5507e-04, -8.0464e-04, -1.6098e-04,  3.0901e-03,
        -6.3042e-04,  3.6311e-03,  1.2511e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 2.2846e-02,  8.7842e-01,  8.0738e-04, -2.7266e-03,  7.8995e-03,
        -1.2022e-02,  4.3915e-03, -8.5772e-04, -1.1901e-03, -8.9180e-03,
         1.0796e-03, -6.1850e-03, -4.6956e-03, -2.3462e-04, -5.7835e-04,
        -3.8492e-04, -2.0210e-02, -2.3113e-03,  6.8461e-04, -2.5682e-03,
        -5.3787e-03, -1.6374e-04, -1.1460e-04, -1.6886e-03,  5.3399e-03,
        -8.3074e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.7128e-03,  9.3258e-01,  5.9694e-03, -1.2041e-03,  4.2973e-03,
        -5.8163e-03,  3.2912e-03,  7.5825e-03, -2.3884e-03,  1.4435e-03,
         6.5073e-03,  1.0991e-03,  1.9247e-04,  1.6941e-03,  1.1127e-03,
        -1.3878e-03, -2.0353e-03, -3.4680e-03,  7.2616e-04,  6.4044e-03,
        -2.0870e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4514e-02,  8.9188e-01, -6.5057e-03, -2.1055e-02,  2.9440e-03,
        -1.2407e-03, -9.2847e-04,  5.5076e-04,  4.7124e-04, -9.8013e-04,
         6.8106e-04,  1.8875e-03, -2.9795e-03, -4.3634e-03,  5.3873e-03,
        -2.2138e-03, -1.8035e-03, -2.4733e-03,  8.7887e-04, -2.7665e-03,
         2.7408e-03, -2.4774e-03,  6.0161e-05, -3.1388e-04,  8.4089e-04,
        -7.5307e-04,  2.4475e-03, -1.2622e-04, -2.4480e-04,  7.4603e-04,
        -1.0733e-03, -1.9352e-04,  1.2694e-03, -3.1799e-03, -4.2312e-03,
        -2.8014e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0755e-02,  9.0806e-01, -1.2650e-02, -1.0011e-03,  1.0413e-02,
         9.3649e-03, -9.5227e-04,  3.0291e-03,  4.6077e-03,  4.8760e-03,
         4.2147e-03,  2.3422e-03, -4.2765e-03, -7.1579e-03,  6.3895e-04,
         3.8491e-04,  2.4549e-03,  1.5750e-03,  5.0417e-03, -6.2087e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3293e-02,  9.2486e-01,  5.3998e-03,  4.4930e-03,  4.1936e-03,
        -9.3165e-04, -1.7945e-03,  1.0762e-04, -1.3969e-03, -2.6854e-03,
         5.3692e-04, -4.4025e-03, -5.8542e-03, -6.2815e-03, -6.9955e-05,
        -1.6047e-03, -3.7955e-03, -3.6957e-03, -7.3319e-03, -4.5012e-03,
         2.7714e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5895e-02,  9.1893e-01, -1.3533e-02,  6.9294e-03,  2.8349e-03,
        -3.5731e-03, -5.3439e-03, -6.1105e-03,  3.7524e-03,  7.3735e-04,
         4.9098e-04,  4.6869e-03, -7.1841e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3071e-02,  8.8441e-01,  9.4440e-03, -2.4726e-03, -1.1956e-03,
        -2.1864e-03, -8.1257e-03, -1.0395e-03, -6.1801e-03, -3.2541e-03,
        -3.4576e-03, -2.8473e-03, -3.2302e-03, -6.8814e-03,  5.0321e-03,
        -1.4077e-04,  1.6957e-04,  5.1484e-04, -1.4411e-03, -1.5135e-03,
         4.1138e-03,  4.7436e-04, -5.7852e-03, -2.4516e-03, -1.2355e-03,
        -3.9357e-03, -5.2336e-03, -3.2130e-03, -2.6452e-04,  2.9250e-04,
        -1.7127e-04, -8.5230e-04, -1.6784e-03, -3.2383e-03,  2.8099e-03,
         5.7805e-03, -1.8572e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4655e-02,  8.9084e-01, -5.4459e-03, -5.2159e-03,  5.5011e-03,
         1.5171e-03,  1.5642e-03,  1.6743e-02, -1.7354e-03, -4.4203e-04,
         1.0410e-03,  3.4165e-03,  1.5218e-03,  3.0813e-03,  4.7681e-03,
        -1.3915e-03,  8.6380e-03,  1.5055e-03, -1.0873e-03,  5.2881e-03,
         2.2619e-03, -7.8976e-04,  1.9010e-02, -2.5406e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8174e-02,  8.8880e-01,  2.3836e-02,  2.5393e-03,  1.5690e-03,
        -2.7739e-03, -4.2781e-04,  1.3828e-03,  4.6778e-03, -1.5300e-03,
        -4.2825e-03,  5.4154e-05, -2.8222e-03,  5.2898e-03,  3.0311e-03,
        -3.7775e-03,  4.5735e-03,  5.6878e-03,  2.3847e-03,  5.8033e-03,
         5.6144e-04, -2.0620e-03,  3.8097e-03, -8.1748e-04,  9.7486e-04,
         9.8790e-05,  1.6345e-03, -5.6853e-05, -2.2975e-03,  1.3865e-03,
        -4.6309e-04, -2.4202e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5890e-02,  8.6550e-01, -4.9869e-03, -7.3214e-03, -4.6969e-03,
        -2.7427e-02,  9.8462e-04, -6.1442e-03, -1.2754e-03,  2.2915e-03,
         1.2720e-03, -5.9826e-04, -1.4365e-03, -3.9215e-04, -1.6473e-03,
        -1.2738e-03, -2.5083e-03, -2.2185e-04,  7.3823e-04, -1.8326e-03,
        -8.7321e-04, -9.1622e-03, -1.5359e-03, -2.1438e-03, -4.6030e-03,
         5.1255e-03,  1.2860e-03, -3.7343e-03,  3.6260e-03,  5.9791e-03,
         3.4942e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4412e-04,  9.1789e-01, -8.2127e-03,  4.8186e-03,  3.1256e-03,
         3.8432e-03,  9.7169e-03, -7.9407e-03,  1.4111e-03,  1.5739e-03,
         2.2816e-03,  2.5520e-03, -2.3427e-03, -5.3450e-03,  6.4858e-03,
         1.7516e-03,  2.0468e-03,  5.0361e-03,  1.5340e-03, -7.6198e-04,
         6.7056e-04,  3.5335e-03, -6.6813e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5634e-03,  9.2167e-01, -2.6312e-02, -2.8754e-03,  1.3709e-04,
         3.3953e-04, -5.3402e-03, -2.3420e-03, -9.2430e-04,  1.5037e-03,
         2.1383e-03, -1.5992e-04,  6.8116e-04, -1.1809e-03, -4.6959e-04,
        -3.3663e-04, -3.8539e-03, -4.6492e-03, -3.0455e-03,  1.4104e-03,
        -3.7534e-03, -5.9699e-03,  3.2379e-03, -1.0737e-03,  1.4689e-03,
         6.6054e-04,  9.6658e-05, -1.8030e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-1.0223e-02,  9.0898e-01,  2.8173e-03,  3.5814e-03, -4.7549e-03,
        -6.4093e-03,  3.8105e-03, -9.0056e-04,  2.3850e-04, -2.3486e-03,
         7.4197e-04, -4.1004e-03,  2.6571e-03,  2.8289e-03, -2.1148e-03,
         2.8696e-03, -1.0449e-03, -5.6547e-03, -3.2132e-03,  7.8267e-03,
         5.4723e-03, -1.3683e-03, -8.8638e-04, -2.9261e-03,  1.0345e-03,
         6.1309e-03,  1.1831e-03,  1.9817e-04, -3.6875e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.1020e-03,  9.2479e-01, -1.1694e-02,  1.2057e-03, -4.0166e-03,
         1.7231e-03,  2.9656e-03, -1.7085e-03,  1.1583e-03, -1.3774e-03,
        -5.8148e-03, -3.9345e-03, -2.5500e-04,  2.8619e-03,  2.2512e-03,
        -1.0523e-03,  1.9681e-03, -2.5157e-03,  5.3985e-03, -6.7367e-04,
         8.7622e-04,  9.9853e-04, -6.5774e-03,  7.0788e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2438e-04,  8.9295e-01,  1.7838e-02, -8.8761e-03,  1.0561e-02,
         6.6381e-03,  8.3230e-05, -1.4921e-03, -1.5948e-03,  2.6276e-03,
         5.2986e-03, -2.3615e-03,  3.6678e-03, -1.8323e-03, -7.9345e-04,
         8.5647e-05,  1.5306e-05,  1.5869e-03,  9.6422e-04,  1.5122e-03,
        -2.0787e-03, -8.8798e-05, -3.2276e-04,  5.0362e-03, -2.3116e-03,
        -2.4177e-03, -8.9896e-03,  6.1383e-04, -1.0095e-03,  2.4352e-03,
         1.6892e-03, -4.2325e-03, -4.9382e-03,  2.7295e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3245e-03,  9.3926e-01, -1.4237e-02,  6.6521e-03,  2.7569e-03,
        -3.2075e-03,  4.7412e-03,  1.8122e-04,  3.9179e-03,  4.5902e-04,
         1.4677e-04,  1.7779e-03,  1.2996e-03,  8.5777e-04,  1.5993e-03,
        -5.6061e-05, -2.5817e-03,  8.4848e-04, -3.3975e-04,  1.7640e-03,
         6.9311e-03,  5.0615e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1140e-02,  9.2980e-01,  6.1677e-04,  4.7816e-03,  4.8473e-03,
         7.3505e-04,  1.5499e-03, -5.8450e-04, -1.8989e-03, -4.4582e-04,
         1.5630e-02, -2.0239e-03, -4.4019e-03,  1.8170e-03, -6.6484e-04,
        -2.0314e-03,  5.6549e-03,  1.1380e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2645e-03,  9.3551e-01, -2.9253e-03, -4.8156e-04, -1.2097e-03,
        -1.7525e-03,  4.6984e-05, -3.2040e-03, -5.2792e-03, -4.5945e-03,
        -1.6736e-03,  1.8964e-05, -1.9171e-03,  6.3460e-03, -2.6389e-04,
         1.9418e-03, -7.9297e-03, -1.5032e-03, -5.0560e-03, -3.3806e-03,
         4.7209e-04,  2.4391e-03,  8.7880e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2618e-02,  9.3030e-01,  3.3572e-03, -1.0407e-02, -1.2102e-02,
        -1.9271e-03,  8.5898e-04,  5.0503e-04, -9.3893e-03,  1.0761e-03,
         3.3970e-04, -2.5274e-04, -4.9393e-04, -3.7749e-04,  1.8822e-03,
         2.1187e-03, -1.9987e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.4387e-03,  9.5449e-01,  1.5688e-03,  3.5186e-03,  1.5868e-03,
        -1.9144e-03,  6.9850e-03, -8.3598e-04,  6.0043e-03, -1.2694e-03,
        -1.5587e-03,  9.5821e-04, -7.8078e-03, -3.0371e-04,  2.0886e-03,
         1.1788e-04, -2.0534e-03,  1.4960e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7716e-02,  8.4921e-01, -1.9457e-02, -9.3349e-03, -4.0294e-03,
        -5.8859e-03, -1.2747e-03, -1.8554e-03, -5.3104e-03,  2.5849e-02,
         3.1559e-03,  1.2268e-03,  2.4699e-03, -1.7001e-03, -9.8899e-04,
        -2.3527e-03, -1.1094e-02,  2.6437e-03,  9.0496e-04,  3.2437e-04,
        -1.1820e-04, -3.3038e-03,  2.9087e-03,  3.2144e-03,  6.5534e-03,
        -1.7119e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6557e-02,  8.5566e-01, -8.0211e-03, -3.4180e-03,  2.0476e-03,
        -3.3162e-03,  2.9738e-03,  5.3585e-03,  1.1937e-02, -1.5825e-02,
         1.0115e-03, -4.2083e-03, -1.6121e-03, -9.5839e-03,  4.0812e-04,
         2.9990e-03, -6.5394e-04, -2.4968e-03,  2.2096e-03, -4.6587e-03,
        -1.4585e-03, -2.0012e-03,  9.0322e-03,  7.7796e-03, -8.7204e-04,
         1.2774e-03, -1.3033e-03,  1.1082e-03, -2.6659e-03,  6.2315e-03,
         1.1312e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1647e-02,  8.8266e-01,  4.8410e-03,  4.2858e-03,  5.1120e-03,
        -5.6788e-03, -3.2704e-03,  1.0767e-02,  6.2754e-03, -8.6749e-03,
        -8.1135e-04, -3.6975e-03, -7.9529e-03,  3.0685e-03, -1.5831e-02,
        -2.0955e-03,  1.3522e-04,  1.3195e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2203e-02,  9.0924e-01,  8.7840e-03,  9.4230e-04,  7.2578e-03,
        -1.1315e-02, -4.6132e-04, -1.9797e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 2.8758e-02,  9.0939e-01,  6.7136e-03, -5.1435e-04, -1.1063e-03,
         2.7972e-03, -2.2969e-03, -1.3349e-03,  9.5550e-04,  4.5539e-04,
         1.6970e-04, -1.5336e-03,  3.1965e-03,  1.8114e-04,  2.5176e-03,
        -5.4366e-04, -9.5086e-04, -1.8452e-03,  1.2866e-03, -5.1784e-04,
         1.0300e-04,  1.6204e-03, -9.6347e-04, -5.4949e-04, -1.0545e-03,
        -1.6669e-03, -8.6135e-04,  1.7064e-03,  3.3411e-03,  6.5025e-04,
        -1.9899e-03,  2.0182e-03, -9.4710e-04, -1.1150e-03,  1.6437e-03,
        -3.5938e-05, -1.2305e-04,  3.5872e-03,  8.9542e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5243e-03,  9.1294e-01, -1.1456e-02, -3.6550e-03,  1.4787e-04,
         1.4501e-03,  3.1908e-04, -5.7294e-04, -7.2698e-03,  8.4437e-04,
        -1.6287e-03,  4.6427e-04,  9.9837e-03, -3.2346e-03,  1.7066e-03,
        -3.1243e-03, -1.5068e-03, -5.6311e-03, -4.5748e-05, -1.6767e-03,
        -3.6131e-03, -3.0983e-03,  7.7682e-04, -1.2191e-03, -8.6621e-04,
         3.3270e-03, -9.8249e-03, -5.0929e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0894e-02,  9.3465e-01,  4.0870e-03, -7.7083e-03,  4.1620e-03,
         6.1978e-03, -2.8057e-03, -1.0775e-03,  1.5002e-03,  2.6963e-03,
         4.7303e-03,  7.9292e-04,  8.6969e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5770e-03,  9.4594e-01, -1.3512e-02,  5.2461e-03,  2.4992e-03,
         2.2452e-03, -1.2759e-03, -5.3881e-04,  7.3249e-03, -7.0042e-04,
         4.1873e-03, -5.8158e-03, -1.2481e-03,  5.6695e-03, -2.2328e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9287e-04,  8.9977e-01,  2.1282e-02,  3.6195e-03, -1.7883e-03,
         8.5442e-04,  5.1329e-03, -2.7030e-03,  5.4049e-04, -9.7156e-03,
         7.2178e-04,  1.8528e-03, -1.1033e-02, -9.4441e-05, -2.7352e-03,
        -3.6038e-04,  7.7128e-04, -2.0577e-03,  1.8942e-05, -2.6209e-04,
         3.3513e-03,  2.5053e-04, -5.6979e-04,  1.7996e-04, -3.4606e-03,
         1.6336e-03, -1.3557e-03,  3.6949e-03, -3.8682e-04,  4.1227e-04,
         1.4148e-03, -2.3911e-04,  1.0445e-03,  1.0905e-03,  9.2816e-04,
         6.6211e-04, -5.5380e-04,  2.7576e-04,  2.6493e-04,  1.3274e-03,
         5.4280e-03,  1.5437e-04, -1.4951e-04,  2.3405e-04, -1.9828e-03,
         2.5996e-03, -7.5268e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2870e-03,  9.0812e-01, -8.9956e-03,  1.7181e-03, -1.0036e-02,
        -1.3494e-02, -1.9670e-03,  3.8067e-04, -9.2675e-03,  7.8599e-04,
         1.0050e-03,  1.7957e-04, -2.4641e-03, -6.9772e-04, -5.0052e-04,
        -2.1437e-03,  1.4522e-03,  1.9039e-03, -1.2762e-03, -1.0335e-03,
        -1.3109e-03, -2.0649e-03, -3.2106e-04, -2.7796e-03, -2.9529e-04,
         2.6835e-04,  7.3194e-04, -1.5141e-03,  6.4860e-04,  8.5248e-05,
        -3.4007e-03,  1.4224e-04,  3.3306e-03, -4.1596e-04,  9.6045e-04,
        -8.2875e-04, -2.5851e-03, -1.9668e-03, -1.1861e-03, -6.3081e-04,
         2.5509e-03,  1.2722e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1237e-02,  8.8158e-01,  2.0842e-02, -3.2573e-03,  1.8215e-03,
        -1.0738e-02,  3.1106e-04,  2.0746e-03,  2.4723e-03,  8.8874e-03,
        -5.7378e-03,  2.9638e-03, -2.5393e-03, -7.5939e-04,  2.1928e-03,
        -9.2875e-03,  2.1319e-03,  1.2751e-03, -1.6413e-03,  1.9451e-04,
        -1.1943e-03, -2.9360e-03, -8.1474e-03, -5.1049e-03, -6.6774e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.9740e-03,  8.8071e-01,  2.5526e-02, -1.3745e-03, -3.4341e-03,
        -3.2429e-03,  3.7523e-03, -6.8983e-04, -6.0646e-03,  1.4463e-03,
         2.8970e-03, -9.9728e-03, -2.0306e-03, -1.0837e-02, -3.1675e-03,
         6.5014e-04,  8.3142e-04,  1.1649e-02, -2.4477e-04,  5.7579e-03,
         5.7997e-03,  3.9636e-03, -8.9808e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2212e-02,  9.0184e-01,  1.6549e-02,  8.4236e-03,  5.1510e-03,
         1.8427e-03,  2.2014e-03, -3.5752e-04, -4.9475e-03,  3.6087e-03,
         2.9170e-03,  2.9294e-03, -2.1004e-04, -1.4964e-03, -7.8500e-04,
         1.3453e-03,  2.2521e-04,  4.8458e-03,  1.3476e-04, -3.0552e-03,
        -2.5023e-03, -6.6087e-03,  5.8127e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0408,  0.8826, -0.0111, -0.0032, -0.0012,  0.0092,  0.0046,  0.0203,
         0.0035,  0.0014, -0.0161,  0.0014,  0.0045,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5364e-02,  8.9926e-01,  2.1461e-02, -1.2852e-02, -1.6979e-03,
        -7.0882e-03, -5.3839e-04, -1.5346e-03, -3.0651e-03,  1.6472e-04,
        -8.8938e-03, -1.1764e-02,  2.4425e-03, -3.8687e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1527e-02,  9.1450e-01,  1.2656e-02, -4.6602e-03, -3.8865e-03,
        -1.3433e-03,  9.2107e-04, -1.1670e-03,  2.1644e-03, -3.5530e-03,
        -1.8786e-03, -4.4263e-04, -1.0768e-02, -2.9489e-03, -1.6480e-03,
        -5.5977e-05, -5.7750e-03,  5.4210e-03,  4.6876e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 1.3596e-02,  9.1380e-01,  1.6550e-03, -1.4118e-03, -5.0648e-03,
        -6.0020e-03, -1.0865e-03, -8.2568e-03, -1.0194e-03, -1.9209e-03,
        -1.5399e-04, -1.3104e-03, -2.0546e-04,  8.4184e-03, -2.7723e-03,
        -5.6273e-03,  2.9022e-03, -1.2947e-03,  2.1199e-03,  1.8753e-03,
        -1.2912e-03, -1.8367e-03,  5.5717e-04,  4.2954e-03, -4.0061e-03,
        -7.5248e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0158,  0.9058, -0.0097, -0.0112, -0.0018,  0.0036, -0.0065, -0.0060,
         0.0060, -0.0089, -0.0025, -0.0010, -0.0029,  0.0069,  0.0040,  0.0074,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0177e-02,  8.5236e-01,  4.0761e-02,  2.3749e-02,  1.3550e-03,
         1.6576e-02,  1.9690e-03, -3.5197e-03, -4.7550e-03,  4.8562e-04,
        -9.4180e-05, -1.7815e-03,  1.1983e-03,  3.2222e-03,  7.1342e-03,
         9.3317e-03, -1.1529e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0066,  0.9302,  0.0201,  0.0088, -0.0036, -0.0078,  0.0044,  0.0151,
         0.0035,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8200e-02,  8.4217e-01, -7.3185e-04, -9.4982e-03, -1.0302e-02,
         5.2087e-03,  5.4613e-03,  4.9314e-03,  5.8949e-03, -8.0158e-04,
         2.4000e-03,  8.9386e-03,  5.4209e-05,  4.1432e-03, -2.1904e-04,
         2.9333e-03,  5.2425e-03,  2.9362e-03,  1.3299e-03,  3.1111e-03,
         1.2289e-03,  2.6240e-04,  1.9666e-04,  5.3903e-04, -8.9555e-04,
        -4.3140e-03, -5.1007e-04, -3.2840e-03,  4.9441e-04, -1.2040e-03,
         6.1906e-04, -5.9666e-04,  1.5259e-03, -3.0273e-03, -3.5264e-04,
         1.3611e-03, -3.2858e-03, -5.1200e-04, -5.2752e-03, -2.5356e-03,
        -2.3821e-03, -6.3158e-03, -4.0539e-03, -2.8490e-03,  6.9043e-03,
         1.0971e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2937e-03,  9.1147e-01,  9.1256e-03, -3.2417e-03, -6.5147e-03,
        -1.3596e-03, -4.8797e-03, -7.7712e-05, -3.5410e-03, -3.4851e-03,
        -3.6169e-03, -5.6387e-04, -8.9407e-03, -4.0673e-03,  7.4173e-04,
        -1.8500e-03,  1.9440e-03, -2.0753e-03, -4.1269e-04,  1.2133e-03,
         5.8349e-03, -7.3433e-04, -4.2290e-03, -1.2196e-05, -1.6998e-03,
        -2.6813e-03,  1.5425e-03, -3.0937e-04, -1.4144e-03,  4.1240e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3820e-02,  9.5371e-01,  1.1172e-02,  8.3972e-03, -7.5176e-04,
         1.4731e-03,  6.7081e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6830e-03,  9.5549e-01,  3.3224e-03, -4.7755e-04, -1.3799e-03,
         4.8766e-05,  7.9883e-03,  7.1320e-03,  1.2154e-02, -6.4871e-03,
        -3.8404e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8872e-02,  8.9182e-01,  1.7460e-02,  6.9775e-03,  4.4258e-03,
        -5.2880e-03,  3.7939e-05, -3.7110e-03,  6.5984e-03, -6.9180e-04,
        -3.7133e-03,  4.3913e-03, -1.0127e-02,  5.8910e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8872e-02,  9.1197e-01,  2.9425e-03, -4.9637e-04, -2.6726e-03,
        -2.7639e-04,  2.4743e-03,  2.2568e-03, -6.7470e-04, -4.2530e-03,
        -1.5822e-04, -5.6132e-05, -1.8086e-03, -3.4092e-03, -3.3686e-04,
         3.0863e-03,  1.0580e-03,  1.6279e-03,  2.4867e-03, -2.7252e-03,
        -2.0505e-03, -5.7603e-03, -2.0404e-03, -2.0609e-03, -1.7129e-03,
         1.9987e-03, -6.3527e-04,  1.9606e-03,  4.2649e-03,  4.3373e-03,
        -1.7945e-03, -2.3893e-03, -2.7559e-03,  2.5950e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9352e-02,  8.7582e-01,  1.5864e-02,  2.1016e-02, -1.7487e-04,
         1.7481e-03, -5.0477e-03, -2.9657e-03, -2.2113e-03,  4.9068e-03,
        -2.6466e-03,  3.0936e-04,  3.3193e-03,  2.3292e-03, -2.7791e-03,
        -2.2637e-04, -6.8834e-03,  2.1364e-03, -2.3297e-03,  4.4056e-04,
         4.7338e-03,  9.8914e-04,  6.6031e-03,  9.9920e-05,  2.2326e-03,
        -4.2982e-03, -2.8242e-04,  8.2524e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0905e-02,  8.7132e-01,  7.8905e-03, -5.4734e-03,  1.7268e-02,
         5.3739e-03, -1.9150e-03, -1.8064e-03, -2.4440e-03, -2.3991e-03,
        -3.9467e-03, -6.9116e-03, -5.6512e-03,  3.7340e-03,  6.5700e-03,
        -3.3429e-03,  3.4953e-04, -3.2343e-03,  4.4095e-03, -9.5184e-04,
        -1.4051e-03,  1.3112e-03, -7.9756e-03, -3.4096e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 1.0462e-02,  8.8408e-01,  1.1383e-02,  1.8083e-02, -9.1563e-03,
        -1.2778e-03, -2.3301e-03, -3.5446e-03,  3.0543e-03, -1.9964e-03,
        -3.5580e-03,  4.7786e-04,  2.6418e-03,  2.5451e-03,  1.5411e-03,
         3.7355e-04, -5.9480e-03,  2.0695e-03, -1.6507e-03,  6.3736e-03,
        -4.9312e-04,  4.0174e-03, -4.7800e-04,  1.0552e-04,  7.4647e-04,
         7.5573e-03, -9.3640e-04, -6.0301e-03, -3.3547e-03, -2.5365e-03,
         1.1954e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9731e-02,  9.2868e-01, -5.6366e-03,  9.5181e-04, -3.6959e-03,
         1.3338e-02,  1.9604e-03, -3.8431e-03,  5.7834e-03,  8.2880e-04,
         8.5119e-03, -7.0387e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0413e-03,  9.3575e-01, -4.3257e-04, -1.1987e-03,  1.5535e-02,
         8.1028e-04, -4.2278e-04,  2.5129e-03,  7.6948e-03,  1.4878e-02,
         4.0067e-03, -1.8074e-03, -2.5620e-03,  2.6308e-03, -5.7169e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7021e-03,  9.1156e-01, -2.7954e-03,  3.9493e-04, -2.7379e-03,
         1.1753e-05,  4.2952e-03,  2.1141e-03,  2.2461e-03,  1.4280e-03,
         2.7830e-03,  3.5600e-04,  8.6613e-03, -1.0622e-03, -4.6034e-03,
        -1.4761e-03,  1.0588e-03,  5.7627e-04,  1.7023e-03, -5.5683e-03,
        -5.7774e-04, -8.8147e-04, -7.0210e-03, -2.2174e-03,  5.6622e-03,
         4.4398e-03, -8.5145e-04, -1.1689e-03, -4.9478e-03,  4.4889e-04,
         1.0695e-03, -1.5242e-03,  9.0547e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3589e-02,  8.8699e-01,  1.0837e-02, -6.2098e-03,  1.0639e-02,
        -3.0403e-03, -4.3134e-03, -5.2320e-03, -7.8969e-03, -4.1052e-03,
        -1.0785e-02,  8.5019e-05, -9.4236e-03,  5.5710e-03, -7.4624e-03,
         1.3819e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4634e-02,  9.1849e-01,  1.3027e-02,  1.5932e-02,  1.2639e-04,
        -8.0488e-03, -9.8671e-03, -2.8364e-03,  1.7038e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0021,  0.9108,  0.0190,  0.0162,  0.0519,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3002e-02,  9.1799e-01,  1.5443e-04, -5.0740e-03,  2.5909e-03,
         6.4727e-04, -8.9584e-04,  1.3904e-03, -7.4818e-04, -2.0669e-03,
        -7.4380e-04,  1.7409e-03, -4.4422e-03, -1.2126e-03,  1.4061e-03,
        -3.6765e-03,  1.6417e-03,  1.1682e-03,  1.0522e-03, -2.1120e-04,
         7.6163e-03,  7.3441e-04,  1.4786e-05, -8.8389e-04,  1.0836e-03,
         3.4560e-03, -4.3554e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0513,  0.8991,  0.0165,  0.0050,  0.0021,  0.0117,  0.0142,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5862e-02,  8.9197e-01,  4.2829e-03,  1.9705e-03,  4.2850e-03,
         1.8189e-03, -5.2697e-03,  5.7644e-04,  5.8114e-03,  9.3724e-03,
         3.9336e-03, -7.5211e-03, -2.1150e-03,  3.9306e-04, -3.0138e-03,
         4.8132e-04,  7.7536e-04,  4.8848e-03,  1.6237e-03, -5.5826e-03,
         6.4812e-04,  7.8045e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9165e-02,  9.2740e-01, -4.4326e-03, -2.0052e-03, -3.8770e-03,
         5.0320e-03,  7.8816e-04,  6.3431e-03, -2.8186e-03, -6.0886e-03,
         4.5065e-03,  9.6382e-04, -4.6309e-03,  7.2243e-04,  2.0126e-03,
        -2.7479e-03, -1.4892e-03, -1.4730e-03, -3.5009e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1200e-02,  9.1447e-01,  6.0874e-03,  7.5940e-03, -2.0460e-03,
        -5.4472e-05,  2.3642e-03,  3.4532e-03,  5.2623e-03,  3.5849e-03,
         3.6236e-04,  3.2081e-03, -5.3780e-04,  3.7842e-03, -1.2935e-03,
         5.8700e-04,  2.6250e-04,  1.0751e-03,  3.6685e-04, -1.1510e-03,
        -1.0389e-03,  4.7486e-03,  1.7200e-03,  8.0901e-04,  4.2716e-03,
        -9.3251e-04, -1.9550e-03,  4.1003e-04,  6.4896e-04,  1.1288e-02,
        -3.4280e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-3.2259e-03,  8.5262e-01,  6.1182e-02,  1.0234e-02,  3.5117e-04,
         1.6131e-03,  2.6501e-04,  3.5448e-03,  5.6599e-03,  2.5316e-03,
         2.1724e-03,  3.2197e-03, -1.5769e-03,  6.8121e-03,  2.5172e-03,
         5.0474e-04, -5.9726e-03,  2.9296e-03,  1.2274e-03,  1.4392e-03,
        -5.4444e-04, -2.0483e-03,  1.4882e-03, -1.4280e-03, -3.0545e-03,
        -5.7006e-04,  7.4600e-04, -7.0328e-03, -2.3938e-03,  2.9600e-04,
         4.9825e-03,  5.8201e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3271e-03,  9.2538e-01, -4.3809e-03, -1.6292e-03,  5.5737e-04,
         4.1613e-03, -2.1666e-03, -4.7473e-04,  8.1387e-03, -3.0979e-03,
        -1.2084e-02,  2.2539e-03, -6.4689e-03, -3.4347e-03, -4.0660e-04,
        -9.5027e-04,  1.6840e-03, -4.0245e-03, -8.1810e-03, -2.0053e-04,
        -4.2727e-04,  5.5667e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7152e-02,  9.1377e-01,  2.6216e-03,  6.6053e-04,  5.2442e-04,
         3.7470e-03, -2.0196e-03,  1.5014e-03,  2.5036e-03,  6.5820e-04,
         1.1752e-03, -3.2668e-03, -3.3324e-03,  6.0735e-04, -1.0613e-03,
         1.2238e-04, -7.7105e-05,  1.1207e-02, -1.7212e-03, -7.4789e-04,
        -3.5257e-03, -5.5797e-03, -1.6462e-03, -2.8282e-03, -4.8393e-03,
         3.3966e-03, -1.2939e-03, -6.5008e-03,  8.0293e-04, -1.1130e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4361e-03,  9.4692e-01,  7.2038e-03, -2.7136e-03, -1.6613e-03,
        -1.3423e-03,  9.0486e-03,  2.7523e-03, -1.8856e-03, -4.1173e-03,
        -2.5083e-03,  5.0614e-03,  8.2799e-04,  1.6417e-04, -3.8854e-03,
         1.7652e-03, -5.6413e-03,  1.0688e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.9667e-03,  8.9929e-01,  1.3382e-02,  8.0866e-04,  7.9558e-04,
        -4.6010e-03,  1.7806e-04,  3.6584e-03, -2.3423e-03,  1.7570e-03,
         2.0528e-03,  3.7134e-04,  1.9663e-03,  1.2154e-03, -7.5817e-05,
        -2.0301e-03,  3.9102e-03, -5.3977e-04, -3.9717e-03,  6.3093e-04,
        -2.4787e-03,  1.5866e-03,  3.5005e-03,  8.2171e-04,  1.5104e-03,
        -1.5656e-03, -9.9110e-03,  4.6338e-05, -6.3878e-04,  6.7293e-04,
        -6.0429e-04, -2.4773e-03, -8.3334e-05, -5.8005e-03,  6.0865e-04,
        -2.8285e-03,  1.3325e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.5508e-03,  9.3477e-01, -1.0681e-02,  8.3945e-03,  2.8157e-05,
        -1.2302e-03,  4.4134e-03, -3.3216e-03, -7.7367e-04,  1.2204e-03,
         1.7574e-03, -1.6015e-03, -4.2639e-03,  3.7614e-03,  4.5938e-03,
         5.9174e-04, -4.2211e-03, -3.3024e-04,  8.7183e-04,  1.1292e-03,
        -1.4209e-03, -2.2213e-03,  4.7972e-04, -3.7660e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3079e-03,  9.2277e-01, -5.5946e-04,  8.1195e-04, -2.3776e-03,
         2.5980e-04, -1.6698e-03, -4.4957e-04,  1.3188e-03, -4.1036e-03,
        -1.7780e-03, -1.4130e-03, -2.3554e-03, -4.5462e-03,  1.5388e-04,
         6.7294e-04, -2.0618e-03, -2.5260e-03, -3.5056e-06, -6.0014e-04,
        -1.6460e-03, -4.9570e-03,  4.3410e-04, -2.8336e-03, -3.2484e-03,
         1.4954e-03,  3.0699e-04, -9.4012e-04,  6.4821e-04, -4.1198e-04,
         7.1913e-03,  1.4408e-04,  2.2193e-03,  2.2932e-03,  1.4959e-03,
        -2.5328e-04,  4.1602e-04, -7.9316e-04, -9.6298e-04, -6.0567e-05,
         1.1550e-03, -9.7918e-04, -4.2858e-05,  3.5217e-03, -1.5903e-03,
         2.1995e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7710e-03,  8.5408e-01, -1.5102e-02, -1.0400e-02, -2.1607e-02,
        -2.4866e-03, -1.4842e-02, -2.6699e-03, -1.4983e-03, -1.0360e-02,
        -3.9344e-04,  1.1520e-03, -1.5723e-04, -3.9769e-03,  4.3177e-03,
         1.1266e-03, -8.5884e-03,  2.0796e-02,  5.1404e-03,  4.0560e-03,
         6.2143e-03, -1.8041e-03,  7.4626e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8736e-02,  9.1337e-01, -1.0881e-02,  2.8511e-03,  1.5725e-02,
         4.5870e-03,  9.0668e-04,  1.1310e-02,  1.3954e-03, -5.3389e-03,
         4.8954e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0190,  0.9191,  0.0067,  0.0097,  0.0045, -0.0032,  0.0042, -0.0043,
         0.0027,  0.0039, -0.0065, -0.0057,  0.0083, -0.0023,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6166e-02,  9.0421e-01, -1.2841e-02,  8.0512e-03, -3.6462e-03,
        -6.3312e-04, -7.4192e-03,  3.0396e-03,  9.8068e-04, -1.7280e-02,
         2.4032e-04,  5.4978e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4707e-02,  8.8348e-01,  4.3637e-03,  8.4106e-03,  2.0484e-03,
         4.8557e-04, -4.2916e-03,  5.1189e-04,  8.8855e-04,  5.8732e-03,
         2.2129e-03,  9.5766e-03,  1.2549e-03,  2.0581e-03, -9.8377e-04,
         1.2847e-03, -7.3587e-04,  2.0570e-03,  2.6127e-05, -3.6478e-03,
        -9.5582e-04, -9.1436e-05,  8.8944e-04, -2.1579e-03,  4.7765e-04,
        -3.2175e-03,  2.8015e-03, -2.3371e-03,  5.8716e-04, -1.1757e-03,
         1.1855e-03,  9.2610e-03,  3.3785e-04,  1.2461e-03,  1.3799e-03,
        -1.8359e-03, -7.7171e-04,  7.2642e-04,  7.2662e-04, -1.4680e-03,
        -5.9014e-04,  4.0362e-04, -1.0215e-03, -6.2576e-04,  9.5188e-04,
         2.6557e-03,  5.7005e-03,  4.6239e-04, -4.3333e-04, -3.9567e-04,
         7.4382e-04,  3.2027e-05, -9.4862e-04, -2.2813e-04, -2.2741e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-4.8882e-03,  9.2325e-01,  6.6226e-03, -4.3967e-03,  8.8523e-03,
         2.9005e-03, -2.7313e-03,  5.3611e-03,  3.4039e-03, -1.1074e-03,
        -8.4328e-03, -2.0796e-03, -2.1346e-03,  4.8045e-04, -4.3826e-03,
         9.8308e-04, -5.2248e-03,  9.0112e-04, -1.9505e-03, -6.3643e-04,
         2.5116e-03, -4.2885e-03, -2.4756e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.2616e-03,  9.2604e-01, -1.3837e-02,  7.0274e-03,  9.4455e-03,
        -1.9765e-03, -5.4495e-03, -3.0943e-03, -8.7991e-04, -3.0555e-04,
        -6.4751e-03, -3.8355e-03, -3.3122e-03, -2.0210e-03, -2.3095e-03,
        -3.4670e-03, -2.2453e-03,  1.0196e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2297e-02,  8.7434e-01,  7.2940e-03,  7.1425e-03, -5.9357e-03,
         1.9651e-03, -8.7230e-03, -8.1770e-04,  5.4634e-03, -2.9969e-03,
         5.2712e-03,  9.4994e-03, -5.5535e-03,  9.7080e-03,  4.3484e-03,
        -2.2060e-03, -1.2361e-02,  4.0754e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.3581e-03,  9.2359e-01,  3.1516e-03, -7.6573e-03, -1.1297e-03,
         5.5573e-04, -9.2593e-03,  4.1122e-03,  4.0758e-03, -1.7710e-03,
        -7.8076e-04, -6.7231e-03,  4.2317e-03,  1.5602e-02, -1.2002e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7818e-02,  8.9902e-01,  7.5041e-03, -1.0016e-03,  8.4907e-04,
        -1.7976e-03, -5.3543e-03, -2.9594e-03, -9.2329e-04, -3.3259e-04,
        -5.4424e-03, -5.6404e-03,  1.4560e-03,  2.0142e-03,  1.8197e-03,
         4.5072e-04, -2.9474e-04, -5.8544e-03, -2.6680e-03,  8.7038e-03,
         3.4484e-03, -2.1094e-03, -6.8771e-04, -2.8827e-03, -6.5263e-04,
         1.2934e-03,  3.2607e-03, -1.4004e-03,  3.4715e-03,  2.5688e-03,
        -5.9558e-03,  3.6161e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0413e-02,  9.1168e-01,  1.5495e-04, -1.5363e-03,  2.1579e-03,
        -1.0479e-03, -4.2749e-03, -4.7508e-03, -4.2796e-03, -3.8311e-04,
        -1.5889e-03, -4.7764e-03, -9.5983e-03, -5.4517e-05, -1.7305e-03,
        -5.4542e-04, -1.7601e-03, -3.6924e-03, -5.8763e-03, -3.2457e-03,
        -3.4116e-03,  1.5889e-03, -3.8083e-03, -3.2390e-03, -1.1401e-03,
        -2.3354e-04,  9.7286e-04,  1.0506e-03, -1.8871e-03, -1.5594e-03,
         1.6051e-03, -2.9556e-03, -2.9974e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.3308e-03,  8.5695e-01, -9.3646e-05,  1.9370e-02,  1.4323e-03,
         5.8797e-05,  4.8380e-03, -6.6320e-03,  1.9954e-03,  7.9498e-05,
        -1.2011e-03, -2.3336e-03, -4.3604e-03, -4.6364e-03, -7.8165e-03,
        -3.4404e-03,  1.0226e-03, -6.8594e-04, -2.8044e-03, -8.7293e-04,
        -1.2326e-03, -1.6384e-03,  7.5442e-05,  9.3997e-04, -5.1065e-04,
         1.2868e-03, -3.4772e-03, -1.9493e-03,  1.0016e-03, -2.5592e-03,
         1.7761e-03,  2.6070e-03, -1.2109e-03, -9.3023e-04,  1.6981e-03,
        -6.2606e-03, -7.0048e-04, -1.1546e-03, -1.6117e-03, -1.4460e-04,
         1.7267e-03, -1.2274e-03, -7.3082e-04, -5.8322e-04, -9.5830e-04,
        -2.4809e-04,  5.2812e-04, -4.6004e-04, -2.0512e-03, -1.4517e-03,
         4.8271e-03,  3.1332e-04, -1.3091e-03, -9.8842e-04,  2.4098e-03,
         2.0845e-03, -1.2064e-02, -4.3126e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8066e-03,  9.0246e-01, -7.6753e-03, -8.9820e-03, -1.3752e-03,
        -3.4184e-03, -6.0951e-03,  5.3162e-03, -6.7953e-03, -9.3292e-04,
         5.0430e-03, -3.3181e-03, -3.5917e-03, -7.6372e-03,  1.4192e-04,
        -1.9570e-03,  2.0580e-03,  3.0225e-03, -3.7032e-03,  4.6701e-03,
        -3.7107e-03, -3.6471e-06,  3.4614e-03,  3.8169e-04,  5.7461e-03,
        -6.6970e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5085e-02,  8.9534e-01, -1.2149e-02,  4.5922e-03, -9.0985e-03,
         4.6172e-03,  4.8982e-03,  9.3636e-04,  2.2118e-03, -2.7319e-03,
        -2.7350e-03, -2.5432e-04,  8.4898e-03,  1.1152e-04, -6.9982e-04,
        -1.0616e-03,  1.9741e-03,  3.3419e-03,  3.2125e-03,  7.1129e-04,
        -1.5731e-03, -5.0674e-03, -2.9263e-03,  3.6522e-03,  2.5338e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2654e-02,  8.8069e-01,  9.5216e-03,  9.2209e-03, -2.0407e-03,
         1.4426e-04, -1.8708e-03,  1.7886e-03, -1.2265e-03, -3.6149e-03,
        -2.6804e-03, -8.5475e-04,  1.1545e-03,  5.8425e-03,  1.3322e-03,
        -3.5819e-03, -6.5321e-03,  2.3606e-03,  1.1960e-03,  3.9655e-03,
        -6.3056e-04,  1.4644e-03,  8.5041e-04, -3.6713e-03, -4.4074e-04,
        -3.3386e-03,  3.2920e-03,  1.1489e-04, -1.2629e-03,  1.3882e-03,
         3.6547e-05, -2.8907e-03,  4.3868e-03, -8.1701e-05, -3.5889e-03,
        -3.6292e-03, -9.9460e-04, -2.8167e-03, -3.7507e-04,  1.3381e-03,
        -1.3947e-03, -4.9037e-04,  1.6296e-03,  1.2792e-03, -6.3390e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9961e-02,  9.0398e-01, -4.8811e-03, -1.9556e-03, -1.3838e-03,
         2.3683e-03, -3.0831e-03, -1.7528e-03,  1.6330e-03,  2.7034e-03,
        -6.7727e-03,  1.1748e-03, -1.4648e-03, -2.3455e-03,  8.9013e-03,
        -9.4745e-03,  1.2373e-02, -4.3192e-03,  8.0480e-04,  9.5528e-04,
         4.4857e-03,  1.4442e-03,  1.7859e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0168,  0.9063,  0.0049, -0.0058, -0.0019,  0.0018,  0.0060,  0.0035,
        -0.0175, -0.0106,  0.0095, -0.0155,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 3.8116e-03,  9.2191e-01,  1.2837e-02, -7.5362e-03, -1.1369e-02,
         6.7595e-04,  2.9482e-03, -1.0785e-03,  3.3715e-03, -3.4865e-03,
        -1.2910e-03, -2.6644e-03,  1.9142e-03, -2.4392e-03, -3.8348e-03,
         4.5977e-03,  3.3447e-03,  4.9913e-03, -1.0833e-03,  4.0698e-07,
        -1.7833e-05,  2.8739e-03,  1.9194e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5664e-02,  9.0978e-01, -2.9831e-05, -2.7745e-03, -6.4150e-03,
        -3.7957e-03, -1.4681e-03,  2.5535e-03,  9.0898e-04,  3.3064e-03,
         2.2715e-03, -3.4543e-03, -2.2063e-05, -6.3607e-04,  4.8164e-03,
        -2.1541e-03, -7.3439e-04,  2.8481e-03,  7.2496e-04, -3.5170e-04,
         5.3474e-03, -1.4015e-03, -1.1230e-02, -2.7128e-03,  7.3606e-04,
        -1.6878e-03, -3.2725e-03,  1.1297e-04, -3.6399e-03,  4.6001e-05,
        -2.3046e-05, -4.0507e-03,  1.0277e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.2204e-03,  9.3828e-01, -1.3745e-02, -9.0201e-03, -1.5151e-03,
        -3.6886e-03, -1.6411e-03, -1.3358e-03, -3.8659e-03, -6.9224e-04,
         7.8790e-04,  8.4406e-05,  7.0917e-05, -3.3538e-03,  1.3309e-03,
        -2.9368e-03, -1.8849e-03, -3.9716e-03, -6.5771e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7188e-02,  8.9671e-01, -1.5371e-02, -4.5772e-03, -9.1492e-03,
         9.0561e-03,  9.4561e-03,  6.1124e-03, -2.8216e-03,  1.2594e-03,
        -1.2083e-02, -9.5771e-03, -5.6648e-04, -6.0750e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3471e-04,  9.6150e-01, -2.7343e-03, -2.3783e-04, -5.8835e-03,
         4.6670e-03,  3.3207e-03, -3.5656e-03, -6.1853e-04, -5.8020e-04,
         3.5880e-05,  4.2009e-04, -9.5885e-04,  1.8466e-03, -3.7118e-04,
        -2.3435e-03,  2.1890e-03,  9.1736e-04,  8.3305e-04, -7.6191e-04,
         1.9750e-03,  3.7088e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2648e-03,  9.1798e-01,  4.8305e-04, -2.4722e-03,  2.0789e-03,
        -4.0725e-03,  4.7711e-05, -1.4695e-02,  1.3804e-03,  3.2606e-03,
        -1.0448e-03,  2.7678e-03,  1.0573e-02,  8.0310e-04,  1.6483e-04,
        -1.7621e-03, -3.8470e-03,  3.3200e-03,  1.7725e-03, -2.0212e-03,
        -2.4095e-03,  1.7412e-03,  1.9069e-03, -3.6022e-03, -2.6591e-03,
         1.1391e-03,  2.7744e-03, -4.2405e-03,  4.4991e-04,  2.2619e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0657e-02,  9.1815e-01, -3.8247e-03,  1.5994e-03, -6.3135e-03,
         1.2370e-02,  5.1732e-03, -8.0215e-03,  2.7913e-03,  2.4845e-03,
         3.6991e-03, -4.2071e-03,  9.4871e-04, -3.3577e-04, -4.9390e-03,
         1.8547e-03, -2.6336e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.2237e-02,  8.7460e-01,  1.7749e-02, -1.0989e-03, -2.4928e-04,
        -1.1881e-03, -4.9725e-03,  8.0260e-03,  1.3673e-02,  5.1235e-03,
         1.1078e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2170e-02,  9.2963e-01, -1.9378e-03, -5.7529e-04,  6.1898e-03,
         4.8850e-03, -3.3579e-04,  9.2342e-04, -3.1934e-03,  3.9487e-04,
         3.2989e-03,  3.6622e-03,  2.5525e-03,  1.3092e-02,  6.2881e-03,
        -1.4167e-03, -9.4533e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.3710e-04,  8.9497e-01, -1.5644e-02,  1.0022e-02, -5.7619e-03,
         7.8559e-04,  1.3046e-03, -2.7435e-03,  1.0286e-02,  1.0181e-04,
         3.2063e-03,  1.5588e-03,  4.2322e-04,  6.3141e-03,  9.5972e-03,
        -1.2155e-03, -9.1875e-04,  1.1604e-03, -7.0352e-04, -8.3154e-05,
        -1.8210e-03, -4.1311e-04,  5.2868e-03,  1.1477e-03, -2.0050e-02,
        -3.5431e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5506e-02,  8.6811e-01,  3.1423e-02,  3.8455e-03, -1.4279e-03,
         9.9826e-03,  1.6553e-03,  9.5737e-03, -2.0052e-03,  3.2576e-03,
        -6.2185e-03, -3.0246e-03, -8.1514e-04, -1.7870e-03, -3.1898e-03,
         2.9603e-03,  6.7195e-04, -2.6279e-03,  6.2030e-04,  3.5305e-04,
         6.9275e-04, -1.0382e-03, -2.8514e-03,  1.1443e-03, -1.4711e-03,
         1.3288e-03,  1.0859e-03, -3.4887e-03, -6.8553e-03,  9.8906e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1143e-02,  8.8149e-01, -4.8205e-03, -8.5287e-03,  2.4735e-03,
        -6.7374e-03, -9.7081e-03, -6.1004e-03, -4.3597e-03, -8.1159e-03,
        -1.5588e-03, -7.0161e-03, -1.3691e-03, -6.6289e-03, -3.9552e-03,
         2.2450e-05, -2.0348e-03,  1.1305e-02,  2.6284e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 9.4055e-04,  9.1575e-01, -5.8696e-04,  1.7023e-02,  4.7057e-03,
        -1.1961e-03,  1.9756e-02,  4.5761e-04, -4.2950e-03,  2.5146e-03,
         4.9567e-03, -1.4596e-04,  1.3866e-03,  8.9235e-04, -1.1878e-03,
         9.9584e-04,  3.6517e-05,  9.5899e-03,  2.6738e-03,  3.4473e-03,
        -2.7651e-03, -1.7638e-04,  3.8322e-04,  3.7942e-04, -1.0224e-03,
         1.3525e-04,  2.6025e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.4077e-04,  9.3297e-01, -2.0909e-02,  4.2195e-03, -2.0909e-03,
         2.0444e-03, -1.4913e-03,  2.4466e-03,  2.1751e-03, -2.5983e-03,
        -1.3396e-03,  1.4323e-02, -5.4176e-03, -7.3364e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6277e-03,  9.2109e-01, -1.8003e-02,  7.3801e-03,  6.6540e-04,
        -2.0847e-03,  2.0387e-03,  1.2067e-03, -4.2938e-03, -6.7659e-04,
        -1.1563e-03,  6.4275e-04, -7.1611e-04,  1.8656e-03, -5.8518e-03,
        -2.3138e-04, -4.3490e-04, -4.0459e-03,  3.5759e-03, -6.5107e-03,
        -1.7438e-03, -3.3681e-03, -2.4662e-03, -4.2503e-04, -4.5106e-03,
        -3.3851e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0220,  0.9217,  0.0090,  0.0074, -0.0014, -0.0065, -0.0013,  0.0018,
         0.0047,  0.0012, -0.0035,  0.0071,  0.0111, -0.0011,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.8415e-03,  9.4969e-01,  5.8586e-03, -8.4107e-04, -2.8699e-03,
         3.1983e-04,  2.7282e-03,  2.9935e-04,  1.7654e-03,  4.3385e-04,
         3.5781e-04,  2.3023e-04,  3.8519e-03,  2.7463e-04,  1.4151e-04,
         2.6832e-03,  1.0014e-03, -1.0329e-04,  2.0401e-03,  9.2157e-04,
        -4.6282e-04, -1.0823e-03,  1.4527e-03, -1.5951e-03,  1.4709e-03,
         4.3961e-03, -7.5513e-04,  3.3804e-03, -1.7829e-03,  3.9156e-04,
         7.4431e-04,  2.2294e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.8726e-03,  9.2951e-01,  3.4841e-03,  2.0774e-03,  3.6429e-04,
         6.9053e-03, -4.8843e-03, -7.5503e-03,  4.6637e-03,  1.9200e-03,
         4.0370e-03,  3.1526e-03, -1.2966e-03, -5.0595e-04, -3.0309e-05,
         2.7116e-03,  1.8603e-04,  1.3454e-03, -7.7060e-04,  3.4988e-03,
         9.9648e-04,  2.0537e-03,  1.2644e-03,  6.1490e-04, -6.3030e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0494,  0.9077,  0.0279,  0.0071, -0.0059,  0.0019,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3331e-02,  8.9441e-01,  1.4398e-02, -5.5798e-03,  1.0826e-03,
        -3.7736e-03,  4.4184e-03, -3.6658e-05, -9.7716e-04, -7.1866e-04,
        -1.6455e-03,  1.5606e-04, -1.1608e-03,  1.1511e-02, -3.6113e-03,
        -4.3320e-03,  1.7478e-03, -4.1795e-05,  2.0046e-04,  1.3194e-04,
        -4.0926e-03,  2.2566e-03,  3.2106e-04,  1.0709e-03,  1.1897e-03,
         1.7882e-03, -3.6185e-04,  2.9363e-03, -2.6442e-04,  3.7917e-03,
         1.9750e-03,  7.9181e-04, -2.1883e-04, -2.7614e-03,  2.7062e-03,
         7.2732e-04,  9.4797e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8249e-02, -9.1021e-01,  3.0643e-03,  1.1883e-03,  6.2758e-04,
         2.6370e-03,  1.4244e-03,  1.7611e-03,  2.3566e-03, -5.1660e-04,
         1.2269e-03,  1.2675e-04,  1.5058e-04, -1.1517e-03,  1.1145e-03,
         3.5580e-03,  1.0939e-03, -3.1149e-03,  3.7945e-04, -3.2429e-03,
         3.8866e-03,  1.6123e-03,  6.6075e-03,  2.0221e-03,  2.4362e-03,
        -8.0520e-04,  4.6357e-03, -1.2569e-03, -1.2563e-04, -3.2142e-03,
        -7.2363e-04,  2.8768e-04,  1.6658e-03, -1.4804e-03,  2.0505e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.5328e-03,  8.7563e-01, -3.9420e-02, -1.5269e-02,  1.7867e-03,
         1.3532e-03, -4.9843e-03, -5.1943e-03, -2.3470e-03,  6.6294e-03,
         8.6198e-04, -4.3749e-03,  1.3393e-02,  1.0608e-02, -8.6140e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.5467e-03,  9.5369e-01, -3.1890e-04,  2.1985e-03,  4.9968e-03,
         2.4405e-03,  1.8861e-03,  2.9282e-03,  1.1165e-03,  3.9342e-03,
         1.1793e-05,  3.2261e-04, -3.5255e-04, -9.8954e-04,  8.5039e-05,
         1.0253e-04,  1.8873e-04, -4.4520e-04,  2.5471e-03,  8.0701e-04,
         3.9682e-04,  1.6169e-03, -6.6433e-04,  5.0752e-04, -2.2356e-03,
        -1.3566e-03,  1.9471e-04,  1.8051e-03, -5.1470e-04,  1.0513e-03,
        -4.7009e-04, -1.9371e-03, -1.3403e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2243e-02, -9.0609e-01, -1.1286e-03,  5.0208e-03,  4.2555e-03,
         1.1709e-02, -7.8790e-04, -6.5799e-04,  6.6885e-03,  4.2222e-04,
         8.3727e-05,  2.2380e-03,  4.2022e-03,  8.0843e-03, -2.5477e-03,
         7.9826e-04, -7.7003e-04,  1.8325e-03, -1.2244e-03, -1.2941e-03,
         6.7695e-04,  1.3142e-03,  1.1117e-03,  5.4496e-04, -9.2038e-04,
         3.0984e-03, -1.2871e-03,  5.4079e-03,  3.5545e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 2.5327e-02,  8.9220e-01,  3.3377e-04, -7.8246e-03, -2.3483e-02,
         2.2823e-03, -2.1091e-04,  4.7758e-03, -2.0170e-04, -2.8437e-02,
        -2.4535e-03,  4.0940e-03, -9.7800e-04,  5.5519e-03,  1.8430e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6160e-02,  9.0829e-01, -6.5573e-03, -2.0792e-03, -3.7575e-03,
         5.3758e-04,  2.6771e-03,  5.9314e-04,  2.8112e-03, -8.8263e-04,
         4.2334e-04,  2.3572e-04,  1.4694e-03,  1.9074e-03,  4.3277e-04,
        -3.8890e-04,  1.0985e-03,  4.8907e-04,  1.5395e-04,  2.6451e-03,
        -3.4235e-04,  6.7074e-04,  6.7034e-04,  8.9439e-04,  5.7539e-04,
        -8.6386e-04, -2.3073e-03,  2.4824e-03, -4.1267e-03,  3.9155e-04,
         1.2557e-03, -4.0737e-04, -1.0008e-05, -1.0518e-03,  1.4841e-03,
        -2.2854e-04,  2.0308e-03, -3.0341e-04,  1.8187e-04, -1.5880e-03,
        -1.2515e-04,  3.0305e-03,  5.4247e-03, -4.8739e-04,  7.9782e-05,
        -2.3803e-03,  6.3663e-05, -4.3862e-03, -8.5632e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1124e-03, -9.2258e-01, -3.2254e-03, -4.2089e-04, -6.3414e-03,
        -3.0985e-03,  3.1177e-04, -2.1983e-03,  1.5859e-03, -2.8090e-03,
         1.5150e-03, -2.8450e-04, -2.5080e-03, -1.4245e-03,  5.9042e-04,
         2.5691e-03, -1.8914e-02, -6.7418e-04, -2.4981e-03, -4.9908e-04,
        -7.2484e-04, -5.8750e-04,  4.8229e-03, -1.7722e-03, -4.8499e-04,
         2.2472e-03,  2.4544e-03,  9.7445e-04, -3.6263e-04, -4.6745e-04,
        -6.8975e-04,  1.4433e-04, -5.8580e-04, -6.8595e-04,  4.8360e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6568e-03,  9.4310e-01,  1.1703e-03,  5.4640e-03,  3.3729e-03,
         3.4532e-03, -5.0307e-03,  1.4962e-03,  2.5244e-04, -1.5546e-03,
        -4.1783e-03, -1.9120e-03,  3.0122e-03,  8.0363e-04,  1.2286e-03,
        -7.6016e-03,  3.2134e-03,  9.1477e-04,  9.5861e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.6910e-03,  9.0950e-01, -1.2801e-02, -1.5664e-02, -6.2072e-03,
        -4.0270e-03, -7.3857e-03,  8.7769e-05,  6.7354e-03, -5.4371e-03,
        -4.5521e-03,  1.0140e-03,  2.3173e-03, -9.2010e-03,  9.3819e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.1017e-04,  9.0465e-01, -9.1019e-03,  1.0690e-02, -5.5614e-03,
         2.0980e-03,  5.0997e-03, -1.1215e-03,  2.0165e-03, -3.8574e-03,
        -3.0651e-03,  8.0620e-04, -4.8676e-03, -3.7041e-03, -3.9183e-03,
         2.9820e-03, -8.7851e-04, -1.2590e-03,  5.9636e-03,  3.3930e-04,
         2.6985e-02,  2.2034e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6450e-03,  9.2171e-01, -7.0386e-03, -8.6807e-03, -1.3338e-03,
         5.8397e-03,  1.1366e-03,  3.0369e-03, -8.9177e-03, -1.2546e-03,
        -1.5726e-03,  1.5824e-03, -5.0991e-03, -3.6910e-03,  1.6361e-03,
         5.0834e-04, -8.3553e-05,  3.1024e-03,  7.2606e-04,  2.8958e-04,
        -5.3115e-04, -3.2321e-03,  7.6469e-04,  1.9078e-03,  1.4999e-03,
         3.5925e-03, -1.0693e-03,  3.1003e-04, -1.8229e-03,  5.3802e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0031,  0.9027, -0.0117, -0.0145,  0.0147, -0.0044,  0.0057, -0.0200,
        -0.0106, -0.0087, -0.0038,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3195e-02,  9.1619e-01,  1.1678e-02,  1.1044e-03,  2.8687e-03,
         4.8455e-03,  4.8605e-03,  2.5039e-04, -4.0145e-04,  2.6573e-03,
         1.9233e-03,  1.2162e-03,  2.0575e-03, -2.0673e-03, -1.3197e-03,
         2.4905e-03,  1.9195e-04,  1.4481e-03, -2.4597e-03, -3.4105e-03,
         5.8588e-03, -7.5038e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3632e-02,  9.0896e-01, -2.4723e-03, -1.3615e-03,  3.5810e-03,
         1.3392e-02,  2.7357e-03,  1.5300e-03,  1.8287e-03,  1.0382e-03,
         1.5198e-03,  2.6285e-03,  6.1439e-03, -3.3423e-03, -1.6282e-03,
        -7.1444e-04,  6.7725e-04,  2.5405e-03, -1.3147e-03,  1.7379e-03,
         1.2174e-04, -1.5629e-03, -7.8609e-04, -9.6198e-04,  3.0409e-03,
        -1.5852e-03,  1.1763e-03, -5.5962e-03,  6.0646e-03, -3.5061e-04,
         4.6803e-03,  5.4150e-04, -7.5397e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.5075e-03,  8.8658e-01,  5.6153e-03,  2.9261e-03,  6.9270e-03,
         5.8145e-03,  8.1927e-03,  1.7566e-03, -2.7368e-03, -2.6596e-03,
        -1.3333e-03, -1.1872e-03,  4.5397e-04,  1.3114e-03, -1.8128e-03,
         1.6338e-02,  1.2466e-03, -4.0322e-03,  7.0463e-04,  1.7596e-03,
         6.4975e-03,  3.9867e-04, -1.9709e-04, -2.4158e-03, -9.3603e-03,
         1.0667e-03,  1.6563e-02, -3.6089e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3127e-02,  8.8803e-01, -5.5385e-03,  4.3373e-03, -3.2912e-03,
         3.3510e-04,  1.1203e-02,  1.0451e-03, -9.2012e-03,  6.8209e-04,
        -3.3421e-03,  7.4972e-04, -7.2900e-03,  1.7179e-03, -3.0399e-03,
         1.4695e-03, -1.1306e-03, -6.1869e-03, -2.9110e-03, -1.9069e-03,
         1.1202e-03,  1.2100e-04,  6.1070e-03, -7.2132e-03,  1.8898e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 8.8235e-03,  8.9060e-01,  1.3572e-02,  2.6293e-03, -1.1535e-03,
         7.4715e-03,  3.0058e-03, -9.2799e-04, -2.7174e-03,  1.4695e-03,
         8.2304e-03,  1.2751e-03, -1.0973e-03,  2.2420e-04,  5.9001e-04,
        -1.6421e-02,  1.7691e-03, -7.0113e-04,  1.2714e-04, -1.0776e-04,
         1.4551e-03,  7.6846e-04, -2.4296e-04, -1.5813e-03,  2.5756e-03,
         3.8370e-04,  1.4115e-03, -4.4828e-04,  3.0579e-04,  3.8713e-03,
        -5.2226e-04, -7.3805e-04, -5.2692e-03, -2.3320e-04,  3.7992e-04,
        -1.0183e-03, -1.1753e-03, -1.9007e-03,  9.5506e-05, -1.0728e-03,
         1.1642e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.4894e-02,  8.7242e-01, -1.1760e-02, -4.0413e-03, -1.2216e-03,
        -4.2191e-03, -2.7479e-03,  1.8686e-03, -3.4374e-03,  6.0509e-04,
        -1.7891e-03, -1.6218e-03,  4.3708e-03, -3.3403e-03,  2.4490e-03,
         5.7180e-05,  1.1811e-03, -3.1944e-03,  9.1943e-04,  7.2705e-03,
        -8.1651e-04, -2.0457e-04,  2.0833e-04, -1.6175e-03,  1.3745e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7178e-02,  9.1755e-01,  1.0881e-02, -1.4262e-02, -1.9531e-03,
        -3.9922e-05, -4.7663e-03,  2.9399e-03, -4.7172e-04, -8.4103e-04,
         3.0321e-03,  5.3736e-03, -4.5854e-04,  2.5582e-04,  5.5074e-05,
        -1.2998e-03, -1.9193e-04,  5.4369e-04, -1.7816e-03,  5.9455e-03,
        -1.8150e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5748e-02,  8.9519e-01,  1.3261e-02, -8.4331e-03, -5.7624e-04,
         3.1279e-04, -1.6790e-03, -4.0575e-04, -3.0961e-03, -5.3945e-03,
        -1.4214e-03, -1.5918e-02,  6.2952e-03, -4.7420e-03, -9.5130e-03,
        -1.3772e-04,  5.9062e-03, -1.1967e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.9980e-03,  9.1842e-01, -1.0203e-02,  8.0722e-03, -6.7611e-04,
         7.6756e-03,  2.6454e-03, -2.8680e-04, -4.8078e-04,  5.3965e-04,
         1.1693e-03, -1.2213e-04, -2.8373e-03,  3.1950e-03, -4.2564e-03,
         5.4985e-03,  1.7269e-03,  8.7681e-04,  3.3168e-03,  3.7194e-04,
         2.5098e-03, -4.7920e-04, -8.8228e-04, -1.7503e-03,  2.2302e-03,
         6.1859e-04,  9.6012e-04, -6.8365e-05,  1.2826e-04,  5.0039e-04,
        -2.2378e-03,  2.6486e-03, -3.9941e-04, -9.0628e-04,  1.0940e-03,
         3.4726e-03,  1.5268e-03, -2.2213e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5667e-02,  8.9614e-01, -2.4121e-02,  8.9338e-03,  3.3110e-03,
         1.5737e-03, -5.0696e-03,  6.7446e-03,  4.0171e-03, -1.8218e-03,
         1.4008e-03, -6.4187e-03, -6.0604e-04,  3.6033e-03,  4.2922e-03,
         1.1881e-03,  2.0109e-03, -1.1201e-02,  6.5739e-04, -1.2168e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0787e-03,  9.4542e-01, -1.7705e-03,  8.8041e-03,  2.7849e-03,
        -3.5766e-03,  7.4753e-03,  5.6690e-03,  2.8484e-03,  3.2662e-04,
        -6.5026e-03, -7.6387e-05, -7.4994e-03, -4.1709e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.7150e-02,  8.5905e-01,  6.6421e-03,  2.3975e-03,  5.9236e-04,
        -3.3280e-04, -3.1684e-03, -1.8599e-03, -5.9358e-04,  4.5679e-04,
        -4.2403e-03,  8.0514e-04,  1.0281e-03,  4.9747e-03,  1.4948e-03,
        -2.0920e-03, -2.5476e-03,  4.4410e-04,  3.2240e-03,  2.5668e-03,
         9.6534e-04,  6.1580e-03, -9.7116e-04, -2.3331e-03, -4.0479e-04,
        -3.5919e-03,  2.1586e-03, -4.0972e-03, -4.1266e-03, -1.9180e-03,
        -1.3872e-03,  6.2276e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1725e-02,  9.0122e-01, -2.1219e-04,  1.0223e-02, -2.2734e-02,
        -2.6294e-03,  2.3166e-03, -5.7561e-03, -2.9044e-03, -4.4081e-03,
        -1.3681e-03, -4.1160e-03,  1.1640e-04,  1.5065e-04,  1.7937e-03,
        -1.9700e-03, -1.0957e-03, -1.9238e-03, -4.3591e-03, -1.8974e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.5888e-03,  9.1817e-01,  4.2624e-03, -1.6076e-03,  3.8099e-03,
         1.1671e-03,  6.2611e-05, -2.0624e-03,  3.1485e-03, -1.3667e-03,
        -1.5904e-03,  2.9365e-04, -2.4508e-03, -7.9694e-04, -3.5241e-03,
        -1.6081e-03,  2.9965e-03, -1.2744e-04,  2.4526e-03, -2.0543e-03,
         8.6502e-04, -7.3918e-04, -1.1088e-03,  3.8239e-04, -3.0831e-04,
        -8.5633e-04, -4.2618e-03,  2.5843e-03, -6.5031e-04,  2.1343e-03,
        -3.8881e-04,  1.0320e-03, -3.8289e-04,  2.9927e-03, -1.3469e-03,
        -1.3531e-05,  2.8476e-05,  3.7758e-03,  2.0476e-03,  1.4945e-03,
         1.5093e-03, -6.0635e-04,  1.4616e-03, -2.5997e-03,  2.7559e-03,
         1.0997e-03,  1.4320e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0600e-03,  9.3911e-01, -1.4044e-03,  6.0936e-03,  1.2134e-02,
         1.2911e-02, -1.2837e-03, -9.8223e-04,  1.2806e-03, -9.0706e-03,
        -3.0896e-03, -7.3557e-04, -1.3563e-03, -7.8959e-03, -5.8763e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5352e-02,  8.9847e-01,  1.9362e-03,  3.0819e-03,  3.2078e-04,
        -3.5526e-03,  1.6079e-03,  5.1038e-03,  1.2719e-03, -8.2938e-04,
        -1.7517e-03, -6.3397e-03, -2.0694e-03,  1.3533e-03, -1.6905e-03,
        -5.2054e-03, -4.1255e-04, -5.1692e-05,  5.2561e-04, -2.0995e-03,
        -1.4422e-03, -5.4064e-03,  7.8259e-04, -4.0175e-04, -7.6634e-04,
        -3.8769e-03,  1.8756e-03,  3.6915e-04, -2.8768e-03,  1.9203e-03,
        -4.4293e-04,  1.5029e-03,  1.2656e-04,  1.2538e-03,  6.5323e-04,
         3.5027e-03, -2.1499e-03, -1.2196e-04, -4.4563e-03, -1.1555e-03,
        -4.5228e-04, -3.7570e-04, -1.1718e-03, -2.9759e-05, -3.0278e-03,
         6.1734e-04, -3.9518e-03, -2.2628e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-9.5134e-03,  9.2203e-01, -1.8386e-03,  6.4151e-04,  9.7926e-03,
        -6.5613e-03,  3.9531e-03,  1.6226e-03,  1.0488e-03, -7.5647e-03,
        -1.6507e-03, -5.6639e-03, -1.0442e-03,  2.7320e-04,  7.9518e-04,
        -2.5168e-03, -7.4200e-03, -3.7916e-03, -6.6632e-04, -5.7007e-05,
         1.2310e-03, -1.7149e-03,  1.8916e-03, -5.0342e-03, -1.9308e-04,
         1.4873e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9689e-03,  9.3393e-01,  2.2487e-05,  3.3303e-03, -1.5435e-03,
        -5.5467e-03,  2.7566e-03,  6.7260e-03,  1.6508e-03, -6.8658e-04,
        -5.8196e-03, -2.1769e-04,  1.2909e-03, -4.0695e-03, -2.1523e-03,
         2.4660e-03,  4.2878e-03,  2.1395e-03,  1.2568e-02, -4.3400e-03,
        -2.4823e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.5175e-03,  8.8476e-01, -2.0042e-03, -8.1400e-03, -5.4654e-03,
        -1.3018e-02, -3.8766e-05, -4.6115e-03, -1.4175e-03, -2.6878e-03,
        -2.7266e-04, -6.4721e-04, -3.1524e-03, -1.2154e-02, -4.5310e-04,
        -3.0192e-03,  1.2641e-03, -7.1272e-03,  2.9683e-03, -1.2623e-03,
        -2.6766e-03, -2.0748e-03,  1.3485e-03,  1.0288e-03,  2.3552e-03,
        -7.5022e-04,  7.1169e-04, -1.9453e-03,  1.9968e-04,  1.3616e-03,
        -5.1650e-03, -1.8307e-03,  3.2899e-03, -1.3369e-03, -7.3591e-03,
         2.5893e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5719e-03,  8.6268e-01, -2.4616e-03,  1.2356e-02,  2.5494e-03,
         6.5837e-03,  6.2299e-05,  1.1457e-02,  9.2197e-03,  2.4464e-04,
         1.2124e-02, -1.8969e-03,  8.9251e-04, -9.8605e-03, -6.6054e-03,
         9.2940e-03,  3.2560e-03, -3.4250e-04,  2.0880e-02, -2.1663e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1240e-03,  9.4891e-01,  3.8168e-03,  6.9402e-05,  5.3659e-03,
        -1.0698e-03, -5.5082e-03, -3.1543e-03, -1.6363e-03,  3.3409e-03,
         5.1356e-03, -6.4693e-03, -7.2573e-04, -1.9990e-04,  4.2596e-03,
         1.6584e-03,  5.9476e-04, -1.6311e-03,  8.3619e-04,  3.0609e-04,
        -4.1909e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.9257e-03,  9.3058e-01, -8.1352e-04, -8.7091e-03, -1.0558e-02,
        -6.5622e-03, -3.1811e-03, -8.6724e-03, -6.9959e-04,  6.7814e-03,
         3.3268e-03,  8.1696e-03, -2.0210e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0116e-02,  9.0856e-01, -4.4182e-03,  5.1434e-06,  2.0865e-03,
        -1.9449e-03,  6.1057e-03,  4.7310e-04,  2.9983e-03, -6.5800e-04,
        -8.1563e-03, -1.0403e-02, -1.5471e-03, -3.4928e-04,  3.3956e-03,
        -3.1525e-03,  1.7528e-03, -1.9421e-03, -1.1549e-03, -1.1495e-03,
        -6.5647e-04,  1.5753e-03,  2.5910e-04, -1.0739e-03, -6.1284e-04,
        -1.0414e-04,  2.3205e-03, -2.9741e-03,  5.1997e-04, -1.5115e-03,
         1.6689e-03, -1.7724e-06, -7.3391e-04, -4.1109e-03,  1.2699e-03,
        -1.6171e-05,  2.2339e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.0615e-03,  9.1259e-01,  3.2709e-02, -7.8701e-03,  9.8353e-04,
         1.2571e-03,  5.1713e-04,  2.1188e-03,  9.4906e-04,  1.7249e-03,
         9.6880e-04,  2.9795e-03,  6.5084e-03,  1.4458e-03, -1.4875e-03,
         2.5387e-03,  4.6908e-03,  9.5500e-04,  1.1397e-03, -2.0239e-03,
         1.0187e-03,  6.0017e-03,  8.4947e-05, -3.3766e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0135e-02,  8.4590e-01,  1.1858e-02,  7.6857e-03,  3.4552e-03,
         3.1645e-03,  8.4954e-04,  3.7082e-03,  7.1832e-03, -4.0531e-03,
        -5.3297e-03,  1.1032e-02,  4.0424e-04,  3.2523e-03,  6.6515e-03,
         2.8454e-03,  2.4901e-03,  5.6354e-04, -3.0038e-04, -2.8878e-03,
        -8.2145e-04,  2.2314e-03,  1.9206e-03, -8.7808e-04, -3.4133e-04,
        -1.3709e-03, -1.0759e-03, -2.7780e-03,  6.9803e-04,  1.7534e-03,
        -3.8824e-03, -8.4985e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.2469e-03,  9.0818e-01,  1.6670e-02, -1.5762e-03, -9.0870e-03,
        -8.3886e-03, -1.6880e-03, -1.5549e-03,  3.0495e-03,  4.5749e-03,
         2.3924e-04,  7.4450e-04,  4.9126e-04, -1.4778e-04,  8.6766e-04,
         6.6610e-04,  2.8094e-03, -1.0457e-03,  3.2823e-03, -2.2881e-03,
         2.3027e-03,  3.7110e-04,  2.9013e-04, -1.7418e-03, -1.1374e-03,
         6.7837e-04,  2.9044e-03, -9.3040e-04,  3.8216e-03, -6.5779e-03,
        -2.6482e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1221e-03,  8.9843e-01,  1.5415e-03, -3.9984e-04,  2.0319e-03,
         5.2678e-03, -6.3321e-04, -3.2371e-02, -4.1480e-03, -4.3913e-03,
         4.0660e-03,  2.0293e-03,  6.7234e-03, -9.6950e-03, -5.9520e-04,
        -4.8426e-03,  1.3595e-04,  2.0669e-03,  2.6197e-03, -2.0413e-03,
        -4.7754e-03, -2.5746e-03, -2.5016e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.8221e-03,  9.0893e-01,  1.5793e-02, -7.1664e-03, -7.0954e-03,
        -1.5261e-04, -1.6221e-03, -3.6273e-03,  6.4222e-03,  1.7146e-03,
        -1.8636e-03,  1.6030e-04, -2.0844e-03,  2.5880e-03,  3.8053e-04,
        -2.4319e-04, -3.5249e-03, -3.1339e-03,  3.4640e-03,  7.4398e-03,
        -3.7361e-03,  1.7087e-04,  3.6696e-03,  1.0341e-03, -9.9671e-04,
        -2.0576e-04, -6.1157e-04, -2.3489e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-1.7908e-02,  9.2556e-01, -3.4035e-04,  2.4069e-03,  4.9384e-04,
         3.3286e-03,  2.9579e-03, -6.7724e-03, -1.1283e-03, -5.1542e-03,
        -2.5645e-03,  1.2768e-04,  2.7397e-03,  1.0956e-03,  2.3623e-04,
         5.0257e-03, -1.1612e-03, -2.3618e-03,  2.5980e-04, -4.0496e-03,
        -2.4775e-04, -6.4001e-04, -4.6906e-03,  1.2006e-03, -3.6147e-04,
        -1.9424e-03, -2.8139e-05,  3.9438e-03, -1.2744e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.8778e-03,  9.3834e-01, -4.4098e-04, -3.3680e-03, -3.0941e-03,
         4.9771e-03, -3.3441e-03,  3.0214e-03,  5.6444e-05,  7.3080e-04,
        -3.9609e-03, -1.1093e-03, -3.9809e-03,  1.2260e-03,  4.7333e-03,
         2.1084e-03,  1.0533e-03, -2.7922e-03, -1.3263e-03,  3.9867e-03,
         1.1709e-03,  1.2693e-04, -4.0963e-03,  3.0814e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7306e-02,  8.9865e-01,  8.0888e-03,  7.2252e-03,  2.8649e-03,
         2.8480e-03, -4.4513e-03,  1.1512e-03,  5.0341e-04,  1.2069e-04,
         1.9181e-03,  2.6669e-03, -2.4344e-04,  2.1904e-03, -5.9868e-04,
         3.2227e-03,  4.8327e-03,  7.9361e-04,  4.5842e-04,  1.1706e-03,
         4.4758e-03, -1.2880e-03,  2.5490e-04,  1.2722e-03,  5.9970e-03,
         7.0680e-05, -6.1045e-03,  2.5135e-03, -2.2781e-03,  4.1055e-04,
         1.6152e-03,  3.8692e-03, -2.3204e-03,  6.2243e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3850e-04,  9.2704e-01, -1.5214e-02,  3.5129e-03,  4.2009e-04,
         1.9271e-03,  4.5508e-03,  8.9069e-03,  6.0693e-03,  6.2058e-05,
         4.1043e-03,  4.1332e-03, -3.5265e-04,  3.2898e-03,  1.7032e-03,
        -6.0313e-04, -2.1887e-03,  8.5634e-04,  1.5141e-03,  3.6148e-03,
        -3.5548e-03, -6.2441e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1962e-03,  9.3253e-01, -2.6731e-03,  6.8780e-03,  2.0715e-03,
         5.4356e-03, -2.5105e-03, -1.8341e-03,  2.1089e-03,  2.4069e-03,
         1.6791e-02, -3.8185e-03,  9.2725e-04, -5.3445e-03,  1.6708e-03,
         4.5768e-03,  1.1596e-03,  4.0666e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7964e-03,  9.4126e-01, -9.7023e-03, -1.3592e-03,  1.4231e-03,
         9.3951e-03,  3.3759e-04, -2.6561e-03, -4.5618e-03,  3.3462e-03,
         6.1535e-04, -3.7335e-03, -1.3619e-03,  1.5103e-03,  9.7215e-04,
         1.4008e-03, -3.2805e-03, -3.5797e-03, -2.9818e-03, -2.0409e-03,
         2.0516e-03, -4.7021e-04,  1.5971e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6477e-02,  9.3410e-01, -1.2402e-02, -6.3124e-03, -8.1984e-03,
        -3.9792e-03,  2.2428e-03,  3.6069e-04, -3.1072e-03, -2.2166e-03,
         7.1597e-04,  7.8137e-05,  1.4407e-03, -7.8305e-04,  1.6203e-03,
         1.7906e-03,  4.1753e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4466e-02,  9.3197e-01, -1.3726e-02,  6.7761e-03,  7.3595e-03,
         9.1916e-05,  1.0670e-04, -2.0514e-03,  2.5602e-04, -5.2779e-04,
         1.9616e-03,  3.1292e-04, -3.9458e-03,  1.5168e-03, -1.5425e-03,
        -3.0690e-03,  2.9605e-03,  7.3626e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6868e-03,  9.1317e-01,  5.3567e-04,  2.3963e-03,  4.1060e-03,
         2.8325e-03, -1.7342e-03, -3.7614e-03, -2.4580e-04,  1.2614e-02,
        -3.4279e-03, -1.0768e-02,  3.4206e-03, -4.2526e-03,  1.9028e-03,
         9.1512e-03, -4.7024e-03,  3.0393e-04,  2.2891e-03, -1.0589e-03,
        -9.4513e-04,  8.4314e-05,  4.5156e-03, -2.1624e-04,  5.0597e-03,
         3.8150e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2606e-02,  9.1823e-01, -1.1900e-02,  1.5367e-03,  3.9306e-03,
         3.4679e-03,  9.5680e-05,  5.2085e-03,  4.3512e-03, -3.1863e-03,
        -3.4750e-03,  2.6409e-03,  3.0025e-04,  3.4594e-03,  6.0483e-04,
         1.2187e-03, -6.5237e-04, -2.3807e-03,  1.0078e-03, -5.8701e-04,
        -1.2110e-03, -1.6232e-03,  1.3254e-03,  1.7742e-03,  1.0665e-03,
         5.6261e-04, -8.9454e-04,  8.9526e-04,  1.3690e-05, -4.0276e-03,
         5.7646e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0530e-03,  9.2119e-01,  3.4387e-03,  4.2968e-03, -9.1424e-03,
        -4.5896e-03, -2.1900e-03, -3.6382e-03,  1.1215e-05, -2.5016e-03,
        -5.9298e-03, -6.7223e-03, -5.1286e-04, -5.2725e-03, -1.1603e-02,
        -3.0831e-03, -2.3175e-03,  1.0510e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0333,  0.9246, -0.0253,  0.0037,  0.0025,  0.0084, -0.0011,  0.0010,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([-2.1529e-02,  9.1308e-01, -6.0214e-03, -4.0973e-03,  2.1959e-03,
        -2.9279e-03, -3.2226e-03, -2.9229e-03, -1.4739e-03, -9.2039e-04,
        -2.2840e-04, -6.7572e-04, -2.3900e-03, -3.9399e-03, -5.1705e-04,
         9.1723e-04, -2.6835e-03, -2.4825e-03,  1.4881e-04,  1.5059e-03,
        -9.3252e-04, -9.2238e-04, -4.1533e-03, -1.7379e-04, -8.2813e-04,
        -2.3044e-03,  1.6619e-03,  2.3926e-03, -6.2986e-04,  4.6784e-04,
         2.7688e-04, -1.0001e-03,  3.3429e-04,  1.6552e-03,  1.8092e-03,
         8.8139e-04, -3.0253e-03, -3.9660e-04,  2.2786e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.5401e-03,  9.4852e-01, -1.0459e-02, -2.8001e-03,  6.4196e-04,
         2.2714e-03, -2.8221e-03,  1.0875e-03, -3.5871e-05,  7.5894e-04,
         1.7273e-04,  1.6216e-06, -2.4305e-03, -3.0839e-03,  3.5977e-03,
         1.3918e-03,  1.4399e-03, -2.0252e-03,  1.9960e-03,  2.1444e-04,
         3.3531e-03,  1.0939e-04, -3.6963e-04, -6.8520e-04,  8.4002e-04,
        -6.4576e-04, -5.9331e-04, -1.1069e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8301e-03,  9.2614e-01,  8.8476e-03,  1.1333e-02,  9.1569e-04,
         3.3670e-03, -1.1496e-03,  1.4225e-03,  8.8639e-03,  3.5258e-06,
         4.8594e-03, -1.4468e-02, -1.4799e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7545e-02,  9.2756e-01, -1.9659e-04, -1.3203e-03, -4.6061e-03,
         6.8913e-04, -4.1580e-03,  7.9992e-03, -1.2204e-04, -8.2412e-04,
        -5.8022e-03, -3.0958e-03, -1.5015e-03,  3.1809e-03,  1.1402e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1864e-02,  9.0293e-01, -8.5531e-03,  2.3930e-03,  5.9634e-03,
         2.4236e-03,  5.9862e-03,  7.6442e-05,  1.5329e-04, -4.1586e-03,
         4.0672e-04,  6.5845e-04,  2.9158e-03, -9.8857e-04,  1.8690e-03,
         1.2492e-03,  1.4471e-03,  1.6272e-03, -2.4669e-03, -1.0329e-05,
         2.3030e-04, -1.4018e-03,  1.5640e-03,  2.8693e-03, -9.0323e-04,
         1.6662e-03, -1.9906e-03,  7.4653e-04, -2.2157e-03,  1.1318e-03,
         1.5994e-03, -8.6316e-04, -5.3335e-04,  2.4297e-03,  4.3196e-03,
         1.4521e-04, -6.8494e-04,  1.0460e-03, -1.3797e-03,  1.6045e-03,
         3.4225e-03,  1.6615e-04, -8.2400e-04,  1.2288e-03, -1.2545e-03,
        -1.8707e-03,  3.7626e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1683e-02,  9.1620e-01,  4.0450e-03,  5.9461e-03,  9.5809e-04,
         1.2968e-03,  3.8455e-03, -2.2840e-03, -2.4370e-03,  5.4302e-04,
        -5.8185e-04,  1.9295e-03, -1.0277e-04, -1.0778e-04,  1.8440e-04,
        -8.4290e-04,  1.1987e-03,  2.2912e-03, -1.8900e-03,  6.1077e-05,
         1.0931e-03,  2.9962e-03,  6.9410e-04, -2.0019e-03, -1.9879e-03,
        -3.2577e-03,  2.3256e-03,  9.7153e-04,  1.4309e-04,  4.4349e-03,
        -2.3684e-03,  2.2558e-04,  1.8652e-04,  5.7488e-04, -5.3721e-04,
        -2.2327e-04, -1.5207e-03, -6.1668e-04, -1.4519e-03, -3.8822e-04,
        -1.0326e-03,  2.5354e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1744e-04,  8.7678e-01,  4.6924e-03, -3.0725e-02,  2.3933e-03,
        -8.6570e-03,  2.1180e-02, -2.2210e-03, -1.7419e-03,  5.2546e-03,
        -1.3627e-03,  1.9532e-03, -2.0113e-03, -2.5842e-03,  5.1279e-03,
         4.8190e-03,  5.7326e-03, -1.7291e-03, -3.9516e-03, -1.3445e-04,
        -3.5641e-03,  1.4429e-03, -5.6005e-03, -2.8331e-03, -3.3862e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.1938e-03,  9.0044e-01,  2.6585e-02,  1.1111e-04, -1.0486e-03,
        -5.0900e-03,  9.2014e-03, -1.3089e-03,  3.4948e-03, -8.1559e-04,
         1.2234e-03, -3.6330e-03,  2.8652e-03, -1.7577e-03,  3.4410e-03,
        -1.5131e-03,  2.0092e-03,  1.2058e-02, -3.0658e-03,  3.5490e-03,
         2.4589e-03,  2.4050e-03, -4.7358e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1046e-03,  8.8888e-01,  2.0019e-02, -3.2749e-04, -9.2273e-03,
        -4.0611e-03,  1.9848e-03, -2.2381e-03,  2.5387e-03,  1.3152e-02,
         3.3741e-03, -5.2590e-03,  3.0646e-03, -5.2040e-03,  1.9836e-04,
        -1.4146e-03,  6.9794e-03,  5.7976e-03,  1.6798e-03, -8.3815e-03,
         2.0427e-03, -2.9905e-03,  6.0823e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0174,  0.8828,  0.0162, -0.0112, -0.0076,  0.0088,  0.0060,  0.0147,
        -0.0017,  0.0144,  0.0057,  0.0066, -0.0068,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1012e-02,  9.3835e-01,  3.2633e-04, -1.8905e-03,  1.4759e-03,
        -1.4305e-03,  1.8554e-03,  3.1627e-03,  2.6960e-03,  2.7882e-03,
        -2.5427e-03, -5.3228e-03, -5.8721e-03,  1.2790e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2045e-03,  9.3336e-01,  1.6061e-02,  3.3815e-03,  3.3065e-03,
         1.4673e-03,  4.0094e-04, -2.2155e-04,  2.2935e-03,  8.7086e-03,
         2.5745e-03,  3.6623e-03, -3.9926e-03,  2.1642e-03,  2.0585e-03,
         1.6055e-03, -4.8016e-03,  3.1448e-03, -5.5904e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-8.2770e-03,  9.2209e-01,  7.8155e-03, -6.4011e-03,  5.4310e-05,
        -1.1313e-03,  4.9718e-04, -9.0628e-04, -2.3686e-03,  4.4439e-03,
        -1.4972e-03, -4.5680e-04,  8.0958e-04,  5.6327e-03, -2.3359e-03,
        -1.0242e-03,  3.2018e-03, -2.0562e-03,  2.5580e-03,  5.9790e-03,
        -1.2212e-03,  3.6978e-04, -4.2424e-03,  3.2919e-03, -3.9634e-03,
        -7.3734e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0272e-03,  9.1191e-01, -1.5831e-02, -3.0100e-03, -6.4693e-03,
        -1.3464e-02, -2.0320e-03, -5.0646e-03,  4.4541e-03, -1.1924e-02,
        -2.4623e-04, -3.2717e-03,  2.3742e-03,  5.9609e-03, -1.5926e-03,
         1.0370e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2430e-02,  8.7130e-01,  4.3172e-03,  2.7793e-02,  2.4990e-03,
         1.6742e-02,  6.1303e-04, -2.2266e-03,  5.9385e-03,  2.5039e-03,
        -8.7675e-03,  5.2842e-04,  8.3929e-04,  1.4095e-02,  1.4478e-03,
        -7.5168e-03, -4.4531e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3603e-03,  9.2031e-01, -1.5453e-02, -1.3293e-02, -1.9651e-02,
         7.6286e-03, -1.6599e-05,  1.1492e-02, -7.7910e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2594e-02,  8.3946e-01,  1.7418e-02, -1.2702e-03, -1.0530e-02,
         1.1563e-03,  4.2985e-03, -3.1566e-03,  3.1961e-03, -1.5252e-03,
         9.3000e-03, -5.5220e-03, -3.2163e-03,  7.1970e-04, -1.4394e-05,
         2.8628e-03,  2.8630e-03,  4.1902e-03,  1.2454e-03,  1.0800e-03,
         6.2547e-04,  1.9605e-03, -5.9834e-04, -1.3639e-03, -3.2694e-04,
        -4.1944e-03,  7.1203e-03,  2.2007e-04, -4.5012e-04, -1.9707e-03,
         3.3312e-03,  1.1587e-03, -4.9024e-05,  1.8380e-04, -2.0443e-03,
        -2.5024e-03, -2.1871e-03, -2.6830e-03, -4.7139e-03, -1.9834e-04,
        -4.2594e-03, -3.5080e-03, -1.1033e-03, -2.7268e-03,  2.1374e-03,
         2.7608e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.6457e-04,  9.5420e-01, -5.0758e-04, -1.4869e-03,  2.1576e-04,
         2.6652e-03, -2.8680e-03, -3.6900e-05,  5.6480e-04,  3.2027e-03,
         6.0837e-04,  1.2367e-03,  7.5946e-04,  3.5090e-04, -9.5372e-04,
         6.7109e-04,  9.4035e-04,  1.5001e-03, -1.7771e-03,  1.8645e-03,
         3.0641e-03, -1.1904e-03, -4.7236e-04, -9.5615e-04,  1.7045e-03,
        -2.4965e-03,  1.5875e-03,  1.2312e-04,  3.3777e-03, -7.7489e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0024,  0.9692, -0.0101, -0.0119,  0.0023, -0.0027,  0.0014,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7207e-02,  9.4097e-01, -1.3789e-03,  6.7100e-03,  3.3013e-04,
         2.7177e-04,  5.2662e-03,  7.5900e-03,  7.9793e-03, -3.1626e-04,
        -1.9854e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5453e-03,  9.1494e-01,  8.6153e-03,  1.3273e-02,  1.8946e-03,
        -5.0691e-03,  4.5473e-04, -4.4223e-03,  1.1945e-02,  3.7984e-03,
         6.9221e-04,  6.8168e-03, -2.0947e-02, -2.5890e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9618e-03,  9.0367e-01, -6.7548e-03,  3.4891e-03,  2.0478e-03,
         3.5475e-06,  6.5215e-03, -6.6990e-04, -1.2430e-04,  1.9406e-03,
        -1.8494e-03,  9.4888e-04, -5.0328e-03, -1.1923e-02,  3.4274e-04,
         3.4844e-03,  2.1328e-03,  5.2334e-04,  9.9407e-04, -5.9989e-05,
         5.5367e-03,  2.4939e-03, -6.4519e-03, -2.0689e-03, -2.3949e-03,
        -1.5108e-03,  3.6593e-03,  7.3564e-04, -1.7012e-03, -3.4852e-04,
         2.8379e-03,  2.1907e-03, -1.1880e-03,  1.2404e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4778e-02,  8.8945e-01,  2.4595e-04,  1.0521e-02, -1.5446e-03,
        -4.4580e-06,  3.9548e-03, -8.4573e-03,  1.1331e-03,  1.2669e-03,
        -2.2740e-03, -1.1046e-03,  2.3400e-03,  2.7842e-03,  2.0236e-02,
         5.0470e-03, -1.6543e-02,  1.1112e-03,  2.4018e-03,  2.0274e-04,
         3.1508e-04,  4.2543e-03,  2.5874e-03, -2.9062e-03,  1.7596e-04,
         1.9497e-03,  2.0439e-03, -3.6946e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2546e-02,  8.9832e-01,  1.7171e-02,  1.6930e-03, -4.2568e-03,
         2.1123e-03, -3.9643e-03,  2.0817e-03, -3.5472e-03, -7.7681e-04,
        -1.7668e-03, -2.4937e-03, -3.3569e-03,  5.1690e-03,  1.6750e-03,
        -3.8903e-03,  6.5876e-03, -1.4304e-02, -1.2053e-03,  2.4541e-03,
         2.2861e-03,  2.9633e-03,  2.0831e-03,  3.2921e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 4.4032e-03,  9.3428e-01,  6.7772e-03,  7.2841e-04,  3.3912e-03,
         3.0566e-06,  1.0603e-02,  6.8773e-04,  1.1243e-03,  8.6101e-05,
        -1.4839e-04,  5.7797e-04,  1.2401e-03,  4.8406e-04,  1.0071e-03,
         1.6620e-03, -4.5801e-04,  1.1177e-03,  2.1636e-03,  1.3831e-03,
         3.4611e-03,  5.1980e-04, -4.1328e-04,  1.2896e-03,  5.1351e-04,
         7.6992e-03,  7.7548e-04, -2.3665e-03,  3.3813e-03, -7.0992e-03,
        -1.5866e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0198,  0.8977, -0.0015, -0.0171, -0.0049, -0.0109, -0.0055,  0.0023,
         0.0077,  0.0043, -0.0097, -0.0184,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4123e-02,  8.9674e-01, -3.1447e-03, -1.9373e-03,  4.7564e-03,
        -8.7884e-03, -4.7669e-03, -3.8161e-04, -7.3148e-04,  4.1740e-03,
         9.2527e-03, -1.5356e-02, -6.0828e-03, -4.9337e-03,  4.8269e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7642e-03, -9.3482e-01, -1.6717e-02,  3.5100e-03, -9.8400e-04,
         1.1123e-03, -4.4900e-03,  1.7737e-03, -1.0792e-03,  6.3273e-04,
         4.5769e-04, -1.6567e-03, -4.3050e-03,  1.3483e-03,  3.1279e-04,
