Iter #50: [tensor([-0.1802,  0.0366, -0.0020,  0.0063,  0.0462, -0.0017, -0.0011, -0.0033,
         0.0478, -0.0284, -0.0030,  0.0025,  0.0047, -0.0047,  0.0044, -0.0006,
         0.0044, -0.0204,  0.1279,  0.0638,  0.1106,  0.0493,  0.0267,  0.0547,
         0.0406,  0.0399,  0.0402,  0.0477,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0279e-01,  3.1352e-02, -1.4848e-03, -1.6143e-02, -3.4920e-03,
        -1.0074e-02,  3.9944e-02, -1.3820e-02,  2.7419e-02, -1.6526e-04,
         3.4396e-06,  3.2322e-03,  2.0316e-02,  3.2333e-02, -3.2321e-03,
         1.4600e-02,  2.9952e-02,  5.1899e-03, -7.9907e-03, -2.3744e-03,
        -3.6135e-03,  4.5404e-02, -2.6946e-03, -1.1050e-02, -1.0750e-02,
         3.4535e-02, -1.1592e-02, -4.5780e-03,  8.8628e-03,  4.0565e-02,
         1.2418e-03,  2.7859e-02,  2.1375e-02,  1.0143e-02, -2.4809e-02,
         8.0832e-03,  2.9168e-02, -9.8133e-03, -7.4347e-03, -8.6042e-03,
         3.2798e-02,  3.3890e-02,  4.3857e-02,  3.4773e-02,  3.8002e-02,
         2.8598e-02], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1259,  0.0409,  0.0072, -0.0043, -0.0021, -0.0054,  0.0506,  0.0003,
         0.0355,  0.0106,  0.0008,  0.0125,  0.0210,  0.0441,  0.0045,  0.0106,
         0.0447,  0.0187, -0.0130,  0.0103,  0.0013,  0.0444,  0.0013, -0.0172,
        -0.0070,  0.0399,  0.0012, -0.0006,  0.0109,  0.0428,  0.0034,  0.0237,
         0.0274,  0.0057, -0.0220,  0.0108,  0.0314, -0.0004,  0.0028, -0.0068,
         0.0357,  0.0709,  0.0229,  0.0610,  0.0454,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3655e-01,  2.6254e-02,  7.9345e-04, -1.1378e-03, -4.0380e-03,
        -2.9287e-03,  4.7838e-02, -7.3234e-03,  2.2598e-02,  7.9606e-03,
         1.3074e-04,  7.6765e-03,  1.3544e-02,  3.5526e-02,  5.4177e-03,
         8.4481e-03,  3.2612e-02,  3.3400e-03, -6.7978e-03,  9.3579e-03,
        -8.0143e-04,  4.2716e-02,  2.8795e-03, -9.6135e-03, -1.4209e-02,
         3.7508e-02, -5.7417e-03, -3.9311e-04,  5.5931e-03,  3.9390e-02,
        -1.8019e-04,  1.5587e-02,  2.1565e-02,  5.6766e-03, -2.4663e-02,
         1.0317e-02,  1.6985e-02,  3.4210e-04, -5.5857e-03, -7.5969e-03,
         3.0181e-02,  6.2669e-02,  3.3417e-02,  2.6113e-02,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1095,  0.0064,  0.0043, -0.0302, -0.0122,  0.0139,  0.0073,  0.0586,
         0.0119, -0.0060, -0.0089,  0.0128,  0.0106, -0.0141,  0.0023, -0.0105,
        -0.0043, -0.0019,  0.0171,  0.0014, -0.0037,  0.1066,  0.0403,  0.0817,
         0.0647,  0.0706,  0.1428,  0.0410,  0.0552,  0.0490,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8015e-01, -5.0789e-03, -3.9952e-03, -2.8462e-02, -2.1579e-03,
         6.8498e-03,  1.2235e-02,  5.1816e-02,  5.4670e-03, -2.2155e-02,
         4.6994e-03,  5.7665e-03,  6.5404e-03, -1.5691e-02, -5.8508e-04,
        -6.9964e-03, -1.0846e-04, -8.5242e-03,  2.0964e-02,  1.3033e-03,
        -1.9225e-02,  6.0343e-02,  4.1875e-02,  4.4623e-02,  7.7161e-02,
         1.4542e-01,  7.4316e-02,  5.1373e-02,  5.0376e-02,  4.5747e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2540, -0.0096,  0.0009, -0.0142, -0.0126,  0.0047,  0.0102,  0.0421,
         0.0077, -0.0063, -0.0012,  0.0072, -0.0004, -0.0054,  0.0091, -0.0108,
         0.0088, -0.0032,  0.0147, -0.0042, -0.0158,  0.0850,  0.0231,  0.0339,
         0.0877,  0.0637,  0.0587,  0.0752,  0.0337,  0.0334,  0.0306,  0.0317,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.2735,  0.0542,  0.0058, -0.0078, -0.0040,  0.0029,  0.0011, -0.0124,
         0.0160, -0.0144,  0.0026,  0.0616, -0.0031, -0.0081,  0.0044, -0.0088,
         0.0447,  0.0555,  0.0871,  0.0710,  0.1244,  0.0566,  0.0438,  0.0365,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1206,  0.0509, -0.0027, -0.0164,  0.0048,  0.0006,  0.0057, -0.0132,
         0.0111, -0.0193,  0.0099,  0.0625,  0.0113, -0.0147,  0.0055, -0.0083,
         0.0624,  0.0785,  0.0990,  0.0626,  0.0797,  0.1132,  0.0443,  0.0595,
         0.0435,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0146,  0.0553,  0.0093, -0.0044,  0.0094,  0.0041,  0.0019, -0.0156,
         0.0249, -0.0235,  0.0081,  0.0774,  0.0179, -0.0040, -0.0017, -0.0166,
         0.0647,  0.0874,  0.1190,  0.0934,  0.1211,  0.0888,  0.0531,  0.0536,
         0.0302,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0340,  0.0381,  0.0002, -0.0062,  0.0412,  0.0010,  0.0137,  0.0367,
        -0.0031, -0.0035,  0.0027,  0.0423, -0.0034,  0.0028, -0.0035, -0.0085,
         0.0918,  0.0438,  0.0901,  0.0452,  0.0782,  0.0335,  0.0499,  0.0799,
         0.0319,  0.0392,  0.0801,  0.0338,  0.0386,  0.0232,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0037,  0.0574,  0.0116, -0.0051,  0.0738,  0.0021,  0.0364,  0.0619,
        -0.0104, -0.0077, -0.0180,  0.0692, -0.0036,  0.0069, -0.0050, -0.0161,
         0.1582,  0.0711,  0.1997,  0.0609,  0.0685,  0.0526,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 0.0199, -0.0204, -0.0040,  0.0323, -0.0162, -0.0006,  0.0490,  0.0104,
        -0.0015, -0.0073,  0.0465, -0.0116,  0.0065,  0.0004, -0.0146,  0.0496,
         0.0838,  0.1272,  0.0862,  0.0661,  0.1814,  0.0350,  0.0466,  0.0466,
         0.0363,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.2164, -0.0165,  0.0006,  0.0354, -0.0106,  0.0162,  0.0629,  0.0064,
        -0.0051,  0.0007,  0.0600, -0.0168,  0.0138, -0.0070, -0.0155,  0.0625,
         0.0966,  0.1613,  0.1409,  0.0241,  0.0306,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4521e-01, -8.5760e-03,  1.4192e-03,  2.3227e-02, -1.4892e-02,
         2.1238e-03,  5.6962e-02,  1.0149e-02,  1.8610e-06,  5.5267e-03,
         4.4174e-02, -1.4097e-02,  1.0507e-02,  9.2086e-04, -2.1159e-02,
         5.3736e-02,  5.6308e-02,  1.0792e-01,  6.3065e-02,  1.8971e-01,
