Iter #50: [tensor([-0.0883, -0.2572, -0.0153, -0.0124, -0.0400, -0.0523, -0.0654, -0.0503,
        -0.0163, -0.0213,  0.0062, -0.1468,  0.0045,  0.0220, -0.0032,  0.0293,
        -0.1444, -0.0055,  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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1513, -0.1524, -0.0094, -0.0421, -0.1370, -0.0193, -0.0378, -0.0360,
        -0.0056, -0.1318, -0.0047, -0.0043, -0.0134, -0.0115, -0.1009,  0.0101,
        -0.0046, -0.0074, -0.0884,  0.0038,  0.0027, -0.0096,  0.0100,  0.0059,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1786,  0.1388,  0.0081,  0.0336,  0.1402,  0.0109,  0.0332,  0.0383,
         0.0096,  0.1237,  0.0093,  0.0103,  0.0040,  0.0012,  0.0745, -0.0137,
        -0.0048,  0.0392,  0.0018, -0.0007,  0.0839, -0.0199, -0.0023, -0.0125,
        -0.0070,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0625, -0.1475, -0.0190, -0.0454, -0.1428, -0.0131, -0.0331, -0.0358,
        -0.0066, -0.1553, -0.0078, -0.0195, -0.0087, -0.0048, -0.1118,  0.0094,
         0.0024, -0.0405, -0.0096, -0.0083, -0.0956, -0.0127,  0.0029,  0.0047,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0821,  0.1841,  0.0144,  0.0813,  0.1954,  0.0437,  0.0312,  0.0141,
         0.0047,  0.1768, -0.0002,  0.0236,  0.0084,  0.0139, -0.0043,  0.0098,
         0.0286,  0.0118, -0.0099, -0.0199,  0.0127,  0.0134, -0.0156,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0965, -0.1450, -0.0122, -0.0670, -0.1568, -0.0374, -0.0127, -0.0137,
        -0.0033, -0.1386, -0.0056, -0.0052, -0.0031, -0.0033, -0.0147,  0.0204,
         0.0054, -0.0089, -0.0842,  0.0122, -0.0283, -0.0976, -0.0261,  0.0009,
         0.0008,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2462,  0.1317,  0.0109,  0.0655,  0.1638,  0.0281,  0.0236,  0.0105,
         0.0112,  0.1307,  0.0056,  0.0143,  0.0159,  0.0034, -0.0153, -0.0129,
         0.0120,  0.0063,  0.0109, -0.0181,  0.0083,  0.0095, -0.0070,  0.0081,
         0.0097, -0.0090, -0.0100, -0.0019,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1638,  0.0181,  0.0354,  0.0117,  0.0433,  0.0486,  0.1513,  0.0109,
         0.0197,  0.0188,  0.0238,  0.0259, -0.0018,  0.0199,  0.1680,  0.0119,
         0.0114,  0.0138,  0.0071,  0.0038,  0.1199, -0.0046,  0.0169,  0.0030,
         0.0117, -0.0088,  0.0048, -0.0072, -0.0141,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1506,  0.0196,  0.0384, -0.0043,  0.0291,  0.0519,  0.1744,  0.0094,
         0.0223,  0.0193,  0.0185,  0.0241, -0.0017,  0.0187,  0.1768,  0.0097,
        -0.0005,  0.0135,  0.0081,  0.0045,  0.1193, -0.0029, -0.0123, -0.0062,
         0.0099,  0.0148, -0.0033,  0.0008, -0.0101, -0.0155, -0.0092],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0566,  0.0247,  0.0410, -0.0042,  0.0360,  0.0525,  0.1712,  0.0138,
         0.0236,  0.0238,  0.0199,  0.0209,  0.0045, -0.0058,  0.1445,  0.0020,
         0.0030,  0.0197,  0.0058, -0.0023,  0.1100, -0.0089,  0.0191,  0.0086,
         0.1356, -0.0189,  0.0015, -0.0104, -0.0108,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2820,  0.1231,  0.0064,  0.0534,  0.1381,  0.0067,  0.0247,  0.0195,
         0.0129,  0.0187, -0.0118,  0.0366,  0.0254,  0.0280,  0.0279,  0.0922,
         0.0093,  0.0073,  0.0029,  0.0080,  0.0045,  0.0133, -0.0042, -0.0003,
         0.0041,  0.0046,  0.0135,  0.0090, -0.0116,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1186,  0.1563,  0.0110,  0.0686,  0.1450,  0.0109,  0.0244,  0.0109,
         0.0201,  0.0241, -0.0223,  0.0397,  0.0304,  0.0242,  0.0275,  0.1088,
         0.0053,  0.0156,  0.0043,  0.0729, -0.0069,  0.0180,  0.0117,  0.0022,
        -0.0052, -0.0152,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-0.1524, -0.2531, -0.0009, -0.0034, -0.0168, -0.0022, -0.0110, -0.0111,
        -0.0288, -0.1838, -0.0306,  0.0077, -0.0021, -0.1992, -0.0135,  0.0012,
        -0.0007, -0.0052,  0.0063,  0.0019,  0.0022, -0.0061, -0.0020, -0.0015,
         0.0079, -0.0027, -0.0056,  0.0003,  0.0075,  0.0024,  0.0043,  0.0117,
        -0.0042,  0.0081, -0.0015,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0228,  0.2934, -0.0019,  0.0042,  0.0281,  0.0010,  0.0125,  0.0209,
         0.0230,  0.1744,  0.0286, -0.0058, -0.0032,  0.2106,  0.0165,  0.0015,
        -0.0069,  0.0061, -0.0031,  0.0038,  0.0012, -0.0021,  0.0973, -0.0007,
         0.0005, -0.0122,  0.0004,  0.0065, -0.0068, -0.0004, -0.0037,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1781e-02,  2.9617e-01,  2.8588e-04,  4.0634e-04,  2.6132e-02,
         8.5316e-03,  1.1423e-02,  1.3858e-02,  2.1989e-02,  1.9482e-01,
         3.4610e-02, -8.4074e-03,  5.0657e-03,  1.7622e-01,  4.1913e-03,
         2.0496e-03, -8.2157e-03,  8.2945e-03,  1.0799e-02, -1.2737e-03,
        -1.3663e-03,  1.7485e-03, -6.5905e-04, -8.3353e-03,  9.5945e-02,
         1.4604e-02, -7.9524e-03,  3.9744e-04, -4.4652e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+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,  0.5185,  0.0191,  0.0281,  0.0165,  0.0753,  0.0256,  0.0591,
        -0.0086,  0.0323, -0.0044,  0.0139,  0.0257,  0.0119,  0.0061,  0.0030,
         0.0146, -0.0115,  0.0141, -0.0053,  0.0207, -0.0147, -0.0085,  0.0365,
         0.0025, -0.0049, -0.0024,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3463e-04,  4.6366e-01, -1.0085e-03,  4.4035e-02,  6.6387e-03,
         7.7035e-02,  8.4139e-03,  3.4211e-02,  6.1137e-03,  3.8684e-02,
         8.1940e-03,  5.4263e-03,  1.9768e-02,  1.9726e-02,  7.2576e-03,
        -4.7101e-03, -4.4469e-07,  1.4405e-01, -6.8418e-03,  1.0149e-02,
        -3.2560e-02,  5.4108e-04,  4.8167e-02,  8.9491e-03,  3.3188e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0802,  0.3781,  0.0015,  0.0285,  0.0084,  0.0721,  0.0138,  0.0372,
         0.0081,  0.0413,  0.0018,  0.0155,  0.0216,  0.0178,  0.0011, -0.0093,
         0.0036,  0.1477, -0.0128,  0.0063, -0.0090,  0.0470, -0.0056, -0.0149,
         0.0114,  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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0422,  0.0283, -0.0167,  0.0140,  0.0245,  0.0755,  0.0497,  0.0222,
         0.0112,  0.0200,  0.0155,  0.4760,  0.0081,  0.0385,  0.0459,  0.0240,
         0.0031,  0.0160, -0.0047, -0.0030, -0.0379,  0.0087,  0.0127, -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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0739,  0.0184, -0.0063,  0.0191,  0.0213,  0.0667,  0.0517,  0.0120,
         0.0101,  0.0104,  0.0110,  0.3554,  0.0229,  0.0233,  0.0324,  0.0141,
        -0.0009, -0.0021, -0.0062, -0.0114,  0.0079,  0.1688,  0.0144, -0.0079,
        -0.0091, -0.0136,  0.0007, -0.0038,  0.0044,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6770e-02,  2.6446e-02, -2.2041e-02,  2.0228e-02,  1.3695e-02,
         6.0235e-02,  4.2738e-02,  1.8519e-02, -2.7699e-04,  1.7572e-03,
         5.4708e-03,  2.5425e-01,  1.7499e-02,  2.3066e-02,  3.0818e-02,
         2.4417e-02,  4.6066e-03,  1.1077e-02,  1.0555e-02,  1.3596e-01,
         2.2727e-02, -1.4379e-02,  1.1490e-02,  3.5530e-02,  4.6259e-03,
         1.3905e-04,  2.5765e-02, -1.8261e-02, -1.6003e-02, -1.6981e-02,
        -3.0177e-04, -1.0238e-02, -2.4564e-02, -1.0056e-02, -6.2229e-03,
        -1.3714e-02, -7.9260e-03,  1.0644e-02], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0201,  0.0285, -0.0259,  0.0122,  0.0154,  0.0596,  0.0570,  0.0157,
        -0.0034, -0.0007,  0.0097,  0.3607,  0.0181,  0.0218,  0.0323,  0.0120,
         0.0116,  0.0073,  0.0046, -0.0281, -0.0287, -0.0035,  0.0227,  0.1679,
         0.0055,  0.0225, -0.0039, -0.0006,  0.0000,  0.0000,  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,  0.0350, -0.0107,  0.0230,  0.0270,  0.0833,  0.0842,  0.0155,
        -0.0005,  0.0033,  0.0139,  0.3749,  0.0341,  0.0429,  0.0461,  0.0129,
         0.0073,  0.0137,  0.0193,  0.0030, -0.0364, -0.0044,  0.0233, -0.0020,
        -0.0035,  0.0005,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0056,  0.0279, -0.0261,  0.0147,  0.0214,  0.0711,  0.0549,  0.0119,
        -0.0061, -0.0014,  0.0136,  0.3658,  0.0238,  0.0293,  0.0314,  0.0186,
         0.0141, -0.0028, -0.0004, -0.0063, -0.0245,  0.0058,  0.0097, -0.0007,
        -0.0006, -0.0169,  0.1417,  0.0337, -0.0080, -0.0111,  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([ 1.8262e-02, -4.6958e-01,  1.3161e-04,  2.7386e-05, -4.0166e-03,
        -1.9311e-01,  2.2033e-04,  7.1101e-04, -3.2635e-04, -1.7338e-01,
         1.3497e-04,  3.5836e-04, -1.2206e-01, -4.4627e-04,  1.4223e-04,
        -2.9784e-03, -2.3698e-03, -4.9525e-03,  2.4910e-04, -4.0028e-04,
        -2.5175e-04, -1.1022e-03,  3.1012e-04,  1.2044e-03, -9.8727e-05,
         5.1362e-05,  7.1673e-04,  8.3974e-04, -7.6641e-04, -1.0220e-04,
         7.0065e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0181e-02, -3.7765e-01, -5.4584e-04, -1.2735e-03, -2.8031e-03,
        -2.5744e-01, -8.8078e-04,  1.3342e-03, -1.9624e-04, -1.7098e-01,
         2.0069e-03,  9.9090e-04, -1.4277e-01,  1.9517e-03,  2.1262e-03,
        -2.8745e-03, -1.7323e-03, -4.5583e-03,  1.5988e-03, -5.4173e-04,
        -2.5582e-04,  8.0086e-04,  2.7560e-04,  1.1813e-03,  2.1360e-03,
         3.6555e-03, -4.3950e-05,  2.8384e-04, -2.8696e-03, -6.5207e-04,
        -1.0562e-03,  1.2816e-03,  1.0738e-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([-4.2547e-02, -3.7424e-01, -8.7459e-04, -1.0018e-03, -3.1439e-03,
        -2.6200e-01, -1.0627e-03,  1.1769e-03,  5.0252e-04, -1.6222e-01,
         5.4510e-04,  8.9247e-05, -9.4938e-02, -2.8613e-04,  9.3627e-04,
        -4.2265e-03, -1.7470e-03, -3.6720e-03, -8.2310e-04,  2.0557e-03,
        -1.0256e-03,  6.1931e-04,  1.6778e-05, -1.1276e-04,  1.4987e-03,
         1.1308e-03, -3.3307e-04,  1.1964e-03, -2.5458e-03, -2.9864e-02,
         1.4860e-03,  1.9231e-03,  1.5354e-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([ 1.6415e-02,  4.3048e-01, -4.8645e-04,  8.8443e-04,  3.8310e-03,
         1.8186e-01,  3.5489e-04,  6.9699e-04,  5.2236e-04,  2.0680e-01,
         5.2875e-04, -2.2301e-04,  1.2904e-01, -9.9849e-04,  3.0852e-05,
         3.2221e-03,  5.4119e-03,  4.4073e-03, -3.3584e-04, -3.3132e-05,
        -3.9997e-05,  1.5171e-03,  5.8417e-04, -1.9427e-04,  6.8490e-04,
        -1.2590e-03, -3.3253e-04,  3.4280e-04,  2.1600e-03, -3.0084e-04,
        -3.9131e-03,  9.8911e-04,  3.5824e-04,  6.0948e-04,  1.4704e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2540e-02, -3.8940e-01, -1.8309e-05,  1.6735e-05, -3.4488e-03,
        -2.4534e-01, -3.6692e-05,  6.3183e-04,  1.7039e-04, -9.5709e-02,
         3.8457e-04,  8.0514e-04, -1.7826e-01,  1.4029e-04,  9.6930e-04,
        -2.3423e-03,  3.3169e-04, -3.7578e-03,  4.8618e-04,  9.9042e-04,
        -1.1948e-04, -1.4774e-05,  9.6200e-05, -7.4923e-04, -2.5586e-02,
        -1.5827e-03, -2.6540e-03,  6.0413e-04,  2.2069e-04,  5.0670e-04,
         3.0090e-04,  1.4407e-03,  7.1561e-04, -2.3477e-02, -2.3486e-03,
         1.1001e-03,  1.5541e-03, -2.4697e-05,  1.1302e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.8556e-03, -4.8519e-01, -2.1024e-04, -6.0675e-04, -5.8339e-03,
        -1.8011e-01, -2.5210e-03,  1.3267e-04,  3.0460e-04, -1.4701e-01,
         8.6458e-04, -5.5119e-04, -1.4332e-01,  2.0854e-03,  5.4982e-04,
        -3.7237e-03, -2.7165e-03, -6.1980e-03,  2.5942e-04,  5.0017e-04,
        -1.5359e-03, -1.4906e-04, -1.2163e-03,  4.3359e-04,  1.0929e-03,
        -6.3743e-05,  1.2288e-03,  5.7933e-04,  9.1498e-04, -1.6053e-03,
        -6.2962e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7860e-03, -4.7078e-01,  1.3028e-03, -3.7865e-04, -2.3345e-03,
        -2.3042e-01, -1.0565e-03,  4.3962e-04,  1.3817e-04, -1.4585e-01,
         1.2140e-03,  1.3032e-03, -1.1461e-01,  2.4871e-04,  6.6744e-04,
        -3.1731e-03, -3.0801e-03, -2.8960e-03,  1.8205e-04,  3.2063e-04,
        -1.1389e-03, -4.8070e-04,  7.2240e-04,  1.4793e-03, -8.2635e-04,
         4.1884e-04, -8.1332e-04,  7.4375e-04, -9.2420e-04,  7.7961e-04,
         2.5404e-04,  1.2428e-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([-2.9560e-02, -4.0029e-01,  8.7712e-04, -8.3498e-04, -4.0179e-03,
        -2.2453e-01, -6.1130e-04,  1.1870e-03,  2.1626e-04, -1.5612e-01,
         1.1247e-03, -1.1217e-04, -1.5975e-01,  5.2831e-04,  7.6411e-05,
        -3.9367e-03, -1.7071e-03, -3.2667e-03,  1.3420e-03,  4.2651e-04,
        -1.5207e-03, -1.0240e-03, -5.0065e-04,  4.4585e-04,  2.9898e-03,
         6.6470e-04, -1.0513e-03, -3.0966e-04,  1.1524e-04, -8.6463e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6978e-03, -4.5610e-01,  1.4391e-03, -6.9131e-04, -3.3765e-03,
        -2.1800e-01, -9.3436e-04,  1.9875e-05,  9.1434e-04, -1.6674e-01,
         3.0582e-04,  2.5331e-03, -1.1980e-01,  7.3499e-04,  8.1163e-04,
        -4.8418e-03, -3.2507e-03, -4.7292e-03,  9.3591e-05, -1.6353e-04,
        -8.4397e-04, -1.2770e-03, -1.0325e-03, -1.0462e-04,  1.5572e-03,
        -6.3804e-04,  6.6321e-04,  1.3551e-03,  1.0118e-03, -1.0034e-03,
        -1.0436e-04, -1.2298e-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([ 2.4347e-02,  1.9932e-03, -7.3375e-03,  4.4773e-03,  9.9884e-03,
         3.0313e-02,  1.4473e-02, -3.5708e-03,  4.8552e-02,  7.8376e-01,
        -2.9916e-03, -2.9974e-03,  2.0471e-04, -7.2886e-03, -1.6134e-03,
        -1.7621e-02, -2.5576e-04, -6.9515e-03,  2.1043e-02, -8.8861e-03,
         7.9818e-04, -5.3544e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.5458e-03,  1.4497e-03, -3.6794e-03,  1.2345e-03,  3.1331e-03,
         2.7574e-02,  1.0132e-02, -3.5256e-03,  4.0293e-02,  8.5771e-01,
        -4.2703e-03,  9.6919e-04,  2.7988e-04, -3.4270e-03, -4.7847e-03,
        -1.0223e-02, -6.6270e-03, -1.0041e-03, -4.6149e-03,  4.1014e-03,
         3.6333e-03, -7.8951e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7327e-03,  5.1591e-03, -2.5642e-03,  1.5699e-04,  5.1415e-03,
         3.7431e-02,  8.8609e-03, -5.3186e-03,  4.7607e-02,  7.3267e-01,
        -7.5464e-03, -1.6363e-03, -2.4396e-03,  8.4381e-04, -1.5995e-04,
        -8.2731e-03, -5.4833e-03,  7.3995e-03, -1.6256e-03, -5.1454e-03,
         9.2906e-02, -1.6424e-02,  5.0489e-04,  2.9691e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  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([ 5.9256e-03,  7.5413e-04,  2.1108e-03, -1.2304e-03, -6.3600e-03,
        -2.6243e-04,  9.3674e-01,  3.0750e-03, -1.5571e-03, -1.4531e-04,
         5.6915e-03,  3.6043e-03,  4.6090e-04, -3.7986e-04, -1.2868e-03,
        -3.8487e-04, -2.9580e-04, -4.4261e-03,  1.4577e-02,  2.9015e-03,
         4.4369e-03,  1.4762e-03, -1.4694e-03, -2.6647e-04, -1.8248e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1137e-02, -1.6151e-03, -2.4148e-03,  1.7511e-03,  7.5512e-03,
         6.7148e-04, -9.3090e-01, -1.1287e-03,  1.0644e-03,  5.4205e-04,
        -1.1370e-02, -2.9333e-03,  1.6734e-03,  7.5987e-04, -9.1138e-04,
         2.4986e-04, -1.7820e-04,  9.5652e-04,  4.1411e-03, -2.3900e-03,
         4.1041e-03, -1.5394e-03, -1.5303e-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([ 3.5436e-03, -7.0231e-01,  6.8608e-04, -2.0065e-03, -2.6971e-01,
         1.5793e-04, -1.0514e-03,  1.5155e-03, -3.3202e-04, -1.7865e-03,
        -1.1559e-03, -9.2771e-04,  1.6797e-03, -2.6601e-03, -1.1841e-03,
         9.1073e-04,  5.5260e-04, -3.3944e-04, -1.1023e-03,  8.1753e-04,
        -4.9517e-05, -6.6127e-04, -8.3316e-04,  1.9879e-05,  5.4996e-04,
        -1.3627e-03,  1.0312e-03,  2.9970e-04, -7.6490e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.0731e-03, -6.4990e-01,  4.3531e-04, -3.1111e-03, -3.1362e-01,
         3.5939e-04, -4.6296e-06,  1.3605e-03, -1.3779e-03, -1.4562e-03,
        -1.0720e-03, -6.0717e-04,  1.6906e-03, -4.4387e-03, -1.9154e-04,
         4.3750e-04, -3.2460e-04, -4.3847e-05, -2.5068e-03,  2.7192e-05,
        -1.0601e-04,  8.9468e-04, -5.1621e-04, -4.5668e-03,  1.9214e-03,
        -9.1498e-04,  4.4934e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.9859e-03, -5.7859e-01, -3.9628e-04, -4.8289e-03, -3.8813e-01,
        -6.3652e-04, -7.7450e-04,  4.8958e-04, -7.7725e-04, -1.0466e-03,
        -1.6065e-03, -8.7223e-04,  1.6210e-03, -5.2348e-03, -6.5309e-04,
         4.1538e-04, -9.5787e-04, -3.9914e-04, -1.8296e-03,  1.4019e-04,
         7.5888e-05, -3.3393e-04, -1.0067e-03, -1.2249e-05, -1.8371e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8817e-03, -7.0475e-01, -8.3153e-04,  1.1016e-03, -1.4584e-04,
        -4.5669e-03, -1.3370e-03, -1.6460e-03,  9.1791e-05, -2.7064e-01,
        -1.0370e-04,  1.0381e-03,  6.4412e-04, -7.1862e-04, -8.0670e-04,
        -8.0993e-04,  5.9889e-05,  5.0353e-04,  1.0797e-03,  2.1701e-03,
        -1.8980e-03, -8.2613e-04,  1.1603e-06, -3.5032e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7004e-03, -5.9196e-01, -5.9046e-04,  1.0306e-03, -3.5932e-04,
        -3.7182e-03,  1.4423e-03, -2.3900e-04,  2.9278e-04, -3.8170e-01,
         3.9607e-04,  1.0829e-03, -7.4633e-04, -1.0062e-04, -2.8953e-04,
        -1.7887e-03,  5.7118e-04, -5.5293e-05,  5.9357e-04, -5.8926e-04,
        -6.9361e-03, -2.5841e-04, -7.6569e-05,  4.7821e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5083e-03, -7.4696e-01,  2.4005e-04,  9.8721e-04, -6.9688e-04,
        -3.4412e-03, -5.6390e-04,  1.4007e-04, -2.0029e-04, -2.3949e-01,
         4.7856e-04, -6.6237e-04, -4.4341e-04, -7.4676e-04, -7.7467e-04,
        -4.6851e-04, -1.2543e-05,  5.3911e-04, -1.5645e-04,  1.2615e-03,
         1.9965e-05, -2.0425e-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([-3.3176e-03, -4.2301e-01, -2.1678e-04, -4.0490e-03, -3.4903e-01,
        -3.7246e-04, -1.1020e-03,  1.0144e-03, -2.4715e-03, -2.0244e-01,
         2.3378e-03,  6.9776e-04,  3.8017e-04, -2.6346e-03,  2.1953e-04,
        -5.2137e-05,  1.4228e-03, -3.3617e-04, -2.3132e-03,  2.0607e-03,
         1.2317e-04,  4.0357e-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([-5.7167e-03, -5.9710e-01,  1.0564e-03, -1.5892e-03, -2.2818e-01,
        -4.5893e-04,  1.8370e-05,  1.5098e-03, -2.4888e-03, -1.5006e-01,
         1.7978e-03, -6.9010e-04,  2.9790e-05, -1.9302e-03,  4.2163e-04,
        -1.6173e-04,  1.1513e-03,  1.8931e-04, -1.8941e-04,  1.3745e-03,
        -1.0690e-03, -5.4033e-04,  1.6589e-03, -1.2326e-04,  4.8965e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.3655e-03, -4.7911e-01,  1.2544e-03, -1.6720e-03, -2.4207e-01,
        -6.0774e-04,  5.8779e-04,  1.0464e-03, -6.1966e-04, -2.5265e-01,
         1.6982e-03,  2.5426e-04, -1.3922e-03, -2.1922e-03,  6.9895e-04,
        -5.7972e-05,  9.2839e-04,  3.0218e-04, -1.2631e-03,  1.8563e-03,
         4.8856e-04, -8.8275e-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([ 0.2264,  0.0011,  0.0217,  0.1632,  0.0242, -0.0156, -0.0630,  0.0258,
        -0.0356, -0.0209, -0.0074, -0.0121,  0.0244,  0.3227, -0.0078, -0.0207,
        -0.0023, -0.0052,  0.0000,  0.0000,  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([ 2.5253e-03, -9.8207e-01, -1.7613e-04,  1.6717e-04,  2.0817e-04,
         6.2594e-05, -1.5262e-03, -8.8315e-04, -2.8504e-03, -8.4148e-05,
        -8.3252e-04, -9.2326e-04, -1.6081e-04,  3.6791e-04,  7.0729e-04,