         7.6672e-04, -5.1713e-04,  2.3747e-03, -3.0089e-03,  2.0077e-03,
         1.4850e-04, -9.2762e-04,  4.9720e-03,  4.3616e-04, -5.8343e-04,
        -3.3519e-04, -1.6075e-04,  4.4974e-04,  2.8432e-03, -8.5269e-04,
        -1.0245e-03,  2.7620e-04,  1.3479e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8653e-03,  9.2499e-01, -1.6238e-02, -4.0764e-03,  5.5085e-03,
        -2.3853e-03, -5.0571e-03, -7.2418e-04, -2.3080e-03, -1.2528e-03,
        -8.0005e-03, -3.5919e-03, -5.8438e-03,  1.5402e-02,  9.6661e-04,
         7.8565e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2787e-02,  9.5108e-01,  3.8647e-03, -3.1407e-03, -5.4534e-04,
        -1.5341e-02,  4.6769e-03,  7.9455e-03, -6.2168e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8295e-02,  9.4257e-01,  2.5392e-02, -1.2996e-02, -7.4805e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.3184e-03,  9.1437e-01,  1.3570e-02,  7.2235e-03, -2.2991e-04,
        -6.4807e-03, -1.4800e-04, -4.5068e-04,  1.6515e-03, -2.1686e-03,
         4.3140e-03, -3.2865e-03, -3.7972e-03, -8.8059e-04, -5.1311e-04,
        -5.6034e-03, -2.3833e-03,  3.5236e-04, -3.6400e-04,  1.6092e-03,
        -2.1081e-03,  7.9153e-04, -4.4607e-04,  3.6147e-03,  3.7066e-04,
        -3.6761e-03, -1.2276e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.3898e-02,  9.1016e-01,  2.2699e-02, -5.3286e-04, -1.0579e-04,
         5.2807e-03,  7.3235e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7093e-04,  9.1511e-01, -3.5544e-02,  2.4422e-03,  1.1678e-03,
         1.8228e-03, -4.0250e-03,  1.6773e-04, -7.9817e-04, -7.6524e-03,
         1.8174e-03, -1.1900e-02, -8.3143e-04, -6.1740e-04, -3.9654e-03,
        -1.9229e-03,  7.1956e-04, -5.5521e-03, -5.9406e-04,  7.8532e-04,
        -1.2468e-03, -3.4405e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.8308e-03,  9.3082e-01,  2.2309e-03, -5.2849e-04, -1.3639e-03,
         7.1920e-03, -4.5482e-03, -9.4658e-04, -4.8855e-03, -2.4384e-03,
        -1.2128e-02,  5.6414e-03, -4.3017e-03,  6.9276e-03,  5.3553e-04,
        -1.8797e-03,  2.6829e-04,  1.6951e-03,  4.8344e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9571e-02,  9.2437e-01,  4.0532e-03,  5.8005e-03,  1.5850e-03,
         3.8399e-03, -8.5558e-04, -2.4861e-03,  3.0873e-03,  1.9717e-03,
        -2.1298e-03, -1.8240e-03, -5.7489e-06,  4.8158e-04, -3.4383e-03,
        -1.7940e-03, -7.3592e-03, -2.5977e-03, -3.0359e-04,  4.2479e-03,
        -1.1377e-03,  1.9748e-03,  8.5618e-04,  5.7588e-05, -2.9005e-04,
         2.7425e-05, -1.1796e-03, -8.1701e-04,  2.4111e-04,  8.0177e-04,
         8.1297e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-1.4549e-02, -9.1158e-01,  1.0544e-02,  3.6474e-03,  5.9098e-03,
         3.9149e-03,  2.2287e-03,  9.5618e-04, -2.5317e-03, -1.1748e-03,
         3.0733e-04,  1.1236e-03,  2.5906e-03, -9.3345e-03,  6.4371e-05,
        -1.3016e-03, -1.9679e-03,  4.8001e-04,  1.0089e-03,  6.9894e-04,
         9.7329e-04,  1.6740e-03, -7.3892e-04, -4.5531e-04, -2.3552e-03,
         3.4339e-03,  1.1484e-03,  3.2847e-03, -2.4317e-04,  1.1385e-03,
        -3.6682e-03,  4.9677e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.2921e-03,  9.3798e-01, -2.6964e-03,  7.8541e-04,  1.2116e-03,
         9.8831e-04,  2.9807e-06,  1.9839e-04,  1.7000e-02, -1.1765e-03,
         4.6570e-05,  2.7574e-03,  7.1487e-05,  8.0131e-04, -2.8529e-05,
         6.3817e-03,  2.1448e-03,  2.0981e-03, -5.4515e-03,  1.6087e-03,
         4.8957e-03,  2.3866e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6239e-03,  9.1596e-01,  3.4390e-03,  6.1287e-04,  3.3946e-03,
         2.9302e-03, -2.3157e-03, -7.6249e-04,  5.2050e-03,  4.9477e-04,
         1.2930e-03,  1.3050e-05, -1.9886e-03, -5.8107e-04, -5.7612e-03,
         6.5094e-03, -2.3998e-03,  1.1668e-02, -1.3424e-03,  1.8323e-03,
        -3.4930e-03, -3.3317e-03,  2.6522e-04,  7.6066e-03, -5.6708e-04,
         2.0878e-04, -4.1805e-04, -5.2397e-03,  3.4194e-03,  3.3262e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.8298e-03,  9.5604e-01, -1.1058e-02, -6.2690e-04, -1.6577e-03,
         1.1631e-03,  7.3840e-04,  1.2800e-03, -3.2531e-04, -2.2946e-03,
        -4.4605e-03,  1.2429e-03, -2.7913e-03, -1.5028e-03, -4.4104e-03,
        -8.6705e-04, -6.1153e-04, -1.1003e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1496e-02,  9.0036e-01,  1.5306e-02, -1.4564e-03,  1.2838e-03,
        -5.8684e-03, -6.8321e-04,  3.6380e-03, -2.8651e-03,  2.2087e-03,
         2.2317e-04, -1.2834e-03,  6.0251e-04,  1.2112e-03, -1.0697e-03,
         8.4305e-04,  7.3002e-03,  6.3613e-04, -1.2034e-03,  2.1012e-04,
         1.4702e-03, -3.6636e-03,  4.7310e-03, -1.2030e-03,  4.2953e-04,
        -1.7777e-03, -1.1373e-03, -2.7740e-04,  2.4265e-03,  2.9568e-04,
         1.4678e-04,  1.6629e-03, -3.7016e-04, -2.2993e-03, -5.6689e-04,
         2.3803e-04,  7.5515e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7622e-02,  8.9457e-01,  1.2483e-02,  2.0545e-02, -6.6144e-04,
        -1.1160e-03, -5.5666e-04, -5.2973e-03, -1.3717e-03, -1.1171e-03,
         1.4676e-04,  1.3575e-03, -9.6744e-03, -1.7178e-03,  5.1944e-03,
        -1.9084e-03, -1.0115e-02, -1.5317e-03, -5.0512e-04, -4.0311e-03,
        -2.1158e-03, -1.3301e-03,  2.2322e-03,  2.7966e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1267e-04,  9.4738e-01,  5.6654e-04,  2.7529e-04, -2.0967e-03,
         1.3174e-03, -9.1815e-04,  3.6685e-03,  3.3483e-03, -2.8141e-03,
         7.0775e-04, -7.3445e-05,  8.1423e-04,  1.2090e-03, -1.2025e-03,
         7.9198e-04,  6.1837e-04, -1.5734e-03,  6.1918e-04, -2.0592e-03,
         1.3415e-04, -1.4155e-03,  1.1431e-04,  1.1880e-03,  8.0887e-06,
        -9.9528e-04, -1.4844e-03, -1.8418e-03, -2.4126e-04, -1.6913e-03,
        -5.0379e-03, -1.8672e-03,  2.3746e-04,  1.3423e-03,  4.7505e-04,
         8.8756e-04,  1.3221e-04, -7.4472e-04,  4.1633e-04, -1.5021e-04,
         3.4739e-04,  1.2299e-03,  2.0604e-04,  1.9610e-04, -1.5900e-04,
         5.0876e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9891e-03,  9.0214e-01, -2.0911e-02, -3.9132e-04, -7.8060e-03,
         2.9043e-04, -2.2437e-03, -1.6297e-03,  1.1806e-03, -1.5434e-02,
         3.3192e-03, -3.7856e-03, -1.8568e-03, -1.4215e-03,  2.1441e-03,
        -5.7942e-04,  2.3308e-03, -2.2003e-03, -5.8354e-03,  1.4196e-03,
         7.2351e-03, -2.8618e-03,  7.9902e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.5427e-03,  9.4723e-01,  1.3408e-02, -3.7243e-03,  6.1356e-03,
         1.4618e-03,  5.0199e-05,  3.7014e-03,  7.2398e-03, -5.2884e-03,
         2.2213e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7604e-02,  8.9923e-01,  8.9530e-03,  6.5390e-03,  6.8388e-03,
        -2.2240e-03,  4.4021e-03,  8.5517e-03,  5.4322e-03,  3.9445e-03,
         4.5492e-04, -2.9586e-04, -7.8493e-03, -7.6853e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.9374e-03,  9.1991e-01, -8.3774e-03, -2.0426e-02, -1.1133e-03,
        -7.6151e-04, -1.1352e-02,  3.7023e-04,  8.9386e-04, -1.7472e-02,
        -1.0198e-02,  1.8836e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7700e-02,  8.8800e-01,  1.9276e-04,  6.4747e-03,  2.9714e-03,
         2.1587e-03, -3.8466e-03,  5.3403e-04,  2.3922e-03,  3.3988e-03,
        -9.6617e-04,  2.0658e-03, -4.3794e-04,  2.1595e-03, -7.7267e-04,
        -1.3077e-03, -6.6153e-04,  8.5215e-04,  6.6547e-04, -3.0184e-03,
        -8.4467e-04,  1.0837e-04,  1.2792e-03,  2.3984e-03,  6.1406e-04,
         3.5141e-03,  1.1428e-02, -2.2298e-03, -3.2452e-03,  1.0372e-03,
         1.3234e-04,  5.1083e-03, -2.0039e-04, -2.4102e-04,  7.7107e-04,
        -2.0519e-03, -8.1238e-04,  7.0901e-04, -4.5698e-04, -2.2187e-03,
        -1.3364e-03,  6.5006e-04,  1.5521e-03, -1.1052e-03,  1.6862e-03,
        -9.9816e-04,  2.8928e-03, -3.0050e-04,  6.7830e-04, -9.0800e-04,
         9.7364e-04,  6.5258e-04, -1.0202e-03,  2.0634e-03,  3.2002e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 1.3581e-03,  9.5364e-01,  3.0248e-03, -3.8596e-03,  2.9241e-03,
         7.8074e-04,  5.8181e-03,  1.5988e-03,  7.6102e-04, -2.2331e-03,
        -1.3301e-03, -1.7719e-03,  1.1594e-03, -1.1698e-03, -9.4626e-04,
        -7.3406e-04,  2.0154e-05,  6.4589e-03,  9.0262e-04, -1.8056e-03,
         3.6480e-04, -5.4738e-03, -1.8655e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.7912e-04,  9.6452e-01,  2.4235e-03,  2.5842e-03,  2.1466e-03,
        -2.4362e-03, -8.0945e-04,  3.6918e-03, -2.3105e-03, -6.6507e-04,
        -7.5436e-04,  3.4261e-04, -4.3113e-03,  6.7979e-03, -5.1031e-04,
         5.1287e-05, -4.2995e-03,  7.6392e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0064,  0.8922, -0.0159, -0.0037,  0.0079, -0.0080,  0.0047, -0.0012,
         0.0038, -0.0106, -0.0050, -0.0022,  0.0091, -0.0093, -0.0036, -0.0075,
        -0.0032, -0.0058,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3052e-02,  8.9430e-01, -1.0258e-02, -3.5316e-04,  3.4202e-03,
         4.7611e-03, -3.5768e-03, -4.1841e-03, -1.9174e-04, -7.5265e-04,
         2.2926e-03,  9.3894e-03,  6.8871e-03,  8.8725e-03, -2.7712e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3061e-02,  8.9711e-01,  2.9774e-03,  8.7168e-03, -1.4140e-03,
        -2.4178e-03, -3.1423e-03,  2.1237e-03,  5.3103e-03,  5.5304e-04,
        -1.8030e-03, -2.6081e-03,  2.6887e-04, -1.4675e-03, -2.3588e-03,
         6.5393e-03,  1.0001e-03, -1.9504e-03, -2.8838e-03,  5.4935e-03,
         7.0978e-03,  2.2236e-03,  6.6947e-04,  4.0246e-04,  1.2977e-02,
         1.3610e-03,  3.1868e-03, -6.4740e-04,  7.6066e-04,  3.2089e-03,
        -4.0716e-03, -1.9429e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.4679e-03,  9.1317e-01,  1.1354e-04, -1.2420e-02, -1.9188e-03,
         2.8921e-04, -1.3087e-03, -3.7901e-03,  1.2054e-03,  3.8291e-03,
         4.7252e-03, -4.1047e-03,  4.5932e-03,  2.2475e-03, -1.4530e-03,
         2.8700e-03,  1.2439e-03,  2.2543e-03, -9.8370e-04,  8.9394e-04,
         3.7953e-04,  6.6274e-03, -3.4291e-03, -3.7354e-03, -1.0078e-03,
         1.2739e-03,  5.1403e-04,  1.3475e-03, -1.2925e-03,  3.7041e-03,
         7.8145e-05,  3.2763e-03,  3.4522e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.9303e-03,  9.1163e-01, -2.3252e-03,  1.9113e-03, -1.1069e-03,
        -2.1427e-04,  9.3578e-04, -5.0408e-04,  4.2431e-03,  4.5545e-03,
         5.1906e-04, -4.0254e-03, -5.2132e-03, -4.0902e-03, -7.1733e-04,
        -1.3129e-03,  2.7882e-03, -1.3820e-03, -3.1677e-03, -9.8583e-05,
        -2.9649e-04, -4.5870e-03, -1.6399e-03,  1.6871e-03,  3.1818e-05,
         1.3275e-03,  5.7654e-04, -3.9367e-03, -1.1526e-03, -1.0940e-03,
         2.0474e-04,  1.0443e-03, -1.7840e-04, -7.2738e-04,  1.9928e-04,
        -2.2715e-03, -1.1984e-03, -6.4991e-04,  3.5669e-04,  8.8099e-04,
         6.7574e-04, -2.4681e-03, -8.1300e-04,  7.5710e-04, -8.3467e-04,
        -4.9484e-04, -1.1465e-04,  1.4509e-03,  4.0302e-04,  2.0034e-03,
         7.8256e-04,  2.3710e-05, -1.4167e-04, -1.3602e-05, -1.0560e-03,
         1.1276e-03, -3.0472e-03,  8.0876e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0727e-03,  9.0392e-01, -1.2900e-02, -7.5106e-04, -1.9186e-03,
        -4.4354e-03,  1.3616e-03,  6.6436e-04,  2.3815e-03, -7.5179e-03,
         2.3938e-03,  4.4969e-03,  1.0147e-03,  6.7492e-03,  7.3765e-03,
        -5.9145e-04,  2.3646e-03,  9.2935e-05, -3.0214e-03, -6.9496e-03,
        -1.1137e-03, -3.9637e-04, -1.5475e-04, -6.0141e-03,  6.7002e-03,
        -1.2649e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4390e-02,  8.7792e-01, -3.1524e-02, -4.5481e-03, -6.1387e-03,
        -3.6161e-03, -4.1923e-04,  8.7268e-03,  4.7394e-03, -5.0589e-03,
         2.8734e-03, -2.3852e-03, -1.0395e-03, -3.7937e-03, -7.1645e-04,
         3.5292e-04, -1.7292e-03,  6.4126e-04,  2.0299e-03, -1.0236e-03,
        -1.8169e-03, -3.6538e-03, -3.4720e-04, -4.1098e-03, -6.4021e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.1000e-03,  8.7081e-01,  1.2027e-02,  1.0823e-02,  9.4609e-03,
         4.6087e-04,  3.4080e-03, -1.1137e-05,  2.1865e-03, -1.7694e-03,
         9.6456e-04,  3.4850e-03,  1.5898e-03,  6.9377e-03,  1.1717e-04,
        -4.3632e-04, -1.2961e-03, -7.1771e-05,  1.3381e-03,  2.7134e-04,
         1.0096e-04,  5.4935e-03, -9.2792e-04,  2.1051e-03,  7.6559e-04,
        -4.9867e-04,  3.4062e-03,  9.7012e-04,  2.4097e-03,  3.9531e-03,
         3.4653e-03, -1.5632e-03,  3.3379e-04,  2.6646e-03, -1.1358e-03,
        -8.5119e-03,  1.8152e-03,  1.1012e-04, -2.1779e-03,  1.6675e-03,
         1.8760e-03,  1.1818e-03,  4.7960e-03,  2.0705e-03, -1.3439e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8185e-02,  9.3603e-01,  9.7929e-04, -8.9349e-04,  5.4442e-03,
         4.2253e-04, -1.7033e-03,  7.7827e-04,  6.2224e-04,  7.9637e-04,
        -1.9613e-03,  1.9419e-03,  1.8101e-03, -1.3000e-03,  3.8722e-03,
         3.0832e-03,  5.0508e-03,  3.3087e-03,  1.0418e-03,  2.7575e-03,
         2.9120e-03,  4.1217e-03, -9.8313e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0567,  0.9014, -0.0018,  0.0086, -0.0045,  0.0033,  0.0106,  0.0039,
        -0.0044,  0.0012, -0.0025, -0.0012,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-8.7417e-03,  9.3701e-01,  5.5750e-03, -6.0055e-03,  4.5037e-03,
        -2.2541e-03, -9.6267e-04, -1.8550e-03,  4.4119e-03, -3.2688e-03,
        -2.7655e-03, -4.2058e-04,  9.3226e-04, -3.7635e-03,  4.3915e-03,
         2.5133e-03,  7.0786e-04,  2.2019e-03, -1.1017e-03, -5.7575e-03,
        -4.5591e-05,  6.1716e-04, -1.9024e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4576e-02,  9.2230e-01, -5.8282e-03,  2.6130e-03, -2.0162e-03,
        -4.6133e-04, -2.0986e-03,  1.8779e-04, -2.8400e-03,  3.7498e-03,
         1.9765e-03, -2.4361e-03, -2.7916e-03,  1.4364e-03,  7.9056e-04,
         6.7298e-04, -4.5278e-03, -2.6034e-04, -2.2040e-03,  1.2804e-03,
        -2.4378e-04, -1.1873e-03, -5.5573e-03, -3.5787e-04, -2.2813e-03,
        -3.4031e-04, -2.1652e-03,  8.1674e-04, -1.3715e-03,  1.2848e-04,
         2.2965e-04,  8.5430e-03, -1.7340e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0613e-02,  9.2690e-01,  8.5192e-04, -4.9368e-03,  2.1942e-03,
        -2.0640e-03, -1.9491e-02,  1.1929e-03, -2.1328e-04,  3.5589e-03,
         2.5218e-03, -6.1345e-03,  9.5404e-04, -3.6438e-03, -2.7773e-03,
        -7.0884e-03,  3.3350e-06,  4.3814e-03,  4.7709e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0028,  0.9127, -0.0200,  0.0064, -0.0082,  0.0081,  0.0126,  0.0026,
         0.0063, -0.0025, -0.0011, -0.0012,  0.0079, -0.0075,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6693e-02,  9.4043e-01,  2.3499e-03, -4.2210e-03, -2.2363e-03,
         9.3900e-05,  2.1788e-03, -2.0113e-03,  8.0515e-03, -1.3757e-03,
         1.2860e-03, -1.4792e-03,  2.3850e-03, -4.3435e-03, -2.9170e-04,
         3.3607e-04, -3.5494e-04, -8.9027e-04, -1.1699e-03, -3.6650e-03,
        -1.2008e-03, -2.9520e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8773e-03,  9.1834e-01, -1.0029e-02, -2.5442e-03,  3.6021e-04,
         1.6626e-03,  4.2031e-03, -5.6819e-03, -3.1360e-03,  1.8284e-03,
        -9.7695e-04,  2.0431e-03, -4.0926e-03, -5.5763e-04, -1.0964e-03,
         2.1102e-03,  8.3053e-03,  4.8004e-03,  5.4162e-04,  4.1814e-03,
        -2.3458e-03, -3.6046e-04, -5.8805e-04, -1.2144e-03, -5.4527e-03,
        -1.1239e-03,  3.2378e-03, -1.3815e-03,  2.2882e-03,  3.6398e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.2492e-03,  9.0264e-01, -7.8838e-03, -1.0826e-02, -2.2663e-03,
         1.1007e-02, -4.9781e-03, -4.9877e-03, -4.5287e-03,  8.9751e-04,
        -1.1666e-02, -2.0899e-03,  1.1846e-03, -6.1367e-03, -4.3818e-03,
        -1.1140e-02, -5.1418e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.1765e-02,  9.0605e-01,  7.2655e-04, -1.4737e-03,  2.6371e-03,
         8.5030e-03, -1.9015e-03,  6.4376e-03, -7.2957e-03,  1.1179e-02,
         2.0323e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8722e-03,  9.3226e-01,  1.3617e-02,  2.9355e-03,  4.0764e-03,
         2.3322e-02,  1.2187e-03, -6.5343e-04,  5.1722e-04,  2.2037e-03,
        -1.4405e-03,  1.9899e-03,  2.4724e-03,  2.5830e-04,  5.1656e-03,
        -1.6988e-03,  2.3004e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.1807e-03,  8.5678e-01, -6.8375e-02,  1.7222e-02, -2.6449e-03,
        -4.6249e-03, -7.8978e-03,  9.3608e-04,  1.4279e-03, -5.5865e-03,
        -4.8258e-04, -2.4404e-03, -2.7546e-05,  2.4783e-03,  6.1070e-03,
         4.3352e-04, -1.9426e-04, -2.0577e-03, -2.1146e-03, -1.0154e-03,
         2.1980e-05, -4.4233e-04,  3.2872e-03,  1.0964e-03,  3.4859e-03,
        -3.6362e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2530e-02,  8.9827e-01,  1.0867e-02, -1.6359e-03, -2.5566e-03,
         4.7155e-03,  2.5602e-03, -3.2839e-04, -1.3829e-03,  2.5945e-04,
        -8.4223e-03,  8.1302e-04, -1.4269e-03,  5.3204e-04, -1.6827e-03,
        -3.4349e-03, -1.0691e-03,  1.4613e-03, -1.0218e-03, -3.9816e-03,
        -1.7104e-03, -2.5310e-03, -2.2935e-03,  9.2528e-04,  5.2382e-04,
        -2.0003e-03,  1.1717e-03, -2.5856e-03, -1.3452e-03,  5.9570e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1451e-02,  9.4110e-01,  1.1069e-03, -1.6235e-04,  1.3383e-03,
        -1.3953e-03, -1.0968e-02,  2.5139e-03, -4.7093e-03, -2.3776e-03,
         1.4719e-03, -3.9175e-03,  5.7074e-05, -2.8836e-03,  6.4996e-04,
         5.3581e-04, -2.2581e-03,  5.7541e-03, -5.3451e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-1.6160e-02,  8.9978e-01, -1.4496e-02, -3.7570e-03, -9.9469e-04,
        -1.2345e-03,  1.7904e-02, -3.1002e-03, -5.9822e-04,  1.1324e-03,
         1.6533e-04,  3.2930e-03,  8.6814e-04,  7.1510e-03, -8.8790e-04,
         3.0507e-03,  6.3326e-03,  2.5294e-03, -9.5567e-05,  5.5245e-03,
        -2.5463e-03, -8.1254e-05, -4.1216e-05, -3.6006e-04, -7.4253e-04,
        -1.7370e-03, -5.4381e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0023e-02,  9.2440e-01,  2.6828e-03, -3.6542e-03,  7.0237e-03,
         4.6263e-03,  2.2616e-03,  1.0213e-02,  6.1457e-03,  2.7211e-03,
        -4.7324e-03,  1.2770e-02,  4.1653e-04, -8.3317e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0384e-02,  9.3314e-01, -3.0708e-03,  1.4540e-03, -2.1242e-04,
        -2.8939e-03,  1.9178e-03,  1.4024e-03,  2.8513e-03, -7.9158e-04,
         2.0434e-03,  2.0492e-03, -8.6895e-03,  2.2432e-03, -4.6765e-03,
        -2.4278e-03,  2.7124e-04, -1.9061e-03, -2.4589e-03, -2.5966e-03,
        -1.9767e-03, -1.2336e-03, -2.3004e-04,  1.0888e-03, -8.1791e-04,
         7.1706e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1023e-02,  8.9502e-01,  1.7813e-03, -5.7436e-03, -4.3648e-03,
        -1.2257e-02, -6.1679e-04,  2.1416e-03,  5.6305e-05,  1.5533e-02,
         3.1002e-03,  5.7704e-03, -9.7921e-04,  1.6132e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6654e-02,  9.2041e-01,  4.0951e-03, -3.8659e-03, -1.4019e-03,
        -1.7642e-04, -4.3779e-03, -4.0162e-03,  1.7828e-03, -2.7049e-03,
         9.1693e-04, -1.5293e-03,  2.5901e-03,  1.1049e-03,  2.5542e-03,
         1.7003e-03,  2.3070e-03,  1.4585e-03, -8.1586e-04, -2.8626e-03,
        -1.2702e-03, -1.2116e-04,  9.7939e-04,  1.2223e-03, -3.3985e-04,
         1.3316e-03, -4.3729e-03, -2.9911e-03, -6.7657e-04, -1.5108e-03,
         5.8703e-03,  1.9902e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.6331e-03,  9.3812e-01,  5.2005e-03,  2.0168e-03, -4.7579e-04,
        -1.7322e-03, -4.3757e-03,  7.1394e-04,  1.9263e-03,  2.8030e-03,
         3.7154e-03,  6.2682e-03,  3.6307e-03, -6.7237e-05,  1.3440e-03,
         2.2280e-03,  4.2579e-03, -6.6482e-04,  2.5705e-03,  4.1926e-03,
         7.9135e-04,  9.8101e-04,  1.2542e-03, -2.3192e-03, -7.1754e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0021,  0.9791,  0.0084, -0.0016,  0.0051, -0.0038,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1666e-02,  8.8979e-01,  1.1503e-02,  2.7233e-05,  1.7944e-03,
        -1.9118e-03,  4.4823e-05, -2.8109e-03, -6.5082e-04, -1.0470e-03,
        -2.1995e-03,  4.8140e-03, -6.6331e-04,  3.7904e-03, -1.8320e-03,
        -7.4357e-04,  2.8078e-03,  1.4139e-03,  4.5032e-04, -1.1568e-03,
        -1.9188e-03,  2.1226e-03, -1.9948e-03, -2.1692e-04,  7.8673e-04,
         2.9208e-03, -9.3331e-04,  3.1555e-03, -1.3879e-03,  3.5691e-03,
         3.1135e-03, -1.0025e-03,  5.0868e-04, -1.1949e-03,  4.6259e-04,
        -1.9965e-03,  1.1598e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9893e-03,  9.1258e-01, -6.9707e-03, -4.4780e-03,  1.3878e-03,
        -4.2561e-03,  1.6505e-03,  1.1194e-03, -8.9885e-04, -6.5829e-04,
        -2.0294e-03,  9.5976e-04,  2.0622e-03,  2.3503e-03,  1.3265e-04,
        -2.8915e-03, -3.3733e-03,  8.3476e-03, -3.7417e-04,  4.5140e-03,
        -7.5798e-03, -8.8249e-04, -5.2614e-03,  9.2616e-04, -7.5785e-04,
         5.6099e-05,  2.2016e-04, -6.9041e-03, -1.2325e-03,  3.6574e-03,
        -1.8982e-03,  6.0952e-04, -6.9136e-04, -2.6618e-03, -1.6394e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6861e-02,  8.4711e-01, -7.5124e-02, -4.8054e-03, -5.2016e-03,
         6.6104e-04, -6.5319e-03, -4.5005e-03, -3.4282e-03,  1.8333e-02,
        -8.6280e-03, -2.5308e-03,  9.4538e-05,  2.6466e-04,  5.9212e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1675e-02,  9.2996e-01, -1.4368e-02,  3.9684e-03,  2.0589e-04,
         2.6735e-03,  1.5871e-03, -6.7776e-04, -6.3440e-03, -2.1782e-03,
        -4.1814e-03,  1.7105e-03, -1.4793e-03, -2.3312e-03,  1.0072e-03,
        -2.8867e-04, -1.8151e-03, -1.3438e-03, -1.0945e-03,  5.8621e-04,
        -2.9175e-05,  4.6642e-04, -3.6470e-03, -4.0093e-04,  5.5389e-04,
        -1.5300e-04,  2.7957e-04, -1.2187e-04, -2.4872e-04, -9.4237e-04,
        -1.7711e-04, -1.6600e-03,  1.8413e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0862e-02, -9.0362e-01,  3.5456e-03,  3.4465e-03,  4.9869e-03,
         1.9743e-02, -1.3387e-03,  1.3007e-03,  5.3211e-03, -6.7695e-04,
        -3.9534e-04,  2.6534e-03,  8.1672e-04, -3.8736e-03, -2.6136e-03,
        -1.0588e-03, -9.6506e-05,  2.9766e-03, -3.9182e-04,  8.9993e-04,
         2.0786e-03,  6.4386e-04, -3.0483e-04,  9.0987e-04, -1.3826e-03,
         3.5201e-03, -1.9009e-03, -4.3474e-03,  4.2904e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 1.8909e-02,  9.2031e-01, -1.3401e-02, -4.5073e-03, -9.7640e-03,
        -8.9084e-03, -7.7227e-04, -1.5905e-03, -4.6532e-04, -3.5526e-03,
        -4.4265e-03, -4.9786e-03,  3.5164e-03,  5.5955e-04, -4.3387e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7453e-03,  9.1883e-01,  4.2266e-03,  6.3262e-05, -1.6245e-03,
        -3.0194e-03,  1.8671e-03,  2.8284e-03,  7.4513e-04,  2.4934e-03,
        -8.3977e-03, -1.3349e-03, -2.3145e-04,  3.4833e-03,  4.5248e-05,
        -1.8997e-03, -8.9427e-04, -1.0952e-03, -1.4725e-03,  2.2483e-03,
        -1.8894e-04, -2.0010e-03, -1.5726e-03, -1.2882e-03,  3.9541e-04,
        -1.0020e-04, -3.3319e-03,  1.2478e-03, -3.4706e-03, -1.4655e-03,
         6.8343e-04, -3.2158e-04,  4.8576e-04,  8.1684e-04, -1.0673e-03,
        -1.4339e-03,  6.6647e-04, -1.4259e-04,  1.6269e-03, -1.4591e-03,
        -1.5038e-03,  2.1987e-03, -9.7080e-04, -6.2777e-04, -2.1661e-04,
        -3.4005e-03,  8.9006e-04, -5.5018e-03,  2.3769e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.6643e-03,  9.0409e-01,  1.2268e-02,  3.5682e-03,  4.9700e-03,
         1.7654e-03,  3.3502e-03, -3.6195e-04,  8.0715e-03,  5.2206e-03,
         1.8576e-03,  1.1572e-03, -7.5138e-04,  2.5279e-04, -3.3209e-03,
        -3.2923e-04,  6.2354e-04,  1.5468e-03, -5.6053e-05, -5.6758e-04,
        -3.7501e-04,  2.0986e-03, -1.5465e-03, -1.1884e-05, -5.3746e-04,
         3.5082e-04, -4.1546e-03,  1.0078e-02,  7.6223e-04, -8.7076e-04,
         1.4051e-03,  1.4101e-02,  3.0584e-04, -1.1096e-03,  5.0195e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2888e-03,  9.4243e-01, -3.5176e-03,  4.6140e-03,  5.4695e-05,
         5.6804e-03, -7.6969e-03,  2.2522e-03,  8.0051e-04, -3.5286e-03,
        -1.7089e-03, -2.7101e-03, -2.3519e-03, -8.9509e-04,  9.6343e-03,
         9.6113e-04, -1.1401e-03, -1.2140e-03, -6.5228e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.0228e-03,  9.0681e-01, -1.2843e-02, -2.6859e-03,  4.3814e-03,
         2.5279e-03,  6.4253e-04, -5.0747e-03,  3.1415e-05, -2.1436e-03,
        -3.8081e-03, -2.7226e-03,  1.2620e-03, -1.9340e-02, -3.0709e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5596e-02,  9.4340e-01,  3.1489e-04, -1.0708e-03, -2.2936e-03,
         4.1216e-03, -3.4714e-03,  5.9786e-04, -1.3977e-03, -1.1813e-03,
        -2.1347e-03,  1.9023e-03, -3.7528e-03, -1.4886e-03, -4.2365e-04,
        -4.5046e-03, -1.2812e-03, -6.6864e-04, -5.1837e-03, -9.3908e-04,
        -1.1203e-03,  3.1595e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1995e-03,  9.1534e-01, -1.5585e-02, -1.7767e-03, -6.0396e-04,
        -4.0424e-03,  7.5855e-04, -8.9346e-04, -7.1705e-03, -7.1022e-03,
        -5.4727e-04,  5.4323e-03, -4.6317e-03, -1.2491e-03,  1.9579e-03,
         4.4606e-04, -1.2191e-03, -4.4764e-03,  5.2146e-04,  9.6287e-04,
         6.5226e-04,  4.6865e-04, -8.4721e-04, -3.6420e-03,  1.2247e-03,
         3.4886e-04,  1.2574e-03, -2.4092e-03, -5.2379e-03,  7.9986e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0316,  0.8929, -0.0137, -0.0072, -0.0099, -0.0037, -0.0019, -0.0068,
        -0.0052,  0.0043, -0.0228,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5708e-02,  9.0666e-01,  6.0413e-03, -5.6249e-04,  4.6422e-03,
         1.1319e-03,  5.5720e-03, -3.7672e-03,  1.0479e-03, -1.0764e-03,
         7.0228e-03,  5.5112e-03, -8.6963e-04,  3.9634e-03,  2.3126e-03,
        -1.9239e-03, -2.4490e-04,  5.1734e-04, -1.2315e-03, -1.0107e-03,
         1.0552e-03, -8.1270e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.2056e-03,  8.9213e-01,  7.7008e-03, -3.5924e-03, -3.7475e-03,
        -2.1843e-02, -7.8012e-04,  3.0931e-03, -2.7965e-03, -1.7055e-03,
        -6.8902e-03,  2.6069e-03,  2.2291e-03, -4.9440e-03, -3.0525e-03,
        -2.7719e-03, -2.0697e-04, -4.3795e-03,  4.3696e-04, -3.3834e-03,
        -1.0625e-03, -3.2675e-03, -5.0773e-03, -3.5373e-04,  9.1128e-04,
        -1.7658e-03, -2.4175e-04, -4.2699e-05, -6.4788e-03,  2.3240e-03,
         2.5189e-03,  1.4001e-03,  1.0577e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3823e-02,  9.1014e-01,  1.9628e-02, -8.7324e-04,  4.3263e-04,
        -7.4511e-04,  8.8076e-04, -7.7878e-03,  2.6784e-03, -4.7568e-04,
        -6.1433e-03,  4.4858e-04,  1.3589e-03,  9.0735e-04, -1.9253e-04,
         1.1353e-02,  3.5822e-03, -2.1121e-03, -2.5784e-04, -2.1444e-03,
         1.6354e-03, -3.6129e-03,  3.4960e-03, -1.6690e-04, -2.4585e-03,
         1.9269e-04,  1.6753e-03,  7.9722e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2853e-04,  9.2759e-01, -5.1340e-03, -1.0454e-03,  2.6415e-03,
         2.8816e-03,  1.9998e-03,  6.1853e-03, -1.1230e-02,  2.3960e-03,
        -1.1009e-02, -5.4252e-04, -1.1154e-03,  1.4531e-03, -1.1746e-03,
        -6.2957e-04,  1.3921e-03, -2.6925e-03,  7.8011e-04, -1.4002e-03,
        -1.1368e-03, -8.0288e-04,  1.1446e-03,  1.1610e-02,  1.5863e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 5.1353e-03,  9.0425e-01,  2.3782e-02,  2.4434e-03, -9.3084e-04,
        -2.3989e-04, -3.0527e-03,  1.7427e-03,  7.0859e-04, -1.2496e-03,
         1.3553e-03,  2.9899e-03, -1.8238e-03,  9.0035e-03, -7.0010e-04,
        -5.6106e-03, -5.8548e-04,  1.9071e-03, -1.2877e-03, -1.3298e-03,
         6.0054e-03,  1.0944e-03,  3.2951e-04,  3.9872e-04,  4.1543e-03,
        -5.9296e-04, -6.4961e-04, -2.9551e-03,  1.1436e-04,  6.9966e-04,
        -2.3516e-03,  1.1543e-03,  9.3393e-04,  1.3466e-03, -1.6951e-03,
         2.1573e-03, -3.4951e-04,  3.2104e-04, -1.5565e-03,  8.1962e-05,
         9.3428e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8113e-02,  9.0576e-01, -4.3337e-03, -8.5726e-03,  7.5142e-03,
        -7.3237e-03,  1.2915e-03,  4.5526e-03, -1.7455e-03, -8.6523e-04,
        -2.4745e-03, -3.8983e-03,  4.7491e-04, -1.7930e-03,  5.7219e-03,
        -3.5025e-03, -3.9595e-04,  2.0988e-03, -1.2523e-03, -9.0458e-04,
        -1.3805e-03, -6.5922e-03, -1.0246e-03, -1.4234e-03,  6.9928e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.7603e-03,  9.2711e-01,  1.0996e-02, -5.9788e-03, -1.6565e-03,
        -3.2682e-04,  1.2081e-02, -3.9318e-03, -3.8806e-03, -2.3108e-04,
        -2.7093e-03,  3.4719e-03, -1.8804e-04, -4.1137e-04, -2.6115e-03,
        -1.0237e-03,  9.5439e-04,  9.7570e-04, -6.7740e-03,  4.0837e-03,
         3.8410e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7675e-02,  9.1067e-01, -3.5145e-04, -1.3269e-02, -9.1926e-03,
         2.4663e-03, -1.6509e-03,  6.5839e-04, -6.1396e-04, -5.0985e-03,
        -4.9543e-04, -9.4986e-03,  1.1802e-03, -8.4994e-03, -6.4268e-03,
         3.6867e-03,  1.9941e-03, -6.5777e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.8090e-03,  9.1656e-01,  8.6566e-03, -5.7584e-03, -4.3855e-03,
         1.5332e-03, -8.1760e-04, -3.7146e-03, -2.2666e-03, -7.8360e-04,
         2.4766e-04, -2.1596e-03, -9.4651e-04, -2.3377e-03,  1.8246e-04,
         6.1737e-03, -3.9432e-03, -2.9586e-03, -1.5879e-03, -6.0090e-03,
        -1.6361e-03, -4.3607e-04,  2.2293e-04, -7.3098e-04,  2.0254e-03,
        -1.0319e-03, -1.2605e-03,  6.0246e-04, -1.4590e-03, -1.4127e-03,
         1.2025e-04,  4.7510e-04,  8.0780e-05, -9.4341e-05,  4.6341e-03,
         2.1226e-03, -1.9962e-03, -2.8269e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3158e-03,  9.4048e-01, -1.0449e-02, -2.3955e-03, -6.9263e-04,
         4.2097e-03, -2.2548e-03, -4.1230e-04, -2.6126e-03, -1.2241e-03,
         7.3151e-04, -2.8119e-03, -2.0298e-03, -7.9230e-03, -5.2661e-03,
        -3.1795e-03,  1.0526e-03, -6.1621e-03,  3.2105e-04, -4.4731e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0112,  0.9287, -0.0059, -0.0018,  0.0090, -0.0095,  0.0033,  0.0104,
         0.0018, -0.0021,  0.0036, -0.0011, -0.0019,  0.0097,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1835e-02,  9.0378e-01,  1.1421e-03,  1.6623e-03,  2.5355e-03,
        -2.0029e-03, -1.7921e-03,  2.5818e-03,  2.9671e-03,  1.3032e-03,
        -3.1191e-03,  4.6187e-03,  1.5316e-03,  6.5063e-03,  3.4795e-03,
         3.1169e-03,  2.9497e-03,  3.3983e-05, -8.8212e-04,  3.3878e-03,
        -3.2001e-03,  1.2136e-02,  1.6536e-03,  3.3798e-03, -2.0003e-04,
         4.3332e-04,  1.2319e-03, -1.7699e-03, -9.0575e-04, -2.8640e-03,
        -3.1379e-03,  7.8625e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8913e-03,  9.3606e-01, -4.2372e-03, -5.1529e-04, -8.9608e-03,
        -6.4940e-03,  1.8428e-03, -1.4245e-02, -3.5621e-03, -2.5833e-03,
         7.5300e-04,  8.2204e-04,  1.9501e-03,  4.4451e-04, -1.1298e-03,
        -1.7419e-03, -1.7997e-03, -1.7535e-03, -3.9951e-03, -5.2231e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5249e-02,  8.8864e-01,  4.7973e-03, -3.3695e-03,  4.2165e-03,
        -1.5772e-04,  2.1241e-03, -1.5314e-03,  7.1811e-04, -1.8255e-03,
        -9.6945e-05, -1.6420e-03,  1.2091e-04, -1.4593e-04,  7.6678e-04,
        -2.2095e-03,  1.0667e-03, -1.2414e-03, -2.3319e-03, -5.0211e-04,
         9.4509e-04, -1.5978e-03, -4.9727e-03, -1.3273e-03, -2.3947e-03,
         1.1235e-03, -3.8961e-03,  8.4496e-04, -1.4428e-03, -6.8451e-04,
         1.4337e-03,  1.0009e-03,  2.7666e-04,  1.0826e-03,  3.9923e-03,
         4.0441e-03, -2.2428e-03, -1.1662e-03, -4.9792e-04, -1.6410e-03,
        -3.7523e-03, -2.2603e-04, -1.7617e-04,  4.0882e-04, -7.9545e-04,
         5.1422e-03,  1.4131e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7564e-02,  9.1605e-01, -7.2269e-03, -9.4345e-03,  1.1936e-02,
         3.7192e-04,  5.6861e-04, -1.9491e-03, -5.3415e-03, -5.4712e-03,
        -4.1626e-03, -3.7095e-03,  8.5750e-04,  8.5762e-03,  6.7789e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.8910e-03,  8.9685e-01, -1.3871e-02, -1.7717e-03, -4.2850e-03,
        -5.7372e-03, -9.1311e-04, -5.9329e-03,  8.6257e-04, -2.9618e-03,
         1.9091e-04,  9.7908e-04, -3.3177e-04,  3.5590e-03, -1.9232e-03,
        -5.2821e-03,  4.8747e-04,  1.8781e-03,  7.7669e-04, -3.4990e-05,
         5.6177e-04, -3.4481e-03, -2.1678e-03,  1.4742e-03,  3.9643e-04,
        -1.9600e-03, -3.1139e-04,  6.7646e-03,  1.3135e-03,  4.4810e-04,
         2.0095e-03, -1.8058e-03, -3.6318e-04,  2.4442e-04,  1.2266e-03,
         1.1534e-03,  4.9402e-04, -8.4628e-04, -3.5410e-03,  5.0642e-04,
        -3.9234e-04, -1.0209e-03, -1.1368e-03, -4.4820e-04,  4.7358e-04,
        -1.7325e-03, -1.5588e-03,  5.6837e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-2.3022e-02,  9.1675e-01,  1.0943e-03, -1.4248e-03,  6.0486e-03,
        -2.6529e-03,  1.5531e-03, -1.8056e-03, -2.8942e-03, -1.1787e-02,
        -8.2068e-04,  1.4222e-03, -8.0293e-04,  6.6811e-04, -9.0797e-04,
        -1.4442e-03, -9.0316e-03, -3.4674e-03,  2.9465e-04,  1.0819e-03,
         4.1304e-04, -1.3061e-04, -2.9338e-03,  1.0340e-03, -1.1349e-04,
        -6.4008e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5077e-02,  9.3555e-01, -2.3748e-03, -2.5640e-03,  1.1093e-03,
        -2.1823e-03,  2.5044e-03,  6.2717e-03,  2.8143e-03, -1.7658e-03,
         7.9156e-03,  6.5781e-04, -4.6924e-04,  2.3113e-04,  1.0128e-03,
        -3.9334e-04,  5.6603e-03, -1.9522e-03, -1.7306e-03, -1.4990e-03,
        -6.2669e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4427e-02,  8.8277e-01,  1.1909e-02, -1.6355e-02, -4.5848e-03,
        -4.9663e-03,  1.1876e-03, -3.9415e-03,  6.8773e-04, -6.1084e-04,
        -3.9799e-04,  9.2080e-04, -2.0111e-03, -3.6882e-03,  2.7446e-03,
        -2.1845e-03,  5.8360e-03, -5.3405e-03,  1.8234e-03,  1.2067e-03,
        -1.8870e-03,  7.0613e-04, -1.2704e-03,  5.1038e-03,  5.2355e-03,
         1.0291e-04, -1.6481e-03,  7.1711e-06,  1.2053e-03,  7.9040e-04,
        -1.2667e-03, -1.4283e-03,  4.3804e-03, -8.6751e-04, -3.6524e-03,
         2.8545e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4579e-02,  9.2973e-01, -1.0057e-02,  1.4257e-03, -2.3238e-03,
         6.1480e-03, -5.0140e-03,  4.1821e-04, -2.7853e-03, -2.0656e-03,
         3.0197e-03, -3.8520e-04, -2.0792e-04, -1.3812e-03, -2.2018e-03,
         8.5292e-03,  4.1924e-03, -4.8863e-05,  2.5117e-03, -2.9738e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3861e-02,  9.3410e-01, -1.0730e-02,  3.8662e-03,  1.8634e-03,
         5.7248e-03, -1.2714e-03, -1.0864e-03, -9.8802e-04, -5.6379e-04,
         3.6664e-04, -2.0205e-03,  4.1625e-05,  3.0187e-03,  5.9461e-04,
         1.4535e-03, -1.1831e-03, -2.0970e-03, -5.2416e-03,  1.6209e-03,
         8.3051e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.1409e-03,  9.4814e-01, -3.2005e-03, -4.2861e-03, -5.7094e-03,
         3.0698e-03, -9.8443e-04, -2.1424e-03,  5.7976e-03, -7.8503e-04,
        -3.6702e-03, -1.3788e-03, -1.2693e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7037e-03,  9.2823e-01, -4.5513e-03, -3.7978e-03, -2.9518e-03,
        -1.4501e-03,  1.4898e-04, -1.5393e-03, -2.8893e-03, -7.8987e-04,
        -5.1299e-03, -3.1203e-03, -4.7424e-03, -2.5800e-03, -7.9355e-05,
        -1.5304e-04, -9.2610e-05,  9.2557e-04, -3.5637e-03, -2.4339e-03,
         3.3233e-04,  3.1561e-03,  4.2505e-04, -7.9512e-04, -1.2960e-03,
        -5.1184e-05,  1.3880e-03, -4.4486e-03,  7.1606e-04,  9.5773e-04,
         1.3295e-03,  1.0867e-03,  2.8564e-03, -7.7325e-04, -1.0120e-04,
         2.9804e-03, -4.4327e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0071e-02,  9.2193e-01,  2.4192e-02, -4.5627e-03,  1.7976e-03,
        -2.8689e-05, -3.8834e-03,  3.6013e-03,  1.3748e-03, -2.3700e-04,
        -6.7144e-04,  2.4248e-04, -8.3503e-05,  2.8730e-03,  5.5649e-03,
        -3.2335e-03,  3.0470e-03, -4.1008e-04,  2.9028e-03,  1.7585e-03,
         1.9445e-03, -6.3961e-04,  1.5392e-03,  3.4061e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0572e-02,  8.6402e-01,  1.7594e-02,  4.9492e-03,  4.8032e-03,
         1.0368e-02,  4.0997e-04,  2.3047e-03,  7.5183e-03, -2.4108e-03,
        -2.1758e-03,  2.4044e-03,  4.2455e-03,  2.3261e-03,  7.0674e-03,
        -1.8595e-03,  2.5895e-03,  4.1261e-03,  2.0608e-04, -2.9624e-04,
         8.1264e-04, -5.5353e-03,  2.7026e-03, -1.7594e-03,  1.5661e-03,
        -9.9009e-04, -5.0434e-04, -1.8397e-03, -1.2297e-03,  1.0273e-03,
        -5.2403e-03, -4.5482e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8356e-02,  9.2899e-01,  1.5694e-03, -7.9433e-03,  4.7372e-04,
        -5.6082e-03,  2.9637e-03,  1.4225e-03,  7.8579e-03, -2.5528e-04,
         1.8001e-03, -1.2064e-04, -5.7612e-04,  1.2116e-03,  5.2266e-05,
         5.5884e-04,  2.4953e-03,  3.0271e-05, -1.2603e-04,  3.3330e-03,
         9.2707e-05, -3.7851e-03, -1.4071e-03,  3.4527e-04,  6.6971e-04,
         2.8408e-03,  5.9166e-04, -1.2280e-04,  1.2190e-03,  1.0488e-03,
         2.1366e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.0678e-04,  9.2181e-01,  5.8550e-03,  1.4025e-03,  5.1008e-03,
         5.4989e-03,  4.9621e-03, -2.7003e-03,  8.6389e-03,  7.1390e-04,
         4.6457e-03,  2.7098e-03,  2.2482e-04, -1.8857e-03,  4.4669e-03,
        -6.7290e-04,  2.8330e-03,  2.4555e-03,  3.9315e-03, -2.9646e-04,
        -4.4048e-03, -9.7995e-03,  4.0819e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1019e-02,  9.2336e-01, -1.3747e-03, -4.7742e-03, -3.8692e-03,
        -4.7946e-04, -2.3035e-03, -3.2541e-03, -1.3350e-03, -3.7337e-03,
        -8.4268e-04, -4.7823e-03, -1.8646e-03, -4.1683e-03,  8.9138e-04,
         2.5427e-03, -2.1740e-03, -7.5295e-04, -2.5411e-04,  2.9970e-03,
        -4.2674e-03, -1.2884e-03,  8.2373e-04,  7.7376e-04, -2.6425e-04,
         1.7768e-03,  1.3254e-02,  7.8168e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-4.5055e-03,  9.4440e-01,  2.4173e-04,  1.6072e-03, -2.7133e-03,
        -6.6355e-04,  1.0121e-03, -1.0886e-02, -1.6786e-03,  1.7953e-05,
         1.7270e-03, -1.9266e-03,  6.2476e-04,  6.3466e-04,  1.0060e-03,
         2.1521e-03,  4.6190e-04, -1.2888e-04, -2.5003e-03, -3.0391e-03,
         3.5161e-04, -6.3991e-04, -1.3440e-03,  5.5205e-04,  3.8796e-05,
        -1.2976e-03,  2.2423e-03,  9.9101e-03, -1.6948e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0275e-03,  9.2674e-01,  1.2036e-02, -2.6667e-04,  1.3779e-02,
         1.7572e-03,  1.0337e-03, -3.1076e-03,  1.5170e-03, -2.6842e-03,
         4.0060e-04,  1.5005e-04, -1.5574e-03,  3.0463e-03,  6.0719e-03,
        -8.7256e-04,  3.9603e-03,  2.4908e-04, -1.8176e-03,  2.1291e-04,
         4.6820e-04, -4.0208e-03, -3.4903e-03, -6.7375e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.4983e-03,  9.2052e-01, -1.4602e-02, -3.4317e-04, -2.5276e-03,
         2.1401e-03, -2.0742e-03,  1.0020e-03, -7.6130e-03,  1.3502e-03,
        -1.5454e-03, -2.8535e-03,  8.1935e-03, -1.5223e-03,  7.8578e-04,
         2.8690e-03, -3.9245e-04,  1.2267e-03,  3.4772e-04,  2.3719e-03,
        -2.0115e-03, -1.6709e-04,  2.2580e-04,  2.6482e-04,  8.3040e-04,
        -2.3409e-03, -5.1991e-03,  1.5886e-03, -1.1083e-03,  5.7107e-05,
        -6.7967e-04,  6.5011e-04, -2.5577e-04, -2.8385e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.1639e-03,  9.2017e-01, -8.8288e-03, -1.8674e-02,  6.0876e-04,
         1.2693e-03, -8.9106e-04,  3.0603e-03, -9.8556e-05, -2.3655e-03,
        -5.2519e-03, -3.6191e-03, -5.3451e-04, -4.9135e-04, -5.1632e-04,
        -2.7075e-03, -4.2165e-03,  2.2970e-03, -2.6000e-03,  3.0017e-03,
         9.3104e-03,  1.3242e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8242e-02,  9.2266e-01, -6.9400e-03,  1.8253e-03,  1.0712e-04,
        -2.1937e-03,  6.7980e-04, -4.8602e-03,  1.6664e-03,  2.5784e-03,
         1.0414e-02, -1.2971e-03, -1.3928e-03,  3.8076e-04,  6.7379e-04,
        -4.3917e-03,  1.7629e-03,  7.9369e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1908e-02,  9.4556e-01,  2.8978e-03, -4.5143e-04,  1.9947e-03,
        -1.7614e-04,  2.8788e-04,  1.3643e-03,  3.5359e-03,  1.4604e-03,
        -5.9157e-04,  3.0738e-03, -1.2192e-03,  3.5367e-03, -7.9119e-04,
         2.6288e-03, -1.1697e-03, -1.2386e-05,  1.9003e-03, -2.7213e-03,
         3.6959e-03, -7.0380e-03, -1.9867e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.9981e-04,  9.5929e-01, -3.1498e-04,  8.1060e-04, -8.9803e-03,
        -1.0528e-03, -1.4900e-03, -2.4887e-03, -5.8857e-03,  8.4118e-04,
        -3.1523e-04, -2.6129e-03, -1.0650e-03, -4.0235e-03,  6.7617e-04,
         7.2004e-03, -1.9554e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2267e-02,  9.3625e-01,  3.1418e-03,  3.4702e-03, -1.0282e-04,
        -1.7828e-03, -2.0633e-03, -1.0072e-03,  5.4335e-03,  4.7340e-03,
        -2.4325e-03, -4.4857e-03, -2.3054e-03,  1.3770e-03, -3.4887e-03,
         3.0288e-03, -8.8412e-03, -3.7912e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0617e-03,  9.2328e-01,  5.1025e-03, -7.9600e-03, -6.7641e-04,
        -6.8877e-04, -1.3233e-03, -3.2479e-03, -6.4691e-04,  1.4661e-02,
         4.5647e-03, -4.9710e-03,  1.1574e-03, -1.8648e-03,  2.1165e-04,
        -5.1168e-03, -3.7393e-03, -1.0680e-03,  3.4988e-03,  3.9026e-04,
        -2.4562e-03, -5.1844e-03,  3.1494e-03, -1.1408e-04,  1.5087e-03,
         2.3576e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5858e-02,  9.1003e-01,  4.0337e-03, -2.9781e-03, -1.6709e-04,
        -1.2753e-03,  3.2984e-03,  5.1594e-03,  2.8795e-03, -3.7384e-03,
        -1.1902e-05, -2.8357e-03, -1.2109e-03, -9.1478e-04, -1.4370e-03,
         1.0860e-03, -1.2107e-04,  1.5599e-03,  3.7783e-04,  6.0565e-05,
         2.5436e-03, -4.5925e-03, -3.7653e-03, -4.7781e-03, -9.2992e-04,
         7.3484e-04,  2.3639e-03,  2.4976e-03, -1.7006e-03, -2.7190e-04,
         6.7920e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3310e-02,  9.0135e-01,  2.6836e-03, -1.0119e-04, -1.7959e-02,
        -7.0272e-03, -1.4315e-03,  4.4688e-04,  2.7011e-03, -1.0967e-02,
        -7.1190e-03,  4.3674e-05, -6.8710e-03, -4.1712e-03, -6.4126e-03,
        -4.8613e-04, -1.1461e-02,  5.4622e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0191,  0.9481, -0.0101, -0.0037,  0.0035, -0.0100,  0.0021,  0.0035,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([-8.3665e-03,  9.2171e-01,  6.1141e-03, -4.0411e-03,  1.0905e-03,
         2.4831e-03, -1.1201e-03, -2.9390e-03,  2.0064e-03,  5.6783e-03,
        -1.4780e-03, -1.1096e-03,  2.5037e-03,  1.5400e-03,  7.9523e-04,
         1.1785e-03,  1.6102e-03, -5.4431e-04,  1.0388e-03, -1.7485e-03,
        -3.3920e-05,  8.3805e-05, -2.4657e-03,  3.2210e-04,  6.8806e-04,
        -1.2972e-03,  1.6236e-03,  3.2295e-03,  4.9273e-03,  1.0058e-03,
        -2.6388e-03, -1.1442e-03,  2.4725e-03,  1.1553e-03,  3.1750e-03,
         6.3260e-04, -2.0124e-04,  1.6064e-03,  2.2059e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.5125e-03,  9.2050e-01, -1.1429e-02, -4.1334e-03, -4.7969e-03,
         1.9450e-03,  4.6816e-03, -8.7975e-04,  3.0840e-04,  7.4311e-03,
        -7.6816e-04,  2.6712e-03,  1.1208e-02,  3.4434e-04, -3.7429e-04,
         1.2612e-03, -2.2278e-03, -4.9263e-04,  1.4937e-03,  2.2812e-03,
         6.5133e-03,  1.0015e-03, -1.1981e-03, -1.1745e-04,  6.2643e-04,
         1.3133e-03, -9.0360e-04, -5.8547e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3973e-02,  9.2610e-01,  1.2642e-03,  1.1374e-02,  2.2150e-04,
        -5.0963e-03,  1.7785e-03,  1.6342e-03,  2.5221e-03,  5.8769e-03,
         6.0221e-03, -1.2871e-02, -1.2638e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.5197e-03,  9.5345e-01,  8.4159e-03,  2.9824e-03,  3.0084e-03,
         3.9752e-03, -2.7413e-03,  1.8660e-03,  1.6548e-04,  2.2067e-03,
         7.6670e-05, -2.0465e-03, -1.5727e-03,  3.2182e-03, -6.7595e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6258e-02,  8.8197e-01,  1.4580e-02, -2.0505e-03,  5.6831e-04,
         1.6032e-03,  6.2163e-03, -2.8518e-03,  3.7973e-03, -4.2047e-03,
        -1.6209e-03,  5.2058e-05, -2.5894e-03,  2.5613e-03,  2.3272e-03,
         1.9554e-03,  4.4901e-03,  8.4112e-04,  8.9436e-04,  1.8329e-03,
         2.1745e-03,  1.2261e-03,  1.8869e-03, -8.8083e-04,  2.5576e-03,
        -3.5333e-04,  2.6759e-03,  2.0925e-03, -1.2267e-03,  1.0115e-03,
         1.0643e-03,  5.3169e-04,  4.9610e-03,  4.2468e-03,  3.2355e-03,
         8.5208e-04, -3.1191e-05, -4.2634e-04, -2.3908e-04,  2.0081e-04,
         2.1360e-03, -1.7995e-03, -3.7550e-04,  8.3678e-04,  4.0251e-04,
         3.4800e-03, -5.8258e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7417e-04,  9.2113e-01,  8.9835e-03,  4.1385e-03, -2.9268e-03,
        -4.6022e-03,  2.7083e-03, -2.9790e-03,  1.0007e-03, -5.7940e-04,
        -9.9398e-04, -5.9417e-03, -8.2776e-04, -2.5923e-03, -1.4160e-03,
         1.0138e-03, -1.2599e-04,  5.7131e-03, -1.2603e-03,  1.9341e-04,
        -1.0559e-03, -8.4407e-04, -4.0907e-04, -1.4168e-03, -1.1983e-03,
        -1.0973e-03,  1.9133e-03, -1.2631e-03,  6.6551e-05,  7.8875e-04,
        -3.3609e-03, -1.1958e-03, -1.1028e-06, -5.0382e-04,  3.4982e-03,
        -1.4402e-03,  1.8975e-03, -7.2053e-04, -5.2288e-04, -8.2900e-04,
        -2.5477e-03,  4.1280e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3392e-02,  8.8620e-01,  5.4590e-03, -3.8400e-03, -1.7821e-03,
        -3.9846e-03,  1.7968e-02, -2.4979e-03, -3.2979e-03,  2.6876e-03,
         8.6501e-03,  7.0217e-03, -1.3565e-03,  1.3521e-03,  4.2879e-03,
        -2.5279e-03,  5.0241e-03, -1.3573e-03, -4.7938e-03, -6.2655e-04,
        -2.3510e-03, -1.1932e-03, -3.5057e-03, -1.3547e-03, -3.4866e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0394e-02,  8.9288e-01, -7.5413e-03,  2.8706e-04, -6.7147e-04,
        -5.3436e-03,  3.4319e-03, -1.0802e-03,  4.0841e-03,  1.5864e-03,
         1.9744e-03,  3.5319e-03,  1.6548e-03, -2.7487e-03,  3.4930e-03,
         1.4636e-03,  1.6746e-03,  8.0825e-03,  2.8281e-04, -1.4070e-03,
         1.0389e-03, -1.4145e-03, -3.9327e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2850e-03, -9.0308e-01, -1.1406e-02,  3.0287e-03, -1.3097e-03,
        -2.3857e-03, -4.8146e-04,  8.9431e-04,  1.0683e-02, -5.6820e-03,
        -1.2536e-03, -5.5889e-04, -3.8166e-03, -7.8282e-04,  7.7177e-03,
         3.2335e-03, -2.6781e-03, -4.8324e-03,  6.3596e-03,  4.2045e-03,
         4.7432e-03,  7.9839e-03,  8.5998e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9361e-03,  9.4425e-01,  9.7380e-05,  6.3234e-03, -2.5910e-03,
        -1.9822e-04,  2.5875e-03, -5.3873e-03,  5.3134e-03, -2.3909e-03,
        -2.2036e-04, -1.5471e-02,  1.2233e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0017e-04,  9.4410e-01, -1.3140e-02,  1.5422e-02,  1.4262e-03,
        -5.9530e-03,  2.9110e-03,  5.2415e-03, -2.3093e-04,  8.9727e-04,
         7.2411e-04,  5.5596e-03,  2.3716e-03, -1.9255e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7639e-02, -9.3544e-01, -7.3982e-03, -2.4110e-03, -1.0986e-03,
        -1.7269e-03,  2.6217e-03, -7.3013e-04, -1.2264e-03,  1.0642e-03,
        -3.2221e-03, -7.0267e-04,  4.5907e-03, -1.2715e-03, -1.7527e-03,
        -1.0039e-03,  6.9611e-03, -2.0527e-03, -7.0858e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 2.0170e-02,  9.2187e-01, -7.4413e-03, -2.8963e-03, -4.4985e-04,
        -2.4355e-03,  2.9974e-04, -8.8151e-04, -3.9037e-03, -2.4237e-03,
        -1.0698e-03,  1.5338e-03,  2.9836e-04,  5.4666e-03, -3.0258e-03,
        -1.9792e-03,  2.0176e-03, -9.8508e-05,  2.1622e-03,  1.1076e-02,
        -1.6215e-03,  7.7918e-04, -2.1292e-03, -5.7703e-04, -9.0036e-04,
        -2.4909e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.8079e-03,  9.1961e-01, -4.1252e-03, -7.9195e-03, -9.3227e-03,
        -7.9188e-03, -3.6510e-03, -7.3897e-03,  1.0092e-02, -6.4429e-04,
         1.1997e-03, -1.9301e-03, -2.6512e-03,  3.2100e-03, -1.7456e-03,
         9.7783e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9863e-02,  8.8691e-01, -3.4896e-03, -3.5643e-03,  2.3191e-03,
         1.7348e-02,  6.2899e-04,  5.0239e-05, -1.7352e-03,  2.0143e-03,
        -2.4277e-03, -6.3603e-03, -1.7797e-03,  8.5242e-04, -2.6612e-03,
         5.1503e-03,  1.2847e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6700e-02, -9.1520e-01, -2.7357e-02, -5.4651e-03,  1.0759e-02,
        -6.8335e-03,  2.2517e-04, -2.5153e-03, -4.9463e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.2211e-03,  8.6419e-01,  9.5456e-03, -1.3112e-02, -1.6045e-02,
        -5.5889e-03,  1.7527e-04,  2.6858e-03, -1.9715e-03, -2.0009e-03,
         1.3654e-03,  2.3583e-03,  2.4087e-03,  2.0931e-03, -1.1253e-03,
         9.5406e-04,  2.6315e-03, -1.6174e-03, -5.9270e-04,  1.7944e-03,
         4.5247e-03, -3.4493e-03,  2.8027e-04, -3.5882e-03, -3.4760e-03,
        -2.9233e-03, -4.5088e-03, -1.4579e-03, -6.4666e-04,  3.0046e-03,
         8.8947e-04,  4.1667e-04,  5.1932e-03, -1.1046e-05, -1.7450e-05,
        -1.1761e-03, -1.6969e-03,  2.4926e-03, -3.7440e-03,  4.8256e-04,
         1.2655e-03, -3.7739e-03, -2.2843e-03, -1.3555e-04,  2.8587e-03,
         7.2226e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8102e-03,  9.0895e-01,  1.6232e-03, -8.6011e-03, -7.9496e-05,
         7.5048e-03, -6.0071e-03,  2.7409e-03, -4.5378e-03,  8.6406e-04,
         7.1640e-04,  5.5832e-04, -1.1758e-02, -1.9179e-03, -1.0627e-03,
         2.8235e-03,  1.5268e-03, -1.9897e-03,  4.1351e-04,  3.3826e-03,
         1.6836e-03,  4.5343e-04, -3.4917e-03, -2.6858e-04, -7.0556e-04,
        -6.7478e-03, -4.0764e-03, -4.1883e-04,  3.5801e-03,  7.7086e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0266, -0.9479, -0.0055, -0.0040, -0.0019,  0.0066, -0.0075,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1260e-03, -9.6766e-01, -4.6947e-03, -6.4236e-03, -5.2062e-05,
        -1.0626e-03, -2.7109e-03, -2.1345e-03, -2.5983e-03,  5.6020e-03,
         2.9388e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1193e-02,  9.1948e-01,  5.8562e-03,  8.4050e-03,  4.2046e-03,
         6.7949e-03,  4.5462e-03, -1.6264e-03,  2.2515e-03,  1.8925e-03,
        -9.4021e-04, -5.5991e-04, -2.0112e-02, -1.2134e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4757e-03,  9.4071e-01, -4.6179e-04,  3.0099e-03,  2.6831e-04,
        -3.0621e-04,  7.3178e-04, -1.1611e-03, -4.8346e-04,  4.4864e-03,
         2.0971e-04,  3.5226e-04, -1.6016e-03, -1.2504e-03, -1.1422e-03,
         2.4590e-03,  9.0570e-04, -9.1428e-04,  4.5030e-04,  2.9265e-03,
         6.5278e-03,  1.3123e-04, -5.6147e-03, -1.2802e-03, -1.7949e-03,
        -7.5300e-04,  2.6907e-03,  1.4476e-05, -4.2514e-04,  3.5799e-03,
         6.6579e-03, -1.3966e-03, -1.6806e-03, -1.1434e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7315e-02,  9.1019e-01,  7.5395e-03,  9.2707e-03, -9.4987e-04,
         1.0565e-04,  2.0801e-04, -4.7428e-03, -5.3646e-03,  1.0305e-05,
        -3.1452e-03,  1.4221e-04,  1.8655e-03,  2.2065e-03,  1.2967e-02,
         3.1659e-04, -4.8602e-03,  1.1957e-03, -7.5578e-04, -3.1485e-03,
         1.3160e-03, -2.4249e-03, -3.1086e-03, -5.8188e-04, -2.7277e-04,
         1.4877e-03, -4.0792e-04, -4.1047e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0150e-03,  9.1717e-01,  1.9471e-02, -5.3559e-03,  6.8057e-03,
         5.9950e-03, -6.8639e-05,  1.0107e-03, -2.2097e-04,  1.0574e-03,
        -7.7954e-04, -6.0484e-04, -1.1611e-03,  3.7707e-03, -8.4098e-04,
        -2.5936e-03,  3.8996e-03, -1.1182e-02,  8.1414e-04, -1.3648e-03,
        -3.1430e-03,  6.2735e-04, -6.4579e-03,  3.5914e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 2.1223e-02,  9.2587e-01,  6.1976e-03, -2.6591e-03, -1.4762e-03,
        -3.0350e-04,  1.3492e-04, -3.6641e-03,  2.5104e-03,  4.4270e-04,
         1.7700e-03, -2.0418e-03,  8.2475e-04, -1.4612e-03,  4.0273e-04,
        -1.8140e-03,  7.2585e-04,  1.9493e-03,  1.9592e-03, -7.3379e-04,
        -4.8829e-03, -2.1457e-04, -1.5421e-03, -2.5886e-04,  1.2793e-04,
         2.2524e-04, -1.5127e-03, -1.4546e-03, -1.4174e-03, -6.8459e-03,
        -3.3553e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3749e-02,  9.5578e-01, -6.2341e-03,  4.4829e-05,  2.2052e-03,
         3.4267e-03,  1.3571e-03,  2.8713e-03,  4.7902e-03, -1.8374e-03,
         4.5654e-03,  3.1399e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4760e-02,  9.3575e-01, -3.9823e-03, -4.6074e-03,  2.8719e-03,
        -1.0314e-02, -2.2215e-03, -1.8729e-05, -3.2694e-03, -6.3988e-03,
        -1.9106e-03,  4.8023e-03, -6.6535e-04,  1.0009e-05, -8.4189e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0336e-03,  9.4920e-01, -1.8704e-03, -3.9955e-03,  1.1721e-03,
        -3.2157e-03,  5.6243e-03, -4.9070e-04, -5.1630e-04, -6.4029e-04,
         8.9221e-04,  1.5980e-03, -1.9573e-04, -1.1245e-03, -1.4430e-03,
        -1.5763e-03,  3.6203e-04, -1.0520e-03,  2.7337e-03,  1.6441e-03,
        -4.1494e-05,  7.5368e-04, -4.5113e-03, -1.1334e-03,  8.1530e-04,
         1.7852e-03, -7.5599e-04, -1.4358e-03, -2.4094e-03, -1.9802e-03,
        -8.2151e-04, -2.1659e-03,  1.3889e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.7299e-03,  9.0562e-01, -5.6116e-03, -1.4403e-02,  3.6833e-04,
        -2.8636e-03, -6.9164e-03, -4.8555e-03, -5.4591e-03, -5.3871e-03,
        -5.9805e-03, -3.1646e-03, -9.6771e-03,  5.6580e-03, -6.1045e-03,
         8.2022e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8403e-03,  9.3618e-01,  1.6583e-02,  3.5633e-03,  4.6128e-03,
        -7.4292e-04, -1.0881e-02, -8.3827e-03, -1.7212e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0480,  0.8854, -0.0166,  0.0112, -0.0388,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6612e-02,  8.6546e-01, -1.6227e-02, -5.6604e-03,  3.3182e-03,
        -1.0370e-02, -1.7112e-03, -1.7419e-03, -1.9345e-03,  1.4918e-03,
         1.6613e-03, -1.6780e-03, -6.4925e-03, -6.0806e-04,  1.3255e-03,
         2.1877e-03,  3.8802e-03, -3.0630e-04,  4.3582e-03,  1.0184e-03,
        -1.8897e-03,  3.4261e-03,  1.0675e-03,  2.6490e-03,  1.6559e-03,
         8.2184e-03, -3.0527e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0663,  0.8948, -0.0009,  0.0038,  0.0140,  0.0028,  0.0172,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3351e-02, -9.4310e-01, -2.3778e-03, -4.1612e-03, -2.1848e-03,
        -3.1469e-03, -4.8586e-04, -6.5443e-04,  1.5300e-04,  4.4158e-04,
        -3.6194e-04, -3.4264e-03,  6.3519e-04, -5.2851e-03,  3.3593e-03,
        -1.9839e-04,  4.6980e-04,  2.3745e-03, -6.1455e-03, -6.2921e-03,
        -6.3161e-04, -7.5880e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3949e-02,  9.1233e-01,  1.7299e-02,  6.9511e-04,  8.2307e-03,
         6.2306e-03,  1.0850e-03, -1.2365e-03, -2.8266e-03, -6.9698e-03,
        -6.3277e-03,  2.8551e-05, -4.2371e-03, -1.2474e-03, -1.2896e-03,
        -1.6838e-03, -1.2037e-03, -2.4661e-03, -6.5991e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0670e-02,  9.4476e-01,  4.8853e-03,  2.5562e-05,  4.8057e-04,
         4.4315e-03, -1.5361e-03, -3.3204e-03,  3.5033e-03, -1.4363e-03,
        -1.2066e-03,  1.9882e-03,  5.5201e-05,  1.1924e-03,  3.2426e-04,
         7.6281e-04,  1.3652e-03,  1.7018e-04,  8.0237e-04,  1.3811e-03,
         1.3902e-03,  7.2116e-04,  2.0800e-03,  1.3882e-03,  5.6042e-05,
         6.0142e-04, -1.2116e-03, -1.1924e-03,  1.7214e-03,  8.9804e-04,
         4.4411e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-1.1503e-02, -8.9303e-01, -3.8092e-02,  4.5906e-03,  4.2681e-03,
         5.1830e-03,  2.2161e-03, -7.6639e-04,  2.7499e-04, -3.2893e-03,
         5.0582e-04, -2.8338e-04,  1.9890e-03, -6.3300e-03, -1.6226e-04,
         2.4449e-04,  8.8139e-04, -2.1089e-03, -7.7571e-04,  9.2982e-04,
        -1.0465e-03,  1.2635e-03,  1.7054e-04, -2.9733e-03,  2.3757e-03,
         1.9687e-03, -1.0106e-04,  7.0084e-03, -1.4852e-03,  2.1946e-03,
         1.1205e-03, -8.6662e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.5521e-03,  9.3022e-01,  1.8082e-03,  4.7017e-04,  3.8604e-04,
         1.0294e-03, -1.5339e-03,  6.1388e-03,  1.4737e-02,  2.9939e-03,
        -7.4373e-03,  3.5387e-03,  7.1770e-04, -2.1523e-04,  3.8915e-03,
         1.6430e-03,  4.0758e-03,  2.8939e-04, -1.6123e-03,  3.0372e-03,
        -2.1969e-03, -3.4712e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0614e-02,  8.9997e-01,  6.7308e-03, -4.4460e-03,  1.9004e-03,
         3.9108e-03,  1.3378e-03,  1.7454e-04, -3.4368e-03,  1.9693e-03,
        -1.6185e-03, -4.7526e-03,  6.1863e-04,  4.5084e-05, -1.0415e-02,
        -1.3417e-03, -3.4379e-03,  1.0834e-02, -3.5056e-03, -3.4560e-03,
         1.6862e-03, -4.9978e-04, -3.1780e-03, -2.2509e-03,  1.8054e-04,
         3.8000e-03, -1.0323e-03, -7.4042e-03, -7.7747e-04,  4.6790e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.9642e-03,  9.4847e-01, -6.3271e-03, -4.8701e-03, -9.2873e-04,
         2.0978e-03, -3.9868e-04, -1.6046e-03, -5.8149e-03, -1.3293e-03,
        -4.9884e-03, -1.5374e-03, -7.9065e-04, -5.6893e-04, -1.9493e-03,
         1.1700e-03,  8.7141e-04,  7.3228e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9068e-03,  8.8476e-01,  3.2830e-02, -4.9014e-03,  1.9583e-03,
         4.5367e-04, -1.2488e-03, -1.2532e-03, -3.1731e-03,  2.4006e-03,
         2.3099e-03,  1.3548e-03,  3.4629e-03,  3.1299e-03, -3.5112e-03,
        -1.9021e-03,  3.3852e-03,  4.6813e-04, -8.6550e-03,  2.0614e-03,
         1.7235e-03, -2.4806e-03,  1.8498e-03,  2.3488e-03,  1.8839e-03,
         1.0840e-04,  7.6033e-05,  7.3040e-04,  2.7913e-03, -2.3133e-04,
         9.9553e-04,  1.3590e-03,  1.2916e-03, -3.0367e-03, -1.3237e-03,
        -2.8492e-03,  7.7984e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3696e-02,  8.9379e-01,  1.7207e-02,  8.3830e-03,  3.3138e-03,
         8.5304e-05,  1.5557e-03, -2.8346e-03, -1.4519e-03, -2.8201e-03,
        -9.7537e-05,  1.3198e-03, -5.6786e-03, -2.6263e-03,  1.6517e-03,
         9.7181e-04, -4.7459e-04, -6.1277e-03,  1.9873e-03, -2.0686e-05,
         3.9318e-04, -2.0518e-03,  4.3282e-03, -7.1312e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0801e-03,  9.2156e-01,  4.2084e-03, -3.6239e-03,  9.0221e-04,
         1.8355e-03, -2.8032e-03,  4.2825e-03,  3.2761e-03, -1.3460e-03,
         2.1159e-03, -1.0520e-03,  3.0607e-03, -2.5478e-04, -1.0383e-03,
         2.5825e-03,  1.2784e-03, -2.7489e-04, -6.9327e-06,  1.9606e-04,
         1.1225e-04, -2.9371e-03,  3.6748e-04, -1.7352e-04,  2.4501e-03,
         2.7259e-03,  1.0525e-03, -1.9522e-04,  1.4699e-03, -4.9941e-04,
         2.9727e-03,  2.5000e-03,  5.8190e-04,  2.5763e-03,  2.0075e-03,
         6.1205e-03,  3.2146e-03,  1.0836e-03,  9.3005e-04, -3.0999e-04,
         6.8635e-04, -1.0203e-03,  1.4934e-03,  1.3220e-03,  6.5386e-04,
         3.7622e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4466e-03,  8.8942e-01, -2.5524e-02, -1.8672e-03, -1.5663e-02,
         2.0221e-03, -2.2308e-03, -4.5009e-03, -7.5989e-04, -6.7782e-03,
         3.0002e-05, -1.3608e-03, -1.0444e-03, -2.8609e-03, -6.4083e-04,
        -2.5786e-03,  8.3318e-03,  1.4119e-02, -1.9360e-03,  1.5008e-03,
         2.1547e-03,  5.6784e-03,  4.5522e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0211,  0.9262, -0.0162,  0.0082, -0.0054,  0.0014, -0.0017,  0.0085,
         0.0033, -0.0056,  0.0023,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6174e-02,  9.2467e-01, -1.1890e-02,  6.1783e-04,  3.9301e-03,
        -5.8638e-03,  1.3703e-03,  3.1416e-03,  7.8879e-03,  1.5395e-03,
         2.6239e-03, -1.7748e-03, -5.6195e-03, -1.2901e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.0641e-03,  9.4965e-01,  5.3044e-03, -1.1798e-02, -1.8729e-03,
        -2.4585e-04, -1.1547e-02,  5.3894e-03, -1.3482e-03,  9.7605e-05,
        -1.3132e-03,  4.3648e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1435e-03,  9.2024e-01,  5.0328e-03,  6.8514e-03, -3.1739e-04,
         6.0077e-05, -5.6774e-03,  9.8812e-04,  4.4317e-04,  4.2541e-03,
         1.5218e-03, -7.3216e-04,  5.5599e-05,  1.4978e-03,  7.9812e-04,
         1.3653e-03,  1.1158e-04,  7.6125e-04,  1.2425e-03, -3.8843e-03,
        -3.4375e-04, -1.5854e-04,  4.1436e-03, -1.0565e-03,  4.5200e-04,
         1.6804e-03,  8.6231e-03,  4.5396e-05,  6.3956e-04,  1.2768e-03,
        -2.9840e-05,  6.6577e-03,  3.1520e-04, -1.5632e-04,  3.3721e-03,
         4.2385e-05, -1.6704e-05, -6.8048e-04,  8.6534e-04, -6.0073e-04,
         2.2452e-04,  9.6406e-04, -5.4647e-05, -1.0547e-04,  2.1300e-03,
         1.4828e-03,  2.0493e-03, -3.9857e-04, -6.1254e-04, -1.5043e-04,
         1.1265e-03, -4.1984e-04, -1.5079e-03,  3.2526e-04,  3.1481e-04],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-3.3089e-03, -9.3091e-01, -6.7487e-03,  5.1252e-03, -5.1138e-03,
         7.2883e-05, -1.3911e-03, -9.0551e-04,  4.6863e-05,  2.5807e-03,
         9.0300e-03,  1.7943e-03,  4.2813e-05,  4.6507e-03,  1.3458e-03,
        -1.1142e-03, -2.8841e-03,  7.0075e-05, -2.7779e-03, -6.6374e-04,
         2.6090e-03, -1.9780e-03,  1.4837e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0450e-02,  9.4801e-01, -5.0016e-03, -2.2814e-04,  3.8540e-03,
        -2.0559e-03, -3.5678e-03,  1.8003e-03,  2.4058e-03, -1.9743e-03,
        -5.2953e-03, -6.7638e-04, -1.3518e-03, -1.3937e-03, -1.5032e-03,
        -5.9638e-04, -4.8359e-03,  5.0014e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9119e-03,  8.8964e-01, -2.8450e-02, -1.5399e-02,  8.3592e-03,
        -4.2428e-03, -1.1693e-02, -3.0586e-03, -4.2271e-03,  1.1187e-03,
        -8.8289e-04, -6.5420e-03, -3.1654e-03, -1.9815e-03,  2.7785e-03,
        -1.9423e-03, -6.0449e-04, -1.0005e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5191e-03,  9.4939e-01, -3.6298e-03,  7.8889e-03,  8.1960e-04,
        -2.0057e-03,  2.4838e-03,  9.2141e-03,  8.8094e-04,  3.1576e-04,
         8.6382e-04,  4.9777e-03,  2.2472e-03,  2.2262e-03, -9.5359e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0220e-02,  9.0062e-01, -5.8418e-04, -4.0724e-03, -4.6120e-03,
        -4.0401e-03,  3.9437e-04,  1.4910e-04, -2.5475e-03, -6.0046e-04,
        -3.3349e-03, -6.3592e-04,  2.8966e-03, -5.9659e-04, -2.4082e-03,
         9.1042e-04,  3.6147e-04,  8.0010e-04,  6.5369e-04,  8.5345e-03,
         3.6930e-03,  3.1777e-03,  2.3188e-03, -7.5897e-04, -8.3239e-03,
         6.8868e-04,  4.8452e-03,  2.6295e-04,  1.4252e-03,  5.8127e-03,
         5.9227e-04,  9.1284e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7368e-02,  8.9070e-01, -2.1082e-03,  2.3999e-03, -4.5171e-03,
        -7.0402e-05, -2.1273e-03, -1.0088e-03, -2.2134e-03, -5.3604e-04,
         3.2999e-03, -3.7025e-03,  2.1169e-03,  3.4841e-03,  4.3850e-04,
         1.0596e-03,  7.1911e-04, -3.2680e-03, -4.1615e-03,  2.7118e-03,
         5.5345e-04,  4.6558e-03, -3.5164e-03, -2.5242e-03,  1.1784e-03,
         5.1054e-04,  2.9668e-03, -4.2500e-03, -4.8305e-04,  1.2464e-03,
         1.4260e-03,  2.0381e-03,  6.6411e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.0439e-04,  9.1341e-01,  1.0087e-02,  1.6964e-03, -6.1901e-04,
         1.7041e-03,  1.1958e-04,  1.2580e-04,  2.8978e-03,  4.1020e-03,
         1.4967e-04, -3.3077e-03, -1.9207e-03, -5.0684e-03, -2.0341e-03,
         1.2263e-03,  2.5659e-03,  1.4341e-03, -1.6353e-03, -6.7560e-04,
        -1.0943e-03, -3.7220e-04, -7.2176e-04, -5.1445e-03,  1.3185e-03,
        -6.2393e-05,  1.6559e-03, -2.1494e-04, -3.6906e-05, -5.0262e-05,
         1.9377e-03,  3.5814e-04, -1.3228e-03, -4.0059e-04,  5.0312e-04,
         1.2976e-03, -1.8625e-03,  2.0507e-04,  8.5958e-04,  1.4004e-03,
         4.0752e-04, -1.3517e-03,  1.1434e-03, -8.5274e-04,  1.0455e-03,
        -2.1221e-03, -8.6063e-04, -5.1524e-04, -9.3002e-04,  1.6062e-03,
         1.2127e-03, -1.8967e-03, -1.2306e-03,  1.0810e-04,  3.1270e-04,
         2.6868e-04, -5.7241e-03,  2.3040e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3345e-02,  9.1166e-01,  1.7024e-04, -5.0450e-03, -3.1430e-03,
        -4.0952e-03,  5.6610e-05,  2.8194e-03,  5.3277e-03, -8.1077e-03,
         2.0997e-03,  4.0136e-03,  9.3204e-04,  6.6718e-04,  4.8473e-03,
        -5.0954e-03,  3.0072e-03,  6.0896e-04,  1.1161e-03,  2.5333e-03,
         7.3682e-05, -6.0199e-04,  1.9327e-03, -3.8412e-03,  6.9359e-03,
        -7.9194e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9743e-02,  8.9416e-01, -1.5981e-02,  4.1471e-05, -1.6718e-02,
         4.5968e-03, -1.8083e-03, -1.2091e-03, -8.7078e-04, -2.7806e-03,
         5.3937e-03, -3.3349e-04, -6.5915e-05, -1.9330e-03, -3.2597e-03,
        -1.5227e-03, -1.3243e-03,  2.6996e-03, -2.5810e-04, -3.9462e-03,
         1.9987e-05, -5.4308e-03, -1.4251e-03,  9.5759e-04, -3.5187e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2086e-03,  9.0045e-01,  2.1881e-02,  4.6506e-03,  1.9851e-03,
         1.6842e-03,  2.4392e-04,  5.4766e-05,  3.2606e-03,  8.4730e-04,
         2.4925e-03,  2.7453e-04,  3.7826e-03,  5.6898e-03,  1.7403e-03,
        -6.0879e-03, -1.4211e-03,  3.6975e-04,  1.9387e-04, -9.0192e-04,
        -3.4529e-04,  1.6945e-03,  1.0185e-03, -8.7856e-05,  1.5158e-03,
         5.9420e-03,  3.0307e-03,  1.3260e-03,  5.7008e-04,  7.6698e-04,
         1.5629e-03, -3.7003e-04, -3.2646e-03,  3.8610e-05,  2.6670e-03,
        -1.0686e-03,  8.4439e-05,  4.2399e-04,  2.9953e-03,  2.8003e-03,
        -1.8569e-03, -3.5801e-04,  2.6570e-03,  2.2138e-03,  1.1250e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5354e-03,  9.4085e-01,  5.2508e-04,  7.7014e-03, -4.0338e-03,
        -4.8152e-04, -5.6587e-03, -2.6249e-03, -8.4831e-04, -1.8025e-04,
        -2.5211e-03, -1.8988e-03, -5.8933e-04, -1.2886e-03, -1.9127e-03,
        -3.0431e-03,  8.3140e-03, -3.8319e-03,  1.0286e-03, -3.0063e-04,
        -4.8436e-04,  7.8978e-03,  1.4450e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1898e-03,  9.3563e-01,  1.3179e-02,  1.9159e-02, -8.7755e-04,
        -8.9439e-04,  1.7420e-04,  5.1260e-03, -2.6782e-03, -8.4810e-04,
         7.8153e-03,  1.1426e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 1.2673e-02,  9.2245e-01,  1.5345e-02, -9.4801e-03,  2.6783e-03,
        -2.6593e-03, -1.9155e-03,  1.4134e-03,  5.5339e-03, -2.9927e-03,
        -4.4971e-03, -1.9203e-03, -7.0111e-04,  1.2205e-03, -6.0579e-04,
         1.4355e-03,  2.0439e-03, -3.6284e-04,  3.1363e-03,  1.4796e-04,
         1.4460e-03,  2.0097e-03,  3.3273e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5584e-03,  9.4395e-01, -4.7262e-03, -7.3659e-04, -1.0611e-03,
        -3.2130e-03,  2.3405e-03,  1.5232e-03, -5.0234e-04,  4.1576e-03,
         3.0563e-04, -3.8375e-03, -2.7106e-03,  1.9445e-03, -1.1192e-03,
         4.6228e-04, -1.3866e-03, -1.9159e-03, -1.9372e-04,  1.6557e-03,
         2.7847e-03, -1.9718e-04, -4.8089e-03, -2.9325e-04,  9.7634e-04,
        -2.5591e-03, -2.6547e-03, -1.7044e-04, -3.7657e-04, -8.8635e-04,
         2.0146e-03, -8.6350e-05,  2.8877e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7145e-04,  9.5784e-01,  2.7492e-03, -9.1059e-03,  7.3213e-03,
        -6.2981e-05, -3.1386e-03,  5.8991e-04,  1.4807e-04, -1.8945e-04,
         1.7972e-03,  1.5125e-03, -1.4602e-03, -1.2948e-03,  2.7404e-03,
        -4.6357e-03,  2.0709e-03,  2.9493e-03, -2.2113e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9607e-04,  9.1718e-01,  8.5204e-03, -6.5377e-03, -7.9761e-03,
         8.3368e-03,  6.0082e-03,  1.1476e-03,  5.6393e-03,  3.2678e-03,
        -1.2623e-02,  5.4125e-03,  3.4280e-03, -1.3731e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.9714e-04,  9.4803e-01,  1.2826e-02,  2.4192e-03,  8.4991e-04,
         1.3368e-03,  3.2656e-03, -6.2922e-03,  5.7035e-04, -2.1579e-03,
         2.3346e-03, -6.2372e-05,  2.5237e-04,  1.2853e-03, -1.7123e-03,
         9.1214e-04,  2.2528e-03,  2.8844e-03, -2.8711e-03,  2.9277e-03,
         3.9206e-03, -3.3880e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1011e-02,  9.2459e-01,  5.4872e-04, -4.8464e-03, -3.0991e-03,
        -3.8226e-03, -4.5862e-03, -8.3470e-03, -3.3956e-03,  2.2245e-03,
        -4.2007e-03,  2.7090e-04,  1.7954e-03,  9.1458e-04, -2.5911e-03,
        -5.6717e-03, -6.6511e-04,  1.6395e-03, -5.5875e-04,  2.3238e-03,
        -2.6504e-03,  6.6000e-05, -9.3366e-04, -2.0498e-03, -9.7890e-04,
        -9.2386e-04, -4.4630e-04, -9.1258e-04,  2.8933e-03, -1.0467e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.2266e-03,  9.1732e-01, -1.2861e-02, -1.0893e-02, -4.9553e-04,
         1.1613e-02,  3.7576e-03, -7.2857e-03,  1.0886e-03, -3.6751e-03,
        -7.8782e-04,  1.6534e-03,  6.6278e-04,  4.9440e-04,  4.6179e-03,
         7.8013e-04, -1.3786e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0204,  0.9105,  0.0257,  0.0157, -0.0095,  0.0026, -0.0058, -0.0012,
        -0.0019,  0.0010, -0.0057,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7904e-03,  9.3953e-01, -4.0471e-03, -6.4725e-03, -2.1934e-03,
        -3.4622e-03, -2.8277e-03, -2.9843e-03, -5.6202e-03, -2.3734e-03,
        -3.0715e-04, -1.4684e-03, -7.2312e-03, -5.6833e-04, -7.0102e-03,
        -5.5052e-03,  6.6086e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5896e-03,  9.3960e-01,  2.3893e-03,  9.9779e-03, -5.6859e-03,
         5.0948e-03, -3.5450e-04, -1.9802e-03, -1.9648e-04, -2.4601e-04,
        -3.2595e-03,  6.3548e-04, -1.5436e-04,  2.5295e-03,  3.2132e-03,
        -1.9521e-03, -7.1979e-04, -1.7321e-03,  1.5315e-04, -2.6774e-04,
        -1.8006e-03,  1.9634e-03,  3.5148e-03, -1.6844e-03, -3.0252e-03,
         4.2791e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.2891e-03,  9.2839e-01,  3.2670e-03, -3.9981e-03, -8.7975e-03,
         1.1165e-03, -7.3961e-04,  5.1759e-03,  1.0092e-03, -3.8217e-04,
        -6.4272e-04,  8.7853e-04, -3.1829e-03, -5.7709e-05,  3.4429e-03,
         1.9194e-03, -2.8331e-03, -1.2022e-03, -1.4155e-03, -1.2253e-03,
        -1.6660e-03, -7.1955e-04, -1.0207e-03,  3.1093e-04,  5.6569e-03,
        -1.3286e-03,  5.5564e-04, -3.3059e-03, -9.2530e-03, -2.2037e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.9957e-02,  9.3069e-01,  3.8755e-04, -6.8762e-03, -8.3034e-03,
        -1.8802e-03,  4.3078e-03,  3.4367e-03, -1.1414e-03, -2.1056e-03,
        -2.3163e-04, -4.4885e-04, -3.1836e-03,  7.9047e-04, -6.8468e-04,
         8.9518e-04, -2.1546e-03, -1.1625e-03,  1.3639e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 7.6995e-03,  9.2073e-01,  5.5180e-03,  1.0437e-03,  4.1311e-03,
         9.9990e-05,  1.0719e-02,  3.8081e-03,  3.9923e-05,  3.9604e-03,
         2.2662e-03, -9.8924e-04,  1.2151e-03,  1.1934e-03,  4.3346e-04,
        -9.3156e-04,  1.7165e-03,  6.1880e-03, -6.6709e-05,  7.5989e-04,
         1.5013e-03,  3.0693e-03,  1.6405e-03, -7.4112e-03,  2.7310e-03,
        -3.6337e-03,  6.5067e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6063e-02,  9.2933e-01,  2.1143e-03, -1.3235e-02, -2.0025e-03,
        -1.2300e-03, -4.6311e-04,  4.8205e-04,  3.4894e-03,  8.0807e-03,
         8.1357e-03,  6.8244e-03,  3.2702e-03, -5.2751e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9943e-04,  9.3381e-01,  1.4786e-02, -1.9834e-04,  3.7005e-03,
        -2.0561e-03,  2.9577e-03,  2.4333e-04, -3.6932e-03, -1.3925e-03,
         2.5408e-04,  3.9772e-03, -4.7184e-03, -2.3744e-03, -1.6559e-03,
        -4.5346e-03,  1.8430e-03, -6.1834e-04, -1.0371e-03,  4.2821e-03,
        -3.5865e-03, -1.2325e-03,  1.7534e-03, -6.2630e-04, -1.4831e-03,
         2.8824e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2175e-03,  9.1643e-01, -1.5880e-02, -1.5259e-02, -3.7656e-03,
        -5.2546e-03,  9.6979e-03,  8.8603e-04, -5.9336e-03,  6.0459e-03,
         1.5594e-03, -4.9611e-03,  1.4194e-03, -1.0685e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5604e-02,  9.4628e-01, -1.2855e-04, -1.5481e-03, -8.6587e-04,
        -1.7592e-03,  3.6672e-03,  1.4579e-04,  8.7442e-04,  1.4227e-04,
        -4.0982e-04, -3.4089e-03,  2.5705e-03, -3.2551e-04,  1.4108e-05,
         8.7600e-04,  9.3704e-04, -2.2039e-03, -5.8480e-05,  5.8178e-04,
        -3.6304e-05,  8.6021e-04,  8.1006e-04, -2.7625e-03, -2.0607e-03,
         3.6151e-04, -2.5682e-03,  1.6671e-03, -1.2700e-03, -3.2307e-04,
        -2.1046e-03,  2.7747e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.8546e-03,  9.3244e-01,  5.4000e-03, -6.8926e-03,  3.0682e-03,
        -3.4026e-03,  2.9655e-03, -5.4272e-03,  1.1300e-03,  2.4906e-03,
         4.7862e-03,  2.8584e-03,  2.6693e-03,  3.5317e-04,  6.8271e-04,
        -7.8920e-04,  7.7398e-04,  1.6034e-03, -9.9707e-04,  2.9622e-03,
        -2.3711e-03,  2.1932e-03,  4.3188e-04, -7.4851e-03,  9.6907e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0189,  0.9405,  0.0143, -0.0098,  0.0023,  0.0142,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7829e-02,  9.0317e-01,  1.3955e-02, -2.6510e-04,  5.5913e-03,