         8.8178e-02,  5.1136e-02,  3.1002e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0598, -0.0100, -0.0088, -0.0085, -0.0082,  0.0024, -0.0021,  0.0033,
         0.0377,  0.0058,  0.0032,  0.0165, -0.0063, -0.0017, -0.0123, -0.0096,
         0.0364,  0.0404,  0.0426,  0.0411,  0.1147,  0.0532,  0.0503,  0.1087,
         0.0342,  0.0506,  0.1173,  0.0650,  0.0326,  0.0166], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0369, -0.0168,  0.0073, -0.0060, -0.0105,  0.0058, -0.0060, -0.0020,
         0.0575,  0.0102,  0.0016,  0.0296, -0.0117, -0.0135, -0.0026, -0.0210,
         0.0487,  0.0689,  0.0693,  0.0700,  0.2034,  0.1024,  0.0260,  0.0788,
         0.0458,  0.0478,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1377, -0.0101, -0.0083, -0.0031, -0.0080,  0.0049, -0.0016, -0.0069,
         0.0350, -0.0018, -0.0005,  0.0122, -0.0027, -0.0094, -0.0019, -0.0104,
         0.0509,  0.0502,  0.0498,  0.0580,  0.1477,  0.0396,  0.0442,  0.0607,
         0.0997,  0.0408,  0.0528,  0.0396,  0.0116,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0308, -0.0030, -0.0036, -0.0010, -0.0126,  0.0428,  0.0083,  0.0018,
         0.0436,  0.0140, -0.0125, -0.0140,  0.1723,  0.0373,  0.1975,  0.0427,
         0.0592,  0.0505,  0.0729,  0.0413,  0.0428,  0.0539,  0.0164,  0.0252,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0123, -0.0092,  0.0028, -0.0135, -0.0156,  0.0529,  0.0099, -0.0081,
         0.0437,  0.0006, -0.0146, -0.0111,  0.1898,  0.0462,  0.0346,  0.1698,
         0.0632,  0.0508,  0.1484,  0.0637,  0.0184,  0.0207,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0350, -0.0026, -0.0068, -0.0105, -0.0084,  0.0308,  0.0070, -0.0020,
         0.0301,  0.0040, -0.0135, -0.0124,  0.1454,  0.0489,  0.0518,  0.0657,
         0.0535,  0.1174,  0.0638,  0.0704,  0.1064,  0.0649,  0.0227,  0.0258,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0167, -0.0104, -0.0031, -0.0063, -0.0137,  0.0362,  0.0128,  0.0009,
         0.0365, -0.0081, -0.0028, -0.0052, -0.0188,  0.0610,  0.0398,  0.0431,
         0.0457,  0.0901,  0.1364,  0.0447,  0.0310,  0.1882,  0.0446,  0.0413,
         0.0322,  0.0303,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1203, -0.0052, -0.0022, -0.0082, -0.0127,  0.0325,  0.0055, -0.0005,
         0.0281, -0.0039, -0.0002, -0.0017, -0.0079,  0.0417,  0.1378,  0.0480,
         0.0351,  0.0331,  0.0413,  0.1093,  0.0343,  0.0429,  0.0188,  0.0398,
         0.0767,  0.0463,  0.0291,  0.0156,  0.0213,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3242e-02, -3.3719e-03,  1.0157e-04, -9.7421e-03, -1.8983e-02,
         4.9150e-02,  5.9642e-03, -5.9214e-03,  5.0804e-02, -2.9096e-03,
        -6.0622e-03, -1.1300e-02, -8.3060e-03,  7.4460e-02,  6.1598e-02,
         1.6669e-02,  5.8723e-02,  5.7400e-02,  5.0935e-02,  5.8867e-02,
         5.1101e-02,  5.8654e-02,  1.3020e-01,  4.9611e-02,  4.6708e-02,
         3.9431e-02,  2.9784e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 3.6438e-02,  4.3409e-04,  8.5690e-03,  7.5040e-04, -1.0723e-03,
         3.7088e-03, -7.3123e-03,  6.5869e-03,  1.5122e-03,  2.2642e-03,
         1.1465e-02, -2.8455e-03,  3.8708e-03, -4.6392e-03,  3.2208e-03,
         5.2195e-04,  3.4114e-02,  1.5415e-02,  3.9781e-02,  2.8294e-02,
         2.1035e-02,  1.8902e-02,  2.7727e-02,  1.0315e-02,  8.0622e-02,
         4.0867e-02,  8.1131e-02,  4.4995e-01,  6.7738e-03,  3.2394e-02,
         1.2305e-02,  6.7978e-04, -4.4819e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.1487e-02, -5.7990e-04,  1.5066e-03, -5.5725e-03,  6.9878e-03,
         1.5177e-02, -1.1970e-04, -2.9878e-03,  7.7376e-04,  7.9190e-03,
        -2.8418e-03, -6.3534e-04,  2.4264e-03,  8.6485e-04, -8.6777e-04,
         7.5034e-03, -4.1422e-03,  1.6688e-03,  2.6498e-03,  5.7848e-02,
         3.2539e-02,  8.0102e-03,  3.0279e-02,  5.2494e-01,  2.8639e-02,
         2.0627e-02,  1.4024e-02,  1.4459e-02,  5.0388e-02,  1.8052e-02,
         1.6778e-02,  1.3762e-02,  2.6515e-02,  2.8707e-02,  5.5906e-03,
        -6.6311e-04, -1.4714e-03,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1174e-02,  7.6120e-04,  4.5138e-03, -8.5716e-03,  7.3043e-03,
         1.5957e-02, -6.0728e-04, -2.7609e-03,  1.8894e-04,  7.3712e-03,
         3.9909e-04, -5.5430e-04, -2.6163e-04,  4.5141e-03,  4.7865e-03,
         1.0189e-02, -5.8487e-03, -1.8461e-04,  1.0944e-03,  2.3157e-02,
         2.3330e-02,  1.0397e-02,  3.2537e-02,  5.7579e-01,  2.4166e-02,
         1.5893e-02,  1.8976e-02,  1.8251e-02,  3.3153e-02,  1.3079e-02,
         1.9583e-02,  1.4900e-02,  3.3199e-02,  2.9511e-02,  7.7336e-03,
         9.2340e-03, -6.7708e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.0560e-02,  3.7863e-03,  2.2198e-03, -1.0997e-02,  6.3580e-03,
         1.8423e-02,  1.4935e-03, -2.7831e-03,  2.5357e-03,  8.3329e-03,
        -7.8115e-05,  2.4009e-03,  1.1465e-03,  6.5698e-03,  4.4513e-03,
         8.0244e-03, -6.3918e-03,  1.9829e-03,  2.4311e-03,  3.1803e-02,
         1.2745e-02,  6.8452e-03,  3.8604e-02,  4.2429e-01,  4.3244e-02,
         1.6285e-02,  1.7918e-02,  6.2381e-02,  1.0619e-02,  5.8585e-02,
         1.9471e-02,  1.3553e-02,  1.6406e-02,  3.3225e-02,  3.5259e-02,
         9.6619e-03,  6.9636e-03, -1.1783e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7431e-02, -1.7466e-03,  1.0511e-03, -1.8819e-04,  3.4034e-03,
        -1.8164e-03,  2.3500e-03, -3.3693e-03, -8.0250e-04, -8.1402e-03,
         2.7748e-04,  6.8975e-05, -3.5361e-03, -1.6353e-04,  5.8836e-03,
        -3.6441e-04, -6.3689e-04,  2.4382e-01,  1.4488e-02,  1.3141e-01,
         2.8253e-02,  5.5167e-02,  6.5859e-02,  2.6426e-02,  1.9206e-02,
         3.2638e-01,  1.7470e-02, -4.2326e-03, -6.0590e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7256e-02, -1.4407e-03, -1.1258e-03,  2.5108e-03,  5.1273e-03,
        -2.4802e-03,  2.4996e-03, -5.3021e-03,  3.8848e-03, -9.0245e-03,