        -1.2004e-04,  3.8759e-04, -9.0741e-04,  6.6445e-04, -2.0803e-03,
         4.7465e-04, -1.2916e-03,  5.3360e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.8454e-03, -7.8779e-01,  1.3826e-04, -7.5896e-04, -3.7441e-03,
        -2.9677e-04, -2.0766e-03, -2.3535e-04, -1.1965e-02, -2.4460e-04,
        -8.1708e-04, -1.7987e-01, -2.5957e-05, -7.2783e-05,  9.5814e-05,
        -3.9825e-04,  7.3717e-04, -4.1859e-04, -7.6663e-04,  3.8322e-04,
        -4.8629e-04,  6.8688e-04, -1.5749e-04, -1.6245e-04, -8.1654e-05,
         5.8660e-04, -5.7558e-04, -3.7622e-04, -2.1200e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2310e-03,  8.4863e-01,  6.5514e-05, -1.8119e-04,  4.1797e-03,
         2.5374e-04,  5.1237e-04,  7.2131e-04,  2.0428e-02, -1.6479e-05,
         1.2126e-03,  1.1289e-01,  2.8676e-04,  1.2168e-04,  8.3182e-04,
         1.8047e-04, -1.4591e-04,  2.1271e-04,  5.6709e-05,  1.2232e-04,
         2.9343e-05,  3.8439e-04, -3.0614e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0193e-02, -4.5744e-01, -7.5408e-04, -9.1094e-04, -5.5555e-03,
        -2.9032e-04, -4.5119e-03, -8.5217e-04, -4.9954e-02, -5.8156e-04,
        -1.5720e-03, -4.6022e-01,  6.4500e-04, -3.7331e-04, -1.0087e-03,
        -1.6158e-03,  4.5336e-04,  1.0042e-04, -9.0845e-04,  3.6419e-04,
         2.8676e-04,  4.8875e-06,  4.0646e-04, -7.9080e-04, -2.0875e-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([ 3.7165e-02,  5.1059e-03, -1.4357e-05,  8.9571e-04, -4.6938e-03,
        -3.3687e-03,  1.0434e-02, -5.1946e-04, -3.2636e-02, -8.8088e-01,
        -6.1453e-03,  8.4744e-04,  2.5800e-03,  7.3306e-03,  2.2863e-03,
         2.6135e-03,  6.9431e-04,  5.9424e-04,  1.1903e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8518e-04,  2.4199e-03, -3.3119e-04, -8.7419e-04, -2.4559e-03,
        -2.4379e-03,  2.1182e-03, -7.3124e-05, -1.6640e-02, -9.6133e-01,
        -3.2822e-03, -3.0901e-05, -9.3683e-04,  8.6254e-04,  1.9562e-06,
         3.9321e-04, -1.3199e-03, -1.2973e-03, -6.2304e-04, -8.3042e-04,
         7.6458e-04, -7.9019e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0102e-03,  2.5595e-03,  4.3620e-04,  9.5473e-05, -3.4016e-03,
        -2.2818e-03,  3.0629e-03,  3.4177e-04, -5.4300e-03, -9.7381e-01,
        -1.8830e-03, -3.7836e-04,  8.9586e-04,  1.9323e-03,  1.2108e-03,
         2.7904e-04,  9.8945e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+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, -7.6636e-01,  3.4993e-04, -1.0801e-03,  6.5295e-04,
         3.4895e-04, -8.8523e-04, -3.4844e-04, -2.0963e-01, -9.3774e-05,
        -1.8677e-03,  4.5563e-04,  8.8690e-04,  9.0680e-04, -4.7665e-04,
         2.7547e-04,  1.7139e-03, -5.1082e-04,  2.7038e-04,  3.4939e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1227e-03, -6.9795e-01, -1.8693e-04,  4.5446e-04,  6.6619e-04,
         6.4942e-04, -1.2981e-03, -7.5862e-04, -2.8794e-01, -7.9475e-05,
        -1.1397e-03,  4.8933e-04,  3.0933e-04,  2.8285e-03, -5.1147e-04,
        -5.0754e-04, -1.1411e-03,  1.7192e-04, -1.2209e-04,  1.7237e-04,
        -1.5553e-05, -4.7548e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7193e-03, -8.0892e-01, -5.9695e-04, -1.5514e-03,  6.1338e-04,
         1.2447e-03, -1.3494e-03,  2.5800e-04, -1.7152e-01,  2.3837e-04,
        -2.7723e-03,  6.1412e-04,  1.5217e-03,  3.9351e-04, -5.5662e-04,
        -1.1174e-03, -7.8205e-04,  3.5073e-04, -1.2223e-03,  1.3449e-03,
         2.7633e-04,  3.1702e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.0038e-03, -5.8194e-04,  4.9328e-05, -6.2992e-05,  2.4632e-03,
        -4.2434e-03,  7.2121e-04, -3.6120e-03, -7.0632e-01, -4.6738e-04,
         1.6053e-03, -2.1308e-03, -2.5935e-05, -2.6183e-01,  7.1681e-04,
         6.6211e-05, -1.0870e-03, -2.3660e-05,  3.0139e-03, -5.5452e-04,
         1.0008e-04, -2.4020e-04,  2.4540e-04, -5.6611e-04, -2.6293e-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([-7.3353e-03, -1.0480e-03, -4.5936e-05, -5.7200e-04,  3.6406e-03,
        -1.9189e-03, -3.6963e-04, -2.7651e-03, -8.1446e-01, -6.3531e-04,
         1.5579e-04, -1.1205e-03, -2.5280e-03, -1.5413e-01,  9.1541e-04,
        -1.2844e-04, -8.2201e-04,  3.6594e-04,  3.2820e-04, -1.5089e-04,
         6.1937e-04,  9.2045e-04,  4.9798e-04, -7.6215e-04,  2.2765e-03,
        -4.7584e-05,  2.9663e-04,  1.0073e-05, -7.3362e-04, -2.8698e-04,
         1.1037e-04], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #300: [tensor([-8.1650e-03,  9.7844e-01, -6.1132e-04, -1.6779e-03,  1.9295e-03,
         2.6819e-04,  5.1834e-04, -3.2933e-04, -4.7149e-04, -3.6663e-04,
        -6.5756e-04,  4.9128e-04,  5.5169e-04,  3.9459e-04, -9.3077e-04,
         4.7360e-04, -1.9944e-05, -1.2126e-03, -1.6183e-03,  2.7839e-04,
         4.6945e-04,  1.2199e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6094e-03, -9.8844e-01, -2.3613e-04,  1.2313e-03, -1.0472e-03,
         4.5021e-04, -7.5859e-05,  1.5812e-04,  6.1394e-04,  9.2815e-04,
        -2.0623e-04, -1.5105e-05,  7.3532e-05, -3.6147e-04,  9.1762e-04,
        -9.7966e-04,  2.8438e-04,  4.0943e-05, -3.6894e-04,  9.1425e-04,
        -1.3832e-04, -5.6126e-04,  3.5081e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0785e-03,  9.5456e-01,  6.5623e-04,  8.3523e-04,  2.3646e-03,
         2.8722e-03, -2.6223e-04,  2.5280e-02, -1.3943e-04, -1.5744e-03,
         1.0570e-03,  1.1979e-03,  7.0575e-04, -3.9720e-04,  1.7369e-03,
         7.0988e-04,  4.6965e-04, -3.4645e-04, -3.5285e-04,  1.3362e-03,
         1.2221e-03,  1.6442e-04, -6.8391e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4853e-03,  9.7347e-01, -5.3271e-06,  2.6714e-04,  3.2928e-03,
         5.8668e-04, -3.7946e-04,  1.2455e-02, -2.6401e-05,  3.1096e-04,
         8.0395e-04,  4.1385e-04,  3.7739e-04,  4.7281e-04,  5.5800e-04,
         5.2734e-04, -4.0718e-04, -2.8816e-04, -6.8947e-05, -3.2788e-04,
        -2.8759e-04,  2.0297e-05, -1.7258e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.0901e-03, -9.8556e-01,  3.0746e-04,  7.4214e-04, -1.8955e-03,
         2.7471e-04, -6.0844e-04, -2.2956e-03,  6.1022e-04, -3.7964e-04,
         2.0819e-04,  2.9933e-04, -3.3663e-04, -9.8997e-05,  1.2175e-03,
        -6.8688e-04,  1.3546e-04, -7.9861e-05, -7.4148e-05,  9.9230e-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([-4.8545e-03,  9.7873e-01,  4.2287e-04, -5.6386e-04,  1.3623e-03,
         7.5864e-04, -9.6531e-05,  6.4261e-03,  4.7865e-05, -2.7694e-04,
         9.5552e-04,  2.6437e-04, -1.3863e-04,  3.7117e-04,  1.8302e-04,
         2.3599e-04,  4.6723e-04, -4.9821e-04,  1.7533e-04,  5.5644e-04,
         3.3108e-04,  4.3109e-04, -1.4784e-05, -2.5987e-04,  1.1622e-04,
         2.8654e-04, -3.1546e-04,  5.0167e-04,  3.5458e-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([-3.4679e-03, -9.7843e-01,  9.5533e-04,  1.4639e-03, -1.7038e-03,
         1.1180e-03,  8.8805e-04, -5.7603e-03,  1.1598e-03,  1.5123e-03,
        -1.9843e-04, -2.1381e-04,  1.6327e-04, -6.5966e-05,  1.6863e-03,
        -2.5799e-04,  4.5538e-05, -3.5620e-04, -4.7029e-04,  8.3170e-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([ 5.6020e-03,  9.5985e-01, -3.9115e-04,  9.9124e-04,  9.6589e-04,
         1.2269e-03,  3.3200e-04,  2.1574e-02, -3.1106e-04, -1.2828e-03,
         7.2048e-04,  5.5276e-04,  5.8655e-05,  4.3410e-06,  9.3715e-04,
         3.8458e-04, -3.9994e-05, -1.6196e-04,  1.4195e-04,  7.3614e-06,
         4.7148e-04,  2.7434e-04,  8.1571e-05,  1.1405e-04, -5.1157e-04,
         1.3950e-03, -4.0942e-04, -2.2755e-04,  4.3581e-04, -1.8186e-05,
        -5.2637e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4540e-03,  9.7540e-01, -1.0988e-03, -6.4007e-04,  1.0118e-03,
         7.6478e-04, -3.8647e-04,  6.3745e-03, -1.5364e-04, -2.1181e-04,
         7.6705e-04,  5.8342e-04, -2.5404e-04,  7.3017e-04, -2.4655e-04,
         3.0466e-04,  1.4009e-03,  1.8660e-04, -4.3225e-04,  6.2495e-04,
         8.9660e-04,  1.0402e-04,  1.9166e-04, -1.5598e-04, -5.1883e-04,
         1.4816e-03,  1.0841e-04, -2.0800e-04,  1.1190e-04, -1.9559e-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([-5.6722e-03, -9.8102e-01,  7.1948e-04, -3.5359e-04, -2.0440e-03,
        -1.4806e-04,  2.2253e-04, -4.7133e-03,  4.3475e-05, -2.1645e-04,
        -6.3143e-04,  2.5426e-04, -1.5796e-04, -1.2065e-04,  7.8081e-04,
        -8.3081e-04, -1.1415e-04,  7.4124e-05, -2.5589e-04, -3.3956e-04,
         2.1867e-04, -7.7230e-04,  4.4426e-05,  2.5300e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7607e-03,  9.7166e-01, -3.8249e-04,  6.1819e-04,  1.9285e-03,
         6.8014e-04,  1.6054e-03,  1.4382e-02, -9.4176e-04,  3.7740e-04,
         8.5217e-04,  1.2751e-04, -3.2515e-04, -1.2930e-04, -1.0917e-04,
         7.3687e-04, -6.4878e-04, -3.3437e-04,  3.5578e-04,  9.1515e-04,
        -1.2423e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+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,  9.8214e-01, -4.6197e-04, -7.4773e-04,  1.5210e-03,
        -6.2624e-04, -3.3898e-04, -1.1132e-03, -5.3566e-04, -3.8684e-04,
         4.6931e-04,  2.0256e-04, -6.8369e-04, -1.0465e-04,  2.1896e-04,
        -2.6093e-04, -1.9241e-04,  5.6386e-04, -1.4689e-04,  3.5481e-04,
        -4.5064e-04,  1.3286e-03, -2.6103e-05,  9.7001e-05, -2.4790e-04,
        -1.7654e-04, -6.9571e-04, -3.4394e-04,  6.5681e-04,  3.9730e-05,
         7.1990e-04,  3.9800e-04, -3.1547e-04,  2.2994e-04, -2.6184e-04,
         2.9606e-04], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-1.6944e-04, -9.4690e-01,  4.6288e-04, -1.0873e-03, -2.4523e-04,
        -2.0428e-02, -2.4526e-05, -6.7880e-04, -1.9927e-02, -2.6478e-04,
        -8.5846e-04, -1.4715e-04,  1.4939e-04, -1.8910e-03,  1.6215e-04,
        -2.9432e-04,  3.6637e-04, -6.9829e-04, -6.2772e-04, -5.2213e-04,
         2.1909e-04, -1.1917e-04,  6.1497e-06,  3.5774e-05,  1.6670e-04,
        -6.9504e-04,  1.5098e-04,  1.4164e-04, -3.0544e-04, -6.7346e-04,
        -1.6511e-04,  1.0047e-04,  1.6576e-04, -3.0406e-04, -1.6106e-04,
         8.9346e-05, -1.3632e-04, -3.1834e-04, -1.4481e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2044e-03, -8.7667e-01,  1.3270e-04, -2.9461e-03, -1.2723e-04,
        -3.8184e-02, -4.9615e-04, -8.4615e-04, -6.0470e-02, -3.3646e-04,
        -1.4125e-03, -4.0301e-04,  6.2974e-04, -5.4518e-03,  8.9901e-05,
        -4.5369e-04,  7.0090e-04, -6.4564e-04, -9.5594e-04, -6.0607e-04,
        -8.9773e-04, -6.6648e-04, -1.1998e-05,  3.6458e-04, -9.5893e-05,
        -4.0053e-04,  8.1689e-05, -1.0634e-03, -1.0938e-03, -5.9423e-04,
         3.2225e-04, -3.1507e-04, -3.3147e-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([ 2.7329e-03, -9.8786e-01,  4.8077e-04,  7.1919e-04, -9.0053e-04,
         4.6689e-04,  9.3927e-04, -1.5643e-03,  3.3132e-04,  2.1044e-04,
         3.3584e-05, -5.9349e-04, -5.5446e-04,  2.8137e-04, -4.5795e-04,
        -3.5480e-04, -4.8866e-04,  5.2343e-04, -9.4066e-05, -1.6266e-04,
        -1.7987e-05,  2.2928e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.8195e-03, -9.7171e-01, -6.2245e-04, -5.6924e-04, -1.7141e-03,
        -2.3293e-04, -9.9855e-04, -1.4129e-02,  4.8899e-04, -6.0794e-05,
        -1.1241e-03, -7.4673e-05, -5.2354e-04, -3.9098e-04, -6.2108e-04,
        -3.2243e-05, -1.2898e-03,  6.1989e-04, -7.0986e-04, -9.9148e-05,
        -1.6781e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6729e-03, -9.8037e-01,  2.1975e-05, -3.7525e-04, -6.4135e-04,
         8.6535e-04,  8.9053e-04, -7.6296e-03, -4.0118e-04,  1.4487e-04,
         3.8394e-04,  1.6869e-04,  3.1248e-04,  2.8610e-04, -6.6812e-04,
        -1.0737e-04,  5.9697e-04, -1.5837e-04, -1.4805e-04, -2.8550e-04,
        -6.0936e-04, -5.5716e-04, -2.0398e-04, -2.7870e-04, -1.6095e-04,
         5.6469e-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([ 1.4816e-03,  2.2601e-04, -7.6812e-05, -6.5413e-04,  9.2369e-02,
         1.9105e-03, -1.5508e-04,  5.8575e-04,  2.7936e-04, -9.2626e-04,
        -2.0248e-03,  8.9490e-01,  5.7463e-04,  1.3355e-03,  3.4533e-04,
         2.7599e-04, -9.9618e-05, -1.1799e-03, -4.4357e-06, -5.9351e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.1323e-03, -1.0003e-04, -7.3331e-04, -4.0094e-04, -5.6144e-02,
        -9.7830e-04, -5.9953e-04, -6.5228e-04, -1.6786e-03,  1.4260e-03,
        -7.3842e-06, -9.2533e-01, -1.4987e-03, -8.3022e-04, -7.2080e-04,
        -3.6925e-04, -4.3007e-04,  4.4983e-05, -5.9731e-04,  3.0117e-04,
        -3.7865e-04,  2.7878e-04, -3.2838e-04, -2.7964e-05, -1.5794e-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([ 3.4888e-03,  5.5637e-04,  6.3556e-04,  3.4149e-04,  9.1372e-02,
         1.5744e-03,  7.1534e-04,  1.0382e-03,  1.5297e-03,  2.1955e-05,
        -1.0196e-03,  8.9373e-01, -1.0044e-04,  4.9354e-04, -4.7879e-04,
         2.3778e-04,  8.2706e-04, -5.6225e-06, -1.5516e-04,  1.8750e-04,
        -4.5945e-04, -1.1987e-04, -1.3661e-05,  4.9588e-04, -4.0843e-05,
         1.8355e-04, -1.2455e-04,  5.3569e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7932e-03, -3.4813e-01, -3.0148e-04, -6.5476e-02, -4.2480e-01,
         7.3361e-05,  3.0124e-04, -2.2648e-04, -1.9703e-04,  4.5809e-04,
        -1.6871e-04, -2.0528e-04, -9.6215e-02, -4.2183e-04, -1.5660e-04,
        -9.3244e-05, -4.6792e-04,  7.9880e-04, -5.0834e-04, -5.2793e-02,
         4.9124e-05, -4.7053e-04, -1.2954e-04, -2.5426e-04,  3.2041e-05,
        -9.5295e-05, -2.0618e-04,  3.9783e-04,  1.7075e-04, -1.4965e-04,
         2.9161e-05,  4.5526e-04, -6.9066e-04,  1.8762e-04,  9.4294e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4845e-03, -3.9928e-01, -1.6038e-04, -2.2052e-02, -3.8750e-01,
         4.9405e-04, -1.1931e-04, -4.1985e-04, -1.2632e-04,  8.5063e-05,
        -5.2022e-04,  5.2005e-04, -6.4435e-02, -5.3341e-04, -5.2486e-04,
        -3.5050e-04, -8.1420e-04,  4.6257e-04, -5.1543e-04, -1.1236e-01,
         3.0775e-04, -7.8420e-04,  5.2846e-04,  2.4195e-04,  5.0753e-04,
        -3.6222e-04,  1.8613e-04,  3.2422e-04, -3.6911e-05, -6.0063e-04,
        -3.1329e-04,  2.3345e-04,  1.0464e-05, -2.6725e-04,  3.2918e-05,
        -4.2288e-04,  5.3271e-05, -2.5089e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8302e-03, -4.8762e-01, -2.0203e-04, -3.3675e-02, -2.7745e-01,
        -1.4521e-04, -5.4899e-04, -3.9466e-04,  1.7234e-04,  1.1636e-04,
        -6.1700e-05,  1.1262e-04, -5.9737e-02, -4.3433e-04, -4.6542e-04,
        -5.9262e-04, -4.7000e-04,  6.6133e-04, -5.8820e-04, -1.2786e-01,
        -2.5518e-04, -1.9531e-04,  1.0664e-03,  4.6450e-04,  7.4075e-04,
        -5.6931e-04,  7.4699e-05, -6.4136e-05,  3.9124e-05, -9.7325e-04,
        -6.9049e-05, -1.2183e-04, -3.8324e-04, -1.2410e-04,  1.8029e-04,
        -8.7243e-05,  3.2402e-06, -4.5148e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3357e-03, -7.0006e-01,  3.1130e-04, -1.4206e-03, -4.4796e-04,
        -6.1911e-03,  4.1300e-04, -2.7759e-03, -2.7897e-01, -1.0605e-03,
         6.9837e-04, -3.9056e-05,  2.8493e-05,  2.9287e-04,  5.3709e-04,
         1.1342e-04, -1.2395e-03,  2.7863e-04, -8.6355e-04, -1.8977e-04,
        -2.3245e-04, -4.9926e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  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([-1.4043e-02,  5.9079e-04, -5.0696e-04, -5.6617e-04,  3.0746e-04,
        -1.9815e-04,  9.1629e-01,  1.6504e-03,  5.3823e-04,  5.4337e-02,
         4.0827e-04,  8.9663e-04, -2.9869e-03, -3.3960e-04, -1.0450e-03,
         8.3684e-04,  5.7748e-04, -4.2693e-04,  8.8196e-05,  5.5673e-04,
        -4.3818e-04,  8.4096e-04, -8.2070e-04, -7.0632e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.4874e-03,  8.8809e-04, -1.4761e-03,  4.1362e-04,  3.5211e-04,
         1.6565e-03,  5.8347e-01,  1.2298e-03,  3.0649e-03,  3.8998e-01,
        -1.1485e-03,  3.7848e-03, -4.5673e-03,  5.4825e-04, -8.2579e-04,
         1.0071e-03,  2.6792e-04, -6.6966e-04, -4.6876e-04, -2.9715e-05,
         3.7422e-04,  2.9554e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.6135e-03, -1.3662e-04, -6.4205e-05,  1.9431e-04, -2.8716e-01,
        -6.9501e-01, -5.0539e-04,  6.0928e-04,  1.6424e-03,  6.4377e-04,
        -5.5999e-04, -1.4886e-03,  4.2419e-04, -2.1952e-04,  1.5018e-04,
        -8.4518e-04,  1.1427e-03, -8.4433e-04, -6.5831e-04,  2.3314e-04,
         6.2947e-04,  2.2441e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5177e-04,  5.7305e-05,  4.4145e-04,  1.1995e-03, -1.8741e-01,
        -8.0159e-01, -2.1787e-04,  1.7032e-03,  2.0499e-03,  9.3106e-04,
        -3.9846e-04,  1.6803e-04,  1.3886e-03, -1.5109e-04,  5.2269e-04,
        -6.4800e-04, -1.8782e-04,  5.8976e-04,  1.8940e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.5715e-04,  7.3638e-04, -4.8313e-04,  5.1243e-04, -1.7648e-01,
        -8.1375e-01, -4.5727e-04,  1.1901e-03,  1.5487e-03,  5.2748e-04,
         7.7913e-04, -9.0777e-04, -6.4636e-04, -3.7101e-04,  3.5710e-04,
        -4.6558e-04,  4.1761e-05, -2.8928e-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([-1.9266e-04, -3.5222e-05,  9.7235e-01,  2.0364e-04, -5.5980e-03,
        -4.2986e-03,  4.7393e-04, -1.4554e-03,  5.5152e-04,  2.1903e-03,
         1.7387e-03, -3.8154e-03, -1.9498e-03, -1.0076e-03,  3.0574e-03,
        -7.8451e-04,  2.9673e-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([-1.2040e-02,  2.9934e-03,  9.5069e-01,  1.9474e-03, -1.2872e-02,
        -9.2531e-04,  2.1068e-03, -4.8073e-04,  1.2195e-03, -1.6567e-03,
         2.6986e-04,  1.5475e-03, -4.7376e-03, -8.2975e-04, -3.6862e-03,
         1.0701e-03,  9.2938e-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([-4.6151e-04, -1.1907e-04,  9.7282e-01,  2.8491e-03, -8.2049e-03,
        -1.9041e-03,  4.6241e-04, -7.5949e-04,  3.7152e-03, -4.7026e-04,
         1.9462e-03, -1.3916e-03,  2.5148e-03,  1.0158e-04,  2.2823e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7286e-03,  9.5784e-01,  7.0980e-05,  3.6706e-03,  1.1445e-03,
         2.6614e-02, -1.4978e-03, -1.1725e-03, -5.7377e-04, -5.7290e-04,
         9.4647e-05,  3.8039e-04,  7.5361e-05, -6.3564e-04, -6.5312e-04,
         2.7467e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4392e-03,  9.7895e-01,  2.7090e-04, -8.3282e-04, -3.7739e-04,
         1.1086e-02, -1.7784e-03, -4.8517e-04, -1.1601e-04,  1.2969e-04,
        -4.8852e-04,  1.0588e-04, -9.5266e-04, -6.1664e-04,  2.0253e-04,
        -1.7090e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5496e-03,  8.5334e-01, -1.1546e-04,  9.3099e-04, -2.0399e-04,
         1.3814e-01, -1.5121e-03, -1.5104e-03, -1.8510e-04,  3.4613e-04,
         3.4124e-06, -2.1999e-04, -2.2899e-04,  3.6332e-04,  2.9773e-04,
        -5.1357e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1582e-02,  8.6603e-04, -2.1252e-03,  6.9571e-03, -2.4932e-03,
         2.4303e-02,  8.2180e-01, -2.3675e-03, -6.3230e-03,  8.3329e-02,
         1.9075e-03, -3.7899e-03, -7.0272e-05,  1.3154e-03, -3.5001e-04,
        -1.9549e-03,  3.3960e-03, -2.9993e-03,  2.9559e-04,  9.7255e-04,
        -8.0196e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-5.0905e-03, -9.5365e-01, -5.0044e-05, -7.1421e-05,  5.9349e-04,
         2.8553e-04,  7.6589e-04, -1.7839e-04,  6.0028e-04, -2.4087e-04,
         5.6748e-04,  2.9431e-03,  3.0309e-04,  3.9521e-04, -2.9940e-02,
         6.3232e-05, -8.5971e-05,  4.3935e-04,  1.6344e-04, -3.4070e-07,