         1.7335e-03,  4.6676e-04, -1.5990e-03,  1.2635e-03,  5.2717e-03,
         6.7735e-04,  5.6677e-03, -1.7115e-04,  6.7277e-03,  4.1165e-04,
        -6.4689e-04,  1.1047e-03,  2.2208e-03,  1.7970e-03,  3.9379e-03,
        -3.2097e-04, -6.0809e-04,  1.1881e-03,  1.6330e-03,  1.2277e-03,
         4.2595e-03, -2.7855e-04, -7.0545e-04, -6.8232e-04,  6.4822e-03,
         1.3764e-03, -1.5230e-03, -2.4631e-03,  4.1716e-04,  5.9264e-04,
         3.2766e-04,  1.4033e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.0727e-03, -9.3497e-01,  7.2940e-03,  5.5214e-05,  3.0490e-03,
        -2.7146e-03,  1.2504e-03,  7.7820e-03, -4.4944e-04, -2.8904e-03,
        -4.7012e-04, -7.3472e-04, -1.3410e-03,  1.9389e-03, -3.1646e-03,
         1.5567e-03,  6.6884e-04,  1.1256e-03, -1.9177e-04, -3.3356e-03,
         3.0331e-03,  5.8028e-05, -2.0389e-03, -2.9643e-04, -3.0327e-04,
         8.1256e-04,  2.1338e-03, -3.7744e-03,  2.5740e-04, -1.6320e-03,
         5.1475e-04,  1.3274e-03, -4.7697e-04,  9.7441e-05,  1.1888e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9700e-02, -9.2404e-01,  7.2874e-03,  7.7065e-03,  9.1649e-04,
         8.1996e-03,  1.1229e-03,  5.5544e-04,  2.5737e-03, -7.1395e-03,
        -4.3220e-03,  7.7038e-04, -4.5913e-03, -7.6204e-03,  3.4498e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.8873e-03, -9.3577e-01,  6.4408e-03, -2.2219e-03,  1.0238e-03,
        -4.0103e-05,  7.1850e-04, -7.8336e-03, -1.4251e-04, -5.5539e-03,
        -1.6923e-04,  2.1873e-04,  3.7045e-04,  4.7096e-04, -1.7146e-03,
        -2.3046e-03, -9.0252e-04, -1.4557e-03, -4.7280e-03, -1.4138e-03,
         1.7225e-03, -3.0889e-03,  1.7420e-04,  2.8503e-04, -3.9196e-03,
         1.6227e-04, -8.7228e-04,  5.7195e-04, -5.8111e-04, -2.0345e-04,
        -7.5064e-04,  2.8169e-03, -1.4731e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7788e-02, -9.0880e-01,  1.9013e-03, -3.1320e-05,  2.3791e-03,
         5.5462e-04,  6.9048e-04, -1.1641e-03,  3.5371e-03,  1.6913e-03,
        -6.9949e-04,  3.9222e-03,  1.3329e-04,  3.0270e-03, -3.3160e-03,
        -3.8873e-03, -5.7684e-04,  1.6342e-04, -6.2432e-04,  2.3147e-03,
         6.2336e-03,  1.9908e-03, -2.3121e-03,  1.9701e-03, -1.6414e-03,
        -1.9695e-03, -5.5541e-03, -7.4080e-04,  1.0381e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([-5.1565e-03,  9.3531e-01,  8.1000e-03, -2.6936e-03, -4.5907e-03,
        -1.1839e-03,  1.3386e-03,  7.5624e-03, -3.5381e-03, -1.5266e-02,
         1.3345e-03,  4.7901e-06,  2.5131e-03,  1.9268e-03,  9.4794e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8459e-03,  9.3530e-01, -5.4076e-03, -1.2662e-03,  5.2024e-04,
         1.4499e-03,  3.0736e-03,  4.1358e-03,  2.6384e-03,  2.7468e-03,
         8.0156e-04,  3.6766e-04, -6.8789e-04, -6.2103e-06,  9.4048e-04,
         1.1873e-03,  2.6245e-04,  3.6942e-04,  3.3873e-04,  8.7360e-05,
        -1.2571e-03, -1.6114e-03, -4.4889e-04,  5.2092e-04,  9.6142e-04,
         1.3236e-03, -1.3786e-03,  2.5656e-03, -2.6535e-03, -4.5558e-04,
        -6.2637e-04, -2.0693e-03, -1.5062e-03,  3.0207e-04, -9.6955e-04,
         1.1257e-03,  1.1730e-03,  8.3990e-04,  1.1480e-03, -2.0871e-03,
        -1.1099e-03,  3.1168e-03,  9.7309e-04, -1.4135e-04, -2.6662e-04,
        -1.6622e-03, -1.3873e-04,  1.5308e-03, -2.6028e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1182e-02, -9.0108e-01,  2.3900e-03,  4.5320e-03, -4.2752e-03,
        -6.4053e-03, -3.7269e-03,  7.3222e-03,  1.3475e-03, -4.3813e-04,
         3.1781e-04,  1.2270e-03,  7.3579e-03,  3.4528e-04,  1.2504e-03,
        -2.7983e-03, -5.8648e-03,  2.2936e-03, -1.3628e-04, -4.9375e-03,
         1.5995e-03,  3.8025e-03, -1.4331e-04,  5.2173e-04,  1.5478e-03,
        -1.0703e-03, -7.1622e-04,  2.7761e-03,  1.4558e-04,  1.8011e-03,
        -2.0357e-03,  1.6596e-05,  3.3276e-04, -1.7080e-03,  2.5584e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.7076e-03,  9.3900e-01, -2.9806e-03,  2.3261e-03,  6.1137e-03,
        -3.0378e-04,  2.9058e-03,  5.5835e-03, -2.1834e-03, -1.2090e-03,
        -3.1433e-03,  4.8149e-04, -2.3871e-03, -2.2764e-03,  6.6854e-03,
        -8.6296e-03,  4.3721e-04, -2.1169e-03,  1.5329e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2291e-03,  9.3116e-01, -6.2313e-03, -3.7267e-03, -5.7875e-04,
        -1.1822e-03, -7.0299e-03, -4.9798e-03,  5.7379e-03, -2.9715e-03,
        -9.6332e-03, -1.3726e-03, -4.0995e-03, -1.4037e-02, -4.0294e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2989e-03,  9.5493e-01, -2.9459e-03,  2.8759e-03, -9.2478e-04,
        -3.7825e-03,  9.7217e-04, -7.6174e-04, -2.3127e-03,  1.2832e-03,
         4.3349e-03,  1.5081e-03, -4.8982e-05,  3.1300e-05,  2.7019e-04,
        -1.0769e-03,  1.7062e-03,  1.0093e-03, -9.4139e-04,  9.3120e-03,
        -3.0541e-03,  1.6236e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.9160e-03,  9.4369e-01, -1.8518e-03,  7.4456e-04, -2.0129e-03,
        -1.9832e-03, -2.3219e-03, -1.8957e-03, -2.5354e-03, -5.2698e-03,
        -3.1469e-03,  1.0059e-03, -6.1711e-03, -1.1174e-03,  7.1403e-04,
         5.1559e-04, -7.5375e-04, -1.7191e-05, -3.2866e-03, -4.1674e-04,
        -1.7667e-04, -2.2026e-03, -3.0583e-03,  1.7930e-03,  1.2037e-03,
        -3.8410e-04, -1.5486e-03, -2.1829e-03,  2.2354e-03,  8.4977e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6815e-03,  9.4254e-01, -6.3821e-03, -2.2786e-02,  5.4830e-04,
        -6.3056e-03,  3.7398e-04, -3.1330e-03,  4.7131e-03,  3.9893e-03,
         4.5488e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.3081e-03,  9.3350e-01,  3.4315e-03,  7.8095e-04, -5.7891e-03,
         1.4575e-03,  1.3848e-03, -4.2822e-03, -1.5548e-03, -2.8210e-03,
         2.7941e-03,  5.6479e-03, -4.9992e-03, -8.5352e-04,  4.6144e-04,
        -1.3122e-03, -1.6805e-03, -1.3224e-03, -2.9082e-03, -2.4674e-03,
         4.7041e-03,  1.0538e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0680e-02,  8.9255e-01,  1.3369e-02, -1.5304e-03,  3.7802e-03,
         1.1463e-02,  6.8476e-04, -7.2679e-04, -1.6819e-03, -1.8487e-03,
         5.1990e-03,  4.1278e-03,  9.4550e-04, -2.5672e-03,  4.1664e-03,
         3.3361e-04,  1.1227e-03,  1.9420e-03, -1.7143e-03,  2.3730e-03,
         4.9308e-04, -7.7253e-04, -3.0427e-03, -9.5384e-05,  4.0697e-03,
         1.8287e-03,  3.2586e-03,  7.7341e-05, -4.1450e-03, -1.4163e-03,
         3.0370e-03,  6.9613e-04,  4.2595e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.3471e-03,  9.1079e-01,  1.0117e-02,  7.3934e-03, -1.7796e-04,
        -2.9281e-03,  4.4582e-03, -9.3797e-03,  1.2264e-03, -1.3669e-03,
        -1.6279e-03,  1.7532e-03,  2.7860e-03,  1.6679e-03, -1.7690e-03,
         7.1575e-04,  1.9839e-03, -8.3363e-04,  1.8816e-03,  1.6080e-03,
         4.2534e-03, -1.3726e-03, -8.1194e-04, -1.5394e-03, -4.7437e-03,
         7.6939e-03,  6.6523e-03, -2.1234e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8522e-02,  9.2354e-01,  2.2784e-03,  3.8864e-03,  2.0162e-04,
         4.1166e-04,  2.0420e-03,  7.2611e-03,  1.9929e-03,  8.5543e-03,
        -9.6293e-04, -2.0820e-03, -8.6886e-04, -7.4005e-04, -1.7264e-03,
         5.0978e-04,  2.5744e-03, -5.6805e-03, -9.2265e-04, -3.8816e-03,
        -8.5938e-04,  2.3165e-03, -9.8504e-04, -2.8274e-03,  4.3717e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 7.5080e-04,  8.9828e-01,  1.1964e-02,  1.3345e-02, -2.5993e-03,
         5.3242e-05,  2.2680e-03,  1.4590e-03,  9.9719e-04,  2.9806e-03,
         4.3366e-03,  6.2462e-03, -1.4272e-03,  3.5571e-03, -1.7529e-04,
        -3.6671e-03,  2.7105e-05,  1.0264e-03,  9.8237e-04, -2.4305e-03,
         3.0782e-03, -3.2143e-04,  2.7246e-03,  5.0684e-04,  2.5146e-03,
         9.9741e-06,  9.0277e-04,  1.0337e-04, -1.5842e-03, -2.5370e-03,
        -1.1606e-02, -5.5448e-04,  1.8774e-03,  2.1255e-04,  1.9553e-03,
        -1.8696e-03, -1.8649e-04,  2.6024e-03,  2.6415e-03, -1.4584e-03,
        -2.1781e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4659e-02,  9.1872e-01,  1.3597e-02,  2.5993e-03,  3.1726e-03,
        -2.1915e-03, -4.1030e-03, -7.0654e-04,  3.7303e-04, -1.0126e-03,
        -2.9452e-04, -8.7181e-03, -3.2271e-04,  1.0514e-03,  5.7883e-03,
        -3.0395e-03,  1.0192e-03, -1.9403e-03,  1.8830e-03, -1.7331e-03,
        -3.2298e-03, -4.0097e-03, -1.8284e-03, -3.4972e-03, -5.1120e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2225e-02,  9.2929e-01, -7.2901e-03,  1.3522e-03,  7.8419e-03,
         4.2534e-03,  6.0617e-03,  3.4793e-03, -2.2378e-03, -5.2877e-04,
         1.8714e-03,  4.8212e-03,  2.6543e-03, -5.8159e-04,  1.4449e-03,
        -1.5881e-04,  2.7077e-03,  3.2320e-03,  2.2646e-03,  1.6495e-03,
        -4.0582e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9714e-03,  9.4479e-01,  5.7159e-04, -4.0584e-03, -6.0113e-03,
         1.2270e-03,  8.6271e-03, -3.9535e-04,  1.5583e-03, -7.7771e-05,
         3.7866e-03, -8.8571e-03,  2.2768e-03, -1.6884e-03, -1.9044e-03,
        -3.4735e-04, -2.2976e-03, -5.5511e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4887e-02,  9.0731e-01,  7.9499e-03, -1.0685e-02, -1.8198e-03,
         1.3109e-03,  1.5655e-03, -2.2181e-03, -2.6690e-03,  2.1284e-03,
         5.3112e-04,  3.7870e-03, -1.5741e-04,  8.2561e-04,  2.0359e-03,
         5.1671e-03,  2.8538e-03,  6.3653e-04, -6.4572e-04, -2.3691e-03,
        -1.5112e-03, -1.0713e-03, -5.9168e-04, -8.7452e-04,  2.6121e-03,
        -5.8459e-04, -4.4519e-04,  1.7534e-03, -1.4821e-03, -3.3257e-04,
         1.1703e-03,  1.6167e-03, -1.6333e-03,  5.1803e-04,  4.6053e-03,
         4.6284e-03, -1.8062e-03, -1.2117e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0072e-04,  9.6567e-01,  8.2970e-04,  1.5427e-04,  2.1176e-03,
         3.6660e-03,  6.9805e-04, -2.3572e-03,  3.1237e-03,  7.7944e-04,
        -1.9306e-04, -3.9093e-03, -7.8410e-04, -3.3254e-03,  6.4703e-04,
         4.3099e-04,  1.5760e-03, -1.8735e-03,  1.8934e-03,  5.7687e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0422,  0.8981, -0.0117,  0.0154, -0.0021,  0.0036,  0.0019,  0.0049,
         0.0011,  0.0046,  0.0047,  0.0018,  0.0013,  0.0066,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0160e-02,  9.2445e-01,  3.3451e-03,  6.7775e-04,  1.6717e-03,
         2.6462e-04, -3.0378e-03,  3.3905e-04,  1.4260e-03,  1.7896e-03,
        -3.1293e-03,  5.0932e-03,  1.4035e-03,  4.5076e-03,  3.0400e-03,
        -5.5895e-04,  1.9136e-03,  9.2626e-04,  8.8152e-04,  3.2121e-03,
         6.7969e-04,  3.0999e-03, -6.5754e-04, -1.5138e-03,  1.9052e-03,
        -7.8701e-04, -4.7650e-04, -2.3057e-03,  3.2889e-03, -6.7690e-04,
        -2.6436e-04, -2.5209e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2146e-02, -8.9732e-01,  8.2085e-03,  3.0176e-03,  5.0209e-03,
         1.9987e-03,  1.1391e-03,  1.1852e-02, -2.1762e-03,  1.5845e-03,
         4.1305e-04,  2.7919e-03,  1.7643e-03,  7.7017e-04, -1.4548e-03,
         1.1255e-03,  2.1935e-03,  1.1034e-02,  9.3694e-04,  1.3055e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.4725e-03,  9.0438e-01, -6.3023e-03, -5.0925e-03, -7.2894e-05,
        -1.9939e-03,  3.5885e-04, -2.9713e-03, -1.1503e-03, -4.1743e-03,
        -1.2160e-03, -1.0599e-03,  3.1698e-04, -8.5359e-04, -3.2081e-03,
        -3.3212e-03, -4.1787e-03, -5.3184e-04,  1.9767e-04, -2.5641e-03,
        -1.9580e-04,  2.4729e-03, -1.8813e-03,  1.2953e-03,  7.2811e-04,
        -2.4693e-03, -7.8227e-03, -2.0870e-03, -5.0234e-03, -1.9745e-03,
         2.3926e-03, -7.1275e-04, -2.4855e-03,  2.6421e-03,  2.1474e-04,
         2.4043e-03, -1.6561e-03, -1.7094e-03,  2.2886e-04, -1.4499e-03,
         2.7647e-04,  7.9553e-04,  1.2591e-03, -1.2869e-03, -8.8402e-05,
         1.9759e-03,  3.0593e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5505e-02,  9.1579e-01,  8.1947e-03, -1.1758e-03,  9.2238e-03,
         3.6235e-03, -2.0135e-03, -6.9386e-03, -6.8715e-03, -5.1537e-03,
        -7.1573e-03, -7.5353e-04,  7.0886e-03,  5.4320e-03, -5.0821e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0630e-03,  9.1174e-01, -6.3604e-03,  1.5972e-03,  2.5369e-03,
         1.9634e-03,  3.0187e-03, -6.4923e-03,  2.0296e-03, -1.0069e-03,
        -2.7080e-03,  1.2428e-03,  1.4236e-03,  4.7824e-03, -1.3961e-03,
        -2.9632e-03,  1.5725e-04,  7.5380e-04, -2.2173e-04, -6.3977e-04,
         7.6012e-04, -2.6991e-03,  8.0309e-04,  2.9640e-03,  8.4355e-04,
         2.7940e-03,  2.2988e-03, -6.7023e-03,  3.3834e-03, -1.8153e-05,
         1.0089e-03,  1.3720e-03, -6.9787e-04,  8.1775e-04, -5.7645e-04,
         1.4058e-03, -9.8280e-04, -2.0073e-03, -1.3546e-03, -1.0802e-03,
         1.2535e-03, -6.4312e-04, -1.1555e-03, -1.3188e-03, -6.5369e-04,
        -1.3966e-03,  1.3447e-03,  2.5621e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-6.7208e-03,  9.0433e-01,  8.2664e-03, -3.8905e-03, -4.8031e-04,
        -1.4352e-02, -3.0462e-03, -1.1591e-04, -5.4037e-03,  1.6693e-03,
         2.5444e-03, -9.8208e-03,  1.6213e-03,  1.2427e-03, -1.7172e-03,
         7.1989e-04, -9.1243e-03, -1.0742e-02, -2.3473e-05, -3.2576e-03,
         1.9690e-03, -4.0699e-03,  3.7033e-03,  7.4712e-04, -2.9130e-04,
         1.3332e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.4116e-04,  9.5808e-01,  2.3909e-03, -7.5369e-03, -1.9017e-03,
        -1.2036e-03, -1.3898e-03, -3.7967e-05, -7.4437e-04,  2.0353e-04,
         5.6213e-03,  4.0079e-04, -3.7221e-04,  8.0018e-05, -3.8685e-03,
         5.2337e-03,  4.7869e-03, -6.1810e-04, -3.8061e-03, -2.8903e-04,
        -4.9095e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9543e-02,  8.4730e-01, -1.1875e-02, -1.4852e-02, -3.7876e-03,
        -6.8917e-03,  5.0287e-04, -4.6401e-03, -3.9919e-03, -6.6136e-04,
         9.4346e-04,  1.0962e-04, -2.5646e-03, -6.6777e-03,  8.4326e-03,
        -4.2286e-03, -5.3756e-04, -3.8766e-03,  2.4963e-03, -4.6356e-04,
         3.1367e-03, -1.1064e-03, -1.4678e-02,  1.7193e-03,  2.6952e-03,
        -3.7294e-04, -4.5514e-04,  7.3629e-04,  1.5945e-03,  6.9470e-04,
         1.0083e-04, -5.3675e-03,  3.7380e-03, -2.7467e-03,  2.6016e-03,
        -3.8827e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.3084e-03,  9.3918e-01,  9.5028e-04,  3.7144e-03,  1.0215e-02,
         7.1040e-03,  3.2818e-04,  9.1476e-04,  2.0032e-03,  2.1011e-03,
        -2.0799e-03,  7.4540e-04, -7.4003e-04, -5.1019e-03, -1.4233e-03,
         3.1895e-03, -3.9049e-04,  3.7040e-03,  4.0074e-03,  2.8028e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.5469e-03,  9.4094e-01, -9.9119e-03,  6.9765e-03,  1.1608e-03,
        -1.1090e-03, -1.0599e-03, -7.5080e-04, -3.2936e-04, -1.4375e-03,
        -3.8364e-03, -4.2715e-03, -2.3012e-03, -2.0767e-03, -7.5055e-04,
        -5.8373e-03, -9.8900e-04, -1.1154e-03, -6.6203e-04,  1.9904e-04,
         4.7428e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3216e-02,  9.4068e-01, -1.7466e-02,  3.8463e-03, -1.5604e-03,
        -2.1124e-03,  1.1171e-04,  1.8496e-03, -1.7823e-03,  3.9282e-03,
         1.1911e-03,  3.5398e-03,  8.7142e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6376e-03,  9.1911e-01, -2.7412e-03, -5.1695e-03,  1.3414e-03,
        -4.3794e-03,  1.3927e-03, -8.8564e-04, -3.0571e-03, -3.1954e-04,
         5.6235e-04, -8.6920e-03, -3.4902e-03, -2.2518e-03, -1.0367e-03,
         7.3550e-04, -6.0684e-04,  1.8766e-03, -1.3165e-03, -2.6349e-03,
        -3.1175e-03,  1.3678e-03, -3.1846e-03,  4.3844e-04,  2.7491e-04,
         1.8598e-03, -1.7080e-03,  1.0447e-03,  2.0623e-03, -4.1754e-04,
         3.5939e-04,  1.5392e-03,  2.1277e-03, -2.8231e-03,  1.2220e-03,
         5.7465e-03,  6.4673e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.6254e-04,  9.1318e-01, -1.3606e-02, -4.1506e-03,  2.3228e-03,
        -3.3052e-03,  2.0616e-03,  5.7884e-03, -9.0951e-04,  1.3203e-03,
        -1.5304e-03,  6.0999e-03,  6.6414e-04, -2.8615e-04, -4.5750e-03,
        -3.5819e-03,  2.6628e-03,  1.2640e-03, -1.1240e-03,  1.4108e-02,
        -4.8248e-04, -1.2138e-03,  1.4881e-02, -2.1871e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8665e-04,  9.3427e-01,  1.8596e-02,  3.1320e-03,  4.8822e-03,
         1.2602e-03,  6.4278e-04, -2.7633e-03, -2.8695e-03, -1.9264e-03,
        -9.2003e-04, -1.9742e-03, -1.0531e-02,  6.5877e-04,  2.7418e-03,
        -1.1323e-03, -1.7465e-04,  6.1667e-04,  6.1015e-04, -1.7292e-04,
         9.3651e-06,  9.3947e-05, -3.4789e-04, -3.7915e-04,  1.5258e-04,
         1.8289e-04,  9.0834e-04, -4.8530e-04, -1.3259e-03, -3.3899e-04,
        -5.5479e-04,  4.9595e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2571e-03, -9.1845e-01, -7.3859e-03,  4.9357e-03, -1.5217e-03,
         1.9883e-02, -1.8998e-04,  2.7742e-03, -2.7782e-03, -3.7991e-03,
        -4.6890e-04,  1.2050e-03,  1.1830e-03,  8.0650e-04, -2.0212e-04,
        -8.2097e-04,  4.9701e-04,  3.5758e-04, -1.1403e-03, -5.6820e-04,
         8.0014e-03,  5.5924e-03,  6.5352e-04,  3.1304e-03, -5.6455e-04,
        -6.6499e-04, -5.4220e-04,  2.0183e-03,  1.1425e-03,  4.4513e-03,
        -1.2251e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4564e-02,  9.0112e-01,  2.2553e-03,  7.1340e-03,  5.0530e-04,
         3.6703e-03,  7.6303e-03, -9.2980e-03,  1.0216e-03, -1.9685e-03,
         7.5305e-04, -1.9316e-03, -2.1808e-03, -1.8756e-03,  1.2679e-03,
         2.1401e-03,  4.6000e-03, -2.1482e-04,  1.4952e-03, -5.0113e-04,
        -2.5717e-03, -3.9010e-03, -1.7395e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3097e-03,  9.3580e-01, -5.9141e-03, -6.0245e-03, -1.6455e-03,
         4.2308e-03,  1.0519e-03, -1.8411e-03,  3.2091e-03,  2.9582e-03,
         1.4141e-03, -3.7834e-03,  1.4924e-03,  1.4935e-03,  5.1995e-04,
         1.3830e-03, -8.4987e-04, -3.2547e-03, -1.1595e-04,  5.2058e-03,
        -1.5673e-03, -1.2268e-03,  5.5773e-03, -9.5050e-04, -7.8359e-04,
         2.1176e-03,  3.1484e-03,  1.1342e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 1.5497e-03,  9.5070e-01, -2.0829e-03, -2.1847e-03,  1.3137e-03,
        -6.4351e-04,  2.7799e-03, -2.2617e-03, -7.3328e-03, -2.6485e-03,
         7.5944e-04, -9.0310e-04,  1.1949e-03, -1.7095e-03, -1.6912e-03,
         3.6288e-03, -3.0912e-04,  2.6775e-03, -1.5326e-03,  6.7968e-04,
         7.7606e-04, -4.5153e-04, -8.6337e-04,  4.7685e-04, -1.0373e-03,
        -1.1900e-03,  3.2929e-03,  1.8214e-03,  1.5030e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0327e-03,  9.1570e-01, -6.0221e-03,  9.1631e-04,  1.0274e-02,
         1.0740e-02,  3.9588e-03, -2.5943e-03, -1.8639e-03,  2.9465e-03,
         1.6670e-03, -6.3898e-03, -1.6331e-03,  4.6425e-03,  2.7976e-03,
        -3.5759e-04,  3.1257e-03,  2.6252e-03,  2.6702e-03,  1.8181e-03,
        -1.4123e-03,  7.9789e-03,  1.0128e-03,  3.8195e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.2622e-03,  9.2018e-01, -7.3393e-04,  2.9982e-03,  3.6020e-03,
         5.5619e-03, -1.7598e-03,  1.1282e-03, -2.9942e-03, -1.1947e-03,
        -2.9193e-03, -3.7610e-03,  3.6414e-03, -2.6568e-04, -1.3950e-03,
         5.7253e-03, -1.3524e-03,  4.3898e-04,  2.8246e-03,  1.4845e-03,
        -1.5634e-03, -8.8336e-04,  5.1382e-03,  1.2094e-03,  1.3318e-04,
         6.5353e-04, -8.1388e-03, -2.5295e-04, -1.2812e-03,  1.2244e-03,
        -5.9600e-04,  4.2725e-03, -2.1862e-03, -1.2433e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.8800e-03,  9.2670e-01, -9.1199e-03, -8.4802e-03,  1.7926e-03,
        -3.9547e-03, -1.5926e-04,  3.0829e-03,  1.5591e-03, -1.9547e-03,
        -1.7589e-03, -2.6237e-03,  1.5467e-03, -1.9966e-03,  1.3409e-03,
         3.7567e-04, -3.7225e-03,  1.0880e-03, -2.8794e-03, -4.7408e-03,
         8.9722e-03,  3.2728e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1462e-02,  9.3030e-01, -7.9409e-03,  4.2797e-03,  4.1749e-03,
        -1.5619e-03, -8.4243e-06,  8.4997e-04,  1.7358e-03,  3.3874e-03,
         2.2212e-03, -6.0732e-03,  3.5786e-04, -2.9266e-03, -8.3728e-04,
        -1.0469e-02, -7.6002e-03, -3.8136e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6936e-02,  9.1515e-01, -5.1873e-03,  5.9415e-04, -2.3857e-03,
        -3.3406e-03,  1.7213e-03, -3.4775e-03,  2.3172e-03,  7.0821e-03,
        -9.4355e-04,  1.2169e-03, -3.4294e-03,  4.6793e-03, -1.3612e-03,
         5.5842e-04, -2.8887e-05, -6.1063e-03, -7.0035e-03, -1.1884e-03,
         2.2154e-03, -2.0232e-03, -1.0522e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5405e-04,  9.6792e-01,  2.8342e-03,  2.0292e-03,  2.3292e-04,
         5.2274e-04,  2.4279e-03,  1.4171e-03, -1.0460e-02,  5.9468e-04,
        -7.6423e-04, -6.5283e-04,  9.1962e-04, -2.2752e-03, -1.0838e-03,
        -3.8785e-03,  1.7281e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0937e-02, -9.4000e-01, -2.5368e-03,  6.7787e-03, -4.4438e-04,
        -5.6758e-04, -2.1022e-04, -1.4932e-03, -2.4546e-03, -1.2757e-03,
         1.1615e-03,  4.3133e-03, -1.2623e-03,  1.2171e-03, -3.5996e-03,
        -4.0517e-04,  1.2366e-03,  1.0934e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.2817e-04,  8.8469e-01, -1.8678e-02, -2.7956e-03,  3.8975e-03,
        -1.6881e-03, -9.2450e-04,  3.4815e-03, -4.9165e-04,  2.0294e-02,
         3.1693e-03, -3.7165e-03, -2.2324e-03, -6.0197e-03,  9.9513e-05,
         1.1857e-02, -1.4011e-02,  1.1300e-03,  2.6940e-03, -1.5735e-03,
         4.0287e-03, -4.6231e-03, -1.2338e-03, -9.3038e-04,  8.5180e-04,
         4.1576e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7265e-02,  8.8820e-01, -5.4966e-04, -6.4266e-03, -1.4894e-03,
         7.2526e-03,  3.4595e-03,  1.9725e-03,  3.8786e-03, -5.7481e-04,
        -5.8313e-03, -2.1232e-03, -2.2669e-03, -1.1408e-02, -5.7386e-03,
        -4.6698e-05, -2.4755e-03,  2.5719e-03,  2.7774e-03, -2.1124e-03,
        -2.3674e-03, -1.8695e-03, -1.7811e-05,  2.0139e-03, -1.0664e-03,
        -7.6359e-04, -2.6620e-03,  2.2919e-03, -4.0032e-04,  2.2100e-03,
         1.5915e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7046e-03,  9.3432e-01,  1.9403e-02,  4.0666e-03,  2.9870e-04,
        -2.3359e-03,  1.3754e-03,  3.1728e-03, -1.1746e-03,  7.4415e-03,
         7.1133e-04, -2.8379e-03, -6.7346e-03, -2.6904e-03, -3.0958e-03,
         1.7907e-03, -3.7864e-03, -1.0604e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0350,  0.9091, -0.0216, -0.0042, -0.0015, -0.0057, -0.0048,  0.0181,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 1.7405e-02,  8.9201e-01,  9.5858e-03, -3.9020e-03, -4.6628e-03,
         1.6328e-03, -2.0402e-03, -1.2546e-03, -2.4757e-03,  3.4898e-03,
         1.7472e-03,  3.6164e-03, -5.0644e-03, -3.6842e-03, -1.3801e-03,
         1.4186e-04,  7.3178e-05, -3.3213e-03,  6.4718e-04, -7.4996e-05,
         4.6278e-04,  3.2444e-03, -1.3599e-03,  1.6728e-03,  1.4623e-03,
        -3.0793e-04, -2.4081e-03,  2.8905e-03, -6.9211e-04,  3.0066e-03,
         2.3141e-03,  1.4650e-03,  9.6109e-04,  2.2779e-03,  3.5160e-03,
        -2.6503e-03, -5.5399e-04,  2.9013e-03,  7.6428e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2852e-02,  9.4098e-01,  7.2395e-04, -2.6371e-03, -9.1869e-04,
         1.8780e-03,  5.8338e-04,  1.1234e-03, -4.5701e-03,  5.5432e-03,
        -2.5727e-03, -7.8917e-04,  1.0259e-03, -8.5735e-04, -1.3746e-03,
         1.9669e-03,  1.2575e-03, -1.6865e-03, -1.9701e-04,  1.8437e-03,
         3.9768e-03, -2.3776e-03,  1.3799e-03,  4.0547e-04,  3.3586e-05,
         8.2253e-04,  4.9320e-04, -5.1261e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7649e-04, -9.6225e-01, -4.5708e-03, -1.2470e-03, -2.8452e-03,
        -4.3229e-03, -1.6084e-03, -4.1445e-03, -7.4649e-03, -7.4314e-03,
        -2.3454e-03, -3.7190e-04,  8.2625e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0004e-04,  9.3987e-01, -9.6723e-03, -3.6321e-03,  1.7143e-04,
        -1.8772e-03,  4.6433e-03, -5.9490e-03,  6.8648e-03,  1.6730e-04,
         1.0056e-03, -8.8011e-03, -3.0114e-03, -9.8826e-04,  1.2844e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6678e-02,  8.7740e-01, -2.5533e-02,  7.2400e-04,  3.9256e-03,
        -8.4197e-03,  6.1944e-03,  3.7372e-04,  2.9958e-03, -3.4507e-03,
         1.4164e-03,  2.3982e-03,  1.6501e-03,  1.7616e-03,  2.0152e-03,
         2.2826e-04,  3.3408e-03, -1.0320e-03,  9.6399e-04,  1.6235e-05,
         7.4979e-07,  6.4907e-04,  6.7204e-04, -1.3810e-03, -4.5345e-04,
        -1.9380e-03,  2.2322e-03,  6.7112e-03, -1.0526e-03,  3.0012e-03,
         2.5007e-04,  6.3647e-04, -9.8542e-04,  1.7552e-03,  1.7654e-03,
         5.3692e-04,  2.4323e-03,  1.3916e-03, -2.0767e-03,  1.3279e-03,
         2.1058e-03, -5.1795e-04,  4.6170e-04,  2.9421e-04, -2.2763e-03,
        -8.0130e-04,  1.7718e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9818e-03,  9.2603e-01, -1.9192e-03,  8.3209e-04, -4.1539e-03,
        -1.1523e-03, -2.2010e-04,  2.0156e-04, -1.2110e-02,  7.0046e-04,
        -7.0452e-04, -3.9047e-03, -6.0194e-04, -1.1042e-03, -2.4662e-04,
        -5.1493e-04,  2.7394e-03,  2.7697e-03, -2.9628e-03,  5.2188e-04,
         3.0226e-05, -2.3608e-03,  8.7070e-04, -7.7507e-04, -5.9323e-04,
        -4.6899e-04, -5.5410e-03, -5.8323e-04, -1.1407e-03, -2.1972e-03,
        -6.7946e-04, -5.5305e-04,  2.8457e-03, -1.2604e-04,  1.8756e-03,
        -1.4392e-04, -8.5351e-04, -1.6699e-04,  8.2863e-04,  8.4390e-05,
        -2.1855e-04,  1.1691e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3958e-02,  9.2080e-01,  1.8305e-03, -7.1202e-03, -2.4401e-03,
         2.5083e-03,  9.5794e-03, -2.1199e-03, -5.1196e-03,  9.3995e-04,
        -3.7993e-03,  1.0047e-03, -2.8790e-03, -2.3437e-03,  6.9046e-03,
         8.6381e-05,  2.5772e-03,  3.7597e-04, -8.1921e-04,  1.7097e-03,
        -1.8777e-03,  4.6939e-04, -2.6379e-03,  4.6598e-04, -5.6345e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8607e-02,  9.0970e-01,  8.4879e-03, -4.6882e-04,  1.3057e-03,
        -1.5093e-02,  1.2263e-03, -7.4354e-04,  3.1239e-03,  1.6924e-04,
         4.4405e-04, -3.7349e-03, -2.7320e-05, -6.3692e-03,  4.9931e-03,
        -3.4054e-04,  1.2361e-03,  9.6026e-03, -1.1126e-03,  2.1924e-03,
        -9.9017e-04, -1.3459e-03, -8.6855e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0884e-03,  9.0435e-01, -6.4166e-03,  1.3965e-02, -3.0774e-03,
        -2.1865e-03,  4.4141e-03,  3.9597e-04,  3.1789e-04,  1.5763e-02,
        -2.7015e-03,  1.7930e-03,  7.6894e-03,  9.3143e-04,  6.4996e-03,
        -1.1294e-03,  4.2335e-03,  2.2193e-03,  4.6853e-03, -2.6413e-03,
         2.8367e-03, -5.7536e-03, -4.9075e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6549e-02,  8.9329e-01,  1.2698e-02, -8.4308e-03, -4.4652e-03,
        -1.8984e-03,  9.2898e-04,  8.0746e-03,  3.9312e-03,  8.0917e-04,
        -1.7656e-02,  3.8606e-03, -7.4117e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.5045e-03,  9.6142e-01,  2.4711e-03, -2.9269e-03,  2.5772e-04,
        -7.2556e-04,  1.7379e-03, -8.9613e-04,  1.5562e-04,  3.5416e-03,
        -6.0744e-03,  1.0600e-03, -4.9930e-04, -1.1725e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4871e-03,  9.5131e-01, -8.5943e-03, -7.8765e-03,  2.2929e-03,
        -2.6945e-03, -1.1593e-03, -1.5441e-03,  1.6777e-03,  3.3428e-03,
         2.9119e-03,  1.1988e-03, -4.7734e-04,  8.9876e-04, -4.2847e-04,
        -3.0348e-03, -4.8754e-03, -9.9141e-05,  3.0914e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-9.1437e-04, -9.4923e-01,  9.0564e-04,  2.9427e-03,  5.3945e-04,
        -1.4870e-03,  4.0907e-04,  1.0007e-03, -3.4730e-04, -4.6596e-04,
        -6.3774e-04, -4.3076e-03,  8.1459e-04, -3.8019e-03, -3.0522e-03,
         6.4948e-03, -2.9941e-03,  8.2158e-04, -1.8243e-03, -6.1950e-03,
        -2.8534e-03, -4.1119e-04,  1.1418e-03, -3.8036e-03, -4.7552e-04,
        -2.1260e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0156,  0.9196,  0.0067,  0.0101, -0.0097,  0.0013, -0.0026, -0.0035,
         0.0049, -0.0019,  0.0036, -0.0039,  0.0022,  0.0050,  0.0061,  0.0033,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0112,  0.8858,  0.0162,  0.0417,  0.0057,  0.0014,  0.0011,  0.0081,
        -0.0009,  0.0012,  0.0013, -0.0048,  0.0023,  0.0030,  0.0061,  0.0027,
         0.0064,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6190e-02, -9.2376e-01, -2.0250e-02, -1.3012e-03,  5.8366e-05,
        -1.5200e-03, -4.7131e-04, -1.8131e-02, -8.3200e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3311e-03,  8.8995e-01,  3.1412e-03, -1.9489e-03,  1.7442e-03,
         1.4107e-03,  4.3065e-03,  2.6279e-03, -4.1379e-03,  1.6161e-03,
         6.3377e-03,  8.9355e-03,  2.0194e-03,  5.8308e-03,  6.7871e-04,
         3.9053e-03,  5.5946e-03,  4.4415e-03,  1.2122e-03,  3.6256e-03,
         1.8902e-03, -2.7391e-03,  9.2720e-04, -2.1266e-03, -2.9039e-03,
        -1.5204e-03, -2.1610e-03, -3.1149e-03,  2.0698e-03,  3.6401e-04,
         1.7208e-03,  7.1767e-05,  3.3694e-03,  1.9584e-04, -4.4350e-04,
         9.3224e-04, -4.0439e-03, -2.3416e-04, -3.0218e-03, -4.3781e-04,
        -1.2340e-03, -1.5124e-03, -1.9660e-03, -1.9591e-03,  2.8231e-04,
        -1.9582e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1912e-02,  8.8062e-01, -2.7975e-02, -8.9659e-03, -2.8557e-03,
        -3.2753e-03,  6.9018e-03, -2.0211e-04, -1.9646e-03,  2.2966e-03,
        -2.0012e-03, -1.1286e-03, -8.3947e-04, -1.5399e-03,  2.4963e-04,
         1.9657e-03,  1.5965e-03,  2.4893e-03, -2.2935e-03,  4.1842e-04,
        -2.7484e-03, -1.2369e-03, -1.3231e-03, -1.6026e-05,  1.3637e-03,
        -4.0335e-03,  2.0952e-03, -2.1183e-03,  1.1687e-03, -2.4085e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.0273e-03,  9.5810e-01, -1.6869e-02,  4.2095e-03,  1.0729e-04,
         7.9727e-03, -4.7095e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0075, -0.9439, -0.0084, -0.0038, -0.0017,  0.0026, -0.0082, -0.0056,
        -0.0028,  0.0024,  0.0132,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.9064e-03,  9.3858e-01,  1.7031e-02, -1.8307e-03,  1.0159e-03,
         7.4516e-03, -2.6707e-03,  4.2662e-04,  1.0035e-02,  2.8342e-04,
        -1.4842e-03,  5.0221e-03, -1.3922e-03, -3.8648e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5099e-03,  9.5399e-01,  2.6505e-03, -9.5923e-04, -8.2611e-04,
         1.2970e-04, -4.9030e-04, -2.3853e-03,  3.1379e-04, -6.5645e-04,
         4.1718e-04, -1.6878e-03, -3.0638e-03, -7.7193e-04,  1.1708e-03,
         1.9785e-03,  3.0212e-04, -2.3552e-03, -1.0151e-03,  2.6931e-03,
         1.2343e-03,  2.0698e-03, -9.0079e-04, -2.9887e-05, -1.3900e-03,
         4.6369e-04,  1.4664e-03,  7.4458e-04, -3.0764e-03, -8.6863e-05,
        -1.3336e-03, -1.9527e-03,  3.4630e-03,  1.4172e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.5819e-03,  9.0256e-01,  3.0483e-03,  1.0306e-02, -3.8376e-04,
        -1.9599e-03, -1.9831e-03, -7.5844e-03, -3.3727e-03,  2.0293e-03,
        -4.3466e-03, -1.8562e-03,  6.9200e-04,  4.6506e-04,  1.0611e-02,
        -3.8713e-03, -1.6634e-02,  9.8457e-04, -2.7168e-03,  3.3384e-04,
        -9.6966e-04,  6.4220e-04, -3.7204e-03,  4.3816e-04, -1.8955e-03,
         3.5456e-05, -5.0769e-03, -1.9022e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3243e-02,  9.1498e-01,  1.3804e-02, -1.1163e-03,  5.9653e-03,
        -6.9154e-03, -2.2270e-04,  9.2717e-04, -1.6965e-03, -5.8418e-03,
        -1.5279e-03, -3.7759e-04, -3.0661e-03,  4.6476e-03, -5.7699e-03,
        -6.2154e-06, -3.4270e-03, -2.8582e-04,  1.0876e-03, -5.0176e-04,
        -3.0509e-03, -1.4694e-03, -6.1279e-03,  3.9382e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 7.3395e-03,  9.1019e-01, -5.6625e-03,  6.9248e-03, -3.1654e-03,
        -9.0138e-03,  3.6397e-03,  5.7584e-04,  5.1797e-03, -7.4974e-04,
        -5.8928e-04,  2.2383e-03,  1.4185e-03,  1.2933e-03,  2.4846e-05,
        -1.8328e-03, -1.3483e-03, -2.3373e-03,  2.5713e-03,  1.1902e-03,
         2.3961e-03, -1.2854e-03,  8.5357e-04, -3.3179e-03, -3.0579e-03,
         3.6912e-03,  6.8412e-03, -4.9789e-04,  2.7615e-04,  3.6708e-03,
         6.8316e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0458e-03,  9.6993e-01,  1.7564e-03, -4.1394e-03,  2.9150e-05,
        -8.8930e-03, -9.2481e-04,  4.3066e-04,  7.8523e-03, -2.3395e-04,
         5.1771e-05, -1.7089e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.7712e-03,  9.5545e-01,  1.7453e-03,  7.9849e-04,  9.9370e-04,
        -1.7020e-03, -6.7470e-04, -1.5364e-03,  2.5758e-03,  9.0452e-03,
        -2.9990e-03,  5.7563e-03,  5.1108e-03, -3.9686e-03, -8.6832e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.5744e-03,  9.4944e-01, -2.8682e-03,  1.0845e-02, -3.6033e-03,
         9.7558e-04,  3.7026e-03, -4.5934e-04,  2.0269e-04,  2.0871e-03,
         1.1697e-03,  1.9526e-03, -2.4814e-03,  2.8294e-03,  7.0023e-04,
         4.7148e-04,  1.9614e-04,  2.8769e-03,  2.8052e-05, -8.5274e-04,
         3.0671e-04,  8.7291e-04, -3.9058e-04, -2.7654e-04,  1.3149e-05,
         1.3956e-03, -7.8978e-04, -7.8951e-05,  1.3106e-03,  9.4604e-06,
        -8.1512e-04, -4.1882e-04,  1.0050e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.0536e-03,  9.0126e-01,  2.7452e-02, -9.7546e-03,  4.5508e-03,
        -2.8962e-04, -2.1182e-03, -9.6655e-03, -1.9654e-03, -3.5220e-03,
        -7.5971e-03, -5.2439e-04, -3.5792e-03,  2.8287e-03,  4.9413e-03,
         1.0897e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.8980e-04,  9.4602e-01,  5.3042e-03, -1.3105e-02, -2.7841e-03,
        -5.3754e-03, -1.8874e-02, -7.5577e-03, -2.9068e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0205,  0.9134,  0.0088, -0.0303, -0.0269,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0054e-02,  9.3900e-01,  1.3176e-03, -2.2115e-04,  6.8404e-04,
        -5.9947e-03,  7.7269e-05,  6.2995e-04,  3.3470e-04, -3.1073e-03,
        -5.5931e-04, -2.9678e-03, -3.2888e-03, -3.0474e-03, -6.9431e-05,
        -1.8276e-03, -5.4255e-03, -2.6514e-03, -1.7576e-03, -3.3927e-04,
        -2.1740e-03,  1.1236e-03, -1.8639e-03, -3.4318e-03, -2.2325e-03,
        -5.1824e-03,  6.3936e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1691e-02,  9.4813e-01,  6.7835e-03, -9.0619e-05,  4.9980e-03,
         9.0521e-04,  7.3979e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5740e-03, -9.2455e-01, -4.7950e-03, -2.6004e-03, -1.7066e-03,
        -7.5957e-03,  8.5938e-03, -1.5828e-03, -6.3976e-04, -1.3350e-02,
         1.9148e-03,  4.9902e-03,  2.8012e-03,  2.0299e-03,  5.2225e-03,
         3.8431e-03, -6.1619e-04,  2.0503e-04, -7.2477e-04,  6.6169e-04,
        -4.7716e-03, -3.2346e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.8169e-03,  9.1664e-01,  1.4012e-02,  6.1464e-03,  8.4124e-03,
        -1.3939e-03, -2.6574e-03, -7.6664e-04, -2.4757e-03, -8.7143e-03,
        -7.0567e-04, -1.7636e-03, -5.4189e-03, -7.5712e-03, -1.2597e-03,
        -1.3534e-03,  3.7747e-03, -1.7819e-03, -6.3329e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.4709e-03,  9.4456e-01,  1.1575e-04,  3.0026e-03,  1.8271e-03,
        -1.8490e-03,  4.2532e-03,  2.9028e-03,  3.1049e-03,  2.9359e-03,
        -4.3058e-04, -4.6466e-06,  2.5860e-04, -3.7774e-04,  1.6172e-03,
        -1.5410e-03, -2.2225e-03, -6.9913e-04,  1.2492e-03,  1.1352e-03,
        -1.7025e-03, -3.7101e-03,  2.9846e-03, -1.6221e-03,  1.0416e-04,
         6.3771e-05, -1.5430e-03, -5.7190e-04,  5.2434e-04,  1.1461e-03,
        -3.4654e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 2.8497e-02, -8.8602e-01, -9.5094e-03, -4.3992e-03, -2.0606e-03,
         1.2975e-03, -1.6102e-03, -1.1174e-03,  1.7272e-03, -3.8778e-05,
        -1.7211e-03,  1.5094e-03,  8.0359e-04, -1.0870e-02,  3.0460e-03,
        -1.7748e-04, -7.1925e-04,  9.8293e-04, -2.8093e-03,  6.1777e-04,
        -4.2791e-03,  2.1878e-05, -2.5597e-03, -1.1056e-03,  3.7930e-04,
        -1.0505e-03,  9.7806e-04,  1.1526e-02,  2.4784e-03, -3.4267e-03,
        -7.6852e-03, -4.9739e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0912e-03, -9.3458e-01,  9.2464e-03,  1.0582e-03,  5.0062e-04,
        -6.2176e-05, -2.6385e-03, -2.6568e-03, -5.7571e-03, -2.0036e-03,
         3.8990e-03, -1.3635e-03, -3.0751e-03, -2.8728e-03,  1.9307e-03,
         2.8253e-03, -1.8583e-03, -1.4692e-03,  5.7232e-03,  9.8611e-04,
         3.0966e-03, -1.0309e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7847e-02,  9.1725e-01, -5.8321e-03, -9.6742e-03, -1.8613e-03,
         2.1898e-03, -1.4796e-03, -2.3850e-04,  1.8160e-03, -3.8850e-04,
        -1.0421e-03, -1.2537e-03, -4.8299e-03, -7.4754e-04, -8.9394e-03,
         1.7950e-03, -1.7178e-03, -8.7491e-04, -4.2446e-03,  1.9781e-03,
        -1.1067e-03, -1.3788e-03,  2.4220e-03,  6.3276e-04, -1.7342e-03,
         2.9645e-04, -7.3455e-04, -4.0615e-03,  1.3748e-03, -2.6118e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1810e-03,  9.5318e-01, -4.8885e-03, -4.4033e-03, -5.0923e-03,
        -2.7005e-03, -2.6112e-04, -7.3403e-05, -4.9093e-03, -1.9529e-03,
        -4.3639e-03,  1.2997e-03, -3.9895e-03,  9.3909e-04, -1.8203e-03,
        -2.6106e-03, -2.5326e-03, -8.0054e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3125e-02,  9.0064e-01,  2.1918e-02,  2.5448e-03, -1.8464e-03,
        -1.9014e-05,  2.4320e-03, -9.0137e-05, -1.5578e-03,  2.7107e-03,
         1.8780e-03,  5.0231e-04,  3.2605e-03,  1.8454e-03,  1.1930e-03,
         1.3518e-04,  5.5862e-05, -3.2965e-03, -3.7473e-03,  5.1446e-04,
        -8.6591e-04, -1.3447e-03,  1.8287e-03, -1.9821e-03,  1.6661e-03,
        -1.6142e-03, -4.0838e-03,  1.0586e-03, -9.4764e-05,  8.9002e-04,
         8.7789e-04,  7.1004e-04,  4.3319e-04, -6.1795e-03, -1.1087e-03,
        -1.3136e-03,  1.0639e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2478e-02,  8.9994e-01,  3.4258e-02,  2.4707e-03,  6.4576e-03,
         1.8887e-03,  4.5816e-03, -7.1031e-03,  6.2599e-04,  1.5543e-03,
         3.3792e-03, -4.2580e-04,  1.0389e-03,  3.3085e-03,  6.2529e-03,
         1.3122e-03,  1.8907e-03, -5.2964e-04, -1.4768e-04,  1.8878e-03,
        -1.5207e-03,  1.3612e-03,  5.5030e-03, -7.9422e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.6843e-03,  9.2137e-01, -6.2167e-03, -7.2397e-03, -2.6087e-03,
         4.8538e-03, -4.0654e-03,  1.5181e-03,  3.5300e-03, -1.9118e-03,
        -1.0037e-03,  1.9963e-03,  9.9354e-05, -4.2090e-03, -1.0519e-03,
         1.5224e-03, -1.3067e-03, -7.2186e-04, -8.4692e-05, -1.2193e-03,
        -3.1271e-04, -3.5528e-03, -5.2679e-04,  1.7637e-03,  1.5330e-03,
         1.6728e-03, -1.2691e-03, -6.4092e-04,  8.0536e-04, -3.6562e-04,
         2.5524e-03,  4.3814e-04,  6.2990e-05,  9.8227e-05, -3.8125e-04,
         1.2468e-03, -1.5967e-03, -1.8005e-04, -1.0526e-03, -4.8056e-04,
         5.9628e-04, -7.1356e-04, -1.1292e-03, -8.4220e-04,  1.5024e-03,
         4.7354e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.7048e-03,  9.3229e-01, -3.3587e-03, -6.3753e-03, -8.2454e-03,
        -9.3010e-04,  3.1202e-04, -1.1608e-03,  1.8652e-03, -8.1363e-04,
        -2.0681e-03,  2.1565e-03, -5.2864e-04, -2.6777e-03,  3.8150e-03,
        -2.3118e-03,  1.9740e-03, -7.7681e-03, -1.4642e-03, -4.7138e-04,
         3.1362e-03,  1.5761e-03,  9.9955e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4344e-02,  9.3280e-01, -1.4419e-02,  5.9956e-03,  9.9629e-04,
         6.0685e-04,  6.1177e-03, -9.7706e-04,  7.4280e-03,  4.7234e-03,
         1.1595e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6735e-02,  9.3980e-01,  1.0629e-02,  6.7122e-04,  3.5573e-03,
        -7.5567e-03, -2.6150e-03, -2.7780e-03,  3.6286e-03,  2.7171e-03,
        -4.4435e-04,  6.7216e-06,  4.0217e-03, -4.8367e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5148e-02,  9.2016e-01, -6.5037e-03,  1.2865e-03,  3.3545e-03,
        -8.0037e-04,  1.5245e-03,  1.4997e-03,  7.0982e-04, -1.5724e-02,
        -1.1201e-02, -2.0918e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9876e-02,  8.6962e-01,  1.5193e-02,  8.6115e-03,  1.8522e-04,
         5.6939e-03, -4.3857e-03,  2.6092e-03,  3.5827e-04,  2.9457e-03,
        -2.3057e-04, -6.4231e-04, -1.3010e-04, -1.2871e-03,  1.0370e-04,
        -1.6666e-03,  4.3676e-04, -2.7696e-03, -7.8731e-04,  5.7342e-05,
        -1.1671e-03, -2.1312e-03,  9.1559e-03, -3.2450e-03,  3.8104e-04,
        -1.5505e-04,  1.5170e-02,  2.7734e-04, -1.2784e-03,  7.6584e-05,
        -2.2146e-03,  4.2096e-03, -1.1137e-03,  2.8658e-03, -2.2956e-04,
        -1.9414e-03, -7.0035e-04,  9.5287e-04,  2.3966e-04, -9.4707e-04,
         1.3706e-04,  5.2122e-04, -3.2484e-03,  5.6730e-04,  1.9532e-04,
        -1.2870e-03,  1.8872e-03, -2.6984e-04, -6.4569e-04, -7.0212e-04,
         3.9815e-05,  5.8521e-04, -2.1016e-04,  1.4581e-03, -2.1979e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-2.1486e-03,  9.3745e-01, -9.8832e-03,  2.0841e-03, -2.0833e-03,
         5.5663e-04,  2.6065e-03,  4.1059e-03,  8.0120e-05, -6.5059e-03,
        -6.9796e-03,  4.4915e-04,  6.0372e-03, -1.1563e-04, -1.7140e-03,
         1.2680e-03, -1.7148e-03,  1.8240e-03,  1.0627e-03, -2.0631e-03,
        -7.1526e-04, -6.4255e-03,  2.1259e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.4741e-04,  9.5455e-01,  2.5469e-03,  6.1085e-03, -2.6646e-04,
         2.2769e-03, -1.0150e-03, -5.4952e-03, -1.4487e-03, -4.1783e-03,
         3.5072e-03,  6.4536e-04, -2.3804e-03,  3.8980e-03,  2.3734e-03,
         3.6726e-04,  5.7821e-04,  8.0147e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8231e-02,  8.9447e-01,  4.7600e-03,  5.7817e-03,  1.4241e-02,
         5.4052e-04, -3.1599e-03, -4.4031e-03, -2.8306e-04,  2.8731e-03,
         5.6171e-03, -1.2071e-03, -1.5740e-04,  1.9783e-03,  3.8875e-03,
        -4.2077e-04,  1.9866e-02,  8.1198e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1037e-02,  9.3119e-01, -6.3031e-03, -3.9634e-03, -7.2059e-03,
         5.8150e-03,  4.2468e-03,  1.2551e-03,  3.3834e-03,  4.1374e-03,
         3.6639e-04,  2.1430e-03,  1.4854e-03,  1.7336e-03,  5.7322e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4987e-02,  9.0005e-01, -6.8364e-03, -3.0139e-03,  1.1513e-03,
        -2.2241e-03, -3.0639e-03,  1.4199e-03, -2.5862e-03, -1.6977e-05,
        -3.8254e-03, -3.4988e-03,  1.8855e-03,  1.1297e-03, -3.3305e-03,
        -3.2220e-03, -4.2904e-03, -4.6481e-03,  2.4571e-03,  1.0442e-02,
         1.0175e-03,  3.9055e-03, -1.3457e-04, -2.7806e-04, -2.1283e-03,
        -1.1698e-03,  9.2144e-04, -6.3208e-04, -5.5613e-03,  1.1257e-03,
        -3.7563e-03,  5.2855e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.1771e-04,  9.0625e-01, -6.6520e-03, -4.7449e-03,  4.0136e-03,
        -2.7871e-03, -2.5573e-03, -2.5627e-03, -7.6048e-03, -2.2022e-03,
        -2.9047e-03, -2.4649e-03,  3.4103e-03, -1.0354e-02, -3.2133e-04,
         8.9121e-04, -2.5286e-04, -7.0013e-03, -2.9624e-03, -4.0080e-04,
        -2.7178e-03, -4.9150e-04, -2.9015e-03, -2.5525e-03,  8.2932e-04,
        -3.0773e-04, -1.2591e-03, -4.6505e-03, -3.2476e-03,  2.9364e-03,
         1.0950e-03, -9.7289e-04, -4.8863e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9670e-03,  9.2957e-01, -8.9868e-03,  7.2832e-04, -3.4907e-03,
        -7.8020e-04,  1.2493e-03, -2.0996e-03,  2.0907e-03,  1.5901e-03,
        -5.0502e-04, -3.9063e-03, -9.4765e-04, -5.0039e-04, -2.1968e-04,
         5.0966e-04,  2.3850e-03,  7.4385e-04, -2.3839e-03, -5.2495e-04,
         2.9509e-04, -1.5472e-03, -1.0815e-04,  1.0202e-03, -1.6128e-03,
         2.7453e-04, -5.4704e-04, -1.2070e-03,  2.0316e-04, -1.7290e-03,
         1.1042e-03, -1.0615e-03,  3.4850e-04, -1.6992e-03,  2.8652e-04,
        -9.1574e-04, -8.1167e-04,  8.1185e-04, -4.0361e-04, -4.2668e-04,
         4.0224e-04, -7.8978e-04,  2.5015e-04,  6.2235e-05, -6.1923e-04,
        -1.7984e-03, -4.4425e-04, -3.4828e-04,  1.2235e-04, -3.4462e-03,
         5.2575e-04, -1.0041e-03,  2.6182e-04, -3.6077e-04, -2.1223e-04,
         6.9432e-04,  1.2789e-03, -3.7815e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9448e-03,  9.2482e-01,  1.0105e-02,  9.7045e-03,  1.9818e-03,
        -2.0910e-03, -1.2272e-05, -5.5009e-04,  1.7447e-03,  4.6729e-04,
         4.7976e-03,  2.9223e-03, -7.9406e-03,  2.5375e-03,  2.4354e-03,
        -1.6093e-03,  3.2865e-03,  9.7447e-04,  8.2715e-04,  2.1476e-03,
         1.2968e-03,  3.5262e-04, -1.3278e-03,  5.9239e-03,  6.2448e-04,
        -5.5783e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.8082e-03,  9.0193e-01, -1.0671e-02,  1.0778e-02, -4.0425e-03,
         9.9534e-03,  5.7052e-03,  1.1005e-03,  3.4220e-03,  8.8780e-03,
         5.4811e-03,  1.5735e-03,  1.6331e-03, -6.9543e-04, -1.6779e-04,
         1.8454e-03, -2.4596e-03,  1.1931e-03,  2.4168e-03,  3.0396e-03,
        -1.4280e-04, -4.6713e-03, -1.4641e-03, -1.3468e-03, -1.0583e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.5708e-03,  9.1561e-01,  1.4056e-03,  6.1841e-03, -2.3343e-03,
         7.1393e-04,  2.7689e-03, -2.4345e-03,  1.8366e-04, -2.3956e-04,
         4.3444e-03,  1.2551e-03,  1.7268e-03,  7.1114e-03,  4.7872e-04,
         1.2093e-03, -4.2409e-04,  5.0647e-04, -2.9744e-04,  2.3670e-03,
         4.5114e-04,  8.1862e-04,  1.4301e-04,  1.1139e-03,  1.1122e-03,
        -1.7321e-04, -6.5317e-04,  6.9824e-04,  1.1210e-04,  1.3994e-04,
        -3.4808e-03, -2.5848e-04, -5.3291e-03,  3.8878e-04,  2.1122e-03,
        -4.1303e-03,  5.2822e-04,  2.3135e-03, -3.0168e-04,  3.0846e-04,
         4.3690e-04, -2.0911e-03,  3.8734e-03,  1.7469e-03, -1.1122e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2216e-02,  8.9451e-01,  2.5230e-03,  1.3223e-02,  3.3368e-03,
         1.7234e-03, -8.5270e-03, -6.3553e-03, -2.8008e-03,  1.6966e-03,
        -1.5602e-04, -1.7661e-03, -9.7420e-04, -6.8273e-05,  1.1920e-03,
        -1.5817e-03,  5.4211e-03,  1.6050e-03,  7.4335e-04, -4.6870e-04,
         1.0675e-02, -2.2140e-03, -6.2245e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3593e-02,  9.0916e-01,  7.4004e-03,  9.4153e-03, -6.1742e-03,
         8.6013e-03, -3.1658e-03,  3.3364e-03, -1.1201e-03, -4.6781e-03,
        -1.1524e-05,  3.3489e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 5.2346e-03,  9.4072e-01,  2.4925e-03, -8.0780e-03,  1.4467e-03,
         4.3473e-03,  2.3443e-03,  2.2472e-03,  4.9395e-03, -1.1315e-03,
        -3.7971e-03, -2.4667e-03, -1.5146e-03,  3.2442e-03, -1.9169e-03,
         3.0704e-03,  8.7124e-04,  1.5128e-03,  2.2004e-03,  2.0157e-04,
        -1.2098e-03,  2.5370e-03, -2.4760e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2122e-03,  9.4163e-01, -4.8667e-04,  3.6002e-04, -3.8949e-03,
        -1.5685e-03,  2.9858e-03,  2.6381e-05, -5.1905e-04,  8.2651e-04,
        -3.6813e-04, -2.3007e-03,  1.1890e-03, -1.1439e-03,  1.6880e-03,
        -1.6458e-04, -2.0693e-04,  1.2157e-03,  1.6893e-03,  1.3901e-03,
         5.1132e-03, -1.8940e-03, -2.9604e-03, -1.3584e-03,  3.5657e-03,
        -9.9681e-04, -2.8702e-03, -6.2430e-04, -1.7474e-03,  5.9372e-05,
        -5.7880e-04, -2.4294e-03,  7.9338e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2186e-02,  9.2202e-01,  6.5852e-03, -5.2649e-03,  4.4288e-03,
         4.8195e-04, -2.4187e-03,  2.3674e-03,  4.1437e-03, -4.2177e-03,
         6.0701e-03, -2.3859e-03,  2.1470e-03,  2.9515e-03,  4.7955e-03,
        -2.4511e-03, -8.5906e-04, -2.0009e-03, -2.2280e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.3757e-03,  9.3879e-01,  9.8319e-03, -5.6933e-03,  8.0333e-03,
         5.1713e-03,  8.4056e-03,  3.3241e-03,  4.8175e-04,  8.7562e-03,
         7.9401e-04,  4.4913e-04, -3.7457e-04,  1.5146e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4436e-02,  9.4819e-01,  6.7432e-03, -4.2641e-03, -6.5883e-04,
         1.4189e-03,  2.1139e-03, -9.7316e-05,  1.5644e-03, -2.4142e-03,
        -1.5780e-03, -2.1681e-03,  6.7231e-04,  2.9973e-03,  1.2916e-03,
        -1.5005e-03,  2.6570e-03, -1.6007e-04, -4.2740e-04,  1.0936e-03,
        -1.4324e-03,  2.1208e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2076e-02,  9.2557e-01,  6.6502e-03, -5.4235e-03,  2.8238e-03,
        -5.6560e-04, -2.7442e-03, -6.8285e-03, -7.8435e-04, -3.7611e-03,
        -3.0041e-03,  4.7570e-03,  2.8884e-03,  1.7514e-04, -1.1200e-03,
         1.0905e-03,  2.4123e-04,  2.7903e-03,  1.6558e-03,  3.4045e-04,
        -1.4212e-03,  1.1960e-03,  9.7556e-04, -1.3834e-03, -2.4511e-03,
         2.9656e-03,  9.1984e-04, -1.1020e-03,  2.2459e-03, -5.3635e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7094e-02,  9.2569e-01,  1.3938e-03, -1.7467e-03, -1.0433e-03,
         9.8031e-03,  7.0478e-03, -3.8679e-04, -1.8081e-03,  1.8084e-03,
        -7.4219e-03,  5.7500e-04, -2.3238e-03,  1.1984e-03,  3.7320e-03,
        -6.8472e-03, -7.6894e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4097e-02,  9.2697e-01, -3.6218e-03,  6.2357e-03, -7.3230e-05,
        -1.6641e-03, -1.3360e-03,  9.4152e-03,  1.8444e-04,  2.9168e-03,
         3.4807e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5108e-02,  9.2271e-01, -2.3125e-03, -3.7174e-03,  5.4068e-03,
        -5.8353e-03, -2.1439e-03, -2.1719e-04, -4.4526e-03,  7.8149e-04,
        -2.7333e-03,  2.6826e-03, -5.8458e-03, -7.7708e-03, -6.5385e-04,
        -7.1688e-03, -4.6375e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9707e-03,  9.4161e-01, -6.9800e-03,  2.8935e-03,  1.1288e-03,
         5.6155e-03,  1.2727e-03,  1.7881e-04, -6.5304e-04,  1.0795e-03,
         1.6181e-03,  1.6666e-03, -8.2069e-04, -1.6485e-03,  8.6256e-03,
        -4.3901e-04,  6.9962e-04, -6.1055e-04, -3.4485e-03,  3.5214e-04,
        -3.1370e-04,  2.4482e-03,  5.7493e-03, -1.1043e-03,  5.2341e-04,
         4.5490e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.9693e-02,  8.8925e-01,  2.9873e-02,  2.7789e-03,  1.0790e-03,
        -5.3542e-04, -1.9856e-03,  6.1409e-03, -2.1166e-03, -8.2556e-04,
        -4.6441e-04, -4.1483e-03,  3.2336e-03,  1.1418e-03,  1.4315e-03,
        -4.7584e-04, -1.9096e-03, -6.6475e-04,  1.4847e-03, -5.3183e-04,
        -1.1177e-03, -1.9882e-03, -9.5327e-04,  5.0003e-04, -2.0913e-04,
        -4.9660e-04,  3.7913e-03, -1.1316e-03, -5.1787e-03, -4.8694e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.0201e-03,  9.5132e-01,  7.6456e-04,  1.1606e-03, -8.6003e-05,
        -5.7678e-03, -1.6938e-03, -5.9617e-05, -6.6928e-04, -5.9228e-03,
         9.0387e-04,  8.4760e-03, -3.3634e-03, -6.2700e-05,  5.5741e-04,
        -1.4742e-03,  6.0721e-04, -3.4689e-03, -7.6218e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 1.7987e-02,  9.2696e-01, -2.0305e-03, -7.2553e-03, -1.4153e-03,
        -1.9139e-03,  4.8354e-03,  2.4759e-03, -2.3740e-03,  5.8248e-04,
        -9.6089e-04, -2.6508e-04,  1.2939e-05, -4.1813e-04, -7.9263e-03,
        -1.8550e-03,  2.4804e-03,  7.0631e-03,  4.0835e-05, -1.0847e-03,
        -2.1230e-04, -1.6101e-04,  1.7658e-03, -1.3935e-03, -1.9034e-03,
         7.0548e-04,  3.9221e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5453e-03,  9.5630e-01,  3.5299e-03, -4.5145e-03, -3.5228e-03,
         7.1002e-03,  4.0673e-03, -2.2536e-03,  3.2907e-03,  5.5165e-04,
         2.1745e-04,  8.2940e-03,  2.5926e-03,  2.2441e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3321e-03,  9.3268e-01, -1.6301e-04,  3.8207e-03,  5.5832e-04,
        -6.6868e-03,  3.5993e-03,  6.2454e-03,  1.3313e-03,  1.4315e-03,
         3.0595e-03, -2.9473e-03,  2.5290e-03,  1.2345e-03, -1.3233e-03,
         8.7226e-04, -8.3788e-04, -3.5115e-03,  2.4276e-03,  5.7909e-03,
        -1.3549e-03, -4.2059e-03, -6.4200e-04,  1.5180e-03, -1.1684e-03,
         8.7327e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5123e-03,  9.3646e-01,  1.3542e-03,  4.9570e-03,  8.0618e-03,
        -1.2550e-02,  8.4226e-05,  3.5142e-03,  6.1509e-03,  1.7225e-03,
         4.1606e-03, -9.1953e-03, -3.2173e-04, -8.9524e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.3138e-03,  9.5243e-01, -2.3857e-04, -3.4778e-04, -1.0210e-03,
         3.5473e-04,  3.4574e-03, -7.4719e-04,  1.2217e-04,  3.1967e-03,
         5.5675e-04,  6.3850e-04,  2.2565e-03, -6.2773e-05,  2.2492e-03,
         2.4367e-03, -3.9961e-03,  9.8481e-04, -2.4871e-04,  3.0560e-03,
         5.9917e-04,  3.5923e-03, -1.3078e-04,  2.2333e-04,  1.8846e-04,
         1.6812e-03,  4.8933e-04,  1.1685e-03, -2.1261e-05, -5.6079e-04,
        -6.7326e-04,  3.9568e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1663e-02,  9.1185e-01,  1.0112e-04,  1.8460e-02, -1.5987e-03,
         4.3139e-03,  1.1562e-03, -8.0997e-03,  2.3444e-03,  5.2179e-03,
         1.7209e-03,  1.6577e-03,  2.0522e-03, -1.4046e-03,  2.2754e-03,
        -2.5452e-03,  1.2288e-03,  2.4358e-03,  2.7924e-03,  4.2188e-03,
         2.4497e-03, -1.9835e-03,  6.1276e-04,  1.2311e-03,  6.5873e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4886e-03,  9.6963e-01,  1.9232e-02, -4.9863e-03,  7.4208e-04,
         2.9231e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6029e-02,  8.9934e-01,  5.9332e-03, -1.3349e-02,  3.5336e-03,
        -3.4645e-03,  2.2647e-03, -4.6831e-03, -1.2206e-03,  3.1405e-03,
        -2.3316e-04,  1.3343e-03,  3.4916e-04,  6.4593e-03, -1.9097e-03,
        -1.7915e-03,  6.5225e-04,  6.6880e-04, -1.3617e-03,  2.4882e-03,
        -3.1963e-03,  6.4600e-04, -8.0361e-04, -4.8806e-04, -1.2998e-03,
         1.9036e-03, -5.6316e-04, -2.1738e-05,  6.8950e-04,  1.9101e-03,
         2.1233e-03, -8.7188e-04,  5.1394e-04, -2.3462e-03,  1.4399e-03,
         8.4442e-04,  1.3141e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.6930e-03, -9.0875e-01,  9.1862e-03,  7.9442e-04, -3.8112e-03,
         3.4389e-04,  2.2110e-03, -4.8174e-03,  2.1452e-03,  4.6149e-04,
         2.7409e-03, -2.5549e-03,  3.5496e-03,  3.8480e-03, -2.7383e-04,
         2.8716e-03,  2.9163e-03, -1.1267e-03,  2.2431e-03, -1.5188e-03,
         6.9378e-03, -3.5032e-04,  6.3241e-03, -2.4565e-05, -8.7658e-04,
        -4.3594e-04,  1.0359e-03, -4.6457e-03, -2.0283e-03, -2.2752e-03,
         7.5165e-04,  1.5170e-03,  4.0233e-04,  4.7611e-03,  2.7768e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2649e-02,  9.3940e-01, -4.4964e-03, -5.0868e-03, -1.9539e-03,
        -1.8021e-03,  5.6604e-03, -1.5291e-03, -5.1390e-03,  2.5724e-03,
         2.4908e-03, -7.8928e-03,  8.0300e-03,  1.1969e-03,  1.0496e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4697e-03,  9.3148e-01, -1.2704e-02,  6.9044e-05,  1.4600e-03,
         4.9752e-03,  2.5705e-03,  1.4246e-03,  6.4449e-04,  4.8823e-04,
        -3.9614e-03,  1.4666e-03,  2.6293e-03,  1.5186e-04, -4.1239e-04,
        -1.7019e-03, -1.4474e-03, -1.3384e-03,  2.7353e-03, -3.2296e-04,
         3.4529e-04,  2.2334e-03,  3.7475e-04, -5.7985e-04, -6.3637e-04,
         8.9298e-04,  3.9030e-04,  3.7353e-04,  1.1960e-03,  2.7141e-03,
         1.0175e-03,  8.9793e-03,  6.8119e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3592e-02, -9.0089e-01, -3.0555e-03, -1.8373e-04,  6.9372e-04,
        -3.5959e-04, -1.7840e-03,  5.7891e-04,  1.0570e-03, -2.2975e-03,
         4.1174e-03,  7.1084e-03,  2.8437e-03,  1.5188e-02,  3.9703e-03,
        -2.7217e-04,  9.5413e-04,  6.5606e-04,  2.4035e-04, -5.6001e-04,
        -8.1516e-04,  5.8953e-04,  2.2418e-04, -3.5733e-04, -1.1272e-03,
         3.0220e-03,  1.9434e-03, -3.2788e-03,  8.2350e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 2.4677e-02,  9.0467e-01, -1.5450e-03, -5.1743e-03, -1.0958e-02,
        -6.7997e-03, -1.0563e-03,  8.5457e-03,  2.2634e-03, -1.6088e-02,
        -2.5874e-03, -2.6565e-04,  6.1476e-04,  6.0328e-03,  8.7189e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1847e-02,  8.9650e-01, -2.3289e-02, -1.0868e-02, -3.6740e-03,
        -1.4510e-03, -1.9202e-03,  1.7491e-03,  2.8709e-03, -9.9375e-04,
         2.9390e-04, -1.1671e-03, -1.3954e-03, -6.7508e-04,  2.7413e-03,
         4.3251e-03,  1.9381e-04,  2.1496e-03, -7.6000e-04, -3.7898e-04,
        -8.6984e-04, -2.1896e-03, -1.3049e-03, -5.4427e-04,  9.8833e-05,
         3.1307e-04, -1.1584e-03, -2.4126e-04, -2.1837e-03,  1.4624e-04,
        -4.4592e-04, -8.7085e-04, -3.9238e-04, -6.1666e-04, -8.4372e-04,
         1.4444e-03,  7.2549e-04,  3.6763e-04, -2.6760e-04, -1.2473e-03,
        -1.3609e-03,  1.2973e-04, -2.5128e-03, -1.0556e-03, -3.6804e-04,
        -2.5961e-03,  8.5784e-04, -4.1609e-05, -5.5634e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.6928e-03, -8.7805e-01, -1.9163e-02, -5.4951e-03, -4.0517e-03,
        -7.4629e-03, -6.6775e-03, -2.0744e-04,  7.1770e-03, -5.9645e-03,
        -3.0815e-03, -3.9454e-03,  3.1761e-04, -2.4423e-03, -1.2916e-03,
        -5.8193e-04, -9.2564e-03, -4.0286e-03, -1.7150e-03, -9.2308e-04,
        -4.4464e-04,  1.5623e-03, -2.2294e-04, -1.4154e-03, -2.7168e-03,
        -1.1081e-03, -3.5523e-03, -5.1238e-03, -4.1141e-03,  7.8997e-04,
        -1.7578e-03,  1.8610e-04, -3.4113e-04, -1.8401e-03,  8.2984e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.2729e-03,  9.4518e-01,  4.6594e-03, -4.6837e-03, -2.9518e-04,
         1.5930e-03, -5.4247e-03,  1.3323e-03,  1.6966e-03, -2.6129e-04,
        -6.1438e-03,  2.3668e-03, -1.0163e-03, -7.7093e-04,  4.3622e-03,
        -1.5652e-03, -1.2348e-03, -5.2096e-03, -4.9344e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3201e-02,  9.2688e-01, -3.0048e-05,  5.6250e-03,  4.3133e-03,
         3.4386e-03, -1.1290e-03, -1.2477e-03,  7.6610e-03, -7.3378e-04,
        -2.8583e-03, -3.7049e-03,  6.6888e-04, -2.1165e-02,  7.3420e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.4936e-03,  9.2602e-01, -8.0968e-03, -6.6291e-03, -1.0660e-03,
        -8.2123e-03, -1.9233e-03, -3.0570e-03, -5.9042e-03, -1.9660e-03,
        -9.6874e-04,  4.2260e-04, -2.6409e-03, -2.7562e-03,  1.8857e-03,
        -2.0751e-03, -1.5815e-03, -9.3827e-03, -6.6613e-04,  3.5527e-03,
        -1.2795e-03, -4.4153e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8204e-02,  9.0491e-01, -1.4190e-02,  1.8377e-04,  2.9272e-03,
         2.6896e-03, -2.5164e-03, -2.1976e-03, -9.0887e-03, -1.3571e-04,
        -1.9449e-03,  7.8576e-05,  5.1221e-03, -1.6704e-03, -1.3163e-03,
         1.1218e-03, -6.2583e-04, -2.3414e-03,  2.4689e-03,  1.3231e-05,
        -5.7482e-04,  6.8690e-04, -1.1202e-03,  1.9760e-03,  4.7482e-03,
        -1.4467e-03, -2.4342e-03, -4.2901e-03,  3.4479e-03,  5.5297e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.4898e-04,  9.2851e-01, -1.0022e-02, -1.7352e-02, -1.9215e-03,
        -4.4049e-03,  3.1587e-03, -3.9364e-03,  1.8316e-02,  4.3996e-03,
         7.1315e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2404e-02,  9.0757e-01,  4.7851e-03,  6.2590e-03,  9.1515e-03,
         2.5774e-03,  4.2254e-03,  1.2626e-03,  3.8522e-04, -1.4841e-03,
         6.6759e-03,  8.2584e-03,  1.5788e-03,  1.1014e-03,  3.5389e-03,
         6.0936e-03, -3.5624e-03,  9.7298e-04, -1.7018e-03, -3.0952e-04,
         1.1461e-03, -4.9586e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6173e-02,  9.0536e-01, -2.9230e-03, -6.3482e-03,  3.3985e-03,
        -1.1690e-02, -2.8611e-03,  1.6542e-03,  7.3401e-04, -5.6591e-04,
        -9.1010e-04,  3.2389e-03,  2.9604e-05, -4.6863e-05, -4.5367e-04,
        -2.8465e-03, -8.5953e-04, -4.0605e-04, -3.8526e-04, -4.4812e-04,
        -9.6068e-04, -2.3621e-03, -8.8139e-04,  3.6629e-04,  6.9412e-04,
        -6.4641e-04,  2.1388e-04, -1.3744e-03, -1.0423e-02, -1.0787e-03,
         3.2683e-03, -1.3463e-03,  5.0539e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.1344e-03,  8.8789e-01,  2.0677e-02,  1.0381e-02,  1.7400e-03,
         4.6377e-03,  2.1265e-03,  2.7481e-04, -4.7787e-03, -2.9392e-03,
        -3.3444e-03,  3.6494e-04, -7.5573e-04,  3.8021e-04, -3.4583e-03,
         1.4838e-02,  3.4960e-03,  5.3648e-05,  7.9411e-05, -1.1868e-03,
         5.8771e-03, -2.1459e-03,  4.3316e-03,  6.2523e-04,  2.2689e-04,
         9.7733e-03, -2.0346e-03, -2.4477e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4411e-02, -9.4188e-01, -1.4647e-03, -2.5589e-03,  4.6594e-04,
        -2.3319e-03, -1.0212e-03, -6.6902e-03,  7.9833e-03, -4.2729e-03,
         2.1357e-03, -1.0422e-03,  7.3747e-04, -8.1965e-04,  4.4978e-04,
         5.5895e-04, -9.6863e-04,  1.9020e-03, -3.6613e-05, -2.7318e-05,
         1.7071e-03, -9.8664e-05,  4.5849e-03, -4.5299e-04,  1.3975e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 1.4335e-02,  8.9379e-01,  1.9495e-02, -4.8839e-03,  6.2374e-04,
         3.4266e-03, -2.9085e-03, -2.4240e-03, -5.5361e-03, -4.0033e-04,
         2.4486e-03,  3.0677e-03, -7.3434e-04, -1.4921e-03, -2.6480e-03,
        -5.6981e-03, -1.2345e-03,  9.2199e-04,  5.8013e-04, -1.6812e-03,
         1.4038e-03, -5.6763e-04, -2.7273e-03,  5.6899e-06,  6.3608e-04,
        -5.1084e-04,  1.5585e-03, -3.1031e-03,  6.4569e-04, -5.9536e-03,
        -3.9645e-03, -9.7783e-04, -2.1792e-03,  7.6854e-04, -9.2136e-05,
         1.6053e-03, -1.5523e-03, -7.7486e-04, -5.9022e-04, -4.8274e-04,
        -1.5688e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.2806e-03,  9.1893e-01,  3.9548e-03, -3.1121e-03,  2.1447e-03,
        -4.9884e-03, -2.1066e-03, -5.0233e-03, -5.6115e-03, -3.2079e-03,
        -7.0972e-05, -6.4150e-03,  1.0569e-03,  1.5552e-03,  2.6566e-04,
         2.2067e-03,  9.0236e-05, -2.8321e-03, -9.9034e-04, -3.9354e-03,
        -4.3668e-03, -9.6802e-04, -1.2749e-04,  6.0914e-04,  1.7148e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5931e-02,  9.3387e-01, -6.6274e-03, -7.2102e-03, -6.2239e-03,
         7.8713e-04,  9.9327e-04,  4.0235e-03, -4.1001e-03,  1.0602e-04,
        -1.3565e-03,  3.2195e-03,  1.2588e-03,  2.3310e-03, -2.8887e-04,
        -8.6105e-04,  4.9258e-03,  1.8368e-03, -1.7611e-04,  1.4833e-03,
         2.3939e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2554e-03,  9.3121e-01, -8.9336e-03, -8.6852e-03,  1.3135e-03,
         4.1365e-03,  5.9990e-03,  4.1010e-03, -2.3319e-03, -4.4896e-04,
         3.6696e-03, -3.3412e-03, -1.3530e-03, -1.2850e-03, -6.6189e-03,
        -5.7214e-03, -2.8178e-03, -4.7761e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4099e-02,  9.1518e-01, -6.3303e-03, -1.3465e-03, -1.0396e-03,
         1.4681e-03, -1.1633e-03, -4.1695e-03, -6.7310e-03,  1.2114e-03,
         3.0172e-03, -2.9388e-03, -4.5851e-03,  1.0216e-03, -2.9536e-03,
         4.3042e-03,  2.4285e-04, -8.7024e-04,  1.0404e-03, -5.9062e-04,
        -1.4785e-03, -1.7725e-04, -1.4839e-03, -4.6650e-04,  3.5285e-03,
         1.1091e-03, -8.6029e-04,  8.7377e-04, -1.9642e-03, -3.6437e-03,
        -2.8144e-05,  3.6027e-04,  2.9156e-04,  7.0110e-04,  9.8003e-04,
         4.1215e-03,  1.7786e-03,  1.8467e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6392e-02,  9.0151e-01, -1.8690e-02, -1.2173e-03,  2.6900e-03,
         1.5665e-03, -2.3154e-03,  9.7925e-06, -6.5283e-03, -3.3294e-03,
        -9.7245e-04, -3.8345e-03, -1.0556e-03,  3.6042e-03, -7.4477e-03,
        -3.3282e-03, -2.0882e-03, -7.0498e-03, -5.8253e-03, -5.4232e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6658e-02,  9.2308e-01, -2.5652e-03,  6.4160e-03,  2.4627e-03,
         1.9209e-03,  1.1676e-03,  4.6101e-04,  2.9566e-03,  4.9638e-03,
         1.0082e-02, -3.8325e-03,  1.0245e-02,  3.1868e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0875e-03,  9.4362e-01, -5.0028e-03,  1.4210e-03,  5.5990e-03,
        -2.4350e-03, -2.4687e-03,  7.3399e-04,  1.7853e-03,  1.7753e-03,
        -1.4408e-03,  1.1618e-03,  2.0420e-03, -2.8566e-04, -4.6667e-04,
        -2.1857e-03, -1.6306e-04, -1.6565e-03,  1.5819e-04, -7.8782e-04,
        -2.6247e-03,  7.8095e-03, -1.2030e-03,  1.7618e-03, -2.9914e-04,
        -9.9368e-04,  1.3993e-03, -4.2923e-04,  7.1039e-04, -1.9285e-03,
         1.5575e-04,  2.4124e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5289e-02,  9.1154e-01, -7.4320e-03,  1.1151e-02, -1.6590e-02,
        -5.5302e-03, -5.7519e-03, -2.8639e-03, -4.8029e-03, -4.0409e-03,
        -4.3409e-04, -5.1691e-04,  6.9540e-04, -2.5338e-03,  1.6572e-04,
        -5.8050e-04,  5.4429e-04, -2.4805e-03, -5.1882e-03,  1.8738e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.7818e-04,  9.2142e-01,  7.5627e-03, -3.7107e-03,  4.5328e-03,
        -4.0819e-03, -1.7505e-03,  1.9613e-04,  3.8470e-04, -1.7220e-03,
        -3.3121e-03,  4.3513e-04, -1.0412e-03,  2.1248e-04, -9.2813e-04,
        -2.2869e-03, -1.1858e-03,  1.2490e-04, -2.0108e-03, -1.7416e-03,
        -8.6248e-05, -5.3871e-04,  6.0063e-04, -2.0597e-03,  1.7174e-03,
        -3.0522e-03, -3.3653e-03, -3.2523e-03, -1.9098e-03, -1.1177e-03,
         6.8450e-04, -5.8369e-04, -3.4603e-04,  4.5022e-03,  1.4911e-03,
        -2.3350e-05,  1.2806e-03, -1.9450e-03,  2.4111e-03,  6.9856e-04,
         7.2462e-04,  8.2901e-04,  9.4587e-04, -1.7999e-03,  3.5106e-04,
        -3.8179e-03, -6.4538e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.5804e-03,  9.4055e-01,  7.3766e-05,  6.2265e-03,  8.1442e-03,
         5.2522e-03,  4.5500e-03, -2.6874e-03,  6.4343e-03, -3.9446e-03,
        -1.6240e-04, -3.4715e-03, -2.9719e-03,  1.3637e-03,  4.5850e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0662e-02,  9.0719e-01, -4.9206e-03, -1.4305e-03, -4.1194e-03,
        -2.4373e-03, -1.5684e-04, -2.4610e-03,  4.8309e-03,  3.5345e-03,
        -2.3390e-03, -2.9805e-04,  6.7679e-05,  1.3704e-03, -2.4537e-03,
        -1.8232e-03, -1.9732e-03,  2.4942e-04, -4.3711e-04, -1.3498e-03,
         2.0658e-04, -3.7760e-03, -1.3252e-03,  7.0524e-04,  4.1531e-05,
        -1.8203e-03,  2.7784e-03,  1.1798e-02,  5.6347e-04,  1.9134e-03,
         5.5826e-06, -2.6413e-03, -7.8085e-04, -4.7906e-04, -4.4336e-05,
        -2.4787e-03,  8.7255e-05, -9.5916e-04, -4.1009e-03, -1.7135e-03,
        -7.4234e-04, -8.4272e-04, -1.3848e-03, -1.8562e-03, -1.5796e-04,
        -7.7375e-04,  6.5876e-04,  1.2605e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-5.8646e-03,  9.1420e-01, -3.6404e-03, -3.9269e-03,  3.6590e-03,
        -1.0053e-02,  9.8331e-04, -5.5628e-04, -2.8631e-03, -1.2037e-02,
         7.0030e-04, -3.8318e-03, -3.9563e-03,  1.1586e-04, -9.4062e-04,
        -3.0082e-03, -9.9997e-03, -5.1798e-03, -1.3723e-03,  1.5740e-03,
        -1.1669e-04,  1.0766e-04,  4.8621e-04, -3.8631e-03, -2.1982e-03,
         4.7613e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5600e-02,  9.2987e-01, -2.2749e-04, -8.9765e-03, -5.1988e-03,
        -3.7510e-03,  1.5943e-03,  1.8637e-03, -7.3137e-03, -2.4108e-03,
         3.0890e-03,  1.8484e-03, -2.8017e-04, -1.3632e-03,  1.0099e-03,
        -7.5266e-03,  2.1489e-03, -1.8099e-03,  2.1571e-03,  5.8413e-04,
        -1.3785e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1450e-03,  8.9831e-01, -3.2254e-03, -1.8272e-02, -5.9720e-03,
        -4.7892e-03,  7.7696e-04, -1.0606e-03, -1.6806e-03, -1.2898e-03,
         3.4582e-04, -1.5002e-03, -1.9102e-03, -2.5623e-03,  8.0988e-03,
        -1.7122e-03,  2.8962e-03, -2.0358e-03,  1.3783e-03, -3.6562e-03,
         4.1968e-04, -1.8584e-03, -3.0876e-03, -9.4782e-05,  6.8143e-03,
        -8.9690e-04, -1.1675e-03, -2.2404e-04, -8.1843e-04,  1.9598e-03,
        -4.0544e-03, -3.9514e-03,  3.1885e-03,  2.5481e-04, -2.4917e-03,
        -1.1020e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7585e-03,  9.2818e-01,  2.7304e-03,  8.3045e-03, -3.2869e-03,
         3.8173e-03, -5.3055e-03,  5.2754e-04,  5.4463e-04,  3.0490e-03,
         8.0397e-03, -3.3279e-03, -2.4615e-03, -5.0366e-04, -5.0464e-04,
         7.2800e-03,  2.0812e-03,  4.9709e-05,  1.0771e-02, -5.4783e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5504e-03,  9.3647e-01, -6.1501e-03,  2.5813e-03,  1.9404e-03,
        -3.1560e-03, -1.7148e-03, -3.4524e-03,  8.7145e-03, -3.0474e-03,
         8.9458e-04, -2.8003e-03, -2.1496e-03,  1.0824e-03,  4.2203e-03,
        -2.0406e-03,  6.9165e-05, -4.5462e-03, -2.1788e-03,  9.2642e-04,
        -6.3117e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5691e-02,  9.3329e-01, -5.4691e-04, -5.1386e-03, -3.5908e-03,
        -9.1115e-03, -3.5251e-03, -5.4492e-03, -6.4962e-03, -3.9748e-03,
        -2.8249e-03,  2.5571e-04, -1.0107e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6362e-02,  8.8815e-01,  7.1604e-03,  3.7952e-04, -1.9340e-04,
        -3.4084e-03,  8.6012e-03, -3.8645e-03, -5.0455e-03, -4.4904e-04,
        -7.6400e-03, -5.5660e-03, -4.5071e-03,  9.8001e-04,  4.7812e-03,
        -2.4373e-03, -2.1151e-03,  1.4984e-03, -3.4514e-03,  2.4382e-04,
         1.0502e-03,  6.9612e-03, -2.6747e-03, -2.6643e-03, -1.4636e-04,
        -1.0760e-03, -1.6029e-03, -1.8442e-03, -6.1989e-04,  4.8059e-04,
         4.5966e-04,  1.3028e-03,  2.5991e-04, -2.8556e-03,  1.0112e-03,
         3.5665e-03,  4.5882e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.0502e-03,  9.3167e-01, -1.7734e-03, -4.0796e-03, -1.7678e-03,
        -2.9641e-03,  2.0371e-03,  3.7496e-03, -1.3981e-03, -6.8041e-03,
        -9.2368e-04,  3.4130e-03, -2.0361e-03,  2.7531e-03,  1.0840e-03,
        -1.8268e-03,  3.2761e-03,  1.8815e-03,  6.3652e-04,  1.0179e-03,
        -2.2896e-04,  5.0309e-05,  9.7808e-03,  5.8019e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9673e-02,  9.2602e-01,  1.0431e-02,  4.5041e-03, -1.1890e-03,
        -2.7393e-03,  9.6526e-05,  3.7920e-03, -4.0243e-04, -4.1315e-03,
        -5.3580e-03,  3.1554e-03,  4.9243e-04,  1.0223e-03, -6.3289e-04,
        -1.2782e-03, -1.2690e-03,  3.2676e-04, -1.7880e-04, -3.2650e-04,
         6.5700e-04, -1.9681e-03,  6.5400e-04, -1.0285e-04, -1.7232e-03,
        -4.7096e-04,  9.1248e-05, -1.1733e-04, -1.4786e-03,  7.3077e-04,
        -3.8165e-03,  1.1707e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2710e-02,  9.0773e-01,  1.9784e-03, -1.6156e-02,  1.6301e-04,
        -9.3971e-03, -2.5184e-03, -4.3131e-03,  7.5613e-05,  5.4746e-03,
         3.4998e-04,  4.9510e-04, -9.2090e-04, -1.0576e-03, -4.2513e-04,
         1.3193e-03, -2.1792e-05, -3.6414e-03,  2.0804e-03, -8.4406e-05,
         2.1836e-03, -8.7372e-03, -1.0008e-03,  2.3923e-03, -1.3098e-03,
         1.0528e-03,  1.6659e-03, -3.7299e-04,  1.5682e-03, -2.1996e-03,
        -6.6018e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.0828e-03,  9.1599e-01,  1.1554e-02,  5.5912e-03,  4.5175e-03,
         4.1460e-03,  2.6235e-04, -1.1654e-02, -3.1305e-03,  1.7702e-03,
         3.7379e-03,  1.7155e-03,  1.5530e-03, -1.1213e-03,  1.9423e-03,
         2.5900e-03,  4.1559e-03,  3.4823e-03,  3.6744e-03,  5.8158e-03,
         2.2073e-03,  2.5238e-04, -1.0581e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7654e-03,  9.4744e-01,  1.2674e-03, -5.9119e-03, -4.8405e-03,
         3.0513e-03, -4.2119e-03, -9.6474e-04, -6.1506e-04,  7.1767e-04,
         1.8362e-03, -4.1425e-03,  1.6837e-04, -8.3336e-04,  1.0611e-03,
         9.3769e-04,  3.4278e-03, -1.7423e-03, -2.1458e-03,  1.0437e-03,
        -2.6487e-03,  1.1001e-03,  1.4710e-03,  2.0674e-03,  3.7938e-04,
         1.8081e-03, -7.1601e-04, -1.6844e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 5.9964e-03,  9.3667e-01, -1.0198e-02,  4.0068e-04, -2.9508e-03,
        -1.0930e-03,  1.7800e-03, -3.2773e-03, -2.0651e-03,  1.2012e-03,
         2.8339e-03,  1.0151e-03,  2.4764e-03, -1.7960e-03, -1.4720e-04,
        -8.0243e-05,  1.3288e-04,  1.0792e-03, -1.3819e-04, -1.5174e-03,
         3.6526e-04,  9.2352e-05, -7.6357e-04, -9.7186e-04,  1.1400e-03,
        -2.7084e-04, -1.2864e-03, -8.3331e-03, -9.9283e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3215e-03,  9.3509e-01,  2.0268e-02, -3.7208e-03, -4.3256e-03,
         2.3106e-03,  1.9118e-03, -9.0004e-04, -1.4612e-03,  7.8126e-04,
         1.2927e-03, -2.2567e-03, -1.8706e-03,  2.3332e-03,  2.4875e-03,
        -1.7536e-03,  3.8053e-03,  1.0706e-03,  2.8454e-03,  9.4607e-05,
         1.1471e-03,  2.7298e-03, -2.9323e-04, -3.9320e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.7873e-03,  9.3523e-01,  5.0735e-03, -3.3111e-03, -4.4213e-03,
         5.3281e-03,  2.3114e-04, -1.4480e-03, -8.7232e-04,  1.0758e-03,
         3.6440e-04, -1.5242e-03, -1.2430e-03, -8.9713e-04, -3.1326e-04,
         8.7989e-04, -1.9311e-03,  1.1169e-03, -3.4218e-03, -5.7345e-04,
         1.2429e-04,  6.0393e-04,  1.6636e-03, -1.0649e-05,  2.0132e-03,
        -2.9277e-04, -1.8382e-03,  5.4554e-04, -1.5990e-03, -5.4500e-04,
        -7.7470e-04,  2.9716e-04, -8.7457e-04, -1.2777e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.7672e-03,  9.5197e-01, -3.5009e-03, -2.3110e-03, -2.1567e-03,
        -4.5250e-03,  6.1436e-05, -4.2665e-04,  1.6403e-04,  9.7913e-05,
         7.5343e-04, -1.9018e-03, -6.2091e-03, -2.1375e-03, -5.1435e-03,
        -3.3386e-03, -2.3498e-03,  1.9294e-03, -9.2513e-04,  6.6638e-04,
         8.1586e-04,  2.8532e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.0097e-04,  9.5420e-01,  6.4451e-04,  4.6814e-03, -1.0622e-03,