        -3.7055e-04,  1.4886e-03, -6.0365e-03,  2.6950e-03,  4.4282e-03,
        -5.8684e-04,  3.4059e-03,  1.0953e-02,  3.7626e-01,  5.5825e-03,
         2.5713e-03,  1.3003e-02,  3.4295e-02,  4.1692e-01,  3.3539e-02,
         1.4769e-02, -1.3217e-03,  1.1296e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.2302e-02, -2.9828e-05, -2.1963e-03,  6.9597e-03,  1.0795e-02,
        -6.3375e-03,  8.4367e-03, -1.1488e-02,  5.6270e-03, -1.9324e-02,
        -2.1308e-03,  7.3815e-03, -3.2853e-03,  1.3669e-02,  1.2053e-02,
         1.5915e-03,  6.9589e-04,  4.9096e-01,  4.1128e-02,  2.4548e-01,
         8.8711e-03,  2.0296e-02,  9.6515e-03,  6.9463e-04,  8.6172e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2816e-06, -6.0459e-03,  2.1947e-02,  3.0760e-03,  8.5327e-03,
         1.4315e-02,  1.3925e-04,  8.3456e-03,  8.3744e-03, -3.0405e-03,
         9.0057e-03,  2.6347e-02,  3.3715e-02,  1.7676e-01,  6.3637e-01,
         3.5992e-02, -7.9952e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.2062, -0.0147,  0.0422,  0.0154,  0.0145,  0.0248,  0.0034,  0.0075,
         0.0166,  0.0016,  0.0152,  0.0938,  0.0856,  0.2291,  0.1332,  0.0910,
        -0.0006,  0.0047,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5338e-02, -4.3941e-03,  1.1520e-02, -1.6991e-03,  6.5779e-03,
         7.8615e-03,  2.5356e-03,  3.6970e-03,  2.2294e-03, -3.3198e-03,
        -2.8468e-03,  3.2680e-02,  1.6651e-02,  1.6582e-01,  6.6904e-01,
         2.4325e-02, -9.2837e-03, -1.8647e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7631e-02,  6.5291e-03,  3.7839e-03, -1.1509e-03,  7.5077e-03,
        -1.8205e-03,  2.6359e-04,  1.2313e-02,  9.6451e-04,  1.8006e-03,
        -1.0014e-03,  1.7559e-03,  3.3547e-01,  1.1255e-02,  2.4682e-02,
         1.6355e-02,  7.9250e-02,  2.0680e-02,  7.8883e-02,  3.5746e-01,
        -3.8280e-03, -3.7654e-03, -1.8504e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.5462e-02,  8.6851e-03,  4.4321e-03, -1.2869e-03,  1.1899e-02,
        -7.1018e-03, -4.1444e-03,  1.6759e-02,  1.8636e-03,  1.6038e-03,
        -1.6237e-03,  1.4459e-03,  5.9232e-01,  2.3360e-02,  4.4135e-02,
         1.9798e-02,  1.8546e-01,  2.6676e-02,  7.3328e-05,  1.8760e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #200: [tensor([-1.2499e-02, -1.5403e-03,  2.4827e-04,  1.6752e-03, -1.8138e-03,
        -1.1917e-03,  1.6947e-03,  9.2523e-04,  3.8692e-03, -9.6975e-04,
         1.3207e-04,  4.9169e-03, -1.9261e-04,  1.3104e-03,  6.4017e-04,
         3.6899e-01,  1.0205e-04, -1.3489e-03, -2.7414e-03, -2.4190e-04,
         5.9236e-01,  3.5032e-04, -1.9565e-04,  4.6964e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4192e-02, -2.0444e-03, -6.5163e-04,  1.5607e-03, -1.0226e-03,
        -1.2122e-03,  7.7494e-04, -3.4713e-04,  4.0206e-03, -3.7439e-04,
         4.6775e-04,  2.1826e-03, -1.2377e-03, -1.9273e-04,  5.8452e-04,
         5.2916e-01, -1.7314e-03, -2.1728e-03,  1.3717e-04, -2.3590e-03,
        -3.5683e-05, -3.2297e-03,  2.4842e-03,  4.2293e-01,  1.5559e-03,
        -2.7325e-03, -2.0453e-04, -4.0177e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2291e-02, -1.9038e-03,  1.4530e-04,  1.1302e-03, -9.4348e-04,
        -1.1059e-03,  1.1465e-03, -8.3625e-04,  4.1384e-03,  4.1955e-04,
         4.2951e-04,  5.4853e-03, -1.1534e-03,  6.1997e-04,  1.7822e-04,
         3.6338e-03,  3.6785e-01,  3.9768e-04, -1.5724e-04,  2.2458e-03,
         5.7761e-01,  2.6542e-03,  7.7532e-04, -2.3649e-03,  3.8217e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7968e-02,  6.7510e-03,  1.9611e-03,  4.5646e-03, -1.4893e-03,
         3.1973e-04,  2.6173e-03,  5.3941e-03,  1.7045e-02, -6.3475e-04,
        -1.8868e-04,  1.9206e-02, -6.9941e-04,  5.0305e-04,  3.6829e-05,
         1.7523e-02,  8.9527e-01,  2.3818e-03,  4.1452e-03,  1.2994e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8393e-02,  1.3974e-04,  1.9502e-03,  3.4114e-03, -8.9222e-04,
         2.1857e-04,  2.5727e-03,  1.5208e-03,  9.9775e-03, -1.4458e-03,
         1.4667e-03,  1.4723e-02, -1.1669e-04,  6.1375e-04,  1.8603e-04,
         1.9263e-03, -7.6201e-04,  9.1053e-01,  6.2671e-03,  5.3221e-03,
        -4.7275e-03, -1.4320e-03,  1.0254e-03, -3.7700e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6512e-02, -1.0473e-03,  9.2939e-03, -1.2379e-03,  3.1286e-03,
        -1.5265e-03, -1.4563e-03,  5.1169e-04,  1.6882e-03,  7.1730e-04,
         6.4942e-01, -5.2330e-04,  2.8983e-01, -4.3220e-03,  1.4996e-02,
        -3.1965e-04, -2.8702e-03,  6.0430e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4934e-03, -1.8076e-03,  2.2983e-03, -5.5812e-04,  4.9378e-04,
        -3.4582e-03, -9.8871e-04,  1.1053e-03,  8.4121e-04,  4.5195e-04,
         3.0292e-03,  8.8940e-02,  6.0987e-01,  4.6725e-04, -7.2365e-04,
         4.6151e-03, -1.8814e-03, -1.2996e-04,  2.7345e-01,  5.5123e-04,
        -1.2372e-03, -1.6130e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.6235e-04, -5.9302e-04,  6.2236e-03,  1.4523e-04,  2.2116e-03,
        -1.4318e-03, -1.7129e-03,  5.1771e-04,  3.7376e-05,  2.2811e-03,
         5.0837e-01,  2.3524e-03,  4.6503e-01, -2.8763e-03, -1.8117e-03,
         5.4055e-04,  4.3561e-04, -1.6711e-03,  9.9306e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9318e-04, -1.5290e-03,  3.6026e-03, -9.3344e-04,  9.2291e-04,
        -4.9701e-03, -1.7681e-03,  2.2444e-03,  6.5423e-04,  7.1575e-04,
         5.2971e-01, -1.5104e-04,  4.4733e-01, -4.6435e-03, -4.1595e-04,
        -1.1176e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0521e-03, -1.5952e-03,  4.7406e-03, -2.0784e-04,  1.1707e-03,
        -2.2487e-03, -1.0334e-03,  9.3591e-04,  2.1078e-03,  1.1804e-04,
        -3.2314e-03,  1.2368e-01, -8.9728e-04,  1.5167e-03,  8.4856e-01,
        -1.8084e-03, -1.7638e-03,  5.7245e-04, -1.7593e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7637e-04, -8.5188e-04,  5.4008e-03,  2.6355e-04,  6.9435e-04,