        -8.7015e-04, -5.8234e-05, -6.2264e-04,  3.0147e-04,  6.3151e-04,
         5.7620e-04, -1.0740e-04,  4.0742e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0217e-03, -9.6849e-01, -3.3463e-05, -2.1279e-04,  3.9383e-05,
         1.0109e-03,  5.0888e-04, -6.1084e-04, -2.1114e-04, -2.3224e-04,
         9.8769e-04,  4.8019e-04,  1.1501e-03, -7.7916e-05, -2.2756e-02,
         1.4085e-04, -1.0161e-04, -4.1294e-04, -1.1241e-04, -1.4229e-04,
        -2.6916e-04, -4.3799e-06, -3.0081e-04, -8.2391e-05, -3.4481e-05,
        -1.4436e-04, -1.5451e-04, -1.7199e-04, -3.5257e-05,  7.2057e-05,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4386e-03, -9.5252e-01, -2.0740e-04,  3.6305e-04,  1.0584e-03,
        -2.9004e-04,  2.7224e-04,  7.0783e-04,  1.3575e-03, -4.5531e-05,
         1.5388e-03,  3.9400e-03,  4.9931e-04, -2.0359e-04, -2.8819e-02,
        -1.8528e-04, -3.1650e-04, -1.4728e-04,  5.3052e-04,  4.5769e-04,
        -2.6865e-05, -3.2136e-04,  4.9428e-04,  5.3477e-04,  4.5869e-05,
         5.8943e-04,  2.4977e-05, -6.2013e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5482e-02,  2.8483e-01,  4.3209e-04,  1.0025e-01,  5.5884e-02,
        -2.5181e-02,  2.1031e-02,  4.9266e-02, -5.5131e-02, -1.1015e-02,
         1.9495e-01,  7.4038e-02,  1.2144e-02, -9.1899e-03, -4.4076e-03,
         1.1775e-02,  3.7260e-02,  1.9333e-03, -7.1658e-05,  1.1157e-03,
        -3.2678e-03, -2.9505e-03,  3.5423e-03,  4.8466e-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([-4.0337e-04,  2.1794e-01, -6.6567e-03,  1.0010e-01,  4.2457e-02,
         1.5726e-02,  4.2118e-03,  1.4896e-02, -6.4967e-02, -6.5324e-03,
         3.3554e-01,  1.2998e-01, -2.2609e-03, -1.0400e-02, -1.1956e-03,
        -6.1593e-03,  1.9919e-02,  2.8359e-03, -2.9349e-03,  5.5657e-03,
        -4.5027e-03,  3.2540e-04, -2.3759e-03,  2.1119e-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([-0.0377,  0.1713,  0.0020,  0.3022,  0.0222,  0.0292,  0.0132,  0.0100,
        -0.0357,  0.0032,  0.1849,  0.0390,  0.0950, -0.0093, -0.0107, -0.0061,
         0.0157,  0.0033,  0.0015, -0.0009, -0.0026, -0.0016,  0.0028,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6777e-04, -7.3193e-01,  2.2619e-04, -2.4808e-02, -2.3768e-01,
        -1.8535e-04, -2.1756e-04,  1.7369e-03,  1.3547e-04,  3.2104e-06,
        -5.6047e-05,  2.0514e-04,  7.9518e-04,  5.1784e-04,  8.1559e-04,
         2.1800e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8755e-03, -3.8240e-01,  6.7881e-04, -1.6002e-01, -4.4802e-01,
         2.5061e-04, -2.9886e-04,  1.8314e-03,  2.1180e-04, -7.2366e-04,
        -3.6009e-05,  3.2855e-05, -4.4794e-04,  5.3332e-04,  7.6496e-04,
        -8.0825e-04,  6.3215e-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([-1.0374e-03, -2.4772e-02,  6.2624e-04, -1.2861e-01, -8.3866e-01,
        -4.5769e-04,  1.3563e-04,  9.2125e-04, -8.9682e-05,  5.2361e-07,
         5.0326e-04, -6.0962e-05, -9.5504e-04, -1.0709e-04, -3.5343e-04,
        -1.4016e-03, -2.5453e-04, -7.5606e-04,  2.9752e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.2664e-03, -6.2974e-03, -3.4268e-04, -8.4690e-04, -2.2166e-03,
        -4.8870e-04, -4.0132e-04, -4.1672e-04,  9.7440e-01,  1.4578e-04,
         1.1732e-04, -3.5155e-04,  2.6687e-04, -8.5976e-04,  4.3224e-04,
        -8.3618e-04, -3.0557e-04, -2.7914e-04,  1.0222e-03, -1.2201e-04,
        -3.5031e-04, -7.1240e-04,  2.6396e-04,  2.5486e-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([ 8.6601e-04, -4.9252e-03,  4.1289e-04,  1.8947e-04, -4.6205e-03,
         6.6554e-05, -1.6772e-03,  2.3519e-03,  9.7113e-01, -6.9411e-04,
        -1.3909e-03,  8.8053e-04, -3.1779e-04,  3.8302e-04,  2.7119e-04,
        -1.2961e-03,  2.0707e-04,  1.0160e-04,  7.2103e-04,  1.5015e-03,
         1.3497e-04,  6.1671e-04, -6.7972e-04, -6.6105e-04,  1.1071e-03,
        -6.7031e-05, -5.2606e-04,  3.0246e-04, -2.3461e-04,  1.3775e-03,
        -2.8692e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3333e-03,  6.6428e-03,  1.0051e-03,  2.1556e-04,  1.0544e-03,
        -8.1672e-04,  1.7530e-03, -1.9342e-03, -9.7848e-01,  5.1141e-04,
         6.3833e-05,  6.6214e-04, -2.4295e-04, -5.7793e-04, -1.7320e-04,
         1.4675e-04, -9.2730e-04, -9.6590e-04, -1.0884e-03,  2.1916e-04,
         1.8571e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         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([ 1.8129e-03,  9.9175e-01,  8.7416e-05, -2.0281e-04, -2.5252e-04,
        -1.3294e-03, -8.4829e-04, -3.2929e-04, -4.2469e-04,  6.3785e-04,
        -1.3201e-05, -1.8692e-04, -4.0874e-04,  3.6923e-04,  2.7779e-04,
         8.5252e-05,  1.2319e-04,  2.0118e-04, -2.5398e-04, -2.9896e-05,
        -3.7318e-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([-4.0954e-03, -9.8685e-01,  3.8816e-04,  1.3728e-04,  1.0499e-03,
         9.2866e-05,  6.3813e-04,  2.0990e-03, -5.3919e-04, -3.0076e-04,
        -3.1132e-04,  4.7979e-06, -2.5850e-04, -4.9125e-04, -1.9180e-04,
         1.7577e-04, -2.6480e-04,  1.3937e-04, -1.8490e-04, -3.4985e-06,
         6.2402e-05,  1.9734e-04, -5.2803e-04,  5.0012e-04,  1.8818e-04,
         3.0384e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.0003e-04,  9.8788e-01, -3.9727e-04,  1.3866e-04, -1.2367e-03,
        -6.2389e-04, -8.5353e-04,  4.4251e-04, -3.7577e-04,  1.0036e-03,
        -2.9827e-04, -1.3287e-04, -1.9750e-03, -1.6979e-03, -3.5306e-04,
        -6.1338e-04, -7.0523e-04,  1.5762e-04, -2.1392e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6382e-03, -9.8979e-01, -6.1974e-04,  2.1738e-04,  2.8741e-04,
        -7.0352e-04, -8.9201e-04, -8.8103e-04, -4.7347e-04, -1.0915e-03,
        -9.0832e-04,  1.4651e-05, -9.3239e-04, -7.6683e-04,  1.4724e-05,
        -7.0263e-04,  6.8076e-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([-1.8436e-03, -9.9080e-01,  3.8455e-04,  1.7128e-03,  1.0409e-03,
         4.4433e-04,  6.1318e-04, -8.9651e-05, -3.9560e-04,  7.2446e-05,
         4.0817e-04,  7.9250e-04,  4.4777e-04,  4.3694e-04,  1.6666e-04,
         1.0036e-04,  7.3924e-05,  1.7408e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2686e-03, -9.8683e-01,  5.5414e-04,  1.2362e-03, -5.6712e-04,
        -1.2034e-04,  4.9051e-03, -1.2757e-04,  4.5479e-04,  8.6230e-04,
        -7.5027e-05,  2.0250e-04, -2.8782e-05,  3.6524e-04,  3.4345e-04,
         5.9282e-04, -5.7443e-04, -3.3373e-04,  5.6251e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2336e-03, -9.8721e-01,  5.8816e-05,  1.2481e-03,  9.5263e-05,
        -2.1987e-06,  2.8454e-03,  2.6019e-03,  6.5704e-04,  1.2815e-03,
         8.2921e-04, -6.9835e-04,  4.1822e-04,  5.5813e-04, -2.1132e-04,
         5.0415e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3250e-03, -9.8688e-01, -3.7322e-04,  5.8560e-04, -5.4835e-04,
        -1.4027e-03, -2.7993e-03, -9.4141e-05,  8.1453e-04,  4.7781e-05,
        -6.8201e-05,  3.2889e-04,  1.5718e-04,  6.9565e-04,  2.9875e-04,
        -4.8677e-04,  4.6796e-04,  6.2948e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.3137e-04, -6.2170e-01, -4.8055e-04, -3.8466e-04, -7.2680e-04,
        -2.4986e-04,  3.0143e-04, -3.6375e-01, -3.6077e-04, -6.6440e-03,
        -6.0779e-04,  1.7581e-04, -3.9140e-04, -2.6408e-04, -7.9561e-05,
        -4.6708e-04, -1.0067e-04,  1.4872e-04,  2.9244e-04, -5.5019e-05,
         6.3241e-05, -5.9843e-04,  4.4922e-04,  6.3992e-04,  1.8880e-04,
         1.2422e-04, -2.7903e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4468e-03,  8.5486e-01, -4.5553e-04,  2.2830e-05, -3.0497e-04,
        -1.0001e-03, -1.2626e-03,  1.3129e-01, -3.8187e-04,  1.4707e-03,
        -1.6266e-04, -7.1170e-04,  5.8491e-06,  6.2525e-05, -1.3542e-04,
        -1.1294e-03, -4.9931e-04, -9.6905e-04, -4.0806e-04, -9.0622e-04,
         3.1270e-05,  2.6925e-04, -2.1858e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3710e-03, -6.7456e-01,  4.3032e-04,  1.8353e-04, -9.1361e-04,
         3.3691e-04,  9.8438e-04, -3.1158e-01,  5.9154e-05,  2.3074e-03,
        -1.5204e-04,  3.5960e-06, -3.4121e-04, -8.8851e-04, -4.6039e-04,
        -4.7487e-04,  7.8817e-05, -1.2024e-03, -3.0950e-05,  5.5857e-04,
         3.6172e-04, -3.9003e-04, -1.8323e-04,  1.4583e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0277e-03, -8.6082e-01,  3.8233e-04, -1.5407e-04, -2.2541e-04,
        -2.8234e-04,  3.7050e-04, -1.1965e-01,  4.9403e-05, -8.6381e-03,
         3.2775e-04,  5.2490e-04,  3.6471e-04, -7.0280e-05, -1.3380e-04,
        -6.1650e-04,  5.7534e-05, -2.5149e-05,  4.7073e-04,  5.3304e-04,
        -3.2995e-04, -3.5853e-04, -2.5699e-04, -3.1877e-04, -2.5769e-04,
         6.3133e-04,  1.2184e-04], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #550: [tensor([-0.1290, -0.1403, -0.0779, -0.0509, -0.0666, -0.0135, -0.0389,  0.0020,
         0.1985, -0.0175, -0.0690,  0.1781,  0.0044,  0.0058, -0.0076,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7200e-03, -4.7742e-01,  1.8692e-04, -7.6163e-02, -2.9156e-01,
        -5.0416e-04, -2.7429e-04, -1.4850e-01,  2.0670e-04,  3.9050e-04,
         7.3051e-05,  4.1571e-05,  2.1488e-04, -2.5379e-04, -2.9859e-04,
        -2.1966e-04, -1.4120e-04, -4.9176e-05, -6.4775e-04,  1.3545e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4978e-03, -6.4176e-01,  3.3673e-05, -1.2171e-01, -1.3011e-01,
         1.0721e-03, -3.7439e-04, -9.8949e-02,  1.0683e-03,  3.1320e-06,
         3.3079e-04,  3.8667e-04, -2.1360e-04,  2.9054e-04,  2.2985e-05,
        -3.5918e-04, -1.0682e-05, -1.4003e-03, -2.4248e-04,  1.3769e-04,
        -2.0875e-05,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7676e-03, -6.4487e-01, -3.5191e-05, -4.1147e-02, -1.1394e-01,
         1.5120e-04, -7.6823e-04, -1.9266e-01,  1.7263e-04, -2.3444e-06,
         2.6350e-04, -4.9853e-04, -2.7320e-07, -1.0617e-04, -1.9137e-04,
        -1.5844e-04,  1.4627e-04, -4.3982e-04, -2.5547e-04,  1.0992e-05,
        -2.3989e-04, -1.7294e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0894e-03,  2.8365e-01,  5.6975e-04, -2.4839e-04,  7.0919e-01,
        -5.5450e-04,  2.0780e-04,  1.3547e-03,  2.4970e-04,  5.1267e-05,
         6.9952e-04, -6.1187e-04,  9.1172e-05, -4.1206e-05,  3.8669e-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([ 1.5713e-03, -1.9084e-01,  1.6675e-03, -1.4173e-04, -7.9826e-01,
         1.0136e-03, -3.9205e-04, -6.8940e-04,  2.2908e-03,  2.5694e-04,
         4.9281e-04, -1.5179e-03, -8.3269e-04, -2.9529e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1260e-03, -5.7224e-01,  1.9723e-04,  9.0729e-04, -4.2072e-01,
         1.2736e-03,  1.2416e-05,  2.9052e-04, -5.3558e-04,  7.6953e-04,
         3.8984e-04, -7.0489e-04, -1.6789e-04, -6.3820e-04,  2.4065e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4219e-03,  9.8785e-01,  4.0899e-04, -2.3276e-03, -8.9996e-05,
         1.2490e-03, -3.4584e-04, -5.0502e-04,  4.3163e-04, -8.4796e-05,
        -1.9932e-03,  2.8851e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1354e-03,  9.8429e-01, -7.4750e-04, -5.3556e-03,  7.4365e-04,
        -8.2366e-04, -9.3093e-04, -7.4092e-04,  1.2184e-04,  4.3481e-04,
        -2.9563e-04, -5.8854e-05, -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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2813e-03, -9.8935e-01,  2.0533e-04,  2.4420e-03,  7.2252e-04,
         1.4448e-03,  3.8357e-04, -2.9229e-04, -3.4695e-04,  9.6994e-05,
        -4.6295e-04, -1.1394e-03,  5.1302e-04, -1.0486e-04, -2.1086e-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([ 1.5185e-02,  7.9010e-03, -2.3664e-03, -1.9679e-03, -2.5674e-03,
         9.4951e-01, -2.6917e-03, -2.6006e-03, -3.5182e-03,  5.1275e-05,
        -7.2708e-04, -1.2714e-03, -2.0252e-03, -1.9239e-03, -3.8082e-03,
         3.1314e-04,  6.3929e-04,  9.3501e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9507e-02,  3.6688e-03,  7.5819e-04,  2.3020e-03,  3.8583e-04,
         9.5727e-01, -1.1037e-03,  2.6273e-04,  1.3749e-03, -8.8850e-04,
        -3.9470e-03, -9.3097e-04, -1.8649e-03,  7.2196e-07, -1.2578e-03,
         1.6466e-03,  2.8271e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #600: [tensor([ 2.4017e-03,  5.5407e-04, -1.6404e-03,  5.6809e-04,  6.9880e-04,
         8.9416e-01,  9.1358e-02,  9.9776e-04, -6.8914e-04,  7.9480e-04,
        -9.0167e-04, -8.3454e-04, -1.0635e-03, -1.3913e-03,  5.5977e-04,
         2.0499e-04,  9.2477e-04, -2.5824e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9809e-03, -1.2266e-03, -6.3206e-04,  1.0564e-03, -1.1810e-03,
         9.7402e-01,  1.2999e-02, -1.2143e-03,  1.0554e-03,  3.5345e-06,
        -4.5523e-04,  8.7769e-04,  5.5109e-04, -3.6023e-04, -3.2187e-04,
        -8.5974e-04, -6.3101e-04,  5.0962e-04, -6.2166e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0532e-02, -1.2104e-03,  7.6465e-05,  4.2307e-04, -3.6569e-04,
         9.0071e-01,  8.0251e-02, -1.5622e-04,  1.2354e-03, -1.0278e-03,
         4.9276e-04, -8.2848e-05,  4.1657e-04,  3.9995e-04, -1.6480e-04,
        -1.0224e-03, -4.4987e-04, -3.5776e-04,  3.3267e-04, -2.8791e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6181e-03,  9.1561e-01,  2.2255e-04,  2.6628e-05, -4.0600e-04,
        -9.1004e-04,  4.9015e-02,  2.5670e-02, -9.7616e-04,  2.1875e-04,
         1.1983e-03,  3.3779e-04,  5.5470e-04,  5.6198e-04,  5.7272e-04,
         1.7147e-04, -1.7011e-04, -4.7532e-04, -2.8277e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6372e-03, -9.7471e-01, -9.5982e-04,  8.4589e-04,  1.1760e-03,
        -1.2329e-04,  7.7719e-03,  7.4502e-03, -2.5580e-04, -5.4227e-04,
        -2.9606e-04,  8.4517e-04, -2.1323e-04,  5.4298e-04, -2.2875e-04,
         4.0480e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.0537e-04,  9.5921e-01,  6.8310e-04, -2.1351e-04,  1.3794e-03,
        -8.9233e-04,  2.4910e-02, -6.8374e-03,  1.0750e-04,  6.7722e-04,
        -2.5861e-03, -6.4228e-04, -5.6582e-04, -8.0162e-05,  2.7063e-04,
         3.7767e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.6816e-03, -8.8728e-04,  1.1514e-04, -9.8105e-05, -2.4015e-04,
        -6.5946e-04, -1.2031e-04, -5.4337e-04,  9.6009e-01, -1.9406e-03,
        -6.9169e-04, -7.7267e-04, -1.6699e-04, -7.5221e-04, -9.6288e-04,
        -1.3846e-02, -2.7610e-04, -6.1092e-04, -6.8139e-04,  5.0110e-04,
        -5.5418e-04, -6.7241e-04, -1.4068e-04,  3.4901e-04, -1.4391e-03,
         3.5824e-04, -6.8147e-04, -3.1797e-04,  1.6562e-05, -3.1802e-04,
        -3.8191e-04, -2.8479e-04, -5.0010e-04,  3.4599e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6905e-04, -6.9166e-04, -2.0487e-05, -1.5869e-04,  3.6654e-04,
         1.6615e-04,  5.1294e-04,  4.9756e-04, -9.6849e-01,  1.4697e-04,
        -2.5241e-04,  2.4716e-04, -2.7650e-04,  1.4127e-04, -2.7843e-05,
        -2.3464e-02, -2.8196e-04, -2.2646e-05, -5.0583e-05, -5.0976e-04,
         1.5420e-04, -2.4176e-04,  2.0569e-04, -1.4412e-05,  1.5720e-04,
        -4.3128e-05, -2.0287e-04, -1.3390e-04, -2.3695e-04, -1.1814e-04,
         1.0842e-04, -6.1629e-04, -2.9883e-04, -2.8169e-05, -4.5220e-04,
        -9.1188e-05,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3078e-03,  2.2365e-04, -6.5372e-04, -7.5068e-04, -1.0362e-03,
        -2.7024e-04, -3.4998e-05, -2.6255e-04,  9.8380e-01, -2.1273e-04,
         2.8053e-04, -2.2487e-04, -1.7532e-04, -6.3110e-04,  4.4583e-04,
        -4.8011e-03,  7.9774e-05,  8.9545e-05, -5.1251e-04,  5.7236e-04,
         3.1338e-04,  1.9784e-04, -6.9126e-05, -2.3153e-04,  4.3665e-04,
         4.2325e-04,  1.2111e-04,  2.3503e-04,  3.8459e-04,  2.2387e-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([ 5.9213e-03,  2.1109e-01, -2.7661e-04,  4.6989e-04,  1.7564e-04,
         1.6699e-02, -6.9315e-06, -1.0796e-04,  3.0676e-01, -1.3078e-03,
         2.7098e-01,  6.9686e-04,  1.9256e-04, -1.0131e-04,  8.0640e-02,
         4.6736e-02,  4.8784e-04, -7.7029e-05,  6.7711e-03, -3.1476e-04,
        -5.8562e-04,  4.3029e-02,  8.4442e-04,  3.9012e-03,  1.0104e-04,
        -2.3354e-04, -3.5218e-04,  2.4049e-04, -1.2791e-04,  6.6549e-05,
         5.7951e-04, -8.8108e-06,  4.5214e-05,  3.9735e-05, -3.4705e-05,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3322e-04,  1.6456e-01,  1.4665e-04, -5.0872e-04,  3.7785e-06,
         6.6681e-03, -3.6332e-05,  2.4703e-05,  4.0349e-01, -1.2743e-03,
         2.8612e-01,  2.4684e-04,  2.3298e-04, -5.2731e-05,  8.0019e-02,
         2.8036e-02, -1.8260e-04, -5.4448e-04,  5.5264e-03, -3.0500e-04,
        -1.0384e-03,  1.5910e-02, -8.9900e-04,  1.2526e-04,  6.1290e-04,
        -1.0924e-04,  4.6438e-05, -5.2776e-04,  1.2173e-04, -2.8287e-04,
        -2.5575e-04,  7.8017e-05,  1.2644e-05, -1.6455e-05, -1.7237e-04,
         5.8426e-04, -2.8985e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2119e-03,  8.9553e-02, -8.7294e-05, -7.0171e-04, -2.7063e-04,
         3.4232e-03,  3.2513e-06,  4.9433e-05,  2.2287e-01, -8.7704e-04,
         5.0528e-01,  3.8212e-04, -4.7999e-04, -3.4148e-05,  3.0495e-02,
         8.1365e-02,  3.1734e-04, -1.1847e-04,  6.0115e-03, -4.8897e-04,
        -5.0223e-04,  4.6584e-02,  3.5911e-04,  1.3689e-03,  1.6741e-04,
         1.4971e-04, -8.5095e-04,  4.0891e-04, -2.5690e-04, -1.4571e-04,
         7.8195e-05, -2.7037e-04, -2.1517e-04, -3.6537e-05,  2.8183e-04,
         2.9968e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #650: [tensor([ 1.1057e-03, -3.0911e-04,  3.3590e-04,  4.2374e-04,  1.4032e-01,
        -7.0817e-05,  2.2911e-04,  1.0895e-04, -1.7464e-04,  3.1514e-05,
        -3.1356e-04, -4.1604e-04, -2.5692e-04,  3.9328e-05, -8.3559e-04,
         1.4673e-04,  8.5110e-01, -3.8968e-04,  1.2066e-04,  7.9795e-04,
        -2.5357e-04, -4.3717e-04, -3.5680e-04,  5.8268e-04,  4.7978e-04,
        -3.5957e-04, -2.8341e-06,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4630e-04,  9.9331e-04,  3.4736e-01,  2.1971e-04, -1.3608e-04,
         1.8940e-04, -1.1154e-03,  6.4564e-01,  5.3629e-04,  2.9351e-04,
        -8.4338e-05, -1.9471e-04, -5.8099e-04, -3.0284e-04, -4.6351e-04,
        -8.7854e-04,  2.3852e-04,  4.6865e-04, -6.2345e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5255e-03, -3.2480e-04, -5.0619e-02, -1.0833e-03,  4.4855e-04,
         1.4054e-04,  2.9234e-04, -9.3741e-01,  4.3489e-04, -1.3388e-04,
        -4.8639e-05,  5.2750e-04, -9.1430e-04, -3.7766e-04, -3.4509e-04,
        -9.6404e-05,  1.3332e-04,  9.1106e-05, -2.7765e-04, -3.2146e-04,
        -1.3680e-03, -8.5241e-04, -9.1590e-04, -3.1168e-04,  2.5204e-06,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.9334e-03,  8.6359e-05,  1.6345e-01,  6.6611e-04,  4.4024e-04,
         3.3039e-04, -1.0479e-03,  8.1953e-01, -4.6557e-04,  1.2813e-04,
         2.7722e-04, -1.5523e-05,  3.9559e-05, -2.9946e-04, -8.0099e-04,
        -5.8526e-04,  4.9758e-04,  1.3758e-04, -1.6966e-04, -2.4627e-04,
        -3.6146e-04,  8.0468e-04, -4.1800e-04, -2.6926e-05, -2.4120e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5321e-03, -4.9494e-01,  1.9979e-04, -2.4395e-01, -1.2062e-01,
         5.0316e-05, -2.9667e-05,  2.2431e-04, -2.3142e-04,  3.4056e-04,
        -2.4442e-03,  4.7623e-05,  9.5557e-04, -1.3105e-01, -1.8390e-04,
        -9.2521e-05, -1.5737e-04, -1.3479e-04, -4.4893e-04, -1.9127e-05,
        -1.2047e-04,  2.2647e-05,  4.1863e-05,  1.2402e-04, -2.4714e-04,
        -2.7428e-05, -9.7104e-05, -5.5091e-05, -2.0489e-04, -2.3679e-04,