        -3.4236e-03,  1.9659e-03, -8.6235e-04,  2.0675e-03,  6.4329e-03,
         8.8637e-03,  2.2467e-03, -1.0803e-03, -7.7120e-04,  1.5959e-03,
        -3.0929e-03, -4.7252e-03, -1.7784e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.5881e-03,  9.5562e-01, -8.0981e-03,  9.9773e-04, -7.4013e-04,
         2.7052e-03,  2.6836e-04,  2.2469e-03, -1.1685e-04,  2.0730e-03,
         1.2754e-03, -3.1354e-04, -6.4396e-04,  1.5443e-03, -3.0014e-03,
         7.8732e-05,  6.5932e-04, -2.7528e-03, -1.0014e-03, -3.9924e-03,
         1.6295e-03, -8.3998e-04,  1.8159e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.7028e-03,  9.4746e-01,  3.8440e-03,  1.4654e-03, -1.2455e-02,
         1.2199e-03,  3.7942e-04,  2.2273e-04, -6.7940e-03,  1.0965e-03,
         7.3582e-04,  2.4989e-03, -1.2558e-03, -1.9780e-03, -2.4023e-03,
         4.0527e-03, -5.4332e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6076e-02, -9.2727e-01, -2.2301e-02, -3.6618e-03,  1.9054e-03,
         1.3289e-03,  3.1470e-03,  1.2642e-04,  9.9869e-05,  1.8068e-04,
        -1.3092e-03,  9.8697e-04,  3.4243e-03, -5.2313e-04,  4.8054e-04,
        -4.4294e-03, -5.4006e-03,  7.3472e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.9211e-03,  9.1288e-01,  4.9529e-03,  1.0808e-03,  3.0035e-03,
         5.4653e-03, -2.7246e-03, -3.0363e-03, -2.5171e-03,  6.5881e-03,
        -4.1146e-03, -4.0558e-03, -1.0234e-02, -2.9456e-03, -1.6013e-03,
        -2.2850e-03,  1.8004e-03, -3.0678e-03,  3.9181e-03, -4.3445e-04,
         7.2748e-04,  3.5582e-04,  1.6975e-03, -3.5419e-04, -5.6101e-03,
        -9.6317e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8844e-02,  9.0350e-01, -5.1093e-03, -2.5876e-04,  3.2009e-03,
         4.9493e-03,  1.4085e-03,  5.8269e-03,  8.5736e-03, -5.5401e-04,
         7.3773e-04,  4.0788e-04,  7.0262e-05, -1.7417e-03, -7.8894e-03,
         1.2934e-03,  4.2439e-03,  2.4779e-03, -1.0776e-04, -3.9981e-04,
         3.3256e-04,  8.1167e-04,  2.0890e-04,  6.9449e-03, -1.4398e-03,
         3.0909e-03,  4.3511e-04,  2.8774e-03,  6.5657e-03,  3.3547e-03,
        -2.3449e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6313e-02,  9.1940e-01,  7.9401e-03,  6.8900e-03, -9.1386e-03,
         2.7584e-03, -2.2144e-03,  4.3002e-03, -3.1725e-03,  1.0755e-03,
         2.2269e-03, -7.0971e-04, -6.0939e-03,  4.4892e-04, -7.1774e-04,
        -6.9142e-04, -1.3431e-02,  2.4781e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0022,  0.9595, -0.0096, -0.0061,  0.0040, -0.0097, -0.0031,  0.0059,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 3.4545e-03,  9.4648e-01,  4.9383e-04, -3.5033e-03,  2.2792e-03,
         4.7938e-03, -3.0175e-03, -3.1197e-04,  1.5768e-03,  1.6423e-03,
        -1.6879e-03,  1.9201e-03,  2.0165e-03, -9.4207e-04, -1.0421e-04,
         1.5881e-03,  2.8606e-03, -2.4450e-03,  2.3122e-03,  3.6734e-04,
        -9.6382e-04, -8.4698e-05, -1.1829e-03, -3.0250e-04,  1.9822e-03,
        -2.5904e-05,  7.7558e-04,  6.5162e-04,  3.2411e-04, -7.3134e-04,
         6.0587e-04,  8.1822e-04, -4.9404e-04,  4.0311e-05,  1.8449e-03,
        -5.5413e-04, -1.9068e-03,  1.6063e-03,  1.3104e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.3452e-03,  9.3026e-01, -1.8757e-02, -4.6792e-03, -1.7400e-03,
         9.0624e-04, -1.0252e-03, -1.8736e-03, -1.9513e-03,  1.4603e-04,
        -6.3128e-04,  2.9680e-05,  6.0690e-03, -3.5466e-03, -1.2824e-03,
        -5.0805e-04,  4.9467e-04, -8.5507e-04, -7.6816e-04,  2.0968e-03,
        -1.7075e-03, -1.6496e-03,  6.9882e-04, -1.5872e-03, -1.0858e-03,
        -1.8906e-03, -4.6676e-03,  7.4393e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1049e-03,  9.6481e-01, -3.5097e-03,  4.0468e-03, -3.2005e-03,
        -1.9946e-03,  1.6587e-04,  1.8757e-03,  1.5412e-03,  4.0222e-03,
         3.9827e-03, -3.0673e-03,  2.6822e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6907e-03,  9.4892e-01,  4.3331e-03, -6.6579e-03, -6.6067e-03,
         3.7080e-03,  3.3821e-03, -8.0143e-04, -4.7026e-03,  4.0162e-03,
        -2.6390e-03, -5.2386e-03, -3.1648e-03, -1.3081e-03, -1.8298e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.6567e-03,  9.1707e-01,  7.5918e-03, -1.2053e-04,  3.0271e-03,
         8.0773e-03, -8.4534e-05, -3.0111e-04,  3.3260e-04,  1.9950e-04,
         1.0919e-03,  5.4182e-05, -5.7030e-03,  6.4371e-03,  2.3676e-03,
         5.6476e-04,  1.3736e-03, -5.4384e-04, -1.2174e-03, -1.2364e-03,
         3.4775e-03,  4.7261e-04,  2.4779e-03, -7.2891e-04, -1.0425e-03,
        -2.4585e-05,  5.3818e-04,  1.4307e-03, -1.9075e-03,  6.3171e-04,
         1.0801e-03,  3.8478e-04,  5.6939e-04,  3.1924e-03,  1.2791e-03,
         9.0235e-04, -1.7276e-03, -4.3078e-04, -6.8598e-04,  1.6176e-03,
         2.9796e-03,  5.4152e-04, -2.0127e-04, -6.6662e-04, -1.3538e-03,
        -8.8964e-05,  3.5117e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.4929e-03,  9.1999e-01,  7.8751e-03,  5.4195e-03,  4.0195e-03,
        -1.6816e-03, -1.2569e-04,  3.9422e-03, -4.1562e-03,  1.1920e-03,
         1.5567e-04,  1.4455e-03, -3.2102e-04, -5.9076e-04,  1.0206e-03,
        -6.4668e-04,  2.2267e-04, -2.4856e-03, -6.2330e-04, -1.0533e-03,
        -7.2315e-04,  5.6336e-04, -1.1855e-04, -8.5697e-04,  4.5152e-04,
         1.6336e-03,  1.3655e-02,  1.3520e-03,  1.0079e-03, -9.6628e-04,
         2.0674e-04,  6.8266e-04,  3.3921e-03,  5.8996e-05, -1.4889e-03,
        -3.1106e-04,  1.3499e-03,  5.1648e-04,  6.3862e-04,  3.6900e-03,
        -7.3046e-05, -1.8052e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5824e-02, -8.9732e-01,  1.3648e-02,  4.8043e-03, -3.6043e-04,
        -2.2608e-03, -1.5663e-02,  1.1150e-03,  1.6869e-03, -1.5621e-03,
         2.3653e-03,  1.8745e-03,  1.0793e-03,  2.0254e-04, -9.3367e-03,
         1.2236e-03, -3.4095e-03, -4.2218e-04,  3.4178e-04,  1.9328e-04,
         4.5357e-03, -2.1160e-04,  1.0028e-02, -4.2910e-04,  1.0075e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9266e-03,  9.3958e-01, -1.9233e-03, -8.0158e-04,  2.4130e-03,
        -2.5137e-03,  8.7481e-03, -8.3964e-05,  3.1347e-03,  1.0342e-05,
         1.4385e-03, -2.5254e-03,  6.7744e-04, -2.4748e-03,  4.0855e-03,
         3.5266e-03,  2.5263e-03,  7.6277e-03, -2.9038e-03,  6.9142e-04,
         2.6744e-04,  1.3610e-03,  1.7583e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0199e-04, -9.3623e-01,  5.6317e-03, -1.3513e-03, -6.0879e-03,
        -2.5662e-03,  1.7581e-03, -1.3516e-03,  3.3953e-03, -1.3830e-03,
        -4.3033e-03, -2.5212e-03, -7.7693e-03,  2.2467e-03,  3.1012e-03,
        -6.8791e-05, -1.0065e-03, -4.9758e-03,  2.6909e-03,  7.2493e-03,
         9.7563e-04, -2.2538e-03, -9.8084e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5009e-03,  9.1018e-01,  1.2257e-02, -3.6104e-03, -4.3326e-03,
         7.7554e-04,  2.1477e-03,  8.4347e-03,  7.5354e-03,  4.3831e-03,
        -1.7209e-02, -1.4633e-02,  1.1002e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.6499e-03, -9.5511e-01, -8.9938e-03,  3.0138e-03, -2.5992e-03,
        -9.7324e-05,  2.8935e-03,  3.8586e-03, -3.7107e-03, -3.0797e-03,
         3.8412e-03, -5.2880e-04, -2.1924e-03,  1.4285e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3418e-03,  9.5332e-01,  5.2252e-03, -1.7763e-03, -1.2512e-03,
        -8.7095e-04,  8.8607e-04,  6.5112e-04, -4.2340e-04,  3.9004e-03,
        -2.6292e-03, -3.1485e-03, -4.6259e-03, -4.0480e-04, -6.3454e-04,
         2.7089e-03, -1.0715e-03, -3.7110e-03,  3.4174e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 2.8870e-03,  9.4810e-01,  1.2519e-03, -4.2627e-03,  3.6374e-03,
         4.0877e-03,  1.3019e-03,  1.0448e-03, -4.1002e-04,  1.2993e-04,
         2.5561e-03,  1.3644e-03, -3.7580e-05,  3.4797e-03, -9.7285e-04,
         1.1242e-04, -1.3288e-03, -8.8519e-04,  4.6551e-03,  4.4250e-03,
        -1.8239e-03, -8.1175e-04, -2.4958e-03,  1.5157e-03,  7.7740e-05,
        -6.3413e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4366e-03,  9.3051e-01,  1.6492e-02,  1.8016e-03, -1.5742e-03,
        -5.5329e-03, -6.1702e-03,  5.2950e-04,  2.6970e-04, -1.6942e-02,
         3.8963e-03, -2.7184e-03,  2.2612e-03,  5.0311e-03, -4.5153e-04,
        -4.3861e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3357e-02,  8.9878e-01, -1.1501e-02,  3.4584e-02,  1.2922e-04,
         3.8532e-03,  5.4746e-03,  3.1236e-03,  1.0648e-03, -4.2769e-04,
        -3.8780e-04, -9.0903e-03, -1.7596e-03,  3.6065e-03,  6.2501e-03,
         3.8551e-03,  2.7535e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0405,  0.9029,  0.0206,  0.0049,  0.0092,  0.0012,  0.0023,  0.0089,
        -0.0094,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6293e-02,  8.9655e-01, -8.8548e-03,  6.3391e-03, -2.6628e-03,
        -1.1092e-03,  1.1990e-03, -2.6659e-03,  8.4560e-05, -5.2404e-04,
         7.5101e-04, -3.6822e-03,  1.0132e-03,  1.6746e-03, -2.4961e-03,
         1.9977e-03,  2.0938e-03,  8.5717e-04,  1.6159e-03,  1.4078e-04,
        -5.0616e-04,  2.0439e-03,  1.6064e-03,  7.2710e-04, -2.2299e-03,
        -1.3896e-03,  4.4785e-03, -1.1183e-03,  1.4602e-03,  9.9384e-04,
         1.3747e-03,  7.0135e-04,  1.3102e-03, -8.9530e-04,  1.7914e-03,
         1.1359e-03, -5.7205e-03, -3.8108e-04, -2.2169e-03, -6.9123e-04,
        -3.7838e-03, -3.8399e-03, -2.3586e-03, -9.3774e-04, -3.4569e-04,
        -3.3605e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3154e-02,  9.0662e-01, -4.7002e-03, -3.8826e-03, -1.8762e-03,
        -1.0809e-03,  2.5943e-03,  7.2155e-04, -9.0378e-04, -2.2790e-03,
        -3.4897e-03,  3.4046e-03,  5.0302e-04, -7.0067e-05, -1.8583e-03,
         2.2187e-03, -1.0333e-03,  1.7005e-04, -1.6584e-03,  1.0397e-03,
         8.5332e-04, -8.0230e-04, -6.9394e-03, -1.0591e-04, -4.1937e-04,
        -1.5886e-03, -3.1154e-03, -2.3831e-03, -3.3123e-03,  1.7224e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6639e-02, -9.2885e-01,  3.4756e-03, -3.0864e-05,  9.7098e-03,
         3.3644e-03,  2.7934e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8191e-03, -9.7677e-01, -3.0751e-03, -6.0657e-05,  1.2869e-03,
        -1.5460e-03, -2.4182e-03, -1.9829e-03,  1.2879e-03,  4.0693e-03,
        -2.6875e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2089e-03,  9.2775e-01,  7.8748e-03,  9.4734e-03,  7.7719e-04,
         2.2984e-03, -5.8072e-03, -6.0822e-03,  8.5035e-03,  6.1994e-04,
        -4.2057e-03,  3.5494e-03,  1.5976e-02, -1.8699e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.2217e-03,  9.3000e-01, -1.5743e-04,  6.8036e-04,  2.8040e-03,
        -5.3371e-04,  1.7971e-03,  3.4970e-03,  2.8386e-03,  2.6185e-03,
         2.0997e-03, -1.5488e-03, -1.9054e-03, -5.8161e-03,  9.5627e-05,
         1.0603e-03, -3.2453e-05,  8.9883e-04,  9.3334e-04,  3.8917e-03,
        -8.1402e-03, -1.1678e-04, -3.5266e-03, -7.8768e-04, -1.9296e-03,
        -1.3150e-03, -2.6178e-05, -1.0880e-04, -7.9860e-04,  4.9251e-03,
        -1.7458e-03,  1.9960e-03,  4.9013e-04,  3.6584e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4668e-02,  9.1064e-01, -1.3335e-02,  1.1746e-02, -5.4649e-06,
         6.3446e-04, -3.1659e-03, -5.7523e-03, -9.8237e-04,  2.4175e-04,
        -1.7628e-03,  2.3111e-04, -3.4551e-03, -1.1933e-03,  6.8776e-03,
        -3.1413e-03, -4.8542e-03, -1.3589e-03, -1.7680e-03, -7.9868e-04,
        -6.2578e-04, -3.9514e-04, -3.6336e-03, -2.2603e-03,  3.4915e-04,
        -1.9267e-03, -2.6851e-03, -1.5101e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.8036e-04,  9.2774e-01, -9.3374e-03,  1.8656e-03,  1.7803e-03,
         3.3834e-03, -3.0175e-03,  2.4456e-03, -2.0402e-03,  1.0741e-03,
        -3.2273e-04, -2.6266e-03,  6.2998e-03,  5.3796e-03,  4.2893e-03,
        -1.0771e-03, -4.6685e-03,  4.4202e-03,  2.4875e-03, -1.3872e-04,
        -7.3457e-05,  1.2350e-03, -5.4534e-03,  8.2650e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-8.7774e-03, -9.1340e-01, -4.8919e-03,  9.9277e-04, -4.4898e-05,
        -3.2384e-03,  1.0030e-05,  4.3162e-03, -4.3168e-03,  8.4055e-04,
         1.2608e-02,  4.5089e-03, -2.7792e-04, -3.3408e-03, -1.1353e-05,
        -3.7124e-03,  8.1958e-03,  1.1761e-03, -2.7461e-04, -2.7903e-03,
        -4.2800e-04,  3.5120e-04, -1.2318e-03, -1.3966e-03, -7.7600e-05,
        -2.7876e-03,  4.8963e-03,  1.4968e-03,  2.7530e-03,  2.3383e-03,
        -4.5134e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0212,  0.9463,  0.0041,  0.0050,  0.0019, -0.0037, -0.0050, -0.0012,
         0.0023,  0.0020,  0.0029,  0.0045,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0044e-02,  9.3682e-01, -2.1432e-03, -1.0144e-02,  1.0071e-02,
        -3.4524e-03, -3.2883e-03, -1.7250e-03, -2.8187e-03,  2.3664e-04,
        -2.0940e-03,  3.4992e-03, -1.4516e-03, -1.5987e-03, -1.0612e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.2673e-03,  9.1182e-01,  1.5762e-02, -4.6989e-04, -9.4677e-04,
         5.1749e-04,  1.5447e-03, -8.4289e-04,  1.8129e-04, -3.1622e-04,
        -6.4531e-03,  5.9718e-05, -7.9929e-03,  3.1835e-04,  6.1314e-04,
        -6.3547e-04, -1.4654e-03, -2.3024e-03,  4.0142e-03, -1.6269e-03,
        -1.0149e-04, -4.0235e-04, -4.2705e-03,  4.5493e-04,  1.0411e-02,
         1.2338e-03, -3.8165e-04, -2.3200e-03, -2.9090e-03,  1.4274e-03,
         9.3303e-04,  8.4017e-03,  3.6026e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7288e-02,  9.4478e-01,  3.6987e-04, -5.3327e-03,  1.6055e-03,
        -9.5518e-04, -7.6978e-06, -9.1976e-04,  1.7124e-03, -4.9444e-03,
        -3.2376e-03, -4.7746e-03, -3.6527e-03, -1.3225e-03, -8.1114e-03,
         9.8121e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5445e-03,  9.6987e-01, -4.1161e-03,  8.6285e-03,  8.2213e-04,
         3.4035e-03, -4.6539e-03, -3.9850e-03, -9.7195e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([0.0163, 0.9131, 0.0026, 0.0136, 0.0544, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3435e-03, -9.3483e-01,  1.8719e-02,  3.7243e-04,  3.1343e-03,
         3.6524e-03,  2.7687e-03, -1.9132e-03,  2.6339e-04, -3.6708e-03,
        -4.1901e-03, -7.7094e-04,  1.0817e-03, -2.6120e-03,  9.0155e-05,
        -6.4582e-04,  1.9402e-04,  1.6131e-06, -1.6060e-03, -1.5562e-03,
        -1.3056e-04, -3.2172e-03,  1.0407e-03, -2.9913e-03, -8.9936e-04,
        -1.4785e-03, -1.8272e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0052, -0.9412,  0.0052,  0.0073,  0.0136, -0.0042, -0.0233,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1257e-02, -8.9302e-01, -7.9555e-03, -4.2529e-03, -2.5461e-03,
        -3.1648e-03,  3.9270e-03, -3.0579e-03,  2.8515e-03,  1.0187e-03,
         7.7345e-04,  2.6883e-03, -4.0261e-03,  2.8197e-03, -3.4139e-03,
        -2.7756e-03, -7.7356e-04,  3.4388e-03, -3.3512e-03, -1.9800e-03,
         2.5003e-03, -8.4059e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.3433e-03,  9.3677e-01,  3.3532e-03,  3.5736e-03,  4.4335e-03,
         4.0748e-03, -9.6239e-04,  3.5651e-04, -3.1245e-05, -1.0881e-02,
         4.7420e-03,  1.7740e-03, -6.9942e-03,  5.3532e-04, -2.0134e-03,
        -6.1997e-04, -1.0433e-03, -6.8782e-03, -4.6217e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.6797e-03,  9.3001e-01,  9.3730e-03,  3.8715e-03,  3.4717e-03,
         2.9991e-04, -4.5555e-03,  3.3342e-03,  3.5217e-03,  1.3373e-03,
        -7.1271e-04,  6.5567e-04, -4.0080e-04,  2.9548e-04,  4.1346e-03,
         9.3808e-04, -3.2824e-03,  1.6208e-03, -1.1796e-03, -8.7669e-04,
         1.7957e-03,  9.7476e-04,  1.7256e-03, -3.6977e-04, -2.9153e-03,
         7.9528e-04,  6.3026e-04, -7.2141e-05, -1.6151e-03, -5.3280e-04,
         7.0265e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 6.5703e-03, -9.0387e-01, -2.4706e-02, -8.5244e-03,  4.1658e-03,
         5.6290e-04, -2.0546e-03, -2.0150e-03,  2.9035e-03,  5.8562e-04,
        -1.1762e-03, -1.0042e-03,  4.6709e-03, -9.8201e-03, -2.8165e-03,
         3.5000e-03,  7.3374e-03, -8.0681e-04, -2.0817e-03, -6.2642e-04,
        -6.2458e-04,  4.6879e-04,  2.6173e-04,  1.0043e-04,  1.0673e-03,
         6.8955e-04, -2.4507e-04,  2.4561e-03,  3.4367e-04,  2.0454e-03,
        -8.0528e-04, -1.0942e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.5341e-03,  9.4189e-01, -2.6494e-03, -2.4993e-03,  1.2581e-04,
         9.1628e-04, -3.3481e-03,  4.2433e-03,  8.6224e-03,  7.7985e-04,
         4.0207e-05,  3.8253e-03, -9.8363e-04, -3.0491e-03,  2.8248e-03,
         2.6688e-03,  6.2782e-03, -2.5887e-05, -3.5374e-03,  4.5258e-05,
         2.3872e-04,  2.8785e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.7039e-03,  9.1777e-01, -6.7709e-03, -1.3854e-03, -6.4105e-05,
         3.0772e-03,  1.8056e-03,  2.0619e-04, -1.2255e-03,  1.5170e-03,
         2.0639e-03,  3.0084e-04, -1.8204e-03,  6.4349e-04,  1.2894e-03,
         2.6824e-03, -1.9065e-03,  1.8611e-02, -4.0558e-03,  4.4236e-03,
        -1.1777e-04,  2.5618e-04, -8.0029e-05, -1.9215e-03, -8.1181e-04,
         2.2655e-04, -1.0786e-03, -5.1386e-03, -3.6563e-04,  1.1683e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2848e-02,  9.4709e-01, -1.0774e-02,  1.0592e-03,  1.3453e-04,
        -6.1244e-03, -3.6485e-03, -1.2572e-03, -2.2105e-03, -6.7626e-04,
         1.9092e-03,  5.0809e-03,  5.2330e-04, -1.9293e-04, -8.8811e-04,
         1.8696e-03,  1.3159e-04, -3.5847e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0622e-03,  9.2876e-01, -1.6385e-03,  2.2063e-04, -5.6699e-03,
        -1.8177e-03, -8.4493e-04,  1.3228e-03,  4.9315e-04,  2.9810e-03,
         1.5098e-03,  5.8103e-04,  3.9157e-04, -8.4340e-04,  7.0988e-04,
        -3.3406e-03,  8.5227e-04,  3.7867e-04, -5.6972e-03,  3.1355e-04,
        -1.8066e-03,  1.2418e-04,  3.7214e-03,  2.2924e-03,  7.2314e-04,
        -1.1322e-03, -4.9388e-03, -9.0158e-04, -1.7270e-03, -3.8458e-04,
         5.6149e-04, -1.9071e-03, -5.4907e-04, -4.0154e-03, -1.3059e-03,
        -2.4744e-03,  1.2007e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.0382e-03, -9.5206e-01,  6.0337e-03, -2.3042e-03, -7.4186e-04,
        -3.2218e-04, -2.0414e-03,  3.9580e-03,  9.2305e-05, -2.1500e-03,
        -2.2711e-03,  2.2887e-03, -1.0096e-03, -1.0502e-03, -2.5250e-03,
        -5.2263e-04,  1.2407e-03, -1.3142e-03,  3.5930e-05, -2.4247e-03,
         1.6764e-03, -1.0118e-03, -3.7225e-03,  2.1607e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9875e-03,  9.3014e-01, -9.7785e-03, -5.8854e-03, -2.6971e-03,
        -6.8785e-04, -2.4164e-04, -1.7084e-03, -9.4401e-04, -1.9708e-03,
        -6.0026e-04, -9.3993e-04, -7.2771e-03, -3.8687e-04,  1.4015e-03,
         1.1243e-04, -2.6966e-03, -2.1904e-03, -1.1657e-03, -5.9977e-04,
         1.4098e-04, -2.2932e-03,  4.9348e-04, -1.4992e-03,  9.8343e-05,
        -1.0983e-03, -1.1839e-03, -5.3185e-04,  2.1516e-03, -4.5950e-04,
         8.4212e-05, -1.0838e-03,  5.4415e-04,  7.9327e-04, -5.4762e-04,
        -1.1875e-03, -1.1835e-03, -3.1208e-04,  3.7649e-04, -1.1660e-03,
         1.1830e-03, -1.9306e-03,  1.0912e-03, -1.4348e-03,  7.4403e-04,
         1.9720e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3882e-02,  9.2525e-01, -5.2627e-03, -8.9828e-03, -4.3114e-03,
         2.5423e-03, -1.1451e-02, -1.1201e-03, -3.0261e-03, -1.7731e-03,
        -2.2328e-04,  3.7843e-04,  6.8627e-04, -1.7333e-03,  4.6014e-04,
         2.0966e-03, -2.7092e-03,  2.1396e-03,  3.2511e-03, -2.7159e-03,
        -1.0396e-03,  3.2002e-04,  4.6500e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3607e-02,  9.4977e-01,  7.8703e-03, -3.8367e-03,  8.7361e-03,
         3.2206e-03,  1.2692e-03,  1.2672e-03,  5.9449e-03, -4.4371e-03,
         3.7649e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0113e-02,  9.2186e-01, -3.8589e-03,  1.6915e-03,  8.1988e-03,
         2.8120e-03,  6.5265e-04, -6.4451e-03,  6.4970e-03,  1.1924e-03,
        -1.9377e-03,  9.4540e-04, -3.2300e-03, -1.0570e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2042e-02,  9.2802e-01,  8.1725e-03,  6.1297e-03, -1.1530e-02,
         2.9420e-03, -9.3265e-03,  3.3574e-03,  2.1729e-05, -1.0017e-02,
        -3.6491e-03, -4.7958e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3938e-03,  9.2704e-01,  1.9031e-03,  3.3979e-03, -2.5877e-03,
         1.3028e-03, -5.0805e-04,  6.6021e-04,  7.4679e-04,  3.4466e-03,
        -9.3440e-04, -2.0910e-03, -7.6069e-04,  2.1949e-03, -1.0246e-03,
        -1.6616e-05, -1.7202e-04,  2.1155e-03,  9.5129e-04, -4.1384e-03,
         7.8457e-04,  1.3702e-03,  2.6587e-03, -2.0544e-03,  4.7495e-04,
         1.1798e-03,  6.2634e-04,  1.9823e-04, -1.2379e-03,  1.4490e-03,
         2.8142e-03,  3.1890e-03,  1.7008e-03,  2.3087e-03,  2.0954e-03,
        -1.4498e-03,  4.4960e-04,  8.8233e-04,  2.1536e-03, -1.5415e-03,
         4.9207e-04,  5.4352e-04, -1.2262e-03,  3.2816e-04, -8.9940e-04,
        -1.8396e-04,  5.4588e-04, -3.4393e-04, -2.5138e-05, -3.7107e-04,
        -2.9888e-04,  1.1600e-03, -1.1698e-03, -7.4588e-04,  2.6559e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-1.3969e-02,  9.3930e-01, -2.2433e-03, -5.2256e-03,  1.3904e-03,
         4.8481e-03, -3.3171e-03,  2.2274e-04, -4.8090e-03, -1.3067e-03,
        -7.5707e-03,  1.6920e-03,  2.4499e-03,  1.0833e-03, -9.8876e-04,
         8.7190e-04, -1.4981e-03,  8.4483e-04,  2.7318e-03, -8.3462e-04,
        -4.6075e-04,  5.2250e-04, -1.8227e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.5599e-03, -9.5301e-01, -1.2283e-02,  7.1126e-04,  1.4737e-04,
         1.1306e-04, -3.3215e-04,  1.1389e-03, -3.8546e-04,  2.9604e-03,
        -5.7725e-04,  3.0584e-03,  1.6563e-03, -1.5148e-03,  3.5221e-03,
         1.3692e-03,  6.0863e-03,  1.5778e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8697e-02,  8.9581e-01, -7.5306e-03, -4.4467e-03,  7.4892e-03,
        -1.0777e-03,  1.7975e-02,  4.1504e-03,  1.5282e-03, -5.6545e-03,
        -6.5324e-03, -3.2418e-03,  1.3701e-02, -2.1867e-03, -4.2338e-03,
         4.3830e-05, -5.1959e-04,  5.1760e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0771e-03,  9.4869e-01,  1.4688e-03, -1.8024e-03, -8.8276e-04,
         1.0209e-02,  9.3945e-03, -3.6664e-03,  4.2554e-03, -3.1821e-05,
         1.4307e-03,  3.7877e-03,  3.0935e-03,  6.3575e-03, -2.8506e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1242e-02,  9.0780e-01,  8.6088e-03, -3.0773e-03, -3.1866e-04,
        -2.8492e-03,  4.1977e-04,  9.4120e-04, -1.8936e-03, -1.9925e-03,
         1.8307e-03, -4.0520e-03, -2.5479e-04,  6.0061e-05, -3.5537e-03,
         6.7016e-03, -9.2654e-03, -2.3605e-03, -1.3617e-03,  5.4721e-03,
         1.3704e-03,  1.0798e-03, -8.2912e-04, -1.4243e-03, -4.3839e-03,
        -3.9091e-04, -2.2076e-04, -5.1357e-04, -2.8621e-04,  1.0358e-03,
        -2.4982e-03,  1.9127e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.8486e-03,  9.3022e-01,  9.1556e-03,  6.1873e-03,  5.2955e-03,
         4.8078e-04,  7.5390e-04,  1.1676e-03, -3.3452e-03, -2.7429e-04,
         4.4034e-03, -8.3824e-04, -3.7042e-03,  1.6307e-03, -8.3780e-04,
        -1.2641e-03, -5.8853e-04,  2.5021e-04, -3.7016e-03,  1.3428e-03,
        -6.2619e-04,  2.2807e-03, -1.6596e-03, -2.4373e-03,  6.8286e-04,
         9.4315e-04,  3.1864e-03,  1.6838e-03,  4.1269e-04, -7.3456e-04,
        -2.6988e-04,  2.9694e-04, -3.4987e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2829e-03, -9.4046e-01,  5.7837e-04,  3.7614e-04, -1.6509e-03,
         2.5635e-04, -2.0460e-03, -5.1149e-04, -2.4433e-03, -1.8641e-03,
        -1.3712e-03,  1.6891e-03,  9.8181e-04,  5.7919e-03,  2.3344e-04,
         7.8434e-05, -2.4015e-03,  4.7025e-04, -1.5110e-04, -6.6326e-05,
         2.0500e-04, -4.1484e-04, -3.5201e-04, -2.7130e-03,  7.7718e-04,
        -1.4224e-03, -7.7478e-04, -6.4131e-04,  1.1885e-04,  1.2082e-03,
        -9.0760e-04,  2.4252e-04, -1.1366e-03,  7.2749e-04, -1.8352e-03,
        -2.3950e-03, -7.3705e-04, -5.2310e-04, -6.6729e-04, -7.6150e-04,
        -1.0090e-03,  7.8602e-04, -1.7278e-04, -1.5142e-03, -1.4158e-04,
        -1.2947e-03,  1.6064e-04, -9.3897e-04, -9.6316e-04,  8.6688e-04,
         1.2470e-04, -1.9793e-03, -8.5236e-04,  5.3430e-04, -9.0250e-04,
        -6.3631e-04, -4.0819e-04,  2.4488e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.6389e-03, -9.2437e-01,  2.1066e-02,  7.8247e-04,  1.1926e-03,
         5.5298e-03,  2.3984e-03, -1.6300e-05,  1.9007e-03,  2.3222e-03,
         1.3759e-03, -2.3696e-04, -7.7293e-04, -3.7239e-03, -4.5399e-03,
         4.4668e-04, -1.3208e-04, -9.1477e-06,  2.0701e-03,  1.0376e-03,
         2.1397e-03,  1.6237e-03,  2.0708e-03,  3.4098e-03,  2.7632e-03,
         5.4346e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8113e-02,  9.0577e-01, -8.2748e-03,  8.3132e-03, -6.9577e-03,
         2.0270e-03,  2.6619e-03,  1.6477e-03,  3.7942e-03, -7.4698e-04,
         4.3044e-03,  2.2801e-03,  6.5455e-03, -4.6402e-04,  2.4870e-03,
         2.3724e-04,  1.2989e-03,  1.8428e-03,  6.7136e-03, -5.4927e-04,
         6.8508e-04, -3.9301e-03, -3.2014e-03,  3.7900e-03, -3.3619e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3345e-02,  9.2578e-01,  1.2613e-03,  5.1468e-03,  5.5549e-04,
        -1.0934e-03, -1.9446e-04, -2.4027e-03,  1.5653e-03,  1.7935e-03,
         4.1123e-03, -1.2446e-04, -3.2993e-03,  8.6238e-05, -1.1628e-03,
        -6.7025e-04, -2.0049e-03, -7.4458e-05, -1.6823e-04,  1.0902e-03,
        -4.0676e-04,  3.4753e-04,  1.1307e-03, -2.1480e-03,  1.7123e-04,
         2.9965e-03,  4.5792e-04,  1.6729e-03, -1.1359e-03, -6.3913e-04,
         7.8668e-04, -2.1211e-03, -2.0362e-03, -8.9199e-04, -1.1585e-03,
        -3.8093e-03, -1.0202e-03, -1.3510e-03, -1.1271e-03,  9.5824e-04,
        -9.2401e-04, -1.0660e-03,  3.3297e-03,  8.7183e-04, -1.5086e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4991e-02,  9.2654e-01, -4.7948e-03,  6.8860e-03, -1.0287e-03,
        -1.8020e-03, -4.0386e-03,  3.6306e-03, -1.1793e-03, -4.3810e-03,
        -4.1560e-03,  2.4211e-03, -8.5228e-04, -1.8803e-03, -2.3463e-03,
         1.3533e-03, -2.6836e-03, -1.3227e-03, -5.9397e-05,  3.6841e-03,
        -2.5219e-04,  4.2450e-03, -5.4675e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.9524e-03,  9.1993e-01, -1.7362e-02, -1.4659e-04, -6.2882e-03,
        -1.4117e-03,  1.1290e-03,  6.2058e-03,  9.2325e-03,  5.9303e-05,
         7.2228e-03,  2.2062e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 2.3989e-03,  9.1895e-01,  1.3212e-02, -7.8161e-03, -8.0591e-03,
        -4.0934e-03, -6.8543e-03, -1.4521e-03, -5.3151e-04, -4.7386e-03,
        -3.2045e-03, -2.0097e-04, -5.4204e-03,  4.6107e-03, -5.8823e-03,
        -1.7243e-03, -8.9560e-04, -1.8638e-03,  1.5400e-03, -3.8784e-03,
         8.8893e-05,  3.5676e-04,  2.2267e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.4646e-03,  9.5105e-01,  5.7023e-03, -6.5061e-04,  1.1432e-03,
        -5.1178e-04,  8.2016e-04,  1.7545e-03, -6.0342e-04,  9.4903e-04,
         1.6263e-03, -6.2603e-04,  2.3574e-03, -3.0511e-04,  6.2127e-04,
         1.6464e-03,  6.2793e-04,  9.3571e-04, -5.1844e-04,  6.3338e-04,
         4.4371e-03, -9.4436e-04, -5.5876e-05,  3.4521e-04,  1.6257e-03,
         1.8429e-04, -2.2148e-03, -5.4642e-04, -1.3638e-03,  8.5569e-04,
         8.0075e-04, -1.8075e-03,  4.2740e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.8127e-03,  9.4514e-01, -6.9932e-03, -5.5826e-04,  9.9072e-03,
        -1.9142e-03, -4.5625e-03,  2.3165e-03,  1.7602e-03,  2.6646e-04,
        -1.4974e-03, -1.8504e-03,  7.4579e-04,  2.7999e-03, -2.4170e-03,
        -2.9618e-03,  8.0534e-04, -8.7173e-04,  2.8222e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.6115e-03,  9.4197e-01, -2.8172e-03, -5.8141e-03, -4.1856e-03,
         7.9476e-04,  1.2854e-03,  2.1424e-04,  2.3102e-03, -5.2576e-03,
        -5.1790e-03,  8.8993e-03, -4.3609e-03,  7.3042e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4420e-03,  9.3310e-01, -4.0564e-03, -2.4464e-03,  8.2936e-04,
         6.3969e-03,  3.2549e-03, -7.7171e-04, -4.6135e-04, -4.6435e-04,
        -3.0890e-03, -4.9918e-04, -6.9546e-03,  4.5425e-03,  1.8716e-04,
        -2.1374e-03,  1.1473e-02, -4.5167e-04, -1.5009e-03, -8.9562e-04,
         1.3069e-02,  1.9801e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9349e-02,  9.2216e-01, -1.7051e-03, -6.6932e-03, -1.6459e-03,
        -1.9824e-03, -3.8039e-03, -8.7852e-03, -4.7815e-03,  4.0316e-04,
        -1.5790e-03,  1.0047e-03, -1.7674e-03, -2.2784e-04, -3.1119e-03,
         2.9790e-03,  1.4137e-03,  1.6869e-03,  1.7008e-03,  2.1060e-05,
        -1.2403e-03,  1.4143e-03, -3.2221e-04, -6.2169e-04, -1.2825e-03,
        -1.0357e-03,  1.5631e-03, -2.3363e-05,  1.8365e-03,  3.8584e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2830e-03,  9.2056e-01,  3.9545e-03, -1.6337e-03, -4.7298e-04,
         4.6348e-03, -3.9350e-03, -1.4008e-02,  6.4467e-03,  5.6279e-03,
        -1.1898e-02,  1.4705e-03,  7.3677e-04,  1.9464e-03,  1.9924e-03,
        -3.8769e-03, -1.2524e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.5606e-03,  9.2873e-01, -6.1511e-04,  1.3557e-02, -3.6945e-03,
        -1.4998e-03, -6.2858e-03,  1.2214e-02, -1.4390e-03, -1.8449e-02,
         8.9599e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6777e-02,  9.2994e-01,  5.4077e-03,  1.2610e-03, -2.2085e-04,
         9.8807e-03,  6.0191e-04, -6.7081e-04, -5.5042e-03,  2.5060e-03,
        -1.5247e-03,  2.6168e-03,  1.0239e-03,  1.1583e-02, -6.8043e-03,
         1.1514e-03,  2.5274e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1527e-03,  9.4734e-01, -7.8778e-03,  3.4119e-03, -2.8055e-03,
        -1.3619e-04,  1.3945e-03, -9.7085e-04, -1.1693e-03, -7.6810e-04,
        -3.4332e-03,  2.5167e-03, -3.0860e-03, -5.1055e-03, -3.1551e-03,
        -3.7186e-04,  1.1855e-03, -1.8289e-03, -7.8871e-04,  1.6887e-03,
        -6.4979e-04, -1.7634e-03,  7.2196e-04, -7.6701e-04,  3.4398e-03,
        -4.7245e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9819e-02,  8.7310e-01,  1.4603e-02,  1.5965e-03,  1.3536e-04,
         3.1817e-03,  2.7839e-03,  6.1310e-04, -4.5001e-03,  2.3718e-03,
        -2.3081e-03, -2.9197e-03,  3.8190e-04,  1.3429e-03, -4.1665e-03,
        -2.5650e-03, -3.9663e-04,  1.3353e-04, -2.3260e-03, -2.1911e-03,
        -5.2383e-03,  5.9882e-06, -1.0877e-03,  6.9435e-04, -1.8952e-03,
        -1.1663e-03,  4.6039e-03, -1.4870e-03, -2.1249e-03, -2.6458e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2106e-02,  9.4785e-01, -2.4599e-03,  1.3729e-03,  8.5519e-03,
         1.1857e-03,  3.7118e-03, -1.9368e-03,  1.3275e-03, -4.3643e-03,
         3.6247e-04,  2.9610e-03, -7.0719e-03,  1.0718e-03, -2.2095e-06,
         6.5780e-04, -2.2526e-04,  8.4299e-04, -1.9370e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 1.4584e-02,  9.2471e-01, -7.0584e-04,  1.3389e-02,  2.9163e-03,
        -1.1047e-03,  1.1710e-02, -4.2592e-04, -3.3526e-03,  2.2761e-03,
        -2.7472e-04,  9.5151e-04, -7.1290e-04, -5.1041e-04,  2.7150e-03,
        -3.3317e-03,  1.4404e-03,  2.3088e-03, -7.1247e-05,  1.0347e-03,
         1.4596e-03, -3.1588e-05,  1.0034e-03, -5.8048e-03, -1.5048e-03,
         1.2385e-03,  4.3020e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.6260e-03,  9.4977e-01,  4.2107e-03, -1.7920e-03,  3.2147e-03,
        -2.1260e-04,  1.0638e-02,  1.4198e-03,  7.2452e-03, -2.2336e-03,
        -2.4112e-03,  3.7707e-03, -5.5918e-04, -5.8968e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3741e-04,  9.4590e-01,  4.1612e-03, -1.7195e-03,  1.3943e-03,
        -6.3426e-03, -2.0382e-03,  4.9895e-04,  1.3807e-03,  1.6865e-03,
        -1.4982e-03, -6.1553e-04, -6.4909e-03,  1.4612e-03,  1.3957e-05,
         1.5003e-04,  4.5577e-04, -2.6160e-04,  6.1713e-03,  4.6961e-03,
        -2.3895e-03, -8.2212e-04,  6.7035e-04, -1.7383e-04, -6.2968e-03,
         2.3750e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9357e-03,  9.5123e-01, -5.3961e-04, -7.7180e-03, -7.4800e-04,
        -8.1295e-03,  2.7966e-03, -2.8606e-03, -6.3244e-03,  1.2945e-03,
         3.3091e-03,  6.4910e-03,  1.0661e-03, -2.5588e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.7693e-04,  9.4439e-01, -1.2989e-03,  1.1398e-04, -2.8229e-03,
         1.1134e-03, -2.4661e-03, -7.6434e-04,  8.0908e-04,  2.9057e-05,
         5.4330e-04, -1.1899e-03,  3.6816e-03, -3.2579e-03,  3.6227e-03,
         7.5421e-04, -5.4120e-04,  7.7429e-04,  1.7892e-04,  5.6005e-03,
        -1.1867e-03,  1.5497e-03,  5.7142e-04,  4.8678e-04,  5.9724e-04,
         2.7181e-03,  4.4619e-03,  2.2331e-03,  1.9967e-03, -3.9690e-04,
        -7.0436e-03,  2.0246e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.4943e-03,  9.4030e-01, -6.2238e-03, -7.6111e-03, -2.5261e-04,
         1.3962e-03, -1.9093e-03, -1.0941e-04,  2.9837e-03, -3.8992e-04,
        -3.2129e-03, -6.0463e-04,  3.2508e-03, -2.1155e-04,  8.8499e-04,
         4.5969e-04,  5.3386e-04, -2.9632e-04, -7.2237e-03,  1.8511e-03,
         1.8799e-03,  1.0614e-03,  7.2699e-04,  1.6655e-03, -9.4692e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0354,  0.9417, -0.0035, -0.0033,  0.0078, -0.0082,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7282e-02,  9.2158e-01,  8.3657e-03, -6.5539e-03,  4.0082e-03,
        -4.3952e-03,  1.1811e-03,  1.1379e-03,  7.4093e-04, -7.4006e-04,
        -3.5333e-03, -1.2489e-03, -4.5119e-08, -4.3518e-03,  2.2803e-04,
         1.5849e-05, -8.0501e-04,  1.3901e-03, -6.3277e-04,  2.9510e-03,
        -7.9900e-04, -1.6447e-03, -3.5623e-04,  3.8414e-04,  1.1784e-03,
        -5.7120e-04, -6.3356e-05, -5.0218e-04, -2.8346e-04,  4.2273e-03,
         1.6817e-04, -8.8286e-04, -3.6299e-03,  4.7832e-04,  1.6945e-03,
        -8.0492e-04,  1.1926e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6514e-03, -9.2187e-01, -3.8371e-04, -2.5084e-02, -1.5079e-04,
        -9.1731e-04,  2.9588e-03, -3.0034e-03,  2.2108e-03, -2.5148e-04,
        -1.7235e-03, -1.2319e-03, -3.9921e-03,  1.0516e-03, -9.0826e-04,
         9.2000e-04,  7.2291e-04, -5.4037e-04,  1.5339e-03, -1.7384e-03,
         8.3309e-06, -1.3267e-04, -3.4916e-03, -5.3683e-04,  8.0937e-04,
         5.3801e-03,  3.3417e-03,  1.5374e-03, -6.0206e-04, -6.4746e-05,
         5.1988e-05,  1.2392e-03,  2.5314e-03, -1.6864e-03, -3.7412e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.4037e-03,  8.7745e-01, -5.4389e-02, -1.3965e-03,  7.2671e-04,
         3.7617e-03, -4.6870e-03, -2.8833e-03, -4.4262e-03, -5.5361e-03,
         2.3948e-03, -2.2950e-02, -4.7268e-03, -2.8094e-03, -5.4558e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.5636e-03,  9.4290e-01, -1.2321e-02, -1.1996e-03,  5.5386e-04,
        -2.4155e-03,  1.0283e-03, -8.6384e-04, -3.9890e-03,  2.6594e-04,
         5.9185e-04, -1.0218e-03, -1.6006e-03, -1.4310e-03, -3.5818e-05,
        -1.9844e-04, -5.6987e-04,  1.3366e-03,  1.2489e-03,  2.2953e-04,
         1.2819e-03,  2.0033e-03, -4.5427e-04,  1.9518e-03,  4.1420e-03,
        -1.5231e-03,  9.5318e-04, -4.3856e-04,  8.0810e-05,  5.2689e-04,
        -7.5154e-04, -6.8818e-04,  1.8356e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1525e-02, -9.0628e-01,  1.6882e-02,  1.2293e-03, -1.4030e-03,
         7.2673e-03, -1.2819e-03,  5.1252e-04, -1.5680e-03, -2.8042e-03,
         4.8959e-03,  5.2874e-03,  3.2840e-03,  2.6864e-03, -1.2263e-04,
         5.4148e-04,  8.3711e-05,  2.0788e-03, -3.5213e-04,  3.1057e-04,
         9.9751e-05, -4.3021e-04,  2.6774e-03, -2.0475e-03, -1.2729e-03,
        -6.2073e-03, -1.5406e-03, -3.2950e-03,  2.0312e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 7.3823e-03,  9.4314e-01,  1.2816e-02, -2.0404e-03, -8.3336e-03,
        -2.7096e-04, -2.3917e-03,  7.1169e-05,  1.3132e-04, -1.0242e-02,
         2.4932e-03,  5.4228e-04, -9.9993e-04, -3.3793e-04, -8.8127e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9120e-02, -9.2364e-01, -6.1760e-03,  3.8120e-03,  2.4944e-03,
        -3.2840e-03,  1.2442e-03,  1.1224e-03, -2.7310e-03,  9.4649e-04,
         2.0276e-03,  8.5563e-05,  7.1715e-04, -3.5362e-05, -7.4598e-04,
         9.1938e-04, -2.8626e-04,  9.0802e-06,  4.6861e-04, -1.5837e-04,
         1.5376e-03,  2.5123e-03,  1.3310e-03, -1.5458e-04,  4.8281e-06,
        -5.0283e-04, -2.2356e-04,  1.1988e-04,  3.8002e-03,  1.7560e-04,
        -5.2594e-04, -6.7730e-05,  1.7326e-03, -7.1958e-04,  1.2115e-03,
        -1.8553e-04,  1.4557e-04, -3.2047e-04, -4.8763e-04,  1.0673e-03,
        -3.1266e-04, -1.3426e-03, -3.2609e-05, -1.2863e-04,  6.5823e-04,
         8.0621e-04,  6.4350e-04, -2.0133e-03,  7.2085e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7749e-03, -9.1489e-01,  1.5584e-02,  2.3845e-04, -4.4475e-04,
        -2.9806e-03, -2.0033e-03,  3.9426e-03,  3.9646e-03,  1.5451e-03,
        -1.4450e-03, -4.9015e-04, -3.7954e-04, -1.3784e-03,  9.2666e-04,
         3.3786e-03, -1.4505e-02,  1.6321e-03, -2.6317e-03, -2.9916e-03,
         2.4883e-03,  1.9310e-04,  7.1022e-04, -1.1228e-04,  9.9522e-04,
         2.2657e-04,  4.1376e-03,  1.2133e-03, -5.9965e-04,  2.6632e-03,
        -5.3250e-04, -2.4058e-03,  8.9642e-04, -2.8102e-03, -1.8885e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0940e-04,  9.4524e-01, -6.5263e-03,  1.4800e-03,  3.2128e-03,
         5.3346e-03, -1.2201e-02,  2.0473e-03,  2.2705e-03, -2.7244e-03,
        -1.2108e-03, -1.4720e-03, -1.5353e-03, -2.9696e-05,  2.9988e-03,
         1.2850e-03,  3.8113e-03, -3.2948e-04, -6.0826e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0791e-02,  9.1279e-01, -3.0175e-03, -1.5587e-02,  2.1310e-03,
        -9.0762e-03, -3.1996e-03,  2.9140e-04, -4.1018e-03, -5.7802e-03,
         1.5645e-03, -4.2664e-03, -1.1852e-02,  5.6511e-04, -4.9854e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2122e-02,  9.3076e-01,  3.0946e-03,  2.6566e-03,  3.3860e-04,
        -2.9323e-04, -2.2325e-03, -2.0902e-04,  1.7633e-03, -1.0862e-03,
        -1.3145e-03,  4.9274e-04,  3.0865e-05,  3.1687e-03,  1.5033e-03,
        -8.7003e-04,  1.1593e-05, -1.5583e-03, -2.2220e-03,  6.4456e-03,
         6.3370e-03,  1.4897e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.6956e-03,  8.9738e-01, -1.4242e-02, -2.8749e-03, -1.0424e-04,
        -3.1179e-04, -5.3692e-03, -1.1224e-03, -1.0467e-02, -5.9315e-04,
         7.1948e-04, -3.6809e-03, -3.0298e-03, -2.8723e-03, -1.6578e-03,
        -1.0187e-03, -1.9702e-03, -1.0097e-02, -4.0959e-03, -3.5704e-03,
        -1.2904e-03, -1.0013e-04, -3.4452e-03, -4.4116e-03,  1.9597e-03,
        -3.0783e-04, -2.2637e-03,  2.8465e-03, -3.3518e-03,  6.1453e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0605,  0.8882,  0.0036,  0.0086,  0.0130,  0.0049,  0.0017,  0.0015,
         0.0042, -0.0019, -0.0118,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.5303e-03,  9.3877e-01,  1.1375e-02,  4.5403e-04, -2.8615e-03,
         2.0052e-03,  4.4879e-03, -2.7274e-03,  2.9859e-03,  1.4856e-03,
        -1.6087e-03,  1.9777e-03, -1.5347e-03, -1.6555e-03, -2.1182e-03,
         3.1691e-03,  1.3400e-03,  3.6795e-04, -6.9582e-04,  3.2788e-05,
         2.1070e-03, -8.7069e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.4447e-03,  9.2268e-01,  2.8023e-03,  4.8260e-03,  1.5720e-03,
        -1.2177e-02, -1.3785e-03,  5.7198e-03, -1.4259e-03, -5.5507e-05,
        -6.0168e-03,  2.1267e-04,  2.7338e-03,  6.2865e-05, -3.4469e-03,
        -2.9702e-04,  1.2654e-03, -5.5285e-04, -4.4921e-03,  2.9049e-04,
         2.0337e-04, -6.2759e-04, -5.3452e-04, -1.0829e-03, -4.8982e-04,
         1.9616e-04, -2.4830e-03, -1.2831e-04, -6.2182e-03,  3.4301e-05,
         2.3586e-03,  1.3076e-03, -2.8838e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0495e-02, -8.7138e-01, -4.3409e-02, -8.4388e-03, -2.6697e-03,
        -2.4718e-03, -9.3390e-04,  5.7263e-03,  5.2135e-03,  2.1852e-03,
        -1.5059e-03, -2.3923e-03, -1.4048e-03, -4.9906e-04,  2.9458e-03,
        -9.3434e-03, -1.1563e-03, -9.2243e-04, -1.2232e-03, -1.2845e-03,
        -6.1058e-03, -2.2182e-03, -8.8694e-04, -2.5755e-03, -1.7975e-03,
        -6.1276e-03,  2.2867e-04,  4.4598e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3311e-02,  9.1572e-01, -1.4843e-02, -9.1442e-03, -1.2478e-03,
         3.3372e-04, -3.8786e-03,  1.6596e-03, -6.7970e-03, -1.0445e-03,
        -5.0564e-03, -4.5185e-04, -4.1926e-03,  1.1762e-03, -4.6126e-04,
        -1.0448e-04, -1.9438e-03, -2.1162e-03, -5.1809e-04, -3.5132e-03,
        -9.7622e-04, -1.9786e-03,  4.8030e-03, -3.3198e-03,  1.4084e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-3.2319e-04,  9.3380e-01, -9.8690e-03,  3.1060e-03, -5.3807e-03,
         2.1922e-03, -8.2657e-04,  1.7036e-03, -3.3399e-04, -1.3371e-03,
         2.0976e-03,  6.0039e-03, -1.3760e-03, -2.1258e-03, -1.3756e-03,
        -2.4309e-03, -1.3955e-03,  1.9188e-03, -6.8454e-04, -3.6996e-04,
         1.0605e-04, -5.4127e-04, -2.0960e-03, -1.0623e-03, -1.4008e-03,
         1.6937e-03,  4.5305e-04, -2.3552e-03,  3.9021e-04,  5.6711e-05,
        -3.2108e-03, -8.1818e-04, -4.7079e-04, -8.0105e-04, -7.8032e-04,
         1.7528e-03, -1.2111e-03, -8.5840e-04,  9.6518e-05,  4.8958e-04,
         7.0747e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1472e-02,  9.2919e-01,  1.0111e-02, -1.1205e-02, -2.3059e-04,
         1.3331e-04, -2.7685e-03, -2.9188e-04,  2.4744e-04,  1.2521e-03,
        -2.0198e-03, -4.0219e-03,  2.4385e-04, -2.8047e-04,  1.1687e-03,
        -3.4745e-03,  1.5437e-03, -7.1914e-04, -1.5737e-03,  4.6805e-03,
        -1.5925e-03,  8.6924e-04, -2.4552e-04,  2.2612e-04,  4.3893e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3825e-03,  9.4080e-01,  7.7414e-03, -9.1609e-03,  2.7235e-03,
         4.8432e-04,  1.3434e-03,  2.1326e-03, -4.4331e-03,  4.6654e-03,
         4.0371e-06,  2.9664e-03,  5.8645e-03,  1.4114e-03, -2.3506e-03,
         7.1642e-04, -1.1687e-03,  1.3772e-03, -5.5552e-04, -3.1903e-04,
        -3.3974e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5536e-03,  9.3199e-01,  3.5306e-03, -2.4472e-03,  2.6454e-03,
         4.6485e-04, -2.4873e-03, -2.2793e-03, -2.0575e-03,  2.3479e-03,
        -5.9789e-03, -1.0509e-02, -2.5850e-03, -5.6214e-03, -1.2366e-03,
        -1.0174e-02, -4.2306e-03,  3.8590e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6457e-03,  9.3981e-01, -9.3515e-03,  3.2461e-03,  9.9652e-04,
         3.7664e-03,  9.2784e-04, -4.7757e-03,  5.6140e-04,  9.6652e-04,
        -1.2772e-03, -2.3741e-04,  3.5928e-04, -1.2406e-04,  2.1308e-03,
         1.7655e-03,  1.4599e-03, -2.2509e-04, -8.7447e-04,  1.1366e-03,
        -1.0516e-03, -7.6639e-04, -7.7794e-04, -3.4181e-05,  1.3553e-03,
         8.5577e-04,  1.0438e-03,  1.0259e-03, -1.9736e-03,  1.0933e-03,
         2.8017e-03,  3.5026e-04, -1.4728e-03,  1.3063e-03,  5.0368e-03,
         1.5633e-03, -6.1409e-04,  1.2349e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9118e-03,  9.6218e-01,  9.9625e-04,  6.7353e-04,  1.9723e-03,
        -1.9739e-03, -2.1350e-03, -5.3503e-04,  2.2430e-03,  1.3323e-03,
         3.9464e-04, -1.8905e-03, -1.5457e-03,  1.8632e-03,  1.2442e-03,
         1.8229e-03,  8.1297e-04, -4.2167e-04,  1.3160e-03, -8.7340e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0056,  0.9221, -0.0116, -0.0103, -0.0022, -0.0060,  0.0037,  0.0047,
         0.0015,  0.0034,  0.0128, -0.0018, -0.0051,  0.0093,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.9736e-03,  9.3397e-01,  4.6404e-03, -1.8716e-03, -5.7215e-04,
        -2.4855e-03, -1.3166e-03,  7.2034e-04, -5.4628e-04,  6.2076e-04,
        -9.0395e-03, -6.1592e-04,  6.2128e-04,  5.3364e-04,  8.1699e-04,
        -9.2896e-04, -4.4462e-04, -1.8457e-03,  5.3956e-04, -1.4471e-04,
        -1.4498e-03,  1.3919e-03,  6.2264e-04, -3.2151e-03,  7.2483e-04,
        -1.0645e-03,  1.5184e-03, -5.7798e-03, -1.6730e-04, -4.9241e-03,
         2.3803e-04,  8.6558e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1432e-02,  9.3534e-01, -4.1418e-03, -9.2098e-04, -3.4942e-03,
        -2.8154e-03, -3.2132e-03, -1.6645e-03, -1.2816e-04, -2.4832e-03,
        -1.6564e-04, -3.6353e-03, -2.6517e-03,  1.6152e-04,  1.6077e-03,
        -1.0452e-03, -1.4610e-03, -8.0911e-03,  1.8913e-04, -5.3619e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.9222e-04,  8.8936e-01, -7.5315e-03,  2.8960e-03,  3.5339e-03,
         2.6867e-02,  4.0536e-03,  4.7453e-03,  5.3516e-04,  1.5267e-03,
         1.7166e-03,  1.1032e-04, -1.1046e-03, -9.0029e-05, -6.4963e-04,
        -3.2460e-03, -8.4470e-04,  4.9371e-04, -1.6698e-04, -1.9476e-03,
         4.9665e-04,  2.0513e-03, -3.2970e-03, -2.9565e-04, -1.5838e-03,
        -2.5367e-03, -2.6833e-03, -2.6373e-03, -1.1892e-03,  1.5210e-03,
        -2.0766e-03,  1.9771e-03, -3.2971e-03,  4.5676e-03,  1.4383e-03,
         2.4358e-05, -1.6390e-03,  1.5573e-03, -1.4695e-03, -1.0803e-03,
         3.9792e-04, -6.4646e-04, -1.5412e-04, -4.4570e-03,  9.1099e-05,
        -1.3822e-03, -3.2419e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4803e-02,  9.4990e-01,  9.8794e-04, -1.3506e-03,  3.4551e-03,
        -6.3588e-04, -2.8866e-03, -1.7162e-03,  2.4504e-03, -4.8296e-03,
        -5.3030e-03, -2.8738e-03,  1.1886e-03, -1.5686e-03,  6.0508e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0908e-03,  9.1216e-01,  9.5694e-03,  3.7495e-03, -1.0376e-02,
        -3.0741e-03,  2.0029e-04,  1.3879e-03,  1.0189e-03,  1.1968e-03,
        -8.5101e-04, -8.6462e-04, -2.9375e-04,  4.0412e-03, -9.2119e-04,
        -1.2953e-03, -5.6953e-04, -6.5390e-04,  7.4903e-04, -1.4369e-03,
        -3.9088e-04, -8.0854e-04, -6.8408e-04,  3.4593e-04, -4.9472e-04,
        -2.3752e-03,  2.5215e-03,  1.1140e-02, -8.5676e-04,  5.2841e-04,
        -2.4634e-03, -8.7334e-04, -3.2200e-05, -6.4907e-04, -1.3764e-03,
         9.0676e-04, -8.5398e-04, -8.9676e-04, -3.4040e-03, -2.3199e-03,
        -6.6326e-04,  4.3185e-04, -5.8551e-05, -6.4063e-04, -3.4507e-04,
        -3.8238e-03,  2.3822e-03, -1.2323e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-1.1603e-02,  9.1233e-01, -3.6952e-03, -6.6658e-04,  2.7655e-03,
        -1.1106e-02,  3.6594e-03,  4.2290e-03,  7.1407e-05, -9.2041e-03,
        -1.3188e-03, -7.9419e-03,  1.9339e-03, -4.3438e-04, -4.7071e-03,
        -1.5671e-03, -7.7906e-03, -2.6025e-03,  1.4231e-03, -8.0102e-04,
         3.7873e-04,  1.7780e-03,  2.5590e-03,  3.1291e-03,  1.1350e-03,
        -1.1716e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0074e-03,  9.4446e-01,  1.0557e-02,  3.8282e-03,  8.1855e-04,
         1.0073e-03, -3.1726e-04,  3.1418e-03,  9.7046e-04,  6.3644e-04,
         5.8669e-03, -1.7279e-04,  1.3423e-03,  1.0432e-03, -3.7796e-03,
        -3.6713e-03,  6.7075e-03, -1.4866e-04,  4.8404e-03,  9.9641e-04,
         2.6863e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4535e-02,  8.8872e-01, -6.9848e-03, -1.0042e-02,  5.9316e-03,
        -4.3323e-03,  6.3678e-04, -1.2863e-03,  1.4051e-03,  8.5997e-05,
         1.3762e-06,  8.1903e-04,  8.1604e-04, -4.7608e-03, -2.7899e-03,
        -2.2287e-03,  1.7346e-03, -2.8370e-03,  5.0806e-03, -5.3409e-04,
        -1.7093e-03, -1.1198e-03, -8.8400e-03,  2.2401e-03,  3.6636e-03,
         1.0143e-03, -9.5647e-04,  2.2967e-03,  1.1173e-03, -1.1764e-04,
        -4.8660e-03,  5.0092e-04,  2.9964e-03,  1.6916e-05, -1.7387e-03,
        -1.2482e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5563e-03,  9.3358e-01, -1.0830e-02, -5.2313e-03, -4.5207e-03,
        -2.6950e-03, -1.6399e-03, -3.1968e-03,  8.2847e-03,  9.4137e-04,
        -5.6260e-05,  3.8726e-04, -9.7011e-04, -3.8443e-03, -1.7045e-03,
        -9.7514e-05, -1.1667e-03, -4.3112e-03,  7.5696e-03, -6.4195e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2056e-03,  9.3334e-01,  1.6035e-02, -2.2811e-04,  4.6445e-03,
        -9.8269e-05,  3.5242e-05, -5.3066e-03,  2.6138e-03,  9.8936e-04,
        -3.5929e-03, -3.4053e-03, -7.3045e-03, -8.7641e-04, -1.4829e-03,
        -2.4514e-04, -3.6000e-03, -2.7995e-03, -2.2936e-03, -8.6963e-04,
         2.0348e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.6988e-03,  9.3519e-01, -6.8813e-04,  9.7971e-03, -3.5114e-04,
        -1.3565e-03, -6.0280e-03, -1.1332e-03,  5.0053e-03,  1.1722e-02,
         1.8719e-03,  1.2513e-02,  8.6409e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4225e-02,  9.1736e-01,  5.4677e-03, -1.2151e-03,  2.0079e-03,
         1.6725e-03,  3.6625e-03,  1.4558e-03, -2.8804e-03, -3.3571e-04,
        -7.0005e-03, -3.5286e-03,  1.2852e-03, -4.9151e-04, -5.8180e-04,
        -2.4546e-03,  1.1391e-03,  1.3466e-04, -2.6385e-03, -1.7466e-03,
         2.5101e-03,  2.2459e-05,  3.6661e-04, -4.9282e-04, -1.0617e-04,
        -9.1504e-04, -2.3520e-03,  5.0679e-04,  1.5826e-03, -1.4586e-03,
        -8.0589e-04,  1.8922e-03,  1.9636e-03, -4.2413e-04, -5.8182e-04,
         2.3356e-03,  3.9646e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.9734e-03,  9.4252e-01, -7.5286e-03, -1.0307e-03,  2.8793e-03,
        -1.6623e-03,  4.4072e-04,  6.5495e-03,  1.8005e-03,  1.1023e-03,
        -3.8417e-04,  7.1542e-04,  9.0652e-04,  1.3799e-03, -6.4341e-03,
        -5.2087e-04,  2.5070e-03,  1.2053e-04, -1.3765e-05,  6.7218e-03,
         2.7018e-03, -7.3213e-04,  2.5853e-03,  2.7852e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2916e-03,  9.2082e-01,  3.9137e-03, -4.0420e-04,  1.5349e-03,
         3.8928e-04, -5.0657e-04, -4.3791e-03, -5.9601e-03,  6.3593e-03,
        -7.2988e-04, -3.8928e-03, -9.1623e-03,  1.8479e-03,  6.8989e-03,
        -1.4679e-03, -9.5980e-04,  1.2470e-03,  2.3689e-03,  1.4350e-03,
         1.0917e-03,  7.3351e-04,  1.8980e-03, -3.3856e-03, -1.3816e-03,
        -2.0744e-04, -1.9290e-03,  2.9942e-04, -3.0228e-03, -1.4084e-03,
        -2.7241e-03, -6.3446e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5663e-03,  9.4211e-01,  1.8599e-03, -5.1631e-03, -9.3643e-04,
        -8.8430e-04,  7.4587e-04, -1.7876e-03,  1.9709e-03, -8.9435e-04,
         4.3584e-04,  2.0039e-04, -1.6883e-03,  1.0410e-03, -1.3240e-03,
        -1.7057e-03,  1.6440e-03, -6.3961e-04,  4.0378e-04,  1.5921e-03,
         9.1792e-04,  2.6656e-03, -1.1516e-03, -2.2096e-03,  1.0924e-03,
         1.9813e-03,  4.8100e-03,  2.2971e-04,  1.8070e-03, -7.4515e-03,
         6.0892e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3079e-02,  9.1378e-01,  7.0068e-03,  1.5575e-03,  5.0085e-04,
        -7.7478e-04,  8.0840e-03, -1.1849e-02, -5.6291e-04, -2.1720e-03,
         9.0288e-04,  1.4342e-03,  1.8187e-03, -4.3225e-03, -8.2681e-04,
        -2.5535e-04, -6.0038e-04, -2.3663e-03,  1.6879e-03,  5.2842e-03,
        -2.1437e-03, -6.5921e-03, -1.2400e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9565e-02,  9.2120e-01,  4.6805e-03,  1.4020e-03, -3.0394e-03,
         6.1283e-04,  5.4415e-03,  1.6662e-03, -9.5489e-04,  4.9446e-04,
         3.1345e-03,  2.3131e-03, -1.6755e-03, -1.6570e-03,  1.2186e-03,
        -1.5973e-03, -1.4732e-03,  2.0998e-05, -3.5292e-04, -2.1777e-04,
        -2.8946e-04, -8.1775e-05,  6.1799e-03,  9.8993e-04,  2.6238e-03,
         2.3360e-03,  1.3081e-02,  1.7016e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-8.2344e-04,  9.5044e-01,  6.9729e-03, -2.7446e-03, -2.5145e-03,
        -3.5646e-03,  2.0694e-03, -1.7206e-03,  1.5431e-03, -1.2353e-03,
         1.6579e-03,  2.0167e-04, -1.3331e-03,  3.1532e-04,  2.0829e-04,
         1.4948e-04,  2.3170e-03,  3.9633e-03, -1.1826e-03,  3.8286e-03,
         9.4479e-04, -4.5778e-04, -3.2914e-03,  6.5030e-04,  1.2052e-03,
        -4.9746e-04,  2.0559e-03, -9.5026e-04, -1.1577e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.6569e-03,  9.4414e-01, -2.1609e-03, -6.5544e-03,  5.0846e-03,
         5.1586e-04,  1.0320e-03, -1.8369e-03, -2.1341e-03,  1.2671e-03,
         2.5543e-03, -2.5884e-03, -7.7668e-04, -5.7081e-04, -5.9976e-04,
         3.3819e-05,  9.5984e-04,  5.5932e-04,  8.2738e-03, -1.1907e-04,
         1.2308e-03,  3.3240e-03, -9.1617e-04, -3.1119e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1055e-02,  9.3592e-01,  3.1398e-03,  1.2351e-03,  1.3542e-03,
         3.5886e-03, -1.5956e-03, -3.4361e-04,  1.6416e-03, -1.6251e-03,
        -4.9380e-04, -1.5329e-03,  3.5764e-03,  3.1629e-03,  2.4548e-03,
         2.7141e-03,  1.0640e-03, -8.9841e-04, -4.8946e-03,  4.9586e-05,
         3.8705e-03, -2.7025e-03, -3.4363e-04,  3.8593e-03,  2.1997e-03,
        -2.0461e-04, -7.0378e-04, -3.2575e-05, -1.6165e-03,  9.3280e-04,
         7.9970e-05,  8.1299e-04,  1.6289e-04, -1.3512e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0118e-02,  9.5897e-01,  4.5798e-03,  1.3912e-03, -2.2050e-03,
        -1.8139e-03,  4.7124e-04,  8.2016e-04,  1.8989e-03,  8.3157e-04,
         1.4746e-04, -2.5084e-03, -1.3672e-03,  2.1885e-04,  1.4357e-03,
        -3.1327e-03, -1.2441e-03,  1.8408e-03,  4.3651e-04,  2.0390e-03,
        -1.3845e-03, -1.1435e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6374e-02,  8.9655e-01,  2.0813e-03,  1.3528e-02, -7.1667e-04,
         1.2618e-03,  3.1031e-03,  4.4271e-03, -3.6813e-04, -1.1814e-03,
         4.5517e-03, -4.5445e-03, -4.2059e-03, -2.2272e-03, -3.8841e-03,
        -3.3716e-03, -1.6524e-02, -1.1006e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3702e-03,  9.5352e-01, -1.4010e-03,  4.3005e-03,  4.8984e-03,
        -1.0349e-05,  6.2696e-04, -1.5375e-03,  2.6891e-03,  1.8384e-03,
         2.4619e-05, -4.2430e-03, -4.7830e-04,  1.5785e-03, -5.6286e-04,
        -1.0637e-03, -2.1242e-04, -4.4612e-03, -7.9123e-04, -5.2813e-03,
        -1.0750e-03,  4.1667e-03, -2.8655e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1294e-02, -9.5040e-01, -1.1076e-02,  3.4997e-05, -3.3579e-03,
        -2.1713e-03,  5.5611e-05,  1.1770e-03, -3.9739e-03,  1.0913e-03,
        -5.8976e-03,  7.1994e-04,  4.6607e-04,  8.7055e-04,  2.2472e-03,
         3.2497e-03,  1.9150e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6912e-02,  8.9308e-01,  3.9639e-02,  8.2019e-04,  1.5049e-03,
         6.8741e-03,  2.9886e-04,  1.9652e-03,  4.4400e-03,  5.1914e-03,
         3.5117e-03, -5.2657e-04, -2.5152e-03, -2.1533e-03, -1.6986e-04,
        -6.3100e-04, -1.1035e-04, -9.6609e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.3658e-03,  8.9932e-01, -8.8101e-03, -1.9306e-03, -2.2740e-03,
        -1.0253e-03, -1.9192e-03, -7.3204e-03,  2.4562e-03,  4.0834e-02,
         8.0712e-04, -1.1800e-03,  1.2070e-03,  4.3300e-04, -1.5411e-03,
        -4.3406e-04,  1.7632e-03,  1.3345e-03,  2.9798e-03, -1.3914e-03,
        -1.4882e-03, -2.6528e-03,  1.2602e-03,  1.8009e-03,  1.2747e-03,
        -3.1930e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5056e-02,  9.2778e-01, -7.9545e-03, -1.8215e-03,  3.9687e-04,
        -4.6811e-03,  3.8276e-04, -1.2249e-03,  1.8194e-03, -2.6968e-03,
        -3.2336e-03, -2.1692e-03, -2.1016e-03, -5.0642e-03, -6.0459e-03,
        -3.2070e-04, -8.3737e-04, -4.6911e-04,  4.3150e-04, -1.0715e-03,
         1.1282e-03,  9.5598e-04,  7.6409e-04, -5.4457e-05, -8.0945e-04,
         1.9161e-03,  9.1879e-05, -8.3876e-04,  1.3071e-03,  4.7133e-03,
        -1.8583e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1958e-02,  9.2506e-01,  6.6055e-04,  2.1839e-03, -6.5187e-03,
        -2.9123e-03, -6.2618e-03, -2.3815e-03, -5.5816e-03, -9.3242e-05,
        -3.8429e-03, -2.3897e-03, -4.3680e-04,  1.4111e-03, -1.0232e-03,
         4.3828e-04, -3.0205e-03, -3.8215e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6148e-02,  9.0790e-01, -3.8002e-02,  4.0599e-04,  5.2246e-03,
        -4.3209e-03,  7.7369e-04,  1.7226e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([-1.1986e-03,  9.3712e-01, -8.0796e-03, -8.9667e-04, -7.6159e-03,
        -1.0227e-03, -1.8738e-03, -8.8373e-04, -3.9690e-03, -7.0677e-04,
         1.3631e-03,  2.1722e-04, -6.4476e-04, -1.2576e-03,  1.5474e-03,
        -3.7682e-04, -2.7425e-03,  3.0712e-04,  2.9740e-03, -1.1011e-03,
        -3.4921e-03, -1.9549e-03, -4.0267e-03, -8.5847e-05, -1.7214e-03,
        -6.0248e-04, -2.7776e-03, -4.2228e-04,  1.3030e-03,  8.6739e-04,
         1.6163e-03, -5.6456e-04, -2.4284e-03, -2.7224e-07,  5.5352e-04,
        -3.5346e-04,  6.4964e-04, -5.7038e-04,  1.1547e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.7831e-03,  9.3653e-01,  3.0890e-03, -1.1124e-03, -1.4583e-03,
         8.6779e-04,  1.6246e-03, -1.1945e-03, -4.9087e-04,  2.4535e-03,
        -8.9430e-04,  5.7490e-04,  1.4137e-02, -8.9571e-04,  7.3292e-05,
        -2.0759e-03, -1.6176e-03, -2.1889e-03,  7.2230e-05,  1.5413e-03,
         6.7045e-03, -5.4985e-04,  2.4952e-03, -4.0822e-04, -6.9915e-04,
        -1.9638e-03, -2.5970e-03, -3.9073e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1823e-02,  9.3072e-01, -8.7957e-03,  1.1974e-02,  2.1778e-03,
         7.0843e-03,  5.5276e-04,  2.4328e-03,  4.9346e-03, -1.5307e-03,
         6.8487e-03,  6.0955e-03,  5.0278e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0976e-02,  9.5177e-01,  1.5341e-03, -6.3711e-03, -7.8637e-03,
        -2.4670e-03, -4.2567e-03, -2.1679e-03, -1.1538e-03, -8.4890e-04,
        -9.5128e-05, -2.8657e-03, -3.1989e-03, -2.4459e-03, -1.9860e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0568e-02, -8.9145e-01, -1.1253e-02,  1.9966e-03, -4.8171e-03,
        -1.0741e-03, -5.1488e-03, -1.7700e-03, -3.3677e-03,  1.3313e-02,
        -5.6595e-04, -1.5603e-03,  5.1829e-03, -3.2597e-03, -7.1141e-05,
        -1.7274e-03,  3.4364e-04, -1.9760e-03,  3.3278e-04, -4.6148e-04,
        -1.2678e-03,  8.1213e-04,  4.6554e-04,  2.1581e-04,  1.7998e-03,
         1.0941e-03, -6.1322e-04, -3.9481e-03, -4.2522e-04,  7.2487e-04,
         4.2952e-03, -7.5789e-04, -6.8222e-04, -9.4315e-05, -2.3026e-04,
        -4.6778e-04,  3.8160e-05, -6.2176e-04, -9.2132e-04,  7.1658e-04,
        -9.2585e-04, -1.6255e-04, -1.4613e-03,  5.4325e-04,  3.9867e-03,
        -6.6581e-04,  1.8205e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.1016e-03,  9.4188e-01,  4.2295e-03,  6.3579e-03, -4.9817e-03,
        -2.5493e-03, -7.7370e-04, -3.7220e-04,  1.6172e-03, -8.0226e-04,
         1.0743e-03, -2.8590e-03, -9.7339e-05, -4.1361e-04,  9.7518e-04,
        -3.9562e-04,  1.0566e-03,  3.2377e-04,  3.4247e-04, -1.1069e-03,
         4.9358e-04, -1.5571e-03,  1.6620e-03,  1.2937e-03,  5.2133e-04,
         8.2936e-04,  3.4836e-03,  1.8973e-03, -6.1385e-04,  7.5057e-04,
        -1.3951e-04,  4.9751e-04,  8.5982e-04, -8.7289e-05,  1.2043e-03,
         3.5873e-07, -3.9090e-04, -7.1162e-04,  1.5563e-04, -1.5076e-04,
         6.9276e-04,  2.6940e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8315e-02, -9.0314e-01, -9.0723e-03, -2.8849e-03, -7.7552e-04,
        -2.3515e-03, -1.8529e-02,  1.8754e-03,  2.4916e-03, -3.6324e-03,
        -2.4917e-03, -3.7473e-03,  1.2196e-03, -1.5411e-03, -3.1013e-03,
        -1.3624e-03, -2.2777e-03, -4.5361e-03, -4.7242e-04, -7.3258e-04,
        -7.7006e-04,  1.4941e-03, -2.3673e-03,  3.4702e-03,  7.3507e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6341e-03,  9.5316e-01, -2.5229e-03, -4.8567e-04,  2.1622e-03,
        -1.1801e-03,  3.8422e-03, -1.4391e-03,  1.6715e-03,  1.1721e-03,
         1.7225e-03,  1.3152e-03,  2.4724e-03, -5.7372e-03,  1.4088e-03,
         2.2218e-03,  3.0413e-04,  3.0162e-03, -2.5300e-03, -2.2481e-03,
         1.3127e-03, -9.1691e-04, -4.5244e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9703e-02,  9.2514e-01,  3.0062e-03, -4.3418e-03, -7.3974e-03,
        -1.6855e-03, -1.2102e-04, -9.2150e-04, -4.9048e-03, -2.0441e-03,
         2.7480e-03, -3.0320e-03,  1.6671e-03,  4.6330e-04, -5.1258e-03,
        -1.1808e-03, -2.7854e-04, -1.3187e-03, -2.4489e-04, -1.2446e-03,
        -1.0105e-04, -8.4006e-04, -2.4903e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.7119e-03,  9.5031e-01,  7.7398e-03, -5.3753e-03, -5.3933e-03,
        -1.1693e-03, -2.0431e-03,  3.3824e-03,  1.9135e-04, -1.8223e-04,
        -9.4193e-03,  2.5658e-03, -5.5201e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1248e-03,  9.6167e-01, -1.0140e-02,  1.0948e-03,  1.5625e-03,
        -2.3964e-03, -1.2048e-03,  3.0820e-03,  2.6254e-04,  3.7711e-03,
         2.2609e-03,  1.6828e-03, -5.1136e-03,  2.6352e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3990e-02, -9.5123e-01, -1.4402e-03,  1.2265e-03,  7.7739e-04,
         9.2357e-04, -4.6560e-04,  3.2554e-03,  7.6152e-04,  6.7598e-04,
         4.5741e-03, -1.1866e-03,  2.0486e-03, -2.3773e-03,  3.1281e-04,
        -1.5068e-03,  3.6713e-03, -1.5427e-03, -8.0318e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-2.7357e-02,  9.1113e-01, -5.9827e-03, -3.8199e-03, -5.7986e-04,
        -4.4898e-04, -3.2649e-03, -1.5094e-04, -1.5845e-03, -2.2588e-04,
        -2.6910e-03,  1.8276e-03,  9.1387e-04,  9.8180e-03,  3.1971e-04,
        -3.2949e-03,  6.7157e-03,  2.3807e-03,  6.5482e-04,  4.7013e-03,
         1.0119e-03,  3.2326e-04, -1.5445e-03,  1.7600e-03, -5.8617e-03,
        -1.6388e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.4221e-03,  9.4134e-01,  2.3089e-03, -6.2543e-03,  3.9512e-03,
         2.9261e-03, -4.0226e-03,  5.0392e-03,  3.4211e-03, -6.6073e-04,
         3.4202e-03, -2.6296e-03,  2.9483e-03,  1.6340e-03,  6.9428e-03,
         3.0802e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6647e-02,  8.9831e-01, -3.1263e-03,  4.3008e-03,  1.9777e-03,
        -8.5365e-04,  6.6917e-03, -1.5869e-03,  2.7130e-03,  1.0163e-03,
         4.8205e-03, -7.0836e-03,  9.3531e-04, -1.9950e-03,  6.0397e-03,
         1.0419e-03,  1.0859e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0092, -0.9458, -0.0115, -0.0011, -0.0048,  0.0114, -0.0014, -0.0030,
        -0.0119,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.6356e-03,  8.7195e-01, -1.0287e-02,  9.4652e-03, -3.7266e-03,
        -3.1742e-04,  7.4475e-04, -4.4789e-03, -1.7766e-03,  5.7411e-04,
         2.7149e-03,  1.8677e-02, -1.5696e-04, -1.2538e-04,  1.3500e-03,
         2.0132e-03, -3.8913e-04,  1.7847e-03, -8.5661e-04,  2.6403e-04,
         1.8800e-03,  1.9793e-03, -1.4196e-03, -2.3364e-03,  1.2698e-03,
         2.3745e-03,  1.5940e-03,  1.2005e-05,  3.5486e-03, -1.7086e-04,
         1.2950e-03,  1.4069e-03,  2.7588e-04, -2.5352e-04,  1.6991e-03,
         2.6607e-03, -1.5939e-03, -1.0406e-03, -2.4425e-03, -7.0839e-04,
        -2.5043e-03,  1.1083e-02, -1.4989e-03, -8.4995e-04,  7.1864e-03,
         7.6311e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8597e-03,  9.2176e-01,  2.2236e-02, -6.1444e-03,  4.4754e-04,
         2.5707e-04,  8.1589e-04, -2.7503e-04, -2.8384e-03,  6.2555e-04,
        -1.7224e-03, -1.0050e-03,  5.7162e-03, -4.9460e-04,  3.8663e-04,
         4.1305e-04,  9.8916e-04,  4.3645e-03,  5.4279e-04,  5.1176e-04,
        -3.9635e-04, -4.1601e-03, -4.6861e-03,  2.1799e-03, -1.7459e-03,
        -2.9992e-03,  2.6815e-03, -1.3527e-03,  3.6229e-03, -2.7654e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0174,  0.9463,  0.0043,  0.0192,  0.0021, -0.0059,  0.0047,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4683e-02, -9.6545e-01, -1.6919e-03,  1.0010e-03,  1.7491e-04,
         2.8508e-03, -3.5266e-03, -5.0704e-03, -3.6472e-03, -1.7783e-04,
        -1.7235e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6594e-02,  9.0066e-01,  1.9165e-02, -2.3294e-02, -3.9630e-03,
         2.6770e-03, -3.5545e-03, -5.2770e-03,  7.2437e-03, -5.7055e-03,
        -3.3192e-03,  6.0247e-03, -3.9184e-04, -2.1276e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8562e-02,  9.0873e-01, -1.4165e-02,  9.1863e-04, -1.5692e-03,
         7.3898e-06,  6.7400e-04, -4.3384e-04,  6.1765e-03, -4.0318e-04,
         6.3994e-04,  2.1509e-03, -2.1440e-03, -1.1186e-02, -1.1689e-03,
         9.0803e-04,  1.4755e-03,  1.6084e-04, -4.6588e-03, -1.6614e-03,
         8.6673e-03, -4.5127e-03, -7.8674e-04, -1.1318e-03,  2.6006e-04,
        -1.1251e-03, -9.8742e-04,  1.6408e-04, -1.1476e-03, -8.5149e-05,
         1.6169e-04,  2.4957e-03, -3.0967e-05,  6.5117e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4979e-02,  8.9446e-01,  1.4798e-02,  3.3805e-03,  2.6489e-03,
        -1.9758e-03, -5.0658e-03, -2.8560e-03, -1.0342e-03, -1.5501e-03,
        -3.1442e-03,  1.3873e-03, -2.5091e-03, -6.8119e-04, -2.3967e-03,
        -2.5053e-03, -9.1682e-03,  3.7821e-03,  3.2232e-04,  2.4215e-04,
         4.2261e-03,  8.3647e-05, -5.0546e-04,  1.2372e-03,  2.5500e-04,
         5.3997e-03, -7.7370e-03, -1.1667e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3502e-02,  9.1756e-01,  1.8981e-03, -2.4313e-03,  2.4181e-03,
        -6.4926e-03, -1.6571e-03,  2.5699e-03, -1.8667e-03, -7.9830e-04,
        -2.4836e-03, -8.3704e-04,  1.2108e-03,  7.1432e-03, -8.2879e-03,
        -1.8809e-03,  4.6358e-03,  3.1872e-03,  7.1932e-04,  6.3883e-04,
        -4.7095e-04,  1.3710e-03, -1.1853e-02, -4.0851e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 2.5783e-02,  8.9980e-01,  4.8225e-03,  5.7297e-03,  1.0057e-03,
         1.0521e-03, -5.3913e-03, -1.3458e-03,  7.3632e-04,  1.8150e-03,
        -5.0241e-03,  6.0642e-04,  7.1833e-03,  3.3405e-03,  5.2460e-04,
         9.9425e-04,  3.7047e-03,  3.5633e-04,  1.7008e-03,  7.8881e-03,
         1.5580e-03,  3.6924e-03, -2.9140e-03,  4.0290e-03,  3.7716e-04,
         2.6027e-03,  1.9498e-03, -9.6267e-04, -7.9629e-04, -4.8643e-04,
        -1.8237e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1240e-03,  9.6971e-01, -1.5633e-03,  2.2547e-04, -7.8890e-03,
        -4.4710e-03, -2.9979e-03, -5.6072e-03, -4.7519e-04, -6.7394e-04,
        -1.4622e-03, -2.7964e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.5203e-03,  9.4080e-01,  6.5461e-05,  1.0092e-02,  2.0198e-03,
        -6.2622e-03,  1.9352e-04,  5.2705e-03,  1.5143e-03,  5.7967e-03,
        -4.1716e-03, -5.0117e-03,  2.5739e-03,  5.5903e-03,  1.1189e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4089e-02,  9.4522e-01,  2.5419e-03, -5.3940e-03, -5.3809e-03,
         1.6398e-03,  7.9890e-04,  1.4155e-03, -9.3157e-04, -2.6051e-04,
        -3.1204e-04, -1.7751e-03, -8.8223e-04,  7.6904e-04, -4.7709e-04,
        -4.1448e-04,  1.9776e-03,  2.0183e-03,  1.7267e-03, -5.9382e-04,
         5.4149e-04,  1.1060e-04, -2.8279e-03,  1.0447e-03,  2.0345e-03,
         1.1072e-03,  1.3924e-04,  4.1007e-04, -1.1075e-03, -9.3496e-04,
         3.8318e-04, -7.1099e-04, -3.3425e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1619e-03,  9.3618e-01, -1.6701e-02, -6.3360e-03,  3.0089e-03,
        -6.3469e-03,  9.7158e-04, -1.1404e-03, -1.0677e-02,  1.8280e-04,
        -8.4557e-04, -1.3547e-03, -3.7768e-04,  9.7709e-03, -1.9261e-03,
        -1.0172e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5979e-02,  9.3093e-01, -1.5323e-02,  1.2407e-02,  1.6977e-04,
        -1.1890e-03,  4.5203e-03, -2.0178e-03, -1.7461e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0208,  0.9581,  0.0124, -0.0047, -0.0040,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4137e-02,  9.1388e-01, -1.5097e-02,  2.6524e-03, -4.1995e-04,
        -4.4130e-03, -2.4676e-03, -1.6630e-03,  3.5208e-04,  2.5021e-04,
        -3.2057e-04, -1.4076e-03,  8.5508e-04, -2.0137e-03,  1.5206e-05,
        -2.8718e-04,  1.1498e-03, -7.2404e-04,  1.4086e-03,  2.3622e-03,
         2.6140e-03,  1.6235e-03,  7.2057e-04,  2.3238e-03, -2.6605e-04,
        -3.0781e-03,  3.4992e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0034, -0.9558,  0.0011,  0.0028,  0.0094, -0.0121,  0.0155,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5422e-02,  9.2512e-01,  3.0628e-03,  9.7057e-03,  1.5557e-03,
         3.6132e-03, -4.9776e-03,  7.5012e-04, -1.1937e-03,  4.5801e-03,
         7.1698e-04, -9.6494e-03, -2.9974e-04,  7.9855e-04, -1.0180e-03,
         3.1796e-03,  9.0359e-04, -1.4064e-04,  9.8699e-04, -1.2372e-03,
        -5.6693e-03,  5.4191e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.0196e-04,  9.5664e-01,  6.0840e-03,  7.0926e-03,  3.8439e-03,
        -1.7632e-04, -9.1517e-04,  2.6998e-03, -9.2218e-04, -5.0863e-03,
        -9.3934e-05,  9.8645e-05,  5.4126e-04,  5.0687e-03, -1.3924e-03,
         2.2657e-03,  1.1227e-03,  1.4401e-04, -4.9129e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3453e-03, -9.4544e-01,  5.1280e-03, -2.9338e-04,  3.2299e-03,
         1.4902e-03,  3.6344e-03,  1.8226e-03, -2.4139e-03, -4.2980e-04,
         2.7933e-03,  9.2299e-04,  4.6600e-04,  3.9835e-04,  2.0320e-03,
        -5.6831e-04,  1.4230e-03,  3.4920e-03,  2.2418e-05,  3.6516e-03,
         6.6201e-04,  1.6234e-03, -6.4656e-04,  3.6467e-04, -5.0556e-04,
         5.4705e-04,  1.0481e-03,  1.2902e-03,  4.8322e-04,  6.9234e-03,
        -2.9058e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 8.8230e-04, -9.0099e-01, -1.9835e-02, -3.7912e-03,  5.7921e-03,
         2.0450e-03, -1.1155e-03, -1.5337e-03, -2.9468e-04, -5.2918e-04,
        -2.0091e-03,  4.4688e-04,  4.2598e-03, -1.8467e-02,  6.6564e-04,
        -3.3500e-04, -8.8228e-03, -4.7590e-04, -1.4162e-03,  2.8894e-04,
         1.6659e-04,  4.6068e-04, -6.6628e-04, -3.0244e-04,  2.2194e-03,
         2.0621e-05,  2.8341e-03,  8.1230e-03, -1.8083e-03, -2.4614e-03,
        -3.6436e-03, -3.2974e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0047e-03, -9.3286e-01,  6.3794e-03,  1.4408e-03, -3.9258e-03,
        -1.0938e-03, -1.8390e-04, -2.8316e-03, -1.3685e-02, -1.4704e-03,
        -1.6634e-03, -3.2616e-03, -6.5996e-04, -1.8841e-03, -3.7106e-03,
        -2.2077e-03, -2.7472e-03,  1.6710e-03,  3.0905e-03, -2.0993e-03,
        -2.2482e-03, -4.8762e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6472e-03, -9.3657e-01, -1.9164e-03, -6.6868e-04, -9.1815e-05,
        -2.7171e-03, -2.1713e-03, -8.1783e-04, -4.0602e-03, -9.3446e-04,
        -5.1103e-04,  3.5857e-04,  2.7677e-03,  1.0639e-03,  1.7995e-03,
        -8.2801e-04,  2.0961e-03, -9.9474e-03,  2.3078e-04, -3.5708e-04,
         5.8497e-04,  4.4419e-03, -8.8550e-04,  1.1999e-03,  1.3177e-04,
        -4.4121e-03,  1.9261e-04,  1.1832e-02,  1.1697e-03,  2.5924e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0918e-03,  9.5774e-01,  2.9723e-03, -1.0795e-03, -3.7940e-03,
        -3.7125e-04,  1.2242e-03, -1.8309e-04, -9.8664e-03, -5.5220e-03,
        -9.3055e-04,  3.6390e-03, -2.6721e-03,  2.0641e-03, -2.0824e-03,
        -1.3879e-03, -1.0177e-03,  2.3573e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1826e-03,  9.1284e-01,  4.6210e-02,  2.3608e-03, -8.8767e-04,
        -5.8902e-04, -4.9149e-04, -1.3533e-03, -2.0012e-03,  2.4938e-04,
        -3.4417e-04, -5.6718e-04,  1.0815e-03, -3.0386e-04, -4.9891e-04,
        -1.9542e-03,  8.1290e-04,  1.5654e-04,  4.1832e-04,  5.7001e-05,
         5.9691e-04, -9.2372e-04,  1.2039e-03,  2.0129e-04,  9.7264e-04,
        -1.3318e-03, -2.7513e-03,  1.5797e-04, -8.0496e-04, -1.6407e-03,
        -9.6556e-04, -4.1922e-04,  1.0572e-03, -2.9057e-03, -4.2396e-04,
        -1.8292e-03,  4.4537e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9787e-02, -8.9090e-01,  1.9007e-02, -2.1398e-02,  9.5759e-04,
         7.9381e-04,  6.9316e-04,  1.8544e-03, -1.7485e-04, -1.7185e-03,
         1.0469e-03,  3.8761e-03,  7.8419e-04,  2.9313e-03,  1.7388e-03,
        -5.2141e-04, -1.4505e-03,  3.0651e-03,  8.2785e-04, -2.7629e-05,
         1.4028e-03,  5.6467e-04, -2.0159e-03,  2.4641e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3497e-03,  9.4522e-01,  4.3729e-03,  1.0112e-03, -2.3277e-04,
         2.1313e-03, -1.0120e-03,  2.5509e-03,  2.1314e-03, -6.1527e-04,
         1.0655e-03, -5.3077e-04,  3.4604e-04, -1.1536e-03,  1.1328e-03,
        -1.3788e-04, -2.2778e-03, -1.4411e-03,  7.8143e-04,  1.1305e-04,
         1.0878e-03, -2.5260e-03, -1.0050e-03, -2.4405e-04, -3.2155e-04,
         2.0669e-03, -8.4322e-04, -1.7102e-05, -9.7178e-04, -4.1002e-04,
         9.5161e-04, -5.5907e-04, -7.1122e-04,  5.6089e-04, -1.2109e-03,
         2.9612e-03, -5.0426e-04, -1.5641e-04,  4.2073e-04, -7.1931e-04,
        -4.7601e-04,  2.7809e-04,  1.0397e-03,  3.7265e-03,  1.8556e-03,
        -7.6342e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.3554e-05,  9.2505e-01, -6.1283e-03, -2.5639e-03, -6.9217e-03,
         5.2233e-04, -1.6492e-02, -9.0410e-04,  3.2731e-04, -3.4597e-03,
         1.0357e-03, -9.4618e-04, -3.1942e-03,  4.2331e-03,  1.9961e-03,
         2.1971e-03,  1.3618e-03,  6.6215e-04,  7.2471e-05, -5.3536e-04,
         1.6456e-03, -1.3770e-02, -5.9359e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4560e-03,  9.5899e-01, -1.0278e-02, -3.7165e-03,  6.2674e-05,
         4.9978e-04, -4.6964e-04, -3.3335e-03, -1.8959e-03, -4.5314e-03,
        -1.2764e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0054e-02,  9.1964e-01, -5.2891e-03,  6.4459e-03,  5.9652e-03,
        -8.0357e-03, -5.2774e-03, -5.7083e-03,  1.5095e-03,  4.3613e-03,
         3.8602e-04,  5.7059e-04,  7.5448e-04,  6.0041e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0110,  0.9510, -0.0022, -0.0079, -0.0016,  0.0011, -0.0086,  0.0074,