        -1.3069e-03, -9.8736e-04,  2.7568e-03,  2.0039e-03,  1.8974e-03,
         5.0034e-03,  3.9130e-01, -1.9517e-04, -2.2487e-03,  3.7021e-03,
         5.7213e-01, -9.9840e-04,  3.2158e-03, -2.3804e-03, -2.1806e-04,
         2.1928e-04, -8.1009e-04, -9.4029e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2880e-02,  1.5886e-03, -2.4153e-03, -3.8249e-04, -2.9224e-03,
         4.7963e-03,  4.0194e-04, -2.3392e-03, -1.7664e-03, -1.8832e-03,
         1.4762e-04, -1.8880e-01,  1.0078e-03,  8.5029e-04, -7.7465e-01,
        -2.0165e-03,  1.1509e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 3.4351e-03,  1.0995e-03,  6.6872e-04,  1.1115e-04, -3.4266e-04,
         1.0250e-03,  4.2045e-04,  6.5958e-04,  9.6346e-01, -2.5539e-03,
         5.3726e-03, -4.0890e-03,  4.7858e-03,  4.0825e-03,  5.2466e-03,
         2.6454e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.3307e-03,  9.1924e-04,  8.5903e-04, -1.6056e-03,  1.7163e-03,
        -1.6830e-03, -6.9176e-04, -2.6721e-03, -9.7249e-01,  1.1145e-03,
         2.2510e-03,  4.7257e-04, -3.6334e-03, -1.5553e-03,  6.5595e-04,
         3.4745e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0734e-03,  1.4695e-03, -1.6696e-03,  5.5860e-04,  3.3644e-04,
         3.7294e-03, -1.0055e-03, -2.7298e-04,  6.6810e-04, -1.1541e-03,
         1.3938e-03, -2.7366e-04,  4.0540e-04,  8.4325e-04,  2.4409e-03,
         1.7972e-03,  9.7480e-01,  1.3160e-03, -8.4802e-04, -1.1166e-03,
         8.3290e-04,  1.9972e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5079e-02,  9.5123e-04,  1.6183e-03,  1.1728e-04, -1.6313e-04,
        -1.9688e-03,  1.1790e-03, -2.4189e-04,  5.5904e-04,  1.1556e-03,
        -1.2712e-03, -7.6328e-04, -1.0951e-04,  3.7212e-04, -5.7415e-01,
         2.2363e-03, -1.7495e-03,  2.0890e-03, -3.9073e-01,  2.5124e-03,
         8.4968e-04, -1.3262e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.8991e-03,  6.6271e-04, -6.1514e-04, -1.1181e-03, -1.5803e-04,
         5.9269e-03,  5.1551e-04, -3.1277e-04, -7.9195e-04, -1.3300e-03,
         5.3554e-03, -1.4681e-03, -8.8813e-04, -8.1689e-04,  1.8546e-03,
         2.3148e-03,  9.6246e-01,  1.6348e-03,  3.4243e-03, -2.8557e-04,
        -1.7014e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3428e-03,  1.1959e-03,  5.9434e-04,  1.3961e-03, -1.1973e-03,
         1.9768e-04,  1.5532e-04,  7.7845e-04,  1.1309e-03, -9.3395e-04,
        -4.2752e-04,  2.3032e-01, -3.0608e-03,  3.6646e-01,  1.5031e-06,
         3.7684e-01,  1.5072e-04, -1.0433e-03,  8.5689e-04,  3.9101e-03,
         1.2476e-04, -8.1327e-04, -7.2785e-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([ 2.7187e-04, -1.3024e-04,  1.2251e-03,  1.2101e-03, -2.0777e-05,
        -7.0693e-04,  1.7286e-03, -4.4445e-04,  4.4869e-04,  1.4325e-04,
         1.0751e-05,  8.8145e-02, -2.9815e-03,  5.1363e-01, -3.1872e-04,
        -1.2550e-03, -3.3674e-04,  3.7958e-01, -1.2049e-03, -1.7352e-03,
        -1.8879e-04,  3.0304e-03, -6.8457e-04, -4.1990e-04,  1.5092e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1601e-04, -1.4130e-03,  3.9194e-04,  1.1574e-04, -1.4616e-04,
        -3.4887e-04,  3.6345e-04, -1.4663e-04,  6.1852e-04, -4.3281e-05,
        -1.8576e-05,  3.1193e-01,  1.1236e-03, -4.3030e-04, -3.4514e-04,
         2.5041e-04,  1.0401e-04,  6.7573e-01,  2.4429e-03, -1.3402e-03,
        -1.1122e-03, -1.4766e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6631e-03,  2.6049e-04,  4.0534e-04, -1.5875e-04,  8.0782e-05,
         2.0768e-04,  7.3829e-04,  9.5559e-04, -5.7949e-04,  7.4848e-04,
         1.4656e-03,  5.2659e-04, -3.6311e-04, -3.5042e-04, -3.2153e-04,
         1.9354e-05,  3.7028e-04,  3.1870e-04,  4.0618e-04, -1.9203e-04,
        -2.6009e-03, -8.8901e-02, -4.9458e-01,  1.1666e-03, -9.3438e-02,
        -3.0193e-01,  9.4045e-04,  9.7929e-04,  1.4157e-03,  8.9636e-04,
        -1.0180e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5649e-03, -7.0062e-05, -1.0209e-03, -1.0296e-04, -7.8896e-04,
        -4.4994e-04, -5.6961e-04, -1.1681e-03,  5.6383e-04, -1.0775e-03,
        -6.7810e-04, -8.6602e-04,  4.3700e-04, -4.3174e-06,  7.2388e-04,
         1.0568e-04, -3.4536e-04, -7.7347e-04, -1.2741e-04, -1.1970e-06,
        -1.2866e-03,  2.8949e-01, -1.4335e-04,  4.9836e-01,  3.1596e-03,
         9.7211e-04,  1.8635e-03, -1.4234e-03,  1.8486e-01,  2.5005e-03,
         8.6067e-04,  1.0461e-03, -1.1224e-03, -4.7067e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0005e-03,  3.2312e-04,  1.0776e-03,  4.3125e-04, -2.9988e-04,
         8.5424e-04,  3.2573e-04,  2.0225e-05, -1.9156e-05,  7.8087e-04,
         9.1855e-04,  9.7840e-04, -6.6779e-04, -4.0391e-04, -4.0337e-04,
         2.5024e-04,  4.5272e-04,  7.8279e-04, -6.4953e-04, -1.4132e-04,
         4.9318e-05,  3.6866e-04, -1.6134e-01,  7.0234e-04, -3.4870e-01,
         9.6864e-04, -2.2399e-04,  9.2035e-05,  1.0006e-04, -4.7140e-01,
         1.5326e-03,  3.8258e-04, -3.5798e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.8832e-03, -5.6678e-05, -1.0445e-03, -4.3062e-04,  3.7954e-04,
        -2.3065e-03,  1.7732e-04, -1.8062e-04,  9.4940e-04,  2.5853e-05,
         3.7540e-01, -3.2820e-03,  5.9961e-01, -2.3643e-04, -3.9183e-03,
         9.4314e-04,  9.4989e-04,  3.9523e-04,  1.0813e-03,  7.4927e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  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 #300: [tensor([ 3.3158e-04,  7.5046e-04,  1.0426e-03, -7.8132e-05,  1.1346e-04,
         4.6088e-04,  6.2340e-04,  3.3834e-04, -1.1261e-04,  4.8952e-05,
         1.1626e-04, -4.6025e-04,  7.6464e-04,  4.8392e-04, -3.6183e-04,
        -1.8981e-04,  2.2933e-04, -3.8165e-04, -5.1340e-04, -4.6165e-04,
         7.8936e-04,  7.8519e-01, -1.4217e-03, -1.2067e-04,  9.0558e-05,
        -2.8879e-04,  2.0200e-01, -3.7844e-07, -6.4689e-04, -4.6511e-04,