        -1.8283e-04,  4.8191e-05,  5.9016e-05, -1.3277e-04, -1.2327e-04,
         1.4112e-04, -2.5886e-04, -2.0586e-05, -1.9341e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.4321e-04, -4.6373e-01,  2.4875e-05, -1.0530e-01, -1.6478e-01,
         1.9390e-04, -8.2439e-05,  4.2012e-04, -3.6698e-05, -1.0990e-04,
        -2.1641e-01, -9.8417e-05,  8.8023e-04, -4.3732e-02, -1.4115e-04,
        -4.2334e-04,  6.2765e-05, -1.7268e-06,  2.8241e-04, -4.8909e-05,
         1.2109e-04, -3.1291e-04,  3.3528e-04, -2.6603e-04,  1.7407e-04,
        -7.0399e-05, -1.3460e-04, -6.4468e-05, -3.9529e-04, -3.8705e-06,
        -2.2348e-04, -1.4331e-04,  2.4713e-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([ 1.3391e-03, -7.3864e-01,  2.3567e-04, -3.0478e-02, -1.7621e-01,
         1.2474e-04,  2.2855e-04,  8.1270e-05,  4.8949e-04,  6.1449e-05,
        -3.5366e-02, -9.9927e-05,  2.5841e-03, -1.0391e-02, -4.3037e-04,
         1.0162e-04, -3.6528e-05, -1.1581e-07,  1.0130e-04,  1.1945e-05,
         2.6702e-04,  6.4415e-04,  1.6277e-04,  4.9999e-04, -3.5280e-05,
         2.2779e-04, -2.3632e-04, -7.5173e-05, -1.3315e-04,  1.5482e-04,
        -2.4232e-04,  5.6530e-05,  2.4638e-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([-1.5871e-03, -1.3144e-01, -4.1654e-04,  3.1474e-04, -1.3725e-01,
        -1.2257e-04, -3.3637e-04,  4.2759e-04, -2.4408e-04, -7.2332e-01,
         4.6432e-04, -8.4385e-04, -1.7389e-04, -4.6452e-04,  3.6465e-05,
        -2.3660e-04, -6.8581e-04, -6.7191e-05,  1.0464e-04,  3.7902e-04,
        -9.4898e-05,  3.2904e-04, -1.6386e-04,  4.9777e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.0165e-04, -7.1437e-01, -5.0517e-04, -2.0073e-04,  2.1871e-03,
         1.4964e-04, -1.5947e-05,  1.1192e-03,  3.7135e-04, -2.7704e-01,
         1.4430e-04, -3.2514e-04, -4.2931e-04, -2.2621e-04, -2.1936e-04,
         3.7409e-04,  4.5838e-04,  5.2419e-04, -8.8286e-05, -2.1093e-04,
        -1.4111e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.4665e-03, -1.6513e-01, -3.0436e-04,  7.3743e-04, -3.8478e-01,
         4.5381e-04,  5.9236e-04,  7.4067e-04, -1.2936e-04, -4.3901e-01,
         2.5947e-04, -1.8312e-04, -1.2158e-04, -3.6103e-04,  1.0249e-04,
        -8.1665e-05,  7.1886e-04, -3.1068e-04, -2.7147e-04, -3.6122e-05,
        -7.7980e-06,  1.2318e-04,  7.6601e-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([-2.4205e-03,  9.1726e-01, -2.8415e-04, -3.5923e-04, -2.8550e-03,
         1.1271e-02,  2.2798e-04, -2.2850e-02,  3.3723e-04,  6.5659e-03,
        -6.9150e-04, -1.5780e-02, -1.5147e-02, -1.3942e-04, -3.2730e-04,
        -8.7311e-05, -3.0015e-04, -2.4729e-04, -2.5079e-04, -3.5381e-04,
        -3.1565e-04, -5.6918e-04, -2.8030e-04, -5.2771e-04, -3.3203e-04,
        -8.0375e-05, -6.2142e-05, -8.0121e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9038e-04, -7.2343e-01, -6.6361e-04, -3.7392e-05,  2.3646e-04,
        -2.4333e-01,  1.0525e-04, -1.3373e-02, -7.9798e-05, -1.0015e-02,
        -8.5624e-05,  4.1372e-05, -5.1763e-03,  4.2549e-04,  4.2027e-04,
         2.3900e-04,  7.0513e-04,  1.6374e-04,  1.9279e-04, -2.4956e-04,
        -8.7814e-05,  3.7360e-05, -2.0313e-05,  4.9470e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  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 #700: [tensor([ 1.5901e-03,  9.7920e-01, -2.3754e-04, -1.2227e-03, -2.6063e-04,
         7.2903e-05,  8.9049e-03, -7.3022e-04,  3.5928e-03,  6.6979e-04,
         2.1081e-04, -2.0559e-05,  5.7459e-05, -5.1400e-04,  8.5376e-04,
        -3.2660e-05, -1.4287e-04,  7.3809e-05,  4.6166e-04,  2.8506e-04,
        -5.4235e-04, -2.7715e-05, -2.9171e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.2110e-03,  8.1787e-01, -8.6429e-04,  4.2491e-04,  4.6021e-04,
        -3.9759e-04,  7.3206e-02,  1.3984e-04,  9.5411e-02,  1.3095e-03,
         1.0937e-04,  1.5007e-04,  1.0382e-04, -1.6871e-04,  1.5933e-04,
         9.6476e-05,  5.2319e-04,  8.8820e-04, -6.0975e-04, -3.1996e-04,
         5.0061e-04, -7.8131e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3200e-03,  7.7556e-01, -2.8402e-04, -3.1256e-04,  8.5345e-04,
         1.7817e-04,  1.7081e-01, -4.7257e-06,  4.3693e-02,  1.6851e-03,
        -3.7403e-04, -4.8132e-04,  2.8358e-04,  5.5381e-05, -4.8518e-04,
        -1.0607e-04, -5.2172e-04, -2.9786e-04, -2.4046e-04,  1.0125e-04,
         5.4800e-05,  2.9974e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6156e-03,  4.5502e-03,  5.2400e-04,  3.1153e-03,  8.7660e-01,
         3.0610e-03,  1.0792e-04, -1.8811e-04,  9.7008e-04,  3.7630e-04,
         1.0229e-01,  2.4994e-04,  1.0994e-03,  7.0216e-04, -3.9028e-05,
        -8.7988e-05, -1.9395e-04, -1.6442e-04, -8.5348e-04, -3.0830e-04,
        -3.1510e-04, -9.4058e-05, -6.1522e-04,  1.8715e-04,  4.3057e-05,
        -1.7331e-04,  4.6947e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.9904e-03,  1.1299e-03,  6.2420e-04,  2.3678e-03,  9.7091e-01,
         2.3018e-03,  1.0911e-03, -1.0497e-03,  4.3334e-04,  9.5741e-04,
         5.6552e-03, -8.7778e-04,  1.6060e-04,  2.8687e-04,  6.6532e-04,
         1.7365e-03,  6.7651e-05,  2.5306e-04,  1.1303e-03,  5.0539e-04,
         8.0487e-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([ 3.1084e-03, -1.0520e-03,  8.5156e-05,  9.3367e-04,  9.8245e-01,
         1.3710e-03,  8.2548e-04, -1.4272e-03, -8.0379e-04,  8.5975e-04,
         4.8374e-04,  1.8154e-05,  5.4862e-04, -3.2039e-04, -6.4515e-04,
        -7.0946e-05, -1.0051e-03,  5.4527e-04,  2.0061e-04, -8.6100e-04,
         1.0060e-03,  1.5317e-04, -6.2574e-04,  9.3227e-05,  5.0602e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9620e-03, -1.8907e-01, -1.1253e-04,  1.1028e-04,  3.8634e-04,
        -6.4922e-01,  5.8365e-04, -4.3821e-05, -1.2011e-01, -3.2322e-02,
         6.5181e-05,  4.5142e-04,  7.9565e-04, -1.0089e-04, -4.8378e-04,
        -2.3526e-04,  3.4151e-04,  4.8292e-04,  1.2482e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4344e-03, -1.4262e-02, -3.1308e-04,  6.5432e-04, -9.5547e-06,
        -9.7133e-01,  6.1973e-04,  1.6199e-04,  5.8643e-03, -1.2497e-03,
         1.7698e-04,  5.0381e-04,  2.1153e-04, -3.3053e-04, -3.4505e-04,
         5.3978e-04,  3.8814e-05,  1.3035e-04, -1.3424e-04,  4.5499e-04,
         2.2067e-04,  1.4387e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3394e-03, -7.0520e-01, -2.4130e-04,  4.3947e-04,  1.0692e-04,
        -2.5685e-01,  3.7039e-04,  3.7428e-04, -2.9667e-02, -2.1046e-03,
        -1.6541e-04, -3.9532e-04,  6.8179e-04,  8.3626e-05, -1.6443e-04,
        -3.7045e-04,  4.0818e-04,  2.5348e-04,  3.2619e-04, -1.2112e-05,
        -3.1812e-04,  1.2407e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4486e-03,  9.6045e-01, -7.2990e-04,  2.1430e-03,  2.9771e-05,
         2.7030e-02,  1.9997e-04, -1.0708e-05,  6.1775e-04, -6.4188e-05,
         2.4796e-04, -1.9247e-04, -4.7539e-04, -5.0742e-04,  3.3906e-04,
        -1.6066e-04,  6.0663e-04,  2.5289e-04, -6.0009e-04,  7.3276e-04,
         7.0922e-06,  3.3480e-04,  4.9296e-04,  1.3760e-04, -6.3513e-05,
        -6.0258e-05,  6.4935e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7790e-03,  9.7859e-01,  7.8438e-05,  3.6493e-04,  7.1423e-04,
         1.2334e-02, -7.0398e-04,  2.8023e-04, -4.0029e-04,  3.2473e-04,
         1.1681e-04,  3.5352e-04, -3.8048e-04, -2.9868e-04,  5.2189e-04,
         9.8462e-04, -8.9043e-04, -5.0175e-04,  3.2176e-04,  6.3960e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2951e-04, -9.8952e-01,  5.3063e-04, -3.8802e-04, -4.7864e-04,
         4.6173e-03, -1.4122e-04,  2.2859e-04,  1.9041e-04, -2.0457e-04,
         4.2039e-04, -3.1507e-04,  1.3161e-04, -1.9420e-04, -3.2427e-04,
        -2.5855e-04,  2.8400e-04, -4.7694e-04,  2.3566e-04,  1.7022e-04,
         1.3598e-04, -3.7480e-04, -2.2700e-05,  1.2263e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #750: [tensor([ 1.5059e-03, -1.4919e-03, -1.5725e-04, -1.0107e-02,  4.3489e-04,
        -9.2723e-04,  9.5977e-01, -3.6387e-04, -1.8369e-02, -1.6743e-03,
         2.6152e-04,  2.3953e-04, -1.1184e-04, -6.1028e-04, -1.4417e-04,
        -9.6697e-04,  7.1149e-04,  9.8337e-04, -7.0491e-04,  3.2941e-04,
         1.3849e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.9356e-04,  2.6506e-05,  1.4810e-04, -1.0734e-03, -7.7140e-04,
         4.1788e-02,  5.2144e-02, -3.7601e-04,  1.3859e-05,  8.8552e-01,
         4.1584e-06, -3.2520e-04,  1.3916e-02, -1.1822e-04, -3.7854e-04,
        -6.0889e-04, -8.7568e-05, -5.4358e-04, -9.6809e-05, -3.9859e-04,
        -3.9589e-04,  2.5405e-04, -2.6229e-04, -5.6639e-05,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1874e-03,  1.9912e-04,  2.1615e-05,  3.4027e-04, -2.5041e-04,
         4.7647e-02,  2.5575e-01,  4.9138e-04,  4.9630e-03,  4.3485e-01,
        -7.3349e-04, -2.8385e-04,  2.4828e-01,  1.9996e-04, -2.5439e-04,
        -3.0720e-04, -9.2117e-05,  2.7112e-04, -1.0827e-04, -4.8904e-04,
        -1.5588e-04,  1.2754e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0050e-03,  6.9196e-04, -1.1042e-04, -9.4478e-04,  6.4745e-04,
         4.8803e-02,  2.6693e-01, -4.5519e-04,  5.7192e-04,  6.5279e-01,
        -1.0210e-03, -4.7497e-04,  2.0893e-02,  3.4440e-04,  1.3938e-04,
         4.7182e-05, -1.2448e-04, -2.5152e-04, -8.1173e-05,  6.5834e-04,
         5.9368e-08, -2.1484e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5576e-03, -1.2957e-04, -1.0578e-03,  3.9018e-05,  5.1340e-04,
        -6.1433e-04, -9.8650e-01,  1.4604e-04,  4.1889e-04, -9.5723e-04,
         3.5921e-04, -3.3046e-04, -2.2917e-03,  6.0026e-05, -1.6792e-03,
        -1.4735e-03,  5.2492e-04, -9.0411e-07, -3.4714e-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([ 1.9585e-03,  3.3822e-04, -2.9981e-05,  5.8317e-04,  4.7612e-04,
         1.8438e-04, -9.9327e-01,  6.1054e-05, -1.0067e-04,  6.9205e-05,
        -1.3994e-04,  2.0034e-04,  4.5182e-04, -3.1571e-04,  3.0248e-04,
        -7.7031e-04,  3.5563e-04, -3.8759e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4899e-03, -2.3466e-04,  8.7188e-04,  1.1262e-04, -4.1676e-04,
        -2.9678e-04,  9.9096e-01, -1.7585e-04,  7.2491e-04,  1.8644e-04,
        -3.9568e-04, -1.7863e-04, -1.7251e-04, -2.3929e-04,  2.7589e-05,
        -4.0159e-04,  1.5876e-04, -1.0762e-04, -2.0417e-04,  4.0761e-04,
        -2.4090e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.1258e-04, -6.8402e-01, -5.1233e-05,  6.9490e-05, -2.0408e-01,
        -1.0601e-01, -2.3623e-04,  4.7404e-05, -9.2038e-04, -8.3466e-04,
        -3.3729e-04,  6.9282e-05,  2.0415e-04, -3.2085e-04,  1.1949e-04,
         2.7845e-04, -6.6890e-04,  1.3043e-04,  3.5730e-04, -5.3634e-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([ 1.6068e-03,  7.4760e-01, -2.4342e-04, -7.5498e-04,  1.6768e-01,
         7.9283e-02, -3.0852e-04,  1.4217e-04,  2.6405e-04,  1.8594e-04,
        -3.4779e-04, -2.9113e-04, -3.8115e-04,  2.6154e-04,  3.3378e-04,
         3.2380e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.6974e-04, -4.9026e-01,  2.9351e-04,  8.5394e-05, -3.6642e-01,
        -1.3935e-01,  4.5012e-04,  1.4871e-04,  1.9102e-04,  4.3357e-05,
        -1.1532e-04,  4.4885e-04,  5.1775e-04, -3.0260e-04,  6.4588e-05,
         1.0323e-04, -2.4342e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1648e-03, -7.4028e-01, -2.1256e-04,  3.5407e-03, -2.4888e-01,
         3.1699e-04, -2.6914e-04,  1.3014e-03, -3.3053e-04, -2.4979e-05,
         6.3537e-04,  1.4340e-04,  5.1912e-04,  1.3722e-04,  9.7416e-05,
        -4.2447e-04,  1.0822e-04,  3.5121e-04, -5.3413e-05,  2.0256e-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([-2.3879e-03, -7.6467e-01, -3.1512e-04, -1.7031e-02, -1.9813e-01,
         9.8382e-05, -1.1307e-04,  1.5799e-04, -1.4432e-04,  2.5489e-04,
        -2.6360e-04, -1.3512e-02,  9.2802e-05, -4.3042e-04, -7.6549e-04,
         3.7192e-04,  2.7839e-05, -3.6328e-05,  1.4261e-04,  2.9918e-05,
         1.9357e-04, -1.8847e-04,  2.4812e-04,  1.5212e-04,  1.1517e-04,
        -1.2603e-04], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #800: [tensor([ 3.7133e-04,  4.0387e-01,  4.7124e-05,  1.1142e-04, -6.5935e-04,
         8.9913e-03, -2.8409e-04, -1.6570e-04, -7.1678e-06,  4.4240e-04,
        -4.0985e-04,  1.5213e-04, -1.0920e-04,  1.5661e-01,  1.7486e-04,
         5.0369e-04,  1.1512e-04,  1.1776e-04,  1.6133e-04,  6.0902e-02,
        -3.4602e-04, -1.8917e-05,  3.6392e-01, -9.2815e-05, -1.6515e-04,
        -3.5921e-04, -5.2774e-05,  8.9871e-05, -2.6582e-04,  1.0088e-04,
        -7.2464e-05,  7.8754e-05, -2.0197e-04,  3.5251e-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([-2.4293e-04,  6.9507e-01, -2.2545e-04,  1.0412e-04, -4.4532e-04,
         6.3662e-03, -3.6152e-04, -1.1356e-04, -3.0976e-04,  2.0002e-04,
        -2.6626e-04,  3.0606e-05, -3.5784e-04,  1.6476e-02,  1.6077e-04,
         2.8306e-04, -2.3490e-04, -1.8506e-04,  1.6561e-04,  9.8495e-03,
        -2.5601e-04, -1.3629e-04,  2.6582e-01, -1.0123e-04,  4.4986e-04,
         1.3057e-04, -1.0357e-05, -1.0094e-04, -1.5371e-04, -4.8073e-05,
         9.8332e-05, -1.9459e-04, -9.9253e-05, -2.1549e-05, -1.2339e-04,
        -4.3708e-04,  7.1057e-05, -2.5942e-04,  2.4623e-05, -1.1807e-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([ 2.4018e-03, -2.8371e-01, -3.8348e-04, -7.1692e-05, -1.6652e-03,
         9.3215e-05, -7.0899e-01, -3.2209e-05, -3.1762e-04,  2.8987e-04,
        -3.0163e-05, -1.5817e-04, -1.4325e-04,  3.0323e-04,  3.8911e-05,
        -1.0277e-03, -2.9420e-04, -4.9699e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.9091e-04,  1.8270e-01, -9.3958e-04, -1.1633e-05,  4.8390e-04,
        -6.1068e-04,  8.1061e-01,  3.5126e-04, -5.3379e-04, -5.9158e-04,
        -3.0228e-04,  2.5321e-04,  9.0664e-05, -2.0289e-04,  2.7478e-04,
        -1.3130e-05, -8.8714e-04,  1.5351e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0823e-03,  7.8971e-01, -4.6509e-04, -1.4299e-04, -1.7146e-05,
         3.3093e-04,  2.0294e-01,  7.1398e-04,  5.3176e-04,  3.1148e-04,
        -6.2504e-04, -3.0697e-04,  1.1667e-04,  5.8308e-04, -4.3924e-04,
         3.5373e-04,  1.8257e-05,  3.1438e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.0361e-04, -4.3483e-01,  3.2645e-04,  1.2384e-04,  1.1194e-05,
         1.6442e-04, -1.1217e-02, -9.5108e-03, -2.4662e-04,  3.0175e-04,
        -5.0211e-03,  1.0310e-04, -1.3351e-01, -3.9965e-01,  2.6040e-04,
        -3.6303e-04, -2.0336e-04,  6.1599e-05,  1.7053e-04,  1.3084e-04,
         5.2085e-05,  2.3129e-05,  1.0697e-05,  1.3654e-05,  5.2431e-05,
        -1.7357e-04,  6.4288e-05, -2.9178e-04,  1.5253e-04, -1.3017e-04,
        -5.8369e-05, -4.0443e-05, -4.2674e-05,  2.7660e-05, -1.2943e-04,
        -1.5083e-04, -2.1553e-04,  5.8607e-05, -1.1988e-04, -7.6868e-05,
        -1.3785e-04,  6.1874e-05,  1.2441e-04, -1.4485e-04, -7.5284e-05,
        -1.7860e-04,  6.3959e-05,  2.5084e-05,  2.6562e-04,  1.7202e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1680e-04, -3.3612e-01,  5.7300e-04, -2.5233e-04,  3.6789e-05,
         3.5499e-04, -6.7767e-03, -1.3546e-01, -1.0289e-04,  2.6977e-04,
        -1.1599e-03,  3.7643e-04, -9.0095e-02, -4.2629e-01,  1.9939e-04,
         4.9070e-06, -8.1472e-05, -1.7550e-04, -1.7766e-04,  2.7672e-04,
         5.4108e-05,  1.7232e-04, -2.6528e-05,  1.2254e-04, -1.1599e-04,
         8.6746e-06, -5.3681e-05, -2.6077e-04,  9.7086e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0286e-03, -5.0497e-01,  3.4558e-05, -2.5252e-05,  2.3622e-05,
        -1.2483e-04, -7.2298e-02, -7.1521e-02,  5.2763e-05,  2.4947e-04,
        -2.9858e-02,  2.1244e-05, -5.1484e-02, -2.6632e-01,  1.1172e-04,
         8.7528e-06, -3.0152e-05,  9.2776e-05, -9.6868e-05,  6.1403e-05,
         1.1101e-04,  1.4591e-04, -3.0424e-04, -1.9968e-04,  7.9125e-05,
        -4.4484e-05,  8.7637e-05, -5.6208e-05,  8.8089e-05,  1.1335e-04,
        -1.1272e-04, -9.8337e-05,  1.4782e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4116e-03,  5.5904e-05,  1.2189e-03,  4.2313e-04, -6.1146e-04,
        -1.4018e-05, -2.7846e-01, -7.1340e-01,  1.3022e-04, -3.3789e-04,
         3.9638e-04,  2.2977e-04,  1.5167e-03,  1.9703e-04, -6.2990e-04,
         9.0486e-05,  4.3346e-04, -4.4445e-04, -8.4706e-07,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3104e-03, -1.2396e-03,  6.3073e-04, -3.4317e-04, -1.2695e-03,
        -1.4712e-04,  3.9834e-02, -9.5026e-01, -1.9123e-04, -6.1114e-04,
        -5.3135e-04,  4.1981e-04,  4.2333e-04, -6.2933e-04,  6.8703e-04,
        -4.7352e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8330e-03, -6.8705e-04,  3.1207e-04,  1.1707e-03,  3.0231e-04,
         7.7234e-04,  3.4148e-03, -9.8646e-01,  4.8847e-04,  1.5739e-04,
         5.3772e-04, -1.4067e-04,  8.4019e-04, -7.3827e-04, -5.5643e-05,
        -3.3415e-04, -4.1881e-04,  1.1106e-04, -4.9020e-04, -2.2590e-04,
        -5.0595e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6877e-03,  7.5725e-05, -1.3648e-03, -5.2308e-05,  1.1101e-03,
         1.0727e-03,  7.6629e-01,  2.2160e-01, -3.5915e-04,  4.4214e-05,
         1.2283e-04, -4.6583e-04, -1.3644e-03, -1.0480e-03,  4.1734e-04,
         6.9228e-04,  1.1741e-03,  5.4780e-05, -9.9791e-06,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  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 #850: [tensor([ 4.5114e-03, -9.9865e-04, -1.2508e-04, -3.8454e-04, -7.0500e-05,
        -5.6473e-04, -7.0351e-04,  2.1577e-04, -7.3697e-05,  6.5050e-05,
         2.3151e-04,  4.7401e-05,  9.1924e-05,  6.1935e-03,  9.7887e-01,
         8.8769e-06, -6.6783e-05, -1.2493e-04,  2.4826e-03,  1.2618e-04,
        -4.3061e-04,  1.3457e-04, -1.0692e-04,  6.0265e-05, -2.9765e-04,
        -1.6198e-04, -3.4002e-04,  2.6338e-04, -3.2661e-04, -6.7933e-05,
        -2.0537e-04,  1.2159e-04, -4.9401e-04, -1.0807e-04,  2.4799e-06,
         4.5340e-04,  1.1702e-04,  1.0402e-04,  1.2536e-04, -1.2568e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0483e-03,  1.1148e-01,  1.0746e-04,  9.4194e-04,  6.9268e-04,
        -1.6462e-03,  2.6359e-01,  3.5244e-04,  1.0215e-03,  3.4863e-04,
         1.5180e-04, -1.6069e-05, -5.0110e-05,  3.0072e-01,  2.9238e-01,
        -3.1038e-04,  2.9054e-04,  2.4747e-04,  1.3233e-02, -8.3843e-05,
         5.5574e-03,  4.4585e-04,  8.0047e-05,  4.6234e-04, -2.1996e-04,
         5.0046e-04, -4.8758e-05, -2.3313e-04, -3.4576e-04,  1.4573e-04,
         9.3523e-05, -9.7474e-05,  5.1364e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4987e-03,  3.4591e-02, -4.5114e-04,  2.8856e-04,  2.0171e-04,
        -1.3837e-03,  5.0191e-01, -3.2134e-04,  1.8248e-04,  4.7445e-04,
         2.6454e-04, -1.4581e-05,  2.4699e-05,  2.7317e-01,  9.3330e-02,
        -2.4065e-04, -1.7119e-04, -1.0358e-03,  8.4800e-02,  1.8882e-04,
        -2.0822e-04, -3.2581e-04, -3.4826e-04,  4.5286e-04, -3.1794e-04,
        -2.7893e-04,  9.5308e-05, -4.9397e-05, -4.6003e-04,  3.3532e-05,
         2.0807e-04, -1.8045e-04, -3.6469e-04,  1.0557e-04,  2.4596e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1895e-03, -7.5348e-01, -1.9526e-04, -7.2701e-02, -9.7860e-02,
        -2.8698e-05,  1.2803e-03, -1.1628e-04, -2.0374e-04, -6.8026e-05,
        -2.2831e-02, -1.7856e-04,  3.5459e-05,  1.7821e-04, -4.0480e-04,
        -1.6543e-04, -2.1301e-04, -1.0741e-03, -1.5582e-04, -1.1257e-02,