        -0.0051, -0.0011, -0.0013,  0.0017,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7247e-02, -9.3093e-01,  9.5116e-04, -3.6756e-03, -1.6946e-03,
        -1.8299e-03,  6.5084e-04,  3.8325e-04, -8.3066e-04, -6.1658e-03,
        -1.2182e-04,  5.8941e-05, -7.6582e-04, -9.4492e-04, -1.1111e-04,
         1.5691e-04, -8.5230e-04,  3.4152e-04, -4.8138e-04,  1.4405e-03,
         9.2076e-04, -7.1153e-04, -3.5877e-03,  1.0028e-03,  2.2456e-03,
         1.0695e-03, -1.3402e-03,  9.7200e-04, -1.3291e-04, -1.5647e-03,
         2.7019e-04, -2.4302e-03,  2.2829e-04, -7.7519e-04,  2.3363e-03,
         7.0351e-04,  6.3980e-04, -4.6466e-04, -6.9956e-06,  4.2890e-04,
         1.0858e-04, -5.8151e-04,  1.2963e-03, -1.6429e-04, -9.8646e-04,
        -9.4859e-05, -5.5745e-04, -2.0568e-04, -1.3218e-04,  2.5082e-04,
         7.7081e-04, -1.5129e-04,  7.9333e-04, -8.5311e-04,  1.5840e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 7.5058e-03, -9.1746e-01, -3.8569e-03,  5.7877e-03,  1.4770e-03,
        -4.4965e-03, -2.6608e-03, -3.0795e-03, -3.0422e-03, -2.0684e-03,
         7.7384e-03,  1.5738e-03, -6.8975e-03,  2.3064e-03,  2.2386e-03,
        -1.0836e-03,  1.8987e-03,  1.5589e-03, -2.6539e-03,  4.6737e-04,
        -3.6705e-03, -6.8423e-03,  9.6310e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6020e-02,  9.3025e-01, -7.4632e-03,  2.0639e-04,  4.3596e-03,
         1.4889e-03, -2.4300e-03, -7.8671e-04, -1.4485e-04, -3.6696e-03,
        -2.4214e-03,  4.0557e-04, -4.6169e-03,  1.6738e-03, -1.7717e-03,
         1.0060e-04, -4.1609e-03,  8.0333e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0040,  0.9283,  0.0046,  0.0014,  0.0124,  0.0029, -0.0067, -0.0011,
         0.0010, -0.0049,  0.0024,  0.0022,  0.0052, -0.0012,  0.0032, -0.0054,
         0.0031, -0.0101,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3132e-03,  9.4844e-01, -3.8089e-03,  4.7105e-03,  4.1682e-03,
        -4.1336e-03, -2.3509e-04, -4.4698e-03,  2.9110e-03,  3.5177e-03,
        -3.4540e-04,  2.3525e-03,  4.6771e-03,  6.0914e-03, -6.8236e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0183e-02,  9.2166e-01, -1.1606e-02,  3.4463e-03, -1.3635e-03,
        -4.8055e-03, -3.4155e-03, -3.4644e-04, -4.3452e-03, -2.9305e-03,
         1.5970e-03, -7.0563e-04,  5.2776e-04, -4.9566e-04, -2.0672e-03,
         2.4991e-03, -4.7201e-03, -4.6224e-04, -1.3231e-04,  5.4799e-03,
         4.6643e-04,  2.5538e-03, -1.0823e-03, -1.1418e-03,  2.2565e-03,
         1.3725e-03, -6.3572e-05,  1.9297e-05,  4.8835e-03,  2.4021e-03,
        -1.1168e-04, -8.6133e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1297e-04,  9.3169e-01, -1.6584e-02, -3.5793e-03,  1.4120e-03,
         8.9273e-04,  5.5789e-04,  4.1966e-04, -8.1852e-04,  1.7082e-03,
         1.8037e-03,  5.7046e-04,  6.7339e-04,  3.9538e-04,  5.4495e-04,
        -2.4553e-03,  4.5218e-04,  1.8803e-03,  1.2607e-03,  1.7090e-03,
         5.6925e-04,  3.9541e-03, -2.7143e-03, -3.7683e-03,  1.8710e-03,
         1.3246e-03,  1.7249e-03,  2.2743e-03,  2.1554e-04,  1.6587e-03,
        -5.1017e-03,  1.0621e-04, -5.0006e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6111e-03,  9.3460e-01,  3.7526e-03,  1.2076e-03, -2.5318e-03,
         3.8765e-04,  1.0480e-03, -1.8935e-03,  1.8956e-03,  2.2551e-03,
        -2.6481e-04,  1.0716e-03, -1.1089e-03, -2.0135e-03, -1.2450e-03,
        -1.1701e-03,  1.2078e-03,  9.7516e-04, -2.0894e-03, -5.0498e-04,
        -8.0655e-04, -1.5569e-03,  4.8259e-04, -6.3689e-05,  4.4600e-04,
        -3.7126e-06, -7.5997e-05, -6.5074e-04, -3.4441e-04, -3.0767e-03,
         2.2494e-03, -3.4668e-04,  1.0216e-03, -4.9887e-03,  3.3829e-04,
         1.5325e-03,  7.9385e-05, -1.2016e-03,  1.1662e-04, -6.2854e-04,
         1.5028e-04, -1.4234e-03,  2.7753e-04, -2.7530e-04,  5.8685e-04,
        -1.8913e-03, -3.4742e-04,  9.8337e-05,  1.3804e-04,  1.3651e-04,
         1.0436e-03, -8.3168e-05,  5.6276e-04,  7.0329e-04, -2.6172e-04,
         1.5551e-03,  1.3520e-04,  5.4901e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.0015e-03, -9.3538e-01,  4.0994e-03, -2.8809e-03,  2.1303e-03,
         2.3556e-03, -1.3838e-03, -3.7197e-04, -1.5246e-03,  1.3743e-03,
        -3.0382e-03, -1.8890e-03,  3.6003e-04,  4.1708e-03, -4.7355e-04,
         1.0119e-03, -1.3739e-03, -1.6097e-03,  7.7712e-04,  1.4652e-03,
         5.0293e-04,  2.1531e-03,  1.3675e-03,  3.6111e-03,  1.1607e-02,
        -5.0902e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2967e-02,  9.3942e-01,  1.0979e-03,  7.5979e-03,  5.6892e-05,
         5.5839e-03, -2.2834e-03,  1.5964e-04, -2.9193e-03, -1.9611e-03,
         2.5498e-03,  1.8034e-03,  9.2161e-04, -1.0083e-03,  5.8744e-04,
        -2.6208e-05, -3.4086e-04,  6.1588e-04, -1.4867e-03,  1.7281e-04,
         3.2283e-03, -1.1545e-02, -3.1636e-04, -1.4231e-04, -1.2074e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.2946e-03, -9.3017e-01, -2.1170e-03, -2.0133e-04, -2.0864e-03,
         4.6598e-04, -7.8951e-05,  1.0017e-03, -1.9190e-03,  4.2104e-04,
         1.5817e-03, -2.4732e-04,  5.2780e-04, -4.5745e-03,  1.2937e-03,
         9.0666e-04,  3.4526e-03,  1.3940e-03, -6.2416e-04,  1.2055e-04,
         1.2474e-03,  4.3052e-04, -1.0744e-03,  2.9001e-03,  7.8512e-04,
         8.9028e-04,  2.3082e-04, -3.1790e-04,  3.6298e-03,  4.6650e-04,
        -1.5796e-03,  1.0592e-03, -2.4662e-03,  5.8736e-04, -1.9543e-03,
         1.5460e-03,  1.7331e-04,  1.1068e-03,  7.0217e-04, -2.5999e-03,
         1.4945e-03, -5.2682e-04, -8.1054e-04, -3.3296e-03, -5.6157e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0965e-02,  9.4388e-01, -2.7831e-03,  1.5666e-03,  1.0483e-04,
        -4.9045e-03, -6.4559e-03,  9.6800e-04, -1.7622e-03,  1.9898e-03,
        -4.8060e-03,  7.1941e-04, -3.0021e-04,  4.3345e-04, -3.2514e-03,
        -9.9057e-04,  4.6587e-03, -7.9028e-04, -1.0515e-03,  1.6938e-03,
        -3.8734e-04,  3.9562e-03,  1.5838e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2890e-02,  9.4116e-01, -3.8034e-03,  1.5266e-03, -2.9643e-04,
         1.3096e-02,  3.2470e-03, -1.5607e-03, -1.9185e-04, -5.3236e-04,
         7.8298e-03,  3.8705e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 6.8130e-03,  9.4726e-01, -1.8701e-03, -1.4178e-03, -7.8411e-03,
        -1.6147e-03,  2.7915e-03, -7.4575e-04,  3.1138e-03, -3.0212e-03,
         9.5584e-04, -2.7572e-03,  7.2868e-04, -4.4054e-03, -1.8350e-03,
         1.9758e-03,  1.6014e-03,  2.9058e-04, -1.1722e-03,  1.3352e-03,
         1.8972e-03,  2.3744e-03, -2.1831e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0039e-03,  9.3646e-01, -4.5963e-03, -1.1871e-03, -3.2062e-03,
        -9.2483e-04,  3.3184e-03,  2.8343e-03, -6.8857e-04,  8.1033e-04,
         3.7452e-04, -2.0683e-03,  1.3207e-03,  1.9394e-04, -3.2898e-04,
         8.1433e-04,  2.2629e-03, -7.4833e-04,  2.2769e-03,  1.4892e-03,
         1.3191e-03, -1.1605e-03, -1.6965e-03,  2.7820e-04,  1.1544e-03,
        -6.9970e-04, -2.6877e-03,  2.0420e-03, -1.1386e-03,  2.4796e-04,
         1.7015e-03,  1.5010e-02, -3.9585e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2837e-02,  9.5371e-01,  1.7647e-03, -3.3025e-03,  1.4898e-03,
         1.5141e-03,  4.0006e-03,  1.5859e-03,  1.1982e-03, -1.2679e-03,
         3.9067e-03,  6.8075e-05,  1.9420e-03,  3.5054e-03, -6.1934e-04,
        -2.1712e-03,  7.2250e-04, -3.5753e-03, -8.2118e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1684e-03,  9.6622e-01,  2.1381e-03, -8.0312e-04, -3.4524e-04,
         4.2957e-03, -4.0229e-03, -1.2980e-03,  5.5091e-03,  3.0407e-03,
         1.2382e-03,  3.3110e-03,  1.5286e-03, -7.8093e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.8467e-03, -9.5261e-01, -1.5456e-02,  2.7083e-03,  2.8748e-03,
         6.4040e-04, -5.2550e-04, -6.8282e-04, -1.5335e-03, -5.6138e-04,
        -2.1128e-03, -2.7222e-04,  7.7395e-04, -3.2835e-03, -2.3121e-04,
         5.9904e-04, -2.2817e-04,  5.4587e-05,  2.8362e-03, -1.2503e-03,
         3.1420e-03, -3.7792e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3280e-02,  9.4315e-01,  8.6558e-05, -4.3080e-03, -3.8650e-03,
        -2.0280e-03,  4.3312e-04, -3.7090e-03, -8.0833e-04, -3.1563e-04,
        -1.6323e-03,  1.6651e-03,  5.6790e-03,  8.7321e-04, -9.9465e-04,
        -5.0406e-03, -1.0809e-03,  1.6968e-03,  9.5390e-04, -9.2898e-04,
        -3.1738e-04,  1.1381e-03,  5.5806e-05, -1.2828e-03, -2.8120e-03,
         4.6540e-04,  1.1942e-03, -1.0425e-04,  8.1184e-05,  2.3593e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9847e-02,  9.4400e-01,  3.0193e-03, -3.5518e-03, -3.3923e-03,
         8.7881e-04,  5.6379e-04, -1.7614e-03, -1.5416e-03,  8.5928e-04,
        -2.6688e-03,  1.1838e-03, -2.7901e-03,  2.2388e-03,  3.0987e-03,
        -5.5841e-03,  3.0182e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9787e-03,  9.2835e-01,  4.8102e-03,  3.2133e-02, -7.7230e-04,
        -4.9006e-04, -5.7398e-03,  6.7164e-03, -5.0101e-03, -3.2166e-03,
        -8.7814e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2568e-02, -9.3704e-01, -1.2657e-03,  1.4739e-04,  8.6265e-04,
        -6.3828e-03, -1.7565e-03, -1.2452e-03, -3.4165e-05,  1.2326e-03,
         1.4602e-03, -2.0781e-03,  1.1031e-03,  3.1529e-03, -2.0043e-03,
        -1.8243e-03,  5.8380e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.7314e-03,  9.4046e-01,  3.4583e-03,  7.2458e-03, -9.7674e-04,
         2.3884e-03,  1.7138e-03,  2.2793e-04,  1.2418e-03,  1.0873e-03,
         8.6590e-04, -4.9666e-04, -2.5193e-03, -1.8122e-03,  2.5050e-03,
         4.6105e-04, -1.3212e-03,  3.0013e-04, -5.8274e-03,  2.6301e-03,
         3.2900e-04, -3.5482e-03,  2.1137e-03, -2.0182e-05, -2.4053e-03,
        -4.3119e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7031e-02,  8.9115e-01,  2.6587e-02, -1.2647e-03,  8.7204e-04,
         2.9709e-04,  1.9810e-03,  4.8118e-03, -2.3779e-04,  2.5844e-03,
        -1.8175e-04, -8.2122e-04,  6.6992e-04, -1.5377e-03, -3.4087e-04,
         2.8325e-04, -7.9050e-04, -1.4862e-03,  1.1939e-03, -1.2003e-03,
         5.7370e-04, -2.3229e-04, -6.1336e-04, -1.9343e-03, -1.4157e-04,
        -1.1071e-03,  2.4156e-03, -3.3955e-03, -2.6552e-03, -1.6078e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2181e-02,  9.5031e-01, -5.9424e-03, -3.4116e-03,  2.6248e-03,
         1.9419e-03, -7.9805e-04, -3.4315e-04,  2.2384e-04, -4.6501e-03,
        -1.0783e-03, -1.0792e-03,  5.2038e-03, -2.9158e-03,  7.0854e-04,
        -1.4286e-03,  1.6314e-03,  5.2914e-04,  2.9967e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-1.9297e-02,  9.0905e-01,  3.2378e-03, -7.9827e-04,  6.7839e-04,
        -1.3058e-03,  5.9249e-03,  2.0431e-03,  1.0087e-03,  3.9171e-03,
         4.9434e-04,  1.0883e-03,  1.0455e-03,  3.7194e-04,  7.5497e-03,
         3.5538e-04,  1.7238e-03,  4.9277e-03,  6.5486e-03,  3.1372e-03,
        -2.6716e-03,  2.7484e-03,  9.2321e-04,  3.2885e-03,  6.7102e-04,
         5.7839e-03,  9.4075e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4997e-05,  9.6061e-01,  1.0594e-02, -6.6380e-03, -3.2083e-03,
        -3.1099e-03, -1.7661e-03, -1.2012e-03, -1.3315e-03, -1.1967e-03,
        -5.3930e-03,  3.2623e-03, -5.4331e-04,  1.0886e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5528e-02,  9.3392e-01,  1.0790e-02,  3.8816e-03,  3.7774e-03,
         3.3088e-03,  6.4130e-04,  1.9372e-04,  1.0511e-04,  2.7879e-03,
        -5.9951e-04,  2.9488e-03,  1.9484e-03, -2.4353e-04,  2.1817e-03,
        -5.6001e-04,  5.4356e-04,  8.0204e-04, -1.2857e-03,  2.2174e-03,
        -1.1391e-03, -1.3650e-03,  2.1477e-04,  1.2279e-03, -7.0706e-03,
        -7.2233e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0149, -0.9150, -0.0092,  0.0056,  0.0011,  0.0128,  0.0038, -0.0069,
         0.0028, -0.0032, -0.0021,  0.0124,  0.0016,  0.0087,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.0432e-03,  9.4954e-01, -4.1324e-03, -2.8500e-03,  1.6943e-04,
        -6.0365e-04,  2.8498e-03,  8.3979e-05,  2.1155e-03,  6.7957e-04,
         2.2803e-03, -4.5385e-05,  3.0083e-03,  7.7994e-04,  2.2915e-03,
         5.8779e-04,  1.2054e-03, -4.5035e-05,  1.7440e-03, -1.0041e-03,
        -3.6520e-04,  2.5438e-03,  8.6407e-04,  1.7671e-03,  2.5472e-03,
         1.4809e-03,  4.2285e-05,  1.8926e-03, -7.2549e-04, -9.3723e-04,
         3.0192e-03,  2.7544e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.4149e-03,  9.3018e-01,  1.6065e-03,  5.1007e-04, -5.3853e-03,
         4.2456e-04,  1.7114e-03, -1.9722e-03,  5.5166e-04, -2.5854e-03,
        -7.2923e-03,  2.4665e-03,  2.4219e-03, -1.1555e-03,  5.4351e-04,
         6.0868e-03,  1.3079e-03, -2.9601e-03,  1.7044e-03,  2.6970e-03,
         1.5983e-04,  1.6290e-04,  2.0027e-04, -1.4400e-02,  3.0973e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0197,  0.9659,  0.0073,  0.0012,  0.0018,  0.0041,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7148e-03,  9.4104e-01, -2.3521e-03, -1.1162e-03,  1.7687e-03,
        -4.0376e-03,  1.5842e-04, -6.3174e-04, -1.7444e-03,  2.0142e-03,
        -2.3266e-03, -6.9890e-04,  1.1571e-04,  7.0321e-03, -2.6314e-04,
        -9.1530e-06, -3.2410e-04,  1.2299e-03,  2.7993e-03,  1.1352e-03,
        -3.2423e-03, -8.2100e-04, -1.4922e-03, -9.2425e-04, -6.7747e-04,
         2.0376e-03, -1.0478e-04,  2.0884e-03, -5.4826e-04,  5.5467e-05,
         9.1585e-04, -3.5825e-03, -1.7778e-03, -2.8585e-05,  6.6596e-04,
        -3.7894e-03, -1.7387e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.4451e-03,  9.1306e-01,  1.1074e-02,  2.5532e-03, -1.4484e-03,
         2.2578e-03,  3.6174e-04,  1.8023e-03, -7.4152e-04,  4.7299e-04,
        -1.9256e-03,  3.2427e-03, -3.4116e-04, -1.0386e-03, -2.0851e-03,
        -1.6705e-03, -3.7222e-04,  5.3242e-03, -6.5810e-04,  4.3152e-03,
        -8.2024e-03, -1.2119e-03,  1.3466e-03,  2.1988e-05, -2.2956e-04,
        -6.6430e-03, -7.1683e-03, -4.1036e-03,  5.5347e-04, -1.5059e-03,
        -3.5713e-04,  3.0194e-04,  2.0072e-04,  1.1178e-03,  3.8405e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2245e-02, -9.2312e-01,  1.3706e-02, -6.3886e-03,  1.4116e-03,
        -5.0412e-03,  3.0506e-03,  1.9548e-06, -7.1416e-04, -3.4788e-04,
         1.8450e-03,  6.5640e-03,  3.6891e-03,  9.1648e-04,  9.5517e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7182e-03,  9.4153e-01, -1.4655e-02, -9.2613e-04,  5.8683e-03,
        -1.3517e-03, -1.9520e-03,  1.1736e-03,  2.3205e-03,  1.2684e-03,
        -1.4245e-03, -1.3571e-03,  5.6843e-05, -1.8677e-03,  6.5697e-04,
         1.2088e-03,  1.6668e-03,  6.0255e-04,  1.8372e-03,  7.9007e-04,
        -7.5799e-04,  1.5907e-03, -1.9581e-03,  9.4570e-04, -8.0555e-04,
         6.9833e-05,  1.6144e-03, -4.1190e-05, -4.3138e-04,  1.5816e-03,
        -2.6796e-04,  9.4156e-04,  3.7645e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4366e-02,  9.1026e-01,  4.1529e-03, -3.7640e-03, -1.7718e-03,
         1.2961e-03,  1.5108e-03, -1.3317e-03, -6.1789e-04,  9.8560e-04,
         1.2621e-05, -1.8382e-03, -3.7287e-03,  4.0107e-03, -6.4155e-04,
         7.0614e-04,  2.7508e-03, -1.9153e-03,  4.1663e-05,  1.9428e-03,
         4.4048e-03, -8.6476e-04,  4.3097e-04, -3.0779e-04,  1.3623e-03,
        -1.1147e-04,  3.9768e-03, -8.9516e-05,  8.0526e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 1.9229e-02,  9.4257e-01,  5.2947e-03, -3.2081e-03,  2.8226e-04,
        -2.4662e-03,  1.4591e-03, -6.0047e-03, -2.2502e-03, -5.3103e-03,
        -2.9732e-03,  3.5513e-03,  1.3112e-03, -1.3698e-04, -3.9559e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.7944e-03,  9.4169e-01,  1.2838e-02, -1.0483e-03, -1.5447e-03,
         7.0840e-04,  6.8917e-04,  8.3063e-04,  8.1271e-05, -2.2592e-03,
         1.7450e-03,  6.2131e-04,  1.0235e-03,  2.2620e-04, -1.8176e-04,
        -1.1664e-03,  5.3012e-06, -2.2432e-04,  4.5813e-04,  1.2555e-03,
        -2.0023e-05, -3.2665e-04, -6.7808e-04,  1.6812e-03,  1.4547e-04,
         1.0999e-03,  1.5831e-04, -8.5963e-04, -1.5723e-03, -2.4670e-04,
         1.0722e-03, -5.6443e-04, -1.2166e-03,  4.7536e-04,  9.7561e-04,
         1.1989e-03,  8.4093e-04,  1.1842e-03,  3.6367e-03, -2.0351e-04,
         9.2737e-05,  5.7520e-04,  3.0270e-04, -9.3635e-05,  4.7869e-04,
        -9.3194e-04,  1.1100e-03, -2.2711e-03, -6.0068e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6451e-02, -9.2002e-01, -7.6519e-03,  4.9452e-03, -1.5747e-03,
        -9.7919e-03,  2.3036e-03, -3.9167e-04,  4.8906e-03, -1.4788e-03,
         3.4322e-04, -8.2691e-04,  3.0761e-03,  1.3591e-03, -3.4047e-04,
         1.0377e-03, -2.6550e-03,  3.6855e-04, -3.7457e-04, -2.6971e-03,
         1.3968e-03,  9.3706e-04,  1.7067e-03, -6.9893e-05,  1.0594e-03,
         3.3396e-03, -9.6370e-06,  3.3369e-04, -3.0331e-04,  2.7142e-03,
         3.5235e-04, -2.1493e-03,  1.3634e-03, -7.7422e-04, -9.1106e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.3525e-03,  9.3844e-01,  1.5454e-02, -6.3083e-04,  1.3657e-03,
        -1.8031e-03, -3.2464e-03,  1.4254e-03, -1.1441e-03, -1.8210e-03,
        -2.0777e-03, -2.1636e-03, -1.6954e-03, -1.7769e-03,  8.1944e-03,
         6.4973e-03,  3.0983e-03, -2.2691e-03,  1.5475e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1997e-03,  9.4619e-01, -1.0708e-02,  3.5711e-03, -1.0905e-03,
        -2.9611e-03, -1.6998e-03,  5.3240e-04,  8.1830e-03, -3.1018e-03,
        -5.8625e-03, -1.6887e-03, -6.2312e-05, -1.0273e-02, -2.8812e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5016e-03,  9.6745e-01, -4.9530e-03,  3.3276e-03, -5.5492e-04,
        -1.6773e-03, -5.1218e-05, -7.0171e-04,  8.6012e-04, -1.7821e-03,
        -1.9262e-04,  2.0879e-03, -6.1293e-04, -9.2532e-04,  1.3439e-03,
        -1.7411e-03, -2.2778e-05, -1.5464e-03,  5.6400e-04,  1.0962e-03,
        -2.8180e-03,  1.8797e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2236e-03,  9.5718e-01,  3.2456e-03, -4.1069e-03,  7.5904e-04,
        -1.4071e-03, -2.6407e-03, -3.8043e-04, -8.6094e-04, -3.1337e-03,
        -8.1780e-04, -1.4781e-04,  5.9455e-04,  7.2763e-05, -2.1239e-05,
         9.5393e-04,  1.4423e-03,  2.7189e-03,  2.4499e-04, -1.1557e-03,
         2.3345e-04, -2.8853e-04, -2.2272e-03,  1.5533e-03,  2.4585e-03,
         8.1812e-04, -1.7678e-03,  7.8327e-04, -3.9175e-05,  4.7264e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6285e-02,  9.3740e-01,  5.4136e-04, -2.5337e-03,  2.1879e-02,
        -3.3627e-03,  3.8059e-03,  4.3099e-03,  4.8845e-03, -1.0510e-03,
        -3.9456e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.6991e-03,  9.2219e-01,  5.8136e-03,  2.5622e-03,  1.7853e-03,
         9.1188e-04,  6.9283e-03, -1.1570e-03, -2.0709e-03,  5.1100e-03,
         1.3723e-03,  5.2552e-03, -2.4892e-03,  3.7233e-04, -6.1435e-04,
         3.7541e-03, -1.4138e-03, -1.0594e-04,  7.2953e-04, -8.5772e-04,
        -1.7648e-02, -9.1615e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0465e-03, -9.2717e-01, -7.6632e-03,  3.3316e-03, -5.5758e-04,
         4.0600e-03, -5.7751e-05, -5.5811e-05,  2.0384e-03,  4.6531e-04,
         1.3703e-02, -9.2148e-04, -1.0424e-03,  4.6456e-04,  1.2948e-03,
         8.0343e-04, -2.9396e-03,  1.0894e-03,  2.2039e-03,  8.6601e-05,
         1.2226e-03,  1.4309e-03,  3.3742e-03, -2.1345e-04, -1.7545e-04,
         5.7440e-04,  4.5657e-03, -1.1180e-03,  6.4851e-03,  7.3432e-04,
        -4.3608e-03, -5.3780e-06, -4.7408e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.2258e-03, -9.3306e-01, -3.7099e-03,  2.6638e-03, -8.2888e-04,
        -3.3690e-03,  1.8265e-03,  5.2249e-03, -3.0362e-03,  7.3979e-04,
        -5.2397e-05, -1.2702e-03, -9.4716e-04,  2.6781e-05,  3.3319e-04,
        -6.7853e-03, -3.8641e-03,  1.5967e-03, -5.4219e-04,  1.5219e-03,
        -1.8330e-03,  5.8521e-03, -7.3804e-04,  7.9846e-04,  2.0032e-03,
        -4.7351e-03, -3.9612e-03,  3.4587e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.4671e-03, -9.3960e-01,  6.7412e-03,  3.2734e-03,  2.1699e-03,
        -5.1475e-04,  1.0972e-03, -1.4332e-03,  4.6548e-03,  2.3806e-04,
         2.6210e-03, -3.9348e-03,  4.5767e-04, -2.3039e-03,  2.2653e-03,
         2.8352e-04, -2.3354e-03, -1.1767e-03, -1.0753e-06,  4.5102e-03,
        -1.7092e-03,  5.6396e-04,  2.9000e-03, -5.2854e-03,  4.5987e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 1.4513e-02,  9.1664e-01, -1.4483e-03, -9.5253e-03,  3.9506e-03,
         1.0456e-03, -4.7623e-03,  8.2772e-04, -7.2566e-04,  1.8504e-03,
         3.1377e-03,  1.6623e-03, -2.8842e-04,  2.9700e-03, -7.1413e-04,
        -4.2097e-03,  1.2148e-04,  1.0028e-03,  1.0574e-03,  2.1862e-03,
         5.8351e-04, -2.7326e-04, -2.1285e-03, -3.6719e-04, -2.3195e-04,
        -5.9008e-04,  3.4350e-03, -1.0594e-03, -1.1931e-03, -4.4631e-04,
        -4.0006e-03, -1.4885e-04, -4.1022e-03, -8.8635e-04,  1.0348e-04,
         7.0150e-04,  1.9519e-04, -2.5986e-03,  2.0046e-03, -1.0017e-03,
         1.3053e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6740e-02,  9.2101e-01, -1.4094e-02, -5.7164e-03,  2.6063e-03,
        -3.3436e-03, -1.1308e-03, -7.4413e-04, -3.3332e-03, -8.4569e-05,
        -3.0232e-03,  6.9551e-04,  2.4360e-03,  3.7221e-04,  4.2973e-03,
        -4.0931e-03,  3.9844e-03,  1.7442e-03,  4.2590e-04,  3.8683e-03,
        -1.0684e-03, -1.9354e-04, -7.6008e-04,  1.7046e-03,  2.5261e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3766e-02,  9.5069e-01, -1.9896e-03, -1.9071e-03, -1.2675e-03,
         5.1965e-03, -2.5341e-04,  8.0694e-03, -1.8879e-03,  1.0155e-03,
         2.2912e-04,  4.1339e-03,  2.7154e-03,  2.0038e-03, -5.3610e-04,
         4.1312e-04,  4.8700e-04,  5.4243e-04,  1.1131e-03,  1.6064e-04,
         1.6249e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0719e-02,  9.4501e-01, -2.4052e-03, -5.5722e-03, -4.0017e-04,
        -3.6602e-03,  2.4087e-04,  1.7406e-03, -1.4671e-03, -1.7341e-06,
         4.8776e-03, -1.0065e-03, -1.9624e-03, -1.7759e-03, -3.4962e-03,
         5.1575e-03,  8.9122e-03, -1.5942e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3828e-02,  9.1898e-01,  6.5960e-03, -1.5144e-03,  2.4353e-03,
         1.7312e-03,  2.3111e-03, -1.0349e-03,  6.2164e-03,  2.8390e-03,
        -1.1927e-03, -1.0768e-03, -4.0705e-03,  6.8713e-04, -8.5332e-04,
         4.4993e-04,  7.2824e-03,  9.7315e-04, -3.7579e-05, -2.0611e-03,
         1.7965e-03, -1.2816e-03,  3.6372e-04, -1.2270e-03,  3.0681e-04,
        -1.3703e-03, -4.0610e-04, -5.2209e-04, -1.2069e-04, -4.8806e-04,
         1.5838e-03,  2.4374e-03, -2.1365e-03,  1.2965e-03,  1.0069e-03,
         2.1086e-04, -3.2530e-03,  4.0194e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.1841e-03,  9.2969e-01, -1.3096e-02, -4.1055e-03, -4.0250e-03,
        -1.1060e-04, -3.9345e-03, -1.8575e-03,  6.0424e-04, -1.0516e-03,
         3.9116e-03, -3.5190e-03,  9.3638e-04,  7.5490e-04, -3.0132e-04,
        -2.3546e-03,  5.8042e-03, -6.0143e-03,  6.0337e-03,  3.7155e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0012e-03,  9.3964e-01, -8.7361e-03, -8.4588e-03, -3.8768e-03,
        -9.0487e-04, -9.2781e-03,  3.7781e-03, -4.0743e-05, -1.1311e-03,
         1.5411e-03, -8.7034e-03, -1.4354e-03,  6.4714e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3215e-02,  9.2669e-01, -2.7386e-03,  2.2863e-03, -1.7647e-03,
        -1.5630e-03, -1.7164e-03,  4.5183e-04,  2.5449e-03,  9.1655e-04,
         1.8769e-03,  7.7196e-03,  1.1815e-03, -5.9865e-04, -8.6240e-04,
        -8.0711e-04,  5.7240e-04,  2.6318e-03,  1.6784e-03,  1.2684e-03,
         2.5561e-04,  2.6819e-03,  2.6123e-03, -4.1369e-04,  6.8071e-04,
        -1.8278e-03,  1.0901e-03,  2.2015e-03,  2.6891e-03,  2.6502e-03,
         7.9774e-04,  9.0150e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.5825e-03,  9.1133e-01,  5.0977e-04,  1.3149e-02, -9.8085e-03,
        -7.9038e-04,  3.8664e-03, -2.7457e-03,  1.1012e-03,  2.8313e-03,
         1.6497e-03,  1.0313e-02, -2.5131e-04, -4.9915e-04, -2.1530e-03,
         7.5771e-04,  5.2581e-04, -6.4758e-03, -6.8719e-03, -1.6790e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2653e-02,  9.2838e-01,  8.8678e-04, -5.5391e-03,  2.3927e-03,
        -1.3486e-04,  7.3197e-05,  2.5591e-04, -2.8844e-03, -5.5471e-05,
         3.2332e-03,  3.9108e-04,  2.1854e-04,  5.3488e-04, -8.3517e-04,
        -3.5947e-04, -4.1154e-04, -5.4177e-04, -2.0466e-03, -1.7054e-04,
         1.3381e-03, -2.7839e-03, -1.9002e-03,  1.6845e-03,  2.8954e-03,
        -2.3216e-04,  2.2798e-03,  9.0004e-05, -3.9993e-03, -1.2337e-03,
         4.5377e-04,  1.4461e-03,  6.8888e-04,  2.3657e-03, -4.7783e-04,
         3.4844e-03, -2.4308e-04,  8.8768e-04,  9.0237e-04, -4.0557e-04,
        -1.5361e-03, -4.2345e-04,  1.7579e-03, -3.7845e-04, -2.9982e-04,
         3.1173e-03,  6.9507e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.8027e-03,  9.6162e-01,  7.4128e-03, -1.0286e-03,  1.1432e-03,
         3.0621e-04, -9.9553e-04, -2.7185e-03,  5.1105e-03,  4.2837e-03,
         1.8633e-03,  3.5943e-03,  2.2680e-03, -1.1521e-03, -1.7028e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7469e-02,  9.0104e-01, -4.0823e-03, -4.7814e-03, -4.0776e-03,
        -3.3550e-03, -9.7703e-04,  6.9250e-04, -9.9878e-04,  3.9589e-04,
        -1.5448e-03, -8.7250e-04, -1.0764e-03,  1.9807e-03, -1.8697e-03,
        -2.2357e-03, -3.0827e-03,  7.4028e-04,  8.2361e-04, -7.8511e-04,
        -3.4128e-04, -1.7334e-03, -1.8523e-04,  1.3818e-03, -2.0707e-03,
        -1.6208e-03,  7.5427e-04,  1.9270e-02,  1.8315e-03,  1.3728e-04,
        -2.3761e-04, -4.5841e-03, -2.8428e-04,  3.1371e-04, -2.0364e-05,
        -9.0423e-04, -5.8814e-04, -6.1692e-04, -5.1746e-04, -6.8737e-04,
         8.2762e-05,  2.9725e-04,  1.3253e-04, -1.0730e-03,  1.3940e-03,
        -6.3630e-04,  8.7960e-04,  4.5460e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 2.9905e-03, -9.4244e-01, -2.8675e-03,  7.4450e-03,  3.1675e-04,
         1.1220e-02,  1.8697e-03,  1.3245e-03,  4.1550e-04,  4.5381e-03,
         1.1767e-03,  2.9898e-03,  1.5905e-04,  1.2793e-03,  5.3672e-04,
         1.1990e-03,  2.3911e-03,  2.0676e-03,  3.3165e-04,  1.4004e-03,
         2.2281e-03,  5.9622e-04,  2.3202e-03, -3.9586e-04, -2.1074e-03,
         3.3962e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8604e-03,  9.5754e-01, -1.3749e-03, -9.0465e-04,  6.7651e-04,
        -7.3917e-04,  2.1741e-03,  3.1880e-03,  3.2260e-03, -6.1870e-04,
         5.7662e-03, -8.9544e-04,  7.8287e-04, -7.1625e-04, -5.5297e-04,
        -1.6910e-03,  6.7819e-03, -1.9461e-03,  1.8215e-03,  5.4872e-04,
         5.1944e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5481e-03,  9.3081e-01, -6.1366e-03, -2.9849e-03,  2.0025e-04,
        -7.1293e-03, -1.8992e-03, -1.5794e-03,  1.2237e-03,  8.5369e-05,
         9.4248e-04,  4.9269e-04, -5.0188e-04, -1.7599e-04,  2.9349e-03,
        -3.4583e-03,  2.3504e-03,  1.8741e-03, -1.1421e-03,  1.1460e-03,
        -1.3994e-03,  4.9186e-04, -4.8215e-03,  3.1491e-03,  1.6049e-03,
        -6.9521e-04, -2.2249e-03,  3.3794e-04,  4.8280e-04, -4.9391e-04,
        -3.8093e-03, -1.6597e-03,  2.3303e-03,  1.1007e-03, -2.8632e-03,
        -1.9159e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9300e-03,  9.3600e-01, -1.5871e-03,  1.8797e-03,  3.4743e-04,
         9.1082e-03,  2.2387e-03,  3.7232e-03,  4.9945e-05,  2.2571e-03,
         6.2865e-03,  2.0358e-03, -8.2757e-04,  1.6653e-03, -1.8461e-03,
         8.2765e-04,  3.3278e-03,  4.7671e-03, -3.7112e-03,  8.5884e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1631e-03,  9.5011e-01, -3.5939e-03,  2.0213e-03,  1.5032e-03,
         6.3152e-04, -2.2729e-03,  4.9165e-04,  8.3816e-05,  1.1000e-03,
         5.5664e-03, -5.8706e-03, -3.0077e-03, -2.3214e-03,  4.5347e-03,
         8.9458e-04, -2.9385e-04, -9.6324e-03,  3.0473e-04, -3.3326e-03,
         1.2737e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1443e-02, -9.5558e-01, -6.5881e-04, -5.7688e-03,  4.3563e-03,
         1.8751e-03,  2.1046e-03,  8.3952e-04, -1.0553e-02, -2.1969e-03,
        -1.2883e-04,  9.5218e-05, -4.4008e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0504e-02,  9.1712e-01, -9.4524e-04,  1.0418e-03,  6.3466e-04,
        -2.8820e-03, -4.0628e-03,  1.0526e-03, -1.2746e-03, -2.4835e-03,
        -2.7513e-03, -3.6734e-03,  2.4532e-04,  2.9650e-04, -3.8204e-03,
        -1.8378e-03, -2.0125e-03, -1.4821e-03, -1.6620e-03, -2.1599e-03,
         6.9564e-04,  2.1325e-03, -4.7567e-03, -2.5378e-03,  6.7692e-04,
         8.7472e-05, -4.5878e-03,  1.0966e-04,  6.4839e-04, -6.6342e-04,
         6.3893e-04,  6.1240e-05,  2.9164e-03, -1.1273e-03, -1.0677e-03,
        -2.9867e-04,  5.0476e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2104e-02,  9.3673e-01,  9.9909e-03,  5.4833e-03,  1.3055e-03,
        -3.4522e-04,  4.6341e-04,  1.4568e-04,  1.7493e-04,  4.1879e-04,
        -1.5295e-03,  1.7092e-03, -2.0241e-03, -1.1814e-03,  5.2734e-04,
        -2.6041e-03, -1.4627e-04, -1.4469e-03, -8.4763e-04, -3.4619e-03,
         1.3026e-03, -1.4427e-03, -2.6682e-03,  1.9454e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0344e-02, -9.2081e-01, -1.2311e-02, -4.2843e-03, -1.6363e-03,
         9.2553e-04, -1.1023e-04, -2.5881e-03, -1.5722e-03,  5.1240e-03,
         3.7221e-03, -2.7078e-03,  8.8274e-03, -1.7314e-03, -3.3763e-03,
        -1.6565e-03,  1.9903e-03,  2.4214e-03,  6.3759e-04, -9.1392e-04,
         1.1108e-03,  2.7664e-03,  3.0645e-04,  8.8943e-04,  3.0213e-04,
         1.0161e-03,  1.7572e-03,  7.8178e-05,  1.0133e-03,  1.3901e-03,
         6.7179e-04,  1.0067e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4115e-02,  9.0810e-01,  7.9026e-03,  2.7479e-03,  1.1926e-03,
        -1.4159e-02,  5.8815e-04,  6.7394e-04, -3.6255e-03, -1.5304e-04,
        -2.1649e-03,  1.5387e-03,  1.1338e-04,  1.7583e-03, -1.2440e-03,
         9.7977e-04,  1.9978e-03, -1.3211e-03,  2.6168e-03,  7.8978e-04,
         1.7787e-03,  3.3314e-03, -4.7129e-04, -1.4295e-03,  3.1809e-04,
         4.2974e-03,  2.1425e-03, -5.3569e-04,  8.8859e-04,  4.6503e-04,
         6.5589e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0386e-03,  9.3258e-01, -1.7914e-03, -8.1970e-04, -5.3388e-04,
         4.6812e-04,  5.9844e-03, -7.9461e-03,  1.7399e-03, -2.6925e-03,
        -4.2517e-03, -4.5857e-03,  1.7403e-03, -7.4465e-03, -2.2331e-03,
        -1.4173e-03,  1.3909e-03, -7.7143e-04,  3.6404e-04, -5.1426e-05,
        -6.9986e-03, -5.8346e-04, -7.5750e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0370e-03,  9.4803e-01,  1.0482e-02, -3.4507e-03, -3.8267e-03,
        -1.0070e-03, -2.3320e-03, -1.4635e-03, -3.0148e-03,  6.4769e-04,
        -3.2907e-04, -2.2869e-03,  2.7686e-05, -1.9197e-03, -1.9221e-03,
        -5.4590e-04, -2.4789e-03, -7.6945e-04, -3.8256e-04,  1.8158e-04,
        -1.4325e-03, -4.7645e-04,  1.9615e-03, -3.1805e-04, -1.0806e-03,
        -1.5524e-03,  1.7990e-03,  2.2477e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 2.6091e-03,  9.5494e-01,  7.4353e-03, -2.0485e-03,  2.8301e-04,
        -2.7195e-03,  7.5193e-04, -4.4244e-04, -4.2176e-03, -7.4653e-04,
         2.9690e-03, -2.2543e-03,  1.9580e-03,  6.9249e-04, -5.7531e-04,
         1.3534e-04,  1.0767e-03, -1.5017e-03, -3.3430e-04, -3.4392e-03,
         1.2241e-03,  1.2213e-05, -1.0165e-03, -1.2127e-03,  4.1884e-04,
         7.5547e-04,  2.6780e-03, -1.4227e-03,  1.2864e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0929e-02,  9.4808e-01, -5.3596e-03,  1.6345e-03,  1.1983e-03,
         4.4442e-03,  1.2392e-03,  2.4408e-03, -1.2355e-03,  4.0223e-04,
        -1.5016e-03,  3.8167e-04, -1.7033e-03,  1.8141e-03, -8.5497e-04,
         8.0426e-05,  1.1048e-03,  1.2269e-03,  4.6528e-03,  6.9988e-04,
         7.4926e-04,  6.6455e-03, -8.8901e-04, -7.3358e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.6730e-03,  9.2769e-01, -6.0190e-03,  1.9251e-03, -5.0975e-05,
         5.3883e-03,  1.1536e-03,  2.6381e-03, -3.0866e-03,  1.2050e-03,
         1.5534e-03, -3.9632e-03,  7.1969e-03,  3.1798e-03, -5.5699e-04,
         1.2677e-03,  3.1724e-03,  9.7040e-04,  4.8940e-03,  1.4405e-03,
         1.7441e-03, -5.6096e-04,  1.6340e-03,  3.1814e-03,  2.1057e-03,
         4.4110e-06, -3.4283e-03,  1.6596e-04, -3.4449e-04,  1.0443e-03,
         1.1137e-03,  2.1266e-03,  5.2092e-04, -8.9431e-08], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5795e-03,  9.5986e-01, -2.9918e-03, -3.3615e-03,  1.0735e-04,
        -1.4612e-03,  2.5595e-03, -1.4581e-03, -1.1823e-03, -2.2214e-03,
        -7.1017e-04,  4.4791e-03, -3.2746e-04, -6.6314e-04,  7.9502e-04,
         3.9030e-05, -4.1534e-03, -3.7471e-05,  9.1682e-04,  3.7609e-03,
        -1.1117e-03,  2.2184e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2491e-03,  9.3723e-01, -1.1497e-02,  1.4407e-03, -9.0545e-03,
         1.5102e-03,  3.4018e-04,  3.5011e-03,  2.4477e-03, -1.8976e-03,
         5.7977e-03, -1.9948e-03, -1.6077e-03,  2.4845e-03, -5.4103e-04,
         5.5026e-04,  1.2759e-02,  1.0986e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5955e-02,  9.4878e-01,  4.5560e-03,  1.6393e-03, -1.2166e-03,
        -2.2624e-03,  2.9498e-03,  2.8370e-03,  3.3292e-04,  1.2596e-03,
        -3.6922e-04, -5.2927e-04, -2.5315e-04,  1.7781e-04,  4.2685e-05,
        -9.0504e-04, -1.2048e-03, -3.8807e-03, -5.0656e-04, -3.8462e-03,
         1.5793e-03, -3.8856e-03, -1.0313e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.9079e-03,  9.5871e-01,  7.3987e-03,  2.0188e-03, -4.6803e-03,
         2.9111e-03,  1.4890e-03, -5.9676e-04, -5.8350e-04, -1.1342e-03,
         1.1800e-03, -1.2972e-03, -9.8593e-04,  3.4092e-03, -2.0844e-04,
         7.2331e-04,  5.7665e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.1348e-03,  9.3452e-01, -6.4138e-03,  7.1409e-03,  1.3285e-03,
        -2.0857e-03, -2.3822e-03, -2.8268e-04,  1.1167e-03,  8.0210e-04,
        -2.2145e-03, -3.2441e-03, -2.5550e-03, -1.6284e-03,  6.8293e-03,
        -5.9755e-03, -2.1059e-03, -1.1238e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9797e-02,  9.1894e-01,  2.5450e-03, -1.6071e-03,  1.1590e-03,
         2.7703e-05, -7.1737e-04, -7.4026e-03, -8.7607e-04,  1.3387e-02,
         3.5856e-03, -4.3408e-03, -9.0340e-04, -2.3047e-03,  4.2113e-04,
         1.9018e-03, -6.2996e-04,  8.5606e-03,  2.4110e-04, -7.0172e-04,
        -1.9670e-03, -1.2919e-03,  1.6092e-03, -2.5065e-04,  5.7772e-05,
        -4.7775e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4426e-02,  9.2781e-01, -5.6645e-03, -1.5392e-03, -7.3299e-03,
        -1.7367e-04, -1.6898e-03,  2.4070e-03,  6.8798e-03, -2.9480e-03,
        -1.3064e-03, -1.6965e-03, -1.3213e-03, -5.4067e-03, -1.7168e-03,
        -1.2834e-03, -1.4353e-03, -2.7207e-04,  8.3368e-04, -1.0001e-03,
        -1.1955e-03, -1.3230e-03, -1.9639e-03, -1.0369e-03, -1.4068e-03,
         4.5091e-04, -1.0295e-03, -8.3885e-04,  2.8793e-04, -1.3429e-03,
         1.9861e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9671e-03,  9.4828e-01,  1.9780e-05, -2.7770e-03, -1.1382e-02,
         6.5037e-04, -1.9230e-03,  5.0100e-03,  1.7146e-04, -7.3636e-03,
         2.7417e-03, -2.1948e-03, -2.6548e-03, -2.5709e-04, -1.3237e-03,
         1.2460e-03,  1.4710e-04,  9.8890e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2947e-02,  9.5694e-01,  6.5099e-03,  1.5432e-02,  6.6649e-04,
        -2.6372e-05, -2.9658e-03, -4.5150e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 5.5410e-04, -9.4278e-01, -5.0675e-04,  7.9687e-04, -4.9262e-04,
        -1.2533e-03,  8.3963e-04,  3.9148e-04,  2.2585e-03,  6.0493e-04,
         1.9114e-03,  4.8866e-03,  3.1183e-04,  1.0614e-03,  5.8380e-04,
         9.4064e-04, -1.2088e-03,  3.0467e-03,  7.3911e-04,  2.8866e-03,
         2.7371e-03, -1.1625e-04,  2.3848e-03, -4.9631e-04, -2.2827e-03,
        -1.5627e-03, -2.3204e-03, -3.1636e-03,  5.0627e-04,  2.1920e-04,
        -2.6449e-03, -8.7435e-04, -2.7866e-04,  5.3688e-04, -3.0293e-03,
         3.2958e-04, -1.8960e-04,  3.7074e-04, -7.8986e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3240e-03,  9.3845e-01, -1.1300e-02, -4.4079e-03, -7.4905e-05,
         1.2495e-05, -8.8453e-04, -1.2671e-03, -1.2728e-04,  3.2101e-03,
        -1.4406e-04, -1.5567e-03, -6.6627e-04, -2.6620e-03,  2.4350e-04,
        -4.7418e-03, -1.2429e-03, -6.0098e-04, -1.3701e-04,  5.6314e-04,
         1.2094e-02,  6.8220e-04,  4.3147e-03,  6.4844e-04,  2.2639e-03,
        -1.2567e-03, -3.4801e-03,  1.6463e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0446e-05, -9.6813e-01,  2.0534e-03,  1.8048e-03,  6.3788e-04,
         1.2861e-03,  4.1908e-03, -3.5953e-03, -4.5661e-03, -1.4869e-03,
        -1.9006e-03, -2.2178e-03, -8.1115e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8657e-03,  9.7220e-01,  1.8632e-03, -5.5438e-03,  1.1526e-03,
        -4.6623e-04, -4.1953e-03,  1.7673e-03, -3.2539e-04, -3.5537e-03,
        -3.1898e-03, -9.3714e-04, -3.8047e-04,  1.3935e-03,  1.6747e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2863e-02,  9.2438e-01,  1.6590e-03,  2.0988e-03, -4.0941e-03,
        -3.9835e-04,  3.3670e-03,  9.2704e-05,  8.5413e-04, -7.7497e-04,
        -9.8774e-04, -1.3667e-04, -5.0817e-03,  7.4093e-03,  7.0275e-04,
        -9.5192e-04, -1.1024e-03, -6.1246e-04,  2.5766e-04,  1.6990e-03,
         2.6799e-03, -5.6578e-04,  6.6743e-04, -3.2300e-04, -7.6045e-04,
         1.4026e-03,  1.7184e-03,  2.1459e-03,  7.9702e-04,  1.5362e-04,
        -3.4825e-03,  5.2315e-04,  1.2617e-04,  5.6576e-04,  2.5381e-03,
         7.4944e-04,  4.5429e-04, -2.4057e-04,  1.1267e-03,  6.1064e-04,
         2.2277e-03, -1.6084e-03,  1.5273e-03, -8.9451e-05, -1.3824e-03,
         2.9584e-04, -1.7125e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4821e-03, -9.3297e-01,  1.0215e-02,  4.3493e-03,  9.2477e-04,
         3.2333e-03, -5.2673e-05, -5.7457e-04,  3.0441e-03, -1.3417e-03,
        -2.5702e-03,  1.2157e-03,  4.3886e-04, -7.0553e-05, -7.2565e-04,
        -1.4340e-03, -9.9791e-04,  7.3648e-05,  1.0218e-03, -3.1745e-03,
        -2.8040e-04, -2.0112e-03, -1.5443e-03, -8.5308e-05,  7.6718e-04,
        -1.0024e-03, -5.0553e-03, -1.0434e-03, -8.9151e-04, -9.8184e-04,
         2.4074e-03, -1.1831e-03, -9.9077e-04, -6.9308e-05, -2.7748e-03,
         3.1281e-04, -2.8604e-04,  4.4942e-04, -1.0614e-03, -1.1182e-03,
         2.1442e-04, -3.5554e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8573e-02,  9.2406e-01,  2.7811e-03, -3.6492e-03,  2.3205e-04,
        -7.0628e-03,  2.0637e-03, -4.0105e-03, -7.7995e-04, -3.1020e-03,
        -8.2704e-03, -3.8022e-03, -1.2043e-03,  6.1644e-04,  2.7696e-03,
        -4.3914e-04,  2.9004e-03, -1.6114e-03, -2.1098e-04,  8.0967e-04,
         1.7146e-03, -4.4928e-04, -7.2632e-03, -1.4252e-03, -1.9671e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6590e-02,  9.3137e-01,  5.2757e-03,  1.4113e-03, -1.5285e-03,
        -6.1944e-03,  5.5454e-03, -9.6358e-04,  5.1829e-03,  1.9111e-03,
         1.3710e-03, -2.6175e-03,  8.4603e-04, -3.3416e-03,  5.5813e-04,
         5.2955e-04,  4.5885e-04,  3.8328e-03, -2.3853e-03, -1.0051e-03,
        -3.8435e-03, -2.5566e-03, -6.7921e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4803e-02,  9.4186e-01,  4.4252e-03, -6.4311e-05, -1.3096e-03,
        -5.0046e-03, -1.8602e-03, -5.6773e-04, -1.8462e-03,  1.6342e-03,
        -1.4502e-03,  5.5844e-04,  4.2297e-03,  1.1015e-03,  1.7456e-04,
        -2.6155e-03,  2.5137e-03, -1.0654e-03, -7.0682e-03,  8.5238e-04,
        -1.4654e-03, -1.9016e-03,  1.6272e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0011e-03,  9.3742e-01, -1.0077e-02, -4.9832e-03,  2.5849e-03,
        -2.3635e-03,  3.7190e-04,  1.1782e-03,  4.4833e-03,  2.4711e-03,
        -7.0788e-03, -9.3955e-03, -1.2590e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1979e-03,  9.6208e-01, -1.0679e-02,  2.4311e-03,  1.3860e-04,
        -7.4218e-04, -1.3674e-03,  3.7773e-03,  5.4217e-03,  1.6874e-03,
        -1.7701e-03, -6.0606e-03,  2.4233e-03, -2.2593e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6733e-02,  9.2525e-01,  1.5908e-02, -1.3480e-03, -2.0448e-03,
         9.8230e-04,  7.5529e-03, -2.8566e-03,  1.6901e-03,  5.9656e-03,
         1.1684e-03,  4.5559e-03,  1.4300e-03,  5.3675e-03, -1.5599e-03,
         2.7270e-04, -1.3571e-03, -5.4880e-04,  3.4126e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 1.3630e-02,  9.4499e-01, -5.2581e-03, -1.1628e-04, -1.0146e-03,
         6.5538e-04,  8.9266e-04, -1.4361e-03, -3.9581e-04, -1.0644e-03,
        -5.7257e-04,  3.1356e-03,  1.2173e-03,  2.3862e-03,  3.2787e-03,
        -5.5102e-04,  2.2941e-03, -1.6078e-04,  1.9969e-03,  5.8408e-03,
         1.7701e-03, -1.0288e-03, -2.1380e-03,  1.0315e-03, -2.8149e-03,
         3.3001e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4644e-02,  9.1450e-01, -1.8650e-03, -1.3317e-03, -3.1580e-03,
        -1.4944e-03, -6.3274e-03, -2.7021e-03, -6.7401e-03,  1.4972e-02,
         1.6704e-03, -1.9948e-02, -2.6254e-03, -2.3474e-03, -5.6194e-03,
         5.0995e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.4426e-04,  9.3593e-01,  4.1001e-03,  1.7570e-02,  1.4370e-03,
         8.6434e-03, -2.7902e-04,  9.6729e-04, -4.5019e-03,  5.3883e-04,
        -2.6184e-03, -8.3139e-03,  7.4862e-04,  4.9310e-03, -2.2500e-03,
        -5.1446e-05,  6.4774e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6096e-02, -9.5074e-01,  4.1229e-03, -1.0078e-02,  7.8332e-04,
         5.3390e-03, -2.4769e-03,  7.5157e-04, -9.6076e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.9134e-03,  8.8782e-01, -2.5602e-02, -4.2836e-03, -3.6606e-03,
         1.2260e-03,  1.0513e-03,  1.6673e-03, -6.3264e-03, -7.1693e-04,
         8.6305e-04,  1.9588e-03,  2.2275e-03,  2.6754e-04, -2.3604e-03,
        -5.8950e-04,  2.0839e-03,  2.4454e-03,  2.0436e-03,  2.5844e-03,
         2.2075e-03,  1.1149e-03,  2.2696e-03, -1.5393e-03,  1.8205e-03,
        -4.6533e-04, -8.9305e-04, -1.7851e-04,  6.3983e-04,  1.7175e-03,
         6.6491e-04,  1.4086e-04,  4.2705e-03,  1.7773e-03,  1.9712e-03,
        -3.4880e-05, -1.4406e-03,  1.8767e-03, -2.2081e-04, -1.0833e-03,
        -1.6631e-03, -6.1577e-03, -1.4865e-03, -2.9882e-03, -2.4862e-04,
        -1.4408e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8882e-04,  9.5364e-01, -3.7006e-03, -4.1909e-03,  1.6822e-03,
         5.5691e-04,  3.7408e-05, -1.0327e-03,  1.5237e-03, -7.1167e-04,
        -9.9557e-05, -3.8880e-04, -5.4641e-03, -9.0985e-04,  3.6910e-04,
         3.0417e-03,  1.1477e-03,  3.0270e-03,  4.8247e-04,  9.6502e-04,
        -6.3233e-04,  1.3088e-03,  1.2358e-03, -1.3844e-03,  2.0901e-03,
        -3.7093e-03,  1.4374e-03,  1.2753e-03,  3.3404e-03,  4.2787e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9993e-03, -9.6272e-01, -8.3425e-04, -1.6094e-02,  2.8590e-03,
         6.6342e-03, -5.8546e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.9292e-03,  9.6495e-01, -2.8472e-03, -3.8210e-04,  4.7606e-03,
         2.2716e-03,  1.4788e-03,  8.4858e-03,  5.7471e-04,  3.1518e-03,
        -4.1637e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4225e-03,  9.2131e-01,  1.3581e-02,  2.0491e-03,  1.0424e-02,
        -4.3277e-03,  6.6933e-03, -4.2463e-03,  3.8562e-03, -1.5484e-04,
        -3.5805e-03,  1.9314e-02, -2.4844e-03,  3.5605e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9665e-03, -9.4332e-01,  2.2696e-03,  2.7069e-03, -2.8688e-04,
        -7.5788e-04, -1.2731e-03, -1.3312e-03, -1.6781e-03,  6.8254e-04,
         3.6752e-04, -2.9884e-03,  1.4875e-03,  6.5421e-03,  9.5967e-04,
        -4.8064e-04, -5.6855e-05,  7.7950e-04, -1.3464e-03, -5.0060e-03,
        -5.7120e-03, -2.3963e-03,  2.9124e-04,  1.1409e-03,  5.2762e-04,
        -9.9126e-04,  4.4085e-04,  1.5195e-03, -3.3175e-04, -2.2457e-03,
        -1.5347e-03,  1.6854e-03,  9.5171e-04, -2.9418e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.2043e-03,  9.2161e-01, -9.7980e-03,  1.1627e-02, -1.7156e-03,
         1.0408e-03, -9.0360e-04, -4.9084e-04,  1.1783e-05,  5.7799e-04,
        -2.1157e-04,  9.5602e-04,  1.7964e-03,  4.2037e-04,  1.1973e-02,
        -2.1401e-03, -1.0193e-02,  3.0981e-04, -3.0312e-03, -3.8344e-04,
        -2.1674e-03,  7.9134e-04, -8.0181e-04, -1.7554e-03, -1.0229e-03,
        -1.5713e-03, -1.4740e-03,  2.0195e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6678e-03,  9.3836e-01, -6.0043e-03, -8.4954e-04, -1.6832e-03,
         1.5517e-03, -1.9532e-03,  1.3570e-03, -1.2961e-04,  2.2708e-03,
         7.3819e-04, -3.5992e-03,  3.7710e-03,  8.1362e-03,  4.1634e-03,
        -1.8985e-03,  1.1356e-03, -1.5128e-03,  2.0157e-03, -5.4091e-04,
        -2.6687e-03,  1.7218e-03, -2.5261e-03, -7.7499e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-1.0905e-02, -9.3554e-01,  4.0618e-03, -4.3565e-03, -1.2961e-03,
        -1.8437e-03, -3.2787e-03,  8.7090e-04, -4.3035e-03, -1.5585e-03,
         2.5640e-03,  1.6708e-03, -3.0346e-03, -2.3458e-03, -7.7719e-04,
         2.4405e-04, -6.8717e-04, -1.2102e-03, -8.7281e-04, -4.2333e-03,
         1.4534e-03,  5.3403e-05, -2.4943e-03, -3.1784e-04,  1.0067e-03,
        -2.4469e-03,  2.0070e-04,  6.9880e-04, -7.4639e-04, -6.7211e-04,
        -4.2558e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1231e-02,  9.7165e-01,  7.0894e-03, -2.7807e-03, -3.1057e-04,
        -2.7581e-04, -1.3348e-03, -2.1162e-03,  1.2834e-03,  1.3116e-03,
         4.7243e-04, -1.4676e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.1278e-02,  8.9836e-01,  3.4968e-03, -8.7388e-03, -3.5978e-03,
        -3.0373e-03, -5.0472e-03, -2.9702e-03, -9.8356e-04,  2.9406e-03,
         2.5065e-03, -1.0781e-02,  2.6632e-04, -1.8070e-03,  4.1862e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4486e-02, -9.3633e-01, -9.4570e-03,  5.1616e-03,  1.3336e-03,
        -1.0608e-03, -1.9333e-03,  2.0980e-03, -2.7857e-05,  1.9409e-04,
         1.8563e-03, -2.8347e-03,  8.0550e-04, -4.5368e-04, -3.2312e-03,
         5.2127e-04, -1.8428e-03, -8.5052e-04,  2.9930e-03,  1.3528e-04,
         7.0538e-04, -3.9944e-04, -1.0294e-04, -2.4973e-04, -1.8131e-04,
        -2.8472e-03, -5.9052e-04,  7.2532e-04, -7.8127e-04, -8.9204e-04,
         2.6045e-04, -1.2766e-03, -3.3836e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4309e-02,  9.3995e-01, -2.0937e-02, -2.1865e-04, -2.0543e-03,
        -2.1535e-03, -1.8161e-03,  4.9557e-04, -1.7175e-03, -2.6871e-03,
        -3.7778e-03, -2.3278e-04, -3.0252e-03, -1.5701e-03,  3.7993e-03,
         1.2531e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1151e-02,  9.5402e-01,  6.4400e-03, -5.2027e-03,  6.8977e-03,
        -3.8093e-03,  2.6147e-03,  8.0769e-04, -9.0583e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0073,  0.9560, -0.0062,  0.0134,  0.0171,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9246e-02, -9.3028e-01, -8.3859e-04,  1.4516e-03, -1.8393e-03,
         4.8532e-03,  1.3474e-03,  1.4394e-03, -2.6234e-03, -1.7595e-03,
        -1.4904e-03,  1.6903e-03,  2.3909e-03,  4.3384e-05, -1.8456e-03,
        -1.5004e-03,  7.8060e-04,  5.4696e-05, -2.2328e-03,  8.6542e-04,
        -1.3835e-04, -1.8396e-03,  4.3587e-04, -2.7307e-03,  4.2422e-04,
        -1.5304e-03,  4.3294e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0458, -0.9243,  0.0123, -0.0044, -0.0043, -0.0044, -0.0044,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8430e-02, -9.4626e-01, -1.8571e-03, -1.7746e-03,  8.3533e-05,
         8.1844e-04,  4.6110e-04,  4.4453e-04,  2.3912e-03, -8.8494e-04,
        -9.0065e-04,  2.3851e-03,  1.6977e-03, -2.5783e-03, -2.5395e-04,
         1.1516e-03,  3.0758e-04, -1.1720e-03,  3.2097e-04,  1.0353e-03,
         2.4614e-03, -2.3311e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8993e-03,  9.4454e-01,  9.3936e-03,  1.7541e-03, -3.1006e-03,
         3.2184e-03, -2.3783e-03, -1.5088e-03, -2.7482e-03, -3.6302e-03,
        -5.7312e-03,  6.4277e-04, -4.0359e-03, -5.3386e-03,  3.2875e-04,
        -2.8111e-03, -5.5514e-04, -1.2102e-03,  4.1778e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6120e-02,  9.2714e-01, -6.1567e-03, -9.8149e-04, -3.7798e-03,
         1.1003e-03,  9.2128e-04, -8.3463e-03,  3.0880e-03,  2.2187e-03,
        -1.8164e-03, -2.4892e-04,  3.4689e-04,  1.2829e-03,  2.5956e-03,
        -1.5096e-03, -2.8546e-03, -6.8793e-04,  1.4610e-03,  8.1543e-04,
         2.0348e-03,  3.6383e-04,  2.1687e-03, -1.2249e-03,  1.4137e-03,
        -1.8961e-04, -1.9580e-03, -6.4284e-04, -2.1394e-03,  3.9909e-04,
        -3.9931e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-2.6669e-02,  8.9270e-01, -7.4884e-03, -4.5952e-03, -2.4826e-03,
         5.3553e-04,  1.1963e-03,  6.1255e-04, -2.3987e-03, -2.5337e-04,
        -2.7817e-04, -7.3732e-04,  4.2513e-04,  1.1463e-02,  6.5070e-04,
         1.8118e-03,  9.2963e-03,  2.2960e-03,  5.0747e-03,  6.6275e-04,
         7.7593e-04, -1.0829e-03,  3.4006e-03,  5.4563e-04,  1.6507e-03,
         1.2495e-04, -6.0802e-04, -4.8968e-03, -7.8609e-04,  8.0233e-04,
        -1.1682e-02, -2.0192e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5245e-03, -9.4914e-01, -2.7034e-04,  2.7750e-03, -3.5372e-03,
        -1.9838e-03,  8.6830e-04, -4.0409e-03, -3.6568e-03,  2.5153e-04,
         8.4768e-04, -3.1748e-03, -1.2516e-03, -4.7319e-03, -5.2092e-04,
        -1.3127e-03, -3.7457e-03, -9.1045e-04,  2.9660e-03,  2.2734e-04,
         4.7580e-04, -7.7831e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.4434e-03,  9.3609e-01, -2.9865e-03, -5.2270e-03, -9.3543e-04,
         3.8494e-03, -7.8001e-04, -2.2982e-03, -6.1952e-04,  1.0370e-03,
         1.3763e-03, -8.2087e-04, -2.3380e-04, -1.2872e-03, -8.6641e-04,
         1.6866e-03, -1.2398e-03,  4.6415e-03, -2.4399e-03, -2.6937e-03,
        -2.9932e-03, -4.2338e-03,  5.0836e-04, -2.2138e-03, -2.5581e-04,
        -2.5146e-04, -3.2285e-04, -4.1636e-03, -4.3251e-03,  3.1822e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1354e-02,  9.5635e-01, -2.9334e-03,  1.2476e-04, -5.0811e-04,
         2.6847e-03, -6.2525e-03,  2.3670e-04, -1.1069e-03, -1.5940e-03,
         2.3601e-03,  3.8992e-03,  1.8054e-03,  2.6701e-03, -2.9416e-03,
        -1.8836e-03, -8.3531e-04,  4.5778e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2390e-02,  8.9890e-01,  2.9898e-02, -3.9926e-04, -3.3026e-03,
        -1.9117e-03,  1.1209e-03,  1.8130e-03,  4.9862e-04,  1.8822e-04,
         1.7409e-03,  7.1715e-04,  1.9860e-03,  9.7364e-04,  5.7048e-04,
        -1.5418e-03, -5.7484e-04, -2.0278e-03, -9.0992e-04, -1.1885e-03,
        -4.1460e-04, -4.2175e-03,  1.7420e-03,  1.7871e-03,  8.7784e-04,
         1.5491e-03, -6.5585e-03,  2.5422e-04,  1.9809e-03, -1.3059e-04,
         3.1516e-04,  5.7409e-04,  6.2113e-03,  3.0514e-03,  6.4052e-04,
         3.3804e-03,  3.6643e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.7465e-03,  9.5552e-01,  7.3364e-03, -2.4503e-03, -2.3237e-03,
        -1.5072e-03,  5.6469e-05,  1.8014e-04,  1.7190e-04, -1.1122e-03,
         1.5585e-03, -3.8871e-03,  9.9135e-04,  2.6023e-04,  3.5153e-03,
        -7.2015e-04, -1.5991e-03,  4.2153e-04,  1.0522e-03,  1.1099e-03,
        -1.4808e-03, -1.6528e-03,  1.8463e-03, -2.4961e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3226e-03,  9.5015e-01, -7.2647e-03, -2.5559e-03, -9.0597e-04,
         2.1558e-03, -1.0733e-03, -1.0829e-03,  1.3837e-03, -8.5465e-04,
        -3.6644e-04, -1.1253e-03,  2.2071e-03, -1.4477e-03, -6.3319e-05,
        -9.5356e-04, -2.7115e-04, -2.7827e-04,  7.5083e-04,  5.1174e-05,
         7.8522e-04, -1.4508e-03,  8.2925e-05,  8.1913e-04, -1.5707e-03,
        -6.4997e-04, -2.2276e-04, -1.4778e-04, -1.9693e-04, -7.2295e-04,
        -4.4811e-03,  6.9959e-04,  6.6061e-04, -9.9810e-05,  1.1710e-03,
        -5.3352e-04,  3.4848e-04, -1.3634e-04,  1.0533e-03, -3.4784e-04,
         2.4702e-04,  5.2432e-04,  8.2249e-04,  3.7900e-03, -2.1392e-04,
         9.5779e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.4156e-03,  9.0955e-01, -2.0077e-02, -3.9719e-03, -1.5509e-02,
        -2.7054e-03, -8.3897e-04, -2.0347e-03,  9.2783e-04,  2.2153e-03,
        -1.2848e-03, -2.6092e-03, -3.9016e-03, -1.4018e-03,  4.1472e-03,
         1.2895e-03, -2.4309e-03,  3.9862e-03,  3.2179e-03,  1.2580e-03,
         2.1745e-03,  3.3847e-03,  2.6697e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.9773e-03,  9.2829e-01, -8.1010e-03, -4.5491e-04,  1.6968e-02,
         8.2638e-03,  6.5638e-05,  7.8452e-03,  8.5342e-03, -3.1399e-03,
         8.3586e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1903e-02,  9.5942e-01, -2.7609e-03, -4.3342e-03,  4.1444e-03,
        -3.0095e-03,  1.4095e-03,  4.1916e-03, -1.3021e-03, -1.9683e-03,
         2.0726e-03, -5.7159e-04,  1.7680e-03, -1.1399e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1734e-02,  9.4377e-01,  6.6694e-04,  1.1563e-03, -2.3249e-03,
        -4.7539e-03, -2.1374e-03, -1.3441e-03, -2.5051e-03, -1.1726e-02,
         1.6117e-02,  1.7605e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6472e-03,  9.3452e-01,  6.8609e-05,  1.0992e-02,  7.0981e-04,
         2.3276e-03, -1.3910e-03, -1.4990e-03, -1.2725e-03,  1.2597e-03,
         9.1724e-05, -1.1234e-03,  3.6077e-04,  2.6814e-04, -6.6239e-05,
         7.6081e-04, -8.8068e-06,  1.3872e-04,  5.8174e-04, -2.2485e-03,
        -3.7689e-04,  3.0368e-04,  1.5497e-03, -1.1701e-03,  2.1937e-03,
        -1.1156e-03,  8.0584e-03,  3.2043e-04, -1.6667e-04,  7.1078e-04,
        -4.0389e-05,  1.6138e-03,  9.9631e-04, -7.7070e-04,  7.9670e-04,
        -7.1908e-04, -2.4567e-04,  9.0336e-04,  1.3877e-03,  7.3585e-05,
        -2.3769e-04, -8.4162e-04, -1.9916e-04,  4.4742e-04,  2.1438e-04,
        -2.7394e-04,  9.8343e-05, -1.0836e-03, -5.8823e-04, -1.4231e-03,
         5.8437e-04, -2.4766e-04, -6.2966e-04, -2.6758e-03,  2.6070e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 5.1349e-03, -9.4843e-01,  1.4310e-03,  1.3587e-03, -2.5127e-03,
        -4.5115e-03,  2.1813e-04, -3.0423e-03, -5.8984e-03,  2.7135e-03,
         1.4725e-03, -2.8501e-03,  2.2791e-03,  4.4862e-04, -1.8859e-03,
        -1.0424e-03,  1.6644e-03, -4.6562e-03, -1.9053e-03,  5.1868e-03,
         2.2157e-05, -1.1048e-03, -2.2754e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.7898e-03,  9.4280e-01,  1.5559e-02, -3.8450e-03,  3.3356e-03,
         2.1440e-03,  3.7413e-03, -2.0816e-03, -1.2331e-03,  1.0431e-03,
        -1.4184e-03, -1.4629e-03, -2.6941e-04,  6.1963e-03,  5.8271e-04,
        -8.0729e-04, -1.9441e-03,  2.7435e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.7607e-05,  9.5616e-01, -7.1890e-03, -4.7627e-03, -2.5158e-03,
        -9.7467e-04,  2.4862e-03, -5.1132e-03, -3.6513e-04, -7.0379e-04,
         4.7095e-03, -9.0209e-04,  2.9998e-03, -2.6020e-03,  1.8171e-04,
         5.7642e-03, -1.3614e-03, -1.1173e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.3604e-04,  9.4616e-01, -1.5068e-02,  2.6077e-03, -5.9695e-04,
         3.7940e-03, -3.4834e-03, -5.2450e-05,  3.6103e-03,  2.5357e-03,
        -1.1038e-03,  5.9049e-03,  5.1599e-03,  4.5548e-03, -4.5331e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3401e-02,  9.0456e-01, -7.8299e-03,  9.3323e-03,  5.7231e-04,
        -4.8621e-04, -5.1888e-03, -2.7684e-03, -2.8923e-03, -2.4844e-03,
         8.5240e-04, -1.2530e-03, -4.2236e-04,  7.8455e-04, -2.3285e-03,
         1.9315e-03, -2.9408e-03, -2.2363e-03, -5.6743e-04,  6.2282e-03,
         2.9236e-03,  1.8400e-03,  4.8184e-05, -3.3840e-04,  7.0566e-03,
         1.7212e-03,  1.0670e-03, -5.5978e-06, -1.1202e-02,  1.7433e-03,
        -2.1217e-03, -8.6932e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5171e-03,  9.4362e-01,  1.4069e-03,  1.2654e-03,  2.9542e-03,
        -3.5609e-05,  2.9791e-03,  7.2305e-04, -3.6392e-03,  2.2651e-03,
         3.3050e-04, -2.7916e-03, -1.0087e-03, -4.5657e-04, -1.1217e-03,
        -1.1623e-03, -2.8783e-03,  2.1280e-03,  1.4581e-03,  1.0262e-03,
         2.9559e-03,  7.0606e-04, -1.8966e-03, -6.6600e-04, -7.7625e-04,
        -1.3257e-03,  3.9730e-04,  8.3144e-03,  1.1514e-03, -1.5253e-03,
        -8.5031e-04,  3.2471e-03,  1.4232e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.5148e-03,  9.2577e-01,  3.5319e-03,  6.4055e-03,  5.2277e-03,
         4.7648e-06,  3.3167e-03, -5.5080e-05,  4.6662e-03,  3.0727e-03,
         1.7650e-03,  2.7681e-03,  2.9734e-03,  2.6241e-03,  1.5132e-03,
         7.7785e-04,  2.6076e-03, -1.6331e-05, -1.6108e-03, -3.0293e-05,
         2.8103e-04, -1.0032e-03,  4.7428e-04, -1.0315e-03, -1.7197e-04,
         7.0986e-04, -2.3026e-04, -6.8651e-04, -1.5038e-04, -1.0148e-03,
         6.2320e-04,  5.2304e-04,  2.2410e-03, -1.0566e-04,  3.3932e-04,
         1.4432e-04,  2.3551e-04, -5.3979e-04,  9.1445e-04,  8.5892e-04,
         1.2452e-04, -7.2636e-04, -1.3190e-04,  3.9518e-04,  5.1698e-04,
         5.1372e-04,  1.4646e-04,  5.8338e-04, -6.8128e-05, -1.8437e-03,
        -1.1952e-03, -4.5328e-05, -6.7203e-04,  4.0493e-05,  3.8224e-04,
         5.7755e-05,  2.4989e-04, -4.7731e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.3481e-04, -9.5575e-01, -1.6612e-03,  6.2580e-03,  3.4445e-03,
        -6.8303e-04,  2.0312e-04,  5.9507e-04,  1.2875e-03,  1.6579e-03,
        -1.2778e-03,  3.3348e-04,  5.7133e-04, -4.7404e-03, -7.3367e-03,
         1.6361e-03,  4.4776e-04,  6.4827e-04,  1.3236e-03,  3.4730e-04,
         1.9858e-03, -1.3669e-04,  1.9762e-03,  1.4941e-03,  7.5357e-04,
        -2.7148e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4246e-02,  9.0171e-01, -1.8117e-02, -1.2651e-03, -4.0530e-03,
         6.3225e-03, -3.3555e-03,  3.2072e-03, -3.9955e-03, -8.3373e-05,
         1.6176e-04, -1.0517e-03,  4.7168e-03,  3.4894e-03, -3.2112e-03,
        -2.0388e-03, -1.1670e-03, -4.6528e-04, -1.0013e-03, -3.4997e-03,
         1.3003e-03, -6.0967e-03, -1.7493e-03, -9.1244e-04, -2.7790e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.0539e-03, -9.2565e-01, -1.8480e-02,  2.2685e-03, -3.9335e-05,
        -1.6608e-03,  1.4218e-03,  6.5548e-04,  5.6515e-04,  4.4824e-04,
        -7.3678e-04,  3.9646e-04,  2.9401e-03, -2.9220e-03, -1.0752e-03,
         6.9132e-04,  3.0481e-04, -1.2001e-04,  3.5546e-04,  5.3876e-04,
         4.0796e-04,  1.0951e-03, -2.9165e-04,  1.0846e-04, -9.5990e-04,
         1.3093e-03, -1.0231e-03,  2.4455e-04,  4.6415e-04, -7.6015e-04,
        -8.3508e-04,  1.0937e-03,  7.0685e-03,  4.0754e-04,  1.5901e-03,
         2.6144e-03, -3.0277e-04,  5.7320e-04,  4.6396e-04, -2.0288e-03,
        -1.2186e-04,  1.0775e-03, -3.6645e-04, -1.9529e-03, -6.5169e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6617e-02,  9.0826e-01,  8.3216e-03, -9.4996e-03, -1.6544e-03,
        -6.8278e-04, -3.4315e-03, -1.6060e-02,  3.5519e-03, -3.9835e-04,
        -3.9201e-03, -1.4609e-03,  3.0533e-04,  3.2057e-03,  2.4599e-03,
         4.6447e-03,  4.5941e-03, -3.6991e-03, -1.0521e-04,  2.1583e-03,
        -4.4183e-04, -1.6999e-03,  2.8315e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0442,  0.9097,  0.0060,  0.0015, -0.0020,  0.0017, -0.0075,  0.0039,
         0.0048,  0.0047,  0.0088, -0.0052,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-2.5397e-03,  9.4471e-01, -1.6701e-03, -2.4321e-04, -4.8249e-03,
         6.2687e-04,  1.3328e-02,  4.7008e-03,  1.0756e-03,  4.1566e-04,
         3.4219e-03, -1.6404e-04,  4.1229e-04, -1.8065e-03, -5.6257e-03,
         1.9071e-03, -2.2291e-03, -8.7264e-04,  1.7632e-03, -1.8824e-03,
        -2.0924e-04, -8.5756e-04,  4.7162e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7229e-03, -9.4948e-01, -4.9462e-03,  1.7754e-03,  1.3331e-03,
         2.7221e-03, -4.1781e-03, -1.4414e-03, -2.0355e-03, -1.8491e-03,
        -1.0101e-04,  1.9999e-03, -1.1936e-03, -1.1760e-03, -1.2656e-03,
        -1.2220e-03, -1.1823e-03,  2.7022e-04, -6.7796e-04, -1.7559e-03,
        -7.4102e-04, -1.3803e-03, -6.5609e-04, -1.0930e-03, -1.1996e-03,
        -1.7417e-03,  6.4922e-04, -8.2708e-04, -2.4983e-04, -1.5203e-03,
        -3.2776e-03,  5.5929e-04,  7.8158e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7779e-02,  9.4529e-01,  8.2379e-03, -2.7712e-03,  3.9655e-03,
        -1.8201e-03, -2.6929e-03,  2.0923e-03,  7.0212e-04,  6.2686e-04,
         4.7212e-03,  8.0727e-04, -2.2888e-03, -7.2732e-04,  6.5242e-04,
        -1.2278e-03,  9.4781e-04, -2.2302e-03, -4.1720e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1992e-03,  9.5690e-01, -5.7666e-04, -7.1132e-03, -4.1789e-03,
         2.8037e-03,  3.2439e-03, -2.0934e-03,  1.0448e-03,  1.3526e-03,
        -9.3187e-03, -4.7533e-03, -1.8274e-03, -5.9109e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.4555e-03, -9.5739e-01, -4.0910e-03,  2.6293e-03, -2.8907e-03,
        -4.2941e-03, -1.1116e-03, -3.3935e-03, -2.7864e-04, -1.9069e-04,
        -4.4193e-04, -2.4197e-03, -4.5206e-04, -2.8188e-03, -1.0777e-03,
        -7.2048e-04, -2.4596e-03, -7.7054e-04,  2.0572e-03,  1.0458e-03,
         1.3660e-03, -1.6405e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.6659e-03,  9.4851e-01, -4.6422e-03, -8.6083e-03,  4.1592e-04,
        -1.1762e-03, -2.3797e-03, -3.5890e-03, -9.8756e-04, -8.2627e-04,
        -5.0747e-04, -1.7887e-03, -1.3701e-03,  3.3576e-04, -8.8399e-04,
        -1.3897e-03, -6.8071e-04,  8.7414e-04, -6.6364e-04, -1.1446e-05,
        -1.3830e-03, -3.2620e-05, -5.8940e-04,  5.7351e-04,  5.4339e-04,
        -1.3040e-04,  4.1302e-03, -4.8008e-04, -8.9749e-04,  3.9373e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5706e-03,  9.4907e-01, -5.5881e-03,  3.5658e-04, -1.9434e-03,
         2.6335e-03, -6.2658e-03, -2.8255e-03,  1.3589e-03, -3.0584e-03,
        -6.7447e-03,  2.9130e-04,  5.2386e-04,  7.2659e-04, -2.5250e-04,
        -1.1293e-02, -3.4971e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7067e-02,  9.4894e-01,  1.7265e-03, -6.5371e-03, -2.0348e-04,
        -4.3763e-04, -2.0599e-03,  7.9815e-03, -3.0520e-03, -7.6761e-04,
         1.2220e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4349e-02, -9.4129e-01, -7.7610e-03, -2.1636e-03,  2.7017e-03,
         3.4043e-04,  8.2242e-03,  2.9925e-03,  1.6650e-03,  6.0147e-04,
         3.4159e-03, -1.9575e-03,  4.0363e-03,  2.7371e-03,  2.9189e-04,
         4.7642e-03, -7.1070e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.8561e-03,  9.6172e-01, -4.2346e-03,  4.1312e-03,  1.7935e-03,
        -1.7432e-04,  6.9137e-04, -1.7463e-04, -4.6554e-05, -1.2701e-03,
         4.3266e-04,  2.3412e-04, -2.7970e-04, -3.5816e-04,  4.9155e-03,
        -7.7920e-04,  2.0983e-04, -1.5672e-03, -1.7507e-03,  7.8962e-04,
        -1.4446e-03, -9.9863e-04,  9.2086e-04,  1.9671e-04, -2.8595e-03,
        -3.1709e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5182e-02,  8.9850e-01,  1.3384e-02, -8.7088e-04, -9.7610e-04,
         6.6527e-03, -3.9429e-03,  4.9877e-03, -2.8350e-03,  9.4397e-04,
        -1.3206e-03, -8.2701e-03,  1.1258e-03, -4.1746e-03,  8.2990e-04,
         5.1569e-04, -7.5482e-04, -4.4546e-03,  1.6548e-03, -1.1732e-03,
        -7.6049e-04,  1.7446e-03, -1.1001e-03, -2.4800e-03, -2.3546e-03,
         2.1260e-04,  1.2089e-03, -2.1450e-03,  2.8639e-03, -2.5795e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1875e-02,  9.5189e-01,  2.3423e-03,  1.1612e-03,  1.8961e-03,
        -4.0481e-03,  1.8133e-03,  1.1836e-03,  4.6775e-04,  1.1814e-03,
         1.7309e-03,  1.5958e-03,  1.7888e-03,  5.2945e-03,  1.4911e-03,
         3.0643e-03, -3.0621e-03, -2.5312e-03,  1.5863e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-3.8054e-03, -9.3380e-01, -7.0819e-03, -1.3951e-02, -1.5565e-03,
         4.7931e-03, -6.5739e-03,  3.1255e-04, -4.3388e-04, -9.9850e-04,
         5.2706e-06, -6.4564e-04, -1.7191e-04, -1.5197e-03,  3.3674e-03,
         1.4064e-03, -1.6595e-03, -2.0154e-03, -1.5202e-03, -5.5771e-03,
        -1.0853e-03,  1.9378e-03,  2.6787e-04,  4.6335e-04, -8.8088e-04,
        -4.2246e-04, -3.7431e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.5370e-03,  9.6547e-01, -3.8778e-03, -4.8767e-03,  1.7373e-03,
        -1.4820e-03,  7.0802e-04,  2.4101e-04,  4.5460e-03, -3.4913e-03,
         1.9835e-04,  1.4456e-03, -3.6897e-03, -7.0266e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5304e-03,  9.4047e-01, -5.1122e-03, -3.5280e-03, -1.9586e-03,
        -8.7850e-03, -1.7895e-04,  3.9231e-03, -4.2809e-03, -4.2703e-04,
         1.6336e-03, -1.9433e-03, -4.9681e-04, -3.2201e-03, -1.9457e-03,
        -1.0696e-03, -4.0560e-03, -1.6627e-03,  4.1043e-05,  2.4509e-03,
        -4.4056e-04, -3.7533e-03, -1.1838e-03, -1.1861e-03,  2.8984e-03,
        -1.8261e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5367e-03,  9.5361e-01, -1.1343e-02, -5.0248e-03, -2.8270e-03,
        -9.9618e-03,  3.3610e-03, -1.6432e-04,  2.5654e-04,  1.3782e-03,
         3.3649e-03,  2.0276e-03, -2.0491e-03,  3.0898e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9288e-03,  9.6189e-01, -1.0654e-03,  5.3073e-03, -2.1564e-03,
        -1.6618e-03,  9.0611e-04, -3.1523e-04, -1.0593e-03,  4.0468e-04,
         9.9211e-05,  1.3884e-04,  1.6146e-03, -4.5878e-05,  2.6491e-03,
         1.4011e-03, -5.9382e-04,  2.1367e-04,  2.9294e-05,  4.6166e-04,
         1.2209e-03, -2.2115e-04, -4.2663e-04, -1.0055e-03,  3.9601e-04,
        -6.1965e-04, -1.0596e-03,  8.6079e-04, -8.0405e-04,  5.0290e-04,
         5.2649e-04, -1.4123e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5529e-02,  9.3676e-01, -5.1785e-03,  5.3819e-03, -7.7493e-03,
         5.4071e-04, -6.0209e-03, -1.0966e-03, -2.2814e-03, -5.7373e-04,
        -1.9773e-03,  4.5784e-05,  2.0889e-03, -1.0988e-03,  2.4096e-03,
         7.3509e-04,  6.5625e-04,  1.7581e-03,  1.2689e-03,  2.1205e-03,
         1.0935e-03,  1.1888e-03,  9.1898e-04,  1.4368e-03, -9.3130e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0070,  0.9490,  0.0284,  0.0048,  0.0058, -0.0050,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0250e-03,  9.4163e-01,  6.0318e-03, -4.9541e-03,  2.6606e-03,
        -1.2633e-03,  7.7812e-04,  1.0765e-03, -1.1640e-04, -3.7237e-04,
         7.7540e-04,  9.6085e-04,  1.9643e-04,  6.9958e-03,  1.8341e-03,
         3.2636e-04,  2.0427e-03,  2.0054e-03,  1.7555e-03, -1.7499e-04,
        -2.5182e-03,  5.1222e-04,  1.8664e-04,  7.1209e-04,  7.7976e-04,
         4.0787e-03,  2.5787e-04, -2.8759e-05,  8.4928e-04,  1.0875e-03,
         1.9568e-03, -3.3191e-03, -1.3079e-03, -7.6531e-05,  7.9212e-04,
        -9.2833e-05, -2.4636e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5193e-02, -9.1540e-01,  3.7991e-03, -6.2601e-03,  2.4188e-03,
         2.8155e-03,  9.1356e-03, -8.2254e-04,  1.2157e-03,  6.2588e-04,
         1.2812e-03, -1.7718e-04, -2.9215e-04,  1.9906e-03,  4.5796e-04,
         1.7233e-03,  2.4074e-03, -1.8252e-04,  1.3872e-03, -1.5312e-03,
         9.5931e-03,  1.6083e-03,  5.1260e-03,  3.5648e-04,  9.2922e-04,
         3.9135e-04,  1.0124e-05,  5.1399e-04,  2.7974e-03, -1.3871e-03,
         2.0191e-03,  1.8947e-03,  6.5777e-04,  1.3990e-03,  2.2029e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9815e-03, -9.1577e-01,  2.5773e-02, -1.6561e-03,  1.5889e-04,
         4.2954e-04, -1.1724e-03,  2.1104e-03,  3.0718e-03, -5.1348e-03,
         3.9690e-03,  1.6027e-02, -5.4371e-03,  1.3393e-03, -1.2974e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5340e-02, -9.4382e-01,  3.5718e-03,  1.9687e-03, -3.6323e-03,
         1.2346e-03, -3.1266e-03, -3.6822e-03, -5.6148e-04, -3.2309e-03,
         1.4782e-03, -4.2849e-04,  7.5679e-05,  1.0471e-03, -1.1726e-03,
        -1.0076e-03, -2.0360e-03,  1.7018e-03, -2.3158e-03,  3.0055e-04,
         8.4624e-04, -6.6970e-04, -2.9483e-04, -1.0993e-03, -4.5504e-04,
        -1.4462e-04, -5.8803e-04, -6.7221e-04, -6.1699e-04, -7.3369e-04,
        -9.1696e-04,  5.9270e-04,  6.3758e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2448e-03, -9.3641e-01, -4.9020e-04,  4.7363e-03,  3.8311e-03,
         1.4739e-03,  8.8617e-04, -7.0356e-06,  2.2891e-03,  1.6572e-03,
         1.0124e-03,  1.9226e-03, -5.5734e-04,  3.8190e-03, -6.3963e-04,
         5.2026e-04,  1.3276e-03, -4.0067e-04,  7.0445e-03,  1.7650e-03,
         7.5292e-04,  1.7719e-03, -1.0940e-03, -1.3217e-05, -7.8117e-04,
        -1.9856e-04,  1.6386e-04,  1.2021e-02,  7.1650e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([-4.9713e-03,  9.7136e-01, -1.8024e-03, -9.3445e-04,  1.3750e-03,
         6.6306e-04,  7.4005e-04, -2.2986e-03, -1.7742e-03,  1.5720e-03,
         1.9394e-03,  2.1920e-03,  1.6744e-03,  2.4577e-04, -6.4569e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.2613e-03,  9.4832e-01, -4.1139e-03, -2.5144e-03, -6.3130e-04,
        -1.8878e-03,  7.8082e-05, -8.6231e-04,  1.3037e-03, -1.5423e-03,
         1.4412e-05, -9.4979e-04, -3.0839e-04,  3.0016e-04,  6.4486e-05,
        -1.5130e-03, -6.1211e-05, -7.6340e-05, -6.5414e-04, -9.7938e-04,
        -9.0165e-04, -9.5629e-04, -6.1851e-04,  1.6219e-03,  7.8168e-04,
        -9.6161e-04,  6.2081e-04,  6.2641e-04, -2.3223e-04, -1.5786e-04,
        -2.4709e-05, -5.9655e-04, -1.0956e-03,  3.7645e-04,  3.9048e-04,
         2.6097e-03,  1.4047e-03, -5.2738e-04,  4.5797e-04, -5.2126e-04,
        -3.6138e-06,  1.6199e-03, -1.8139e-03, -1.2506e-03,  3.6184e-04,
        -1.6685e-03,  9.9532e-04, -4.3375e-04, -3.9335e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4193e-02, -9.0645e-01, -1.0521e-02, -2.0639e-03, -5.9586e-05,
        -1.7159e-03,  6.1651e-04, -1.1189e-03, -4.1619e-04, -6.4503e-03,
        -4.0326e-04, -1.8024e-03,  4.7102e-04,  1.1789e-04,  8.1727e-04,
        -5.3488e-04, -3.0088e-03, -8.9276e-04, -2.5198e-04, -1.2799e-03,
        -1.5249e-03, -1.5678e-03,  7.0663e-04, -8.9384e-04, -7.2451e-04,
        -2.2079e-03,  1.2044e-03,  3.0895e-04, -2.0501e-03,  2.0690e-03,
        -2.6791e-04,  1.8641e-04,  2.4786e-03, -5.5059e-03,  5.1137e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.9099e-03,  9.4327e-01,  1.0363e-02,  2.7290e-03, -5.4591e-04,
         2.5749e-03, -5.7041e-03,  5.7669e-03,  7.9360e-04,  1.0340e-03,
        -3.1481e-03, -1.1653e-03, -1.0557e-03,  3.2837e-04,  2.9160e-03,
        -4.3225e-03, -1.0146e-03,  2.4272e-04, -5.1144e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.7532e-03,  9.4474e-01, -7.6778e-03,  4.0148e-03, -4.1080e-03,
        -8.0673e-05, -6.4964e-03, -8.5749e-04,  9.3249e-05, -4.2579e-03,
        -4.0921e-03, -2.2669e-03, -3.2706e-03, -7.4820e-03,  1.8111e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8544e-02,  9.5680e-01, -4.8310e-06, -3.3415e-04,  5.8814e-04,
        -5.4991e-04,  1.7566e-03, -1.7327e-05,  2.8076e-03,  2.4038e-03,
        -1.5176e-03,  1.7783e-04,  6.6770e-04, -7.0922e-04, -2.6551e-03,
         4.2634e-04, -4.8193e-04, -1.1039e-03, -4.5072e-03,  2.2906e-03,
         6.8263e-04, -9.7784e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.8821e-03,  9.3675e-01,  2.9807e-03, -4.6672e-03, -1.0831e-03,
        -7.1430e-03, -4.4627e-03, -4.1636e-04, -2.2058e-03, -1.3313e-03,
        -1.1045e-03,  1.9849e-03, -5.2098e-03, -9.9839e-04,  1.1412e-04,
         7.8293e-04, -7.3300e-04,  6.4532e-04, -2.1540e-03, -1.3149e-03,
        -1.1377e-04, -8.0366e-04, -1.1834e-03,  1.3464e-03,  2.1982e-03,
         1.5200e-03, -1.2684e-03, -1.0101e-03,  1.4012e-04,  5.4495e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5796e-02,  9.3076e-01, -4.7521e-03, -1.1457e-02, -7.7425e-03,
         5.8249e-03,  1.9632e-04, -3.3655e-03, -5.1101e-03, -3.9470e-03,
        -1.0507e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9774e-03,  9.6667e-01,  2.1260e-03,  4.1031e-04, -1.0885e-03,
         8.4205e-04,  4.5722e-03,  2.3254e-04,  2.9458e-03, -1.5561e-03,
        -8.5911e-04,  1.9344e-03,  1.6692e-04,  1.2969e-03,  4.3279e-04,
         2.6001e-03, -2.1746e-03,  1.1417e-03, -7.3023e-04,  6.5356e-04,
         4.0993e-03, -1.4884e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.2200e-03,  9.2367e-01,  2.6292e-03, -3.2551e-03, -4.8149e-03,
         3.6282e-03,  1.9053e-03,  4.0223e-03, -1.9202e-03,  3.9488e-04,
        -6.1862e-03,  2.6787e-03,  1.7582e-03, -3.4915e-03,  2.0576e-04,
         3.4981e-04,  1.4941e-03, -6.7342e-04, -2.2756e-03,  8.5864e-04,
         8.8094e-04, -9.9562e-04, -2.0699e-03, -3.4146e-04,  1.1369e-03,
         3.9461e-04, -1.1846e-03, -7.1599e-04, -7.4445e-03, -3.7994e-04,
         5.6097e-03, -6.9598e-04,  6.7130e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7084e-03, -9.1165e-01, -2.2709e-02, -2.9364e-03, -6.9661e-03,
        -4.3646e-03, -2.4905e-03,  1.0591e-02,  8.2866e-04,  8.4554e-04,
        -1.4406e-04, -1.3333e-03,  4.0645e-04,  8.1027e-04, -1.6027e-03,