         4.7605e-04,  6.5023e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6807e-03,  2.7456e-03,  2.6460e-04,  1.0398e-03,  1.8895e-04,
         1.6831e-03,  8.8660e-04,  4.5695e-04, -3.5394e-04,  9.8484e-04,
         8.9737e-04,  1.6255e-03,  5.4302e-03,  1.6352e-04,  8.8060e-04,
        -2.8304e-04,  7.3247e-04, -3.0369e-04, -1.2190e-03, -1.5557e-04,
        -2.8678e-04,  3.8040e-03,  1.5819e-02,  9.4370e-01,  9.3226e-04,
         2.4236e-03,  3.7769e-03,  2.9266e-03, -5.2315e-05,  1.6539e-03,
        -1.6477e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9523e-03, -2.7137e-04, -1.5873e-03,  5.1345e-05, -7.1012e-04,
         3.7463e-04, -3.3397e-04,  3.5642e-05,  2.5569e-04, -2.1521e-04,
         3.6931e-04, -1.8362e-04,  2.7785e-04,  4.6485e-04,  5.7798e-04,
        -6.2414e-04,  7.2271e-05,  6.3526e-04,  5.0084e-04, -1.4050e-04,
         5.4284e-04, -2.7385e-01,  9.3144e-04, -2.6609e-01,  2.7788e-04,
         2.8807e-03, -4.4009e-01, -1.6889e-03,  2.2285e-03,  7.9139e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7852e-03,  4.8578e-05,  1.0730e-03,  6.7991e-04,  1.2166e-04,
         8.6651e-04,  2.9207e-04, -6.3483e-04, -6.2618e-05,  4.5950e-04,
        -5.4788e-04,  2.8091e-04,  6.8918e-04, -4.0929e-04, -8.2078e-05,
        -5.0453e-04,  1.7917e-04,  2.4725e-04, -5.4121e-04, -6.7653e-04,
        -1.4007e-04,  7.2848e-01, -1.9027e-03,  2.5572e-01, -5.2427e-05,
         6.2332e-04, -6.9678e-04, -2.0174e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7599e-03,  1.3371e-03, -5.1078e-05,  4.3911e-04,  1.1344e-04,
         7.2486e-04, -1.6301e-04,  8.2734e-05,  7.4078e-04,  6.6626e-04,
         4.1030e-05,  4.6963e-04,  2.6456e-03, -6.9576e-05,  6.6424e-04,
        -6.6692e-04, -1.4340e-04,  3.6282e-04, -3.0524e-04, -9.3784e-04,
         1.2148e-04,  5.2254e-01,  9.5698e-04,  1.2953e-03,  1.5565e-04,
         4.4675e-01,  1.7259e-03,  8.2202e-03, -9.6791e-04,  3.3108e-04,
        -1.2141e-03, -3.4095e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5636e-04,  9.1088e-05,  2.4492e-04,  3.4330e-04, -5.6264e-04,
        -3.7911e-04,  4.7888e-04,  4.1050e-04,  4.9250e-04,  3.8328e-04,
        -1.4018e-04,  3.5164e-04,  2.7916e-04,  4.3809e-04, -1.1086e-04,
         1.1320e-04,  5.5783e-06,  1.6600e-04, -1.5155e-04,  9.0255e-05,
        -1.1395e-04,  4.9969e-04,  3.6272e-04,  6.0056e-04, -4.2924e-04,
         3.2957e-04,  1.7285e-04, -3.6212e-04, -1.4867e-04, -4.5090e-04,
        -2.3461e-01, -7.0769e-04, -1.1989e-01,  7.2509e-04, -6.3142e-01,
         4.6255e-04, -9.1725e-04, -1.8251e-03, -3.7138e-04, -7.7973e-06,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1839e-03,  5.6565e-04,  2.9188e-04, -1.6065e-04,  1.4900e-03,
         5.6166e-04, -1.8120e-04, -2.8876e-04, -3.1234e-04,  3.1922e-04,
         4.2640e-04,  1.2982e-04, -1.5853e-04, -1.0090e-03,  6.5240e-04,
         9.4574e-04,  7.7132e-04, -2.1299e-04,  5.5215e-04, -6.1032e-06,
         4.3718e-04,  8.0730e-04,  8.8216e-05, -4.4977e-04,  3.8079e-04,
         5.2514e-04,  9.1537e-04,  4.6089e-04,  9.8074e-05,  3.9544e-05,
        -5.8968e-04,  4.5963e-03,  4.9541e-01,  1.9973e-03,  7.3592e-03,
         3.9076e-03,  3.2879e-04,  4.6478e-01,  3.8183e-04, -6.9370e-04,
         5.3377e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9930e-03,  3.9279e-04, -4.1593e-04, -4.0835e-04,  1.2066e-03,
         2.8647e-04,  1.8695e-04,  4.9748e-05, -4.1441e-04, -3.6675e-04,
         3.4036e-04, -2.3128e-04,  5.2883e-05, -5.3988e-04,  1.3763e-04,
         5.6005e-05,  1.2191e-03, -4.6322e-04,  7.2472e-05,  1.8222e-04,
         3.4653e-04,  1.5654e-04,  3.4065e-05, -7.1322e-04,  5.0455e-04,
         1.6891e-04,  8.0955e-04, -1.2793e-05,  8.3690e-05, -1.8091e-04,
         7.6087e-01,  1.1212e-03,  8.6239e-05,  2.2023e-01,  8.7680e-04,
         1.3095e-03, -3.0737e-05,  8.6971e-04, -6.9253e-05,  2.2467e-04,
         2.9015e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8471e-03,  1.1417e-03, -7.5945e-05, -8.8792e-05,  1.9190e-03,
         1.1645e-03,  9.1563e-04, -3.5187e-04,  1.6297e-04,  3.5725e-04,
         6.3285e-04,  1.3261e-04, -2.9671e-04, -6.0222e-04,  4.9180e-04,
         3.1153e-04,  7.7585e-04, -1.2116e-04,  4.6982e-04,  1.4381e-04,
         5.3388e-04,  7.5922e-04, -6.5015e-04, -3.5493e-04,  4.3013e-04,
         2.4893e-05,  1.7276e-04,  2.4998e-04, -8.8701e-05,  5.2038e-05,
         7.4578e-01, -3.3073e-05,  1.5277e-03,  2.2585e-01,  1.3352e-03,
         3.3872e-03,  1.0115e-03,  5.7787e-04,  6.0088e-04,  1.1828e-03,
         7.8755e-04, -5.1942e-04,  9.0048e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3164e-03,  9.8803e-04,  3.4289e-04,  1.2239e-04,  5.9248e-04,
         1.0743e-03,  3.5145e-04, -3.7379e-04, -2.6512e-04,  4.3076e-05,
         4.1097e-04,  3.4443e-04,  1.4692e-04, -6.2674e-04,  2.5588e-04,
         4.6056e-04,  4.2119e-04,  4.0374e-05,  5.9649e-04,  8.4440e-05,
        -4.6610e-07,  3.1178e-04,  2.8708e-04, -1.0802e-03,  2.3298e-04,
         1.2924e-04,  4.0583e-04,  2.1684e-04, -6.5318e-05,  1.7815e-05,
         3.0748e-01,  4.2710e-04,  5.5519e-04,  3.0236e-01, -6.2567e-04,
         3.7440e-01,  5.9718e-04,  3.4626e-04, -4.7719e-04,  1.2314e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.1657e-03,  3.1230e-04,  1.3112e-04,  2.5900e-04,  3.9799e-04,
         4.9868e-04, -9.4931e-05, -8.0442e-05, -7.2662e-04,  3.6945e-04,
         2.1852e-04,  3.8579e-04, -4.4829e-04, -7.5660e-04,  5.1641e-05,
         4.0312e-04,  7.6062e-04, -1.4188e-04,  2.4354e-04,  4.3905e-04,
         3.2165e-04,  6.8457e-04,  2.6931e-04, -4.0372e-04,  1.2878e-03,
         5.8249e-04,  8.0888e-04,  4.0627e-04, -3.2377e-04, -1.2718e-04,
         2.8254e-01,  7.8169e-04,  5.6034e-04,  5.2718e-04, -1.2604e-04,
         6.9460e-01,  1.0905e-03, -4.7436e-04,  8.7843e-04, -1.2710e-03,