         9.6172e-05, -2.4081e-04, -3.5290e-02, -1.5347e-05, -1.0824e-04,
         1.7070e-04, -2.0527e-05, -1.2945e-06, -8.8425e-05, -5.9613e-05,
        -2.0968e-04,  8.0248e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0039e-03, -3.4864e-03,  1.5852e-04, -4.2413e-03, -6.1631e-01,
        -2.3880e-04, -2.5107e-02,  3.4521e-05,  6.9726e-06, -7.0327e-04,
        -1.8991e-02,  8.0640e-05, -5.2757e-04, -4.5468e-04, -1.9755e-04,
        -1.0677e-04, -6.4562e-07,  5.0353e-04, -9.2323e-05, -1.8406e-01,
         2.1605e-05, -5.9115e-04, -1.3771e-01, -6.5997e-05, -1.7354e-04,
        -3.3595e-05, -3.7874e-04, -2.7877e-05, -2.1066e-04,  3.2199e-04,
        -6.0733e-04, -3.2486e-04,  6.0190e-05, -1.6595e-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([ 6.8312e-03,  1.8214e-01, -1.7172e-04,  5.4157e-01,  1.6898e-01,
         7.3577e-05,  1.1112e-02,  9.5439e-05,  2.4128e-05,  2.4342e-04,
        -3.4780e-04, -5.8844e-04,  7.1615e-04, -4.1925e-04, -5.7422e-04,
        -2.3612e-04,  7.8848e-04, -2.1047e-03, -6.8736e-05,  1.1465e-02,
         6.1402e-05,  6.6182e-04,  6.9420e-02, -2.1981e-04,  1.1528e-04,
        -1.7249e-05, -7.0197e-05,  6.0584e-05,  5.4332e-06,  3.0744e-05,
        -6.2705e-05, -6.3133e-05, -6.8986e-05, -2.3367e-04, -1.1433e-04,
        -1.5302e-04,  9.3153e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4330e-04, -6.7642e-01,  1.7524e-04,  3.5888e-05,  1.9865e-02,
        -2.5578e-04,  2.4218e-03, -2.8036e-04, -1.3378e-04,  2.9546e-05,
        -2.6670e-06,  1.2764e-04, -3.5632e-04, -2.2852e-01,  7.7496e-05,
        -1.2559e-03, -4.4920e-05, -6.7039e-02, -5.1874e-04, -6.0317e-04,
        -2.1423e-04, -1.8024e-05, -6.8460e-05, -1.5624e-04, -1.8686e-04,
        -1.1117e-04,  1.7267e-04, -7.6241e-05,  7.5854e-05, -1.5111e-05,
         1.1232e-04,  1.2008e-04,  6.5902e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4921e-03, -2.6304e-01, -1.9755e-04, -5.7036e-05, -5.3311e-02,
         2.3525e-04, -4.3436e-02, -2.3373e-04,  1.2032e-04,  1.3347e-04,
         3.6007e-05,  4.9674e-04, -1.7990e-04, -4.1426e-01,  9.8764e-05,
        -2.7866e-04,  2.1263e-04, -2.1776e-01, -1.3027e-04, -1.3129e-03,
         2.3461e-05, -4.2650e-04, -1.7847e-04,  1.8520e-04,  1.2919e-04,
         2.8530e-05, -4.1214e-05, -1.1545e-04,  5.6946e-06,  2.0694e-04,
        -3.6464e-05, -1.6891e-04,  3.1953e-04, -1.0532e-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([-8.3824e-04,  1.1472e-03, -2.5916e-05, -2.4290e-04, -7.9283e-01,
        -1.5796e-04, -2.0163e-02, -1.4838e-04,  8.8528e-05, -4.5473e-04,
         2.3862e-05, -2.3850e-05, -3.7353e-04, -4.6451e-02,  1.4969e-04,
        -2.4162e-03, -1.7827e-04, -1.3000e-01, -1.6548e-04, -1.3718e-03,
        -2.9260e-04, -2.4024e-04,  1.6414e-04, -1.7614e-04, -1.0842e-04,
         1.1528e-04, -4.5421e-04,  2.9148e-04, -4.6987e-05, -3.1593e-04,
        -1.8692e-04,  3.2924e-04, -2.2517e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.5021e-04, -6.3139e-01,  7.1290e-06, -6.1409e-04, -3.5873e-01,
        -4.4666e-04, -2.6047e-04,  7.3267e-04, -5.6691e-04,  2.3882e-03,
        -6.5708e-04, -2.8309e-04, -1.8656e-04,  2.6313e-04, -3.8114e-04,
         1.8118e-04,  2.2010e-04, -5.8770e-05,  1.0671e-04,  4.8561e-05,
         4.2094e-04, -3.3705e-05,  3.3798e-04, -1.5727e-04,  2.6955e-04,
         1.0935e-04,  1.1188e-04,  8.5642e-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([-1.9614e-03,  2.4755e-01, -4.9094e-04,  2.1510e-03,  7.1270e-01,
         5.0481e-04,  6.9189e-04,  1.9545e-04, -1.4870e-04, -2.1790e-03,
         2.7481e-05,  2.8405e-02,  4.7719e-04,  1.0915e-04,  2.0514e-04,
        -7.7299e-05, -3.7155e-04, -2.4377e-04,  1.8183e-05, -5.0020e-05,
        -3.3972e-04, -5.4495e-04, -1.7035e-04,  1.1895e-04,  2.1990e-04,
         4.4654e-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([ 2.2192e-03, -5.9409e-01, -7.1535e-06, -1.2535e-02, -3.3556e-01,
        -4.3998e-04, -2.9681e-05,  2.7694e-04,  3.5860e-04, -2.4494e-03,
         2.2059e-04, -4.9872e-02, -1.7611e-04, -1.3828e-04,  8.6644e-05,
        -6.8707e-05, -1.0976e-04, -1.6850e-04,  1.0771e-04,  1.6442e-04,
        -2.4293e-04,  4.7946e-04, -1.2028e-05, -5.2612e-06,  1.6559e-05,
         1.5850e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  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 #900: [tensor([-4.7024e-04, -1.8979e-05,  3.2056e-04,  8.1007e-01, -2.2686e-04,
         2.8058e-02, -3.3201e-04, -5.9898e-05,  8.4228e-02,  3.0622e-04,
        -4.6981e-04, -7.2292e-04,  9.9243e-05,  7.0685e-02, -1.4914e-03,
        -2.7896e-05, -1.0944e-04, -1.4354e-04,  2.1286e-04,  1.3689e-04,
         3.7643e-04, -2.3660e-04, -3.5940e-04, -5.9786e-04, -2.1396e-04,
        -2.7986e-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([ 3.2850e-04, -9.2204e-01, -7.0524e-04, -1.1075e-03, -4.1867e-04,
        -5.0364e-02, -1.8709e-04, -7.1064e-04, -2.0419e-02, -4.0413e-04,
        -1.7767e-04, -1.1391e-03,  1.7322e-04, -6.4779e-04, -7.2528e-04,
        -6.6429e-05, -1.4947e-04, -2.3992e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0005e-03, -9.2948e-01, -6.5195e-04, -7.4959e-04, -1.6168e-04,
        -1.6981e-02, -5.8437e-04,  2.3562e-04, -4.5584e-02, -2.5639e-04,
        -4.2560e-04, -4.3265e-04, -1.2656e-04, -2.7619e-04,  5.8988e-05,
        -3.0968e-04,  2.7487e-04, -4.4736e-04,  1.3531e-04, -1.7123e-04,
        -4.7136e-04,  1.8206e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9500e-04, -9.3170e-01,  4.0197e-04, -2.5716e-04, -7.1649e-04,
        -6.0914e-02,  1.9032e-04, -6.1979e-04, -2.0674e-03, -4.3400e-04,
         1.8972e-04,  1.9939e-04,  2.3503e-04,  3.2384e-04, -3.2261e-04,
        -2.8456e-04, -1.7830e-04, -4.5998e-05,  2.3212e-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([-1.3014e-03, -2.8601e-01, -4.6840e-05, -8.7561e-02,  9.1827e-05,
         3.7837e-04, -2.1009e-04, -5.4236e-01, -8.1244e-05, -6.2574e-05,
        -6.7285e-06,  3.2156e-05, -5.4428e-04, -7.8603e-02, -1.0362e-04,
         1.0036e-04, -5.9256e-05,  2.1771e-04,  7.7511e-06, -4.1680e-04,
         2.0277e-04, -3.0624e-04, -2.2481e-04,  4.3049e-04,  1.1445e-04,
        -1.2891e-04,  6.3716e-05,  1.1551e-04, -1.4167e-04, -7.2987e-05,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7775e-03, -9.3524e-01,  8.4013e-05, -2.6201e-02, -3.6552e-04,
        -8.1227e-05,  4.8668e-04, -1.3020e-02, -4.8676e-04, -6.0136e-05,
         7.3280e-05, -5.0945e-04, -2.9134e-04, -1.8612e-02, -2.3335e-04,
        -3.9360e-04,  1.2923e-05, -1.6766e-04, -5.9311e-04,  7.2368e-05,
         3.8205e-05, -2.9267e-04, -2.1959e-06,  1.0714e-04, -9.1593e-05,
        -3.2153e-04,  2.0385e-04, -1.8027e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4578e-04, -8.0937e-01,  1.1329e-04, -3.3307e-02,  1.9467e-04,
         9.3606e-05,  6.1792e-04,  5.1683e-02,  2.9066e-04, -1.8161e-04,
         1.6520e-04,  5.2095e-04, -3.2292e-05, -1.0030e-01, -2.0447e-04,
        -7.7223e-05,  3.8157e-04, -2.7473e-04,  1.4597e-04,  1.4761e-04,
         2.6493e-04,  1.5176e-04,  1.6001e-04,  1.9621e-04,  1.8288e-04,
         3.6782e-05,  5.0445e-05,  1.4860e-04,  1.9269e-04, -5.7834e-05,
        -1.4921e-04,  5.9933e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0719, -0.0192,  0.0562,  0.1079,  0.0070,  0.1959,  0.0008,  0.4510,
        -0.0168, -0.0017,  0.0042, -0.0011, -0.0029,  0.0007, -0.0023, -0.0068,
        -0.0085, -0.0250, -0.0061,  0.0076,  0.0064,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5092e-02, -1.6884e-02,  7.7526e-02,  1.4117e-01,  7.7705e-02,
         2.4009e-01, -2.3239e-02,  3.0069e-01, -1.2965e-02,  3.8080e-04,
         1.0014e-02, -4.2460e-03, -7.5055e-03, -2.5198e-02, -4.3120e-03,
        -2.8347e-03,  1.2330e-02,  7.7404e-03, -8.2181e-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([ 0.0135, -0.0220,  0.0491,  0.1837, -0.0102,  0.3643, -0.0234,  0.2575,
        -0.0203,  0.0086,  0.0011,  0.0054, -0.0050,  0.0232,  0.0029,  0.0052,
         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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5001e-04, -4.1688e-01,  1.6074e-04,  3.6430e-05, -3.5886e-03,
        -2.3711e-01,  1.0602e-03,  4.9509e-04, -1.2420e-04, -2.0410e-03,
        -3.3407e-01,  1.3310e-03,  8.4448e-04,  2.0743e-04,  2.9780e-04,
        -4.3502e-06, -1.6935e-06, -1.6632e-04, -7.3145e-04,  5.3588e-04,
         1.8396e-05, -4.2242e-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([-5.1599e-04, -1.0811e-01,  2.8263e-04,  3.2229e-04, -3.3839e-02,
        -4.9518e-01, -1.2114e-04, -4.5757e-04,  2.1551e-04, -2.6308e-03,
        -3.5574e-01,  4.3320e-04,  4.0783e-04, -3.8839e-04, -1.9712e-04,
        -2.4578e-05, -1.2731e-04, -7.7318e-06, -2.1659e-04, -1.1402e-04,
         8.6934e-05, -1.8617e-04, -3.2226e-04,  6.9413e-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], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #950: [tensor([-1.4877e-04, -5.4236e-01, -1.9593e-04,  6.1399e-04,  3.0329e-04,
         5.3133e-04,  1.4452e-04,  2.5273e-04,  1.8519e-02, -4.2040e-01,
         1.8709e-04,  5.3501e-04, -3.9149e-04,  6.0196e-04,  1.0902e-02,
        -2.4010e-03,  7.7755e-05,  2.0264e-04, -6.0192e-05,  3.5742e-04,
         8.5314e-05, -2.5477e-04, -1.4910e-04,  1.3105e-04, -1.4966e-04,
         4.5231e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9690e-03, -4.5872e-01, -1.9281e-04,  1.4740e-04,  4.0890e-04,
         7.8268e-04,  3.7097e-04, -2.4820e-05, -1.3303e-01, -3.7960e-01,
        -6.7027e-05,  2.8298e-04,  5.2550e-04,  1.8384e-04,  1.0755e-03,
        -2.0712e-02,  2.9754e-04,  4.6958e-05,  1.7593e-04, -2.6623e-04,
         1.9337e-04,  1.1468e-04, -2.9831e-04, -1.8111e-04,  8.7542e-05,
        -4.0085e-05, -4.6161e-05, -7.9092e-05, -7.9022e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.6992e-05, -6.8073e-01, -1.1407e-04, -3.1916e-05, -4.1827e-05,
         1.5234e-04,  1.9785e-04, -1.7514e-05, -6.9099e-03, -3.0397e-01,
         3.9789e-04,  2.1327e-04, -2.0611e-04, -4.1732e-04,  2.8839e-03,
        -1.3292e-03, -4.9289e-04,  1.3547e-04,  4.8302e-05, -1.0045e-04,
        -8.1801e-05,  6.2876e-05, -9.4110e-06, -3.2715e-05,  2.6600e-04,
         6.5808e-06,  2.2929e-05, -4.6039e-05,  3.3555e-04,  6.9773e-05,
         4.0949e-04, -7.9523e-05,  1.0948e-05, -8.0209e-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([-6.9240e-05, -7.7624e-02, -1.2040e-04, -5.6825e-06,  2.8556e-04,
         4.6988e-04, -3.9446e-04,  3.9839e-04, -1.6391e-02, -8.9013e-01,
         3.9272e-04,  1.6078e-04,  9.2692e-05, -8.7012e-05,  2.0456e-03,
        -8.3398e-03, -5.3604e-04, -3.6250e-04, -3.1137e-04, -1.6026e-04,
        -1.6272e-04, -2.7929e-04,  2.5941e-05, -2.0856e-04,  1.1524e-05,
        -9.2190e-05,  3.8309e-04, -3.5415e-04,  1.0747e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3123e-03, -6.7207e-01, -7.1673e-05,  2.2783e-04, -3.4466e-05,
         5.6049e-05, -2.5236e-04,  4.3732e-04, -5.1686e-02, -1.8719e-01,
         7.6489e-04, -2.7860e-05, -1.2592e-04,  2.4194e-04,  5.3723e-03,
        -7.6749e-02,  3.7440e-04, -4.9757e-05,  2.2253e-04,  3.8825e-04,
        -1.2928e-04, -1.9387e-04,  3.8553e-05,  1.7409e-04, -4.3411e-04,
         3.7507e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8662e-03, -8.2797e-01, -7.3387e-05, -1.7007e-04, -6.9651e-05,
         2.6274e-04, -6.4213e-05,  3.4873e-04, -7.8383e-03, -1.0716e-01,
         5.1922e-04,  5.8732e-04,  3.6140e-04,  2.5284e-04, -1.4013e-03,
        -4.9506e-02,  1.0093e-04,  4.6677e-05,  4.2506e-05,  1.5269e-04,
        -1.0152e-04, -1.4424e-04, -3.6550e-04, -1.3093e-04, -1.7917e-04,
         1.1903e-04,  4.9493e-05,  8.5401e-05, -1.3647e-07, -3.1260e-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([ 1.3690e-03,  2.2517e-01,  2.0613e-04,  2.3359e-04,  3.2517e-04,
        -1.0791e-03, -4.5878e-05,  4.8615e-04, -1.7938e-04,  2.1140e-04,
         1.0449e-04,  8.8085e-04,  7.7560e-04,  7.3332e-01,  3.1016e-02,
        -6.7845e-04, -2.4394e-04, -2.2497e-04,  1.5434e-04, -6.5086e-04,
         1.9204e-05,  5.5595e-05,  4.7667e-06,  1.5428e-04, -1.2664e-04,
        -1.5736e-04, -5.6943e-04,  7.6686e-05, -2.1242e-04, -7.3487e-04,
         1.1876e-04, -3.5470e-04,  6.2350e-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([-1.9864e-03,  2.4194e-02,  9.0586e-05, -1.1134e-04, -8.3930e-04,
        -4.2397e-04, -6.0298e-04, -1.0365e-04, -3.1899e-04,  2.7362e-04,
        -7.6069e-04, -5.1981e-04,  9.4429e-04,  9.5303e-01,  1.0603e-02,
        -1.0487e-03,  6.9410e-04, -4.8994e-05,  2.4130e-04,  2.6238e-04,
         5.7122e-04,  2.6990e-04, -1.9898e-04,  4.0699e-04, -3.4160e-04,
         4.5626e-04, -3.8198e-04,  3.1375e-05,  2.4089e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.3710e-04,  4.2170e-03, -6.8395e-05,  1.5291e-04,  2.6163e-04,
        -4.2013e-04, -4.1729e-04, -1.3006e-04, -3.5517e-04,  1.2705e-04,
        -5.3686e-04, -4.9132e-04, -5.7917e-05,  9.8568e-01, -7.9187e-05,
        -1.4022e-03, -9.3386e-04, -2.1843e-04, -4.6589e-04, -3.1827e-04,
        -1.8027e-04, -4.5841e-04, -4.5160e-04, -2.1967e-04, -1.1309e-04,
        -5.5449e-04, -4.5429e-04, -4.2426e-04, -3.7093e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1242e-03, -6.8285e-01, -1.1780e-04,  4.1169e-04, -1.7274e-04,
        -1.3367e-02,  2.5344e-04,  2.1041e-03,  2.3724e-05,  1.1568e-04,
        -1.9626e-01,  1.1827e-05,  2.5340e-03, -3.0947e-04,  5.0320e-05,
        -1.5804e-04, -2.5891e-03, -2.8201e-04, -2.6856e-02, -1.2646e-02,
        -7.4462e-05, -1.0363e-04, -5.5321e-02, -1.0049e-04,  3.4113e-04,
         4.0065e-04,  1.8552e-04,  2.1095e-04,  8.7261e-05,  1.3558e-04,
        -6.2869e-05,  1.3974e-04,  6.4921e-05, -1.6687e-04,  2.3945e-05,
        -1.4197e-04,  8.7830e-05, -2.9355e-05,  6.7355e-05, -1.9120e-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([ 2.5811e-03, -4.8476e-01,  1.4503e-04,  2.7097e-04,  1.5334e-05,
        -2.7378e-01,  6.4490e-05,  5.9812e-04,  1.3196e-04, -1.0529e-04,
        -5.9927e-03,  1.7152e-04, -1.7301e-03,  4.7035e-05, -6.6337e-05,
         4.2193e-05, -1.0250e-01, -1.5970e-04, -1.5579e-03, -3.4075e-02,
        -2.2071e-04, -2.7237e-04, -8.9634e-02,  5.1092e-06,  9.1401e-05,
         2.3679e-04, -7.3218e-05, -1.2197e-04,  7.7193e-06, -8.5120e-06,
        -1.2986e-04,  6.9218e-05, -9.9698e-05, -4.4435e-05,  3.4697e-05,
         8.9583e-05, -6.5352e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3017e-03, -2.2720e-01, -2.5229e-04,  2.2392e-04,  2.1346e-04,
        -1.7747e-01,  3.9427e-04,  2.8963e-03,  1.9177e-04,  1.2666e-04,
        -6.9044e-02,  8.8014e-05,  1.5725e-03, -8.7567e-05,  7.8886e-05,
         6.1354e-06, -2.1489e-01,  1.9651e-04, -3.3617e-03, -5.0161e-02,
        -4.2629e-05,  6.7566e-04, -2.4571e-01,  1.1510e-04,  2.5548e-04,
         3.5390e-04,  1.1676e-04,  1.4098e-04,  4.0967e-05, -1.2587e-04,
        -2.8153e-04, -6.9444e-05, -1.4193e-05, -1.3457e-04, -1.8283e-04,
        -3.4167e-05,  1.2178e-04, -4.7066e-05,  4.0980e-05,  1.4542e-04,
        -1.8145e-04,  1.0257e-04,  1.0658e-04,  5.0222e-05,  1.0346e-04,
        -4.9515e-05], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1000: [tensor([-9.4919e-04,  7.1963e-01,  8.0023e-06,  1.2678e-02, -1.6717e-05,
         4.3614e-03,  1.5113e-02, -1.0366e-04,  1.8588e-04,  4.7686e-03,
        -8.3671e-05, -3.0042e-04,  1.0682e-04, -5.3260e-05,  2.4037e-01,
         5.4717e-05, -1.6033e-05,  5.5874e-05, -3.2078e-05,  2.5291e-05,
         2.0482e-04, -1.0642e-05, -2.8838e-04,  1.3825e-04, -7.1512e-05,
        -3.1138e-05,  4.3206e-05, -1.6519e-04,  8.6585e-05,  4.6114e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1926e-03,  1.5326e-01, -2.7658e-04,  5.8704e-03,  1.9262e-04,
        -2.2577e-03,  7.8615e-01, -4.3460e-05,  6.0878e-05, -4.9497e-03,
        -2.1966e-04, -2.7135e-04, -5.0027e-05,  1.0804e-04,  4.2015e-02,
         9.2715e-05, -6.8601e-04, -2.0995e-05,  2.3967e-04, -3.8533e-04,
        -1.9823e-04, -2.3777e-05,  1.2611e-04,  1.0726e-04, -5.5745e-05,
        -4.7473e-04, -1.1608e-04, -2.6906e-04, -2.8305e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1536e-03,  4.7489e-02,  2.3870e-05,  3.5240e-04,  1.0559e-01,
         1.3549e-01,  1.6353e-04, -8.7544e-05,  1.1059e-05,  3.7074e-04,
         6.5692e-01, -1.1402e-05,  6.8872e-04,  4.5920e-02,  8.6564e-05,
        -4.8419e-04,  2.0784e-04, -5.9077e-04,  4.3431e-04, -3.7708e-04,
         4.4577e-05,  1.0313e-04,  3.8825e-04, -1.6701e-04, -2.9243e-04,
         1.9480e-04, -2.1655e-04,  1.3958e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9833e-04,  1.4591e-02, -2.3623e-04,  3.0358e-04,  1.0411e-01,
         1.8031e-01, -1.1041e-04,  6.5569e-04, -4.8903e-05,  5.0919e-04,
         1.4468e-02, -6.8093e-05,  7.0593e-04,  6.8155e-01, -2.0498e-04,
        -2.3941e-04,  1.9353e-05, -1.6225e-04,  1.9818e-04,  2.2888e-04,
         1.0420e-04, -3.0121e-04,  1.5542e-04, -5.4966e-05, -1.6514e-04,
        -1.3960e-04,  1.5569e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7446e-03,  2.1800e-02, -4.1522e-04, -2.3892e-04,  1.2494e-01,
         2.0138e-01,  3.2320e-04, -6.7693e-06,  8.3410e-05,  1.1534e-04,
         2.0625e-01, -4.1608e-04,  7.0615e-04,  4.3900e-01, -1.7921e-04,
        -4.9573e-04, -4.0125e-04, -1.9577e-04, -1.1624e-04, -6.1067e-05,
        -1.6057e-04,  4.5975e-05, -1.5092e-04,  3.8427e-04, -2.1249e-06,
         5.4791e-05, -2.6204e-04, -5.3098e-05,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0482e-03,  9.5378e-03, -4.1353e-05,  7.7654e-05,  1.2923e-01,
         3.3865e-04, -2.7088e-04,  1.7979e-04,  6.9715e-03,  8.4979e-01,
        -3.3114e-04,  5.1900e-04, -4.9070e-04, -1.0103e-04, -1.2699e-04,
        -1.1978e-05, -2.6224e-04, -4.0104e-04,  3.4434e-05, -2.3926e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5727e-03,  4.4672e-02,  8.2712e-05, -6.8522e-04,  6.9551e-01,
        -1.7293e-04, -9.6857e-05,  1.0369e-04,  4.3503e-02,  2.1102e-01,
        -3.5784e-04, -4.1995e-04, -3.2350e-04, -3.5809e-05,  3.3122e-05,
         2.3647e-04, -2.4191e-04, -1.4942e-04,  1.7212e-05,  4.9774e-05,
         1.6122e-04, -7.1033e-05, -4.5700e-04, -2.7011e-05,  1.5495e-06,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.3837e-05,  3.4217e-02,  2.5609e-04, -1.5119e-04,  2.1616e-02,
         4.6560e-04,  1.2671e-07, -1.2624e-04,  1.1865e-03,  9.3814e-01,
        -1.6796e-04, -7.2514e-04, -1.7224e-04,  2.3256e-04, -1.1652e-05,
        -2.3361e-04,  3.6659e-05,  3.9074e-04,  1.0292e-04, -1.6550e-04,
        -1.3006e-04, -1.5354e-04, -3.2986e-04, -4.9658e-04, -2.9448e-04,
         1.3312e-05, -1.1762e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8735e-04,  7.6239e-01,  1.7918e-04, -3.8897e-04,  3.6090e-02,
        -2.7902e-04, -1.1335e-05,  3.7426e-05,  2.4797e-03,  1.9335e-01,
         1.9832e-04, -2.7744e-04, -6.5160e-04, -8.5506e-05,  2.6350e-04,
         4.0458e-06,  4.4723e-05, -2.4126e-04, -2.2764e-04, -1.9498e-04,
         4.8807e-05,  1.3274e-05, -5.4784e-04, -3.0282e-04, -1.8108e-04,
        -7.7015e-05, -1.7460e-04, -3.7119e-04, -3.2309e-04, -1.7587e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2956e-03,  4.8467e-03,  1.0444e-04, -4.5745e-04,  6.7916e-02,