         9.0611e-04, -2.6976e-03,  2.5648e-03, -1.8050e-03,  1.4001e-03,
        -2.7084e-03,  4.6489e-03,  3.0147e-04, -2.2993e-04, -5.2919e-03,
        -1.3148e-03,  3.9086e-03,  2.8345e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.4441e-03,  9.4972e-01,  6.2178e-03, -1.0855e-03, -6.7758e-04,
        -6.3228e-03,  3.1303e-03,  3.2097e-03, -5.3154e-03,  4.8011e-03,
         8.0245e-05, -6.4705e-05,  4.6181e-04,  1.8198e-03, -3.0562e-04,
         2.0061e-04, -1.0943e-03, -2.2613e-03, -1.0675e-03, -2.8368e-03,
        -9.8175e-06,  1.7819e-04,  1.5420e-03, -1.7557e-03,  1.3982e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 2.2171e-02,  9.2967e-01, -5.4493e-04, -4.8527e-04, -3.7726e-03,
         2.2489e-03, -1.5584e-03,  5.6811e-04, -3.2662e-04, -1.1385e-03,
         1.5306e-03, -1.6188e-03, -7.5720e-04,  1.6785e-03, -5.9504e-04,
        -5.2264e-03,  1.9808e-04, -2.2223e-03, -1.4139e-03,  1.7766e-03,
         1.7793e-03,  4.3365e-04,  3.9873e-04, -8.7262e-04,  2.0454e-03,
        -3.3495e-05, -4.2661e-04, -1.3539e-03, -1.1309e-03, -1.7584e-03,
         9.2960e-05,  1.5424e-05, -2.9525e-03,  9.0075e-04, -9.4014e-04,
         1.1054e-04, -1.0322e-03, -4.5067e-04,  1.9734e-03, -8.2173e-04,
         9.7862e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1313e-03,  9.6041e-01,  1.8520e-03,  1.5415e-04,  3.3210e-03,
        -6.5253e-04, -2.1042e-03,  1.6980e-03,  2.2304e-03, -2.3481e-04,
         2.2128e-03,  1.2109e-03, -7.2519e-04, -9.5012e-04,  7.9200e-03,
        -3.2666e-03,  1.8098e-03,  7.8601e-04, -1.5319e-03,  3.5848e-05,
         1.0542e-03,  2.3746e-03,  3.8188e-04, -6.1017e-04, -3.3808e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.9433e-03,  9.4787e-01,  6.3749e-03,  2.5682e-04,  4.4499e-03,
         2.4366e-04,  5.9092e-04,  3.8769e-03,  1.3132e-03,  2.5995e-03,
        -1.1973e-03,  3.3636e-03,  1.7777e-03,  6.9707e-05,  2.5709e-03,
        -1.4948e-03,  2.3050e-03,  4.8169e-04, -1.6158e-03,  3.8787e-03,
        -4.7303e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6551e-03,  9.4457e-01,  6.3883e-03, -6.9821e-03,  4.4445e-04,
         1.9577e-03, -1.6709e-04, -2.1116e-03, -1.8432e-03, -4.1824e-03,
         3.5639e-03, -5.7764e-03,  2.2696e-03, -4.5769e-03, -2.1189e-03,
         1.5885e-03,  5.6232e-04, -9.2461e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.9981e-04,  9.2595e-01,  6.3616e-03,  2.3980e-03,  1.3379e-03,
         2.6952e-03,  1.4543e-03, -1.6251e-03,  2.9262e-03,  2.3358e-03,
         7.7647e-04,  7.7241e-04, -3.4946e-03,  2.3596e-03,  5.3702e-03,
         3.1000e-03,  2.4675e-03,  1.0716e-03, -1.2719e-03,  5.5302e-04,
        -1.8469e-03,  9.3915e-04,  2.1286e-05, -2.3544e-04,  1.2452e-04,
         2.9567e-04, -7.7500e-04,  1.8265e-03, -4.9168e-04,  7.0073e-05,
        -6.2592e-04,  1.2640e-03,  3.4894e-04, -5.7169e-03,  1.4250e-03,
         2.5950e-03,  7.6854e-03, -4.4922e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.0734e-03,  9.5335e-01, -1.4093e-02, -3.0876e-03,  9.8787e-05,
         4.0049e-03, -3.3584e-03,  1.8577e-03,  1.5970e-03,  1.1816e-03,
         2.4196e-03, -2.7561e-03,  4.9668e-04, -4.2430e-04,  5.3016e-05,
        -4.8124e-04,  1.0487e-04, -2.3259e-05, -1.5329e-03,  4.0083e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2399e-02,  9.4390e-01, -1.1388e-02, -2.2248e-03, -6.1435e-03,
        -1.4183e-03, -4.5710e-03,  2.1995e-03, -5.9054e-04,  2.2450e-05,
         5.3769e-03, -2.7189e-03, -2.4239e-04,  6.8007e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7417e-02,  9.4394e-01,  5.4411e-03, -6.7047e-04, -1.0603e-03,
        -1.1024e-03,  8.3034e-04,  5.6050e-04, -8.8938e-04,  2.5850e-04,
        -1.4157e-03,  3.2403e-04, -6.1808e-04,  1.4067e-03, -3.1396e-04,
        -3.4007e-03, -1.7084e-04, -5.1582e-04, -1.2835e-03,  3.5459e-04,
        -7.2642e-04,  4.7488e-04,  6.3638e-05,  6.1893e-04, -7.7072e-04,
        -1.3658e-03, -1.2291e-04, -3.5903e-03, -2.8647e-04, -5.0548e-03,
        -1.5496e-03,  3.4012e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.7533e-03, -9.5324e-01,  1.5023e-03, -6.8772e-05,  8.3823e-03,
        -2.4226e-05,  2.0996e-03,  4.2851e-03,  1.5945e-03, -4.5151e-03,
         1.4248e-03,  2.0086e-03,  1.7615e-03,  2.2424e-03, -5.3585e-04,
        -1.9208e-04,  1.7749e-03,  2.2041e-03,  2.8856e-03,  1.5037e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2029e-02,  9.0617e-01,  2.0253e-02, -1.6446e-04,  3.4344e-04,
        -3.9853e-03,  2.6665e-03, -2.0765e-03,  1.4544e-03,  2.0050e-03,
         5.8557e-05,  9.5830e-04, -8.8361e-04, -9.2937e-04, -3.9715e-04,
        -4.1792e-03, -8.1606e-04, -1.3820e-03, -1.9605e-04, -3.0318e-04,
         7.4311e-04, -1.8380e-03, -8.5041e-05,  3.1644e-03, -1.7549e-05,
        -3.4741e-03, -4.6862e-03,  1.6513e-03, -3.0380e-03, -1.3858e-03,
        -4.0323e-04,  4.4929e-04,  6.1044e-04,  3.0486e-03,  1.1639e-03,
         1.8564e-03,  3.9272e-04, -1.8083e-03,  1.4748e-03, -6.1884e-04,
        -4.6580e-04, -6.1367e-04,  8.5733e-04, -2.7972e-03, -7.6106e-04,
         2.0234e-04,  1.1401e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9961e-04,  9.4817e-01, -8.0580e-03,  5.6520e-03,  8.9465e-03,
         2.5121e-03, -2.7747e-03, -4.4781e-03, -2.4470e-03,  1.5967e-03,
        -4.4526e-03, -1.8985e-03, -1.2273e-03,  3.8760e-04,  6.7966e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1819e-02,  9.2443e-01, -6.0975e-03,  5.5586e-04, -1.0742e-03,
         1.9801e-03,  1.0749e-03, -3.1534e-04,  1.5457e-03,  1.2778e-03,
         1.6569e-03,  2.8429e-03,  4.1213e-04,  2.1402e-03,  6.2171e-04,
        -8.0826e-04, -1.1815e-03,  9.7559e-04,  3.0473e-03,  5.7334e-05,
        -2.8136e-04, -2.0040e-03,  1.7893e-04,  8.4777e-04,  9.2511e-04,
         2.3727e-03,  2.2331e-04,  1.2541e-02, -8.0198e-05,  4.3047e-04,
         2.4634e-03, -6.6067e-04,  1.6995e-03,  2.6137e-04,  7.8494e-04,
         6.4163e-06, -5.4354e-04,  2.7750e-04, -2.2744e-03,  4.9457e-04,
        -2.7357e-05,  2.3065e-04, -1.4674e-03,  6.3808e-04, -1.7648e-05,
         1.9067e-03,  1.0786e-03,  1.3647e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-7.8478e-04,  9.4392e-01, -9.2478e-04, -3.6785e-03,  7.8256e-03,
         4.6420e-03,  1.5834e-03, -1.9107e-03, -1.9890e-03, -7.9876e-03,
        -3.8358e-04, -4.4507e-03, -2.8939e-03,  4.3989e-04, -1.5914e-03,
        -4.0415e-04, -4.7955e-03,  1.1144e-03, -4.3321e-04,  1.6041e-03,
         6.3475e-04,  2.7563e-03,  4.0358e-04, -7.4431e-04, -4.1209e-04,
        -1.6968e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.2703e-04,  9.5718e-01, -4.1578e-03,  2.5393e-03,  3.7172e-03,
        -1.0647e-03,  2.2406e-03,  3.8225e-03,  1.6732e-03, -5.2263e-04,
         7.3703e-04, -6.9443e-04,  1.1031e-03,  1.7804e-03,  1.1100e-03,
        -2.9540e-03,  8.3911e-03,  8.5464e-04,  1.2618e-03, -5.7173e-06,
         3.4608e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.0891e-03,  9.3124e-01, -2.6888e-03, -4.3117e-03, -2.1253e-03,
        -7.4760e-03, -5.8683e-04,  1.6909e-04,  1.6932e-03,  6.8152e-04,
        -2.5187e-04,  7.2974e-04,  3.3233e-05, -3.0769e-03,  2.2866e-03,
        -3.8205e-03, -8.9103e-04, -3.6259e-03, -1.1526e-04,  3.1405e-03,
         1.6469e-03,  9.1745e-05, -2.9184e-03,  3.6459e-04, -5.0695e-04,
        -1.9370e-03, -6.0137e-04,  8.7080e-04,  1.1898e-03, -5.8904e-04,
        -4.5770e-03, -2.6830e-04,  8.0003e-04,  2.4915e-04, -6.9876e-03,
        -2.3730e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3767e-02,  9.3527e-01,  1.6385e-03,  6.4049e-03, -2.6590e-03,
         9.4199e-03, -3.2371e-03,  5.7668e-03, -1.8396e-03, -1.0985e-03,
         7.6244e-05,  1.2540e-03,  1.2195e-03, -3.8232e-03, -5.3369e-03,
        -2.0912e-03, -6.5762e-04,  3.7066e-04,  1.6976e-03,  2.3730e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.7798e-04,  9.5832e-01,  2.1591e-03, -9.9494e-04, -1.3910e-03,
        -1.1702e-03,  2.2920e-03,  1.6776e-03,  1.9870e-03, -1.2744e-03,
        -5.1803e-03, -4.1213e-03, -1.8588e-03, -2.0081e-03,  2.4090e-04,
        -1.1065e-03,  1.5262e-03, -2.1040e-03,  2.3216e-03,  2.9228e-03,
         4.6662e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5311e-03,  9.6866e-01, -8.8809e-03,  4.2096e-03, -1.5687e-03,
        -5.6417e-04,  9.8276e-04, -1.0788e-03,  1.2116e-03,  3.8981e-03,
        -1.2421e-03,  4.0422e-03, -2.1330e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8797e-03,  9.3080e-01,  1.8328e-02, -8.8232e-04,  3.3849e-03,
         1.1443e-03,  7.5598e-04, -3.2493e-04, -1.6298e-03, -1.0357e-04,
         7.7269e-04, -5.1116e-03, -4.6291e-04,  1.0865e-03,  2.8385e-03,
         1.8491e-03,  8.2498e-05,  1.4682e-04, -1.6501e-03, -1.9272e-03,
         9.4963e-05,  8.1609e-04, -8.8414e-04, -1.1680e-03,  2.1648e-04,
         5.4812e-04, -2.7941e-03,  3.4577e-04, -2.4228e-04, -7.9003e-04,
         6.8651e-05, -9.1369e-04,  3.6528e-03, -9.2808e-04,  1.8876e-05,
         3.0701e-03,  6.2881e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6605e-03,  9.5790e-01,  6.4760e-03, -1.3126e-03,  1.0837e-03,
         2.1773e-03,  2.8905e-03,  2.4941e-04,  5.9839e-04,  2.1946e-03,
        -2.0033e-04,  1.1151e-03,  9.6618e-04,  3.6575e-03,  1.0270e-03,
         8.5256e-04,  1.1490e-03, -3.3462e-04,  4.8845e-04, -3.5587e-03,
         7.6647e-04,  1.2061e-03, -1.5752e-03, -6.5630e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8120e-03, -9.4240e-01,  3.3957e-03, -6.8868e-03, -1.2569e-03,
        -8.9292e-04,  1.4642e-03, -5.5848e-03,  7.9079e-04,  2.2010e-04,
        -1.0171e-03,  4.1596e-03, -5.1119e-03, -1.2583e-03, -1.9260e-03,
        -3.7695e-04, -2.0069e-04,  1.8604e-03,  2.9286e-04, -4.4609e-03,
        -4.1748e-05,  1.2265e-03,  1.1405e-03,  1.8286e-04, -5.6982e-04,
         6.4818e-04,  4.7896e-04,  3.1797e-03, -7.0689e-04, -1.1237e-03,
         4.7332e-04,  3.8574e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4482e-03,  9.3652e-01, -4.2038e-03, -7.1383e-03, -1.6281e-03,
        -6.6544e-03,  7.9307e-04, -3.0779e-03,  3.1738e-03, -4.0216e-03,
        -1.7929e-03, -8.1101e-05, -2.9248e-04,  6.7329e-05, -6.0883e-04,
        -1.5744e-06, -1.3502e-03, -3.3794e-04,  2.1663e-03, -8.7497e-04,
        -3.6920e-03, -4.5465e-03,  5.7060e-04, -1.0549e-03, -5.2197e-04,
         1.4369e-03,  5.5075e-03, -6.0268e-04,  1.0817e-03, -2.4442e-03,
        -3.0986e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8923e-02,  9.0936e-01, -1.9646e-02, -3.0470e-04,  4.2895e-03,
         1.9617e-03,  3.3718e-03, -4.9154e-03,  4.8350e-04,  2.2575e-04,
         1.0058e-03,  1.3474e-03,  5.2517e-03,  4.7283e-03,  3.7769e-03,
         5.5063e-03,  2.2700e-03,  1.4441e-03,  1.6952e-03, -3.2543e-03,
        -2.1374e-03, -2.6029e-03, -1.4981e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.0967e-03,  9.5417e-01,  3.9226e-03, -5.1283e-03, -8.3642e-04,
         2.0282e-04,  8.2930e-04, -3.6445e-03, -6.0516e-04,  6.2458e-04,
         1.5503e-03, -3.5780e-03, -6.4223e-04,  1.7462e-03, -2.8497e-04,
         7.7285e-04, -1.0444e-03, -2.1149e-04,  7.7426e-06, -1.2542e-03,
        -1.8772e-03,  6.5066e-04, -1.4076e-03, -1.2428e-04,  6.1215e-04,
         2.2862e-03,  4.7846e-04,  6.4137e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 3.6037e-03,  9.5008e-01,  7.7881e-03,  1.0765e-03,  1.2468e-03,
         1.2788e-03,  1.6642e-03,  1.3699e-03, -6.1297e-04, -2.9206e-03,
         8.4028e-04,  1.4820e-03,  3.3039e-03,  7.4456e-04, -4.2875e-04,
         7.4107e-04,  8.6877e-04,  1.7883e-03,  3.5036e-04,  1.7419e-03,
        -2.4530e-04,  7.5398e-04, -2.0774e-03, -1.7619e-04,  2.2604e-04,
        -7.4116e-04,  7.3143e-04, -1.0394e-02, -7.1908e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8220e-03,  9.5303e-01,  9.0466e-06, -8.0724e-03,  2.3215e-03,
         4.6156e-03, -3.0849e-03, -9.4524e-04, -5.8562e-04,  6.0719e-04,
        -1.9746e-03, -2.4535e-03,  4.0609e-04,  2.7418e-03, -7.3342e-05,
        -1.7857e-03,  3.4057e-03,  4.9521e-04,  3.0160e-03, -2.3269e-04,
         2.2191e-04,  2.1865e-03,  2.1244e-03, -3.7850e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9905e-03,  9.3330e-01, -8.9583e-03, -1.4274e-03,  4.4278e-03,
         1.0942e-03,  9.5546e-04,  4.9168e-03, -2.7807e-03, -5.6909e-04,
        -8.8487e-04,  4.7668e-03,  5.0568e-03,  2.9665e-03,  5.8390e-04,
         3.0701e-03,  3.4234e-04,  5.7911e-04,  1.4563e-03,  6.0833e-04,
         2.2076e-03, -8.1586e-04,  1.3908e-03,  2.7047e-03,  1.7168e-03,
         9.5520e-04, -1.0748e-03,  9.7185e-04, -4.3633e-04, -3.8102e-04,
         1.4251e-03,  2.9168e-03, -2.0036e-03,  2.5929e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.1713e-03, -9.5093e-01, -2.2810e-03,  7.5355e-03,  6.9126e-04,
         1.0222e-03, -2.3496e-03,  1.3224e-03,  1.4099e-03, -2.7860e-03,
        -6.4571e-04,  4.6398e-04, -1.4907e-03,  2.7230e-03, -1.0398e-03,
        -1.0503e-03,  4.0176e-03,  2.2814e-03,  1.0911e-03,  8.9036e-04,
        -2.3725e-03, -5.4308e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2869e-03,  9.4726e-01, -9.0165e-03, -6.7812e-03, -3.7109e-03,
        -1.1611e-03, -2.6128e-03,  3.8442e-03,  2.2995e-03,  2.0035e-03,
         6.0668e-03, -1.3205e-03,  1.1205e-03, -2.6111e-03,  1.9905e-04,
         1.7952e-03,  1.4805e-03, -4.4267e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.5696e-03,  9.6133e-01,  2.2843e-03, -2.7876e-04,  1.7156e-03,
         3.4971e-04,  2.7625e-03, -2.0095e-05,  8.3095e-04,  1.9684e-03,
         3.9574e-04,  1.9596e-03, -2.5298e-03,  1.9046e-03, -3.0242e-03,
        -4.6733e-04,  2.3618e-03, -9.1187e-04, -7.7525e-04,  3.8667e-04,
         7.1793e-04, -3.2142e-03,  1.2411e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9453e-03, -9.6552e-01, -8.9594e-04,  8.1220e-04,  3.1207e-03,
         2.5465e-04,  3.0584e-03,  2.2863e-04,  6.7983e-03,  4.2772e-03,
         4.8446e-04,  5.0633e-04,  8.4446e-04, -3.3976e-03,  5.0563e-04,
        -2.1243e-03,  1.2276e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5943e-03,  9.7127e-01,  8.0272e-04, -9.3146e-04,  1.4810e-03,
         1.8805e-03,  1.4702e-03,  7.7905e-04, -2.6697e-03, -1.4932e-03,
         3.5431e-04, -1.1528e-04, -3.3619e-03,  2.5093e-03,  2.5645e-03,
         1.6154e-03,  6.1029e-04, -4.4935e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7864e-03,  9.2800e-01, -1.4800e-02, -2.2927e-03,  9.9864e-05,
         6.9803e-04,  8.0180e-04,  1.8171e-03,  6.0571e-04, -3.3726e-03,
         3.9854e-03, -5.2779e-03,  1.4079e-03, -4.1085e-04,  3.7621e-04,
         6.4527e-03,  1.0632e-03,  3.5029e-03,  2.7504e-04, -8.8636e-04,
         1.2508e-03, -2.8119e-03,  7.5259e-04,  2.3443e-03,  4.5701e-03,
        -7.3545e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7125e-02,  9.3013e-01, -7.6141e-03, -3.7008e-03, -5.3057e-04,
         2.9460e-03, -2.2621e-03,  6.3657e-04,  3.5224e-03, -2.8644e-03,
        -2.2034e-03,  9.8327e-04, -1.4523e-03, -1.5319e-03, -2.6390e-03,
        -2.0112e-03, -3.6852e-04,  3.3542e-04,  5.5018e-04, -1.5682e-03,
         1.4078e-03,  1.8595e-03,  1.1198e-03, -1.1883e-03, -2.8442e-04,
        -1.0705e-03, -1.7458e-03, -3.0593e-04, -8.9983e-04,  4.2750e-03,
        -8.6715e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2290e-03,  9.4858e-01, -5.2735e-04, -4.6132e-04, -8.6946e-03,
        -5.9468e-03, -1.3377e-03,  1.8227e-03, -2.6183e-03, -3.8436e-03,
        -2.3215e-03,  1.6311e-04, -3.9338e-03, -1.3316e-03, -6.8664e-03,
        -1.3419e-03,  3.3936e-03, -4.5913e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0104,  0.9568,  0.0150,  0.0031,  0.0050, -0.0034, -0.0016, -0.0048,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([-2.8506e-02, -9.2581e-01, -8.3335e-03,  2.5723e-03, -2.2941e-03,
         1.2742e-03,  1.2545e-04,  5.4356e-04,  1.0472e-03, -2.3846e-03,
         1.6702e-03,  1.8356e-03,  8.4661e-04,  1.8919e-03, -1.6380e-03,
        -6.6655e-04, -3.4968e-04,  2.3712e-03, -1.3468e-03,  5.5833e-04,
        -3.7931e-04,  5.3287e-04,  2.3135e-03,  1.5982e-04,  1.1922e-04,
        -5.7482e-05,  1.0598e-03, -5.0411e-04, -8.3680e-04,  2.4777e-04,
        -9.8359e-04, -3.7843e-05, -1.0880e-03, -9.5004e-04, -2.3726e-03,
         5.4850e-04, -9.1028e-04,  1.8783e-04, -6.4885e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.8442e-03,  9.5057e-01, -7.2194e-03, -7.3480e-03,  7.8682e-04,
        -1.3105e-03, -8.7248e-04, -3.4057e-04, -2.4560e-03,  1.4536e-03,
         1.9102e-03,  2.3396e-04,  1.8678e-03, -2.2794e-03, -1.0434e-03,
         2.8353e-03, -1.4155e-03, -7.3040e-04, -1.5243e-03, -2.9474e-05,
         2.4237e-04, -1.7045e-03, -1.2129e-03, -3.5513e-05,  9.6979e-06,
        -4.8529e-04,  2.7755e-04,  3.9653e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.5994e-03,  9.6360e-01, -2.6784e-03,  2.7599e-03,  3.4224e-03,
        -1.6600e-03,  3.8191e-03,  1.7475e-03,  2.8755e-03, -3.7950e-04,
         5.2117e-03, -3.0069e-03,  2.3985e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4138e-02,  9.6806e-01, -5.7086e-04,  1.5994e-03, -2.3476e-03,
         1.1937e-04, -2.1882e-03, -1.0207e-04, -3.2815e-03, -1.3201e-04,
        -4.5679e-04,  3.0467e-06, -2.7122e-03,  7.3152e-04,  3.5534e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.6886e-04, -9.4891e-01, -5.5767e-03,  2.3422e-03,  1.1966e-04,
        -3.4015e-04, -1.0872e-03,  3.2001e-04, -3.4710e-03, -2.5488e-03,
        -1.2041e-03, -1.3705e-03, -1.8700e-03, -2.4455e-03, -1.4199e-03,
        -3.1281e-04, -7.0385e-04, -1.6305e-03, -9.8924e-04, -6.2067e-04,
        -4.1133e-04,  8.1493e-04,  2.0642e-04, -3.6524e-04,  3.0909e-04,
        -1.8067e-03, -3.7081e-04, -4.4572e-03, -4.6991e-04, -1.6307e-03,
         4.0871e-04, -8.4868e-04,  3.5112e-05, -1.4626e-03, -3.5567e-04,
        -1.6195e-03, -4.7440e-04, -7.3030e-04, -6.6376e-04,  4.7900e-04,
        -1.3029e-04, -2.5100e-04, -6.0280e-04,  6.1329e-04,  3.2652e-04,
        -1.3499e-03,  6.5027e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9828e-03, -9.4427e-01, -1.1016e-02, -1.9447e-03, -2.0910e-03,
        -1.5084e-03, -3.0379e-03, -1.7839e-03,  2.9654e-03, -2.0967e-03,
        -7.0713e-04,  1.5028e-03, -9.8403e-04, -1.5299e-04, -6.2794e-04,
        -4.1457e-04,  5.0351e-04, -2.8774e-04, -1.2183e-03, -2.1457e-03,
        -1.4013e-04, -5.2650e-05,  3.0489e-04, -1.7257e-03, -1.0164e-03,
        -6.5393e-04,  2.0913e-03, -5.5318e-04, -6.7742e-04,  1.1018e-03,
        -1.4572e-03, -1.0005e-03, -2.0766e-03, -1.1922e-03, -6.9588e-04,
        -5.7591e-04, -3.8914e-04, -5.1000e-04, -1.7056e-03,  3.3486e-04,
        -3.5940e-05,  4.6584e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6470e-02, -9.5006e-01,  1.0052e-05, -6.5751e-04, -2.4134e-03,
        -2.3803e-03, -3.2719e-03, -1.6556e-03, -1.3013e-04, -2.6204e-03,
         1.1974e-03,  1.2594e-03,  2.2821e-03,  8.0083e-04, -9.9732e-04,
         2.2050e-03, -3.8123e-03,  9.9975e-04,  3.6693e-04,  1.1726e-03,
        -3.3293e-04,  2.8097e-04,  3.6944e-03,  6.4227e-04, -2.8547e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2866e-03,  9.4660e-01,  6.0653e-03, -2.7401e-04, -1.2792e-03,
        -3.1686e-03,  6.6754e-03, -6.4504e-04, -1.2466e-03,  5.7379e-04,
         1.1215e-03, -1.2720e-03, -2.6142e-04, -2.1086e-03,  1.9242e-03,
         2.4779e-03,  1.1546e-03,  7.4890e-03, -1.7091e-03, -2.4074e-03,
        -1.6407e-03, -1.8214e-04, -1.4358e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7378e-03,  9.5113e-01,  3.3959e-03,  5.6010e-05, -1.1949e-03,
        -2.7602e-03,  1.2969e-03,  5.5306e-04,  1.0060e-03,  6.9197e-03,
         2.1464e-03,  1.7560e-04,  2.8536e-03,  2.0049e-03, -6.9698e-04,
         2.0004e-03, -6.7369e-04,  4.7525e-03, -3.4136e-03,  1.3059e-03,
        -9.4381e-07, -1.3759e-03,  7.5482e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9385e-02,  9.2915e-01,  7.2043e-03, -6.3915e-03,  1.3529e-03,
         1.4005e-03,  7.3645e-03, -6.2815e-03,  1.3917e-03,  8.5897e-04,
         3.4090e-03, -3.3847e-03, -2.4253e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3297e-03, -9.7289e-01, -1.7369e-03, -5.3962e-03, -2.4872e-03,
        -3.0137e-04, -5.5031e-04, -1.9878e-03,  2.5691e-03, -9.2688e-06,
         3.9889e-03, -4.8520e-03, -1.2824e-03,  6.1816e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4214e-02, -9.6509e-01, -2.0399e-03, -1.1733e-03,  2.2306e-03,
        -1.9390e-03, -1.4753e-03,  1.5795e-03, -8.0674e-04, -2.5042e-03,
         2.1679e-04, -7.6986e-04, -7.1334e-04, -1.1230e-03,  1.7930e-03,
         2.3969e-04, -1.1219e-03,  4.3269e-04, -5.3935e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-1.2845e-02, -9.3529e-01, -6.0664e-03,  8.8637e-04, -6.4042e-04,
         3.7951e-04, -1.0561e-03,  2.1736e-04, -1.2965e-03,  4.6225e-04,
         1.1815e-03, -4.5023e-03, -1.4643e-03, -2.3055e-03,  3.3280e-03,
         2.0301e-03, -1.9385e-03, -2.0976e-03, -2.3016e-03, -5.0969e-03,
        -2.7044e-03, -5.1681e-04,  3.9736e-04, -2.4416e-03,  4.0519e-03,
         4.5057e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2497e-02,  9.3102e-01,  1.2721e-03, -5.0249e-03,  1.0714e-03,
        -4.1243e-03, -8.4500e-04, -3.5755e-04, -3.7995e-03,  3.9735e-03,
         2.7303e-03, -5.0291e-03,  3.1520e-03,  6.4386e-04,  6.2520e-04,
         3.8364e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0415e-02,  9.2877e-01,  1.8459e-03,  1.3910e-02,  1.8522e-03,
        -1.2436e-05,  1.5877e-03, -1.5207e-03,  5.6577e-03,  4.8527e-03,
         3.5088e-03, -3.6221e-03, -3.9885e-04, -4.0582e-03,  5.0131e-03,
        -2.8451e-03, -1.3256e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0176, -0.9443,  0.0073, -0.0077,  0.0013, -0.0058,  0.0011, -0.0013,
        -0.0137,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4165e-04,  9.3719e-01, -8.3834e-04,  3.1430e-03, -1.2451e-02,
         1.3093e-03,  2.1533e-03,  2.5651e-04,  6.0013e-05, -1.5283e-03,
         4.1733e-04, -3.1588e-04,  2.0502e-03,  5.0811e-04,  7.9274e-04,
         1.1327e-04,  3.3174e-04, -3.7330e-04,  5.7394e-04, -3.5849e-04,
        -1.1358e-03, -1.1108e-03,  6.4920e-04, -1.3189e-03, -1.6342e-04,
        -2.7265e-03,  3.4071e-03, -1.9985e-03, -9.6821e-04,  1.8740e-04,
        -9.3659e-04, -6.7896e-04,  1.4225e-03, -5.0168e-04,  1.5929e-03,
        -2.7381e-04,  6.0931e-04, -1.2163e-03, -2.0591e-03, -1.3579e-03,
         1.0455e-04,  6.7775e-04, -1.2083e-03,  1.5484e-04,  1.8303e-03,
         6.3029e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3814e-03,  9.2759e-01, -1.3138e-02, -4.6066e-03, -1.2086e-03,
        -2.3607e-03, -1.3106e-03, -2.6347e-03, -4.6832e-03,  2.4725e-03,
        -4.4089e-04,  2.4717e-03, -5.3770e-03, -7.0172e-04,  1.4723e-03,
        -5.7797e-05, -5.9940e-05, -3.7337e-04, -1.1833e-03,  7.7713e-04,
        -1.3423e-03, -2.3694e-03,  3.2657e-04,  1.1234e-03,  4.2477e-03,
        -2.2645e-03, -2.7721e-03,  5.1028e-04,  6.9361e-03,  2.8017e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.8987e-03, -9.6559e-01,  5.0903e-03,  3.9689e-04,  2.8938e-03,
         6.3939e-03,  1.2735e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0128e-03,  9.6288e-01, -7.8830e-03,  3.8025e-03,  3.6955e-04,
         6.1590e-03, -5.3977e-03,  2.8950e-03,  5.3466e-04,  1.3460e-03,
         3.7185e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0305e-02,  9.3756e-01,  4.6490e-03, -3.1755e-03, -4.1491e-03,
         9.1215e-04,  5.8599e-03, -1.0629e-03,  2.6525e-03,  6.2402e-03,
        -3.6718e-03,  4.9547e-03,  4.8589e-03, -9.9517e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3952e-03,  9.4424e-01,  2.5466e-03, -1.6725e-03, -1.1095e-03,
         5.3129e-04,  2.6532e-03, -1.9804e-03, -2.2718e-03, -4.4832e-03,
        -4.8883e-04,  1.5112e-03, -1.7402e-04, -2.8931e-03, -8.4421e-04,
         3.8026e-03,  4.9091e-04,  1.3742e-03,  1.5865e-03,  1.3334e-03,
        -8.4623e-05, -2.6746e-03, -1.4897e-03,  2.4593e-04,  4.7518e-04,
         1.6233e-03,  5.3575e-04,  5.1435e-04,  9.9398e-04,  1.2669e-03,
        -5.4663e-04,  4.1342e-03, -1.2302e-03,  5.8025e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3798e-04,  9.3840e-01,  1.4544e-03,  5.8748e-03, -1.4296e-04,
         1.1594e-03, -3.3381e-05, -6.4744e-03, -1.4074e-03,  2.9232e-03,
        -6.5856e-03,  1.4335e-03, -6.5119e-05,  1.3542e-03,  6.0365e-03,
        -9.6412e-04, -2.0447e-03, -6.8432e-03, -3.0004e-03,  6.4999e-04,
        -5.1369e-04,  1.2493e-04, -5.8199e-03, -1.3419e-03,  1.2109e-03,
         1.1928e-03, -7.4800e-04, -1.9622e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1905e-02,  9.4488e-01, -3.3723e-03, -1.4320e-03,  1.0526e-03,
         1.3036e-03, -7.1340e-04,  1.2521e-03, -1.4744e-04,  1.2812e-03,
        -2.8055e-04, -2.9097e-03, -1.6584e-03,  3.7121e-04, -5.0421e-04,
        -2.6501e-03, -2.7891e-03,  2.0819e-03, -1.1355e-03, -7.1533e-04,
        -5.9505e-04,  2.9269e-03, -2.6407e-04,  3.7825e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-1.6163e-02,  9.3263e-01, -5.9132e-03,  4.7021e-03,  7.1233e-04,
        -4.0533e-03,  5.0649e-04, -3.0182e-03, -6.9748e-04, -1.8720e-03,
         8.4402e-04, -1.0775e-03, -2.7085e-04, -1.4991e-03,  4.7782e-04,
         1.4246e-05, -7.5108e-04,  1.6209e-03,  3.6963e-04,  5.9066e-03,
        -1.9024e-03, -1.2338e-03,  3.4647e-03,  2.9890e-04, -1.4352e-03,
         3.6661e-03, -3.0075e-05,  5.0292e-04, -1.0367e-03, -1.6878e-03,
        -1.6402e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0022,  0.9517, -0.0077,  0.0060, -0.0017, -0.0077,  0.0012, -0.0031,
         0.0055,  0.0037,  0.0012, -0.0083,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.1903e-04,  9.6902e-01,  1.5136e-04, -4.8579e-03,  2.0002e-03,
         1.2254e-03, -8.5494e-04,  2.2091e-03, -2.9674e-04,  4.8371e-03,
         2.8305e-03,  5.9157e-04, -4.9378e-03, -4.2237e-03,  1.2454e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0480e-02,  9.4382e-01,  8.6830e-03, -3.8288e-03, -2.5606e-03,
         1.9371e-03, -1.9453e-03, -1.2153e-03,  7.9941e-04, -1.0912e-03,
        -1.2288e-03,  2.4553e-04,  1.9138e-04,  2.1454e-04,  1.1680e-03,
        -3.0444e-04,  8.4392e-04, -6.9802e-05, -1.8635e-03, -3.9620e-03,
         2.2758e-04,  4.7270e-04, -3.9999e-04, -2.7821e-03, -3.7637e-04,
         1.6256e-03,  7.0209e-04, -1.4953e-03, -5.6523e-04, -3.9979e-04,
         1.3430e-03, -1.1356e-03,  2.0236e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4964e-03,  9.6625e-01, -6.8483e-03, -7.0345e-03,  1.1366e-03,
         2.0083e-03,  1.5200e-03,  8.4788e-04, -7.2479e-04, -4.7841e-04,
        -2.7390e-03, -1.3432e-04, -7.2641e-04, -4.3251e-04,  5.2760e-03,
        -1.3511e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8830e-02,  9.4246e-01,  5.0130e-03,  3.0927e-04, -1.2565e-03,
        -8.6674e-03, -1.4103e-02,  1.6936e-03, -7.6681e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0242,  0.9433, -0.0026,  0.0080,  0.0220,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1466e-03,  9.4865e-01, -4.2476e-03, -6.8902e-03, -3.2259e-03,
        -2.0833e-03, -3.7227e-03, -1.6337e-03, -2.0974e-04, -2.7356e-03,
        -1.9558e-03, -2.1730e-03, -3.4376e-03,  3.3741e-03, -3.3630e-04,
        -2.5304e-04,  1.2825e-03, -1.0517e-03, -1.4552e-03, -8.4455e-04,
        -1.6865e-03, -2.9742e-04, -9.2659e-04, -1.0538e-03, -6.5272e-05,
        -7.9250e-04, -4.4732e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0045, -0.9681,  0.0039,  0.0067,  0.0083,  0.0026,  0.0060,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3499e-02, -9.1956e-01, -2.3015e-03,  1.7109e-04, -1.0621e-03,
        -2.8675e-03,  2.4551e-03,  1.7426e-03, -4.0155e-03, -1.0656e-02,
        -1.6422e-03, -7.2132e-03,  1.1663e-03, -9.8107e-04,  2.5723e-03,
         3.6564e-03, -7.6318e-04,  4.3920e-04, -7.5011e-04, -1.1215e-03,
        -3.0854e-03, -8.2811e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4141e-03,  9.4394e-01,  7.6578e-03, -2.2804e-03,  8.9157e-03,
        -3.4725e-03, -2.0570e-03, -4.2443e-04, -1.6097e-04, -7.8876e-03,
         5.4874e-04, -1.7479e-03, -5.9301e-03, -6.4397e-03, -1.0654e-04,
        -1.6626e-03, -6.1931e-04, -2.7998e-03, -1.9358e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.7599e-03,  9.5648e-01, -3.8978e-03,  1.1311e-05, -1.1991e-03,
         5.9539e-04,  1.8778e-03, -9.6460e-04,  2.3597e-03, -9.9288e-04,
         1.0594e-03, -1.6484e-03, -5.7873e-05,  1.3788e-03,  1.0125e-03,
        -6.8199e-04, -2.8957e-03, -1.6357e-03, -5.9267e-04,  6.1542e-04,
        -1.3943e-03, -2.9161e-03, -1.0411e-03,  5.9262e-04,  4.0688e-04,
         6.1548e-04, -1.4508e-03, -1.6214e-03,  1.0065e-03,  7.8301e-06,
         1.2288e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-3.1501e-03, -9.2362e-01,  1.0234e-02,  2.8020e-03,  4.3456e-03,
         4.6293e-03, -6.3158e-04,  5.9192e-04, -7.5536e-04,  1.2787e-04,
         2.0379e-03,  3.9717e-03,  1.0859e-03,  1.2271e-02,  1.7709e-03,
        -3.2353e-03,  3.0547e-03, -1.7412e-03, -3.5665e-03,  6.0843e-04,
         9.0791e-05, -9.1577e-04,  1.3827e-04, -1.6952e-03,  1.3839e-04,
        -2.3890e-05, -9.5764e-04, -2.4919e-03, -8.5045e-04,  1.4869e-05,
        -1.6387e-03,  6.8137e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.0019e-03,  9.5556e-01, -6.0205e-05, -2.0788e-03,  1.5988e-03,
         5.7354e-04, -9.9132e-04,  1.1193e-03,  1.8411e-02, -7.4365e-04,
        -1.6568e-03,  1.8706e-03,  2.4316e-03, -1.2975e-04, -3.5333e-04,
         6.2473e-04,  9.3557e-04,  3.0058e-04, -4.3160e-03,  1.7654e-04,
        -2.4467e-04,  8.1816e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5827e-03,  9.4812e-01, -6.5482e-03, -4.9518e-03, -5.3514e-04,
         2.8800e-03,  1.0047e-03,  2.2709e-03,  3.8943e-04,  1.6482e-03,
        -1.2007e-03, -1.5746e-04, -6.4512e-05,  6.6565e-04, -1.8585e-03,
         1.1592e-04, -6.4842e-04,  6.8559e-03, -2.5946e-03, -2.4978e-04,
         1.9669e-05, -3.3430e-03,  1.1375e-03, -2.7735e-04, -1.3379e-03,
         1.5709e-03,  7.8501e-04, -1.9973e-03,  1.0569e-03, -1.3201e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3444e-05,  9.5451e-01, -8.9795e-03,  1.2116e-03, -3.2788e-03,
        -5.4657e-03,  2.7334e-04,  7.5382e-04, -1.7304e-03, -1.8606e-03,
         9.5564e-04,  2.6713e-03, -2.4626e-03, -4.4987e-04, -2.9718e-04,
        -1.6618e-03, -5.4097e-03, -8.0133e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.4726e-03,  9.3328e-01,  1.4036e-02, -1.8573e-03, -1.5662e-03,
        -2.7153e-03,  4.0598e-04, -8.4043e-05, -3.6590e-03, -1.0198e-03,
        -1.0809e-03, -4.1689e-04,  8.2239e-04, -2.2450e-04,  1.4626e-03,
        -1.4064e-04,  9.7470e-04, -3.2593e-03, -2.3459e-03, -3.2514e-04,
        -1.3112e-03, -4.5212e-03,  4.2336e-04,  9.2741e-04,  8.7713e-04,
        -4.0750e-05, -2.0818e-03,  3.9661e-04, -3.1421e-05,  4.6434e-04,
         1.1448e-03, -1.4934e-03,  8.6976e-04, -1.3079e-04,  7.8973e-04,
         4.1611e-03,  3.1899e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5465e-02,  9.3069e-01,  7.9828e-03,  5.6346e-04, -5.0009e-04,
        -1.5793e-04, -1.8016e-03, -1.1429e-03, -1.6683e-03,  3.0806e-04,
        -9.1622e-04, -2.7246e-03, -4.6475e-03,  1.2369e-03,  5.6134e-03,
         1.7302e-03, -7.0101e-05,  2.2168e-04,  1.4303e-03, -3.2534e-03,
         1.1229e-03, -1.6773e-03,  4.1429e-03, -9.2786e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2619e-03,  9.5215e-01, -6.0133e-03, -1.4379e-03, -1.6396e-03,
         1.3577e-03, -1.2691e-03,  1.6356e-03,  1.5893e-04, -1.1842e-03,
         3.5597e-04,  2.5775e-04,  2.4864e-04,  6.9908e-04, -1.1835e-03,
         6.7347e-04, -2.5855e-04, -7.7510e-04, -1.2169e-03,  8.2332e-04,
        -8.6966e-04, -4.9926e-04,  5.3744e-04,  1.4301e-03,  2.4394e-04,
         7.3351e-03,  9.7689e-04,  1.3779e-04, -3.7463e-04,  2.7299e-04,
        -3.1507e-04,  4.5112e-04,  7.7731e-04,  4.2575e-04, -6.3332e-04,
        -5.0237e-05,  1.0768e-03,  4.9481e-04, -3.0492e-04, -6.0545e-05,
         3.0654e-04,  2.3273e-04,  1.2530e-03,  2.6208e-03,  2.3864e-04,
         1.4828e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5366e-03,  9.4996e-01, -1.0795e-02,  4.1427e-04, -3.2969e-03,
        -1.1144e-03, -4.0617e-03,  1.3729e-03,  1.2088e-03, -1.7334e-03,
         2.0220e-03, -2.1053e-03, -2.9842e-04, -1.6400e-03,  1.3071e-03,
         1.3357e-04,  1.5710e-04,  8.1534e-04,  2.4421e-03,  5.4121e-04,
         2.1990e-03,  2.2646e-04,  6.6166e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2781e-02,  9.6263e-01,  6.4534e-03, -6.5630e-03,  2.7385e-03,
        -3.2137e-03, -1.2099e-03, -1.7939e-03, -1.2659e-03, -8.8826e-04,
        -4.6696e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3258e-03,  9.6872e-01,  3.3528e-03,  3.0561e-03,  5.4734e-04,
        -1.2070e-03, -2.8665e-03,  3.7584e-03,  4.6305e-03, -2.1825e-04,
        -1.5379e-04, -2.1418e-03, -1.6711e-04, -6.8527e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2283e-03,  9.5419e-01, -5.8295e-03, -1.0617e-02, -3.5335e-03,
         6.5103e-05, -2.8421e-03,  7.0691e-04, -1.2116e-03, -1.0660e-02,
         3.3258e-04, -6.7847e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.4188e-03,  9.3269e-01,  3.6623e-03,  5.1284e-03,  5.3646e-04,
         2.9347e-03,  4.6263e-04,  1.3744e-03,  3.9347e-05,  2.7836e-03,
        -6.7880e-04,  7.2348e-04, -1.4222e-04,  2.7327e-03, -4.7502e-05,
         5.7351e-04,  1.5170e-04, -8.8000e-04,  1.7879e-03, -7.6604e-04,
        -4.1434e-04,  2.8467e-04,  1.9819e-03, -3.2991e-04,  1.3049e-03,
        -3.2697e-04,  4.7190e-03,  9.2121e-04, -2.7241e-05,  2.1163e-03,
        -1.2502e-03,  5.0218e-03,  3.7840e-04,  2.4886e-03,  9.6831e-04,
         2.4878e-04,  2.8996e-04,  1.1620e-03,  1.0340e-03, -6.2680e-04,
         6.4825e-04, -3.4736e-05, -4.0088e-04, -5.2822e-04, -1.5663e-04,
         2.9861e-03,  6.9489e-04, -1.2132e-05, -2.3400e-04, -8.7598e-04,
         2.2501e-04,  1.0653e-03, -2.3092e-04, -5.2804e-04,  9.6675e-04],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 8.2632e-03, -9.4097e-01,  8.3425e-03,  2.7246e-03, -3.1361e-03,
        -6.3213e-04,  4.3060e-04, -2.2956e-04, -1.4271e-04,  2.8024e-03,
         6.0428e-03,  1.1215e-03, -4.2264e-03, -2.0226e-03,  1.5233e-03,
         2.9290e-04,  9.9059e-04, -4.4271e-03, -1.8060e-03,  1.3732e-03,
         1.2266e-03,  2.1016e-03, -5.1740e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.3118e-03,  9.6791e-01,  6.6214e-03, -1.4930e-03,  8.6303e-04,
        -1.1469e-03, -2.9833e-03,  2.2952e-04, -4.0568e-04, -3.7924e-03,
        -7.6873e-04, -2.0429e-03, -1.4066e-03, -8.1152e-04, -8.9474e-04,
         2.4006e-04, -2.9725e-04, -1.7762e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4713e-03,  9.2662e-01,  2.2547e-02,  4.0509e-03,  4.5139e-04,
         4.5164e-04,  4.1675e-03,  2.8852e-03, -1.2508e-03, -4.4493e-03,
         4.6944e-04, -1.4273e-03,  6.1707e-03, -1.2489e-03,  8.7438e-03,
         5.6538e-04,  5.0140e-03,  4.0174e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2910e-02,  9.2281e-01,  9.1248e-03,  2.6092e-02,  2.6458e-03,
         1.5675e-03, -3.3978e-04,  2.2797e-04,  2.7976e-03,  1.0157e-03,
         1.5492e-03,  4.2537e-03,  4.7368e-03,  7.2811e-03,  2.6464e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.9676e-03,  9.2606e-01,  4.6602e-03, -1.3555e-03,  2.9829e-03,
        -3.3076e-03, -5.7289e-03,  2.0853e-03, -3.7512e-04, -1.5624e-03,
         3.6004e-03, -1.6537e-03,  1.8764e-03,  6.7592e-04, -2.7003e-03,
         7.4617e-04, -3.9332e-03, -1.2228e-03,  9.9684e-04,  3.9067e-03,
         3.4884e-03, -2.6273e-03, -1.0352e-03,  3.1075e-04,  2.8901e-03,
         1.4844e-03,  1.8239e-03,  6.3897e-04,  4.9201e-03,  1.7106e-03,
        -1.0573e-03, -1.6197e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.1438e-03,  9.4153e-01, -1.6404e-03, -5.2218e-03,  7.3790e-03,
         6.3105e-04,  1.4293e-03, -6.2339e-04,  2.7439e-03,  1.6160e-03,
        -5.6166e-04, -1.6358e-03, -3.7541e-03,  1.8993e-03,  9.1471e-04,
         5.8400e-04, -4.7511e-04, -6.1641e-06, -8.9316e-04,  1.8154e-04,
        -5.0712e-04,  2.0329e-03, -1.4444e-03, -2.2555e-03,  6.7599e-04,
        -2.3383e-06,  2.1903e-03, -1.4154e-03, -9.8670e-04,  1.1775e-03,
        -1.6566e-03,  1.9191e-03, -8.6722e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1137e-02,  9.3994e-01,  4.6839e-04, -1.2051e-03, -2.1878e-03,
        -2.4133e-03,  2.4747e-05,  1.2790e-03,  3.5564e-03,  2.6596e-03,
         1.5744e-04, -3.6191e-03, -3.7290e-04, -1.1727e-03, -1.8811e-05,
        -1.5956e-04,  1.6528e-03,  9.4402e-04,  1.4311e-03, -1.5347e-04,
         8.0931e-04, -1.1635e-03,  1.0588e-03, -4.7279e-04,  2.0597e-04,
         3.6437e-04,  5.4023e-05, -1.0704e-03, -1.6863e-03, -1.9310e-03,
         2.3155e-03,  3.9481e-04,  1.9852e-03,  3.9947e-05,  2.8275e-04,
         1.2837e-03,  2.2175e-04,  3.7202e-04,  5.8085e-04, -4.0882e-04,
         2.7519e-04, -9.3866e-04,  7.5958e-04,  2.1229e-05,  4.1393e-04,
         3.8670e-05,  3.7822e-04, -1.6674e-04,  5.5527e-04,  8.7856e-04,
        -4.6376e-04,  2.8354e-04, -5.0633e-04,  5.1996e-04,  4.6054e-04,
         1.1317e-03,  2.5785e-05, -9.2725e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.6871e-03, -9.4598e-01,  8.8889e-03,  5.6637e-03,  1.7614e-03,
         7.5011e-04,  3.0414e-04, -1.3002e-04,  1.6606e-03,  4.9315e-03,
         7.2122e-04,  2.9659e-04, -2.6427e-03, -2.9107e-03, -4.4282e-03,
         6.3893e-04, -1.1233e-03, -1.6046e-03, -8.4293e-04, -2.4392e-03,
        -8.7010e-04, -3.1010e-03,  1.3141e-03,  2.9002e-05,  1.2041e-03,
         1.0789e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.6603e-03,  9.4109e-01, -7.1972e-03,  6.2364e-03, -7.5241e-04,
         3.6235e-03,  3.2018e-03,  1.8129e-03,  6.4950e-04,  1.1369e-03,
         6.0419e-03,  3.7816e-04,  3.1433e-03,  3.5421e-03, -7.5822e-04,
         8.5072e-04, -3.6603e-04,  8.0143e-04,  7.8060e-04,  1.3966e-03,
        -8.1844e-04, -1.5240e-04, -1.0050e-03, -2.2806e-03, -2.3258e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.7375e-03,  9.3271e-01,  1.2647e-02, -2.5824e-03,  1.2081e-03,
         1.1684e-05, -3.6715e-04, -3.5128e-03,  1.5318e-03,  2.3491e-03,
        -1.1329e-03, -4.8615e-04,  7.8865e-04,  2.0455e-03,  1.1457e-03,
         3.1362e-04,  1.3549e-03, -1.0298e-03, -1.4919e-04,  7.8475e-04,
        -4.0463e-05, -1.8369e-03, -2.9030e-04,  4.4298e-04, -4.6037e-04,
         1.4010e-03,  1.1571e-03, -2.1583e-04, -2.4984e-04, -2.4191e-04,
         6.6119e-04, -1.2671e-03, -2.2729e-03, -7.6002e-04,  2.8952e-03,
        -3.0997e-03, -2.6247e-04, -6.7688e-04, -1.5883e-03, -1.7306e-03,
        -5.6927e-04, -1.0118e-03, -3.9149e-04,  1.6774e-03, -2.9126e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2472e-02,  9.3841e-01,  2.5159e-03,  2.7620e-03, -1.5810e-03,
        -5.8364e-03, -2.9369e-04, -2.9793e-03, -7.2295e-04, -5.1324e-03,
        -3.2845e-03,  2.1626e-04, -3.4142e-03, -4.3155e-04,  1.3034e-04,
         7.5345e-04,  2.7068e-03, -4.1884e-03, -1.0872e-03, -1.8444e-04,
         3.9725e-03, -1.0485e-03,  5.8780e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0518e-02,  9.2279e-01,  2.4147e-02,  9.1850e-03, -4.2787e-03,
         2.0678e-02,  1.2846e-04, -2.8534e-03, -1.1846e-03,  2.5509e-04,
         2.6131e-04,  3.7181e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([-6.6552e-03,  9.4110e-01, -1.5687e-02, -3.3054e-03,  1.0252e-03,
         1.4520e-03,  3.3435e-03, -1.0665e-03,  5.8567e-04, -1.7767e-03,
         3.7698e-03, -3.9578e-03, -7.9189e-04,  9.5865e-04,  4.2422e-04,
         2.5821e-03, -2.7910e-04, -5.7712e-06,  9.5904e-04, -5.6387e-03,
        -9.9399e-06,  1.1475e-03,  3.4815e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2711e-02, -9.5319e-01,  1.9947e-03, -1.6424e-03,  1.3461e-03,
         1.0198e-03, -8.3162e-04,  9.5597e-04,  9.5636e-04,  6.7702e-04,
        -4.8260e-04,  3.2604e-03,  2.4956e-04,  2.0220e-03,  6.8732e-04,
         1.2784e-03,  2.1508e-04, -1.1353e-03, -7.1866e-04, -1.0342e-03,
        -7.7233e-04,  1.1690e-03,  2.4847e-03, -1.2546e-04, -4.5551e-04,
         7.0067e-04,  5.4947e-04, -4.9079e-04,  9.3562e-04,  2.9386e-04,
        -8.4224e-04, -1.8237e-03, -2.9492e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4659e-02, -9.5118e-01,  1.5320e-03,  1.9338e-03, -2.0355e-03,
         3.7694e-04,  2.1673e-03,  3.5006e-03,  1.5150e-03, -7.2456e-05,
        -9.1558e-04,  2.2704e-03,  2.9847e-03,  3.4083e-03, -1.3804e-03,
         5.4144e-04,  1.9720e-04,  1.7448e-03,  7.5830e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.4507e-03,  9.4657e-01, -2.4609e-03,  1.6893e-02,  1.9389e-04,
         7.4705e-03, -2.7382e-03, -1.5117e-03, -2.2360e-03, -6.2691e-04,
         1.5043e-04, -2.8704e-03,  4.4721e-03, -4.3507e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.1356e-03, -9.4967e-01,  7.8786e-03,  5.1326e-03,  4.2430e-04,
        -1.9579e-03, -2.0455e-03, -2.7835e-04, -1.3591e-03, -5.6187e-04,
        -2.0490e-03, -2.7740e-04, -3.9245e-03, -3.1638e-03,  7.6335e-04,
         1.1696e-03, -3.3204e-03, -4.7363e-04, -9.5116e-04,  1.0082e-03,
        -1.9301e-03, -7.5290e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1308e-05,  9.5285e-01,  2.0251e-03, -6.0282e-05, -1.8319e-04,
         1.1261e-03,  1.8796e-03, -3.2400e-03, -9.9177e-04,  5.3790e-04,
        -2.0985e-04,  5.1324e-03,  6.5353e-03,  2.2453e-04, -3.5682e-03,
        -1.4531e-03,  3.0041e-03,  3.0764e-03,  1.2718e-04,  1.3008e-03,
        -3.2798e-04,  8.6574e-04,  8.0717e-04,  1.7163e-03, -1.2167e-03,
         5.6210e-04,  1.3291e-03, -1.3024e-03,  2.5838e-04,  4.0540e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2706e-02, -9.3032e-01,  9.5315e-03,  1.2555e-02, -2.9281e-03,
        -2.0150e-03,  2.7985e-04,  6.8100e-03,  1.7365e-04,  2.1204e-03,
         2.7460e-03,  2.1905e-03,  3.9925e-03,  2.6219e-03,  1.5792e-03,
         1.0182e-03,  6.4091e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0103e-02,  9.5306e-01,  5.7067e-04,  2.7553e-03, -2.2726e-05,
         1.4484e-03,  1.0615e-03,  2.9480e-03, -2.7153e-04,  2.0244e-03,
        -5.7326e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7862e-02, -9.4980e-01, -3.4860e-03, -1.5435e-04,  4.9306e-04,
         1.2117e-03,  7.4281e-04,  1.5953e-03,  4.2562e-03, -1.1637e-03,
         2.7528e-03,  1.8198e-05,  4.0452e-03,  3.2277e-03, -9.1089e-04,
         2.0573e-03,  6.2243e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0622e-02,  9.3502e-01,  1.0861e-02, -4.1784e-04,  1.9883e-03,
        -1.5539e-03,  3.4942e-03,  5.6241e-04,  5.6384e-04, -1.7233e-04,
         2.2395e-03,  1.6166e-03,  1.7483e-03, -3.7315e-03,  5.0348e-03,
        -1.0819e-03,  2.0045e-04,  2.7564e-03,  9.7664e-04,  3.0857e-03,
        -1.4336e-03,  1.0226e-03,  4.3949e-03,  1.0140e-03,  3.0801e-04,
         4.0987e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8865e-03,  9.5214e-01, -6.2502e-03, -1.1580e-03,  4.1820e-04,
         3.1762e-04,  1.8639e-03,  3.6337e-03,  2.4479e-03,  2.1146e-04,
         5.1831e-04,  2.4350e-03, -3.4665e-04, -1.8355e-03,  3.9113e-05,
        -5.5151e-04, -8.0459e-04,  3.7606e-04,  1.0552e-03,  5.6619e-04,
        -1.5325e-03,  3.8077e-03, -2.4648e-04, -1.3038e-03, -1.9773e-03,
        -7.0568e-04,  1.7850e-03, -2.9706e-03, -3.3038e-03, -1.5120e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4675e-03,  9.5825e-01, -1.1849e-03, -1.3703e-03,  3.4460e-03,
        -4.9794e-04,  3.8120e-03, -1.8612e-03,  8.9243e-04,  4.2196e-03,
        -1.9052e-03,  4.2937e-03, -2.4630e-03, -1.1663e-03, -1.9224e-03,
        -3.0479e-04, -2.6238e-03, -4.5157e-04, -6.8693e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-7.5465e-03, -9.4253e-01,  3.8248e-04,  5.2738e-03,  2.5490e-03,
         4.0422e-04, -1.4308e-02,  2.4137e-03,  2.8382e-03, -1.2508e-03,
        -3.0933e-03,  6.7034e-04, -2.0554e-04,  1.3539e-03, -1.2522e-03,
        -3.4567e-04,  7.4868e-04,  4.1192e-04, -1.4995e-03,  5.7681e-05,
        -4.2961e-04,  3.9575e-03,  1.1505e-03, -1.0328e-03,  2.0032e-03,
        -7.0807e-04,  1.5845e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.9137e-03,  9.5095e-01, -1.9991e-03, -1.3123e-02, -3.8500e-03,
        -3.9242e-03, -1.4460e-03,  3.4782e-04,  2.2492e-03, -1.2499e-03,
        -1.1217e-03, -3.1814e-03,  2.3271e-03, -7.3132e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.0549e-03,  9.4751e-01, -4.3256e-04, -1.1521e-03,  1.6813e-03,
         8.2703e-04,  2.5845e-03,  5.2175e-03,  1.5418e-03,  1.7410e-03,
         9.4598e-04,  7.4113e-04,  1.9497e-03,  2.0905e-03,  1.8807e-03,
         3.4079e-03,  1.6467e-03, -2.1158e-03,  1.5914e-03,  4.7973e-03,
        -2.2189e-03, -2.4554e-03,  2.3875e-04,  5.2318e-04, -4.1355e-04,
        -4.2395e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5587e-02,  9.4946e-01,  5.3204e-03, -3.8486e-03, -2.9394e-03,
        -9.2171e-03,  2.4799e-03,  2.9405e-04,  1.7013e-03, -5.9060e-03,
         1.0093e-04,  1.6373e-03, -3.2776e-04, -1.1806e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.1409e-04,  9.7166e-01, -7.4502e-04,  6.6177e-04, -1.9583e-03,
        -3.0393e-04, -5.2731e-04,  4.9139e-05, -4.2643e-04, -1.7405e-03,
         2.9000e-04, -9.9036e-04,  2.0229e-04, -1.6577e-03,  1.1201e-03,
         9.5860e-04, -3.5557e-04, -1.4882e-03,  4.6665e-04, -3.1595e-03,
        -5.6370e-04,  1.5532e-04, -1.0993e-03, -2.5824e-03,  1.3051e-03,
        -9.1411e-04,  2.6538e-04, -1.4015e-03, -1.1425e-04, -1.1583e-03,
         6.6145e-05, -8.0034e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1021e-03,  9.6083e-01,  3.1466e-03, -6.1054e-03,  3.9069e-03,
        -5.9436e-04, -5.0271e-03, -2.9785e-03, -3.3320e-04, -1.4333e-06,
        -2.8798e-03, -4.2506e-04, -5.1677e-05, -8.3739e-04,  7.7398e-04,
        -2.8678e-03, -1.6204e-03,  8.3838e-04, -1.4851e-03,  4.3697e-05,
         1.0625e-05,  7.8732e-04, -1.0321e-03,  4.5448e-05,  2.2772e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2331e-02,  9.5079e-01,  1.6169e-02, -1.8350e-04,  4.2310e-03,
         6.2978e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1218e-02,  9.3957e-01, -1.7499e-03,  2.5178e-03,  2.6751e-03,
        -2.4608e-03, -2.0176e-03, -1.6201e-03, -1.2308e-03,  1.5244e-03,
        -5.0632e-04,  1.2930e-03,  1.5040e-03, -2.1356e-04, -8.1500e-04,
         5.4435e-04,  7.4771e-04,  5.0723e-04, -2.4335e-04,  1.2707e-03,
        -8.9075e-04,  6.6654e-05, -2.1618e-04, -1.7457e-04, -1.1177e-03,
         9.8633e-04,  7.5142e-04,  9.6064e-04, -5.6931e-04, -1.3779e-04,
         5.4827e-04, -1.0621e-05, -5.5541e-04, -1.3159e-04,  8.8551e-04,
         6.7793e-03,  9.9295e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8432e-03, -9.3863e-01,  4.0973e-03, -1.9904e-03,  2.3148e-03,
        -1.4260e-03,  5.8004e-04,  8.0852e-04, -7.0563e-04,  5.4295e-04,
         8.3144e-04, -3.0384e-03,  6.7953e-04,  3.0837e-04,  2.4563e-03,
         3.4226e-03,  2.0776e-03,  4.9301e-04, -3.9663e-04, -4.7735e-03,
         1.1700e-02,  8.2074e-04,  3.5106e-03,  1.1535e-03, -4.1733e-07,
        -1.1821e-03,  3.1955e-03,  7.0862e-04, -1.7103e-04, -2.3649e-03,
         6.8027e-04,  5.7550e-04,  6.1987e-04,  2.3268e-04, -1.6721e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2431e-02, -9.0185e-01,  5.6452e-02,  4.3366e-03,  5.3310e-03,
        -2.5103e-04,  2.0245e-03,  6.4931e-04, -4.9431e-04,  1.0568e-03,
         6.1910e-04,  3.8878e-03,  2.4413e-03,  6.2211e-04,  7.5578e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2738e-04, -9.5400e-01, -1.3756e-04,  5.0077e-03, -3.4938e-04,
         1.2054e-03, -9.7733e-04, -1.0584e-03, -1.8787e-03, -2.5083e-03,
        -1.6789e-03,  7.8750e-05, -2.1758e-03,  1.7884e-03, -2.1107e-03,
         4.1429e-04, -1.3983e-03, -1.2586e-04, -5.6842e-03, -1.3250e-03,
         7.3330e-04, -1.0469e-03, -2.1682e-03, -1.3592e-03,  3.4463e-04,
        -1.4691e-03, -6.1259e-04, -1.0946e-03, -2.8340e-04, -1.4493e-03,
        -1.6619e-03,  4.7640e-04, -3.0688e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7679e-02, -9.2569e-01,  5.1396e-03, -1.2539e-03, -2.9006e-04,
         5.2325e-03, -7.8160e-04, -1.0991e-03, -1.8132e-03, -6.5207e-04,
         9.6911e-04, -3.4672e-03, -6.0307e-05,  2.1875e-04, -1.8224e-03,
        -2.1548e-03, -3.4159e-03, -6.1976e-06, -9.1282e-04,  9.2040e-06,
         2.1039e-04, -3.1045e-04, -1.7539e-03, -6.9836e-04, -1.8728e-03,
         2.0733e-03,  3.5234e-04, -4.4976e-03,  5.5594e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([-3.4189e-03,  9.7003e-01, -3.2897e-03, -1.9592e-03, -4.5435e-03,
         2.2100e-03,  1.7037e-03, -1.6841e-03, -1.9180e-03, -3.3770e-03,
        -9.7651e-04,  2.0677e-03,  2.0335e-04, -4.0553e-04, -2.2165e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0279e-03, -9.4191e-01, -1.1003e-02, -2.2065e-03,  8.2031e-04,
        -6.7065e-04, -1.3377e-03,  3.8649e-03, -1.6383e-03, -9.8319e-04,
        -7.8091e-05, -5.2880e-04, -1.0075e-03,  7.2914e-06,  7.1257e-04,
        -3.0523e-03,  7.7113e-05, -2.0198e-03, -7.2214e-04,  1.6066e-04,
        -8.6353e-04,  1.0315e-03, -2.8770e-04, -6.5383e-04, -4.8194e-04,
        -6.8796e-04, -3.3822e-04, -1.5484e-03,  9.7334e-04, -2.7177e-04,
        -8.7292e-04, -5.7157e-05,  2.2737e-04, -4.5467e-05, -8.5595e-04,
        -1.1632e-03, -1.1255e-03,  4.3440e-04, -9.8977e-05,  9.2284e-04,
        -6.6425e-04, -1.3930e-03,  1.8789e-03, -5.1353e-04, -5.4208e-04,
        -7.4084e-04,  3.7833e-04, -4.5742e-04,  3.6568e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.7720e-04, -9.3399e-01,  6.0549e-03, -2.2053e-03,  3.5856e-03,
        -8.6458e-03,  3.6183e-03,  1.8703e-03, -1.3525e-03, -1.8224e-03,
        -2.7983e-03, -5.6783e-04,  4.3290e-05, -1.6387e-03, -7.2499e-04,
        -2.3146e-03, -4.4676e-03,  2.1060e-05,  4.1255e-04, -8.1390e-04,
         2.5916e-04, -6.0488e-04, -1.5887e-03, -9.7668e-04, -1.1407e-03,
         7.4880e-04,  1.4058e-03, -6.9793e-03, -5.1510e-04,  2.1103e-03,
        -1.1198e-03,  1.0071e-04, -1.0818e-03, -1.6904e-04,  3.4783e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.5809e-03,  9.4946e-01, -1.1666e-02, -6.9642e-03, -1.2098e-03,
         1.1873e-03, -3.8955e-03,  1.1044e-03, -2.6752e-03, -1.2322e-03,
        -7.2494e-04, -1.3497e-03, -1.5547e-04, -1.0298e-03,  4.1498e-04,
        -3.5103e-03,  1.2496e-03, -3.4637e-03,  1.2672e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2828e-02, -9.6138e-01,  1.0082e-03, -6.8961e-04,  2.3208e-03,
         2.8792e-03,  2.5667e-03, -6.0374e-04,  6.4774e-03,  7.2394e-04,
        -2.0878e-03, -3.1347e-04, -3.1375e-03,  2.8960e-03,  9.0936e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1121e-02, -9.5589e-01,  4.8838e-03, -1.2558e-03, -1.0866e-03,
        -4.3874e-04, -4.8249e-04, -2.6627e-05, -4.6084e-04, -3.8020e-03,
        -1.1568e-03,  2.1042e-04, -1.6029e-03,  7.5420e-04, -1.2553e-03,
        -2.5476e-04, -1.0141e-04,  2.7505e-03, -6.7540e-04, -5.3670e-03,
        -2.6401e-03,  3.7791e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5487e-02, -9.4450e-01,  6.5959e-04, -2.2239e-03, -1.2311e-03,
        -1.9507e-03,  2.3919e-03,  7.1369e-04,  5.0636e-03,  3.7737e-04,
         7.3295e-04, -2.0783e-03,  3.0769e-03,  2.2484e-03, -1.2757e-05,
         5.1054e-04,  4.3982e-03,  9.9589e-04,  1.4824e-03,  1.4348e-03,
         5.0941e-05,  5.8571e-04,  1.8185e-03,  4.6775e-04, -1.7280e-03,
         2.7581e-04,  1.1875e-05,  1.2418e-03,  5.0739e-04,  1.7449e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1041e-04,  9.6759e-01, -7.3082e-03, -3.7182e-03,  1.5570e-03,
        -5.0591e-03, -2.3823e-03, -9.3621e-04, -7.9914e-04, -3.5929e-04,
        -1.0078e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1897e-02,  9.5742e-01, -1.2951e-03,  1.0544e-03,  7.3110e-04,
         2.0392e-05,  1.2561e-03,  1.3502e-03,  1.3664e-03, -4.8169e-03,
        -3.2683e-04,  2.5212e-03,  4.8288e-04, -4.2577e-05, -6.8895e-04,
         1.4554e-03, -2.6566e-03,  5.4836e-04, -1.8392e-03,  1.1055e-03,
        -1.3205e-03, -5.8032e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5596e-02, -9.2771e-01, -5.8878e-03,  1.0610e-04, -8.3681e-04,
         4.5790e-03, -5.0309e-04,  1.3094e-03,  2.0394e-03, -3.7577e-04,
        -2.1110e-03, -1.6427e-03, -6.0083e-04,  1.7044e-03, -1.1905e-04,
         2.0229e-03, -1.8948e-03,  8.5803e-04,  2.0410e-03, -5.0178e-04,
        -1.6043e-04,  2.0162e-03, -1.9245e-03,  1.8206e-04, -9.4098e-04,
        -3.6796e-04,  2.4198e-05, -7.0862e-04,  6.6454e-03,  7.2820e-05,
        -2.3507e-03, -4.1421e-04,  1.7547e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6441e-03, -9.4893e-01,  7.5104e-04,  1.3516e-04, -1.1927e-03,
        -3.2282e-03, -3.0944e-03,  5.9933e-03, -2.9043e-03,  1.4270e-04,
        -5.5967e-05, -1.5411e-03, -9.1720e-04,  1.5687e-04, -5.6980e-04,
        -1.0598e-02, -2.7937e-03,  1.2966e-04, -9.8161e-04, -2.0027e-04,
         5.3833e-04,  2.7582e-03, -2.1962e-04, -2.0389e-03,  3.4199e-03,
        -5.1754e-04, -2.3205e-03, -2.2929e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3112e-02, -9.3482e-01,  4.6854e-03, -1.8747e-03,  1.7462e-03,
        -1.8689e-03, -1.1821e-03,  2.2552e-03,  8.8558e-03, -1.7938e-04,
        -5.7386e-04,  4.7569e-04,  2.3693e-03,  5.0314e-04,  1.1180e-03,
        -6.9975e-04,  2.6780e-03,  1.8270e-03,  1.5906e-03,  8.3305e-03,
         3.2590e-04,  1.0617e-05,  1.5612e-04,  8.2835e-03,  4.7852e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 2.8198e-03,  9.3088e-01,  8.2188e-03, -4.9440e-03, -5.2182e-04,
         2.7278e-03, -3.1525e-03, -1.5650e-04, -8.1996e-04, -2.8831e-04,
         2.7076e-03,  2.4343e-03,  4.5641e-04, -3.0417e-03,  1.4439e-03,
        -4.9852e-03, -8.0731e-04, -4.5149e-04,  3.3781e-04, -9.7530e-04,
         1.5580e-03,  3.6807e-04, -2.3729e-03, -7.8548e-04,  2.5936e-03,
         1.2156e-04,  8.7081e-04, -3.6617e-03,  1.3225e-04,  1.7830e-03,
        -1.9242e-03, -6.6490e-04, -1.5821e-03, -1.0069e-04, -8.4143e-04,
         1.0729e-03, -6.5258e-04, -4.6078e-04,  5.9880e-04, -2.9085e-03,
         2.7777e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3075e-03,  9.5070e-01, -5.9224e-04,  4.0090e-04, -1.9568e-03,
        -1.2324e-03, -1.3541e-03,  7.7596e-04, -2.0524e-03, -2.9137e-04,
         4.1637e-04, -4.1485e-03,  6.0608e-03, -1.1680e-03,  3.1421e-03,
         5.6951e-04, -1.2203e-03, -3.8826e-04,  2.5441e-05,  2.9605e-03,
        -3.6278e-04, -6.9598e-03, -5.8216e-04,  3.1225e-03,  8.2061e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9199e-02, -9.4381e-01, -1.1290e-02,  3.0111e-03, -2.0235e-03,
        -4.0968e-04,  2.1242e-03,  7.8268e-04,  8.0437e-04, -2.0473e-03,
         7.7041e-04, -1.0251e-03, -1.7803e-03,  1.0123e-03, -6.4254e-05,
         2.6432e-04,  2.5147e-03,  5.3506e-05,  6.4320e-04, -1.1294e-03,
        -5.2402e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7246e-03,  9.2889e-01,  1.9260e-02, -5.6237e-03,  5.6939e-03,
         4.8768e-03,  5.2759e-04,  1.9312e-03, -2.9067e-03, -6.2609e-03,
         4.8400e-03, -3.3192e-03,  1.0930e-03, -1.5728e-03, -2.0099e-03,
        -2.2956e-04, -5.0103e-03,  3.2302e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.7692e-03,  9.2786e-01,  2.1036e-02, -4.2536e-03,  3.2857e-03,
        -2.9660e-04,  4.1506e-03, -2.2113e-04,  1.3488e-03,  1.2623e-03,
        -8.5122e-04, -1.2956e-04, -5.5371e-04,  2.4544e-03, -7.2492e-04,
         2.1028e-03,  9.3874e-03, -7.8822e-04, -2.7851e-04, -3.5528e-04,
         1.1649e-03, -1.1720e-04, -1.2674e-03,  1.0952e-03, -1.0598e-04,
         2.4138e-04,  8.1897e-04, -1.0316e-03,  3.9135e-05, -8.8207e-04,
        -8.8098e-05,  2.9358e-05, -1.3391e-03,  6.8278e-04, -2.6010e-04,
         1.1977e-03, -8.6247e-04, -6.6803e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7141e-02, -9.2632e-01,  1.7584e-02,  6.5708e-04, -5.8560e-03,
        -1.8084e-03,  6.5767e-03,  7.2899e-04, -4.8735e-03,  2.0491e-04,
        -2.1944e-03,  6.3480e-04, -5.3715e-04,  2.6439e-03, -1.7047e-03,
         6.1445e-04, -1.7201e-03, -3.3813e-03,  6.4566e-04, -4.1723e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2987e-02,  9.3891e-01, -7.7529e-03,  7.0445e-03, -2.7336e-04,
        -3.5561e-03,  4.1481e-03,  5.7478e-04,  2.6469e-03,  2.9629e-03,
         6.3069e-03, -4.4233e-04,  3.7376e-03,  8.6582e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8222e-02,  9.4883e-01, -1.8851e-03,  2.0402e-03,  3.0143e-05,
        -1.9258e-05, -1.8318e-03,  5.8028e-05, -1.3069e-03,  6.3553e-04,
         1.2972e-03,  2.5115e-03,  1.0883e-03,  1.3003e-04, -6.6578e-04,
        -1.6705e-03, -1.2435e-03, -2.3009e-04,  5.9543e-04,  1.1297e-04,
        -2.1113e-03,  2.7334e-03,  3.4345e-04,  1.4275e-03, -2.0416e-04,
        -8.5538e-04, -5.3692e-04, -5.5841e-04, -1.1336e-03, -1.4662e-03,
        -1.9465e-03,  2.2755e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5444e-03,  9.3828e-01, -3.2606e-03, -5.7449e-03,  3.0446e-03,
        -1.7248e-03, -4.8670e-03, -1.7697e-03, -1.4444e-03, -2.9974e-03,
         1.1702e-03,  2.6724e-03, -9.1982e-04, -4.0078e-03,  2.3998e-03,
         6.3253e-04, -5.9390e-04, -8.3251e-03, -5.0709e-03, -7.5318e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.2200e-03,  9.1407e-01,  1.3756e-02, -1.8481e-03,  3.5262e-03,
         8.7033e-04,  1.0973e-03, -6.9170e-04, -5.9411e-04,  2.8763e-04,
        -1.7683e-03,  5.0193e-04, -1.9807e-03, -8.8059e-04, -2.3436e-03,
        -1.7255e-03, -1.0693e-03,  1.9038e-04, -1.9925e-03, -4.5972e-05,
        -2.2235e-03,  1.5377e-04, -1.8651e-04,  1.7000e-05,  6.6742e-04,
        -5.3596e-04, -3.4277e-03,  6.2145e-04, -6.1533e-03, -1.4452e-03,
         1.1882e-04,  1.4992e-03, -3.7882e-04,  3.5850e-03,  2.5164e-04,
        -8.0366e-04, -1.6449e-03, -1.8512e-03, -2.4464e-03, -6.0898e-05,
        -1.8589e-03, -4.9309e-04,  1.3172e-03,  1.4995e-03,  2.8109e-03,
        -8.4093e-04, -6.6426e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5481e-02,  9.5310e-01, -2.4016e-03, -3.2046e-03, -4.2196e-04,
        -1.1554e-03, -4.5956e-03,  3.9370e-03,  7.6924e-03, -1.0764e-03,
        -1.4666e-03, -2.1492e-03, -7.5708e-06, -5.2942e-04,  2.7821e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.6966e-03,  9.1495e-01, -8.3728e-03,  2.9362e-04, -3.7499e-03,
        -2.0291e-03,  5.6858e-04, -3.0116e-03,  7.8015e-04,  2.4621e-04,
        -3.1873e-03, -3.1664e-04,  1.4528e-06,  2.1342e-03, -3.6092e-03,
        -4.7749e-03, -3.2025e-03, -8.0243e-04, -1.2372e-04, -3.8706e-04,
        -5.5861e-04, -2.3349e-03, -4.7418e-04,  3.7109e-04,  4.8476e-04,
        -2.5493e-03,  2.4665e-04,  1.0938e-02, -1.3433e-03, -2.3587e-04,
         1.0257e-03, -2.7863e-04, -6.6515e-04, -2.7360e-04, -6.0297e-04,
        -1.2157e-03, -1.3126e-03, -1.1401e-03, -2.6035e-03, -1.0499e-03,
        -3.0664e-04,  1.6262e-03, -8.2736e-04, -1.2542e-03,  8.3037e-04,
        -4.6934e-04, -4.3758e-03, -3.6903e-04], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 7.1761e-03, -9.4316e-01, -8.8552e-03,  1.4292e-03, -1.1282e-03,
         6.8139e-03,  3.3610e-03, -1.9140e-03, -8.7537e-04,  5.0293e-04,
        -7.3751e-04,  3.7666e-03, -1.6644e-03, -8.7108e-04,  1.7899e-04,
        -9.2095e-04,  7.9071e-04, -5.2981e-04, -1.2893e-03,  1.5210e-03,
        -5.2094e-04,  2.9980e-04, -1.8107e-03,  7.7380e-04, -4.1344e-03,
        -4.9711e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1206e-04, -9.5326e-01, -3.4780e-03, -3.8430e-03, -2.7299e-03,
         1.7149e-03, -2.2618e-03, -7.9999e-03, -3.7632e-03, -2.6136e-03,
        -6.5526e-03,  4.2914e-04,  1.2879e-04, -3.5790e-03, -1.5683e-03,
        -9.6610e-04, -1.9877e-03, -7.3515e-04, -4.1393e-04,  3.8290e-04,
         1.2825e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3961e-02,  9.3646e-01,  4.4637e-03, -4.6305e-03, -2.9927e-03,
        -3.7542e-03,  1.2685e-04,  2.0001e-03,  2.8185e-04,  4.6059e-04,
        -9.3030e-04,  1.5412e-03,  1.7367e-03,  1.2834e-03,  3.6095e-03,
        -1.5784e-03,  7.5304e-05,  8.1955e-04,  2.0207e-04,  1.0159e-04,
         8.7366e-04, -1.8558e-03, -3.0281e-03,  1.2646e-03,  1.8469e-03,
         7.7196e-05, -1.9612e-03, -2.1672e-04,  1.8809e-04, -4.5079e-05,
        -1.1956e-03, -6.9579e-04,  2.7533e-03,  1.0461e-05, -1.3473e-03,
        -1.6340e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7868e-03,  9.5567e-01,  1.0802e-03, -1.1219e-03,  3.4978e-03,
        -3.6792e-03,  1.1464e-04,  1.5813e-03, -2.0430e-03,  3.8267e-03,
         3.0168e-03, -3.2167e-04,  1.2671e-03, -6.0312e-04,  4.4311e-04,
         2.8011e-03, -1.2384e-03,  3.7055e-03, -6.5389e-03,  4.6646e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6898e-03,  9.6286e-01, -4.8116e-03,  4.3006e-03,  7.6212e-04,
         8.8638e-04, -8.5980e-04,  1.4333e-03,  7.1340e-04, -8.7735e-04,
         2.6127e-03, -2.6008e-03,  2.6291e-03, -9.9967e-04,  2.6453e-03,
        -1.8293e-03, -1.9086e-03, -8.5280e-04,  5.6055e-04, -1.0424e-03,
         2.1287e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.5553e-03,  9.6331e-01,  5.2003e-03, -7.0420e-03,  1.8860e-03,
        -2.8996e-04, -1.2704e-04,  5.1784e-04,  5.2878e-03,  4.7562e-03,
        -7.8765e-04, -3.4701e-03,  7.7417e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7542e-02,  9.0635e-01,  1.0435e-02,  5.0624e-04,  6.0940e-03,
        -4.2336e-04,  3.6709e-03,  8.4944e-05,  1.0376e-04, -2.8236e-03,
         7.1408e-04, -4.5053e-03,  6.0167e-04, -1.3144e-03,  6.7565e-04,
        -4.7475e-04,  1.0499e-03,  2.6869e-03,  4.2501e-05, -6.7405e-04,
         1.1591e-03,  2.3489e-03, -9.6289e-04, -2.0520e-03,  5.8450e-04,
        -1.1940e-03, -1.0930e-03, -1.2684e-03, -8.0222e-04, -1.1972e-03,
        -1.1034e-02,  6.0176e-04, -1.1770e-04, -2.1390e-03, -8.2141e-05,
        -9.6902e-03,  2.8984e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9475e-03, -9.4806e-01, -1.8170e-02, -2.3596e-03,  2.3861e-03,
        -5.4483e-04, -1.3127e-03, -3.5832e-03, -1.0364e-03, -1.2859e-03,
        -2.6836e-03, -1.0798e-03,  1.3556e-03, -1.3761e-03, -7.4194e-04,
        -1.5167e-03, -1.7997e-03, -1.1376e-03, -5.7875e-04,  9.0739e-04,
        -1.4744e-03, -2.7483e-04, -2.8071e-03,  5.7894e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6896e-03,  9.4231e-01,  1.0280e-02,  4.0979e-03, -1.3814e-03,
        -1.2794e-04,  1.0202e-03,  4.0247e-04,  1.0037e-04, -1.0324e-03,
        -4.9289e-03, -4.3322e-03, -6.1418e-03,  2.7662e-03,  2.3512e-04,
         6.4372e-05,  1.5485e-04,  6.7937e-04, -8.0615e-04,  2.1478e-03,
         4.5214e-05, -3.4397e-04, -1.0580e-03, -1.5092e-03,  1.6584e-03,
        -1.5195e-04, -1.4004e-03, -2.1763e-03, -6.5945e-05,  8.0429e-04,
        -1.3753e-03, -1.7114e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6820e-03,  9.5018e-01,  6.0858e-03, -2.6864e-03,  2.5146e-03,
        -6.6352e-03, -4.6804e-04, -1.0804e-03, -2.1829e-03,  5.5894e-04,
        -1.7301e-04, -3.5090e-04, -1.0688e-05,  1.8961e-03,  1.1997e-04,
         3.2215e-04,  1.6736e-03,  5.2450e-04,  2.3196e-03,  1.4163e-03,
         4.4666e-03, -6.2928e-04, -5.0740e-04,  4.6486e-05,  5.1393e-04,
        -2.3310e-03,  4.2020e-04,  8.5723e-04,  1.6087e-03, -2.0849e-03,
         2.6493e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8983e-02, -9.1719e-01, -8.6408e-03, -7.8545e-04, -1.6037e-04,
         2.3043e-03, -3.1050e-03,  1.4401e-02,  6.4325e-03,  8.0438e-04,
         1.5326e-03, -1.3071e-03, -3.6139e-04,  5.1044e-03, -2.7329e-04,
        -1.1923e-03,  1.8730e-03, -2.4847e-03, -5.0238e-04, -5.3039e-03,
         1.4261e-03,  7.4259e-04,  5.0873e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3590e-03,  9.5818e-01, -2.4097e-03, -2.9177e-03,  1.0174e-03,
        -3.2979e-04, -6.3488e-03, -2.2890e-03,  8.2974e-04,  5.4135e-04,
         3.1298e-03, -2.1011e-03,  2.0507e-05,  9.5339e-04, -6.3202e-04,
         1.3621e-03, -9.0424e-04, -1.6330e-03,  1.5292e-04, -1.0621e-03,
        -2.6119e-03, -4.5812e-04,  9.8689e-04,  1.5188e-03,  2.5750e-04,
         8.5063e-04,  5.7154e-04,  1.5720e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 2.7803e-03, -9.6624e-01,  3.0939e-03,  3.1695e-03,  2.0349e-03,
        -5.5202e-04,  1.0260e-03, -1.2844e-03,  1.8570e-03,  2.8680e-03,
         1.4893e-04,  2.2821e-04, -1.6161e-04,  2.2304e-04,  1.1633e-03,
         4.8061e-04,  4.0698e-04,  1.6634e-03, -4.5143e-04,  3.1397e-03,
         5.4337e-04,  2.9866e-04,  1.7077e-03,  1.5065e-03, -7.2368e-05,
        -6.3856e-05, -6.2771e-04,  3.1615e-04, -1.8859e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1449e-03,  9.5364e-01,  3.2720e-03, -9.5406e-04,  4.7005e-03,
        -1.6375e-03,  1.5380e-03,  1.6300e-03,  1.5116e-03,  3.1204e-04,
        -1.0429e-03, -1.1094e-03,  4.6166e-04,  6.0970e-04,  1.5443e-04,
        -1.7878e-03,  3.7694e-03, -1.2888e-03,  6.3876e-03,  1.4807e-03,
        -2.2651e-04, -4.2716e-04,  1.0444e-04,  8.8078e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9355e-03,  9.4209e-01,  2.9316e-03,  1.5106e-04,  3.4619e-03,
         7.1657e-03, -2.5966e-03,  1.3324e-03,  2.0820e-03, -5.4126e-04,
         4.0557e-05,  4.3058e-03,  1.1258e-03,  2.8787e-03, -9.9700e-04,
        -3.8799e-04,  1.4963e-03, -1.6807e-03,  1.8989e-03,  8.2313e-04,
        -4.9781e-03, -1.4890e-03, -8.9162e-06,  1.4763e-03,  1.1708e-04,
        -1.1348e-03, -2.0305e-03, -3.3631e-05, -1.6595e-06,  1.8332e-03,
         1.2436e-03,  4.5461e-04, -1.0788e-03,  1.9524e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3553e-03,  9.7246e-01, -7.7332e-03, -2.6855e-03,  2.8013e-04,
        -2.8379e-04,  3.9638e-03, -3.8705e-04, -1.5852e-03,  3.5239e-04,
        -2.2355e-03,  3.0624e-04,  3.1561e-04, -2.3195e-04, -4.4531e-04,
         4.9615e-04, -1.7549e-04,  1.2193e-03, -1.4082e-03, -4.9741e-04,
        -6.8178e-04, -8.9643e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0475e-02,  9.3959e-01,  5.2472e-03,  4.5251e-03,  2.4663e-03,
        -2.1239e-03,  2.0202e-03, -6.9064e-03,  2.6094e-03,  2.6839e-03,
         2.8067e-03,  6.4720e-05, -5.2461e-05, -1.8984e-03,  1.4713e-03,
         1.3334e-03,  2.5272e-03, -1.2009e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.4662e-03,  9.4636e-01,  7.9240e-03, -3.0366e-03, -2.0003e-03,
        -4.2690e-03,  1.6701e-03,  1.5770e-03,  1.0508e-03,  9.6393e-04,
        -9.0069e-04, -1.1135e-03,  2.5840e-03,  5.1234e-03, -1.5906e-03,
         1.7041e-04, -6.2148e-04, -3.8897e-03, -8.5637e-04,  7.9694e-05,
        -3.3789e-04, -2.9927e-03, -4.4199e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.2388e-03,  9.5725e-01,  4.1116e-04,  7.2504e-03,  6.5045e-03,
         1.2616e-03,  2.4361e-03,  1.5735e-03,  1.1785e-03, -1.3689e-04,
         1.8951e-03,  1.7345e-03,  1.8801e-03,  7.2364e-04,  2.2859e-03,
         2.7887e-03, -2.4529e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0238e-02, -9.4488e-01,  8.3879e-03,  1.3875e-03, -3.0040e-03,
        -3.9330e-03, -3.3819e-03, -7.7191e-05,  4.2082e-03, -1.9917e-03,
        -1.0003e-03,  5.0062e-03, -1.5178e-03, -2.3011e-03, -1.2108e-03,
         1.5525e-03, -1.0665e-03,  4.8563e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1765e-02,  9.3164e-01,  2.8510e-03, -1.0781e-03,  3.0256e-03,
        -9.2705e-04, -1.7059e-04, -1.6538e-04, -9.2687e-04,  2.6138e-02,
        -9.3208e-04, -3.1103e-03, -8.3673e-04, -1.7672e-03,  1.4296e-03,
        -1.4754e-04, -2.7919e-03,  2.4180e-03,  1.7843e-03,  1.5917e-04,
        -4.6713e-04, -7.1914e-04,  1.3365e-03,  1.7089e-03, -1.6369e-04,
        -1.5406e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.9736e-03,  9.2854e-01, -3.5413e-03,  1.4216e-03, -2.8558e-03,
        -2.7711e-03, -8.6872e-04, -2.0553e-03,  7.7758e-04, -1.9048e-03,
        -5.9748e-04,  1.1735e-03,  3.3083e-03, -4.2795e-03, -2.2670e-03,
         2.4035e-03, -9.7981e-04,  2.1829e-04,  1.0045e-03, -2.3928e-03,
        -8.4758e-04, -4.0999e-03, -1.4180e-03,  1.4733e-03, -3.2846e-03,
         1.6255e-03,  3.0235e-04,  1.1947e-03, -3.0109e-03,  4.0291e-03,
         5.3805e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8601e-02,  9.5894e-01,  1.7348e-03,  7.1323e-04, -1.5154e-03,
         3.9732e-03,  5.1075e-04,  5.4975e-04, -3.3168e-04,  8.3609e-04,
         2.3421e-03, -1.3715e-03, -2.3107e-03,  1.0707e-03,  3.3388e-03,
         2.4663e-04, -1.1739e-03, -4.4037e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0282,  0.9238, -0.0063, -0.0119, -0.0046, -0.0063, -0.0079, -0.0110,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([-7.5562e-03,  9.5074e-01, -4.8326e-03, -2.6536e-03,  1.7896e-03,
        -9.9570e-04,  1.1059e-03, -3.8182e-04, -1.0046e-03,  1.5949e-03,
        -2.8393e-04,  3.3988e-04, -1.2973e-03, -6.7353e-04,  2.0612e-03,
        -3.1798e-04, -3.8843e-04, -2.3597e-03,  1.1445e-03,  2.2348e-04,
         7.4525e-05, -4.9974e-04, -3.1674e-03,  7.0521e-04,  1.7318e-03,
         5.5396e-05,  2.1809e-03,  3.6850e-03,  1.7918e-05,  7.0457e-04,
         5.0647e-04, -7.6313e-04,  7.4951e-04, -5.5035e-04, -1.0237e-03,
        -4.8171e-05, -3.2918e-04, -4.7737e-04, -9.8256e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.6652e-03,  9.3951e-01, -1.0326e-02, -3.4401e-03, -1.1661e-03,
         7.2425e-04, -1.3850e-03,  2.3868e-03, -5.7539e-04, -1.2457e-03,
        -1.5222e-03,  1.1160e-03,  3.2559e-04, -1.0989e-03,  1.4965e-03,
        -1.9320e-05,  7.0582e-04,  1.2303e-03,  1.4506e-03,  7.5661e-04,
        -2.1955e-03, -6.6100e-04, -6.7726e-04,  2.6524e-04, -1.7881e-03,
         7.9227e-04, -1.4583e-02,  1.8880e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.9327e-03,  9.6255e-01, -3.8105e-03,  3.3337e-03,  1.7896e-03,
        -4.1980e-04,  1.0936e-03,  2.4217e-03,  5.4012e-03,  1.2955e-04,
         7.7654e-03,  3.6947e-03, -6.5704e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.3801e-03,  9.7174e-01, -3.4660e-04, -1.8021e-03, -1.7074e-03,
         9.4403e-04, -1.3994e-03,  4.6721e-04, -9.7758e-04, -1.0077e-03,
         4.2104e-03, -1.2464e-03, -1.3223e-03,  2.3688e-03,  4.0832e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3310e-02,  9.1214e-01, -3.2399e-03, -2.7262e-03,  8.1656e-03,
         1.7538e-03,  1.8624e-03, -1.4979e-03,  1.8792e-03,  3.6171e-03,
         2.4861e-04, -2.8006e-04,  3.3789e-03,  1.2544e-03,  2.3582e-03,
        -5.0872e-04, -1.0705e-03,  1.3568e-04, -5.2566e-04,  5.1254e-03,
         5.0904e-05,  1.7632e-04,  2.9999e-03, -2.9187e-04,  1.7194e-03,
         1.4215e-03, -3.6305e-05,  3.0318e-03, -1.1080e-03, -4.0868e-03,
        -2.6970e-03,  9.6848e-04,  1.4327e-03,  2.4410e-04,  2.1217e-07,
         9.4683e-04, -3.8388e-04, -4.4492e-04,  5.6144e-04, -1.0016e-04,
         1.4054e-03, -2.2694e-04, -6.0368e-04, -1.6089e-04, -1.8265e-03,
        -4.3816e-03,  3.6132e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7748e-02,  9.2026e-01,  7.9898e-03, -1.0871e-03,  5.2693e-04,
        -3.5459e-03, -1.0250e-04, -4.9295e-05, -4.6891e-03, -1.1332e-03,
        -1.1083e-03, -1.5726e-03, -4.7906e-04, -1.3277e-03,  1.7370e-04,
         2.1437e-03, -1.5764e-04, -6.9995e-04, -2.6685e-03, -5.1083e-04,
        -1.2796e-04,  7.9194e-04, -3.4126e-04, -2.2106e-03, -1.0466e-03,
        -1.7261e-03,  1.8330e-03, -2.6157e-04, -5.7183e-04,  1.8617e-03,
         6.3461e-05,  7.2736e-04,  1.3932e-03, -7.0711e-04, -9.2577e-04,
        -2.1189e-03,  3.8171e-04,  7.1094e-04, -9.3555e-04, -1.1098e-03,
        -9.4501e-03, -2.7328e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.4663e-04, -9.3133e-01, -2.9707e-03,  1.2255e-02, -4.4336e-03,
        -1.9881e-03, -4.8095e-03,  9.2676e-04, -3.1189e-04, -6.3612e-03,
        -6.2608e-04, -5.3837e-03,  1.3750e-03, -7.1848e-04, -4.9231e-03,
        -9.0115e-03, -2.9384e-03,  7.5818e-04,  6.5871e-04,  3.3555e-04,
        -5.9578e-04, -4.1818e-04,  2.3794e-03, -8.1139e-04,  2.8278e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3190e-02,  9.3830e-01, -4.0169e-03,  3.8572e-05,  3.3081e-04,
        -6.5141e-04,  3.2712e-03, -1.9061e-04,  3.0473e-03, -2.0788e-04,
         2.1513e-04,  7.2208e-04,  9.3099e-04, -8.3809e-03,  1.5606e-03,
        -1.0779e-03,  1.0041e-03,  7.2175e-04, -8.4099e-04,  2.4618e-03,
         2.5121e-03,  8.6279e-04, -5.4627e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.5987e-03,  9.5483e-01,  3.6853e-03,  4.4490e-04, -5.0701e-03,
        -4.7810e-03, -3.8784e-04, -5.4973e-04, -5.7888e-04,  3.6421e-04,
         1.1767e-03,  1.7296e-03,  1.1258e-03, -1.9814e-03, -2.1022e-04,