         4.7774e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.6640e-02,  3.6271e-02, -2.0267e-02, -5.5041e-03,  5.3548e-02,
         2.5130e-05,  7.3097e-03, -7.7700e-04,  2.2782e-04,  5.8152e-03,
         2.0164e-02,  6.7503e-03,  8.2498e-03, -6.1143e-04,  9.5975e-03,
         7.6691e-03,  1.6918e-02,  1.2682e-02,  7.6956e-03,  5.9272e-03,
         1.3197e-02,  4.0846e-02, -8.0053e-03,  8.3795e-03, -1.4767e-03,
         1.3050e-02,  4.4540e-02,  1.1115e-02, -4.0880e-03, -2.2520e-03,
         6.2745e-02,  2.9324e-01,  8.4685e-02,  4.0874e-02, -2.8484e-02,
         2.0370e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #350: [tensor([-2.9181e-03,  3.0522e-04, -6.7220e-04, -1.5196e-04, -3.8924e-04,
        -1.0275e-03, -2.1589e-04, -2.5672e-04, -1.7818e-04,  9.2838e-06,
         2.3022e-04, -5.7064e-04, -1.0258e-05,  7.3757e-05, -4.8596e-04,
        -3.4000e-04,  9.8371e-06, -2.1393e-04,  2.8243e-01,  1.8486e-04,
         1.8043e-03,  1.1943e-04,  7.0571e-01, -2.4466e-04,  7.3972e-04,
         7.1128e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8552e-03, -8.0441e-04,  1.3204e-04,  4.4230e-04,  1.2809e-03,
         9.5589e-04,  3.7472e-05,  1.4590e-03,  1.5296e-03, -7.2391e-05,
         5.1287e-04, -9.6638e-04,  7.0662e-04, -2.3400e-04,  9.8460e-01,
         3.0858e-04, -1.1067e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1175e-03,  3.5390e-05,  1.2389e-04,  3.9637e-05, -2.3543e-04,
        -6.3398e-05,  2.3880e-04,  5.3163e-04, -1.1881e-03, -6.5081e-05,
         4.3032e-05,  2.9753e-04,  3.2724e-04,  6.1333e-04,  5.4616e-01,
         2.9244e-04,  4.2484e-03,  4.3867e-01,  6.9807e-04,  9.7743e-06,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.7707e-03, -7.0674e-05,  6.6698e-04, -3.2834e-04, -8.0643e-04,
         3.2470e-04,  2.7850e-04,  8.9299e-05,  8.8091e-04,  6.3977e-04,
         3.0698e-04,  5.2244e-04,  1.4603e-04,  1.6727e-03, -2.7057e-01,
         6.3619e-04, -2.1424e-03, -7.1200e-01, -1.0694e-04, -3.8117e-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([ 2.5576e-04, -1.2468e-04,  2.6560e-04, -2.3157e-04,  5.4512e-04,
         7.0491e-04, -4.3126e-04, -5.1871e-04, -6.2032e-04, -6.7707e-04,
        -3.9838e-04,  1.0765e-03,  6.4472e-05, -9.9497e-06, -9.8743e-01,
         5.4440e-04, -4.0009e-04, -2.8826e-03,  6.4588e-04, -1.9807e-03,
        -1.2241e-04,  7.1674e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3177e-04,  1.8409e-04,  1.6951e-04,  9.9549e-04,  6.2392e-04,
         6.3357e-04, -4.0863e-04,  8.2370e-05, -4.6103e-04,  5.5444e-04,
        -7.2571e-04,  7.8356e-04, -3.7014e-05,  4.7782e-04, -9.8824e-01,
        -3.9067e-04, -1.1683e-03,  6.9282e-04,  1.2266e-03, -1.9084e-04,
        -6.1501e-04,  9.0479e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3622e-04, -1.2697e-03, -4.7419e-04, -1.0137e-03,  1.3573e-04,
        -8.2680e-04, -1.0955e-03, -1.0359e-03, -9.4054e-04,  1.2353e-03,
        -7.5149e-04, -5.3017e-04, -1.4794e-05, -1.1593e-03, -9.5179e-01,
        -6.6906e-04, -3.1088e-03, -2.3361e-02, -7.8052e-03, -3.2574e-04,
        -8.2110e-04,  1.4014e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4692e-02, -1.4626e-03, -4.1738e-04, -1.7796e-03, -1.3318e-03,
         5.3505e-04,  6.1427e-04,  6.9694e-04, -5.8809e-04, -4.7748e-04,
         6.2963e-04,  2.4288e-04,  1.7851e-03, -4.6488e-03, -9.4528e-01,
        -7.0880e-03, -1.5704e-03, -5.3113e-03, -7.0765e-03, -3.7759e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.8155e-03, -9.4736e-04, -5.8470e-04,  6.6229e-04, -4.5363e-04,
        -6.5868e-05, -7.3993e-04,  1.9151e-04,  1.4074e-03, -4.4456e-04,
         3.9940e-04, -1.6396e-04, -9.0933e-02, -1.4689e-03, -8.9007e-01,
        -4.8087e-04,  8.8202e-04, -2.8858e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8499e-03,  1.9991e-04,  3.4451e-04, -4.8251e-05,  2.5228e-04,
         7.9251e-04, -3.9985e-04,  2.4226e-04, -1.2231e-04, -3.5475e-05,
         4.8530e-05, -5.7315e-04, -4.0843e-01, -6.5759e-04, -1.4240e-01,
        -2.7011e-04,  2.6633e-04, -1.4309e-03, -4.3702e-01, -1.2316e-03,
         9.3287e-04, -4.5635e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8266e-03,  3.8586e-04,  4.7654e-04, -2.0717e-04,  4.3357e-05,
        -5.2176e-05, -4.2901e-04, -1.7324e-04, -6.4626e-05,  4.8545e-04,
        -2.0782e-04, -7.1707e-05,  1.9451e-05, -6.8550e-01,  1.7478e-04,
        -8.4897e-04, -5.3630e-04, -2.9898e-05, -1.8278e-03, -3.0443e-01,
        -4.3240e-04,  5.5614e-04,  1.2211e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7195e-03, -7.5626e-04, -6.5277e-04,  2.6954e-04, -2.1774e-05,
         4.5984e-05, -3.1057e-04,  4.8729e-04,  2.1923e-05,  7.9214e-04,
        -8.1273e-04,  6.9118e-04, -3.3657e-04, -6.7774e-01, -8.8323e-04,
        -2.0926e-03, -5.0311e-04, -1.9189e-03, -3.0666e-01,  7.5124e-04,
         8.6778e-04, -6.6354e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 4.7819e-03,  1.8329e-05, -3.0040e-04, -3.2455e-05, -1.1566e-04,
         1.2439e-04,  2.1238e-05, -1.4029e-04,  1.4309e-05,  1.6370e-04,
         1.3725e-04,  1.3973e-04, -2.7292e-04,  4.6086e-05,  8.6935e-05,
        -2.1639e-04,  7.4402e-04, -6.0359e-06,  8.5731e-05, -4.8193e-05,
        -1.4165e-04, -1.7925e-04,  1.7438e-05,  2.1582e-04,  4.2040e-05,
         4.7381e-04,  2.4540e-04,  2.6656e-04,  2.0333e-04, -1.6702e-05,
        -3.9679e-04,  3.9464e-04,  9.4215e-04, -1.6994e-04,  3.9579e-01,
        -3.6363e-04,  1.2574e-03,  5.8847e-01,  1.7614e-04, -3.0120e-04,
         1.9559e-04,  7.0950e-05,  9.5704e-04,  8.2007e-04, -2.9869e-04,
        -1.0122e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1347e-03,  2.3802e-04,  2.5750e-04,  7.4443e-05,  2.3990e-04,
        -2.4576e-04, -1.2191e-05, -3.9243e-04,  4.3004e-04, -7.0818e-05,
        -1.1836e-04,  2.7391e-04,  3.2561e-04,  1.7679e-04,  3.1117e-04,
         3.1855e-04,  3.2715e-04,  4.3760e-06,  7.1765e-05, -1.3896e-04,