         2.8586e-04,  1.9578e-04, -5.9435e-05,  4.8330e-03,  9.1432e-01,
        -1.8180e-04, -5.5276e-04, -4.5351e-04,  1.5311e-04, -1.9338e-04,
         2.0327e-04, -2.3898e-04, -7.5809e-04, -3.4132e-04,  2.0293e-05,
        -4.4620e-04,  4.6461e-04, -2.2612e-04, -4.1150e-04, -4.0251e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0110e-03, -1.3346e-02, -1.8014e-05, -2.3393e-04,  9.4300e-01,
         6.5746e-04, -3.5199e-04, -1.4558e-04, -4.2434e-04,  3.5962e-02,
         1.7970e-04, -7.7897e-04,  1.0061e-04, -2.4898e-04, -5.9580e-04,
         5.6813e-05, -2.2009e-04, -2.3032e-04,  4.4687e-04,  2.4565e-04,
         4.4710e-04,  6.8960e-06, -4.3693e-04, -1.4706e-04,  3.8670e-04,
         1.6380e-04, -1.5715e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8634e-03,  1.2731e-01,  1.2628e-05, -6.1741e-04,  8.7194e-02,
        -3.2500e-05, -2.0116e-05,  4.3815e-04,  2.1248e-03,  7.7750e-01,
        -1.3310e-04, -4.0518e-04, -4.2220e-04,  1.1812e-04, -2.7231e-04,
         4.5867e-05,  1.1976e-04, -8.9873e-06,  4.7964e-04,  8.9081e-05,
         4.0259e-04, -1.8813e-04,  3.2813e-05, -1.6919e-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>)]
Iter #1050: [tensor([-6.9871e-04, -8.6735e-05, -9.7403e-05,  6.6674e-04, -5.6954e-01,
         2.7983e-04,  3.0334e-04, -4.1768e-01,  4.4644e-04,  7.3579e-04,
         6.5297e-03, -6.5881e-04, -1.0973e-04, -4.0701e-05,  1.7647e-04,
         6.3563e-06,  4.3621e-04,  1.0859e-04,  1.6273e-04, -6.0475e-04,
         1.8231e-04,  2.1610e-04, -7.8261e-05, -4.8352e-05, -1.0509e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.3983e-04,  6.9020e-04, -4.3646e-05,  2.8803e-05, -8.0365e-01,
         4.8659e-04, -4.9379e-04, -1.8848e-01,  3.1159e-04,  7.1383e-04,
         1.7279e-03,  4.8042e-04,  4.0661e-04,  7.8610e-04, -3.1079e-04,
        -5.4091e-05,  4.4612e-04, -8.3323e-05, -1.7417e-05,  2.9598e-06,
         5.0737e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2593e-03,  1.8609e-06,  2.8756e-04, -6.5829e-04, -7.2168e-02,
        -1.8598e-04, -2.4193e-04, -9.1718e-01,  2.5043e-04,  6.2988e-04,
         4.6697e-03,  2.8329e-05,  2.4827e-04, -1.5653e-04,  6.0175e-04,
        -1.5669e-04,  6.1619e-04,  6.0098e-05, -2.2612e-04,  1.8269e-04,
        -4.9241e-05, -8.5237e-05,  5.9428e-05,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3734e-03,  1.5767e-04,  3.9901e-04,  1.4095e-04, -1.1913e-02,
        -5.4181e-04, -3.2414e-04, -9.7916e-01,  6.4168e-04,  9.0142e-04,
         1.2874e-03, -2.9723e-04, -3.0474e-05,  1.5143e-04,  6.5284e-04,
        -2.1233e-04,  2.5807e-04, -4.3417e-04,  5.5835e-04,  2.2734e-04,
         3.3431e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.7676e-04,  4.1674e-04,  1.4760e-04,  1.7846e-05,  2.6244e-02,
        -1.0910e-04, -9.2927e-05, -9.4620e-01,  2.1670e-04, -7.9202e-04,
        -2.3245e-02,  2.3159e-05, -3.9878e-04, -1.3214e-04, -2.3900e-04,
         1.4119e-04, -2.6688e-05,  4.7007e-04, -2.0166e-04,  1.0594e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7729e-04, -3.2202e-04, -3.9044e-04, -3.1729e-04, -3.3922e-01,
        -2.6393e-04, -1.3027e-04, -6.5202e-01,  5.7625e-04,  2.3439e-04,
         3.7727e-03,  1.5196e-04,  1.2364e-05, -1.4297e-04, -3.3829e-04,
         2.1749e-04,  3.9346e-05,  4.0080e-04,  1.7127e-04, -2.1432e-05,
         2.2980e-04,  5.3452e-04, -1.1774e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6965e-03, -1.1343e-04, -4.0006e-04, -5.6565e-04,  3.3115e-03,
         1.6938e-04,  4.6651e-04, -9.7211e-01, -1.4684e-04,  1.5022e-03,
         1.5967e-02,  3.6391e-04, -2.5452e-04, -5.5375e-04, -6.2628e-04,
         8.8298e-06,  5.7006e-04, -1.3177e-04,  3.4310e-04,  6.2735e-05,
        -6.3390e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3971e-04, -4.2859e-04,  2.3912e-04, -1.3657e-05, -6.7106e-02,
         2.1531e-04, -1.1516e-03, -9.1896e-01,  5.6533e-04, -5.4111e-04,
        -6.9697e-03, -3.8895e-04, -4.9266e-04, -3.3352e-04,  5.2997e-04,
         2.8607e-05,  3.3094e-04, -8.4004e-04,  1.4387e-04, -1.9063e-04,
        -9.4511e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.4854e-04,  4.7573e-04,  4.1144e-04, -1.4928e-04, -4.2919e-01,
         2.6945e-04,  3.4552e-04, -5.2764e-01,  2.5077e-04,  1.0323e-04,
        -3.9093e-02, -8.4522e-05, -5.3690e-05,  2.7845e-04, -5.7253e-05,
         8.6041e-06,  2.1953e-04, -1.0309e-04,  3.3153e-05,  1.3326e-04,
        -7.6795e-05,  7.3102e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.7445e-04, -4.0847e-04, -9.9820e-05,  3.0473e-04, -3.2828e-01,
         1.2058e-04, -1.0401e-04, -6.5999e-01,  7.0422e-05,  5.6218e-04,
         8.0917e-03, -9.9536e-05, -1.7910e-04, -2.5373e-04, -1.2103e-05,
         2.1664e-04,  8.6463e-05, -4.7298e-04, -2.3999e-05,  8.0388e-05,
        -6.5596e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0806e-04,  2.2133e-04,  1.4713e-04,  3.8415e-04, -2.1245e-02,
        -3.9443e-05, -3.9592e-04, -9.6841e-01, -5.0897e-04, -5.1723e-04,
         3.4057e-03,  2.9616e-04, -7.7648e-04, -5.2671e-04, -6.5200e-04,
        -6.7919e-05, -3.4971e-04,  1.7557e-04, -4.8738e-04,  1.6946e-04,
        -2.8010e-04, -1.4179e-04, -3.9739e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.8991e-04, -5.2084e-04, -2.4875e-04, -7.4675e-04,  1.8240e-02,
        -1.7833e-04, -2.6794e-04, -9.7040e-01,  5.2485e-04,  7.0363e-04,
         4.5208e-03, -1.0129e-04, -4.9754e-04, -2.8730e-04, -4.3157e-04,
        -3.6562e-04, -4.7632e-04, -6.0322e-04, -2.0270e-05, -9.0185e-05,
        -8.7461e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1100: [tensor([-5.0884e-04,  9.1579e-04, -4.0085e-04, -2.8690e-04,  1.6799e-04,
        -2.5719e-04,  4.2971e-03,  8.2226e-05,  5.2899e-04, -2.8187e-04,
         1.7863e-04,  1.4479e-04, -4.4340e-01, -5.4642e-01,  7.9835e-05,
         4.7271e-05,  1.8387e-04, -2.8127e-05,  2.7057e-04, -3.3311e-04,
         3.5161e-04, -9.5921e-05, -4.9265e-05, -3.1081e-04,  1.4787e-04,
        -2.1239e-04,  2.4008e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4591e-03, -3.9850e-03, -7.8746e-05,  4.0056e-04, -9.6375e-04,
         2.1736e-04, -6.9831e-04,  2.1773e-05,  6.0361e-05,  1.9846e-04,
         2.3151e-04,  1.2891e-04,  3.5588e-01,  6.3091e-01, -1.8445e-05,
         1.4954e-04, -1.8394e-04,  6.8466e-06,  3.5255e-04, -3.4580e-04,
        -2.0246e-04, -1.4700e-05, -1.9183e-04, -2.1799e-04,  5.3013e-04,
        -3.0118e-05, -2.6670e-04,  8.0138e-07, -2.6141e-05, -1.9899e-04,
         2.7634e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5287e-03,  7.5881e-03,  3.6191e-04,  1.9471e-04, -2.0421e-04,
         1.6882e-04,  5.4838e-03,  9.1325e-05, -5.6647e-06,  3.0481e-05,
         3.4782e-04,  1.0512e-04, -2.4427e-01, -7.3558e-01,  5.3695e-04,
         4.9681e-04,  7.3801e-04, -3.5704e-04,  1.8022e-05,  4.1904e-04,
         2.3517e-04, -1.2780e-04,  8.9553e-05,  3.4939e-04, -3.4440e-04,
         3.2332e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1266e-04,  1.1315e-02,  5.4645e-04,  1.1151e-03,  4.6283e-03,
         1.6246e-03, -5.0963e-01, -1.7660e-04,  4.4500e-04, -3.6602e-01,
        -1.0001e-01,  4.8786e-04,  3.2499e-04,  1.3570e-06, -4.4900e-05,
         3.1359e-04,  1.2150e-03, -2.6103e-04,  5.6295e-04, -1.0539e-04,
        -3.0456e-04, -4.4104e-05,  2.1529e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7596e-04,  1.3299e-02,  3.9843e-04,  3.8791e-04,  7.8733e-03,
         2.5619e-03,  1.0156e-02,  1.2595e-03,  6.6014e-05, -4.9247e-01,
        -4.6400e-01,  3.0747e-04, -3.3072e-04,  1.1888e-03,  5.6127e-04,
         4.9178e-04,  1.2667e-03,  2.1640e-04,  4.8528e-04,  6.8378e-04,
         1.0135e-03, -1.6649e-05,  2.7787e-04, -1.4483e-04,  1.7338e-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([-3.8800e-03,  9.7548e-03,  5.3869e-04,  1.7568e-04,  2.3432e-03,
         1.1291e-03,  3.7049e-02,  8.7600e-05, -4.6858e-05, -8.8765e-01,
         5.2434e-02, -4.5126e-05, -2.1625e-04, -9.6557e-05,  3.4225e-04,
         3.8417e-04,  1.2541e-03,  5.6554e-04,  4.9152e-04,  8.7316e-04,
         1.6000e-04, -4.9910e-05,  4.3001e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.8789e-04, -9.9441e-01,  2.0829e-04,  2.2496e-04, -6.3752e-05,
        -1.9338e-04,  3.2103e-05,  8.6576e-04,  5.5314e-04, -2.1181e-04,
         8.9772e-05, -7.4873e-05,  2.3192e-04,  1.0055e-04, -2.3737e-05,
         5.0886e-05,  3.5090e-04, -5.4371e-05, -1.2660e-04,  3.7515e-04,
        -3.6262e-05, -6.5010e-05,  3.3980e-04, -2.2892e-04, -9.5817e-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([-8.9404e-04,  9.9401e-01, -5.6327e-04, -2.4439e-04, -6.2597e-04,
        -2.5458e-05,  7.8365e-04, -1.4391e-04, -2.4792e-04, -3.2573e-04,
         1.2556e-04, -7.8445e-04, -2.4019e-05, -1.2925e-04,  1.0936e-04,
        -6.0286e-04, -2.2228e-04,  1.0763e-04,  2.6161e-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([ 6.7641e-04, -9.9275e-01,  2.8611e-04,  4.5883e-04,  1.0087e-03,
         3.3904e-04,  6.7191e-04,  7.9464e-05,  2.0560e-04,  1.6241e-04,
         5.7398e-04, -4.3408e-04,  6.0869e-04,  2.1127e-04, -6.9864e-05,
         5.4728e-04,  3.3965e-04, -2.6108e-04,  1.8320e-04,  1.3091e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5735e-03,  2.5341e-01,  3.5973e-04,  6.9066e-04,  4.8499e-01,
        -8.4276e-04, -5.1463e-04,  4.1930e-03,  5.3732e-04,  1.6939e-01,
        -9.4685e-04,  7.9277e-02, -1.7745e-04,  2.5891e-04,  4.5369e-04,
        -3.3128e-04, -3.9287e-04,  2.4036e-04,  2.5362e-04, -3.2488e-04,
         4.2734e-04,  1.7107e-04, -1.2979e-04,  1.2029e-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([-3.1283e-03,  6.7085e-01, -6.5359e-04,  4.6650e-04,  1.0757e-01,
        -2.8299e-04, -1.1513e-03,  1.5923e-03, -5.4839e-05,  1.4159e-01,
        -6.1648e-04,  6.5327e-02,  6.9719e-04,  3.8842e-04, -1.2556e-03,
        -1.0953e-03,  1.1306e-04,  4.7009e-04,  1.2597e-04, -2.1084e-04,
         2.0074e-04, -9.3218e-05,  9.1557e-04,  2.1219e-04,  1.4833e-04,
         5.9056e-04,  1.8921e-04, -1.8428e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0740e-03,  1.9172e-01,  7.4711e-05, -3.8589e-04,  1.8652e-01,
        -7.9143e-04, -1.3446e-03,  2.4394e-03, -6.6353e-04,  5.6189e-01,
        -5.8673e-06,  4.6235e-02, -2.1323e-04,  7.4483e-04, -9.1060e-04,
        -3.1874e-04, -9.2711e-04, -2.4290e-04, -1.4812e-04, -1.8651e-04,
        -4.3004e-04,  2.0393e-04, -4.4596e-04, -1.2309e-03,  6.6466e-04,
         7.3881e-05,  1.1523e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1150: [tensor([ 3.9650e-05, -2.2273e-01, -3.7967e-05,  8.5700e-05, -2.4430e-04,
        -5.8161e-02, -4.1242e-01, -1.7165e-04, -4.1668e-04, -1.3218e-01,
         5.7522e-05, -3.9306e-04,  7.5343e-05, -1.7117e-01,  1.6212e-04,
        -2.7910e-04,  7.1116e-05, -7.0614e-05, -6.8028e-05,  6.7838e-08,
        -1.0502e-04, -1.6581e-04, -1.6362e-04, -3.9932e-04,  1.6934e-04,
        -4.9744e-05, -1.2070e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3460e-03,  3.7407e-02, -1.6203e-04,  2.2389e-03, -8.1439e-05,
        -1.2771e-01, -3.3673e-01,  6.1865e-05,  6.0678e-03, -3.8525e-01,
         6.2271e-05, -6.6806e-04,  5.9021e-04, -9.9628e-02,  5.3570e-05,
        -2.5060e-04, -6.6529e-05,  2.4951e-04,  1.2014e-04,  2.4193e-05,
        -2.3455e-04, -7.8302e-05, -1.7597e-04, -7.8359e-05,  2.9850e-04,
         3.6571e-04,  4.6679e-06,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1272e-03, -7.3824e-03, -1.8354e-04,  5.8265e-04,  5.9103e-05,
        -2.1244e-01,  1.8500e-02, -6.8362e-04,  3.2504e-03, -2.8888e-01,
        -6.3309e-05, -1.1778e-03, -3.4624e-04, -4.6251e-01,  2.8695e-04,
        -8.6088e-04, -3.3075e-04, -2.5999e-04, -2.6053e-04, -2.0549e-04,
         3.1287e-05,  1.5390e-04,  7.1943e-05, -3.2868e-05, -9.8175e-05,
         1.4316e-04,  7.5869e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9717e-03,  3.5160e-04,  1.5865e-04, -5.7422e-04,  1.7422e-01,
        -6.0325e-05,  1.2091e-03,  7.8239e-01,  8.3650e-04,  2.6790e-04,
         3.0901e-02,  8.8541e-04, -2.6257e-04,  1.6065e-04,  2.0447e-04,
        -2.7134e-04, -5.0321e-04,  8.4664e-04,  1.8304e-04, -1.4713e-05,
        -5.0792e-04,  6.2346e-04, -2.6690e-04, -3.3414e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4888e-03, -3.9977e-05,  1.7002e-04,  9.9770e-04,  4.1898e-01,
        -1.6731e-05,  3.9996e-04,  5.5209e-01,  2.0993e-04, -2.3863e-04,
         1.8386e-02,  1.3702e-03,  1.7437e-04,  5.8766e-04,  7.3021e-04,
        -5.1698e-04,  5.5905e-05, -4.8207e-04, -1.9711e-04, -2.1557e-04,
         5.0910e-04,  1.4778e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.6949e-03,  3.2226e-04, -1.9509e-04,  3.9280e-04,  4.3083e-01,
        -3.8073e-04,  4.6908e-04,  5.2256e-01,  3.4681e-04,  2.1375e-04,
         3.4107e-02,  5.4769e-04,  7.2532e-06,  6.7439e-04,  3.4417e-04,
         1.5682e-04, -1.6578e-04,  4.4749e-05,  5.3008e-04,  3.9864e-04,
        -4.3525e-04, -3.6041e-04,  3.2506e-04,  4.9597e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3843e-03,  3.4225e-04,  1.1900e-04,  8.6514e-01,  6.4637e-04,
        -5.2155e-06,  6.5499e-05,  2.7359e-05,  1.3730e-04,  8.2568e-04,
         6.2459e-04,  5.8385e-03, -9.2771e-06,  3.6115e-02,  5.5538e-05,
        -7.5187e-04,  5.0294e-02,  2.1651e-05,  8.4950e-04,  2.0888e-04,
         2.2706e-04, -3.5570e-04,  3.0248e-02,  6.3124e-04,  3.0441e-04,
        -2.8412e-04, -2.2317e-05, -1.1346e-04,  2.1704e-04,  5.7434e-04,
         4.2169e-04, -1.3598e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.1960e-03, -2.0558e-04,  3.3950e-04, -9.1865e-01,  4.3121e-04,
        -9.8729e-05, -1.7231e-04,  8.0041e-04, -2.7015e-05,  2.4453e-03,
        -2.6286e-04,  1.0420e-02,  4.1564e-04, -4.1546e-02,  5.5613e-04,
         1.2115e-03,  2.4021e-03, -4.3377e-04,  3.1183e-04, -4.1450e-04,
         2.6000e-03, -5.3938e-04,  6.7957e-03, -4.7401e-04, -1.0663e-05,
         3.9211e-04,  1.4145e-04,  8.3742e-06, -4.6710e-05,  3.0240e-05,
        -4.2753e-04, -3.1566e-05,  1.6173e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5363e-04,  3.0894e-04,  5.1243e-05,  9.6008e-01,  3.4801e-04,
        -7.8529e-05,  3.3986e-04, -2.7409e-04,  1.8756e-04,  3.7946e-03,
         4.1118e-05,  1.7430e-02, -3.6460e-04, -1.1952e-03,  2.3098e-04,
        -5.2850e-04,  2.0830e-04,  5.9520e-05,  6.9311e-05,  5.1520e-04,
        -4.2766e-04, -4.7979e-04,  1.1333e-02, -1.4988e-06, -6.9295e-05,
         1.2434e-04,  9.0390e-05, -1.5181e-04,  3.9089e-05, -1.2751e-04,
        -1.7099e-06,  3.6441e-04, -2.9432e-04, -1.3157e-04, -1.5959e-06],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.5084e-04, -8.3985e-01, -9.4779e-05, -1.5995e-04,  2.4723e-04,
        -1.9045e-04, -1.4058e-01, -1.4466e-02, -1.8224e-04, -3.0945e-04,
        -3.4185e-04, -3.2507e-04, -4.9157e-04, -2.8839e-04, -2.4493e-04,
         1.8370e-05,  9.2964e-05, -5.3401e-04, -4.1247e-04, -4.1511e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.4303e-04, -2.1051e-01,  1.6707e-06,  9.6842e-05,  1.0365e-03,
        -3.5396e-04, -4.2981e-01, -3.5483e-01, -2.4192e-04, -1.1018e-04,
         3.5488e-04, -1.3340e-04, -1.5222e-04,  2.2663e-04, -3.4026e-05,
         3.6787e-05,  4.9703e-04,  2.8356e-05,  6.6758e-05,  2.4615e-04,
        -3.1170e-04,  6.3157e-05, -1.3900e-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([-2.4655e-04, -6.8505e-01, -8.3798e-05,  5.3497e-05, -1.4004e-04,
         4.9249e-05, -1.7331e-01, -1.3829e-01, -6.2457e-05,  8.8226e-05,
        -4.0605e-05,  1.6555e-04, -1.4357e-04,  2.5479e-04,  2.3888e-04,
        -1.9073e-04,  4.2065e-04, -1.9878e-04,  3.4552e-04, -7.4042e-05,
        -4.8594e-04, -6.4435e-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>)]
Iter #1200: [tensor([ 3.2152e-03, -2.9755e-01, -1.2864e-04,  5.4640e-04, -7.6361e-02,
        -4.5117e-05,  3.7705e-03, -1.2824e-04, -4.6421e-01, -1.7979e-04,
        -1.2750e-04,  2.6960e-06,  1.5020e-05, -1.1725e-04,  2.5215e-04,
         2.1382e-04,  1.0021e-03, -6.2819e-04, -2.0743e-04,  5.1438e-05,
        -1.4931e-01,  1.0946e-04, -1.1410e-04, -1.2262e-04,  2.1548e-04,
         1.4507e-04, -1.0235e-06,  3.1018e-05,  2.2678e-04,  2.0828e-04,
         5.5117e-05, -1.6811e-04, -6.5803e-05,  1.7054e-04,  1.7276e-04,
         7.9847e-05,  4.6175e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.4354e-04, -7.5775e-02,  5.6736e-05,  3.1616e-04, -4.5913e-01,
        -2.6374e-04,  9.9426e-04, -3.8876e-04, -2.8756e-01,  1.1123e-04,
        -1.3367e-05,  2.0321e-06, -4.8026e-07, -3.6354e-04, -1.4049e-04,
         8.6202e-05,  2.2899e-04,  2.7997e-04,  2.1817e-05,  4.5825e-05,
        -1.7256e-01,  2.0756e-04, -6.0605e-05, -7.7024e-05,  8.8472e-05,
        -2.3887e-05,  9.0780e-05,  1.7408e-04,  1.0614e-04, -5.6195e-05,
        -5.7267e-05, -3.1101e-05, -7.8919e-05,  4.9584e-05,  4.5719e-05,
         1.4848e-04,  2.1188e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1439e-03,  5.9218e-01, -4.4389e-05, -1.3400e-04,  2.9615e-02,
        -4.6695e-05, -2.9412e-03,  2.0941e-04,  2.9935e-01,  2.0422e-04,
        -1.7948e-04, -1.9759e-04,  2.5263e-04, -3.0867e-04,  9.3649e-05,
         9.4886e-05, -7.2916e-06,  1.2229e-04, -1.9008e-04,  3.4986e-05,
         7.1346e-02,  9.2284e-05,  4.2971e-06, -2.1372e-04, -2.6876e-04,
         6.6180e-05,  1.9782e-05, -4.0640e-05,  1.5009e-04, -1.4839e-04,
         2.4437e-05,  6.8505e-05,  4.2822e-05,  1.3014e-05, -5.6593e-05,
         5.0446e-05,  4.0722e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6664e-04, -3.8298e-01,  2.7316e-05,  4.2455e-04, -1.6015e-01,
        -2.5806e-04,  6.0453e-04, -8.5447e-05, -1.4539e-01,  6.2650e-05,
         1.5533e-04, -4.3848e-05,  1.8447e-04, -1.5136e-04, -1.1038e-04,
        -1.4033e-04,  4.0604e-04, -3.8851e-05,  1.1335e-04, -5.6614e-05,
        -3.0729e-01, -5.1901e-05, -1.1453e-04, -1.2150e-04,  2.6378e-05,
         1.0401e-04,  2.3769e-04,  8.7463e-06, -4.5941e-05, -1.2086e-05,
         6.5588e-05,  1.7388e-04, -6.0414e-05, -5.3253e-06,  1.2978e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1606e-03, -5.2890e-01, -1.5272e-04,  2.5629e-04,  1.1067e-02,
        -1.5569e-04,  4.2143e-03, -1.0833e-04, -1.8299e-01,  1.6873e-05,
        -1.9777e-04,  1.4737e-04, -2.4405e-04, -1.0277e-04, -1.5874e-04,
        -1.7321e-04,  1.4963e-03, -1.1268e-04, -1.8230e-04, -8.9452e-05,
        -2.6647e-01,  9.7386e-05, -7.4243e-05, -8.4162e-05, -8.3774e-05,
        -6.9507e-05,  1.4629e-04, -4.8246e-05,  2.2541e-04,  1.3764e-04,
         5.6036e-05, -1.4844e-04,  6.3133e-05, -2.0886e-04, -6.7356e-05,
        -6.5454e-05, -2.9018e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2056e-04, -4.2846e-01, -2.4019e-04,  6.4050e-05, -1.0438e-01,
        -1.9300e-04,  1.6136e-03, -1.1112e-04, -2.5483e-02,  1.3515e-05,
        -5.6569e-05,  1.0350e-04, -5.6116e-05,  1.6078e-05, -8.2740e-05,
        -9.6731e-06,  9.5260e-04, -2.3445e-04, -1.9762e-04,  3.2777e-05,
        -4.3521e-01, -2.2400e-04, -6.4452e-05, -2.6816e-04,  1.7103e-04,
        -3.3465e-05,  2.2358e-04, -1.0340e-06,  1.6416e-04,  1.0074e-04,
        -1.1631e-04, -7.7355e-05, -2.0773e-04,  3.7355e-05, -7.5987e-05,
         1.4591e-04, -1.5642e-04,  1.0953e-04,  1.9973e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7528e-03, -6.4555e-03,  9.3491e-05, -7.9591e-04,  8.4871e-02,