        -8.7032e-04, -1.0182e-03,  1.7418e-03,  1.5528e-04,  1.3986e-03,
        -1.0538e-03,  5.1864e-03,  2.0653e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3558e-02,  9.3185e-01,  1.6416e-02,  4.4655e-03, -4.0992e-03,
         5.8039e-04,  1.0011e-03, -9.4121e-04,  1.0137e-02,  4.9588e-04,
         3.1149e-03,  3.3848e-04,  3.0033e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0811e-02, -9.5955e-01,  5.6817e-03,  2.7625e-03, -1.5970e-04,
         7.9356e-04,  1.8026e-03,  2.3586e-04,  3.2315e-04, -2.9099e-03,
         2.5168e-03, -1.4859e-03,  2.6014e-03,  8.3667e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6296e-04,  9.7353e-01,  2.9001e-03, -2.6007e-03, -1.7997e-03,
         4.6734e-04, -3.7926e-03, -1.0480e-03,  1.2279e-03,  2.1523e-03,
        -1.3077e-03,  3.7348e-04, -3.4656e-03,  9.4786e-05, -4.7489e-04,
        -2.8588e-04, -3.6896e-03,  1.6820e-04, -4.5735e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-1.8905e-03, -9.4978e-01, -3.3636e-03,  2.3177e-03, -7.5800e-04,
         3.5918e-03, -4.5019e-04,  2.5312e-03,  1.7158e-03,  4.0556e-03,
        -1.7945e-04,  6.1774e-04,  1.5307e-03, -3.6474e-03,  2.4848e-04,
        -6.2541e-04,  1.6211e-03,  7.9025e-04, -6.0861e-04, -5.0519e-03,
         1.9337e-03,  1.4438e-03,  1.9697e-03, -7.0621e-04, -1.1608e-03,
         7.4115e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0113,  0.9283,  0.0122, -0.0055,  0.0083, -0.0047,  0.0029,  0.0031,
        -0.0028, -0.0028,  0.0045, -0.0023,  0.0030,  0.0017,  0.0020,  0.0046,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4954e-02,  9.3383e-01,  6.1552e-04,  2.0811e-02, -7.7699e-04,
         7.5607e-03,  2.8207e-03, -4.9730e-03,  3.5564e-03,  7.2783e-04,
         1.7625e-03, -8.5344e-04,  1.6738e-03,  1.4545e-03, -1.9206e-03,
         1.6234e-03, -8.3673e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1601e-02, -9.4597e-01,  6.2300e-03, -7.4636e-03, -4.6394e-03,
         1.1173e-02, -7.9203e-04,  5.7201e-03, -6.4139e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.2520e-03,  9.4168e-01,  1.4309e-03,  5.0261e-03, -2.6390e-03,
        -9.7611e-04,  9.4012e-04,  2.7987e-03, -1.2174e-03,  1.3517e-04,
         4.7007e-04, -2.1233e-03,  1.1621e-03, -6.2259e-04, -9.7954e-04,
         1.7424e-03,  7.5930e-04,  4.1451e-04, -4.6170e-04,  1.1459e-03,
        -1.9579e-04,  9.2775e-04,  1.1976e-03,  5.2419e-04, -7.1767e-04,
        -1.2813e-03, -6.1308e-04, -1.4272e-03,  2.8504e-03, -5.3720e-05,
         7.2694e-04, -5.7729e-06, -8.5471e-05,  1.0349e-03,  1.4804e-03,
         1.1718e-03, -4.8079e-04, -4.6400e-04, -6.0325e-05, -7.5716e-04,
        -1.2090e-03, -3.2525e-03, -2.2399e-03, -3.4830e-04,  7.3457e-05,
         1.8403e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.5314e-03,  9.3856e-01, -3.5105e-03, -2.6560e-03, -1.3093e-03,
         1.1160e-03,  4.3661e-04,  1.3435e-03, -3.6780e-04, -5.6089e-04,
        -2.1717e-04,  1.0896e-03, -1.8365e-04, -1.1115e-03,  7.9019e-05,
        -2.6823e-03,  9.9036e-04,  3.6329e-04,  2.6691e-04, -1.1523e-03,
         9.8747e-04, -3.3811e-03, -1.0091e-02,  9.6294e-05,  3.5248e-03,
        -9.2115e-04,  2.4533e-03,  1.1189e-03,  1.4945e-03,  8.3991e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2498e-03, -9.7321e-01, -3.2566e-03,  6.4860e-03, -8.4661e-04,
         9.0575e-03,  5.8922e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9428e-03,  9.8092e-01, -1.4843e-03,  3.0866e-03,  5.6945e-04,
        -1.0980e-03,  1.7266e-03,  1.3094e-03,  3.7933e-03,  1.1007e-03,
         1.9652e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0026,  0.9294,  0.0287, -0.0047,  0.0062,  0.0035,  0.0011,  0.0014,
         0.0033,  0.0023, -0.0023, -0.0020, -0.0113, -0.0011,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5381e-03,  9.5357e-01, -1.0433e-04,  4.3921e-03,  1.4025e-03,
         6.5949e-04, -7.8441e-04, -2.1727e-03,  1.6903e-03, -2.6189e-03,
        -6.8300e-04,  2.2263e-04, -5.7008e-04, -8.5728e-03,  1.5262e-03,
         2.2558e-03,  1.8596e-03, -3.4310e-04, -2.8506e-04,  8.9052e-04,
        -2.3712e-03,  1.8347e-05, -2.3843e-04, -9.9285e-04,  1.3409e-03,
         1.5961e-03, -9.8510e-04,  3.7757e-04, -6.2784e-05,  9.7360e-04,
         2.5511e-04,  1.8471e-03, -2.1152e-04,  2.5884e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.8643e-03,  9.1836e-01,  1.2105e-02,  2.3309e-02,  2.5673e-03,
         1.1978e-03, -1.2730e-03, -8.0736e-04,  4.5220e-04,  4.8995e-04,
        -1.8720e-03,  3.3684e-04,  4.4039e-04,  4.9030e-04,  1.3075e-02,
        -7.8229e-04, -4.5551e-03, -7.7936e-04, -2.1250e-03, -2.3692e-03,
        -2.6990e-03, -2.8511e-05, -6.0650e-04, -2.4289e-04, -5.4782e-04,
         3.1469e-04, -3.5764e-05,  2.6925e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2188e-02,  9.2514e-01,  2.5335e-03,  3.8381e-03, -2.4590e-03,
        -1.9128e-03,  4.2658e-05,  2.9089e-03,  3.8639e-03,  8.3372e-04,
         4.5617e-04, -4.9346e-03,  1.5381e-03,  7.1200e-03,  6.0141e-03,
         1.3676e-03,  7.5619e-03,  1.4475e-03,  6.8700e-04,  2.1761e-03,
         8.0068e-04,  3.0431e-03,  5.8761e-03,  1.2538e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 3.8493e-03, -9.4063e-01,  7.6074e-03, -6.1207e-03, -4.6343e-03,
         4.1364e-04, -3.7682e-03,  1.4887e-03, -4.5508e-03,  8.7572e-04,
        -8.1421e-05,  3.0529e-03, -9.1533e-04, -4.1075e-04,  3.2044e-04,
        -1.3089e-03, -1.4102e-03, -2.0839e-03, -4.0308e-04, -4.0619e-03,
        -1.0034e-03,  9.7996e-05, -1.0704e-03,  8.8644e-04,  2.6771e-04,
        -5.7878e-04,  1.6173e-03,  1.3458e-03,  2.3138e-04,  1.3598e-03,
        -3.5555e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4335e-03,  9.6693e-01,  2.6682e-03,  8.3674e-03, -1.1191e-03,
         5.7453e-03, -8.7642e-04, -1.7923e-03, -2.6373e-03,  9.9532e-04,
        -5.0206e-03,  1.4191e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3558e-03,  9.5769e-01,  6.3689e-04,  3.0427e-03,  2.2334e-03,
         4.4740e-03,  3.1341e-03,  1.8366e-03, -8.3444e-04,  5.7596e-03,
         1.0098e-02, -1.5033e-03, -3.3345e-03,  9.5651e-04,  3.1073e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.0093e-03, -9.5445e-01, -5.4918e-03,  2.7395e-04,  1.1033e-03,
         2.1952e-03,  1.0772e-03,  2.3832e-03,  4.1945e-04, -1.0748e-03,
        -4.4529e-04,  1.2631e-03,  6.4871e-04, -3.9767e-04, -1.9703e-03,
         1.0668e-03,  1.8424e-03, -1.2656e-03, -3.1597e-04,  1.6166e-03,
         4.6415e-04, -8.0306e-04,  4.0188e-03,  5.0650e-04,  8.0860e-04,
         5.3017e-04,  3.9069e-04,  6.0428e-04, -5.7822e-05,  4.9654e-04,
        -6.2139e-04,  8.0645e-04, -1.5807e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.1992e-03,  9.6621e-01,  9.1052e-04, -5.3639e-03,  1.2277e-03,
         2.5080e-03,  1.4584e-04, -2.6092e-03,  8.7020e-04, -3.7946e-04,
        -1.8832e-03, -2.9496e-04, -3.4753e-05, -4.3175e-03,  2.5927e-03,
        -3.4522e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.4236e-04,  9.5571e-01, -1.5990e-02,  4.6088e-03,  7.3272e-03,
         5.6807e-04, -1.0892e-02, -2.9972e-03, -1.1608e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([0.0099, 0.9560, 0.0026, 0.0099, 0.0217, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.9911e-03, -9.4880e-01,  9.1877e-03, -5.6196e-03, -1.4885e-03,
        -1.8727e-03, -2.8942e-03,  8.2387e-04,  4.8081e-03,  7.6394e-04,
         2.8359e-04,  2.0309e-03, -1.8876e-05, -2.5030e-04, -4.5608e-04,
         1.7210e-04,  1.2882e-04,  6.6539e-04, -3.2657e-04, -1.8223e-03,
        -2.9612e-03, -6.3107e-04,  7.5803e-04,  2.7280e-04, -2.2621e-04,
        -4.0259e-04,  2.3431e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0255, -0.9540, -0.0039, -0.0072, -0.0017, -0.0049, -0.0028,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.5130e-03,  9.5183e-01, -6.8616e-04,  6.0069e-04,  1.3275e-03,
         2.8255e-03,  1.8463e-03,  2.1081e-03, -5.2581e-04, -8.0951e-04,
         1.4473e-04, -2.9407e-03, -2.6770e-03,  6.5335e-05, -5.2159e-04,
         1.6760e-03,  1.4820e-03, -4.4145e-03, -5.7480e-05, -1.4606e-03,
        -2.8716e-03, -1.1617e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8578e-03,  9.6665e-01,  1.4183e-02,  3.6821e-03,  3.1201e-03,
         1.1784e-03, -1.1642e-05, -7.0243e-04, -1.5932e-04,  1.8149e-04,
        -2.6891e-04,  2.0627e-03,  1.0236e-03, -2.8860e-03,  1.2327e-04,
        -2.9307e-04,  8.1832e-04,  3.9707e-04,  4.0392e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.1805e-03,  9.4640e-01,  7.7513e-03, -3.2833e-03,  5.3592e-04,
         4.1165e-03,  3.2947e-03, -3.3561e-03,  6.5508e-04,  2.5182e-03,
        -1.7083e-03,  9.2419e-04, -5.0489e-04,  9.6751e-04,  5.2649e-04,
         1.8657e-04, -6.1445e-04, -1.5125e-03, -5.7145e-04,  4.4807e-04,
        -1.9755e-04,  8.2460e-05,  2.3135e-03, -7.4657e-04,  1.6607e-03,
        -2.0996e-04, -3.8256e-04,  3.9251e-04,  1.1189e-03,  2.1531e-03,
         2.6871e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-1.0809e-02,  9.4858e-01,  5.2002e-05,  9.7567e-04, -2.1176e-03,
         4.6980e-04,  8.7757e-04,  3.6746e-03,  1.4460e-03, -5.5267e-04,
         6.2339e-04,  8.0641e-04, -3.6043e-05,  3.5901e-03, -1.2515e-03,
         2.2464e-03,  2.1333e-04,  1.6257e-03, -1.6476e-03, -1.5268e-03,
        -1.9826e-04, -1.4728e-04,  1.9865e-03, -1.0900e-03, -9.5814e-04,
        -1.8652e-04, -1.7096e-03, -3.9013e-03,  6.8916e-04, -3.1856e-03,
         3.8614e-04, -2.4367e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.6132e-03, -9.4937e-01, -2.2384e-05,  3.3805e-03, -6.6642e-03,
        -3.7593e-03, -2.2234e-03, -2.7281e-03, -3.4906e-03, -9.8577e-04,
        -3.0783e-03, -7.9477e-04, -3.1535e-04,  5.3788e-04, -3.0411e-03,
        -1.1802e-03, -1.6119e-03, -4.9285e-04,  5.5168e-04, -4.5230e-04,
         6.9038e-04, -8.0156e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6255e-03,  9.5037e-01, -8.9619e-04, -3.3205e-03, -1.0024e-03,
         1.7895e-04,  4.8755e-04,  4.0237e-03,  8.2158e-05,  7.4920e-04,
        -3.3606e-03, -3.3692e-03, -8.9725e-04,  9.0975e-04, -1.1491e-03,
        -4.9096e-04, -8.2221e-04, -5.1176e-04, -2.2005e-03, -7.7636e-03,
         3.9465e-04, -1.6580e-03,  3.0824e-04,  2.2047e-03,  1.6162e-03,
         2.4030e-03,  5.2379e-04, -1.6384e-03, -6.6943e-04,  4.3678e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3746e-03,  9.6533e-01, -7.4146e-03,  2.5926e-03,  2.6306e-03,
        -2.7368e-04, -2.6893e-04, -7.1340e-04, -9.9669e-05,  9.6294e-07,
         2.8909e-03,  1.6881e-03,  3.0880e-03,  4.2654e-04, -3.9928e-03,
        -5.5668e-04,  4.0021e-04, -3.2601e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2833e-02,  9.4525e-01, -1.4734e-04,  2.4607e-05,  1.7415e-03,
        -4.7063e-03, -1.7115e-03, -3.9612e-04, -4.5907e-03, -1.0114e-03,
        -1.9048e-04, -1.1532e-03, -1.5486e-04,  8.0837e-04, -5.9281e-04,
        -6.4047e-04,  7.7191e-04, -6.0955e-04, -2.1725e-03, -5.4745e-04,
         2.3022e-04, -2.6574e-03,  2.5635e-03,  4.1975e-04, -2.0702e-03,
        -2.8915e-04, -1.0071e-03,  2.2764e-04, -7.6538e-04, -6.5675e-04,
        -3.0424e-05, -1.7089e-04,  1.6624e-03, -2.3635e-03, -5.8115e-04,
        -7.6521e-04,  3.4839e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.0576e-03,  9.4292e-01,  4.1138e-03,  1.0793e-02,  2.0109e-03,
        -1.4379e-03,  6.5523e-04, -7.7888e-03, -1.0636e-03,  2.7686e-04,
         1.5133e-03, -3.7076e-03, -2.9071e-03,  2.3063e-04,  2.3058e-03,
        -1.3892e-04, -2.7981e-03, -3.7093e-04,  1.5670e-04, -1.0108e-03,
         1.2508e-03,  1.2566e-03,  3.1118e-03, -2.1214e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0131e-03,  9.6334e-01, -4.3875e-03, -3.0470e-05, -1.0325e-03,
        -5.7455e-04, -2.3217e-03,  2.3624e-03, -7.9310e-04,  4.5279e-04,
        -3.8568e-04,  2.4856e-04,  1.3544e-04, -8.2138e-04, -1.6419e-03,
         7.8383e-04,  2.0350e-04, -5.5619e-04, -2.0644e-04, -5.4757e-04,
         8.4590e-04, -2.1005e-04, -6.7398e-04, -2.1093e-04, -8.6210e-04,
        -1.0357e-03, -1.2227e-03,  4.0478e-05, -1.1223e-03, -4.5421e-04,
        -1.2795e-03, -7.7078e-05,  2.5583e-04,  2.5933e-04, -4.2810e-04,
        -7.4475e-04, -4.4436e-04, -9.9352e-04, -2.8062e-04, -1.2849e-03,
         1.3480e-03, -9.6195e-05,  6.4484e-04, -1.4399e-04,  1.1601e-03,
         1.0398e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.7893e-04,  9.5224e-01, -1.0712e-02, -4.0421e-03, -8.8673e-03,
        -1.3931e-03, -5.8359e-04,  2.6094e-04, -3.5598e-04, -1.8607e-03,
         2.0896e-03,  4.9238e-04,  6.7165e-04, -7.4018e-04, -2.1190e-03,
         1.1238e-03, -5.2939e-03, -2.1118e-03,  2.4649e-04, -3.5153e-04,
         1.0067e-04,  2.4252e-03,  1.0359e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1648e-02,  9.5042e-01, -1.0330e-02, -1.7365e-03, -9.1752e-03,
         2.2069e-03,  5.2810e-03,  5.8176e-04,  4.7063e-03, -1.6742e-03,
        -2.2405e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1101e-02, -9.5373e-01, -3.4883e-03, -3.7265e-03, -3.1319e-03,
         2.1585e-03, -3.6450e-05, -9.5286e-03, -2.7127e-04,  2.3112e-04,
        -2.4912e-03, -1.6830e-03, -1.7216e-03,  6.7052e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2132e-02, -9.4431e-01,  2.0109e-03,  4.1346e-04,  4.9223e-04,
        -8.5481e-04,  1.0342e-02,  1.1775e-03,  2.7231e-03, -4.3622e-03,
        -8.5450e-03,  1.2636e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.2034e-03,  9.3144e-01,  4.9005e-03, -1.1134e-03, -1.8286e-03,
         1.2516e-03, -3.4921e-04, -1.5108e-03,  6.5061e-04,  1.1053e-03,
        -2.2648e-03,  2.5639e-03,  8.7335e-04,  1.4566e-03, -2.6365e-04,
        -8.2157e-04,  1.3262e-04, -3.4426e-04,  1.7788e-03, -4.8368e-04,
         2.4087e-04,  3.4288e-04,  2.7638e-03, -1.0847e-03,  1.6920e-03,
         2.8560e-03,  1.1756e-02, -1.6833e-03, -1.2111e-03, -1.1909e-04,
        -6.5194e-04,  2.4877e-04, -5.7774e-04, -4.7946e-04, -3.2363e-04,
        -2.8166e-04, -4.9814e-04, -7.6840e-05, -8.4432e-04,  7.3397e-04,
         1.0274e-04,  1.0592e-03, -1.8585e-03, -1.7906e-04,  2.4858e-04,
        -3.6408e-04,  1.5230e-03, -2.5035e-05, -1.3475e-05, -6.7166e-04,
         6.9858e-04,  2.8152e-04, -8.9957e-05, -5.5742e-04,  1.5279e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-9.3169e-03, -9.3848e-01, -8.9260e-03, -1.4753e-03,  2.9180e-03,
         6.7047e-04,  3.2982e-04,  5.2313e-04,  1.0134e-03,  1.0757e-03,
         1.0995e-02,  4.3540e-04, -6.5457e-04,  3.4199e-03,  2.0454e-03,
         1.6541e-03,  2.9866e-03, -9.0472e-04, -1.9215e-04,  2.0662e-03,
         1.9260e-03,  2.6359e-04, -7.7307e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3390e-03,  9.6907e-01,  3.1758e-03, -2.4583e-04,  1.3413e-03,
         1.2120e-03, -5.4553e-03,  1.2684e-03, -5.0263e-03,  3.6907e-03,
        -8.1936e-04, -1.8604e-04, -1.5399e-03,  1.4483e-03,  1.1897e-04,
        -1.9521e-03, -1.4778e-03,  6.2972e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2339e-02,  9.3559e-01,  9.3301e-04, -6.1596e-03,  2.9914e-03,
         1.7882e-03, -1.2301e-02,  7.5679e-04, -3.9152e-03,  1.4229e-03,
        -2.3531e-03,  4.8372e-05,  8.6742e-04,  1.0295e-03, -3.8240e-03,
        -5.8138e-03, -3.0174e-03, -4.8482e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.5275e-03,  9.5668e-01, -5.6195e-04,  1.1363e-03, -3.5844e-03,
         1.3754e-04,  2.4149e-03, -9.5952e-04,  1.9597e-03, -9.6917e-04,
        -2.8574e-03,  2.9011e-03,  3.9352e-03,  7.5950e-03,  4.7832e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8873e-02,  9.3943e-01, -9.3576e-05,  2.6057e-03,  1.6670e-03,
        -1.4886e-03,  4.6843e-04,  8.2239e-04,  1.5440e-03, -1.0206e-03,
         2.3937e-03, -2.2997e-03,  1.2696e-03,  9.0241e-04, -2.0297e-03,
         2.2707e-04,  1.0275e-04, -1.7143e-03, -7.3270e-04,  1.9236e-03,
         2.3898e-03,  5.5555e-04, -5.8106e-04,  5.1945e-05, -2.6685e-03,
         8.1709e-04,  3.2868e-03,  3.5376e-05,  8.9889e-04,  2.8424e-03,
        -2.3146e-03,  1.9449e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3095e-03,  9.5931e-01, -1.5853e-03, -4.6125e-03, -1.4223e-03,
        -1.1638e-03, -1.8273e-03, -2.5100e-04, -4.1746e-03, -7.8214e-04,
         5.9393e-04, -3.7268e-04, -2.5487e-03, -5.0502e-04, -3.1454e-04,
        -4.5375e-04, -9.8888e-04, -1.7604e-03,  4.9576e-04,  2.4808e-04,
        -4.7999e-04,  3.1547e-04, -2.4102e-03, -1.8299e-03,  4.1165e-04,
         2.3297e-04,  1.5408e-04, -3.2226e-03, -5.6915e-04, -3.8553e-04,
         1.1986e-03, -4.1550e-04, -6.5828e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8456e-03,  9.5669e-01,  4.1649e-03,  2.9130e-03,  1.1665e-03,
        -1.2609e-03,  3.3136e-04, -1.3190e-03,  1.7838e-03,  3.5732e-04,
         4.3202e-04, -3.0601e-04, -1.0215e-05, -1.1750e-03, -2.3354e-04,
         3.5159e-04,  8.1215e-04,  7.0780e-04,  1.8018e-04, -6.2679e-04,
        -1.5539e-04, -8.4012e-04, -4.1761e-04, -1.2277e-03, -5.5523e-04,
        -6.9327e-05,  7.7513e-05,  1.0224e-03,  7.1012e-04,  1.2753e-03,
         8.7878e-04,  1.9864e-04,  3.1747e-04, -5.2358e-04,  2.9580e-04,
        -9.6395e-04, -1.5516e-03,  6.7603e-04,  6.6011e-04, -7.4218e-04,
         7.2981e-05, -6.4410e-04,  6.8032e-04, -9.0610e-05,  1.2765e-03,
         3.9397e-04, -4.5470e-04,  2.2797e-05,  2.5743e-04,  3.1523e-04,
        -6.5215e-04,  7.7916e-05,  1.4524e-04,  4.0969e-04,  1.6341e-03,
         1.4360e-03, -5.7665e-05, -5.5120e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0466e-02, -9.1219e-01,  1.3562e-02,  7.4900e-03,  2.7469e-03,
         3.0874e-03,  9.0766e-04,  1.9401e-03,  4.4253e-03,  4.2846e-03,
         2.9152e-03,  3.5851e-03,  6.6461e-04, -4.5585e-04, -8.2639e-03,
         1.5704e-03,  2.4976e-03,  1.6675e-03,  1.4949e-03, -2.3728e-03,
        -1.9833e-04,  7.5484e-05,  4.6590e-03,  5.0467e-03,  8.4137e-04,
         2.5949e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3900e-03,  9.5268e-01, -4.3843e-03,  3.3000e-03, -3.7568e-03,
         1.7976e-03,  6.4202e-04, -1.1244e-03,  5.7886e-04,  7.5777e-05,
        -4.7073e-04,  1.5675e-03, -4.5736e-03,  2.3110e-03, -5.4703e-04,
        -2.1202e-03, -1.1600e-03,  8.7557e-04,  1.5344e-03,  3.3988e-03,
        -1.7034e-03, -1.0041e-03,  9.6608e-04,  1.7613e-03,  1.2760e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.5025e-03,  9.4151e-01,  6.3069e-03, -3.5845e-03, -2.8095e-03,
        -6.5874e-04, -1.1754e-03, -1.1327e-03,  2.2987e-05,  4.3814e-04,
        -1.1080e-04, -1.8912e-04, -9.0941e-04,  3.8727e-05, -2.0837e-03,
        -8.2413e-04, -1.1479e-03, -4.8740e-04, -4.2647e-04,  5.6385e-04,
        -4.1007e-04, -1.1160e-04,  1.0319e-03,  5.6587e-04,  1.1869e-03,
        -1.6917e-03,  3.7937e-03,  3.2055e-04, -4.4156e-04, -1.2763e-04,
        -2.5127e-03, -1.1708e-04, -1.7212e-03, -1.2439e-03, -6.7736e-04,
        -2.0116e-03, -7.0871e-04,  4.3512e-04, -1.2614e-03, -1.9983e-03,
         1.1825e-03, -3.9585e-04,  5.1742e-04,  1.0888e-03,  2.5273e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9329e-03,  9.4672e-01, -2.8307e-03,  2.2183e-03,  3.1916e-03,
         2.9998e-03,  3.3502e-03,  3.9806e-03, -2.9054e-03, -3.7659e-04,
         1.6780e-03,  2.7530e-03,  4.9280e-04,  3.6534e-03, -1.4844e-03,
         3.5363e-04,  1.0043e-03,  1.1385e-03,  5.8417e-04,  1.3047e-04,
         3.3485e-03,  8.8677e-04,  7.9878e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4596e-03,  9.5051e-01, -5.5879e-03, -4.6600e-03, -2.7597e-03,
         1.0347e-03, -8.6699e-03, -8.9153e-03, -7.5283e-04,  2.2534e-03,
        -2.5802e-03,  8.8152e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([-5.0930e-03,  9.5047e-01,  3.0602e-03, -5.4154e-04, -1.1606e-03,
         3.8940e-03,  6.6382e-03,  1.0756e-03,  6.9892e-04, -4.4559e-05,
         8.5368e-04,  1.4713e-03, -1.0149e-02,  1.7549e-03,  1.4429e-03,
         4.0114e-03,  2.0893e-03, -1.1663e-03,  1.0903e-03,  5.6971e-04,
         4.9690e-04,  2.0016e-03,  2.2292e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3817e-03,  9.5648e-01,  3.0569e-03,  1.4914e-03, -2.2783e-03,
         2.0255e-03, -3.0022e-03, -3.9095e-04, -5.1092e-04,  6.1706e-04,
         1.2969e-03, -7.7433e-04,  1.9640e-03,  2.6612e-04,  2.7527e-03,
         6.2569e-04, -5.5586e-04,  1.5928e-04, -7.0379e-04,  5.9309e-04,
         2.2251e-03, -6.7687e-04, -4.6960e-04, -4.3615e-04, -2.4324e-04,
         7.5996e-05, -1.3847e-03,  2.9642e-04, -1.7119e-03, -7.7109e-04,
         4.8192e-04,  1.1451e-03,  4.1561e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3053e-03,  9.7872e-01, -8.0507e-03, -5.6392e-04,  4.3383e-04,
        -1.0412e-03, -2.9141e-04,  7.9356e-04,  3.2324e-04, -1.2263e-03,
        -5.3152e-04,  1.3922e-05,  3.4880e-04,  1.3881e-04, -3.5176e-03,
        -6.3802e-04, -1.0887e-03, -7.2007e-04,  2.5456e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.5863e-03,  9.4914e-01, -4.8868e-03, -6.3478e-03, -1.0591e-03,
         7.6218e-03,  3.4472e-04, -1.9686e-03,  1.2241e-03,  9.8786e-04,
        -8.1785e-03, -8.7327e-04,  7.0776e-03,  4.6990e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5994e-03,  9.7837e-01,  1.2182e-03, -2.9371e-03, -6.7777e-04,
         1.3809e-03,  2.0413e-03, -2.4100e-04, -1.4778e-03,  4.1795e-05,
         4.8670e-04,  5.9246e-05,  7.3276e-05,  2.9861e-03, -4.0199e-04,
        -3.2536e-04,  4.1629e-05,  1.5110e-04, -1.8751e-06, -1.2465e-04,
         5.7482e-04,  3.7907e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7172e-03,  9.6659e-01,  3.6439e-03, -2.5075e-03,  6.8672e-04,
        -4.2474e-04,  3.8859e-04,  7.4877e-04, -7.1475e-04,  3.0489e-03,
        -2.4307e-03,  1.9238e-03,  9.5541e-04, -4.7426e-04, -8.8132e-04,
        -1.1500e-03, -4.3032e-04,  1.1120e-03,  1.3793e-03,  6.1495e-04,
        -4.3079e-04,  1.8710e-03,  4.4441e-04, -2.0165e-04, -1.3124e-03,
        -6.5495e-04,  3.3039e-05, -3.1764e-04, -6.3091e-04,  2.8513e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5245e-02,  9.5342e-01, -2.9229e-03, -1.6939e-03,  1.6791e-03,
         2.6294e-03, -3.5642e-03, -4.5190e-03,  2.3939e-04,  7.4108e-04,
         3.8121e-03, -4.2625e-04,  1.2403e-03,  3.3444e-03,  1.4088e-03,
         1.1314e-03,  1.9806e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2081e-03,  9.7410e-01, -1.3811e-03,  6.6715e-04,  1.6130e-03,
         9.2659e-04,  1.0634e-03, -6.9507e-03,  7.8234e-03, -6.9952e-05,
         2.1991e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4041e-02, -9.4943e-01, -5.5176e-03,  4.7844e-03,  3.2768e-03,
        -1.8525e-03,  6.1872e-03,  7.5828e-04,  8.7213e-04, -3.8632e-04,
         2.5510e-05, -2.3466e-04,  3.6254e-03,  3.3699e-04,  1.1913e-03,
         4.7701e-03,  2.7137e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1923e-03,  9.6471e-01, -2.2046e-03,  1.6515e-03, -4.6733e-04,
        -4.0975e-04, -5.4884e-04, -1.9877e-03,  1.4661e-03, -1.6085e-03,
         7.6852e-04,  2.0701e-03,  7.6277e-04,  3.9388e-03,  1.6577e-03,
         1.3325e-03,  8.4452e-04,  9.1911e-04, -4.7457e-04,  1.8831e-03,
         1.1013e-04, -1.6080e-04,  2.4300e-03,  4.1264e-04, -1.9813e-04,
        -2.7853e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0975e-02,  9.5555e-01, -5.2064e-04,  8.6468e-04,  1.9095e-03,
        -4.8360e-04, -2.3059e-06,  7.9794e-04,  3.3836e-04,  4.6157e-04,
        -2.7583e-04, -2.9983e-03, -9.9582e-04,  2.5322e-03,  8.0697e-04,
         3.6893e-04, -1.5979e-03, -7.6068e-04,  5.8107e-04,  1.4163e-03,
        -2.9369e-03, -9.9722e-04, -9.2093e-04, -2.4452e-03, -4.7086e-03,
        -1.3257e-03,  2.3257e-04, -1.3067e-03, -4.7788e-04,  4.0656e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.4859e-03,  9.6648e-01, -1.1355e-03, -2.5823e-03,  4.1912e-03,
        -1.9255e-03,  1.9159e-03, -3.9762e-04,  1.7082e-03,  1.3384e-03,
         6.3471e-04, -2.0472e-04, -1.1442e-03, -1.7718e-03, -2.0297e-03,
        -7.8137e-04, -9.4239e-04, -1.8922e-03,  2.4425e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 1.1740e-03,  9.3963e-01,  6.2945e-03,  9.7867e-03,  4.6721e-04,
         3.1677e-04,  7.6530e-03, -8.3037e-04, -2.9301e-03, -1.5592e-03,
        -9.2270e-05,  5.7349e-04, -1.1738e-04, -1.1512e-03, -1.3000e-03,
        -4.8957e-03,  8.7717e-04,  2.7019e-03, -1.0658e-03,  3.6709e-03,
         1.5254e-03, -1.9694e-04, -1.9828e-04, -1.0901e-03,  4.7048e-04,
        -6.5851e-03, -2.8419e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6270e-04,  9.3806e-01, -8.5756e-03,  1.1765e-02,  2.5920e-03,
        -4.9654e-03, -3.5731e-03,  1.1547e-03,  1.0521e-03,  2.2683e-03,
         2.9390e-03,  8.4729e-03,  3.2285e-03, -1.0787e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5369e-03,  9.5382e-01,  2.7924e-03, -2.9073e-03, -2.1796e-03,
        -8.3575e-03,  4.7949e-04,  5.3404e-04,  5.4739e-04, -1.2171e-04,
         7.9422e-04,  6.3847e-04, -5.8319e-04, -1.6580e-03, -1.8032e-03,
        -1.6473e-03, -2.6685e-04, -3.9296e-03, -2.4811e-03,  1.6705e-03,
         5.8147e-04, -1.9338e-03, -1.9367e-03, -1.6851e-03, -1.2126e-03,
        -2.8968e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0420e-02,  9.5845e-01,  3.0197e-04, -4.3886e-03, -2.2549e-03,
        -9.1299e-03, -1.8305e-03, -2.0503e-03, -3.1200e-03, -3.5489e-03,
         9.5395e-04, -8.0824e-04, -1.4026e-03,  1.3404e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7040e-03,  9.5628e-01, -1.3135e-03,  2.1578e-03,  1.8270e-03,
        -3.4569e-04,  5.0929e-04,  3.6277e-03,  9.9368e-04,  3.4423e-04,
         1.2504e-03, -7.8943e-04,  1.7916e-03, -2.6393e-03,  2.5438e-03,
         5.1724e-05, -2.5339e-04, -1.0991e-03,  2.2144e-03, -1.6498e-03,
        -7.9621e-04,  1.7045e-03, -4.8567e-04,  2.3900e-03,  2.1356e-03,
        -5.7236e-04, -3.3003e-03, -1.1931e-04, -1.1271e-03, -5.7231e-04,
        -7.8348e-04,  1.6227e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6146e-03,  9.6016e-01,  6.6171e-03, -1.6121e-03,  7.8557e-04,
         7.6037e-04, -5.8161e-03, -7.2267e-04,  8.0320e-04,  3.2022e-04,
         3.6530e-03, -6.0473e-04,  1.6365e-03, -4.7363e-04, -6.7429e-04,
         8.2702e-04, -6.7982e-04,  2.2080e-03, -2.2859e-03,  1.8699e-03,
        -9.4445e-04,  1.7129e-03,  1.7797e-04,  5.5499e-04, -2.4829e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0025,  0.9526,  0.0286, -0.0034, -0.0011,  0.0118,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.4604e-03,  9.4619e-01, -4.8985e-03, -2.7701e-03,  2.3589e-03,
        -2.7270e-03,  2.8225e-03, -1.3253e-03, -5.4393e-04,  3.6535e-04,
         8.8555e-04,  3.1084e-04,  8.0826e-04,  1.2764e-03,  1.5675e-03,
         3.5526e-04,  6.5902e-04,  5.3974e-04, -9.4403e-04,  1.8982e-04,
        -1.8831e-03,  3.8840e-04,  2.0407e-03,  1.4748e-03, -6.7788e-04,
         2.9885e-03, -1.4702e-03,  1.9007e-03, -4.1955e-04,  5.5677e-04,
         1.4824e-03, -1.4159e-03,  4.9332e-04, -9.0235e-05,  6.6129e-04,
         1.2544e-04, -9.3560e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.7693e-03, -9.4220e-01,  6.0009e-03, -6.8866e-03,  8.8943e-04,
        -7.0567e-04,  1.6485e-03, -7.7849e-04,  2.0144e-03,  7.4325e-05,
         1.2024e-03,  8.2443e-05, -5.7212e-04,  1.0775e-03,  1.3643e-03,
         6.3277e-04,  8.2988e-04, -2.3714e-03,  7.5784e-04, -2.0456e-03,
         7.6956e-03,  7.2983e-04,  1.3201e-03,  1.1534e-03,  5.4069e-04,
         3.9650e-03,  7.9562e-04,  6.9663e-04,  1.8956e-03, -1.4848e-05,
         1.0987e-03, -2.0051e-04, -4.5617e-04,  3.7583e-04,  1.6139e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4542e-02,  9.1023e-01,  2.1384e-02,  1.5322e-03,  1.5172e-03,
         6.6933e-04,  4.3147e-04, -3.1207e-03,  8.3888e-04,  5.4806e-03,
        -2.4450e-03, -7.1857e-03,  7.3147e-03,  5.4763e-03,  7.8297e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4481e-03,  9.6353e-01,  9.9888e-03,  6.5084e-04,  2.1966e-03,
        -1.6641e-03, -1.9164e-04, -2.6766e-04, -1.4393e-03, -2.3081e-04,
        -4.6520e-04,  1.6261e-04, -1.8460e-04, -1.0390e-03,  2.5418e-04,
         7.7284e-04, -3.9557e-04, -8.8350e-04,  1.8528e-03, -2.9120e-04,
        -2.9173e-05,  1.1621e-04,  2.6172e-04,  2.9690e-05, -1.9676e-03,
         1.0791e-04,  1.1231e-03,  2.4556e-04,  2.4139e-04,  5.4418e-04,
        -5.4409e-05, -7.0988e-04,  3.6553e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.0119e-03,  9.3813e-01,  5.3808e-03,  5.6501e-05, -1.1465e-03,
        -3.9804e-04, -1.3858e-03, -8.3076e-04, -2.0172e-03, -4.4879e-04,
        -5.2561e-03, -5.6558e-03, -1.0105e-03, -5.4170e-03, -3.0744e-03,
        -5.1369e-04, -1.4548e-03,  1.8062e-04,  1.2570e-03,  1.3567e-03,
         8.7776e-04, -1.2535e-03, -5.1687e-04, -1.7313e-04,  1.0779e-03,
         2.3117e-03, -5.6337e-03,  2.4899e-03, -6.6838e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 5.1544e-03,  9.7558e-01,  1.8371e-05,  3.9921e-03,  2.1738e-04,
         6.2953e-04,  1.2509e-03,  3.5240e-03, -1.4849e-03,  3.2245e-04,
         8.6174e-04,  1.9573e-03,  1.8683e-03, -2.4109e-03, -7.3017e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.8337e-04,  9.5791e-01, -1.3896e-03,  5.8599e-04, -3.3471e-04,
        -2.6263e-03,  3.6929e-04, -2.4026e-03,  7.3690e-04, -5.8859e-04,
        -1.6881e-03, -1.1005e-03,  1.4933e-05,  9.6769e-04, -1.5461e-04,
        -2.5518e-03, -5.3781e-04,  1.1003e-03, -4.3332e-04, -3.6883e-04,
         4.4773e-04, -1.4203e-03, -9.0009e-04,  4.5735e-04,  3.3965e-04,
         4.1528e-03, -7.0403e-04, -3.9236e-04, -1.3054e-03,  5.0371e-04,
         5.2455e-04, -8.9630e-04,  2.2532e-04,  5.4085e-04,  6.9414e-04,
         1.0289e-03,  2.8844e-04,  9.4486e-04, -3.8751e-04,  2.2291e-04,
         2.1054e-04, -2.6375e-05, -1.9112e-03, -3.6963e-04, -5.1723e-04,
         4.7971e-04, -4.5181e-04, -5.5016e-04, -2.7636e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1965e-02, -9.3864e-01,  1.4314e-04,  3.4934e-03,  2.6707e-03,
         1.2148e-03,  3.8205e-03,  2.3389e-03,  3.1123e-03, -1.7786e-03,
         1.0114e-03, -2.1073e-04,  1.9875e-03,  9.2643e-04,  7.8188e-04,
         1.1484e-03, -3.2756e-03,  1.6611e-04,  3.2913e-04,  2.4611e-03,
         1.0294e-03,  2.9590e-04, -2.9308e-04,  3.0333e-04,  9.6268e-04,
         1.9848e-03,  2.7567e-03, -9.0088e-04, -1.8206e-03,  2.2803e-03,
        -2.9577e-04, -1.1515e-03,  3.3930e-04,  1.4081e-03, -2.6982e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1820e-02,  9.6375e-01, -1.3246e-03,  8.9270e-04, -1.3294e-03,
         1.1271e-03, -2.5124e-04,  1.2763e-03,  1.4439e-05, -3.3549e-03,
        -4.0893e-03, -1.6359e-03,  9.1996e-04, -1.8751e-03,  2.1525e-03,
        -1.1748e-03,  1.1286e-03, -8.4122e-04,  1.0384e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.5357e-03,  9.6802e-01,  2.2117e-03, -3.6653e-04, -1.1556e-03,
         6.2843e-04, -3.0192e-03, -5.6692e-04,  7.6676e-04, -2.8021e-03,
         7.1422e-04,  2.0660e-03, -5.4558e-04, -6.1036e-03, -1.4986e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.2963e-03,  9.5822e-01, -5.7561e-03, -4.4261e-04,  5.2679e-03,
         6.4737e-04,  8.8157e-04,  8.2602e-04,  8.8129e-04,  3.7427e-04,
        -3.1640e-05, -7.9962e-04, -2.6284e-03, -1.9008e-03, -1.9699e-03,
        -1.6324e-03,  9.1185e-05,  3.8560e-04, -8.1311e-04,  2.3121e-03,
        -4.7052e-04,  4.3739e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2648e-02,  9.4840e-01, -3.7628e-03, -2.7913e-03, -1.6186e-03,
        -5.9191e-03, -2.4278e-03,  1.3775e-04, -3.1782e-03, -8.6776e-04,
        -3.8211e-04, -2.5083e-04,  1.2925e-03, -4.8052e-04, -2.6852e-04,
        -1.3205e-03, -2.1157e-03, -2.7939e-03,  1.9633e-04,  7.2976e-04,
        -1.5417e-03,  3.5432e-04, -2.1941e-04, -1.2952e-03, -1.2330e-03,
         1.1610e-04, -3.4677e-04,  4.0850e-04,  2.0150e-04,  2.6990e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1429e-03,  9.6374e-01, -1.2249e-02, -7.2124e-03, -4.1625e-04,
         3.0775e-03,  1.9522e-03, -5.2899e-03, -1.0292e-04, -1.6712e-03,
        -1.1502e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.0583e-03,  9.5412e-01, -3.3755e-03, -1.1576e-03, -1.7498e-03,
        -2.5157e-03, -1.3687e-03, -8.4923e-04,  2.0316e-03, -6.3030e-04,
        -4.3337e-04, -1.7705e-04, -8.6203e-05, -9.4069e-04,  5.3711e-05,
         3.0171e-03, -3.1094e-03,  9.7430e-04, -3.3920e-03, -9.0069e-04,
         3.1219e-03, -6.9410e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5173e-03,  9.3989e-01,  7.3723e-03, -6.9168e-04,  7.3849e-04,
        -8.4958e-03, -6.0669e-04,  1.3662e-03, -2.2785e-03, -1.5471e-03,
        -2.7236e-03,  1.4442e-03,  2.0933e-03, -1.7809e-03, -5.5753e-04,
        -2.1610e-03, -5.6678e-04, -3.1390e-03, -4.5418e-03, -6.1971e-04,
        -1.8126e-03, -1.3146e-04, -1.7163e-03,  4.8908e-04,  2.2948e-03,
        -6.6920e-04,  1.0184e-03, -1.1132e-04, -1.3500e-03, -1.0392e-04,
         1.6044e-03, -1.8788e-03,  1.6890e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.9989e-03,  9.3371e-01,  1.6457e-02,  1.3324e-03,  2.9308e-03,
        -3.1234e-03,  1.8702e-03, -1.9736e-03,  4.3609e-03, -4.6560e-04,
         1.6145e-03,  8.9986e-04,  3.3471e-03,  2.0640e-03, -8.7011e-04,
         4.0587e-03,  3.3618e-03,  2.5888e-04, -4.4376e-04, -7.8845e-04,
        -2.1316e-03, -2.6292e-03,  1.1217e-03,  1.3396e-03, -3.8457e-03,
         1.5216e-03,  2.2549e-04,  2.5724e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.9085e-03,  9.5892e-01, -5.6926e-03, -1.5964e-03, -2.1457e-03,
         1.6879e-04,  2.0575e-03,  4.7721e-04,  4.1379e-05,  2.5748e-03,
        -2.7532e-03,  3.0591e-03,  2.9576e-04,  1.1012e-04,  1.0139e-04,
        -5.0166e-04,  4.1243e-04,  6.5478e-04,  1.0507e-03, -1.1246e-03,
         9.4483e-05,  8.5562e-05,  3.0038e-03,  2.4635e-04,  2.9185e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 8.2922e-03,  9.3580e-01,  1.3750e-03,  2.1660e-03, -6.1254e-03,
         2.1278e-03, -1.1231e-03, -9.0113e-04,  2.1054e-03,  1.9368e-04,
         1.9878e-03,  2.3687e-03,  1.5474e-03,  8.0119e-04, -1.1085e-03,
        -4.5960e-03, -3.9119e-05,  2.1394e-03,  9.7359e-04, -5.1377e-04,
         1.7525e-03,  6.1228e-04, -9.3768e-04, -8.1981e-05,  1.9177e-04,
        -2.0401e-04,  6.3297e-04, -1.6176e-03, -3.4944e-04, -1.2469e-03,
        -1.1345e-03, -1.2219e-03, -1.5858e-03, -7.3673e-04, -5.9323e-04,
        -1.3460e-03, -1.9814e-03, -1.6710e-03,  1.0187e-03, -2.5706e-03,
        -2.2319e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5408e-02,  9.3617e-01,  6.9760e-03,  6.7001e-04,  3.2371e-03,
         1.0414e-03,  3.7746e-03, -2.4944e-04,  6.1394e-04, -1.9201e-04,
        -1.1475e-03, -3.3275e-03,  1.7553e-03, -5.4602e-04,  9.4447e-03,
        -3.1967e-03,  2.1111e-03,  1.1156e-03,  1.8858e-03,  4.0933e-04,
         9.6351e-04,  2.1083e-03, -2.7393e-04,  3.8742e-05, -3.3445e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.8028e-03,  9.6913e-01,  1.3435e-03,  8.9077e-04, -2.7082e-04,
        -1.9344e-03,  3.5378e-03,  1.4066e-04,  1.6495e-04, -9.9273e-05,
        -1.4665e-03,  7.0931e-04,  6.5517e-06, -5.8520e-04, -1.6158e-03,
        -3.1018e-04, -1.9449e-03,  2.5248e-03,  3.8358e-04,  7.8490e-04,
        -2.3492e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4958e-03,  9.4046e-01, -1.4519e-02, -6.0946e-03, -5.5931e-03,
        -1.7061e-03, -3.3465e-03, -2.8197e-03, -1.9732e-03, -5.5693e-03,
        -4.4737e-03, -3.9919e-03, -2.6257e-04, -1.6813e-03, -2.4617e-03,
         1.4678e-03,  1.2353e-03,  8.5316e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.9128e-03,  9.5211e-01, -3.4777e-03,  1.3224e-04,  1.9309e-03,
         1.8569e-04,  4.5755e-03, -2.0267e-03, -1.9435e-03,  4.0921e-04,
         5.0298e-04, -1.6966e-03,  1.6691e-04,  6.4499e-04,  2.4855e-04,
         1.3419e-03,  8.2867e-04, -1.9047e-03,  1.6372e-03, -1.6192e-03,
         7.4439e-04, -5.4843e-04,  3.2427e-04,  8.9357e-04,  1.5204e-04,
         5.8736e-04,  1.4236e-03,  4.2584e-04, -3.1472e-04,  8.0339e-05,
         2.9203e-04,  7.6137e-04,  2.9484e-05, -1.1670e-03,  1.6134e-03,
         1.8329e-03, -2.2705e-03, -4.2453e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2766e-03,  9.6108e-01, -1.3835e-02, -2.5249e-03,  7.0660e-04,
         1.1422e-03, -1.7442e-03, -2.1944e-04,  1.3823e-03,  1.1343e-03,
        -6.6967e-04, -1.7553e-03, -9.1633e-04, -2.2616e-03,  2.0084e-03,
        -6.0147e-04, -3.6940e-04, -6.3351e-04, -2.2248e-03, -1.5122e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0775e-04,  9.6770e-01, -3.2773e-03, -2.6976e-03,  6.6473e-04,
        -3.4292e-03, -1.6539e-03, -4.8908e-03, -2.5349e-05,  2.6819e-03,
         6.2343e-03,  1.7404e-03, -1.6764e-03,  2.9166e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5736e-02,  9.3871e-01,  5.5993e-03,  2.5915e-04, -6.7123e-04,
        -1.4808e-03, -6.1980e-04, -6.5952e-04, -5.7190e-05,  1.0951e-03,
        -6.0346e-05,  4.8023e-03, -4.3620e-05, -6.8438e-04,  3.4239e-03,
         5.0893e-04, -4.8601e-04, -1.4950e-03, -1.5281e-04,  1.8033e-04,
        -2.4895e-04,  2.9269e-03,  1.0898e-03,  2.1038e-03, -9.1478e-04,
        -1.2349e-03,  5.1051e-04,  4.9448e-04,  4.1804e-06, -3.3870e-03,
         8.0702e-06,  1.0356e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0031e-02, -9.6097e-01, -3.6351e-03,  2.1896e-03,  2.8056e-03,
         6.6808e-04,  1.1065e-04,  1.5199e-03,  1.3831e-04,  2.1618e-03,
         2.9323e-04, -2.9624e-03, -4.8139e-04, -1.4779e-03, -1.0203e-04,
        -1.1898e-03, -1.1072e-04,  2.9006e-03,  2.9014e-03,  3.3493e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0948e-04,  9.3675e-01,  9.7906e-03,  5.1724e-04, -1.3779e-03,
        -4.4499e-04, -4.6319e-04, -6.3751e-04, -1.6698e-03, -1.5518e-03,
        -1.8323e-03, -1.0216e-03, -2.1556e-04, -2.1167e-04, -8.2544e-05,
         1.5625e-05, -2.7402e-04, -5.5237e-03,  1.3252e-04, -4.1685e-04,
        -3.3854e-04, -5.9539e-04, -6.5594e-04,  1.1511e-03, -2.1606e-04,
        -2.1737e-03, -7.2436e-03, -6.6400e-04, -5.3109e-04,  1.7845e-04,
         1.5503e-04,  6.8154e-04, -7.5764e-04,  2.5348e-03,  1.5248e-03,
         3.8968e-03,  3.9472e-04,  1.5485e-04,  1.7985e-05, -1.2246e-03,
        -1.8585e-04, -1.2889e-03, -1.6380e-03, -1.9664e-03,  1.0921e-04,
        -3.8924e-03,  2.2859e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.7373e-03,  9.6026e-01, -3.4111e-03,  2.6551e-04,  7.2766e-03,
         6.9669e-04,  1.2494e-03, -2.7538e-03, -2.6114e-03,  1.0952e-03,
        -3.5148e-03, -3.3608e-04,  3.3022e-03,  2.4391e-03,  2.0472e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2129e-03,  9.2838e-01, -4.9427e-03,  2.3877e-04, -1.8872e-03,
        -1.3348e-03,  1.9227e-04,  7.5985e-04, -4.0694e-04, -2.8556e-04,
        -7.7872e-04,  3.1878e-04,  3.0098e-04,  2.1797e-03,  6.9736e-04,
        -5.2542e-04, -2.0746e-03,  1.2932e-03,  2.0216e-03,  1.8858e-04,
         5.0541e-04, -1.9846e-03,  1.0349e-03,  4.1192e-04,  4.6997e-04,
         6.1030e-04,  3.9410e-03,  7.6335e-03,  5.2946e-04,  1.4397e-04,
         3.6068e-03, -2.5688e-04,  1.2536e-03, -9.6154e-04,  1.5981e-03,
        -6.6363e-04,  7.9254e-05, -1.5781e-04, -4.5920e-03,  2.0661e-04,
         1.2387e-03, -4.5096e-04, -2.8110e-03, -1.0826e-03,  9.4981e-04,
         7.1463e-04, -1.0440e-03,  1.0043e-02], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-2.9083e-02,  9.1157e-01, -9.0837e-03,  1.5901e-04,  6.4703e-03,
         1.3553e-03, -4.4946e-03, -2.2918e-04,  1.0032e-03, -9.0167e-03,
         9.1654e-04, -1.7824e-03, -1.0585e-03, -1.0774e-04, -1.4755e-04,
         8.7509e-05, -3.1851e-03, -2.6189e-03,  1.5647e-03,  1.4388e-03,
         5.9322e-03, -9.5726e-05,  2.4717e-03,  3.1194e-03,  2.6221e-03,
         3.8633e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.7018e-03, -9.5793e-01,  2.3984e-03,  2.3557e-03, -2.2225e-03,
        -2.4343e-03, -1.5842e-03, -2.0294e-03, -1.3884e-03, -1.4207e-04,
         4.3558e-04, -5.2800e-04, -7.0602e-04, -2.6047e-03, -8.1119e-05,
        -7.9287e-04, -4.2255e-03,  1.5717e-03,  3.6004e-03,  1.5604e-03,
         6.7053e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6076e-02,  9.2291e-01, -2.3677e-03, -7.1664e-03, -2.4130e-03,
        -4.7176e-03, -6.4700e-04, -2.1781e-03, -1.5484e-03, -1.1518e-03,
         5.9257e-04, -2.3134e-04, -1.9624e-03, -1.7054e-03,  1.5428e-03,
        -3.1191e-04, -2.3246e-03, -8.6919e-04, -1.2027e-03, -1.2099e-03,
        -1.5892e-03, -7.8577e-04,  1.5797e-04, -1.1942e-03, -1.6016e-03,
        -1.0362e-03, -4.3218e-04, -4.2896e-04, -8.5150e-05, -4.6230e-04,
         2.2832e-04,  1.6487e-03,  4.7109e-04,  1.6299e-04, -3.7854e-03,
         2.7973e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6886e-02,  9.0230e-01,  1.1250e-02, -1.0209e-03, -3.8854e-03,
        -7.9259e-03,  9.0423e-04,  8.7288e-05, -6.0918e-03,  5.5681e-03,
         5.9188e-03, -2.6675e-04,  5.5628e-03, -1.3907e-03,  1.3074e-03,
         6.2113e-04, -3.3391e-04,  1.4546e-03,  5.0370e-03, -2.1823e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.6342e-03,  9.3241e-01,  3.8748e-03,  4.3783e-03,  1.8576e-03,
        -2.5040e-03, -1.2758e-03,  5.5364e-03,  2.9691e-03,  4.0094e-03,
         1.2589e-02,  2.2909e-03, -9.1916e-04,  2.6885e-03,  6.1339e-03,
        -5.2008e-04,  2.0877e-03, -8.1612e-04,  2.7971e-03,  9.8915e-04,
        -7.1489e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0080, -0.9590,  0.0132, -0.0012,  0.0047,  0.0028,  0.0010,  0.0015,
         0.0025, -0.0012,  0.0025,  0.0010,  0.0015,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1250e-02,  9.3782e-01,  3.6271e-04, -1.0404e-03, -4.4455e-05,
        -2.3712e-03, -5.7506e-03, -4.0382e-04, -2.2769e-03, -6.0928e-04,
        -2.9429e-03, -3.3742e-03, -2.8159e-03, -1.4131e-03,  3.1277e-04,
        -1.0466e-03, -7.4243e-04, -2.1191e-03, -1.2287e-03, -3.0669e-03,
        -2.0161e-03,  8.6848e-04, -4.3742e-03, -8.4451e-04, -2.0680e-04,
        -7.3425e-04, -1.5789e-03,  1.8864e-03, -3.2337e-04, -9.1845e-04,
        -3.5426e-04, -1.3272e-03, -8.7038e-04, -6.4567e-04, -1.1536e-03,
         7.9319e-04, -1.1250e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.1582e-03, -9.3697e-01, -1.6907e-02,  1.1249e-03, -2.9273e-03,
        -8.2566e-04, -1.0515e-03, -1.8607e-03, -1.2044e-03,  3.6874e-03,
        -2.1397e-04, -1.9064e-03, -2.4911e-03, -3.7117e-03, -1.4142e-03,
         2.0559e-03,  8.9377e-04, -1.1688e-03, -5.6152e-04,  7.3978e-03,
         4.4401e-04,  2.1306e-03, -8.3570e-04,  6.0798e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6544e-02,  8.9964e-01,  1.5601e-02,  6.4769e-03,  2.1498e-04,
         1.9863e-03, -7.7925e-04, -6.5061e-04, -9.2644e-04,  3.4675e-03,
        -3.7245e-03, -5.7358e-03,  2.6485e-03,  8.8899e-04,  3.5554e-03,
         4.4719e-04,  2.5258e-03,  2.2364e-03,  1.1574e-04,  5.5298e-03,
        -1.1888e-04, -3.9114e-03,  2.3127e-03, -2.6053e-04,  1.0873e-03,
         4.4552e-04, -2.0933e-03,  1.9040e-03,  3.9267e-04,  7.3223e-04,
        -2.9232e-03, -1.2591e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5327e-03,  9.3780e-01, -1.0829e-02, -3.4360e-03, -1.0707e-03,
        -9.7357e-03, -1.3812e-03,  7.9853e-04,  3.3024e-04, -5.4828e-03,
        -5.0896e-04, -1.3810e-03,  1.0411e-03,  5.5985e-04, -2.5030e-03,
        -2.2735e-03,  8.1649e-04, -9.4067e-04,  2.1204e-04,  7.6401e-04,
        -9.6923e-04, -2.3805e-03, -2.6671e-03, -1.2863e-03, -1.7132e-03,
         1.3391e-03,  9.7579e-04, -1.7071e-03, -3.7083e-04, -1.5536e-04,
        -3.0416e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.8738e-03,  9.2825e-01,  5.6794e-03,  1.0703e-04,  2.2807e-03,
         6.7097e-04,  4.4119e-03, -6.5584e-03, -2.8134e-03, -3.4574e-03,
         4.0481e-03,  2.9065e-03,  2.1505e-03, -6.0798e-03,  4.8256e-04,
        -2.2435e-03, -6.8673e-04,  4.0877e-03,  1.5952e-03,  5.7628e-03,
        -1.8939e-03, -5.2814e-03,  2.6806e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9151e-02, -9.3812e-01,  4.4421e-03,  1.7176e-03,  3.4236e-03,
        -1.4796e-03, -4.5293e-04,  4.4254e-04, -1.7031e-03, -1.8524e-03,
         7.0739e-05,  4.3439e-03, -8.5184e-04, -1.8788e-03, -2.1571e-03,
         2.5402e-05,  7.9792e-04,  3.2599e-03, -5.1918e-04,  4.4606e-05,
         1.3375e-03, -2.8061e-03, -1.7109e-03,  5.0960e-04, -1.1268e-03,
         5.5149e-04,  2.4194e-03, -2.8054e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 6.2568e-03,  9.6034e-01, -6.0245e-04, -4.3892e-04,  1.3894e-03,
         5.3879e-04, -6.6579e-04,  3.2128e-03,  7.4740e-04, -1.3047e-03,
         2.9732e-04, -5.1308e-05,  8.8287e-04, -1.0608e-03, -4.6868e-04,
         2.3717e-03,  7.5560e-04,  1.0538e-03,  1.5988e-03, -2.5492e-04,
         2.7540e-03, -1.7001e-04,  2.3501e-04,  6.7462e-04,  5.3710e-04,
        -1.1884e-05, -1.6598e-03,  9.5920e-03,  7.1899e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6464e-03,  9.5903e-01, -2.0296e-03,  1.4870e-04,  6.7066e-03,
         7.1627e-03, -1.2684e-03,  1.3008e-03,  1.5483e-03,  3.5122e-03,
        -1.6716e-03, -5.2370e-04,  1.4371e-03,  6.6120e-04, -1.0603e-03,
        -2.4396e-03, -1.6497e-03,  1.7696e-05,  1.9677e-03,  1.2209e-03,
         2.4308e-04,  5.5278e-04, -2.7821e-04,  9.2383e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7245e-03,  9.4209e-01, -1.4692e-02, -4.3642e-03,  4.9767e-03,
         3.7320e-03, -1.4406e-03, -1.9571e-03,  1.7167e-03,  2.3130e-04,
        -8.7212e-04,  8.3646e-04,  6.8535e-04, -2.8335e-05, -9.3271e-04,
         2.0745e-04,  6.0659e-04, -1.8836e-04,  5.3352e-05,  1.5171e-03,
         4.1254e-03, -8.7708e-05,  1.0900e-04,  2.7645e-03, -1.4560e-03,
        -5.1415e-04,  1.1759e-04,  4.7202e-04, -8.9028e-04,  1.8719e-03,
         1.3405e-03,  4.1012e-04, -1.7863e-03, -1.1992e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.9131e-03, -9.6991e-01,  2.3026e-03,  7.0567e-04,  1.9667e-04,
        -5.0854e-04,  4.0571e-04, -1.6206e-03, -1.1465e-03, -9.2605e-04,
        -2.0906e-03,  6.2197e-04, -4.6810e-04,  1.5552e-03,  1.7351e-04,
         2.9028e-04,  3.0681e-03, -2.7969e-03,  4.5942e-04, -9.0728e-06,
        -4.9058e-04, -3.4002e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4106e-02,  9.3806e-01, -8.8380e-03,  5.3748e-03, -1.0116e-03,
        -6.1835e-03,  1.5368e-03,  4.7633e-03,  1.1771e-03,  2.2522e-03,
         5.1976e-03, -1.0278e-03,  1.7417e-03,  1.5120e-03,  1.3081e-03,
         3.5331e-04,  1.7902e-03, -3.7616e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6992e-03,  9.6478e-01,  5.4120e-03,  3.3699e-04,  3.0656e-04,
        -2.0155e-03,  1.4713e-04,  2.2180e-03, -9.5213e-04, -1.6285e-03,
        -3.8629e-04, -4.2495e-03, -3.8945e-03,  5.4804e-04, -2.0188e-03,
        -9.4744e-04, -7.0610e-04, -1.0632e-03,  1.6074e-03, -2.4628e-05,
         2.9319e-04, -6.7282e-04,  1.0875e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0762e-02,  9.5098e-01, -6.1113e-03, -3.2969e-03, -4.9440e-03,
         7.7579e-04, -8.0081e-04, -7.1823e-04, -2.5200e-03, -1.1733e-04,
        -4.7186e-03,  2.6199e-04, -2.1630e-03, -4.6891e-04, -3.8030e-04,
         1.2832e-04, -8.4988e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0364e-02,  9.2568e-01,  8.5915e-03, -1.8233e-03,  1.8953e-03,
         2.6003e-03,  3.2162e-03,  6.8985e-04,  7.8947e-03, -5.4721e-03,
         6.1767e-03,  3.2073e-04,  2.0932e-03,  6.3039e-03,  2.4056e-03,
         3.5168e-04,  2.2683e-03, -1.8560e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5629e-03,  9.0233e-01, -1.6765e-02, -4.1847e-04, -5.3278e-04,
        -1.6521e-04, -1.1939e-03, -1.4074e-03, -1.7029e-03,  3.6495e-02,
         1.3544e-03, -1.2542e-03,  3.2865e-03, -2.2244e-04,  6.6557e-04,
         1.2982e-02,  1.4808e-03,  6.3059e-03, -2.7780e-04,  2.1365e-04,
        -6.5714e-04, -7.4660e-04, -1.1659e-03, -3.7819e-04,  2.9475e-03,
        -1.4898e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6290e-02,  9.1180e-01,  1.6475e-03, -8.7944e-03, -1.0225e-02,
        -3.1942e-03, -3.7009e-03, -3.5279e-03, -8.8876e-04, -5.7998e-03,
        -5.3514e-04,  4.0488e-04, -8.7905e-04, -7.9100e-03, -1.0433e-03,
        -2.5449e-04, -1.7595e-03, -1.0113e-03,  7.9135e-04, -1.2845e-03,
         2.7098e-04,  9.5624e-04,  1.3017e-03, -2.2061e-03, -2.5325e-03,
        -1.1204e-03, -1.5562e-03, -9.2318e-04, -2.5592e-03,  2.6067e-03,
         2.2266e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8584e-02,  9.4077e-01,  6.8413e-03, -4.3124e-03, -1.8801e-03,
        -1.2377e-03,  1.3879e-03,  3.2083e-03, -1.9615e-03, -6.9288e-03,
         2.1573e-05, -2.2798e-03,  2.4815e-03, -1.2992e-03, -2.2301e-03,
         1.6127e-03,  1.4584e-03,  1.5014e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0036,  0.9627,  0.0125,  0.0082,  0.0031, -0.0020,  0.0034, -0.0046,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 3.8135e-03,  9.4420e-01, -1.1824e-03, -2.1857e-03, -1.2967e-03,
        -8.8346e-04, -3.3836e-03, -2.3658e-03,  2.9656e-03, -1.7075e-03,
         1.7722e-03, -1.5828e-03, -1.3462e-03, -2.2108e-03, -2.1253e-04,
         1.6323e-03, -1.6239e-03, -2.1635e-03,  9.3464e-05, -1.2722e-03,
        -4.1423e-04, -9.5735e-04, -3.3766e-03, -1.1247e-04,  3.1882e-04,
        -1.0457e-03, -1.2118e-03,  2.6418e-03, -9.2629e-04, -5.5189e-05,
        -4.1724e-03, -1.3099e-03,  3.4309e-04,  3.5226e-04, -7.0626e-04,
        -1.1295e-03,  8.9517e-04, -1.5288e-03,  6.0476e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5305e-03,  9.4868e-01,  7.5139e-04, -4.9044e-03, -1.1157e-03,
        -7.6309e-05, -2.8843e-06, -1.0719e-03, -2.2074e-03, -2.7867e-03,
        -1.8606e-03, -4.0013e-05,  8.0481e-04, -8.3889e-04,  3.5148e-04,
         5.4408e-04, -2.7545e-03, -1.9169e-03, -3.9687e-04, -2.7077e-03,
         3.8925e-03, -6.0599e-03,  1.2036e-03, -6.9253e-04,  2.6462e-04,
         1.3812e-03, -1.2298e-03,  5.9284e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.7241e-03,  9.7461e-01,  2.2573e-03, -1.6694e-03, -5.9802e-04,
        -3.7902e-03, -2.8258e-03,  5.2733e-05,  4.1228e-03,  5.6610e-04,
        -2.3403e-04,  2.0081e-03, -2.5449e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.4436e-03,  9.7117e-01, -1.7272e-03,  1.7928e-04, -3.0754e-03,
         3.4368e-03,  4.6799e-04,  9.8928e-04, -7.3691e-04,  2.9665e-03,
         2.3578e-03,  2.0426e-04, -1.6222e-03,  6.4630e-04,  9.7208e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1716e-02,  9.3525e-01, -1.0595e-02,  3.0282e-03,  2.9796e-03,
        -6.4490e-04, -1.2011e-03, -6.6546e-04,  4.8546e-05, -2.0076e-03,
         1.4039e-03, -3.8709e-04,  2.8044e-03,  1.7578e-03, -1.3667e-04,
         3.7428e-04,  6.0222e-05,  4.8352e-04, -1.0869e-03, -1.8455e-03,
         1.9961e-03, -4.1867e-04, -1.5992e-03, -2.9635e-04, -1.0647e-03,
         2.4565e-04,  8.7809e-04,  1.8371e-03, -1.3778e-04,  1.5669e-04,
         1.2106e-03,  9.9623e-05, -1.4901e-03,  1.1521e-03,  2.9264e-04,
        -2.9102e-04, -6.6475e-04,  2.7706e-04,  3.2384e-05,  3.6155e-04,
         2.1832e-03,  2.2776e-04, -2.4713e-04, -2.7355e-04, -1.4826e-03,
         1.4196e-03, -1.1860e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4557e-03,  9.4935e-01, -8.6588e-03, -5.5392e-03, -1.1398e-04,
        -6.9273e-04, -5.8893e-05, -5.0267e-03, -4.0348e-03,  8.1609e-04,
        -3.7555e-03,  5.0789e-04, -6.9791e-04,  3.0899e-04, -7.2991e-04,
        -1.7244e-03,  1.6535e-03,  4.1930e-04,  1.1013e-03, -9.2500e-04,
         7.1854e-05,  1.8980e-04,  8.2254e-04,  8.3994e-05,  6.2955e-04,
        -1.0441e-04,  1.2740e-03, -1.6804e-03, -9.3548e-05, -1.4443e-03,
         1.3109e-03, -6.8344e-04,  1.1888e-03,  3.3381e-05,  7.4842e-05,
        -1.8374e-04, -3.4043e-04, -2.8105e-04,  1.5034e-04,  9.9138e-04,
         9.6846e-05,  7.0001e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6441e-03,  9.2687e-01,  5.4818e-03, -3.0026e-03,  6.9117e-03,
        -2.1230e-03,  7.4141e-03, -1.5111e-03, -3.1109e-03,  2.5278e-03,
        -1.0581e-03, -4.2419e-03,  6.4596e-04,  1.2005e-03,  3.3564e-03,
        -9.2301e-03,  3.1636e-03,  2.1731e-04,  4.0960e-04,  8.5956e-04,
        -1.1174e-03,  6.5007e-04, -3.5740e-03,  6.2471e-04, -6.0567e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2753e-03,  9.5740e-01,  8.6397e-03, -3.1848e-03, -1.1377e-03,
         1.3840e-05,  3.0406e-03, -6.5987e-05,  7.6725e-04,  2.4153e-04,
         1.5497e-03, -1.5368e-03,  1.3367e-03, -4.0179e-03, -1.3963e-04,
         3.4053e-04, -5.5361e-04,  4.6536e-04, -1.2293e-03,  2.3199e-03,
         1.5806e-03,  2.6339e-03, -5.5311e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4287e-02,  9.3752e-01,  1.5167e-02,  1.9482e-03, -1.0891e-03,
        -2.3784e-03, -2.9151e-03, -1.5528e-03, -4.8880e-04, -4.8718e-03,
         1.2059e-03, -1.4219e-03,  1.4869e-03, -3.6654e-04, -3.3868e-04,
         1.9428e-04,  2.8115e-03,  3.8997e-04, -2.6629e-03,  2.5207e-03,
        -2.4697e-03,  1.5299e-03, -3.8042e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6143e-02,  9.4945e-01,  7.6036e-03,  4.5714e-03, -6.3622e-04,
        -2.4063e-04, -2.4204e-03,  2.0938e-04,  5.8873e-03, -2.0005e-03,
         3.6816e-03, -5.4372e-03, -1.7168e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4763e-02,  9.5964e-01,  5.3948e-03, -1.0593e-03,  2.7841e-04,
         2.6036e-03, -8.1939e-04,  2.5472e-03,  1.6797e-03,  4.2520e-03,
        -1.4846e-03, -1.8360e-03, -6.8726e-05, -3.5686e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1222e-03,  9.6067e-01,  1.6587e-03, -3.7149e-03, -3.0129e-03,
         1.8013e-04, -2.3690e-03, -1.1604e-03,  2.2903e-03,  2.3906e-03,
         2.1765e-04,  3.9468e-04, -3.8731e-03,  1.1186e-03,  8.5015e-04,
        -6.2194e-04, -3.9910e-03,  2.5579e-03,  5.8036e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-6.7160e-03,  9.5635e-01, -1.7391e-03, -3.6669e-04, -1.4859e-03,
        -6.0208e-04,  1.9494e-03, -5.9139e-04, -7.9178e-04,  1.8582e-03,
         1.3978e-03, -1.8121e-03, -5.9659e-04,  2.7109e-03, -2.4274e-03,
         2.0489e-03,  1.6287e-03,  2.9245e-04,  3.2827e-04,  3.7488e-03,
         8.9904e-04,  1.6942e-03, -4.1543e-03,  1.1121e-03,  2.5587e-04,
         2.4459e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4695e-02,  9.1669e-01, -2.1063e-03, -1.0704e-02,  4.0450e-03,
        -8.3269e-03, -4.1311e-03, -1.1393e-02, -2.5832e-03,  6.7311e-03,
         2.0970e-03, -3.4821e-03,  1.3414e-03,  7.5741e-04,  4.6261e-03,
         6.2866e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.3837e-03,  9.5379e-01, -2.9184e-03,  3.4256e-03, -6.2520e-04,
         3.6222e-03, -3.2185e-03, -4.0747e-04, -1.9172e-04,  1.2729e-03,
         1.2650e-04, -5.7120e-03, -1.0169e-03,  1.8836e-04, -2.0607e-03,
        -6.6530e-03,  8.3919e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0094,  0.9523, -0.0054, -0.0025,  0.0048,  0.0037,  0.0067,  0.0051,
         0.0102,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7817e-03,  9.3487e-01,  3.3502e-03,  2.6601e-03, -1.7224e-03,
         1.6761e-03,  4.8714e-05,  6.2995e-04,  1.1158e-03, -1.3931e-03,
         2.2879e-03,  5.4291e-03,  1.2972e-03, -1.0013e-03, -7.7209e-04,
         8.7757e-04,  1.2106e-03,  1.9742e-03,  1.3037e-03,  2.9052e-04,
         3.9901e-04,  5.2639e-03, -3.9648e-04,  5.6131e-05,  1.7428e-04,
        -5.4403e-03, -7.5926e-04, -1.3944e-03, -4.5551e-04,  6.1880e-04,
        -9.2365e-04, -2.3223e-04,  2.8918e-04,  2.4688e-05, -5.6947e-05,
        -4.8781e-04, -1.9656e-03, -3.3456e-04, -1.4589e-03, -1.6472e-03,
         1.1560e-03,  1.4598e-03, -2.6530e-03,  1.1647e-03, -9.8741e-04,
        -3.5078e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1388e-02,  9.5067e-01,  2.0253e-03,  7.5079e-04,  5.1077e-04,
         1.0503e-03,  2.4038e-03, -1.5065e-04, -5.3001e-04,  1.0244e-03,
         9.9733e-04, -1.0419e-03,  6.4266e-03, -7.3332e-04,  2.8820e-04,
         2.2829e-03,  7.9650e-04,  1.6432e-03, -4.3343e-04,  3.7724e-03,
         1.1211e-03, -4.7589e-04,  1.9347e-04, -1.7976e-03,  1.0618e-04,
        -1.6384e-03, -8.6634e-04,  2.1735e-03,  1.5720e-03,  1.1323e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0057,  0.9468,  0.0023, -0.0132, -0.0105, -0.0106, -0.0109,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.7959e-03,  9.6683e-01,  1.4301e-03, -5.2044e-03, -2.4331e-03,
        -1.4181e-03, -6.0052e-04,  4.7529e-03,  2.7899e-03, -8.1523e-04,
        -3.9274e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.2552e-04,  9.2451e-01,  4.8658e-03,  2.1816e-02,  6.7800e-03,
        -8.4086e-04,  2.3153e-03,  3.7550e-03,  1.1512e-02, -2.2767e-03,
        -3.5500e-03,  8.6273e-04,  1.6491e-03, -1.4743e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5807e-03,  9.4661e-01, -1.3648e-03, -1.8817e-03, -1.3274e-03,
        -5.6690e-04, -7.5672e-04, -1.6735e-03, -5.8035e-05, -3.0617e-05,
        -1.0576e-03, -9.1024e-04, -5.9421e-04, -1.2066e-02, -2.1215e-04,
         1.2133e-03, -1.2545e-04,  8.6390e-04, -1.7658e-03,  3.1977e-03,
        -2.9580e-04,  4.1975e-05, -1.0922e-03, -6.2974e-04, -1.3332e-03,
         1.0519e-03,  2.2426e-03,  4.0378e-04, -1.0644e-03,  7.8658e-04,
        -7.0673e-04, -2.5133e-03, -2.3051e-03,  3.6729e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.9889e-03,  9.2631e-01,  7.0471e-03,  4.6488e-03,  5.1005e-03,
         1.3326e-03,  1.1118e-03, -1.1806e-03,  3.4981e-04,  2.9163e-03,
        -8.0836e-04,  1.0949e-03, -1.1703e-03,  1.0229e-03,  7.0729e-03,
        -3.6357e-03, -1.0954e-02,  2.5138e-04, -1.8740e-03,  1.7800e-03,
        -2.9915e-03,  1.6279e-05,  1.8575e-03, -1.0423e-03, -1.8500e-03,
         8.5224e-05, -2.5467e-04, -2.2535e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4192e-03,  9.5633e-01, -6.3002e-03, -2.3144e-03,  8.6428e-04,
        -2.8650e-03, -7.1135e-04, -1.3860e-03, -1.1278e-03, -8.0730e-04,
        -2.7990e-03, -3.0513e-03, -2.5023e-03,  1.4453e-03, -2.5811e-03,
        -4.6330e-04, -1.6608e-03,  2.9081e-04,  1.2493e-03,  1.5881e-04,
        -5.6200e-04,  3.6375e-04, -3.0319e-03,  4.7119e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-3.3720e-03,  9.4474e-01,  4.6872e-03, -1.1338e-03,  4.1288e-03,
         1.7813e-03, -1.7283e-03, -9.4172e-04,  5.3483e-04, -8.9958e-04,
        -4.4712e-04,  2.8063e-03,  9.2624e-05,  5.0021e-04,  3.5639e-04,
         1.9813e-03,  1.7342e-03,  5.7040e-04,  1.0303e-03, -9.9757e-03,
         9.0993e-04, -4.3959e-04,  2.6206e-03,  1.8221e-03, -2.5369e-03,
        -5.5101e-04, -2.1762e-03, -1.4054e-03, -1.0150e-03,  7.9932e-05,
         3.0063e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2291e-02,  9.6017e-01,  3.8626e-04, -2.1232e-03, -3.5549e-03,
         8.4107e-04, -2.6924e-03, -2.0233e-03, -8.4588e-04, -8.9747e-06,
        -5.3741e-04, -1.4526e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.5366e-03,  9.5877e-01,  4.1987e-03,  7.4035e-04,  4.2749e-03,
         3.4016e-03, -2.5888e-03, -2.0274e-03,  1.5684e-03,  7.0275e-04,
        -1.5002e-03,  6.6488e-03,  5.1146e-04,  1.0622e-04,  7.4220e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.0072e-03,  9.5160e-01, -2.6837e-03, -7.1963e-04,  1.4419e-03,
         1.2162e-03,  2.0552e-03,  1.0004e-03,  2.1362e-03,  9.4814e-04,
        -3.6364e-04,  2.2544e-03,  2.9191e-03,  2.2546e-03,  2.1110e-03,
        -5.4865e-04,  1.2020e-03, -1.2839e-03, -2.0136e-03, -6.1926e-04,
         7.9745e-04,  5.0207e-04, -8.3711e-05,  5.8236e-04,  1.6881e-04,
         1.5744e-03,  1.3225e-03,  3.0457e-04,  2.8187e-04, -3.3423e-04,
         2.4912e-03, -6.8347e-03, -3.3890e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.8955e-03,  9.4949e-01, -1.3401e-02, -1.7583e-03, -1.6253e-04,
        -1.2305e-03,  6.4566e-04, -5.5300e-03, -3.2622e-03, -9.3255e-04,
        -3.4591e-03, -2.7934e-03,  6.9509e-04,  5.1371e-04,  3.5747e-03,
         5.6560e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0111,  0.9552, -0.0111, -0.0060,  0.0044, -0.0026, -0.0014, -0.0043,
        -0.0039,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0200,  0.9029, -0.0412, -0.0220,  0.0139,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2822e-03,  9.4589e-01,  1.1949e-02,  3.2185e-03,  8.9130e-04,
        -1.5143e-03,  1.4967e-03,  3.4848e-03, -8.8917e-04,  4.8055e-03,
         4.7939e-03, -4.3614e-04,  8.8333e-04,  1.1293e-03,  1.5892e-03,
         1.0279e-03,  1.5234e-04, -4.8729e-05,  1.6045e-03,  4.8106e-04,
         7.2376e-04,  1.3778e-03,  9.9504e-05, -2.4213e-03,  5.3818e-04,
        -1.3415e-03,  4.9309e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0272,  0.9383,  0.0017,  0.0144,  0.0053,  0.0118,  0.0014,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5059e-03,  9.5782e-01,  2.7637e-03,  4.2424e-03, -6.0574e-04,
         3.1616e-03, -3.6275e-03, -1.2289e-03, -7.0304e-04,  2.6947e-03,
        -1.1239e-03, -1.0138e-03, -1.6179e-03,  3.6598e-03, -9.7630e-04,
        -1.5487e-03,  1.6656e-03, -9.5621e-05,  4.4386e-04, -1.2382e-03,
         2.3618e-03,  5.9034e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.8848e-04,  9.6799e-01,  2.4513e-03,  1.3335e-03,  3.3679e-03,
         7.2359e-04, -1.7339e-03, -2.5036e-03,  1.4160e-03, -3.0147e-03,
        -2.0524e-03, -9.4086e-04, -3.7655e-03,  1.2265e-03, -1.1331e-03,
         1.7600e-03, -2.9911e-04, -9.6659e-04, -2.8303e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4952e-03,  9.5929e-01, -6.5472e-03, -2.7996e-03,  2.0932e-03,
         8.5161e-04,  1.0205e-03, -5.2304e-04,  3.0369e-04, -1.6789e-03,
        -4.2220e-04,  1.5641e-03,  5.9780e-04, -6.4680e-04,  1.1439e-03,
        -2.0552e-04,  1.6907e-03,  7.0364e-04,  3.9304e-03,  1.0716e-03,
         1.1777e-03,  3.6951e-03,  5.0023e-04,  5.0974e-04, -4.4306e-04,
         1.9433e-04,  1.6947e-04,  4.9598e-04,  2.6699e-04, -3.6108e-04,
         1.6038e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-5.9412e-03,  9.3828e-01,  1.0593e-02,  2.9497e-03,  1.2178e-03,
        -7.2970e-04,  1.5653e-03,  2.6733e-04,  1.5326e-03,  1.6833e-03,
         4.6288e-04,  1.4806e-03,  6.5927e-04,  7.7802e-03,  5.6641e-04,
         7.9546e-04,  1.1039e-03,  1.1721e-03,  3.5449e-03,  4.3677e-04,
         1.2995e-03,  2.0873e-03,  1.2041e-03,  4.4092e-03, -1.5977e-03,
        -1.7243e-03,  2.7443e-04, -8.1035e-04, -3.1225e-04,  2.4854e-03,
         4.2571e-04,  6.0568e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1378e-03,  9.6261e-01,  3.8778e-03,  1.1345e-04,  1.2203e-03,
        -6.8989e-04,  2.6791e-04, -6.0090e-04,  5.1718e-03,  1.4745e-03,
        -6.7194e-05,  6.2060e-03,  1.9226e-03,  1.1444e-03,  1.3238e-03,
        -1.0112e-03,  2.0490e-03, -8.1354e-04, -8.5016e-05, -4.9576e-04,
        -9.0517e-04,  2.8074e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0949e-04,  9.5829e-01, -6.6510e-03, -2.5601e-03,  4.4671e-04,
         2.0812e-03, -6.1928e-04, -9.7539e-05,  6.5452e-04,  3.8702e-04,
         4.8601e-04,  9.9125e-04, -4.5674e-04, -9.2424e-04, -4.4758e-03,
         5.7069e-04, -7.7996e-04,  5.2157e-03, -2.7641e-03,  3.3635e-04,
         2.0900e-03,  6.5162e-04, -8.7647e-05, -7.6345e-04, -7.4751e-04,
         5.6301e-04,  6.6024e-04, -3.2956e-03,  3.9552e-04, -1.1435e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.8045e-04,  9.5945e-01,  6.1466e-03, -2.4206e-04, -6.9192e-04,
        -3.0567e-03,  2.6002e-03, -1.9253e-03,  2.5832e-03, -6.4656e-03,
        -9.5845e-04, -9.4614e-05, -1.6641e-03,  6.7882e-04, -1.5143e-03,
        -3.4985e-03, -4.3876e-03, -3.5586e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0130e-03,  9.3107e-01,  2.4327e-02, -2.1665e-04,  5.2816e-03,
        -3.6691e-03, -1.4245e-03, -1.0419e-03, -1.4721e-03,  1.0359e-04,
         7.7067e-05, -2.0551e-03, -6.9028e-04, -2.8774e-04, -1.0902e-03,
        -4.3786e-04,  7.0229e-05, -3.2773e-04, -2.4851e-03, -1.1026e-03,
         1.2050e-03, -3.7636e-03,  7.8210e-04,  1.6288e-03,  5.8229e-05,
        -1.8290e-04, -2.9033e-04, -1.7973e-03, -3.8852e-04, -2.4008e-03,
         5.8001e-04, -2.0625e-04,  1.7732e-03, -1.9983e-04, -1.1110e-03,
        -1.4584e-03, -2.9328e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5330e-02,  9.3235e-01, -2.3491e-03,  2.2299e-03,  3.2277e-03,
         2.9209e-03,  4.8307e-03, -6.7178e-03,  1.5097e-03,  8.7069e-04,
         4.0174e-04, -4.1828e-03, -2.5141e-03,  2.1707e-03,  3.8055e-03,
         1.7112e-03, -3.5373e-04,  4.0941e-04, -1.4420e-04,  1.7358e-03,
         1.8570e-03,  5.5243e-04,  1.6053e-03, -6.2185e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.6701e-04,  9.5566e-01, -1.1603e-03, -1.9670e-05, -4.5611e-04,
         1.0405e-04, -1.3036e-03,  1.1309e-03, -3.4949e-04, -6.6333e-04,
        -1.1944e-03, -4.9428e-04, -1.4781e-03,  6.3037e-04, -7.9533e-04,
         7.3968e-04, -6.9687e-04, -1.5214e-03,  1.0330e-03,  1.6250e-03,
         9.7021e-04, -1.3467e-03,  4.3167e-04, -1.2972e-03, -5.2486e-04,
         2.1752e-03,  1.1173e-04, -1.0024e-03,  3.7940e-04, -2.6809e-04,
         1.8428e-03,  1.3272e-04,  6.9692e-05,  1.3352e-03,  1.5481e-03,
         1.8671e-03,  9.5505e-04,  2.5005e-04, -1.9553e-04, -6.3637e-04,
         7.4666e-04,  2.6359e-03,  2.0434e-03,  3.9078e-03,  1.0158e-03,
        -2.8361e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3646e-02,  9.4990e-01, -3.2983e-03, -1.6424e-03,  2.4904e-03,
        -8.2527e-04,  1.2631e-03,  3.4453e-04,  1.8157e-03, -3.0173e-03,
         1.2013e-03,  2.1359e-04, -1.8287e-03, -4.5169e-03,  2.5073e-03,
         2.2709e-04, -9.7223e-04,  1.4066e-04, -1.4892e-03, -1.1261e-03,
         1.1908e-03,  2.2271e-03, -4.1130e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.2417e-03,  9.6859e-01, -1.7318e-03,  4.4239e-03, -7.0059e-04,
        -7.2130e-04,  3.2503e-03,  6.0692e-03,  6.7639e-03, -1.6894e-05,
        -1.4937e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3053e-02,  9.5890e-01, -6.5366e-03,  4.8706e-04, -8.8575e-04,
        -2.9693e-03, -1.3300e-03, -6.5623e-04,  1.4698e-03, -1.5622e-03,
         8.4689e-04, -2.7023e-03, -3.1808e-03, -5.4172e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0214,  0.9125, -0.0077, -0.0024, -0.0041, -0.0024, -0.0098, -0.0052,
        -0.0054, -0.0228,  0.0015, -0.0047,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0458e-02,  9.3814e-01, -3.6659e-03,  1.9387e-03,  1.2450e-03,
         4.0180e-03,  1.1851e-03,  9.3941e-04,  1.1421e-03,  4.1718e-03,
         1.7232e-03,  5.6272e-04,  9.9102e-04,  3.8536e-04,  5.0462e-04,
         2.7820e-04,  1.1824e-03,  1.3699e-03,  2.3420e-03,  5.6443e-04,
        -1.4207e-04,  6.9796e-04,  1.5199e-03, -5.9539e-04,  4.4010e-04,
        -1.8284e-03,  1.1188e-03, -5.3532e-04,  3.6968e-04, -8.9597e-04,
        -1.5094e-04,  2.0153e-03, -9.7483e-04, -4.5153e-04,  6.5315e-04,
        -5.9470e-04, -1.8563e-04,  1.1626e-03,  2.7660e-04, -4.1464e-04,
         7.3860e-04,  1.5167e-03, -3.5275e-04,  8.7518e-04,  2.2467e-04,
         1.8415e-04,  5.1643e-04, -2.3410e-05,  7.8711e-04,  8.8800e-04,
        -1.3027e-04,  3.9489e-04, -2.6017e-04,  5.6679e-04,  7.0952e-04],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 4.2366e-03,  9.6606e-01, -1.4207e-03, -3.3051e-03,  7.0945e-05,
        -4.1681e-04, -1.4691e-03,  6.8000e-05,  3.5359e-03, -2.7672e-03,
        -1.9545e-03, -3.5648e-04,  1.0471e-03, -3.4624e-03,  9.6974e-04,
         2.9235e-03, -1.8890e-03,  5.1856e-04,  8.2937e-04,  3.6969e-04,
        -2.9739e-04,  4.3874e-05, -1.9890e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4336e-03,  9.6027e-01, -2.6603e-03,  1.7927e-05,  1.2510e-03,
        -9.2179e-05, -1.3668e-03, -2.7476e-04, -6.5874e-03, -1.3162e-03,
        -1.4012e-03,  8.1745e-04,  1.5641e-03,  1.3054e-03,  2.1496e-03,
         2.5443e-04,  1.0390e-02, -1.8488e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.1734e-03,  9.4290e-01, -9.6952e-03,  1.0618e-03,  4.1723e-03,
        -2.2412e-03, -4.1168e-03,  3.3465e-03,  4.0207e-03,  1.4399e-03,
         4.1564e-03,  1.5489e-03, -2.4004e-04, -7.2897e-04,  5.9263e-03,
        -1.8666e-03,  2.9088e-03, -4.5954e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7289e-03,  9.4042e-01,  3.1448e-03, -3.2067e-04, -3.4057e-03,
         1.7738e-02,  3.6971e-03, -4.5315e-03,  3.9127e-03, -1.3937e-03,
         2.0524e-03,  2.0027e-03,  3.5773e-03, -5.1192e-03, -5.9548e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2684e-03,  9.5513e-01, -2.9407e-03, -4.4298e-03, -1.4433e-03,
        -9.8560e-04, -2.0574e-03, -1.5943e-03, -2.5175e-03,  2.0950e-04,
        -3.8228e-04, -2.5416e-03,  1.6216e-03, -8.2591e-04, -2.7946e-03,
        -1.0231e-03,  4.8786e-03, -5.4605e-04,  1.2133e-03,  4.4811e-03,
        -3.0437e-04,  4.5520e-05,  1.7570e-04,  7.3834e-04, -2.5923e-04,
        -4.3803e-04,  5.9404e-04, -2.4820e-04, -2.9950e-03, -7.4363e-05,
         1.2055e-03,  3.2027e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.2578e-03,  9.4822e-01,  3.1812e-03,  3.7018e-03,  6.3950e-03,
        -3.2642e-04,  1.2466e-03,  2.1861e-03, -2.3266e-04, -1.1781e-03,
         1.4748e-04, -1.1143e-03, -4.0335e-03,  8.9085e-04,  2.7035e-06,
         4.1588e-06, -1.6086e-03, -6.4899e-04,  1.7531e-03,  8.3979e-04,
         2.0655e-03, -2.9023e-03, -2.6761e-03, -6.5212e-05,  1.1798e-03,
         2.4835e-04,  1.5230e-03,  1.7653e-03,  2.0550e-04, -1.1753e-03,
         9.5588e-04,  1.1553e-03,  1.0909e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7808e-03,  9.5726e-01, -5.1578e-04,  4.3031e-03,  3.3368e-03,
        -1.2257e-04, -7.5192e-04,  2.0229e-03,  2.0317e-03,  5.8812e-04,
        -2.5179e-04, -9.6576e-04, -7.6075e-04, -1.0774e-03, -1.4042e-03,
        -1.3005e-03,  7.5208e-04,  2.2272e-04,  6.5165e-04, -5.4585e-04,
         3.0123e-04, -5.2416e-04,  1.5143e-04, -1.4205e-03, -1.2247e-04,
        -2.7481e-05, -4.2740e-04, -5.8071e-04, -1.6646e-03,  8.1821e-04,
        -5.9809e-05,  4.1471e-04,  2.4857e-04, -8.7884e-04,  3.3056e-04,
        -1.0051e-03,  1.2727e-04,  7.8756e-04,  4.3063e-04, -5.6309e-04,
         1.4483e-04, -7.5233e-04,  5.6280e-04, -1.4291e-04,  7.3194e-04,
         2.6270e-04, -7.3324e-05,  8.2067e-04, -1.6665e-04,  2.0771e-03,
         5.3442e-05, -2.7420e-05,  4.9160e-05,  1.4790e-04, -3.4810e-04,
         7.8377e-04,  4.1019e-04, -9.1795e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.4940e-03, -9.5509e-01, -5.4825e-03, -8.6777e-04,  1.9926e-03,
         1.4526e-03,  1.9860e-03, -5.9616e-04, -7.3350e-04,  7.8963e-04,
        -8.4152e-04,  1.8663e-03, -1.1084e-03,  7.5115e-04, -7.4993e-03,
        -5.7740e-04, -9.3333e-04, -1.1592e-03, -6.6014e-04, -1.9345e-03,
         1.2887e-03, -7.6812e-04,  4.1438e-04, -4.5528e-04,  1.4107e-03,
         1.8456e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9723e-03,  9.4432e-01, -1.2405e-03,  5.4652e-03, -2.0577e-03,
         5.8870e-03, -1.1931e-03,  4.3368e-03,  9.0855e-04,  2.8551e-03,
        -8.0525e-05, -2.3316e-03, -5.8984e-04,  2.7104e-03,  3.4751e-04,
         1.6292e-03,  2.4639e-03,  1.2171e-03,  2.5222e-03, -2.1324e-04,
         2.4056e-03, -3.7870e-03,  9.5908e-04,  3.9924e-03, -3.5152e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6236e-02,  9.3618e-01, -4.5099e-03,  1.3107e-03,  4.4408e-03,
        -1.2315e-03, -7.2144e-04, -1.6464e-03,  7.0397e-04, -1.8441e-04,
         3.7367e-04, -6.3660e-04, -3.3018e-03, -9.2478e-04, -4.1195e-04,
         1.5073e-04, -9.8266e-04, -9.9985e-04, -6.4020e-04,  2.9172e-03,
         5.5795e-04, -4.1215e-04,  6.6761e-04,  2.9162e-04,  7.2882e-04,
        -3.0202e-03, -5.3340e-04, -1.3808e-03, -2.5284e-04, -3.2055e-04,
         1.3940e-03, -7.2632e-04, -1.4411e-03, -9.8492e-04,  5.0247e-04,
        -8.8891e-04, -9.7873e-04, -4.9267e-04, -7.6930e-04,  3.2342e-04,
        -9.4696e-04,  7.9079e-04, -3.8842e-04,  9.4016e-04,  1.7578e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.6291e-04,  9.4901e-01, -5.2701e-03,  1.5204e-03,  8.6275e-04,
         8.5606e-04, -7.0186e-04,  1.1713e-05, -1.0436e-03, -3.0780e-03,
        -3.4430e-03,  1.4870e-03, -2.2266e-03,  1.0596e-03, -2.0105e-03,
        -1.3153e-03,  1.0510e-02,  5.2386e-04, -4.8411e-04, -1.1949e-03,
         3.7191e-03, -5.8665e-04, -8.3199e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5484e-03, -9.5941e-01, -1.0180e-03, -4.9025e-03,  2.8053e-03,
        -1.0767e-03,  7.8413e-03,  4.0729e-04, -1.3972e-03, -8.1633e-03,
        -2.2558e-04, -9.2035e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 4.7387e-03,  9.5271e-01,  1.3817e-02, -4.6936e-03,  5.8569e-04,
         1.9664e-03,  2.0382e-03, -1.8302e-04,  9.2820e-04, -4.2740e-03,
        -5.2449e-04, -1.6558e-03,  2.3100e-03, -1.2383e-03, -1.1472e-03,