         2.7422e-04,  1.7359e-04,  5.6110e-04, -1.7008e-04,  1.4821e-04,
        -2.9874e-04,  2.3671e-04,  2.2387e-04, -7.7405e-05,  1.1760e-04,
        -1.9293e-04,  1.1672e-04, -6.1493e-04, -4.6318e-04, -1.2304e-01,
        -1.0465e-04, -5.4800e-03, -3.3540e-01, -6.7869e-04,  4.9163e-04,
         2.1577e-04, -6.0464e-04, -5.1977e-01, -7.2152e-04, -6.5709e-04,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2097e-03,  7.5445e-05,  1.7642e-04,  1.1426e-04, -2.7519e-04,
        -4.8480e-04, -4.1154e-04, -1.0307e-04,  6.1181e-05,  3.0239e-04,
         1.0874e-04, -2.7732e-04,  2.4993e-04, -9.2549e-05, -2.2532e-04,
        -3.8751e-05, -6.4979e-05, -1.6110e-04, -1.5713e-04,  9.0985e-05,
         2.1787e-04, -1.5265e-04, -4.4705e-04, -9.6272e-05, -2.7662e-04,
        -4.3260e-04, -1.2676e-04, -1.3100e-04, -4.5390e-04, -3.1817e-04,
         1.6363e-04,  1.8944e-04, -7.6460e-04, -3.8888e-05, -1.9626e-04,
         6.0060e-04, -7.1180e-01,  1.9305e-03, -1.3967e-04, -2.7180e-01,
        -4.5325e-04,  5.8614e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.1191e-03, -1.6518e-03,  2.1440e-04, -6.2766e-04, -1.3071e-03,
        -1.3627e-04, -6.9789e-04, -5.6258e-04,  6.7722e-04, -1.3162e-04,
        -8.9783e-04, -1.8345e-03,  2.8059e-04, -5.0383e-04, -2.0519e-03,
        -6.4522e-04, -2.1944e-04, -5.0837e-04,  5.9847e-04, -8.4316e-04,
         2.0832e-04,  2.8536e-04, -2.5325e-03,  2.4893e-04, -6.8601e-04,
        -2.9225e-03,  8.0515e-04, -3.8653e-04, -2.0016e-03,  2.4802e-04,
         1.5454e-05,  1.4869e-03, -8.4241e-04,  3.6066e-03, -2.9344e-04,
        -4.1872e-03, -9.5263e-01, -2.1482e-03, -3.8309e-04, -5.7485e-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([ 1.3880e-03,  2.6051e-05,  1.1270e-04, -2.5845e-04,  3.4865e-04,
         2.4365e-04,  1.1174e-04,  1.2149e-05, -6.3012e-05, -4.0408e-04,
         3.0609e-05, -5.2070e-05, -4.6227e-05,  3.0010e-05,  5.1174e-05,
        -1.9177e-04, -2.8310e-04, -1.4744e-05, -9.1416e-05, -4.5403e-05,
        -1.5488e-04, -2.8052e-04,  1.5231e-05,  2.0614e-04,  3.6774e-04,
         2.3520e-04, -4.8043e-04,  8.5192e-05,  5.0800e-05, -3.2876e-04,
        -5.4533e-05, -1.8658e-04, -2.6345e-04,  4.0231e-04,  5.1001e-04,
         5.0718e-04,  6.9975e-01,  1.1815e-03,  2.7738e-03,  2.8750e-01,
        -3.5901e-04,  2.7291e-04, -8.6928e-05,  1.4686e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7696e-03,  1.9415e-04,  1.7456e-04,  2.0880e-04, -1.1785e-04,
         6.3684e-05,  2.5181e-04,  5.4898e-05, -3.6189e-05,  4.3228e-04,
         1.6895e-04, -1.6375e-04,  2.9980e-04,  3.5950e-04,  1.4454e-04,
         4.9777e-04,  2.9603e-04,  9.4194e-05,  1.9709e-04,  2.2677e-04,
         6.0339e-04,  1.5784e-04, -2.4428e-04, -2.4472e-04, -6.6645e-05,
        -1.1277e-04,  1.4227e-04,  2.0097e-04,  1.4270e-04,  4.9117e-04,
         1.5944e-04,  2.8647e-04, -1.0736e-03,  3.0931e-04, -3.2453e-01,
         1.0718e-04, -6.6126e-01,  6.7186e-05, -1.0410e-03,  3.5990e-04,
        -6.5406e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4419e-03, -5.8336e-05,  3.2799e-04, -5.3896e-05, -1.1434e-05,
         2.6236e-05,  7.3843e-04, -2.2381e-05,  4.8972e-04,  1.8266e-04,
         1.6396e-05,  1.8024e-04,  1.0576e-04, -9.0234e-05,  3.8463e-04,
         7.6671e-04, -7.0463e-05, -4.3796e-05, -8.4971e-07,  2.4348e-04,
         8.9434e-05, -3.2561e-04, -9.9608e-06,  2.9852e-04,  4.1598e-04,
         2.5753e-04,  7.2513e-04, -3.5354e-04,  6.7199e-04,  6.0374e-04,
        -7.6499e-05,  9.9313e-05, -5.6787e-05,  1.0140e-03, -9.8417e-01,
        -1.4541e-04,  1.0537e-03, -9.0336e-04,  4.7636e-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([-3.7106e-03, -6.0940e-04,  4.4494e-05, -5.1228e-04, -8.8730e-05,
         5.7580e-06,  2.2558e-04,  5.7716e-05, -1.2677e-04, -1.3808e-04,
        -5.9079e-04, -2.1215e-04, -1.1441e-03,  2.6616e-04, -1.9593e-04,
        -6.6163e-05, -2.0297e-04,  7.3417e-04, -7.5427e-04, -3.6418e-04,
         1.2377e-03, -1.3389e-03, -9.8455e-01,  1.3222e-03, -1.2957e-03,
        -2.0276e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0221,  0.0430,  0.0020,  0.0133,  0.0382,  0.0084,  0.0112,  0.0803,
        -0.0011,  0.0098, -0.0028,  0.0234, -0.0095,  0.0117,  0.0954,  0.0306,
        -0.0050,  0.0004, -0.0085, -0.0170,  0.0242,  0.0767,  0.3133,  0.1288,
        -0.0200, -0.0031,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0267,  0.0597,  0.0076,  0.0132,  0.0656,  0.0286, -0.0030,  0.0610,
        -0.0076, -0.0074, -0.0012,  0.0285, -0.0096,  0.0061,  0.1102,  0.0219,
        -0.0050, -0.0055, -0.0121,  0.0656,  0.2894,  0.0679, -0.0778, -0.0188,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0587e-03, -2.2101e-04,  7.4776e-04,  5.7911e-04,  5.2115e-05,
        -6.4295e-04, -1.5567e-05,  2.3518e-04, -6.0229e-04,  1.6448e-04,
        -4.4352e-04, -1.1945e-04,  8.6891e-04, -2.7516e-04, -1.8649e-04,
         4.7886e-04,  9.9394e-05,  4.7654e-03,  6.1539e-01, -5.4909e-04,
         2.7488e-01,  6.4649e-03,  9.6085e-03,  7.6683e-02,  9.9134e-04,
        -6.5330e-04, -2.2261e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3708e-03, -3.3702e-05, -6.1165e-04, -4.6384e-04,  4.6361e-04,
         6.3537e-04,  9.0238e-04, -4.0942e-05, -2.5387e-04,  6.1749e-05,
        -3.8186e-04,  2.7695e-04, -8.9568e-01, -4.5217e-04,  1.2296e-03,
         4.2443e-04, -9.2150e-02,  4.0060e-04,  2.3688e-03,  4.0446e-04,
         3.9548e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([-2.7718e-03, -7.2660e-04,  2.6073e-04,  4.0560e-05, -3.1222e-04,
        -2.9249e-04, -2.6266e-06, -3.6731e-04,  1.6191e-04, -6.4043e-04,
        -4.0184e-01,  7.5412e-04,  6.8264e-05, -2.5025e-05, -5.3885e-01,
         1.2692e-04, -4.9983e-02,  7.8174e-04,  1.1152e-03,  8.7536e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5003e-03, -1.1491e-04,  1.8265e-04,  3.3031e-04,  5.6894e-05,