        -8.3654e-05, -9.2073e-04,  3.4027e-04,  5.8386e-01,  7.7452e-05,
         2.3960e-05, -3.2341e-04,  1.1247e-05, -3.4193e-04, -1.1764e-04,
        -1.3322e-04, -8.6751e-04,  2.7114e-04,  6.2234e-05,  3.4333e-04,
         3.1352e-01,  7.3320e-05, -7.4803e-05, -1.3291e-04, -9.5170e-05,
         9.7648e-05,  1.1032e-04, -2.7543e-04,  1.6170e-04, -4.1938e-04,
         6.2206e-05,  7.3613e-05, -1.7841e-04,  8.0340e-05,  1.4853e-04,
         2.6263e-04, -3.5914e-04,  1.0761e-04,  2.2586e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.8587e-04, -2.1817e-01,  1.0159e-04,  1.6262e-03, -7.4179e-02,
        -2.6736e-04,  1.7857e-03, -7.6735e-05, -3.6212e-01, -1.8004e-05,
         4.4962e-05,  1.7212e-04,  1.8043e-05,  9.4687e-05, -6.5239e-05,
        -4.0588e-05,  6.2304e-04, -2.2477e-04, -6.0454e-05, -6.3371e-05,
        -3.3776e-01,  1.1548e-04, -1.8640e-04, -1.9902e-04, -3.4350e-05,
         2.3454e-04,  1.5427e-04, -4.9693e-05,  1.7352e-04, -4.4752e-05,
        -7.3609e-07, -2.3353e-04, -1.0739e-05, -5.6273e-05,  1.1750e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.1112e-03,  4.4471e-01,  8.8341e-06, -7.6179e-04, -8.1276e-03,
        -2.3643e-04,  1.8994e-04,  2.5986e-05,  1.9374e-01, -5.0481e-05,
        -1.0727e-04, -2.8783e-05, -1.4625e-04,  1.4184e-04,  8.2782e-06,
        -4.4484e-05,  2.1424e-03,  1.5764e-05,  1.1314e-04,  4.9735e-04,
         3.4159e-01,  2.2160e-04,  1.0601e-04,  3.0551e-04, -2.2800e-05,
        -1.2546e-04, -1.9811e-04,  6.8130e-05, -2.7010e-05, -2.7806e-04,
        -1.4468e-04, -4.9052e-05, -2.3989e-05, -1.9511e-04, -8.5374e-05,
        -1.4571e-04, -2.7736e-05, -8.0339e-05, -1.0178e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6307e-03, -7.8537e-01, -6.0603e-05,  7.2764e-04,  1.1078e-02,
        -3.0104e-04,  1.2793e-03, -3.4813e-04, -2.9587e-02,  1.0283e-05,
        -1.8225e-04,  2.8127e-04, -1.7273e-04,  2.6892e-04,  2.4018e-04,
        -1.4666e-04,  3.6588e-03, -4.6667e-05, -7.7961e-05,  2.4005e-04,
        -1.6205e-01, -5.7999e-05, -3.3843e-05, -1.8205e-04, -1.7909e-05,
         8.8632e-05,  1.0989e-04,  2.0551e-04, -1.4219e-04, -2.9303e-04,
         1.1492e-04, -1.7575e-04, -2.8725e-04, -5.3170e-05,  6.6169e-05,
         2.1992e-04,  1.5972e-04, -3.3475e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3108e-05, -1.9513e-01,  6.6112e-05,  1.5017e-03, -8.1091e-02,
        -1.0725e-04,  1.5523e-03, -1.7924e-04, -6.3371e-01, -1.7696e-05,
        -4.7697e-05,  7.1945e-05,  1.2462e-04, -1.7292e-04,  5.3683e-05,
         2.0927e-04,  7.8827e-05, -1.1285e-04, -1.6944e-04,  4.6972e-04,
        -8.3121e-02, -7.2237e-05, -1.5355e-04, -2.8995e-04,  3.2843e-04,
         1.6931e-04,  4.6681e-05, -8.6362e-05, -2.9686e-05, -5.4800e-05,
        -5.7011e-06, -1.0385e-04,  1.5423e-04, -2.2050e-04,  1.6567e-04,
         9.6809e-06, -1.6344e-05,  4.6623e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4843e-04, -2.5720e-01, -3.4133e-05, -3.7556e-04, -1.3833e-01,
        -9.5841e-05,  1.3648e-03, -1.8148e-04, -3.6848e-01, -2.8020e-04,
        -1.5022e-04, -6.6141e-05, -1.2827e-04, -1.2051e-04, -1.4067e-04,
        -6.3081e-06,  6.2340e-04, -1.9797e-04,  4.9948e-05, -6.4245e-05,
        -2.3076e-01, -3.6165e-06, -1.3083e-04, -1.1508e-04,  2.0342e-04,
        -2.6061e-05,  1.2561e-05, -4.2379e-05,  7.1973e-05, -1.5704e-04,
         6.0533e-05, -5.4100e-05, -5.2275e-05, -5.5151e-05, -4.0084e-05,
         1.6112e-05,  6.8613e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1250: [tensor([ 1.6148e-03,  4.1777e-03,  4.9789e-04,  3.0611e-04,  3.0642e-04,
         3.7538e-02,  7.2828e-01, -6.2313e-05,  3.7875e-02,  3.8867e-04,
        -3.3265e-04,  1.8506e-01,  3.0761e-04,  2.3820e-04,  3.2975e-04,
         2.1961e-04,  1.4430e-04, -7.8133e-05,  2.0697e-04, -2.3394e-04,
         1.0058e-04,  4.2115e-04,  2.8991e-04,  3.1872e-05,  2.7930e-04,
         2.2369e-04, -1.6621e-05, -1.7990e-04,  2.5350e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1871e-03,  3.0317e-02, -3.0904e-04,  3.1319e-04,  7.4135e-04,
         4.5483e-01,  1.0295e-01, -2.7074e-04,  9.3874e-02,  4.8800e-04,
         1.6172e-02,  2.8148e-01,  2.6426e-05, -1.0961e-04,  3.7676e-05,
         8.1114e-03,  7.3279e-04, -3.2976e-04, -5.1600e-04, -1.6646e-03,
        -7.6795e-04,  5.1700e-04, -1.1521e-03,  4.7246e-05,  4.1886e-05,
        -1.9337e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1242,  0.0526,  0.1430, -0.0135,  0.0712,  0.2459, -0.0006, -0.0267,
        -0.0158, -0.0016,  0.0159,  0.0704, -0.0183,  0.0480, -0.0103, -0.0529,
         0.0430,  0.0039, -0.0023,  0.0196, -0.0202,  0.0000,  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, -0.0070,  0.0517, -0.0366, -0.0248,  0.4829, -0.0425, -0.0046,
         0.0879, -0.0016,  0.0115, -0.0179, -0.0240,  0.0877, -0.0183, -0.0048,
         0.0353,  0.0124,  0.0047,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8872e-01,  3.0781e-02,  5.6168e-02, -1.3567e-02,  2.1890e-02,
         1.3654e-01, -5.1133e-02, -1.0926e-02,  2.4178e-01, -5.1840e-05,
        -1.4563e-02, -1.0482e-02,  1.5847e-02,  9.9937e-03,  6.1547e-02,
         1.7456e-02, -4.1948e-03,  1.3570e-03,  4.5840e-03,  8.4138e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3476e-03,  6.6888e-02, -1.0980e-03, -9.2280e-04,  1.5233e-02,
         4.2607e-03,  1.4971e-03,  8.9039e-01,  3.2745e-04, -2.2495e-04,
        -3.2677e-04, -1.7229e-03, -1.7092e-04,  2.3514e-04,  4.6540e-04,
         1.3782e-03, -2.3948e-04, -1.9394e-04, -8.0339e-04,  3.8355e-04,
         3.7487e-04,  4.6055e-05,  3.6256e-04, -6.4444e-04,  7.7356e-04,
         5.0151e-03,  3.1642e-04, -9.0734e-04,  4.2402e-04,  3.1651e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0007,  0.2545, -0.0106,  0.0051,  0.1267,  0.0092,  0.0021,  0.5525,
         0.0093, -0.0018, -0.0033, -0.0137, -0.0013,  0.0044,  0.0026,  0.0006,
        -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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0337e-02,  5.7694e-01, -1.5306e-03,  2.5525e-04,  7.8389e-03,
         4.9266e-03,  5.3739e-03,  3.6500e-01,  1.5186e-03, -4.7398e-04,
        -1.6220e-03, -6.0476e-03, -1.4895e-04, -3.9522e-04,  1.9360e-03,
         5.8053e-05, -4.2048e-03,  2.5706e-04,  5.2836e-04,  5.9140e-04,
        -1.6818e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0006e-03,  9.7633e-01, -2.6303e-04,  7.6780e-04, -1.6099e-03,
         2.2602e-04,  9.7691e-03,  4.2584e-04, -4.9509e-04, -1.1026e-03,
        -5.6784e-04,  2.2613e-04, -1.7370e-04,  2.8398e-04,  1.1935e-03,
         9.3502e-04,  4.7513e-05,  2.3755e-05, -1.1324e-03, -2.7785e-04,
        -7.2167e-05, -4.1153e-04, -6.6250e-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([ 2.2950e-02,  3.0855e-01, -6.1509e-03,  6.0129e-02, -8.1346e-03,
        -2.4684e-03,  5.6157e-01,  8.6597e-03, -3.2838e-03, -1.4338e-03,
        -4.0136e-03,  1.2992e-03,  2.8508e-03, -5.9445e-04, -1.1952e-03,
         1.7745e-04,  1.7467e-03, -3.4172e-03,  6.3198e-04,  1.3658e-04,
        -2.5808e-04,  3.5342e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3143e-03,  4.2264e-01, -3.2209e-03,  2.3128e-04, -1.4720e-02,
        -6.8216e-03,  5.1832e-01, -2.6515e-04, -4.5077e-03, -1.8741e-03,
        -3.3928e-05,  4.5777e-04, -1.2745e-03,  6.6629e-03, -8.8446e-03,
        -3.8810e-03,  3.7938e-04, -1.5508e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2702e-03,  5.3144e-01,  2.8499e-04, -1.8825e-03, -7.7294e-04,
         2.4326e-03,  1.4603e-03,  4.4428e-01, -8.6801e-04,  4.4407e-04,
        -7.5838e-04, -1.4752e-03,  5.2019e-04,  1.4761e-05,  3.5608e-04,
         7.6848e-04, -7.8426e-04, -3.4059e-03,  1.7434e-04,  8.6707e-05,
        -2.7734e-03, -6.6969e-04,  1.9959e-04,  5.7338e-04, -3.0170e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1300: [tensor([ 6.6455e-03, -3.6426e-01, -6.8178e-04,  5.1914e-04, -4.2139e-01,
         2.4175e-04,  8.2206e-04, -1.9539e-01,  4.2779e-04,  8.1494e-04,
         1.8640e-03,  7.0077e-04,  1.5386e-03,  7.3570e-04,  7.9311e-04,
         7.0552e-05,  8.0093e-04, -4.6052e-04,  1.0122e-04,  1.8930e-04,
         5.6282e-04,  3.8990e-04,  2.0672e-04, -3.9465e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7911e-03, -2.1965e-01, -6.2585e-04,  7.0685e-05, -6.2912e-01,
         9.2940e-04,  9.9588e-04, -1.3574e-01, -5.8808e-06,  7.9811e-04,
         2.2558e-03,  2.5872e-04,  5.4751e-04,  1.2149e-03, -3.0548e-04,
        -1.2391e-04,  5.6910e-04, -1.8288e-04, -1.2936e-04,  2.1143e-04,
        -2.8856e-04,  8.7072e-04, -2.6590e-04, -4.9780e-04, -3.7117e-04,
        -1.8393e-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([-2.9794e-04,  1.6874e-03,  3.5700e-04,  1.2649e-03, -1.7060e-01,
        -8.1762e-01,  4.5502e-04,  1.8326e-03,  7.0356e-04,  6.5068e-04,
         9.4558e-04,  6.0234e-04, -1.8292e-05,  8.7369e-04,  1.4598e-04,
         1.5071e-04,  5.3637e-04,  6.4163e-04,  2.0269e-04,  1.0540e-04,
         3.0588e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3011e-03, -5.3830e-05,  4.1338e-04,  9.3867e-04,  3.7790e-02,
        -9.5067e-01,  8.0237e-04,  2.2687e-03,  1.1462e-04,  4.1797e-04,
         1.0770e-03, -1.4385e-04,  5.7670e-04,  4.6946e-04,  4.7624e-04,
         2.8002e-04, -2.2144e-04,  4.8252e-04, -1.8777e-04, -1.3934e-04,
         1.7532e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7654e-03, -1.1439e-03, -4.8310e-04, -1.3631e-03, -2.8991e-02,
         9.5828e-01, -5.8672e-04, -5.5879e-04,  5.2893e-05, -4.3260e-04,
        -2.2209e-04,  3.5764e-05, -1.4692e-03,  2.0632e-04, -1.6709e-03,
        -7.7599e-04, -1.7063e-05, -5.5559e-05, -4.1531e-04,  1.4865e-04,
         3.2609e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8681e-02,  1.1251e-03,  2.5206e-03,  4.1371e-03,  6.6472e-05,
        -9.1772e-01, -1.0257e-03, -2.0462e-03, -1.6850e-02,  2.2323e-03,
        -1.5634e-02,  2.6143e-03,  1.4179e-03, -3.0485e-04,  5.2590e-04,
         8.9538e-04,  2.6848e-04,  3.2355e-04,  2.3979e-03,  8.2089e-04,
         1.0833e-03,  1.5372e-03,  4.4683e-04,  1.0185e-03,  8.0079e-04,
        -5.4089e-04, -7.0403e-04, -3.9862e-04, -4.0550e-04, -1.8222e-05,
        -7.6264e-04,  3.8146e-04,  2.9264e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.9580e-03,  2.2568e-03,  4.4892e-04,  6.1930e-03,  1.3297e-03,
        -8.4403e-01,  9.8754e-04,  1.2772e-03,  5.0265e-02,  1.5836e-03,
         7.0535e-02,  5.9958e-04,  1.8099e-03,  9.1156e-04, -2.3076e-04,
         1.4093e-03,  3.7580e-04,  4.3344e-04, -1.2645e-03,  1.7379e-04,
         6.1930e-04, -2.8576e-04, -5.8912e-04, -2.7806e-04,  2.3680e-03,
         1.8431e-04,  3.1466e-04,  2.8687e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2082e-02,  3.6635e-03,  1.8806e-03,  3.5828e-03, -1.1520e-03,
        -8.5492e-01,  2.5638e-03,  1.9703e-03,  1.1609e-02,  3.8915e-04,
         8.4456e-02, -1.4416e-03,  1.2720e-03,  1.4911e-03, -7.0037e-05,
         1.7474e-03, -5.7328e-05,  1.1805e-03,  1.6288e-03,  1.3240e-03,
         6.9936e-04,  4.3005e-04,  3.8799e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1384e-03, -6.5066e-04, -5.5832e-04,  9.7318e-01, -1.2512e-03,
        -1.1765e-02, -1.2594e-03, -1.3172e-03, -1.5631e-03, -1.1396e-03,
         4.3942e-04, -2.4556e-06, -3.8682e-04, -1.1984e-04, -4.4785e-04,
        -7.5016e-04,  2.8339e-04, -7.9434e-05, -6.7724e-04, -5.3824e-04,
        -8.8661e-04, -6.8038e-04,  5.7810e-04,  3.1154e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0876e-05, -7.9346e-04, -6.7926e-04,  8.4887e-01, -1.1054e-03,
         1.3673e-01, -1.9876e-03, -1.8819e-04, -1.4360e-03, -2.0851e-03,
        -1.0768e-03,  1.3149e-04,  2.4194e-04, -4.0702e-05, -4.0429e-04,
        -1.2185e-03, -1.2640e-03, -1.2155e-03,  2.7938e-04,  2.2937e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0377e-03, -2.2208e-03, -6.1046e-04,  9.7370e-01, -3.6433e-04,
        -3.4613e-03, -1.7815e-03, -1.5313e-03, -2.3229e-03, -3.0052e-03,
        -3.4908e-04, -1.3958e-03, -2.2853e-04, -5.1565e-04, -3.6520e-04,
        -1.7898e-04, -1.4373e-03, -1.9398e-04, -2.9712e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1539e-03, -1.1714e-03, -1.5527e-03,  9.8039e-01, -2.4176e-04,
        -1.4556e-03, -1.6187e-03, -1.9284e-03, -2.2195e-03, -1.2280e-03,
        -7.5069e-04,  6.2831e-04, -2.6849e-04, -1.6141e-04, -4.0646e-04,
        -6.0477e-04, -6.1115e-04, -2.7256e-04, -1.9689e-04,  1.3518e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  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 #1350: [tensor([-6.0381e-03,  6.4040e-04,  5.1039e-04,  6.3970e-05,  3.2946e-03,
        -1.3495e-03,  8.7233e-01,  3.3826e-03,  1.1530e-03,  1.0082e-05,
        -9.8890e-02, -4.0072e-04,  1.2454e-03, -8.5021e-04,  3.6150e-04,
        -2.7015e-04,  2.2928e-03,  1.5273e-05, -3.6526e-03,  1.0112e-03,
         2.2366e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.4176e-03,  7.7968e-04,  1.6765e-03, -5.2381e-03,  1.5098e-03,
        -3.2893e-04,  1.3574e-02,  5.7857e-04,  1.0105e-03,  6.3532e-04,
        -9.5062e-01,  3.6848e-03,  8.2763e-04,  1.3276e-03,  3.5710e-03,
         1.3942e-03,  9.1987e-06,  1.2260e-03,  7.6981e-04,  4.6254e-03,
         5.9935e-04, -6.8615e-05, -4.3731e-04, -8.5836e-05,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8416e-03,  4.4509e-01,  8.2284e-05,  4.0740e-03,  6.5780e-04,
         1.2503e-03, -2.4778e-02, -5.0274e-01,  6.0248e-05,  1.3247e-02,
         8.0647e-04,  1.2482e-04,  4.0627e-04,  9.2666e-04,  7.7992e-04,
         9.3929e-04,  3.3821e-04,  3.5982e-04,  2.2363e-05,  4.7508e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.0930e-03,  3.8975e-01,  6.4988e-04, -9.9413e-03,  1.7716e-03,
         6.0352e-04, -9.2414e-02, -4.7413e-01,  2.5364e-05,  1.7630e-02,
         5.8103e-04,  3.1691e-04, -4.0417e-04,  8.8579e-04,  5.3588e-04,
        -2.3728e-04,  1.1986e-03,  8.0566e-04,  5.1182e-04,  4.1442e-04,
        -1.0198e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.2580e-03,  2.3304e-01,  1.5876e-03,  3.2239e-04,  4.0125e-03,
        -3.4110e-04, -3.3154e-03, -7.3361e-01,  1.2037e-03,  3.2040e-04,
         1.4597e-03, -2.4759e-04,  9.1883e-05,  2.3196e-03,  9.1272e-04,
         2.4513e-04, -4.9812e-04,  1.8021e-03, -9.5194e-05,  1.9292e-03,
         3.0873e-03,  1.0462e-03,  1.1006e-04,  1.0254e-04, -4.2302e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.7524e-04,  2.9190e-03, -4.9237e-04,  2.4998e-03, -1.1117e-03,
        -1.8576e-01,  1.2527e-03, -6.0519e-01,  3.4524e-04, -1.3519e-03,
        -1.9141e-01,  1.5062e-03,  8.3792e-05, -7.6211e-04,  1.0519e-03,
         1.6300e-04,  2.2215e-04,  6.4256e-04,  1.5655e-03,  4.3982e-04,
        -1.9620e-04, -5.5883e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1430e-03,  1.2467e-03, -3.0800e-04,  6.6407e-04, -3.9536e-02,
        -2.3680e-01, -3.9610e-04, -5.5918e-01,  2.0181e-03,  1.0628e-03,
        -1.4918e-01,  4.6075e-04, -2.6879e-04, -3.2568e-05,  1.5951e-03,
         5.7472e-04,  5.0343e-04, -6.0435e-04,  1.8388e-03, -1.1717e-04,
         1.6173e-03, -5.3127e-04, -3.2106e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3781e-03,  1.0299e-03, -2.4576e-03,  1.4867e-04, -1.5953e-02,
        -3.7112e-01,  1.8742e-04, -5.1487e-01,  2.2504e-03,  1.2991e-03,
        -8.3553e-02,  1.0820e-03,  4.8698e-05, -7.2198e-05,  1.8760e-03,
        -5.0608e-05,  9.6199e-04,  4.0408e-04,  1.0605e-03, -1.2474e-04,
        -7.6964e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3404e-03,  9.8176e-01,  5.4166e-04,  1.3885e-04,  1.0713e-03,
        -6.1770e-04, -4.7619e-03,  1.3672e-03, -5.2280e-04, -2.8915e-04,
        -4.7022e-05,  8.8792e-04,  5.0414e-04, -9.1971e-04,  4.6794e-04,
         3.5422e-04,  3.0538e-04,  2.1068e-04,  2.4660e-04, -2.4761e-04,
         2.7402e-05,  7.9412e-05, -2.9106e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2425e-03,  9.7157e-01,  1.9969e-04,  7.4960e-03,  8.4917e-04,
        -7.9620e-04,  6.9650e-04,  1.0634e-03, -2.5901e-04, -5.5276e-04,
         1.3268e-03,  9.2261e-03, -1.0509e-04, -6.6812e-04, -2.1827e-04,
        -6.7662e-04,  4.8704e-04, -1.4790e-04,  5.7859e-04, -4.7637e-04,
        -3.3251e-04, -2.8666e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6256e-03,  4.9915e-01,  1.5495e-03,  4.3853e-03, -4.6293e-06,
         4.6002e-04,  2.2196e-03,  1.3404e-03, -4.2502e-04,  1.9563e-03,
         1.8384e-03, -4.8052e-01,  4.0341e-04, -3.4995e-04, -4.8994e-04,
         7.7366e-04,  2.3116e-04,  2.6777e-05, -5.4708e-04, -2.3969e-04,
         5.6116e-05,  2.1472e-04, -1.9189e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.4772e-03, -7.8922e-01,  1.3265e-03,  2.5127e-03,  3.0020e-03,
         3.5233e-03, -1.8104e-01, -6.9237e-04,  3.7368e-04,  1.7379e-03,
         4.4053e-05,  2.6840e-04,  1.8172e-03,  1.1092e-03, -3.8726e-04,
         1.6067e-03,  2.2723e-03,  5.8820e-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>)]
Iter #1400: [tensor([-7.9980e-04,  3.8012e-01, -3.3258e-04, -1.1623e-01,  1.9595e-01,
        -4.3316e-04,  1.0322e-04, -1.9323e-04,  5.6309e-02,  2.1052e-04,
         1.0581e-02, -4.5706e-05, -9.8345e-04,  2.3592e-01,  7.7615e-06,
        -4.0074e-04,  4.8583e-05, -5.3168e-05,  1.4596e-04, -6.5471e-05,
         1.6429e-04,  3.2772e-05, -1.5475e-04,  5.0270e-05,  2.3004e-04,
        -1.8999e-04, -2.3962e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4711e-03,  5.3282e-01, -1.2354e-04, -3.9070e-02,  2.7303e-01,
        -5.0253e-04, -2.1039e-04,  2.3959e-04,  4.3282e-02, -8.6636e-05,
         7.2295e-04, -1.5033e-04, -3.8900e-04,  1.0428e-01,  3.2889e-04,
        -1.7699e-04, -1.5275e-04,  2.1106e-07,  1.4135e-04,  6.4531e-05,
        -1.9445e-04,  2.5366e-04,  2.4593e-04, -3.6406e-04, -1.1318e-04,
        -5.1127e-04,  1.9601e-04, -3.9026e-04, -9.7395e-05,  2.0989e-04,
         1.8441e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3426e-03,  4.0852e-01,  3.9018e-05,  4.0771e-02,  3.9731e-01,
         6.2380e-05, -2.0632e-04,  3.9850e-05,  2.5888e-02,  1.7212e-05,
         3.1189e-03, -6.4153e-05, -4.7216e-04, -1.1788e-01,  3.7934e-04,
        -7.0314e-04,  6.1307e-05, -1.6761e-04, -1.5465e-04,  5.1292e-04,
        -4.9833e-05, -2.2190e-05, -8.1392e-05,  6.3275e-04,  2.9442e-04,
        -5.2738e-04,  2.3449e-04,  7.7994e-05, -2.4223e-04,  1.2983e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5382e-04,  7.4384e-04,  7.1599e-05, -3.3790e-01, -7.1250e-04,
         2.8408e-02,  3.7206e-01, -3.1992e-04,  4.7794e-04, -3.9824e-04,
         2.5641e-01,  6.9612e-04, -7.0089e-05,  2.8495e-04,  4.9309e-06,
        -5.8934e-06, -2.4218e-04, -3.1480e-04,  2.1416e-04,  5.3087e-05,
         1.5718e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.4704e-04,  1.2756e-03, -2.2357e-04, -4.0292e-01, -1.3584e-03,
         9.8049e-03,  3.7209e-01, -9.4706e-04, -9.4343e-04, -1.6790e-03,
        -2.0339e-01,  3.3749e-04, -3.4721e-04,  3.4351e-04, -4.2626e-05,
        -9.1126e-04, -1.0278e-03, -1.4426e-04, -9.0301e-04,  2.7833e-04,
         8.6425e-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([-6.6595e-04,  9.0329e-04,  4.6355e-04, -6.7423e-01, -1.0050e-03,
         5.9136e-03,  5.8920e-02, -1.6683e-03, -7.1351e-04, -6.0148e-04,