         1.0899e-03,  3.5569e-04,  2.3733e-03,  1.6334e-04, -3.0136e-04,
        -2.3403e-03,  4.0382e-04, -1.6291e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3966e-03,  9.6107e-01,  2.9646e-03,  1.3583e-03, -3.6811e-04,
        -1.5950e-03,  4.1154e-03,  1.4592e-03, -2.6079e-04,  2.0856e-03,
         7.7993e-04, -4.7796e-04,  4.7684e-04,  5.2550e-04,  7.7775e-04,
        -1.5594e-03,  4.1481e-04, -5.3589e-04, -1.1803e-03, -4.0062e-04,
         5.2889e-04,  5.8815e-04, -2.1929e-03, -8.7071e-04,  5.3458e-05,
        -7.4672e-04, -1.8455e-03,  9.5350e-06, -1.2755e-03, -8.7231e-04,
         9.9116e-04, -2.9419e-04, -1.9301e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6115e-02,  9.5383e-01, -8.3497e-03,  5.4809e-05,  3.5772e-03,
        -3.2010e-03,  1.2939e-03, -1.0486e-03,  9.4980e-04,  5.7850e-05,
         1.2317e-03, -4.1047e-04,  3.3372e-03,  1.9261e-03,  3.8804e-04,
         5.9104e-04,  2.2633e-04, -2.4838e-03, -9.2855e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3039e-02,  9.5184e-01,  1.2571e-02, -7.6605e-04, -1.8483e-04,
         4.8379e-03,  1.9661e-03, -3.4571e-04,  1.9177e-04, -8.9127e-04,
        -3.9697e-03,  3.1243e-04, -6.6707e-03, -2.4141e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.3735e-03,  9.6165e-01,  2.1864e-03, -6.5347e-03, -8.3319e-04,
        -1.0720e-03, -8.5973e-04, -1.6090e-03, -5.3267e-04,  1.0014e-03,
        -1.6166e-03, -2.7061e-04, -1.0178e-03,  2.9415e-03,  3.5201e-04,
        -1.2118e-03, -9.5341e-04,  8.5287e-05, -6.3720e-04, -6.0916e-04,
         1.8619e-04, -4.4676e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2592e-03,  9.5166e-01, -4.9681e-03,  1.2843e-03, -1.3153e-03,
        -3.2705e-03, -2.5326e-03, -1.2477e-03,  5.2391e-04,  1.7296e-03,
        -2.4961e-03,  1.0221e-03, -4.4643e-04,  9.7077e-04, -1.1716e-04,
         1.6140e-03, -1.3572e-03, -1.8732e-03, -4.6083e-03, -8.0085e-04,
        -1.8995e-03,  2.4468e-04, -7.9983e-04,  1.9112e-04, -1.7769e-03,
        -8.4838e-04, -1.8713e-03, -1.0540e-03,  2.4162e-03, -2.8014e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9301e-03,  9.6155e-01,  1.4262e-03, -4.8565e-03, -3.7634e-03,
         3.5856e-03,  3.4156e-03,  4.2232e-03,  5.9707e-04,  1.2135e-03,
         3.0956e-03,  1.1763e-03, -1.2505e-03,  2.9085e-03, -3.0027e-03,
         1.4817e-03,  5.2443e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.2214e-03,  9.4965e-01,  3.9435e-04, -6.2405e-03, -4.4529e-03,
        -4.7988e-04, -3.7755e-03, -6.2216e-03, -9.2128e-03, -3.4482e-03,
        -6.9066e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.5715e-03,  9.6948e-01, -2.7721e-03, -3.5401e-03,  2.0059e-03,
         1.0502e-03, -1.2603e-03,  6.7724e-04,  3.3911e-04,  1.6241e-03,
        -1.6103e-03,  2.8992e-03,  6.2217e-04,  2.7908e-03,  8.7157e-04,
        -2.2629e-03, -1.6252e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9756e-03,  9.5232e-01,  5.7566e-03,  3.8166e-03, -4.6046e-04,
        -3.8826e-04, -1.3002e-03,  5.6461e-04, -2.7299e-04,  8.3283e-04,
         3.2193e-03, -1.5178e-03,  2.5640e-03, -2.3190e-03,  2.3717e-03,
         2.4949e-05,  3.0331e-04, -5.8370e-04, -1.6673e-03,  2.6138e-03,
        -1.1633e-03,  2.1099e-04,  3.2788e-03,  1.1340e-03,  5.9468e-03,
        -3.3908e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2298e-02,  9.0549e-01,  2.3662e-02,  2.3143e-03,  5.4219e-04,
        -4.7851e-03,  3.5320e-03,  1.8797e-03, -2.1222e-03, -1.3901e-03,
        -8.5465e-04,  2.7969e-03, -4.0117e-04,  6.9198e-04, -9.4630e-04,
         9.4141e-04, -1.3112e-03,  8.2451e-04, -9.1028e-04, -2.8798e-03,
        -7.7864e-04, -4.9798e-04, -3.9917e-05,  3.6468e-04, -2.2838e-03,
        -8.1149e-04,  1.2547e-04, -1.8920e-03, -2.5136e-03,  1.2213e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8634e-03,  9.5785e-01, -5.1546e-03,  2.9703e-03,  7.3008e-04,
        -1.8636e-03,  5.6704e-03,  3.0424e-03, -5.5293e-04,  1.0809e-03,
        -5.5938e-04,  1.0535e-03,  2.5691e-03,  3.6448e-03, -2.1355e-03,
        -2.1681e-03,  1.6867e-04,  1.5768e-03, -3.3432e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 1.8566e-03,  9.4734e-01,  3.2208e-03,  5.8406e-03,  8.9548e-04,
        -2.2524e-03,  3.3669e-03,  1.9401e-03,  1.2606e-04,  5.3737e-03,
         8.3079e-03,  1.9200e-04,  1.3618e-03,  1.9404e-03,  2.3088e-03,
        -6.2884e-04,  1.8446e-03,  1.0651e-03,  1.2932e-03,  2.7908e-03,
         1.3047e-04,  8.2391e-04,  3.2554e-04, -1.2143e-03, -3.3900e-04,
        -1.6054e-03, -1.6154e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2698e-02,  9.6203e-01, -4.8555e-03, -4.0105e-04, -2.0454e-03,
        -2.3518e-03, -1.3032e-03, -1.4479e-03,  2.1716e-03,  2.6584e-03,
        -1.9464e-03,  4.5976e-03,  8.3000e-04,  6.6766e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2693e-03,  9.4077e-01,  1.7921e-02,  8.8101e-05, -1.3168e-03,
         3.2789e-03, -1.9825e-03,  6.2980e-03, -5.6130e-04, -2.6086e-04,
         2.0868e-03, -1.0215e-03,  5.4257e-04,  2.1177e-03, -8.1418e-04,
         2.4805e-04, -2.2923e-03, -1.2440e-03,  1.0431e-03, -2.0403e-03,
        -2.1596e-03, -2.2043e-03,  1.3809e-03, -7.7334e-04,  9.6298e-04,
        -4.3208e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0409e-03,  9.6559e-01,  5.2470e-03, -3.8195e-03, -1.6011e-03,
        -6.2430e-03, -7.8285e-04,  1.3265e-03,  1.9245e-03,  5.2052e-03,
         8.2945e-04, -1.5480e-03,  1.2467e-03,  1.5989e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6290e-03,  9.5634e-01, -6.8378e-03, -2.1011e-03,  7.0217e-04,
        -3.1719e-04, -2.5943e-04,  9.9611e-04,  1.6805e-04, -1.7943e-04,
         2.5892e-04,  4.8914e-04,  1.5670e-03,  5.3284e-05,  2.7421e-03,
        -9.5107e-04,  1.0526e-03,  5.4589e-05, -9.9062e-04,  2.3190e-03,
        -7.6113e-04, -7.7475e-04, -5.9671e-04,  1.1204e-04,  1.7525e-03,
        -3.1573e-03, -4.6471e-03,  1.3541e-03, -2.4892e-04, -2.2458e-05,
        -1.3272e-03,  1.2359e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1421e-03,  9.5363e-01,  9.4524e-03, -5.2510e-03,  2.9235e-03,
        -7.2508e-04, -1.4512e-03, -1.3633e-03,  2.1370e-03, -8.9038e-04,
         1.1334e-03, -3.9981e-04,  2.3960e-03, -3.9714e-04,  2.0473e-04,
         1.0830e-03,  1.5094e-03,  8.8522e-04,  5.1228e-04,  2.4462e-04,
         1.3918e-03,  2.9915e-04, -4.1800e-05,  4.2226e-03, -5.3099e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([0.0028, 0.9842, 0.0010, 0.0059, 0.0035, 0.0026, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000, 0.0000,
        0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.8140e-03,  9.5564e-01, -8.4489e-04, -2.4518e-03,  9.7960e-04,
        -9.4872e-04, -2.4694e-03, -2.7001e-03, -1.9175e-04,  1.0433e-03,
         5.0843e-05, -1.7040e-03,  3.9163e-04,  6.5758e-04,  6.7807e-04,
        -1.1215e-04,  6.5421e-04, -2.4117e-03, -1.8845e-03,  1.9033e-03,
        -1.1331e-03,  1.4065e-04, -6.8770e-04,  3.4309e-04, -7.0883e-05,
         3.1497e-03, -1.9869e-04,  9.1098e-06, -9.3369e-04,  1.5175e-03,
        -1.9694e-04, -1.9759e-03, -1.1415e-03, -3.5748e-04, -3.7575e-04,
         1.5127e-03,  1.7286e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0853e-03,  9.3472e-01, -4.5221e-04,  7.5203e-03,  5.3239e-04,
         8.6572e-04, -1.7480e-03,  3.9575e-03, -3.4973e-03,  6.3770e-04,
        -1.8615e-04,  1.2659e-03, -4.1071e-04, -7.7790e-04, -1.5919e-03,
        -3.4768e-03, -1.0842e-03,  1.7361e-03,  7.6698e-04,  3.3252e-03,
        -5.2501e-03, -6.1396e-06,  1.7795e-03, -4.1791e-04, -1.2148e-03,
        -7.9487e-04, -3.3684e-03,  5.0888e-03,  3.8346e-04,  2.0400e-03,
        -1.4453e-03, -3.6069e-04, -3.6427e-04, -6.4604e-04,  4.2050e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3597e-02,  9.2156e-01, -1.4229e-02, -2.7450e-03, -6.9059e-03,
         2.5446e-04, -3.9222e-03, -2.8037e-03, -5.2027e-03, -5.8728e-03,
        -3.5086e-03, -5.4530e-03, -2.4412e-04, -2.1935e-05,  3.6764e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0409e-02,  9.5970e-01,  1.0735e-02, -1.2688e-03, -3.6327e-04,
         6.9060e-04,  3.1502e-04,  5.8156e-04, -1.3004e-03,  4.9571e-04,
         1.3061e-04, -1.0917e-03, -4.5876e-04, -4.7660e-04,  5.2557e-04,
        -6.2470e-04,  4.9924e-05, -1.6839e-03,  1.2924e-03,  1.2872e-03,
         4.2223e-04, -2.0980e-04,  6.4700e-05, -5.4577e-04,  1.1839e-03,
        -1.3883e-03,  4.3415e-04,  2.4232e-04, -5.9190e-06,  4.2423e-04,
         1.9786e-04,  4.9366e-05, -1.3510e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4151e-02, -9.4900e-01,  1.7719e-03,  3.0297e-03,  1.2812e-03,
         2.5994e-03,  1.4957e-03,  6.8706e-04, -3.8222e-03, -8.9529e-04,
         1.3340e-03, -3.6806e-03,  1.3573e-03,  1.3201e-03, -2.1488e-04,
         2.0556e-04, -2.2179e-04,  1.3029e-03, -1.8354e-03, -3.2641e-04,
        -3.7116e-04,  1.0731e-03,  1.6492e-03, -3.2463e-04,  4.8563e-04,
         1.0362e-03,  1.8679e-03, -2.2755e-03,  3.8336e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([-1.3228e-02,  9.5445e-01, -5.9526e-03,  3.7270e-03, -3.6839e-03,
         2.6927e-03,  3.2559e-03, -3.5780e-03,  9.2096e-04, -9.6185e-04,
        -3.1733e-04,  1.1810e-03,  5.0154e-04,  3.1020e-04, -5.2355e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.9648e-05,  8.3641e-01,  1.2416e-02,  1.5246e-03,  1.2770e-03,
        -1.4359e-03, -3.3500e-03, -9.3577e-03, -1.8361e-03,  6.0318e-03,
        -4.1049e-03, -2.1826e-03, -7.7731e-03,  1.0276e-04, -3.1878e-04,
         7.0863e-04,  1.8772e-03,  5.3756e-03,  6.7976e-03, -3.8778e-03,
        -1.0700e-03,  1.5457e-03,  1.3481e-03, -8.2281e-04, -1.6914e-04,
        -4.9703e-04,  4.7793e-03,  3.4726e-03,  4.3735e-03,  4.6271e-03,
        -1.6046e-03, -4.7087e-04,  1.0013e-03,  1.2431e-03,  1.8800e-04,
         3.4583e-03,  2.8661e-03,  2.9905e-04,  1.7616e-03,  8.6029e-04,
        -2.1751e-04, -1.4758e-03,  1.9597e-04, -1.4207e-03,  3.9209e-04,
         3.0588e-03,  5.6652e-03,  6.1013e-03, -3.8210e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4926e-02, -9.0625e-01,  2.2305e-04,  2.9591e-04,  2.6471e-03,
        -3.4316e-03,  1.7427e-03,  1.0828e-03,  3.8613e-03, -3.0767e-03,
         2.0242e-03,  2.5850e-03,  3.8509e-03,  2.1084e-03,  1.3291e-03,
        -3.5774e-05, -5.3803e-03, -5.7310e-04, -1.8451e-03, -6.5176e-03,
         1.6122e-03,  2.0439e-03,  1.7162e-03, -2.3361e-03,  1.0580e-03,
         4.5282e-03,  3.4333e-03, -2.8002e-03,  1.0024e-03,  1.7190e-03,
        -2.9775e-03, -1.1914e-03, -5.9258e-03,  1.0236e-03, -2.8436e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1975e-03,  9.5893e-01, -9.4589e-03,  1.1349e-03,  5.9288e-03,
        -1.3036e-03, -3.1030e-04,  1.1524e-03,  4.7836e-03, -1.2590e-03,
        -1.0552e-03, -2.5113e-04, -7.9536e-04, -1.4436e-03,  3.5819e-03,
         1.6248e-03, -1.5124e-03, -4.0529e-03, -2.2215e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8474e-02,  9.3651e-01, -7.3420e-03, -6.8378e-03, -1.4326e-03,
         4.8010e-03, -2.3086e-03, -2.2100e-03,  1.7562e-03,  1.9309e-04,
         5.3001e-03,  3.8566e-03, -4.3542e-03, -2.7638e-03, -1.8589e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.9487e-04,  9.4414e-01, -1.3217e-02,  1.1657e-03, -3.4383e-03,
        -8.1632e-04, -3.1093e-03, -1.9458e-03, -3.6349e-05,  1.6226e-03,
        -5.7105e-03,  6.5238e-04,  2.4127e-04, -4.1094e-03,  1.1401e-03,
        -3.8322e-03, -2.8969e-04, -5.4856e-03,  2.2056e-04,  4.0047e-03,
         3.1015e-03, -9.2607e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3885e-02,  8.6173e-01, -1.2207e-02,  1.1081e-03, -5.1901e-03,
        -2.0411e-03, -1.1096e-02, -6.7481e-03, -5.0478e-03,  4.1095e-03,
        -2.4363e-03, -8.3534e-03, -8.4906e-04, -2.7440e-03, -4.2721e-04,
         3.2724e-03, -9.5149e-04, -6.3233e-03,  4.5785e-03, -2.3278e-03,
         2.6215e-03,  2.9465e-03,  4.0695e-04, -2.3712e-04, -5.9115e-03,
         1.1985e-03, -5.5225e-04, -7.8573e-03,  4.8942e-03, -1.7953e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0057,  0.9230, -0.0141, -0.0042,  0.0029, -0.0022, -0.0082,  0.0018,
         0.0125,  0.0009, -0.0245,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6739e-02,  9.2895e-01,  4.8745e-03, -4.4901e-03, -1.7621e-03,
        -1.6342e-03,  3.9688e-03, -2.7626e-03,  1.2719e-03, -1.7584e-03,
        -3.3766e-03,  1.6535e-03, -1.2501e-03, -2.1473e-03,  1.4237e-04,
         3.7531e-03, -1.4727e-03,  3.6737e-04, -1.1962e-03, -2.4915e-03,
        -2.7488e-03, -1.1933e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3483e-02,  9.2495e-01,  1.0268e-02, -2.8656e-03, -1.3197e-03,
         2.2561e-03, -8.9040e-04,  1.1070e-04, -2.0700e-03, -1.5821e-03,
        -1.2585e-03, -2.6462e-03,  2.5362e-04, -9.5236e-04, -9.2901e-04,
        -3.0908e-03, -3.8663e-04, -1.8334e-03, -1.8435e-03,  1.3790e-04,
        -1.9770e-03, -1.9141e-03, -5.6483e-04, -3.9050e-04,  5.4867e-03,
        -6.2225e-04, -3.5000e-03, -4.1331e-04, -3.0644e-03, -3.6104e-03,
         1.9204e-03, -1.8338e-03,  1.5741e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1436e-02,  9.2133e-01, -4.9630e-04, -1.0813e-02, -3.8655e-03,
        -8.4155e-04, -1.7052e-03, -6.5404e-03, -1.5743e-03, -1.8827e-03,
        -6.2313e-03, -1.3951e-03, -2.4202e-04,  1.1967e-03, -1.0018e-03,
        -1.8066e-03,  3.6542e-03,  1.7935e-03, -7.6065e-04, -1.8058e-04,
        -1.0999e-03, -6.8880e-03,  1.3758e-03,  2.2503e-04,  2.7999e-03,
         4.0733e-03,  1.3469e-03, -3.4411e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9370e-03,  9.1279e-01,  2.0243e-02, -1.4619e-03, -5.7730e-04,
        -3.7912e-03,  1.4209e-02,  3.7233e-03, -2.1921e-03,  2.9095e-03,
        -9.5492e-06, -2.2723e-03, -2.0597e-03,  7.3044e-04, -1.5046e-03,
        -9.8879e-04, -2.0280e-03, -1.7332e-03, -2.9917e-03, -5.6098e-03,
        -1.7217e-03,  2.4008e-03, -9.3766e-04,  6.7486e-03, -4.3093e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-2.4552e-03,  9.3124e-01,  6.2729e-03, -3.1228e-03, -3.0541e-03,
         1.7569e-03, -1.2469e-03, -1.4606e-03,  1.2037e-03,  3.4854e-04,
         2.1872e-03,  2.6424e-03,  1.0344e-04,  3.5928e-03, -1.5029e-03,
        -3.6865e-03, -2.6025e-03, -1.0472e-03,  3.4545e-04,  1.1012e-03,
        -1.3777e-03,  6.2352e-04,  5.0715e-04,  1.0063e-03,  4.6591e-03,
         8.0360e-04, -1.2252e-03, -7.8909e-04, -1.1852e-03,  8.6323e-04,
        -7.6865e-04, -4.8712e-04, -1.0705e-03, -4.8495e-04, -1.3080e-03,
        -8.1682e-04, -1.4349e-03, -2.0077e-03,  1.6125e-03, -3.2450e-03,
         2.7486e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.0833e-03,  9.4585e-01, -1.1916e-03, -2.7950e-03, -3.7770e-04,
         2.1162e-03,  3.0134e-03, -8.0897e-04,  1.2019e-03,  2.0526e-03,
         1.4260e-03, -4.2443e-03, -1.7730e-03,  1.4524e-03,  4.9792e-03,
        -5.3383e-04,  1.1332e-03,  3.6555e-03,  1.0953e-03, -5.9917e-03,
        -1.2974e-03, -7.5518e-04,  4.0882e-05,  1.0260e-03, -4.1091e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3391e-02,  9.4446e-01,  5.7977e-03, -2.3371e-03,  3.8207e-04,
        -2.0276e-03, -1.5180e-03, -1.1197e-03, -3.4472e-03,  1.9308e-03,
        -2.1410e-03,  7.6844e-04,  1.3568e-03,  1.0640e-03, -3.5252e-04,
        -1.4823e-03, -2.5551e-03, -1.1637e-04,  1.4408e-03,  7.5533e-03,
         4.7621e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6133e-02,  9.2601e-01, -1.2573e-02, -1.8821e-03, -2.8165e-03,
        -1.7198e-03, -3.9068e-03, -3.9242e-03, -2.4981e-03, -3.2795e-03,
        -5.9011e-04, -7.4857e-03, -1.4706e-03, -1.1323e-03, -2.4529e-03,
        -3.5915e-03, -5.9440e-03, -2.5867e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2821e-03,  9.2811e-01,  7.0255e-03,  2.1110e-03, -2.8149e-03,
         3.4448e-05,  4.5074e-03, -3.9637e-03, -6.1620e-03,  9.7267e-04,
        -1.2788e-03,  2.8288e-04, -1.3643e-03,  2.6388e-03, -1.2162e-03,
        -2.0687e-03, -4.6888e-03, -2.6931e-03, -1.0156e-03, -1.6807e-03,
        -1.3689e-03, -6.2801e-04, -1.2701e-03,  1.6757e-04, -9.9878e-04,
        -3.9084e-04, -2.8497e-03, -1.4289e-03, -7.6758e-04, -3.6101e-04,
        -1.1848e-03,  5.0298e-04, -2.3240e-03, -1.5508e-03,  1.5111e-03,
        -2.4785e-03, -2.6392e-03, -6.6345e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.1560e-03,  9.5304e-01, -3.2705e-03,  1.1943e-03,  1.2270e-03,
        -1.1853e-03, -3.5082e-04,  3.2841e-03, -1.3904e-03,  1.7907e-03,
         3.7403e-03, -6.6268e-03, -8.9913e-04,  1.6029e-03, -1.5553e-03,
         9.3888e-04,  1.7446e-03,  1.1863e-03, -3.2353e-04,  5.4958e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.6083e-03,  9.6626e-01,  2.1894e-03,  2.6641e-03, -2.3560e-03,
         3.2664e-03,  2.4940e-03,  3.7127e-03,  2.6884e-03, -5.5986e-04,
        -6.8377e-04, -3.1435e-03,  1.1711e-03,  2.2030e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7057e-03,  9.1290e-01,  3.8183e-03, -5.5301e-04, -5.5174e-03,
        -3.6953e-03, -7.0499e-03, -1.8114e-03, -9.4434e-04, -3.9049e-03,
        -3.1561e-03, -3.9138e-04, -2.1336e-03,  1.5063e-04,  2.5413e-03,
        -1.1947e-03, -6.0688e-03, -1.0928e-03, -4.3449e-04,  1.2697e-03,
        -4.0020e-03, -4.3674e-03,  9.3300e-04, -1.1368e-03, -2.3064e-03,
        -5.2311e-03, -2.5026e-03, -4.0861e-03, -2.3156e-03,  6.0768e-04,
        -7.1662e-04,  9.4618e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1124e-03,  9.1493e-01,  1.1357e-03, -3.5425e-03, -8.6039e-03,
        -5.6363e-03,  5.3193e-03, -2.0856e-03, -1.1003e-04,  1.5513e-03,
        -1.0307e-04,  5.7187e-03, -7.5041e-04,  1.2928e-03, -1.1985e-03,
         9.2222e-05, -4.4846e-04, -1.1087e-02, -4.9931e-03, -2.7293e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.3123e-03,  9.4069e-01,  4.5437e-04, -7.6431e-03,  8.2647e-04,
        -1.1310e-03, -2.2204e-03, -1.4285e-03, -2.5354e-04,  1.0857e-03,
        -5.5743e-04, -4.0803e-04,  1.4349e-04,  1.4916e-04,  6.3122e-04,
        -8.6723e-04,  6.0865e-04, -2.1381e-03,  4.5747e-04,  4.0268e-04,
        -5.2318e-04,  3.7673e-04, -3.4305e-04,  1.4860e-04,  1.9195e-03,
        -2.6956e-03,  1.6403e-03, -2.7475e-04,  2.0736e-04, -1.8417e-03,
         8.4150e-04,  2.2659e-04, -1.6660e-03,  3.2038e-03, -4.5263e-04,
         9.1883e-05, -9.2214e-05,  8.6341e-04,  5.6735e-04, -1.0016e-03,
        -2.0932e-03, -3.9449e-04, -1.6710e-03, -5.7602e-04,  6.1338e-04,
         1.2455e-03, -3.0195e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0579e-02,  9.5043e-01,  3.1183e-03, -3.3599e-03, -2.4719e-03,
         9.7275e-04,  7.9493e-04, -2.9259e-03,  3.8700e-03,  1.9084e-03,
         2.7707e-03, -4.2126e-04, -9.4413e-04,  3.0154e-03,  2.4211e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2605e-03,  9.3214e-01, -4.3144e-03, -1.8822e-03, -1.4724e-03,
         7.3567e-05,  6.3636e-04,  2.0963e-03,  5.7233e-04, -8.5486e-04,
         4.8610e-04,  4.0372e-04, -7.3755e-04,  3.2982e-03, -4.3124e-03,
        -1.3737e-03,  9.1630e-04, -3.2012e-03, -1.0919e-03, -1.4120e-03,
        -1.0942e-03, -1.8380e-03, -7.9301e-04,  2.2438e-03, -7.5244e-04,
         4.8023e-04,  3.8736e-03, -4.8095e-03,  3.2081e-03,  1.2487e-03,
        -8.2128e-05, -3.1533e-03,  1.3586e-04, -4.9010e-04,  8.8539e-04,
        -7.4940e-05, -3.9581e-04,  6.2389e-04, -2.9584e-04, -6.2260e-05,
        -1.3530e-03,  4.4139e-04, -1.6721e-05,  7.2789e-05,  1.6390e-03,
         8.9753e-04,  1.0502e-03, -4.4552e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #50: [tensor([-6.4639e-03,  9.2807e-01,  9.9073e-03, -8.9514e-04,  4.9150e-03,
         4.7938e-03,  4.0653e-03,  1.2787e-03,  8.9710e-04, -3.6539e-03,
         3.5586e-04,  2.9699e-03,  3.0301e-04,  1.0284e-03, -7.7397e-04,
        -4.5478e-03, -3.3755e-03, -3.9773e-03, -8.8401e-04,  6.4898e-04,
        -1.5393e-03, -4.1143e-03, -2.9604e-03,  4.2812e-03,  1.0664e-03,
        -2.2353e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0901e-03,  9.7270e-01, -2.3358e-03, -9.6368e-04,  1.5116e-03,
         2.4320e-04, -2.0965e-03, -1.8017e-03, -2.1844e-04,  1.5716e-04,
        -4.7146e-03, -1.1488e-03, -6.9838e-04, -1.0833e-05, -2.9316e-05,
         2.8008e-04,  6.0419e-04, -5.7493e-04,  1.7347e-03, -1.8670e-03,
        -4.2193e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6527e-02,  9.1583e-01, -4.6960e-03, -8.1834e-03,  5.8205e-04,
        -3.2776e-03, -1.5894e-05,  3.6903e-03, -6.5278e-04, -1.7269e-03,
        -1.8793e-03,  6.9681e-04, -8.3539e-04, -6.5311e-04,  5.4265e-04,
        -1.1681e-04, -2.2960e-03, -9.3873e-04, -7.0412e-05,  2.5595e-03,
        -1.5552e-03, -2.3609e-03,  2.0730e-03,  8.2402e-04, -3.9794e-04,
        -2.7871e-03, -4.8762e-04,  5.4436e-04,  3.8085e-04,  3.2103e-04,
        -6.8797e-04, -1.8350e-03,  2.0945e-03,  1.0337e-03, -3.5547e-03,
         3.2945e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3643e-02,  9.3823e-01,  3.1484e-03, -8.0136e-04, -3.2568e-03,
        -5.0991e-03, -1.0727e-03,  7.5881e-03,  3.6494e-04, -1.6833e-03,
         3.7696e-03,  3.3335e-04, -3.2999e-04, -6.3946e-03,  1.0517e-03,
         1.4898e-03,  3.6883e-03, -4.8167e-03,  2.5213e-03, -7.2105e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.2119e-04,  8.8192e-01,  1.1285e-02,  1.0992e-03, -9.7639e-03,
        -2.7220e-03, -1.4618e-03,  6.4372e-03,  4.0333e-03, -1.0764e-02,
         4.0140e-03, -2.5116e-03, -1.5355e-03,  7.1527e-03,  4.2168e-03,
        -2.1806e-03,  1.9110e-03, -1.4960e-03,  9.3231e-04, -2.0318e-02,
        -2.3421e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.1802e-03,  9.6234e-01,  4.5015e-03, -1.3270e-03,  7.5660e-05,
         1.6406e-03,  4.0269e-04, -2.9039e-03,  5.7208e-03,  1.6352e-03,
         7.2167e-04, -3.6890e-03,  6.8594e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3040e-04,  9.1068e-01,  1.7428e-02, -1.8345e-03,  6.5492e-04,
        -1.3509e-03, -4.6744e-04, -5.5965e-04, -1.7670e-03,  1.3417e-03,
        -1.9350e-03, -5.6521e-04, -8.9769e-04,  7.5987e-03,  2.2324e-04,
         1.5087e-03,  9.6434e-03, -2.0236e-03, -1.6936e-04, -3.4401e-04,
        -8.1205e-04, -2.0373e-03, -7.0882e-04, -1.0476e-03,  1.4050e-03,
        -1.4543e-03, -7.9808e-04,  6.6782e-03,  1.8757e-05,  5.3419e-04,
        -1.3099e-03, -2.0251e-03,  7.1450e-03,  5.3607e-04,  4.3368e-04,
        -5.3057e-03,  6.6260e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3792e-04,  9.5654e-01, -2.3698e-03, -1.9997e-03,  2.1284e-03,
         2.3115e-03,  6.1944e-04, -6.6254e-03,  2.0759e-03,  1.9553e-03,
         1.4000e-03,  2.9904e-03,  2.3447e-03,  8.9481e-04, -1.2596e-03,
         2.2249e-04,  1.0268e-03,  1.7794e-03,  1.8233e-03, -1.6412e-03,
         2.9145e-03,  1.7448e-03,  8.9307e-05, -2.3100e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5240e-02, -9.1265e-01, -1.2739e-02, -2.9404e-03, -5.1530e-04,
         5.1296e-04,  1.3558e-03, -1.6017e-03,  2.2492e-03,  3.8793e-03,
         2.6147e-03, -5.3751e-03, -8.6775e-03, -5.5332e-03, -6.3778e-04,
         4.9353e-03, -1.2954e-03,  9.7508e-04,  1.3196e-04,  9.7895e-04,
        -5.4503e-04,  1.2341e-03, -3.9940e-04,  2.4844e-04, -1.7875e-03,
         5.3738e-04,  1.9301e-03,  3.2729e-03,  1.6839e-03, -2.6542e-03,
        -1.6483e-04, -7.0385e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2060e-03,  9.5178e-01,  4.4949e-03, -1.2946e-03, -4.2172e-03,
        -4.3639e-03, -4.8146e-03, -3.2296e-03,  1.5193e-03, -4.9546e-04,
         1.2178e-03, -1.0126e-03, -1.6345e-03, -1.2324e-03, -1.8621e-03,
        -2.2653e-03,  2.4474e-04, -1.0091e-03, -7.6602e-04,  9.6061e-04,
        -3.5283e-03,  1.9986e-04, -2.1900e-03, -1.7185e-03,  1.1633e-04,
        -1.2831e-03, -3.9915e-04, -3.1984e-04, -2.6756e-04, -2.5543e-04,
        -1.0194e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9145e-02,  9.1155e-01,  1.4326e-02,  1.8359e-03,  5.1006e-03,
         2.3376e-03,  4.6764e-03, -5.4178e-03,  2.7903e-03,  3.4095e-03,
         1.5808e-03,  6.6212e-03,  3.3376e-04, -3.0137e-04,  2.6816e-03,
         1.9164e-03,  1.7919e-03,  5.5367e-03,  2.6271e-03,  1.0268e-03,
        -5.5286e-04, -1.8752e-03, -2.5626e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4322e-02,  9.4408e-01, -7.5467e-03, -4.3365e-03, -2.7527e-03,
        -1.8454e-03, -2.2055e-03, -3.7740e-04, -3.1751e-04, -5.9636e-04,
        -3.5446e-04, -1.5484e-03, -5.6202e-04,  4.7374e-04, -5.7674e-04,
        -1.3688e-03, -8.5405e-04, -1.2120e-03,  1.0957e-03,  1.6923e-04,
        -1.8138e-03, -1.0608e-03,  9.7101e-04,  1.9275e-03,  5.1810e-04,
         4.3094e-04, -2.7054e-03,  3.9792e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-2.2967e-04,  9.6180e-01, -5.6592e-03, -8.9195e-06, -1.6324e-03,
        -1.9804e-03, -6.0780e-04,  2.7928e-03, -2.4596e-03, -4.0768e-05,
         4.1833e-04, -1.8879e-04,  2.1142e-03,  3.5157e-04, -2.0968e-03,
         2.6518e-03, -5.8378e-04, -5.5604e-03,  3.5451e-05,  3.1420e-04,
        -2.6013e-04, -1.0292e-03, -4.3050e-04, -1.7010e-03, -7.3021e-04,
        -2.0070e-03, -6.6886e-04, -6.7320e-04, -9.7128e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7719e-03,  9.5142e-01, -4.3624e-03,  1.3213e-03,  1.5536e-03,
        -9.4503e-04, -3.6980e-04, -2.4330e-03,  2.6735e-03,  8.9531e-04,
         2.1204e-03, -3.0614e-03, -1.3952e-03,  4.0207e-03,  5.2203e-05,
        -4.9830e-04,  3.1215e-03, -1.5678e-03,  7.7125e-03, -1.1245e-03,
        -1.0464e-04, -1.5865e-03,  1.0240e-03,  2.8685e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3107e-02,  9.3325e-01,  1.3379e-02,  7.0424e-04, -1.6875e-04,
        -7.4281e-04, -8.9409e-05, -7.3350e-04, -3.9628e-03,  9.2509e-04,
         4.9242e-04, -7.3898e-04,  3.6504e-03,  4.2451e-03, -3.7322e-04,
         1.3362e-03, -8.7552e-04, -5.7767e-04,  8.4981e-04, -9.5874e-04,
        -2.9089e-03, -1.0631e-05,  1.5808e-03,  1.9148e-03,  6.4199e-04,
        -5.2538e-04, -9.2662e-06, -2.9549e-06,  1.3800e-03,  1.3443e-03,
         2.9009e-04, -2.9150e-04, -4.7640e-03, -3.1703e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2429e-02, -9.4500e-01,  4.6919e-03, -3.0180e-04,  3.1710e-03,
         4.3296e-04, -3.4192e-03, -6.1791e-04,  2.2969e-04, -5.1622e-03,
        -2.4977e-03, -3.7639e-03, -1.9510e-03, -3.0464e-04, -3.7235e-03,
         4.8887e-04,  3.2631e-03, -1.6633e-04, -1.7462e-03, -4.9926e-03,
        -1.3953e-03, -2.5143e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0789e-03,  9.6147e-01, -8.7069e-03, -4.6414e-04, -4.2228e-03,
        -3.4446e-03,  2.5990e-03,  5.8756e-03,  3.7731e-04,  1.1674e-04,
         2.1526e-03, -8.1741e-04,  2.3781e-03,  6.0316e-04, -1.5352e-03,
        -1.6838e-04, -2.6001e-03, -1.3935e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7397e-03,  9.5197e-01, -1.5338e-02, -8.4614e-04, -1.1198e-03,
        -4.6113e-04,  6.8682e-04, -1.9925e-04, -4.7773e-03,  2.7549e-03,
         1.0375e-04, -4.8396e-03, -1.1794e-03,  1.8744e-03, -1.8307e-04,
        -1.4325e-05,  5.9684e-04, -3.0202e-03,  9.8250e-04, -3.4076e-04,
         5.4389e-04,  1.8254e-03,  4.5990e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7085e-03, -9.5967e-01,  5.1052e-03, -5.7600e-03,  1.2399e-03,
        -1.8131e-03, -2.2326e-03,  9.4272e-04,  1.4563e-03,  5.3670e-03,
        -2.4477e-03, -3.8913e-04,  4.5510e-04,  1.4733e-03, -2.1963e-04,
        -4.7284e-04,  1.2463e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4170e-02,  9.2855e-01,  1.5242e-02, -3.2493e-03, -4.6889e-03,
         4.6557e-03,  5.7452e-03, -1.3458e-03,  3.5232e-03,  8.2003e-04,
         9.6399e-04, -1.7329e-03, -3.0597e-03, -6.9570e-04,  6.6545e-03,
        -6.3042e-04,  3.3503e-03,  9.1900e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3578e-04,  9.1766e-01,  1.0425e-03, -2.1758e-03,  3.4062e-03,
        -1.3051e-03,  1.0846e-03, -5.2264e-03, -1.4656e-03,  2.5568e-02,
        -1.9781e-03, -6.1144e-04, -7.0914e-03,  1.7633e-04,  3.8731e-03,
         4.3395e-03, -1.9548e-03,  5.9190e-03,  3.0119e-04,  2.9332e-03,
         1.4143e-03,  1.6760e-03,  2.9674e-03,  2.6095e-03,  1.5263e-03,
         1.0584e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4256e-02,  9.1013e-01,  1.0984e-02, -2.2628e-03, -6.3005e-03,
        -3.6008e-03, -9.7303e-03,  2.7546e-03,  1.4722e-03, -2.5829e-03,
        -1.3813e-03,  2.6231e-03, -3.1997e-03, -5.9852e-03, -5.3088e-03,
         1.4788e-03, -9.7480e-04,  1.3679e-03, -9.5887e-05,  1.0863e-04,
        -5.6578e-04, -7.8151e-04, -6.4717e-04,  1.6737e-04, -1.6518e-03,
         1.2205e-03, -2.8698e-05,  1.5158e-05, -1.5965e-03,  5.6076e-03,
        -1.1236e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3191e-03,  9.3551e-01,  3.0259e-03,  3.0940e-03, -4.9872e-03,
        -5.0945e-03, -6.6481e-04, -3.9453e-03, -1.6659e-02, -2.4836e-03,
        -3.4880e-03, -1.0362e-03,  1.4847e-03, -1.1159e-03, -6.2053e-03,
        -2.6266e-04, -6.5401e-03, -3.0796e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0153,  0.9521, -0.0016,  0.0107, -0.0048, -0.0013, -0.0075, -0.0066,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 2.9226e-03,  9.4457e-01,  3.0583e-03, -3.9846e-03,  5.2029e-04,
        -4.8832e-04, -3.4187e-03, -2.3921e-03,  3.1548e-03, -7.3432e-04,
         8.1736e-04,  9.4171e-04, -7.4210e-04, -1.5788e-03, -2.6082e-04,
         4.9083e-04,  8.5258e-04, -2.3570e-03,  4.8326e-04, -6.2125e-04,
         1.5528e-03,  8.3159e-04, -3.7406e-03,  7.2508e-05,  9.4510e-04,
        -9.1638e-04,  4.4245e-04,  2.2736e-03, -2.2607e-03,  1.9068e-03,
        -2.6690e-03, -1.7551e-03,  2.7411e-04,  7.4855e-04, -2.1353e-05,
        -2.9720e-04,  5.0939e-04,  1.7356e-03,  2.6616e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9605e-03,  9.6656e-01, -7.5843e-04, -5.8420e-04, -1.5418e-03,
         3.9074e-04, -1.5668e-03,  5.5819e-04,  6.2275e-04, -1.6002e-03,
        -7.9409e-04,  4.7106e-04,  1.7890e-04, -7.4131e-04,  1.2768e-03,
        -5.9267e-03, -5.8411e-04,  7.1700e-05,  3.5485e-04,  2.2075e-04,
         1.6466e-03,  1.1104e-03, -5.1702e-04,  4.4598e-04, -9.6110e-04,
         1.1766e-03, -4.1307e-03, -2.2496e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6328e-02,  9.3941e-01,  2.8131e-03,  9.7034e-03,  3.1480e-03,
         2.7055e-03,  5.7365e-04,  1.2554e-03,  3.2168e-03,  4.5380e-03,
         1.6620e-03,  3.1467e-03,  1.1497e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0185e-05, -9.6212e-01,  8.8687e-03, -8.3623e-04,  3.1591e-03,
         1.1267e-03,  3.1228e-03,  1.9446e-03,  3.4359e-03, -4.3556e-04,
        -2.1126e-04,  4.5811e-03,  4.8924e-03,  4.3785e-03, -8.8136e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0115e-02,  9.3125e-01,  8.2064e-03, -3.3790e-03,  5.2903e-03,
         4.4664e-04, -4.2736e-04, -7.8363e-05,  6.9659e-04, -4.6626e-04,
        -1.4657e-03, -6.0502e-04,  1.1954e-03,  7.7676e-04,  2.1301e-04,
        -1.3819e-03,  3.1571e-04, -4.4992e-04,  9.1317e-04, -1.1106e-03,
         9.9382e-04,  2.9605e-04,  6.0304e-03, -4.9429e-04,  1.4388e-03,
         7.2912e-04, -6.3111e-04,  2.4923e-03, -7.1074e-04,  1.1230e-04,
        -7.7746e-04,  3.5810e-04, -2.2824e-03, -5.7921e-05,  7.2558e-04,
         5.6567e-04, -2.2085e-04, -2.9246e-04,  1.1692e-03,  3.3958e-04,
         5.3598e-03, -1.0787e-04,  2.8397e-04, -7.3810e-04, -9.0700e-04,
         3.0894e-03, -9.2802e-06], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.7338e-03,  9.3348e-01,  8.2363e-04,  2.5915e-03,  2.3724e-03,
         1.1948e-03, -8.3376e-04, -1.9118e-03,  4.2152e-03,  5.5285e-04,
        -1.3519e-03,  6.1311e-05, -2.4752e-03, -2.0194e-03,  8.1147e-04,
         1.4881e-03,  3.8499e-03, -9.0912e-04, -5.9265e-04,  1.4547e-03,
         8.6782e-04,  9.3233e-04,  8.9493e-04,  1.7112e-04, -9.3820e-04,
        -3.9336e-04,  3.1101e-03,  9.3600e-05, -3.3479e-04, -7.4495e-04,
        -1.8348e-03, -3.9795e-04,  2.0089e-03, -1.9781e-03, -8.2736e-05,
        -9.1360e-04,  4.3617e-04, -1.3631e-03, -1.0695e-03,  4.5781e-04,
        -2.2991e-03,  8.9508e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.6799e-03,  9.3514e-01,  8.9011e-04, -5.9108e-03, -3.0663e-03,
         2.8761e-04,  3.9147e-03, -3.0107e-03, -2.3670e-04, -3.7526e-03,
        -7.7923e-03, -1.9814e-03, -2.4998e-03,  8.1034e-04,  3.1637e-03,
        -3.6010e-03,  3.3718e-03, -2.5115e-03, -5.7123e-04, -1.5040e-03,
        -1.4975e-03, -1.0002e-03, -3.3559e-03,  4.3970e-04,  2.0094e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1572e-02,  9.5728e-01, -8.2869e-04, -9.0454e-04, -3.7563e-03,
        -1.4613e-03,  3.0936e-03, -3.6250e-04,  6.7934e-04,  1.6501e-03,
         7.8954e-04, -1.5888e-03,  1.3062e-03, -8.5990e-04,  5.1986e-04,
         2.2945e-03, -2.3172e-05,  4.9992e-03,  3.4610e-04, -3.6915e-04,
        -9.9851e-04,  8.5161e-04, -3.4642e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0766e-02,  9.4313e-01,  1.5978e-02,  2.5451e-03,  5.8863e-04,
        -3.8849e-03, -2.4603e-03, -2.2165e-03,  2.2659e-04, -1.5464e-03,
        -8.3211e-05, -9.2160e-04,  3.9348e-03,  1.0938e-03, -1.1330e-03,
        -8.0388e-04, -2.8318e-04,  1.4262e-03,  2.0636e-03,  1.6307e-03,
        -1.8796e-03,  1.1489e-03, -2.5493e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0694e-02,  9.4723e-01,  6.2779e-03,  1.5010e-03, -7.9598e-03,
         3.8640e-03,  1.6260e-03, -2.1309e-03,  4.5393e-03, -2.0073e-03,
        -1.2395e-03, -1.0093e-02, -8.3604e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6672e-02,  9.4958e-01, -4.1162e-03,  3.6674e-03, -3.8593e-03,
        -7.7788e-04,  2.0994e-03,  2.1357e-03,  7.9194e-03,  2.2337e-03,
        -1.2991e-03, -1.8265e-03,  3.5164e-03, -2.9906e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.9214e-03,  9.5682e-01, -1.4315e-03,  1.9412e-03, -3.1591e-03,
        -5.6360e-04,  4.5152e-04, -9.7824e-04,  3.4492e-03,  2.0156e-03,
         9.7823e-05,  2.7926e-03,  1.5621e-05,  6.5705e-03,  2.8077e-04,
         1.1646e-03,  1.2332e-03,  2.3230e-03, -5.7883e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-6.7971e-04,  9.5507e-01,  1.4913e-03, -7.4242e-03, -5.5601e-03,
        -6.5667e-04, -7.7511e-04,  9.7449e-04,  4.6011e-04, -4.9235e-03,
         1.3147e-03, -1.8819e-03,  4.2005e-04,  6.5159e-04, -1.2843e-03,
        -3.0899e-03,  2.0272e-03,  9.1753e-04,  4.1385e-04, -1.0700e-03,
         3.0338e-04,  8.1787e-04, -2.9301e-03,  2.4828e-03,  6.6093e-04,
         1.7162e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7386e-02,  9.3247e-01,  1.6176e-02, -1.7367e-03,  4.1939e-04,
        -2.8556e-03, -3.9761e-04, -6.0999e-03,  7.4241e-04,  3.1431e-03,
         2.7078e-03, -2.3621e-03,  3.3639e-03, -1.8023e-03, -3.1604e-03,
        -5.1726e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7593e-02,  9.0091e-01, -1.3926e-03,  9.3103e-03,  2.0535e-03,
        -1.4386e-03,  4.0062e-03, -6.0734e-03,  6.1032e-04,  6.8785e-03,
        -3.6648e-05, -5.1027e-03,  1.0942e-04,  6.7344e-03,  7.0344e-03,
         1.3573e-03, -9.3579e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3465e-02,  9.3200e-01,  9.4349e-03, -6.0494e-04,  8.4510e-03,
         6.0966e-03, -1.5208e-03, -5.7334e-03,  2.6915e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.6670e-03,  9.2748e-01, -3.8446e-03, -3.2210e-03, -7.3604e-03,
        -1.0115e-03,  1.6227e-03,  1.5878e-03, -1.2071e-04, -7.2576e-05,
         3.3201e-04,  1.0116e-02, -7.3960e-04, -1.6053e-03,  2.2166e-04,
         1.3227e-04,  3.5010e-03,  1.4291e-03, -1.1016e-03, -9.9173e-05,
        -6.5307e-04,  1.3689e-03,  1.9535e-03, -1.6976e-03, -1.9202e-04,
         1.6732e-04,  2.6874e-04, -4.5610e-04,  9.2523e-04,  4.2497e-04,
         1.2334e-03, -2.7819e-04, -1.8795e-04, -3.8257e-04,  1.0546e-05,
        -1.0160e-03, -1.3217e-03,  6.1098e-04, -1.4763e-03, -1.8982e-03,
        -8.7493e-04, -1.2341e-03, -2.5320e-03, -7.5123e-04, -1.3195e-03,
        -2.4993e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5128e-02,  9.4458e-01,  7.8492e-03, -2.0558e-04, -4.6797e-04,
        -2.5149e-03,  1.1716e-04, -2.3185e-04, -5.1946e-04, -5.3763e-04,
         4.5092e-04,  1.5019e-04, -2.4468e-03, -9.3592e-04,  3.9676e-04,
         4.5499e-04,  2.6460e-03,  2.5485e-03, -1.1724e-04,  2.7318e-03,
         3.8219e-03,  3.6818e-04,  9.3254e-04, -8.7643e-05,  2.3071e-03,
         5.1829e-04, -6.9635e-04,  4.6169e-04,  3.5671e-03,  2.2068e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7393e-02,  9.4484e-01,  1.4389e-02,  8.6312e-03, -1.2321e-02,
        -2.4145e-03,  1.0120e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2028e-03,  9.4760e-01,  1.3389e-02,  7.9614e-03, -6.3521e-04,
         3.1727e-03, -6.2977e-03,  3.9214e-03,  3.3604e-03, -4.1705e-03,
         1.2913e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5885e-04,  9.4776e-01,  9.4611e-03, -1.8664e-03,  2.0331e-03,
        -7.3734e-03, -1.5570e-03,  4.6301e-05,  2.8628e-03,  2.1836e-04,
        -1.4594e-03,  6.6522e-03, -2.6265e-03, -1.5525e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2667e-02,  9.5732e-01, -2.6583e-04,  2.8798e-04,  2.0394e-04,
         2.2737e-04,  2.1785e-04,  1.0631e-03,  3.1526e-03, -2.9494e-04,
        -2.8355e-03, -2.4056e-04, -1.8595e-03,  6.2272e-04,  8.4344e-05,
         3.4728e-04,  5.5515e-04,  4.1880e-04,  2.9947e-05,  2.5738e-03,
         5.5023e-04, -1.6395e-04, -6.2629e-04, -1.2981e-03, -2.5523e-03,
        -4.3500e-04,  1.2429e-03,  1.4560e-04,  1.0309e-03,  7.2776e-04,
         1.0967e-03, -1.0380e-03,  7.5944e-04,  3.0627e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0160e-02,  9.1198e-01, -5.8181e-03,  5.2643e-04, -5.8036e-04,
         1.6284e-03, -3.3183e-03, -3.2692e-03,  7.0074e-04, -1.5427e-03,
        -3.0801e-03,  2.6110e-03, -7.7841e-04, -2.6916e-03,  1.0563e-02,
         2.1703e-03, -1.4786e-02, -4.5501e-03, -3.2846e-03, -2.4033e-03,
        -2.6576e-03, -1.6701e-03,  4.9898e-03, -1.2212e-03, -1.4469e-03,
         4.3442e-04, -8.6838e-04, -2.6865e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0384e-02,  9.5471e-01,  1.3137e-03,  6.1238e-04, -2.0447e-03,
        -7.9543e-04, -5.1532e-04,  3.1442e-04, -7.5552e-04,  5.6993e-04,
        -7.8013e-04, -1.4243e-03,  2.0325e-03, -1.0396e-03, -8.6664e-03,
        -3.6407e-03,  1.3062e-03,  1.8157e-03, -4.5517e-04, -6.4879e-04,
        -5.8940e-04,  1.1695e-03, -3.2154e-03, -1.1966e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-2.2930e-02,  9.0927e-01, -1.2753e-02, -5.8729e-03, -3.5481e-03,
         1.2842e-03, -5.3076e-03,  1.6033e-04,  1.2054e-04,  8.8395e-04,
        -2.0431e-03, -1.6804e-03,  3.0484e-03,  2.2510e-04,  3.8056e-04,
         1.9658e-03, -2.8445e-03,  1.2630e-03, -3.4174e-03, -2.0714e-03,
         7.6869e-05, -9.8464e-04, -1.6271e-03,  1.7057e-03, -1.1628e-03,
        -3.9196e-03,  6.1053e-04, -3.2990e-03, -2.1846e-03, -2.6089e-03,
        -7.4974e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.6771e-03,  9.5074e-01, -1.1286e-02,  5.4183e-03, -1.3103e-03,
         1.1201e-02,  6.0487e-04,  1.6899e-03,  1.8398e-03,  3.2538e-03,
         3.0170e-03, -3.9664e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8571e-02,  9.2356e-01,  1.8554e-03,  6.1876e-03,  5.9049e-03,
        -3.2495e-03, -2.0339e-03, -4.2683e-03, -3.9187e-03,  1.2273e-02,
         2.2351e-03,  2.8829e-03, -2.2441e-04,  2.9486e-03,  9.8821e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5096e-03,  9.5845e-01, -5.7060e-04, -2.8393e-03, -6.9715e-04,
        -2.4643e-03,  7.0433e-04, -1.3309e-03, -1.5789e-04, -3.4574e-04,
         3.2311e-04,  2.3093e-04,  2.7685e-03,  8.0329e-04,  7.0249e-05,
        -2.4053e-03,  4.9021e-04, -2.0433e-03,  6.0668e-03,  1.2309e-03,
         1.2101e-04,  3.5571e-04, -1.9566e-03, -4.4563e-04,  2.1118e-03,
         5.8696e-04,  4.1872e-04, -1.1336e-03, -1.0941e-03, -1.2917e-03,
         1.8965e-03, -6.8241e-04,  1.4033e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1252e-02,  9.5885e-01,  1.2692e-03, -8.9464e-04, -3.9604e-03,
        -6.3208e-03,  1.9638e-03, -8.0519e-04,  1.5923e-03,  4.0290e-04,
        -2.8664e-03, -1.0525e-03, -1.4819e-03, -2.6871e-03, -2.4715e-03,
         2.1274e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([6.2719e-03, 9.5807e-01, 6.3311e-03, 7.3483e-03, 6.7656e-03, 1.6145e-03,
        1.2404e-02, 7.9499e-04, 3.9502e-04, 0.0000e+00, 0.0000e+00, 0.0000e+00,
        0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00,
        0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00,
        0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00,
        0.0000e+00, 0.0000e+00, 0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7352e-02,  9.7547e-01, -4.1006e-03, -2.3967e-03,  6.7743e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0956e-02,  9.3785e-01,  2.3558e-03,  8.2143e-04, -6.5467e-04,
        -4.8076e-04,  9.2957e-04,  6.1820e-04, -9.4442e-05,  6.1128e-03,
         6.9255e-03, -1.7872e-03,  6.3070e-04,  8.3349e-05,  8.1135e-04,
         9.6797e-04, -7.3061e-04, -1.9930e-03,  4.5927e-04,  3.7175e-03,
        -1.1392e-03,  2.6412e-03, -1.3090e-03,  1.0481e-03,  7.8503e-04,
         3.0387e-05,  4.0717e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0253,  0.9404, -0.0070,  0.0048,  0.0066,  0.0087,  0.0073,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8314e-03,  9.5311e-01, -2.0602e-03,  2.5040e-03,  1.7938e-03,
         1.2995e-03, -8.1010e-04,  6.9460e-04,  1.4018e-04,  3.8416e-03,
         2.2455e-03, -4.6365e-03, -8.3967e-04,  1.3654e-03, -1.5282e-03,
        -5.7938e-05,  5.4743e-03,  2.8274e-03,  6.3810e-04, -1.0812e-04,
         1.5660e-03,  9.6288e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0861e-03,  9.4955e-01, -1.4947e-02, -3.4826e-03,  2.0342e-03,
         9.0506e-04,  1.2166e-03, -4.7977e-04, -3.5251e-05, -2.2899e-03,
        -4.4596e-03,  2.6567e-03, -4.0806e-03,  5.3692e-03,  6.0104e-04,
         6.4749e-04,  3.8269e-03,  7.0896e-04, -6.1945e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2921e-02,  9.3840e-01, -6.5667e-03,  1.0855e-03,  2.2350e-03,
         4.6658e-03,  3.5875e-03, -2.6536e-03, -2.4989e-04,  1.2753e-03,
        -1.1192e-03,  1.1187e-03,  4.0181e-04,  1.4384e-03,  4.0325e-04,
        -1.2773e-04, -9.9074e-04,  5.7592e-04,  1.6575e-05,  3.2988e-04,
        -1.9988e-04,  6.0174e-04,  2.8969e-04,  2.2187e-03, -1.2691e-03,
         7.9187e-04, -1.0029e-03, -2.1028e-04,  1.5215e-03, -8.1578e-03,
         3.5744e-03,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-5.8738e-03,  9.4744e-01, -2.7023e-04, -2.4976e-04, -1.1199e-03,
        -1.8412e-03,  1.2600e-03,  2.3345e-04,  1.3011e-03, -8.7192e-04,
        -1.4050e-03, -1.0723e-03, -1.6702e-03, -1.2920e-03, -4.0015e-03,
        -2.0959e-03, -2.9128e-03,  2.4672e-04, -1.3681e-03, -6.8719e-04,
        -1.1003e-03,  3.6586e-05,  1.5927e-03, -3.2761e-03,  4.5518e-04,
        -1.1491e-03, -3.3329e-03, -7.8329e-04,  4.0238e-05, -1.9403e-03,
        -3.5870e-04, -8.7222e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3543e-04,  9.6291e-01, -4.9586e-03, -1.7302e-03,  2.8576e-03,
         7.5179e-04,  1.8380e-03,  3.2027e-03,  1.7923e-03,  6.3299e-04,
        -5.8329e-04,  4.3648e-03, -1.9899e-03,  2.1766e-03,  7.6849e-04,
         3.7910e-04,  9.9608e-04,  2.3376e-05,  1.7169e-03,  1.6892e-03,
        -3.5381e-03, -7.6805e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.3742e-04,  9.4644e-01,  2.2196e-03,  1.8256e-05, -3.1477e-04,
         2.9673e-03,  2.0459e-03,  5.1140e-04,  1.6705e-03, -1.7425e-04,
        -3.1095e-04, -2.0962e-03,  2.4857e-03,  1.4550e-04,  2.5030e-04,
        -5.7511e-03, -2.6141e-03,  4.8711e-03, -3.5074e-03,  2.1096e-03,
         5.5014e-04, -2.4779e-03, -7.1179e-04, -2.9792e-03, -9.8829e-04,
         1.0312e-03,  1.4626e-03, -3.5628e-03,  8.1517e-04,  4.1780e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4376e-03,  9.5110e-01, -4.0226e-03,  4.1593e-04, -9.9512e-04,
         1.4193e-03, -4.5639e-03, -1.6192e-03,  1.0132e-02, -4.1715e-03,
         2.5849e-03, -7.0690e-04, -9.2353e-04, -6.2058e-04, -1.4884e-03,
        -2.0005e-03, -2.0729e-03, -8.7226e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6006e-03,  9.2308e-01,  2.1150e-02, -1.5288e-03, -4.2494e-04,
        -2.7881e-03, -7.5534e-04,  1.2687e-03, -2.3235e-03,  2.3342e-04,
        -4.4085e-04, -1.2879e-04, -6.7680e-04,  4.9031e-04,  1.2917e-03,
         2.0259e-03,  2.3202e-03,  3.8322e-04, -8.5889e-06,  1.6667e-05,
        -2.3663e-04, -4.7467e-03,  4.0281e-03, -1.7462e-03,  3.0570e-03,
        -1.1733e-03,  4.3092e-03,  3.8314e-04,  3.2925e-04,  6.0172e-04,
        -5.5063e-05,  5.1046e-04,  1.3109e-03, -2.1338e-03,  5.6800e-04,
         1.8574e-03,  7.0212e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1997e-02,  9.5351e-01, -5.9563e-03,  2.6182e-03,  6.0076e-04,
         2.1995e-03,  9.5428e-04, -1.3099e-05, -1.2695e-03, -2.3428e-03,
         1.4197e-04, -1.3413e-03,  3.9277e-04,  2.2547e-03,  1.4167e-03,
         6.9734e-04, -3.2786e-03, -1.2859e-03, -6.1072e-04,  2.5849e-04,
        -1.1084e-03, -1.8364e-03,  3.3580e-03, -5.5837e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1342e-03,  9.5035e-01, -2.8724e-03, -2.7810e-03, -8.3920e-04,
         2.7447e-04,  1.3722e-04, -4.1234e-04,  7.9939e-04,  4.9618e-04,
         9.6056e-04,  8.0620e-05,  6.7358e-04, -9.0531e-04,  2.1875e-04,
        -1.1164e-03,  1.0589e-03, -5.7911e-04, -9.5432e-04, -1.0586e-03,
         8.5296e-04, -1.4497e-03, -6.2509e-04, -1.5128e-03, -1.3578e-03,
        -4.7915e-04,  6.8811e-04, -3.4696e-04, -2.0033e-04, -9.2143e-04,
         3.5203e-03, -1.4698e-03, -4.6233e-05,  7.5569e-04, -5.6845e-04,
        -1.9017e-03, -1.3547e-03,  4.9481e-04, -2.6682e-04, -2.8851e-03,
         5.5748e-04, -6.3449e-04,  4.5986e-04,  1.4205e-03, -9.7906e-04,
         3.5472e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3711e-03,  9.5910e-01, -2.8930e-03, -1.9551e-03, -2.3750e-04,
         2.1323e-03, -2.0884e-03,  1.6594e-05,  5.6891e-04, -7.6600e-04,
         1.2179e-03, -2.8985e-04, -1.0110e-03,  1.5784e-03,  1.8828e-03,
        -2.2125e-03,  1.6368e-03, -1.7717e-03,  1.2415e-03, -4.2933e-03,
         3.1425e-04,  6.5821e-04,  2.7615e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.6241e-03,  9.4530e-01,  1.3898e-02, -2.2339e-03, -7.8037e-04,
         1.9068e-03,  4.8829e-03,  1.3006e-03,  3.1967e-03,  4.6280e-03,
         1.4247e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0123,  0.9451, -0.0110,  0.0053,  0.0040, -0.0034, -0.0025, -0.0012,
         0.0044,  0.0020,  0.0030,  0.0017,  0.0024, -0.0017,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8846e-02,  9.5466e-01, -2.5226e-03, -2.3063e-03, -1.8476e-03,
         5.7278e-03,  1.5263e-03,  3.8527e-03,  2.9983e-03, -1.2081e-04,
         3.2787e-03, -2.3145e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3017e-03,  9.3261e-01, -1.3750e-03,  4.4764e-03,  1.7235e-03,
         2.5236e-03, -1.8979e-04,  1.4061e-03,  6.6376e-04,  1.6766e-03,
        -4.3566e-05,  1.4480e-03,  1.0572e-03,  1.7778e-03, -4.9013e-04,
         5.5689e-04,  1.1402e-03, -7.0888e-04,  2.3594e-03, -1.5236e-03,
        -2.6489e-04,  1.5626e-03,  3.6198e-03, -2.4326e-04,  3.9394e-04,
         4.1642e-04,  4.8922e-03, -6.2178e-04, -1.8031e-04,  1.8823e-03,
        -1.5409e-04,  3.9193e-03, -2.0085e-03,  8.2050e-04,  1.6823e-03,
        -1.1787e-03, -9.7424e-04,  9.9952e-04, -2.7983e-04, -1.3097e-04,
        -5.0826e-04,  4.3682e-04, -1.8147e-04, -5.5668e-04,  7.0498e-04,
        -9.0657e-04,  7.5259e-04, -1.5897e-04,  1.3212e-04,  1.3160e-04,
         6.0973e-04, -2.5009e-04, -1.2364e-03, -2.1140e-03, -2.0426e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-1.8335e-02,  9.4430e-01, -2.5701e-03, -1.9700e-03,  1.4432e-03,
         1.5009e-03, -1.5977e-03,  1.6405e-03,  2.7981e-04, -9.1302e-03,
        -5.2984e-03,  3.7216e-04, -2.9596e-04, -4.0131e-04,  2.7575e-03,
         2.2177e-04, -9.4617e-04,  1.2797e-03,  1.8071e-03,  4.9486e-04,
        -5.5840e-04,  2.2705e-03,  5.3132e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1457e-03,  9.7697e-01,  8.5142e-04,  3.1323e-05,  7.0366e-04,
         1.3726e-03, -4.8033e-04, -4.6147e-04, -8.0429e-04, -4.0273e-04,
        -1.3397e-03,  1.5451e-04, -1.2315e-04,  1.4606e-03,  1.3006e-03,
        -9.8126e-04, -2.0275e-03, -7.3886e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.7994e-03,  9.1724e-01,  9.1995e-03,  1.1403e-02,  8.4963e-03,
         4.6147e-04,  1.3050e-03,  4.4939e-03,  9.6755e-04, -2.7865e-03,
        -6.2314e-03,  2.0642e-03, -3.9940e-04, -2.0566e-03,  3.0845e-03,
         3.3291e-03,  6.3650e-03, -1.4318e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.6853e-04,  9.5204e-01, -2.0472e-03,  2.8697e-03, -8.0606e-03,
        -3.3362e-03,  5.2820e-03, -5.2634e-03,  1.5321e-03,  2.8182e-03,
        -9.4094e-04,  4.4409e-03,  5.3375e-03, -1.1820e-04,  5.3405e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1440e-02,  9.4116e-01, -1.1609e-03, -1.3000e-03,  2.6594e-03,
        -3.7402e-03,  5.3879e-04, -1.3188e-03, -5.7135e-04,  5.6014e-04,
         1.1552e-03, -6.7159e-04,  1.8312e-03, -6.8045e-05, -5.1688e-03,
        -4.9086e-04, -1.2805e-04, -1.5613e-03,  2.9204e-04,  1.8581e-03,
         2.3321e-03, -1.7807e-03, -6.4595e-04, -1.3995e-03, -4.4040e-03,
        -8.8832e-04,  6.2284e-04, -5.2802e-04, -4.0720e-03,  2.4520e-03,
        -1.4549e-03, -1.7475e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1831e-03,  9.4485e-01, -8.8439e-03,  2.4875e-03,  1.9827e-03,
        -1.0045e-04, -4.9073e-04,  8.4416e-04, -1.0003e-03, -2.5152e-03,
         1.7414e-03, -1.0728e-03, -2.3573e-04,  2.8214e-04, -1.6587e-03,
         6.4268e-04, -7.4161e-04,  1.7633e-03,  3.1603e-03, -1.5714e-03,
         1.4907e-04,  3.2743e-03, -4.1488e-03, -1.0295e-03, -6.8079e-04,
        -4.7616e-04,  3.8454e-04,  1.0118e-03,  2.1046e-03,  8.5091e-04,
         6.6710e-04,  5.5731e-03,  4.7684e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.2280e-03,  9.4136e-01, -6.2974e-04,  8.3829e-03,  1.0132e-03,
        -2.7862e-03,  9.9174e-04,  6.2517e-04,  2.3544e-03,  1.1347e-03,
         5.7897e-04,  2.1675e-04,  9.8845e-04,  8.1756e-04,  6.8296e-04,
        -1.0283e-03,  5.0616e-04, -1.0713e-03,  3.0771e-04, -4.7529e-04,
         4.5128e-04, -8.3181e-04,  3.0794e-04,  1.3652e-03,  3.7374e-04,
        -5.2699e-05,  8.0561e-04, -3.7146e-04,  1.6113e-04, -8.1680e-04,
         9.2810e-04,  2.4216e-04,  2.1279e-04, -3.3151e-04,  4.4204e-04,
        -6.1279e-04, -5.4565e-04,  5.5139e-04,  7.0513e-04,  2.8302e-04,
         4.8383e-04, -6.0243e-04,  2.2171e-04,  1.4505e-03, -2.5796e-04,
         1.1998e-03,  1.2059e-04,  9.5642e-07, -2.1983e-04,  6.1654e-05,
        -2.5104e-04,  5.7346e-04,  9.6995e-04,  1.1592e-04,  1.1768e-03,
         2.3587e-03,  8.7262e-04,  4.4932e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8381e-02,  9.3319e-01, -3.3082e-03, -1.8654e-03, -2.3781e-04,
        -1.1568e-03, -1.1067e-03, -8.5305e-04, -7.8298e-04, -1.2943e-03,
         2.0842e-04,  6.3167e-03,  4.7266e-04,  2.3326e-03, -3.5624e-03,
        -8.0757e-04,  1.2817e-03, -1.0157e-04, -4.9503e-04, -1.4645e-03,
        -8.3039e-04, -1.1089e-03, -2.2694e-03, -2.6151e-03, -7.2926e-03,
        -6.6672e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7658e-03,  9.4832e-01, -7.0404e-03, -2.2148e-03, -9.5709e-04,
         2.8119e-03,  7.6401e-04,  3.0221e-03,  2.7724e-03, -1.0054e-04,
         3.0729e-03,  7.4249e-04,  4.0113e-04,  1.3497e-03, -1.6056e-04,
         1.7112e-03,  2.0693e-04,  1.4285e-04, -4.9845e-04, -6.4821e-04,
         6.4709e-04, -2.9015e-03,  3.4570e-03, -4.8646e-03, -6.4267e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3932e-03,  9.3345e-01, -1.3342e-02, -3.4251e-03,  3.1159e-03,
         2.4016e-04, -1.1021e-03, -9.0883e-04, -3.8642e-04, -8.1148e-04,
         1.0479e-03,  4.4932e-04, -1.2237e-03,  2.4330e-03,  1.1467e-03,
        -1.8000e-03,  2.0449e-04,  2.5640e-04, -6.4897e-05,  1.7654e-03,
         1.3171e-04,  1.2983e-04,  1.4614e-04, -4.7070e-04,  7.0458e-04,
        -2.1169e-03,  1.7771e-03, -1.6311e-04, -3.3961e-04, -4.8401e-04,
         1.3905e-03, -5.6390e-04,  2.3876e-03, -4.4171e-04,  2.3647e-03,
        -2.7362e-03,  2.4148e-04, -1.1210e-03,  8.0972e-04,  2.2958e-03,
        -9.0875e-04,  5.8280e-04, -5.2355e-04,  1.5057e-03,  7.0930e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.6313e-03,  9.3929e-01,  8.5943e-03, -4.0339e-03,  4.7066e-04,
        -6.3545e-03, -3.3567e-03,  9.6048e-05,  1.1277e-04, -5.9465e-03,
        -6.2621e-03, -6.7621e-04,  9.3432e-04,  3.6228e-04, -1.6969e-03,
         2.2636e-03,  2.7153e-03, -1.1630e-04, -3.0944e-04, -9.8425e-04,
        -2.2119e-03, -7.8863e-04,  2.7911e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.9656e-03,  9.3517e-01,  1.3206e-02,  2.4639e-03,  3.4267e-04,
        -6.1290e-03, -1.0935e-02, -1.9028e-03, -3.3489e-03, -1.0461e-03,
         9.1025e-03,  8.3917e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([-9.3472e-03,  9.5002e-01, -7.2166e-03, -1.5470e-03, -3.8653e-04,
        -1.8416e-03, -5.4171e-03, -3.0306e-04,  1.4764e-04, -2.2443e-03,
         1.4710e-03, -2.0951e-03, -4.9137e-03,  2.5499e-04, -1.4523e-03,
        -2.6488e-03, -2.4968e-03,  2.1856e-03, -7.9215e-05,  7.1094e-04,
        -9.5325e-04, -4.3945e-04, -1.8232e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.3154e-03,  9.6275e-01, -1.2040e-04, -9.1833e-04, -5.9547e-04,
        -4.5236e-03, -1.2259e-03,  6.9906e-04, -1.8275e-03,  2.0789e-03,
        -2.5238e-04, -3.3265e-04, -1.2976e-04,  9.5338e-04,  3.6105e-04,
         9.7128e-05,  2.1141e-03, -1.3174e-03, -1.3467e-03, -3.4740e-04,
        -5.7830e-04, -2.3604e-03, -1.1227e-04, -1.5649e-03, -8.9257e-05,
        -3.0949e-04, -1.6012e-03,  6.0530e-04, -8.1828e-04, -5.4825e-04,
         9.1376e-05,  2.4678e-03, -5.4281e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8069e-03,  9.6344e-01,  1.3787e-03, -7.7310e-03, -2.6419e-03,
        -1.9021e-03, -3.7315e-05, -4.8244e-04,  2.7228e-04, -2.3100e-03,
        -1.6742e-03, -1.3622e-03,  7.1665e-04, -2.7657e-03,  2.4941e-03,
         1.2714e-03,  3.1845e-03, -2.0785e-03, -2.4506e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3198e-02,  9.5104e-01,  4.1135e-03, -4.7951e-03, -4.4797e-03,
         7.1451e-04,  2.1697e-04, -2.6417e-03, -1.4476e-03,  1.3954e-03,
        -4.2121e-03, -4.1705e-04,  1.6636e-04,  1.1163e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.6535e-03,  9.6548e-01, -1.1862e-02, -4.8561e-05,  1.2259e-03,
        -2.7103e-03, -1.6635e-04,  6.3068e-04, -9.4372e-04,  2.6077e-04,
        -5.0721e-04, -2.6692e-04,  1.1335e-03, -2.1927e-03,  8.9395e-04,
         5.3501e-04,  1.6017e-04,  8.7018e-04, -9.2179e-04, -1.2200e-03,
         1.3353e-03,  9.8361e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4819e-03,  9.4249e-01,  8.7596e-03, -8.6707e-04, -2.0958e-04,
        -1.1407e-04,  4.1785e-03, -2.9878e-05, -1.8812e-03,  7.4269e-04,
         8.0305e-04, -1.5863e-04, -3.3536e-03, -1.7257e-04, -4.7863e-04,
        -5.4297e-03, -1.3023e-03,  1.4541e-03, -3.8613e-04,  6.6108e-03,
         1.3616e-04, -6.8034e-04,  1.2709e-03, -9.1463e-05, -4.4763e-04,
        -1.4896e-03,  3.6745e-04, -4.7128e-03, -2.7239e-03,  7.1730e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0702e-03,  9.6011e-01, -2.9251e-03, -2.7359e-03, -3.5299e-04,
         3.2267e-03, -4.0796e-03, -2.6450e-03,  1.5363e-03, -8.2941e-04,
         5.1709e-04,  3.5930e-03, -4.9831e-04, -1.8376e-03, -1.7392e-03,
        -1.1008e-03, -6.2037e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.6900e-03,  9.6258e-01,  1.0398e-02,  2.1045e-03, -5.0908e-03,
        -5.7033e-04, -1.4176e-03,  4.3972e-03, -2.1785e-03, -1.5080e-03,
         1.0668e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0912e-02,  9.4079e-01,  3.2518e-05, -3.3853e-03, -3.5521e-04,
        -2.5408e-03,  4.1610e-03, -7.6656e-04,  1.0343e-03,  1.2314e-03,
        -4.4277e-03, -4.9352e-04, -3.7719e-03, -5.0944e-03, -2.3043e-03,
        -5.2502e-03, -3.4536e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9583e-03,  9.4081e-01, -1.9955e-03,  6.8719e-03,  1.9724e-03,
        -3.6316e-03,  1.3816e-03, -1.7603e-04, -1.4119e-03,  1.7600e-04,
         1.1961e-03,  2.3315e-03,  2.0529e-03, -4.3550e-03,  1.6495e-03,
         1.4301e-03,  5.3812e-04,  1.1546e-03,  2.0921e-03,  6.8662e-03,
         1.8835e-03, -1.1661e-03,  2.6487e-03,  1.5099e-03, -1.7994e-03,
         2.9384e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1497e-02,  8.7683e-01,  1.3988e-02, -2.9077e-03,  3.0957e-03,
        -1.8085e-03,  4.3780e-03,  3.3213e-03, -2.5723e-03,  5.4544e-03,
        -2.4633e-03,  3.2955e-03,  1.6635e-04, -2.4298e-03,  2.0477e-04,
         1.8550e-03, -8.9077e-04, -5.0110e-03,  1.7932e-03,  1.7010e-04,
        -1.4216e-03, -2.7110e-03, -2.8263e-04, -5.8423e-03,  2.0171e-03,
        -2.4956e-03,  5.2790e-04, -7.1888e-04,  1.3828e-02, -6.0212e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6690e-03,  9.4786e-01,  9.5946e-03,  9.8065e-04, -8.4641e-04,
        -1.7494e-03,  5.5918e-03,  5.0526e-03, -2.6970e-03,  7.9147e-04,
         2.4991e-03,  7.1679e-03,  9.4953e-05,  4.0557e-03, -2.6313e-03,
         8.9644e-04, -5.3090e-04,  1.0649e-03, -3.2228e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 1.1007e-03,  9.4137e-01, -6.5192e-04,  4.1863e-03,  5.2642e-03,
        -2.3088e-04,  5.6156e-03,  3.2360e-03, -3.7800e-03,  1.3280e-03,
         2.8255e-03,  3.8190e-04,  8.7917e-04,  3.4324e-03,  1.0062e-03,
        -8.4702e-04,  3.4125e-03, -4.4312e-03,  1.3837e-03,  8.5711e-04,
        -7.9670e-04,  9.0780e-04, -1.0948e-03,  2.5222e-04, -4.8841e-04,
         3.1095e-03,  7.1311e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0506,  0.7939, -0.0212, -0.0029,  0.0039,  0.0013, -0.0099,  0.0020,
        -0.0157, -0.0029,  0.0022, -0.0110, -0.0064, -0.0762,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.6967e-04,  9.5086e-01, -9.4340e-04, -1.0563e-03, -2.0516e-03,
         4.3741e-03, -1.9032e-03,  5.4490e-04, -3.4040e-03, -1.0635e-03,
         1.2490e-03, -2.9114e-04, -3.6160e-03,  1.2095e-04, -1.8370e-03,
        -3.2051e-03, -2.4626e-03, -5.2777e-03, -9.7486e-04, -1.0539e-03,
        -1.0124e-03, -1.3531e-03, -3.4005e-04, -1.6856e-03, -1.3912e-03,
        -6.9555e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6165e-02,  9.0737e-01,  1.0206e-02,  5.5698e-03,  1.6561e-04,
        -5.7383e-03,  3.9690e-04, -1.0619e-03, -2.7195e-03, -1.5372e-03,
        -1.1430e-03,  7.8054e-03, -9.8997e-03,  2.1865e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1546e-02,  9.3350e-01, -3.7371e-03, -7.6131e-04,  7.9986e-04,
        -3.1119e-04,  1.2703e-03,  1.1101e-03,  2.6595e-03,  1.0117e-05,
        -5.4786e-04,  4.2736e-04,  5.7253e-04, -2.3626e-04,  3.2700e-03,
        -6.5057e-04,  8.0963e-04, -8.3300e-04, -9.4638e-04,  6.6721e-03,
         6.6442e-04,  3.9087e-03, -2.1310e-03, -3.0492e-03,  1.1812e-03,
        -3.1697e-04, -7.2300e-03,  1.4626e-03,  8.7803e-05, -4.0200e-04,
         6.0737e-03,  2.8201e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1031e-02,  9.5358e-01,  8.0580e-04, -3.6217e-03,  1.7821e-03,
        -1.1862e-03, -4.0045e-03, -7.4741e-04,  2.5183e-03, -1.4891e-03,
        -2.6236e-03, -9.9482e-04, -1.5715e-03, -1.5829e-03, -4.7318e-04,
        -4.1087e-03, -9.7809e-04,  9.3510e-04, -1.8927e-04, -2.6203e-04,
         7.1713e-04, -9.5028e-05, -9.2964e-05,  4.3595e-03, -2.5465e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0464e-03,  9.6848e-01, -2.0889e-02, -6.9053e-03,  2.1669e-04,
        -1.4647e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.9145e-03,  9.4100e-01,  2.9151e-03, -4.8953e-03, -9.8770e-04,
        -2.4718e-03,  1.7771e-03, -3.3869e-04, -6.5571e-04,  1.0790e-03,
        -4.2427e-04, -7.5826e-04, -2.8995e-04, -1.2396e-03, -1.8392e-03,
        -8.0379e-04, -2.0614e-03, -1.8559e-03, -2.9340e-04,  3.5050e-04,
        -1.2465e-03,  2.3706e-04, -2.5773e-03, -7.5997e-04, -8.1477e-04,
         2.7272e-03,  5.0670e-04,  2.2976e-03,  1.4632e-03, -2.3356e-03,
         1.6750e-03, -1.7150e-03, -4.3636e-04,  8.4974e-04,  1.2913e-03,
        -1.1072e-03, -2.0120e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2215e-02, -9.1775e-01, -9.6945e-03, -8.8861e-03, -1.5662e-04,
         3.7824e-04, -1.1238e-04,  4.3673e-03,  2.7017e-03, -2.4852e-04,
         4.8533e-04, -1.7927e-05,  1.7003e-03,  2.5429e-03,  5.4696e-03,
         1.8631e-03,  2.7091e-03, -2.7442e-03,  1.6475e-03, -4.1931e-04,
         3.0176e-03, -1.2362e-03,  2.9998e-03,  2.4386e-03, -5.7569e-04,
         1.1904e-03,  2.0122e-03, -5.0634e-04,  1.5070e-03, -2.7752e-03,
         7.9276e-04, -2.1734e-04,  6.3732e-04,  1.7806e-03, -2.1997e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.5464e-03, -9.2124e-01, -1.7634e-02, -1.0281e-02, -3.5595e-03,
         6.3248e-03,  1.1169e-02, -1.6771e-03, -2.2308e-03, -8.2946e-04,
        -1.7288e-03,  2.6341e-03, -6.7956e-03, -3.4420e-03,  9.0638e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.8382e-03,  9.4981e-01, -1.1904e-02, -2.1397e-03,  1.0851e-03,
        -1.4261e-03, -4.6504e-04, -3.3018e-04, -7.3457e-04, -1.4103e-03,
        -6.9292e-04, -1.4260e-04, -3.2818e-04, -8.1994e-04,  4.8264e-04,
         1.6235e-03,  3.6868e-04,  7.2458e-04,  4.0345e-04,  1.4177e-03,
         7.2090e-04, -1.0851e-03, -2.7597e-03,  1.7553e-03,  3.0997e-03,
        -1.5733e-04,  6.7156e-04,  3.4446e-04,  5.4468e-05, -1.4514e-03,
        -4.3282e-05,  9.6606e-05,  3.6162e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6345e-04,  9.6366e-01, -1.9821e-03, -1.6112e-03, -3.8404e-04,
        -3.1526e-03, -4.9770e-05,  3.7527e-03,  5.5822e-03,  1.6198e-03,
         4.5753e-03,  3.9155e-04,  1.4806e-03,  4.6174e-04, -4.6911e-05,
         1.6154e-04,  2.7086e-04, -7.5135e-04, -5.0840e-04,  1.3912e-03,
        -1.5786e-04, -4.1058e-05,  2.1898e-04,  4.8393e-04,  5.6533e-04,
         1.2391e-04,  2.1922e-03,  1.3554e-03,  2.5649e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([-2.5350e-03,  9.7254e-01, -1.1608e-03, -3.1261e-03, -6.0152e-04,
        -1.1170e-03, -4.4518e-05, -2.1932e-03,  8.2543e-04, -3.4097e-03,
        -1.0039e-04,  2.0274e-03,  1.5182e-03, -2.7124e-03,  6.0866e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7173e-04,  9.6707e-01,  1.6986e-03,  1.3461e-03, -4.7455e-04,
         8.5940e-04,  6.5813e-04, -2.4616e-03,  8.2797e-04,  6.4818e-04,
         1.9315e-05,  4.9199e-04, -1.0357e-03, -4.3585e-04, -1.0113e-03,
        -3.2306e-03,  6.9277e-04, -3.5829e-04, -2.6738e-04, -6.4342e-04,
        -6.7840e-04,  1.6641e-04, -1.4012e-03, -2.6152e-04, -7.0883e-04,
        -5.9045e-04, -3.4889e-04,  1.2121e-04, -4.1456e-04, -2.5414e-04,
        -1.1999e-03, -3.4307e-04, -4.4479e-04, -7.1594e-04, -6.5726e-04,
         8.0503e-05,  6.2040e-04, -2.2685e-04,  4.8783e-04, -9.4736e-04,
        -1.5821e-04,  5.2449e-04,  9.8410e-04, -2.0223e-04, -5.9586e-04,
        -1.1695e-03, -6.6847e-04,  1.2901e-04, -1.9185e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.8310e-03, -9.3547e-01, -3.8980e-03, -5.1578e-03, -7.0554e-03,
         1.0963e-03, -2.7595e-03, -3.0863e-04, -2.0200e-03, -2.5528e-03,
         2.3506e-03, -2.5079e-03, -2.7516e-04, -3.1285e-04,  4.0439e-04,
        -2.4492e-04, -2.8408e-03, -3.3891e-04, -2.5186e-04, -1.5723e-03,
         1.3391e-03, -4.0285e-04, -1.0562e-03,  1.0255e-03,  1.4225e-04,
         2.4785e-03,  2.9971e-03, -2.1901e-03, -1.2309e-03,  2.3707e-03,
         5.8497e-04,  3.3111e-04, -1.0650e-03,  3.3529e-03, -3.1860e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.6509e-03,  9.6836e-01, -2.0195e-03, -5.2622e-04, -9.8737e-04,
         3.6666e-03, -4.2792e-03,  1.4951e-03,  2.4097e-04,  6.4276e-04,
        -1.1071e-03,  1.0989e-03, -1.3256e-03,  3.6742e-04, -1.7294e-03,
        -1.2285e-03,  4.0847e-05, -8.3860e-04,  5.3995e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.8373e-04,  9.3810e-01, -1.1877e-02, -7.1377e-03, -3.6259e-03,
         7.2067e-03, -4.0763e-03,  1.4473e-03,  1.0529e-03,  6.3452e-04,
         7.7478e-03, -1.6075e-03,  2.8369e-03, -8.2285e-03,  3.5403e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.1250e-04,  9.6638e-01,  3.3664e-03, -1.6686e-03, -2.8344e-03,
        -2.2903e-03,  1.3277e-03,  1.7318e-03, -2.2579e-03, -1.5514e-04,
        -2.0679e-03,  2.2419e-03,  2.1697e-03, -4.8000e-04,  7.8809e-04,
        -2.3025e-03, -1.2565e-03, -1.2726e-03, -1.3775e-04,  6.4884e-04,
         1.9753e-03,  1.8310e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8558e-03,  9.5131e-01, -1.2840e-02,  4.2133e-04, -8.8333e-04,
        -2.7473e-03, -2.5049e-03, -5.7013e-04, -2.6342e-03, -2.0692e-03,
         1.0680e-04,  2.3444e-03,  5.3401e-04, -1.8645e-04,  3.6815e-04,
         1.4948e-03, -1.2000e-03, -8.4845e-04,  1.2289e-03,  1.3068e-03,
         2.3168e-03, -2.5335e-04,  9.7144e-05,  5.4090e-05, -1.2884e-03,
         5.8146e-04,  1.1003e-03,  4.9805e-04,  2.3646e-03,  3.9933e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2330e-02,  9.4858e-01, -3.5896e-03, -3.1028e-03, -1.0174e-04,
        -2.3330e-03, -5.2923e-03, -6.5763e-03, -1.3693e-03, -6.5100e-03,
        -2.1660e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9038e-03,  9.5100e-01, -2.6852e-03, -2.1216e-03, -6.4979e-04,
         1.2196e-03,  5.3078e-03,  2.3140e-03,  1.4142e-03,  2.1254e-03,
         9.4486e-06,  7.7449e-03, -7.0922e-04,  1.3368e-03, -3.1671e-03,
         3.3855e-03, -1.5645e-03,  5.8165e-04,  1.3507e-03, -7.5886e-04,
        -4.6020e-03, -2.0460e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1747e-03,  9.3115e-01,  3.6597e-03, -2.3656e-03,  1.6828e-03,
        -1.3899e-02, -1.9976e-03, -1.4928e-03, -1.5349e-03, -2.0837e-03,
        -1.0431e-03, -1.0766e-03, -2.5981e-04, -1.0017e-03,  2.9164e-04,
        -8.0523e-04, -1.6477e-03, -1.1179e-03, -5.0395e-03,  3.7992e-04,
        -8.7552e-04, -1.3880e-03, -7.7426e-03,  8.4321e-04,  2.0362e-03,
         5.9905e-04, -8.3519e-04,  1.5434e-04, -2.5794e-03,  8.2794e-04,
         5.0063e-03,  1.5951e-03,  8.0853e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.5843e-03,  9.4079e-01, -2.5705e-04, -2.9073e-03,  1.4254e-04,
        -3.8521e-04,  3.8753e-03, -6.0807e-03,  4.0455e-03, -9.6259e-04,
        -1.6509e-03,  5.3211e-04,  2.4977e-03,  1.2337e-03, -1.1731e-03,
         1.6987e-03,  1.0277e-03,  9.6513e-04,  1.5165e-03,  3.0734e-04,
         2.6412e-03, -3.7407e-03,  4.2616e-04,  1.0474e-03,  7.8109e-04,
         2.0971e-03,  2.0888e-03,  5.5470e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1221e-02,  9.5033e-01, -6.1037e-04,  1.0355e-03, -3.9075e-03,
         2.8723e-03, -1.1032e-03,  3.0066e-03, -6.0105e-03,  2.3379e-03,
        -3.6638e-03, -1.0752e-03,  6.1016e-04,  7.3790e-04, -2.6964e-04,
         7.9115e-04,  2.2756e-03, -7.8159e-05,  7.7948e-04, -6.7916e-05,
         1.1978e-03,  2.4930e-04,  3.4487e-03,  3.8544e-04,  1.9354e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 3.5524e-03,  9.2990e-01,  1.0037e-03,  1.0213e-02, -3.0955e-03,
         2.2627e-03, -2.2496e-03, -1.6790e-03, -5.6599e-04, -1.5138e-03,
         1.6118e-03,  1.8772e-03,  5.4714e-06,  5.2555e-03, -1.4716e-03,
        -3.9760e-03, -9.9465e-06,  7.8099e-04, -9.5375e-04, -1.8260e-03,
         1.8677e-03, -1.4215e-03, -1.8574e-03,  1.4477e-04, -1.3856e-03,
        -5.6471e-04,  6.5865e-04, -1.2503e-03, -9.7567e-04, -2.0283e-03,
        -7.6601e-04,  1.5633e-04, -5.1447e-03,  4.2912e-04,  4.0699e-04,
         3.8179e-04, -5.0266e-04,  1.1081e-03,  6.6582e-04, -1.0362e-03,
         3.4478e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8115e-03,  9.5381e-01,  2.6190e-03, -2.2074e-03,  9.2285e-04,
         1.5135e-04,  1.9935e-03,  2.3905e-03, -4.9527e-04,  2.8430e-04,
        -1.0125e-03, -3.0759e-03,  1.2914e-03, -2.4568e-04,  6.1473e-03,
        -2.9826e-03,  1.7865e-03,  2.8408e-03,  1.0647e-03, -8.4429e-03,
        -1.2088e-03, -2.0839e-04,  8.1618e-04,  1.2324e-03,  9.5960e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5288e-03,  9.6990e-01,  6.8104e-04,  1.8434e-03,  2.0032e-03,
        -1.9344e-04, -2.1543e-04,  5.2490e-04,  8.3946e-04,  2.0608e-04,
        -3.7462e-04,  2.3961e-03, -1.4292e-03,  5.3676e-04, -2.0089e-03,
        -6.5667e-04, -1.2412e-03,  3.6228e-04, -1.2175e-03,  1.6666e-03,
        -8.1715e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9445e-02,  9.4745e-01, -7.1018e-04, -1.6573e-03, -5.1641e-04,
        -1.8330e-03, -3.7896e-03,  2.5871e-04, -1.7138e-03, -2.9879e-03,
        -5.6264e-03, -2.1456e-03, -4.3610e-05, -9.7106e-04, -4.6542e-03,
        -1.8477e-03, -3.3419e-03, -1.0127e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4113e-03,  9.4101e-01,  4.2895e-03, -4.0018e-04, -2.5965e-03,
        -1.9904e-03,  7.1513e-04, -2.7189e-03, -2.4797e-03, -5.8640e-04,
        -3.7330e-04, -1.5270e-04,  2.1725e-03,  3.0474e-03,  4.6936e-04,
         3.4129e-04,  1.0109e-02, -6.3849e-04,  2.5620e-03,  4.8639e-04,
        -5.2700e-04,  3.8960e-04, -8.9423e-04,  1.5447e-04, -3.5996e-04,
        -1.1615e-04, -1.5333e-03,  1.7298e-03, -1.5156e-03,  7.8334e-04,
        -2.7541e-04,  3.3516e-04, -9.1178e-04,  2.2711e-03,  8.8615e-04,
        -1.1222e-04, -3.8322e-03, -3.8254e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.4980e-03,  9.6621e-01,  4.8519e-03, -4.2703e-03, -1.0096e-03,
         1.5981e-03, -3.6743e-03, -7.6742e-04, -2.0301e-03, -2.0624e-03,
        -2.7953e-04, -2.1082e-03, -3.5545e-04,  9.7738e-04, -7.5060e-04,
         9.3883e-04,  3.6943e-04,  2.0122e-03,  1.1315e-04, -1.1239e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.1142e-03,  9.6231e-01,  2.4287e-04, -1.4650e-03, -1.9684e-03,
         4.4196e-03, -3.5517e-04,  3.7285e-03,  2.5686e-03, -9.7757e-04,
         3.8142e-03, -2.5101e-04, -4.1186e-03, -4.6648e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2238e-02,  9.3211e-01, -1.9668e-03, -2.5693e-03, -7.1200e-04,
         1.3867e-03, -1.3729e-04, -6.5463e-04, -7.9911e-04, -5.6864e-04,
         6.5174e-03,  1.0218e-03, -1.2800e-03,  8.3998e-04, -1.4238e-03,
         1.4184e-03,  1.5322e-03, -1.2072e-04,  1.1514e-03,  1.4545e-04,
        -1.8399e-03,  2.6913e-03,  4.5825e-04, -3.4446e-04, -1.5855e-03,
        -2.2509e-03,  1.5361e-03,  8.3272e-04, -2.9678e-04, -2.3468e-03,
        -1.4802e-04,  7.0746e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5170e-02,  9.1499e-01, -2.3717e-03,  1.4800e-03, -1.2292e-02,
        -3.3032e-03, -2.3112e-03, -3.3745e-03, -4.9866e-03, -3.8535e-04,
        -1.2867e-03, -6.9621e-03,  6.5550e-04,  4.5431e-05, -5.2713e-03,
        -6.6054e-04, -3.9108e-03, -3.1866e-03, -3.9696e-03, -1.3385e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4531e-04,  9.3578e-01, -1.1065e-02, -4.8646e-03,  7.3682e-04,
         2.4241e-03,  5.2321e-03, -4.2349e-04,  9.5089e-04,  2.3536e-04,
        -1.2120e-03,  1.0294e-05,  1.5134e-04, -1.5357e-04,  3.9303e-04,
        -1.3412e-03, -1.9133e-03, -7.7902e-04, -1.2983e-04, -3.4159e-04,
        -6.4068e-04, -1.4585e-03, -3.5151e-04, -1.7223e-04,  1.5410e-03,
        -2.6429e-03, -1.8032e-03, -1.0770e-03, -7.7677e-04,  8.5820e-04,
        -6.0915e-04, -1.2582e-04, -3.5620e-03,  5.2074e-03, -2.1923e-03,
         5.3949e-04,  1.9211e-03, -2.1962e-04,  2.1029e-03,  3.0326e-04,
         4.3750e-05, -8.3876e-04,  3.9813e-04,  4.3258e-04, -3.8658e-04,
        -1.1991e-03, -1.0793e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7342e-02,  9.4699e-01, -1.6710e-03,  8.1493e-05,  5.1454e-04,
        -1.6855e-03,  2.5396e-03,  3.5817e-03,  2.5357e-03,  4.0961e-03,
        -5.7533e-03,  3.9441e-04, -8.0670e-04,  5.7491e-03, -6.2624e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8088e-02,  9.3480e-01, -2.8215e-03,  1.3284e-03, -7.3012e-04,
        -1.7335e-04, -7.7413e-05,  4.8932e-04,  8.1554e-04, -1.1131e-03,
         5.2169e-04, -7.4954e-04, -1.3978e-03,  2.4098e-03, -1.5678e-03,
        -2.4031e-03, -8.2982e-04, -1.3873e-03, -2.8213e-03, -9.2834e-04,
        -7.5505e-04, -1.3153e-03, -3.0715e-04,  4.6472e-04,  1.3953e-03,
        -2.3597e-03,  5.6676e-04, -1.1621e-03,  4.9515e-04,  9.4079e-04,
        -1.6260e-03, -6.4604e-04,  2.9661e-04, -7.8837e-05, -1.3344e-03,
        -2.0801e-04, -4.8198e-04,  5.8830e-04, -1.0768e-03,  3.1568e-04,
         5.9599e-04, -1.8080e-04,  2.2991e-04,  3.4590e-04, -7.2727e-04,
        -1.6097e-03,  2.4848e-03, -1.9557e-03], device='cuda:0',
       grad_fn=<DivBackward0>)]