         2.5106e-01, -7.1644e-04,  7.5892e-05, -5.7104e-04, -1.0808e-04,
        -7.2913e-04,  1.9855e-04, -4.3021e-04, -1.9864e-04,  4.3826e-04,
        -3.9002e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2552e-03,  3.9413e-03,  8.7100e-04, -5.4830e-01, -4.3378e-01,
        -1.4856e-03,  5.0018e-04, -4.0107e-04,  2.0862e-03, -1.0465e-03,
         6.6946e-04,  1.9503e-03, -1.0618e-03, -2.4381e-04, -2.5772e-04,
        -2.0035e-04, -9.4710e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6988e-04,  1.6162e-03, -5.4433e-04, -5.2428e-01, -4.6636e-01,
        -5.9708e-04, -3.8827e-04, -1.0935e-04,  4.6165e-04, -2.8653e-04,
        -3.3690e-05, -5.6247e-04, -7.8298e-04, -1.0674e-03,  7.2109e-05,
        -2.0406e-04, -1.4467e-03, -7.3739e-05, -7.5375e-04, -9.2761e-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([ 1.6938e-03,  2.2012e-03, -6.7624e-05, -3.3883e-01, -6.4738e-01,
        -1.5556e-03,  7.4432e-04,  5.2569e-04, -5.2025e-04, -9.0832e-04,
        -1.7472e-03, -1.1893e-03,  5.1176e-04, -7.3104e-05, -7.0356e-05,
        -1.3087e-03, -5.0690e-04, -1.7323e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6431e-03,  9.6460e-02, -1.7540e-04, -8.4569e-02,  2.7499e-01,
        -1.1540e-04, -1.8592e-04,  2.1752e-04,  1.4339e-05, -5.3084e-05,
         1.5385e-02,  2.0292e-01,  1.6437e-05, -3.0170e-05,  3.5216e-04,
         1.1086e-01,  2.7003e-05,  1.0432e-03, -8.7840e-05, -3.2126e-05,
        -5.6115e-05,  1.3459e-01,  5.8797e-07, -4.1005e-05, -6.9350e-04,
         7.2007e-02, -8.9100e-05,  2.2319e-04, -7.1904e-05,  5.9234e-05,
         1.2737e-04,  1.7573e-04, -2.4726e-04, -3.0905e-04, -7.9396e-05,
        -1.4419e-05, -2.9706e-04, -8.6863e-05, -7.0846e-05,  3.2142e-05,
        -1.6545e-04,  1.0446e-04,  3.5474e-05,  5.6915e-05, -6.5175e-05,
        -1.5356e-05, -4.7478e-04, -1.4386e-04, -3.1434e-04, -1.5843e-04,
        -1.4458e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0334e-03,  5.0870e-02, -2.6680e-04,  1.9286e-02,  3.9301e-01,
        -2.7295e-04, -2.8421e-04,  2.4028e-03,  1.9949e-04, -4.6222e-05,
         1.3640e-02,  1.2512e-01, -2.7326e-04,  5.4117e-05,  5.8197e-04,
         7.1007e-02, -2.7970e-04, -1.6453e-03,  6.2291e-05,  1.8759e-04,
         1.4452e-04,  1.5525e-01,  1.7578e-04,  9.3222e-05,  1.6064e-02,
         1.4404e-01,  9.2501e-06,  4.5533e-04, -2.7282e-04, -1.8353e-04,
         7.3420e-05,  3.3161e-05, -2.8233e-04, -4.0754e-05, -8.5788e-05,
         2.6744e-04, -3.9702e-05, -1.4003e-04, -2.3656e-05, -8.0437e-05,
         9.5509e-05,  1.4318e-04, -1.9347e-04, -1.8581e-04,  1.5672e-07,
        -3.4999e-04, -3.3475e-04,  3.7932e-05, -2.6353e-04, -1.1608e-04,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1980e-03,  1.9857e-01, -2.0610e-04, -1.5358e-02,  2.1864e-01,
        -2.0130e-05, -3.5896e-05,  1.0294e-03,  4.9758e-05,  1.8484e-04,
         1.1329e-02, -7.3002e-03,  2.6006e-04,  1.1440e-04,  3.4462e-04,
         1.1965e-01, -1.0091e-04,  3.7737e-04,  2.3918e-04,  2.2668e-04,
         1.4919e-04,  2.2503e-01,  1.3230e-04,  2.0718e-05,  1.2944e-02,
         1.8264e-01, -1.1783e-05,  2.6970e-04, -6.1411e-05,  8.1122e-05,
         2.1937e-04,  1.4543e-04, -1.1120e-04, -3.7463e-06,  8.7972e-05,
         1.2182e-04, -1.1155e-04, -7.1169e-05,  4.5944e-05,  1.6875e-04,
        -3.1666e-04, -2.6406e-04,  7.1189e-06,  8.1638e-05, -7.0914e-04,
        -3.9437e-04, -2.4628e-04, -2.5673e-04, -1.2520e-05, -4.9049e-05,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1450: [tensor([ 9.4664e-04, -2.6355e-04,  5.7519e-04,  1.6581e-03,  9.8999e-01,
        -6.2342e-04,  3.9210e-04,  7.8345e-04, -5.2930e-04,  8.5738e-04,
         8.2549e-04, -1.3248e-03,  4.0788e-05, -3.8333e-04,  7.7326e-04,
         3.1194e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2933e-03,  1.4460e-04,  4.5820e-04,  1.2150e-03,  9.8202e-01,
        -6.0348e-04, -8.8075e-03, -6.1113e-04, -3.0175e-04,  2.2391e-04,
         5.8591e-04, -9.3190e-04,  5.4576e-04, -1.1611e-03, -2.4568e-04,
        -4.8362e-04,  1.6494e-04, -2.0436e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7565e-04, -1.1329e-03,  5.2228e-04,  3.5604e-04,  9.9124e-01,
        -9.4784e-04, -2.4144e-03,  5.1321e-05,  2.7041e-04,  3.2866e-04,
         1.0541e-03,  1.8872e-04,  4.5717e-04,  5.6073e-04,  2.3579e-04,
        -6.6736e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.7626e-04,  1.4367e-03,  1.0574e-03, -9.7627e-01,  9.8901e-04,
        -8.2064e-03,  1.6175e-03,  1.3867e-03,  4.3349e-04,  4.3698e-04,
         1.5684e-04,  1.0027e-03, -8.5461e-04, -3.1468e-04,  1.2484e-03,
         6.1443e-04,  4.2806e-04, -3.0955e-04,  7.7301e-04, -6.8620e-04,
         6.8345e-05,  2.1514e-04, -5.1504e-04,  9.9011e-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([ 1.5200e-03, -6.3332e-04, -4.4992e-05,  9.8630e-01,  1.0774e-03,
         4.5702e-03, -6.3388e-04, -1.4334e-03, -2.1317e-04,  1.8261e-04,
         2.7398e-04,  7.1601e-04,  1.7524e-04, -6.1734e-04, -4.3481e-04,
        -9.8702e-04,  1.8239e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3404e-03, -1.2706e-03,  2.9757e-04,  9.7469e-01,  1.3916e-03,
         1.2937e-02, -5.3884e-04,  1.7776e-04, -8.0758e-04, -8.4894e-04,
        -3.0461e-04, -1.2712e-04, -3.1824e-04,  7.7285e-05, -2.8923e-04,
        -2.6446e-04, -1.1521e-03,  8.8503e-06, -1.6231e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5641e-03,  1.8996e-01, -1.5509e-05, -1.1550e-03,  2.7607e-06,
         1.4061e-04,  1.8425e-02, -2.7724e-06, -4.0367e-05,  2.4863e-01,
         1.0747e-04, -8.5028e-05, -4.4648e-04,  1.2794e-04,  2.4778e-04,
         1.4874e-02,  3.6296e-05, -1.6871e-03,  1.5990e-04, -1.0086e-04,
         1.2146e-04, -7.1797e-03,  9.9641e-05, -5.4358e-02, -2.6075e-04,
        -1.8220e-04, -3.6246e-01,  1.0633e-04,  4.2522e-04, -2.2119e-02,
         1.9654e-04, -1.9391e-04, -1.8946e-04,  7.1136e-02, -6.2971e-05,
         2.2631e-04,  1.0625e-04,  8.0921e-05, -1.2471e-04, -2.6868e-04,
        -3.6128e-04,  1.4899e-04, -2.8089e-04,  2.4820e-04, -1.0470e-04,
        -1.2678e-04, -9.8981e-06,  1.4671e-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([ 1.2068e-03,  1.0073e-01, -1.9248e-04, -8.0262e-04,  6.3373e-05,
        -5.0522e-05, -1.0928e-01, -3.3433e-05, -1.5148e-04,  4.9038e-01,
        -5.7926e-05,  3.1820e-05, -4.9740e-04,  2.5682e-04,  1.0792e-04,
        -2.2205e-02, -3.1312e-04,  1.9172e-02,  1.2784e-05, -1.2253e-04,
        -8.1732e-05,  1.4079e-03, -9.1356e-06,  3.6734e-03, -1.7004e-04,
        -8.6375e-05, -2.9524e-02,  9.5620e-05,  9.8114e-05,  9.7555e-04,
         7.9433e-05, -1.2178e-04,  8.8774e-05,  2.1459e-01,  1.4299e-04,
        -4.1727e-05, -9.7069e-05, -2.1643e-05,  4.9654e-05, -2.9211e-04,
        -3.6578e-04,  2.3539e-04,  1.4535e-04,  2.2103e-05,  1.6519e-04,
         5.1416e-05,  1.2168e-04,  2.1338e-04,  2.0251e-04, -1.2416e-04,
        -1.2947e-04, -1.5161e-04, -1.7552e-04,  3.4267e-05, -8.7363e-05,
        -1.0531e-04, -2.0997e-04, -1.5543e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2380e-03,  1.5828e-01, -8.3380e-05, -1.3928e-04,  1.2488e-04,
        -7.0617e-05, -2.0242e-01, -4.1773e-06, -3.4171e-04,  2.2044e-01,
        -2.4552e-04,  6.4562e-05, -2.2380e-04,  2.2169e-04,  1.2456e-04,
        -1.0642e-01, -1.9129e-05,  7.1138e-03,  6.2283e-05, -2.6763e-05,
        -1.8629e-04, -3.1158e-02,  2.5500e-04,  5.9543e-03, -6.3630e-05,
        -5.4348e-05, -1.8249e-01,  1.1184e-04,  6.4638e-05,  1.6647e-03,
         2.6278e-04, -2.5907e-04, -3.4008e-05,  7.8299e-02, -1.4567e-04,
         9.3194e-05,  7.9198e-05,  1.0947e-04, -2.3507e-04,  3.1661e-04,
        -4.8875e-05, -1.3917e-04, -6.1173e-05,  1.0626e-04,  6.0164e-05,
         8.5654e-05,  1.1507e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9458e-03,  2.3880e-04, -9.8501e-01,  5.9212e-04, -9.1315e-04,
         9.1489e-04, -8.7830e-05, -1.7251e-03,  3.4476e-04,  1.5766e-03,
        -8.3149e-04, -8.8710e-04,  1.2752e-03, -4.7339e-05, -5.2674e-04,
        -4.8917e-04,  4.3084e-04,  3.8722e-05, -1.3142e-03,  6.7382e-04,
        -1.3195e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.9747e-04,  4.9586e-04,  9.8788e-01, -1.7714e-03, -1.0452e-03,
         1.6317e-03, -1.9803e-04,  1.3292e-03, -4.2954e-05,  1.5987e-03,
         9.6841e-06, -2.1545e-04, -7.1878e-04,  7.7763e-04, -5.4957e-04,
         2.4082e-04,  6.7533e-04, -1.1937e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2826e-04,  6.6824e-04,  9.8488e-01,  1.9474e-04, -4.8097e-04,
        -6.0441e-04, -3.2573e-04,  2.5119e-03, -4.1840e-04, -8.3643e-04,
         1.3486e-05, -3.1337e-04,  8.2389e-04,  1.1504e-03,  1.5594e-03,
         7.4457e-04,  1.2678e-03,  1.2654e-03,  5.0030e-04,  7.8792e-04,
         4.2686e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  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 #1500: [tensor([ 5.5384e-05,  3.5216e-01, -2.6335e-05,  4.9388e-05,  2.1296e-04,
         1.9382e-04, -1.7294e-01, -4.7047e-05, -2.4408e-04, -1.8507e-04,
         3.4960e-04,  4.0333e-01, -2.6019e-04, -5.2955e-04,  3.2547e-05,
        -1.6515e-04, -1.5396e-04,  6.6375e-02, -4.5627e-04, -8.9519e-05,
        -8.0710e-05, -2.4211e-04,  3.3135e-04,  4.9423e-09,  3.6307e-04,
        -6.3531e-04,  1.4251e-04,  1.2735e-04, -1.9479e-04,  2.2722e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9341e-03,  5.1710e-01,  3.8410e-04, -9.9824e-05,  2.3283e-05,
         1.1361e-04,  1.7720e-01,  1.1504e-04, -1.0548e-04,  2.3845e-04,
         2.8965e-04,  2.1117e-01,  3.7830e-04, -2.5805e-04,  2.2049e-04,
        -2.1412e-05,  5.0917e-05,  8.6649e-02,  3.0318e-05,  1.8963e-04,
         3.7137e-04, -3.4308e-04,  2.1904e-04,  3.9039e-04,  2.6679e-04,
        -3.6955e-04,  1.9409e-04,  2.8080e-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([ 1.3472e-04,  4.4036e-01,  2.3246e-05,  3.7567e-04,  2.0126e-04,
        -2.5504e-04, -2.9122e-01, -4.8310e-06,  4.1823e-06, -1.4602e-04,
         4.2664e-04,  2.3599e-01,  1.9261e-04, -2.8800e-04,  2.4710e-04,
         1.8924e-04,  2.5097e-04,  2.7212e-02,  3.3361e-05,  1.6014e-04,
         2.9074e-04, -4.5919e-04,  4.2089e-05,  3.2770e-04, -1.8034e-04,
        -2.8148e-04,  7.0206e-05, -3.8008e-04,  1.5901e-04, -2.3820e-06,
         9.2480e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6666e-02, -1.6065e-04, -9.9754e-04,  2.5289e-01,  6.4888e-01,
         2.0853e-03, -4.2188e-04,  5.6532e-04,  2.3527e-04, -3.5003e-02,
        -9.6934e-04,  3.3840e-03, -1.4451e-02,  9.4837e-04,  1.1630e-03,
        -1.3748e-03,  3.5391e-04,  1.4767e-03,  9.0259e-04,  1.6964e-03,
         1.2576e-04,  1.0378e-03, -9.3856e-04,  6.1916e-04,  1.9028e-03,
         1.4893e-03,  1.5272e-03,  1.2758e-03, -1.2297e-03,  2.2235e-03,
         2.2701e-03, -2.2496e-04, -5.0262e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6317e-03,  7.7215e-04,  2.3238e-04, -5.8467e-02,  9.2488e-01,
         2.6910e-04,  4.2934e-04,  4.1008e-04,  3.6648e-04, -2.9664e-03,
         6.0333e-04,  3.9663e-04,  2.6974e-04,  3.6268e-04,  6.8796e-04,
         5.0803e-04,  3.1535e-04,  1.3897e-04, -5.5954e-05,  5.1416e-04,
         6.0189e-04, -2.2521e-05, -3.1584e-04,  4.5118e-04,  3.5689e-04,
         6.1916e-04,  3.8310e-04,  2.8124e-04,  5.6954e-04,  5.1441e-05,
         6.5741e-04, -2.2580e-04,  1.8282e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5680e-03,  4.0574e-04,  3.5507e-04, -3.4855e-01,  6.4003e-01,
        -1.8334e-04, -6.6663e-05,  2.9046e-04,  1.2268e-04, -1.3516e-03,
        -1.7612e-05,  4.0682e-04,  2.6823e-03, -6.0548e-05,  2.8139e-04,
        -1.1420e-04, -2.4697e-05,  3.5510e-04, -8.2523e-05,  2.2612e-04,
         1.2288e-04, -2.2505e-05, -9.2002e-05, -1.4280e-04,  4.0140e-04,
         4.9994e-04, -3.2187e-05,  2.8877e-04, -1.1245e-05, -5.4735e-04,
         2.1415e-04,  1.6278e-04, -2.0186e-04, -8.0335e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.1509e-04,  6.1906e-01, -1.8195e-04,  9.4160e-05, -1.5667e-03,
         3.3193e-04, -4.3777e-02, -3.7635e-04, -3.4752e-04,  2.8660e-05,
         5.9198e-05, -1.9156e-01, -1.9024e-02,  2.3868e-05,  1.5882e-03,
         1.1982e-01,  3.6063e-04,  8.0743e-05, -2.8975e-04, -2.0940e-04,
        -7.0279e-05,  3.5697e-04, -5.3369e-05, -1.2896e-04,  1.3765e-05,
        -7.5080e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1996e-04,  7.1032e-01, -7.1237e-06, -9.3995e-05, -2.7437e-04,
        -4.0868e-05,  1.5050e-01, -9.0101e-06, -3.0698e-05,  1.5273e-04,
         4.9148e-05,  7.0844e-03,  1.0806e-01, -3.9423e-04,  1.1379e-03,
        -1.8222e-02,  4.6486e-04,  2.1108e-04,  6.7842e-05,  4.3644e-05,
         6.0862e-05,  2.6710e-04,  1.8838e-04, -1.2617e-04,  1.3013e-04,
        -9.1199e-05,  1.6964e-04,  1.6409e-04, -3.4686e-05, -3.3154e-04,
         3.3879e-04, -3.1530e-04,  1.8682e-04, -1.8540e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4482e-05,  7.5500e-01,  1.8474e-04,  4.0162e-04,  2.3929e-03,
        -1.9740e-04,  3.3314e-02, -4.1677e-05,  6.3668e-05,  7.3473e-05,
         6.1034e-04,  8.0195e-03, -2.5127e-02,  2.0241e-04,  1.0658e-02,
         1.6188e-01,  7.9616e-04, -4.7778e-05,  2.2159e-06, -3.6834e-04,
         3.4005e-04, -1.4023e-04,  7.8466e-05,  3.6680e-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([ 3.7054e-04,  4.2461e-04,  1.0751e-04,  5.3982e-04,  5.7219e-04,
         9.9093e-01, -3.2128e-04,  1.3861e-04, -8.3313e-04, -7.4524e-04,
        -1.4545e-03, -2.7439e-04, -1.9306e-04, -1.4299e-03, -6.1594e-04,
         2.0977e-04, -6.5677e-04, -1.8151e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2188e-04, -8.0722e-04,  6.2351e-04, -1.7983e-03,  9.5323e-04,
         9.8915e-01,  1.2213e-03, -1.8291e-04, -7.7469e-04, -5.2713e-04,
        -1.1419e-03,  1.6497e-03, -6.9942e-04,  2.7833e-05, -1.1635e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8382e-04,  7.5419e-04, -8.5603e-04, -1.5172e-03, -1.3654e-03,
         9.8786e-01, -1.4084e-03, -6.0171e-04, -1.8703e-03, -9.1055e-04,
        -1.3889e-04, -3.4361e-05, -1.4545e-03, -8.1766e-04,  2.7820e-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>)]
Iter #1550: [tensor([ 5.2638e-04,  4.4202e-01,  2.7988e-04, -2.3119e-01,  2.3290e-04,
         6.0253e-03, -1.2246e-05,  3.0654e-01, -6.1139e-04, -1.0015e-02,
         4.4577e-04, -2.4880e-04,  3.0494e-04, -4.7453e-05, -6.2174e-05,
        -4.5419e-05,  4.3950e-05,  1.3662e-04, -1.8558e-04,  5.5089e-04,
        -3.2538e-04,  6.3755e-05,  8.9688e-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([ 7.1436e-04,  1.2811e-01,  3.8709e-04, -7.3527e-02,  5.5478e-05,
         1.3414e-03,  7.7477e-05,  7.9070e-01, -1.4545e-04, -2.3606e-03,
        -1.4175e-05,  1.0809e-04,  1.5991e-04, -1.9912e-04,  1.6582e-04,
         1.6930e-04, -1.8145e-04,  1.9145e-04, -1.6027e-04, -2.5992e-04,
        -2.3311e-04,  5.9599e-05,  1.5651e-04, -2.3467e-05,  3.9359e-04,
        -9.9170e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.4101e-04, -1.7209e-01, -3.0345e-04,  5.6950e-01, -3.7078e-04,
        -1.1024e-04, -5.1677e-04, -2.1357e-01, -3.9009e-05,  3.9919e-02,
        -1.3986e-04, -1.4974e-04,  1.9760e-04,  2.4794e-05, -5.8625e-04,
         2.4960e-04, -1.7137e-04,  9.0545e-05, -9.0426e-05, -1.2635e-04,
         2.7246e-05, -6.5719e-04,  2.7567e-04,  5.8156e-05, -1.8596e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0699e-03, -3.9393e-01, -5.7605e-04,  3.6610e-06, -2.0595e-04,
         5.5601e-03,  1.6447e-04,  3.8557e-04, -9.5325e-05,  3.4174e-04,
        -7.6725e-05,  4.8955e-04, -2.3053e-04, -3.4950e-02, -6.5698e-05,
         4.4677e-03,  1.8967e-04,  5.5342e-01, -2.6015e-04, -4.0831e-04,
        -3.7459e-04, -8.2458e-05,  2.7940e-04,  8.5056e-05,  7.2141e-05,
         1.4285e-04,  9.8769e-04,  2.4219e-04, -2.7702e-04, -2.8034e-05,
        -2.8640e-04, -2.5238e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2355e-03,  7.7477e-02,  2.9215e-04,  2.3231e-04,  5.1213e-04,
        -4.2007e-01, -5.2280e-04, -1.9497e-04, -5.2469e-05,  3.9708e-04,
        -2.2746e-05,  6.6673e-04, -2.8319e-04,  3.1751e-02,  8.8098e-04,
        -3.8195e-03, -8.0503e-05, -4.5948e-01, -3.3467e-04, -1.0428e-04,
         3.0116e-05,  2.5651e-04, -5.6495e-05, -2.8059e-04,  1.5227e-04,
         3.5494e-04,  5.8442e-05,  5.6828e-05, -1.9663e-04,  1.4480e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3639e-03,  5.1392e-01, -1.6114e-05, -1.3789e-04,  2.8513e-04,
         3.7269e-02, -1.6508e-04, -3.3913e-05, -3.6866e-04,  8.8161e-05,
         1.5635e-06,  2.8200e-05,  2.9431e-04,  1.5806e-01,  1.3120e-04,
        -3.2548e-03, -1.2991e-04, -2.8139e-01,  5.9763e-04,  2.0902e-04,
         2.5211e-04,  1.0662e-04, -1.1096e-05,  9.1856e-05, -5.6558e-04,
         2.3488e-05,  5.5074e-05,  1.1872e-04, -3.0551e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.2393e-04,  1.6004e-01,  3.0579e-05,  1.6696e-02,  6.3239e-01,
        -2.4092e-04, -5.4743e-05,  4.0943e-05,  9.9735e-05, -2.9097e-05,
        -2.6509e-03,  1.8214e-01,  1.7608e-04, -2.7100e-03,  2.8767e-04,
         2.1219e-04,  2.6301e-04,  1.4165e-04,  2.3032e-04,  5.4890e-05,
        -2.8962e-04, -2.5814e-04, -7.5934e-05,  1.1083e-04,  1.5831e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.5353e-04, -6.2073e-02,  1.1058e-04, -8.5427e-02, -5.9892e-01,
        -2.2888e-04,  1.1579e-04,  1.2923e-04,  6.9377e-04,  1.2780e-04,
         1.5841e-03, -2.4637e-01,  4.5707e-04,  3.7474e-04, -6.3106e-04,
        -2.5992e-04,  1.5499e-04,  7.2440e-04,  3.9030e-05, -9.5244e-05,
         1.8264e-05,  3.5537e-04, -2.3867e-04, -1.9196e-04, -2.2423e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.2316e-04,  8.6269e-03, -2.0001e-04, -2.5599e-01,  5.9845e-01,
        -8.4633e-05, -7.6731e-05,  1.0374e-04,  3.2051e-04,  8.2754e-05,
        -1.7225e-03,  1.2889e-01, -1.6163e-05, -1.9238e-03,  1.8786e-04,
         2.2603e-04,  1.7143e-04,  3.7912e-04, -1.0555e-04, -5.5680e-05,
         1.0726e-04,  2.3064e-05, -1.2382e-04, -2.2833e-04, -1.0688e-04,
        -3.8801e-04,  5.9675e-05,  8.8002e-05,  4.5602e-04,  8.7471e-05,
         1.0067e-04,  1.3173e-06,  8.9236e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3466e-04,  1.0387e-03,  1.4793e-04, -3.7685e-01,  5.3952e-04,
        -8.5720e-05, -1.2551e-04,  1.3099e-04, -3.5251e-04, -6.1590e-01,
         3.8227e-04, -3.1048e-04, -4.7059e-04,  4.0984e-04, -1.1289e-04,
        -1.2144e-05, -2.2072e-04, -1.1421e-04, -2.2074e-04, -2.3174e-04,
         5.2051e-04,  3.7206e-04, -2.9502e-04, -7.2799e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3580e-03, -4.9661e-05,  4.0876e-04, -8.9925e-01, -1.2244e-03,
        -1.0252e-03, -3.6124e-04,  6.6009e-04, -6.5182e-04,  8.4579e-02,
        -9.4502e-04, -4.1057e-04,  9.7268e-04, -1.0078e-04,  1.4239e-03,
         1.5391e-03, -1.2779e-03,  1.0428e-03,  2.5373e-04,  4.6672e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1115e-03,  5.9526e-05, -1.2273e-04,  7.2434e-01,  2.6806e-04,
         8.3872e-04,  1.1372e-04, -3.2972e-04,  4.6159e-04,  2.6760e-01,
        -5.1106e-04,  7.5320e-04, -9.3283e-05, -2.0900e-04, -3.9621e-04,
         4.7046e-05, -4.4577e-04, -7.0550e-04, -4.1295e-04,  1.8302e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