         2.9158e-05,  2.0185e-04, -1.8289e-04,  1.2158e-05, -3.4528e-04,
         1.8851e-04, -1.7796e-04, -4.4506e-04,  1.8447e-04, -3.2390e-04,
         1.1130e-04,  1.4350e-04, -4.7696e-05, -4.4320e-04,  3.5844e-04,
        -2.9724e-04,  2.1627e-04, -1.7277e-04,  5.4660e-05, -2.5222e-04,
        -3.0078e-01, -3.2779e-04, -1.0714e-03,  3.5774e-04, -6.8618e-01,
        -7.8967e-04, -5.8791e-04,  8.0734e-05,  9.1421e-05,  3.6272e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4691e-03,  4.9890e-05, -8.7634e-05, -1.6615e-04,  3.6075e-05,
        -3.3304e-04,  1.9516e-04, -9.3199e-05,  2.7165e-04, -8.2777e-04,
        -4.7881e-04, -1.8281e-04, -3.9800e-04, -5.9865e-04, -2.9902e-04,
        -4.3503e-04, -1.3295e-04,  1.9686e-04, -1.0357e-04,  4.6313e-04,
        -2.1301e-04,  1.3929e-04, -9.6294e-05, -2.0653e-04, -3.4628e-04,
        -3.4336e-01,  1.1915e-03, -6.4345e-01, -1.0574e-03,  4.5773e-04,
        -4.5089e-04, -4.8944e-04, -1.1437e-03,  5.7215e-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([ 3.6262e-03,  8.1713e-05,  2.4906e-04,  2.7584e-04,  1.9745e-04,
        -1.8862e-04, -4.2884e-05,  2.5599e-05, -3.7115e-04, -3.7483e-04,
         7.6118e-05,  1.0259e-04, -4.3392e-04, -1.6489e-06, -1.3903e-04,
        -3.4431e-04,  1.4773e-04,  4.2630e-04,  1.0312e-04, -1.5811e-04,
        -2.6977e-04,  2.5003e-04, -4.5204e-04,  2.2630e-04,  8.5449e-05,
        -1.9386e-01,  2.9778e-04, -4.0601e-01,  2.4707e-04, -3.9033e-01,
         2.4982e-04,  3.4864e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0500e-03, -4.8164e-05, -1.1244e-05,  2.9499e-04,  1.6247e-04,
         1.6001e-04,  1.3017e-04,  2.2882e-04,  9.6946e-05,  6.9945e-04,
        -2.4630e-05, -5.9340e-05, -9.6476e-05, -1.0552e-05,  1.6255e-04,
         3.4308e-05,  2.5874e-04,  1.5692e-05, -1.2961e-04, -8.7113e-05,
         6.4387e-05,  3.4379e-04, -2.4925e-05, -8.8917e-05,  8.3220e-05,
         1.2635e-01, -1.5731e-04,  1.0512e-03,  3.1394e-01, -2.5170e-04,
        -4.7286e-04, -2.7252e-04,  1.1956e-03,  2.2901e-01, -3.6551e-04,
        -2.4326e-04,  3.1997e-01,  1.7924e-04, -7.5796e-05, -1.0225e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1539e-03, -6.4002e-05,  2.0266e-05,  2.5447e-04,  2.8951e-04,
        -5.4341e-04, -1.1587e-04, -3.6757e-04, -2.5883e-04,  4.0998e-06,
        -6.9198e-04,  1.6156e-04,  1.0405e-04,  2.0199e-04, -1.8765e-04,
         5.1101e-04, -2.3998e-04,  1.2793e-06, -1.8442e-05,  1.1636e-04,
         2.6522e-04,  2.8191e-05,  6.5202e-04, -1.9843e-04, -9.0529e-05,
        -5.6773e-01, -1.9035e-04, -2.5603e-01, -5.5333e-04, -8.1237e-02,
        -8.5516e-02, -4.0485e-04, -7.9187e-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([-1.6096e-03,  6.4000e-04, -6.7205e-06, -3.2452e-04,  1.5420e-04,
         1.8405e-05, -4.6783e-04,  3.3641e-04,  2.0245e-04,  3.7457e-05,
        -2.1762e-04,  4.7802e-04,  1.5747e-05, -8.5582e-05, -3.4669e-04,
        -3.7337e-05, -2.2331e-05, -4.4146e-04,  5.1495e-04, -2.5354e-04,
         2.4816e-04, -7.8418e-05, -8.2680e-04, -2.0496e-04,  4.5617e-05,
         2.6444e-02,  2.8235e-04,  9.6124e-01, -1.8995e-03,  1.2389e-03,
         1.1996e-03,  8.4760e-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([-6.6333e-04,  2.4178e-04, -8.5800e-05, -1.4788e-04, -1.4625e-04,
         1.7429e-04,  6.6228e-05,  2.2854e-04,  6.2823e-05,  4.1838e-04,
         4.8803e-04,  5.8032e-05, -4.1295e-06, -6.9978e-05,  4.1572e-04,
         3.2022e-04, -2.0960e-05, -7.5976e-05,  2.8311e-04, -9.4580e-06,
         1.0556e-04, -7.2977e-06,  2.1947e-04, -1.7281e-04, -2.0112e-05,
         4.0933e-01, -2.8948e-04,  1.5117e-01,  1.0511e-03,  5.2100e-04,
         4.3116e-01,  1.1485e-03, -3.2194e-04,  1.3506e-04,  3.5834e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0240e-04, -2.6837e-04,  5.9614e-05, -7.8973e-05,  2.1353e-04,
        -3.7274e-05, -3.5386e-04, -8.2939e-05, -3.9801e-04, -8.9163e-04,
        -6.0950e-04, -4.1419e-04, -5.9096e-05, -2.1274e-05, -3.0598e-05,
        -5.0760e-04, -1.5924e-04, -5.5215e-04, -3.3315e-05, -1.4220e-04,
        -2.4027e-04, -5.5722e-04, -2.6017e-04,  7.3591e-05,  1.5911e-04,
        -5.7636e-01,  2.0660e-04, -1.5664e-03, -8.2068e-05, -4.1397e-01,
        -1.0298e-03,  2.3163e-04, -4.6541e-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([-4.3056e-03,  3.5982e-04,  2.2214e-04,  2.8620e-04,  5.4377e-06,
        -4.9010e-04,  1.5306e-04, -4.3072e-05, -6.8899e-04,  5.5905e-04,
         7.3199e-04,  4.4839e-05, -2.8237e-04, -5.3336e-05,  6.2334e-04,
         8.8064e-04,  7.1801e-04, -1.4465e-04,  3.8546e-04, -2.8917e-05,
        -2.0245e-06, -4.2636e-04, -2.0614e-04, -1.6207e-04,  1.1052e-04,
         9.8536e-01, -5.1406e-05, -7.3641e-05, -5.4999e-04,  1.1043e-03,
         9.4481e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8772e-03, -4.1907e-04, -5.8807e-05, -1.1413e-05, -5.1646e-04,
         3.0101e-05,  3.1272e-04,  3.5304e-04,  5.3329e-04,  4.6754e-04,
         6.4285e-04,  2.8453e-04, -1.4770e-04,  1.3061e-04,  5.3474e-05,
        -1.0056e-04, -3.1983e-04, -4.3398e-04, -3.1525e-04,  5.4364e-05,
        -4.0875e-05,  1.6011e-05, -3.5201e-04, -4.9509e-04, -1.7595e-04,
         7.7537e-01,  9.4163e-04,  2.9907e-03,  2.4372e-04, -2.8760e-04,
         3.8251e-04,  1.5550e-03,  8.4664e-03,  1.9922e-01,  8.6505e-05,
        -8.4911e-04, -4.6099e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4301e-03, -4.2980e-04,  1.3890e-04, -4.3729e-04, -5.0486e-04,
        -2.4186e-04, -1.9874e-04, -1.6951e-04, -1.9551e-04, -8.0164e-04,
        -4.4606e-04,  7.0512e-05, -5.6006e-04, -2.5013e-04, -4.6096e-04,
        -5.6428e-04,  6.9538e-06, -2.0191e-04, -5.7814e-04,  4.1025e-05,
        -1.4630e-04,  2.6663e-04, -4.2507e-04, -1.6340e-04, -1.5362e-04,
        -7.7998e-01,  1.2498e-04, -4.2968e-03,  3.7834e-06, -3.9307e-04,
        -1.5528e-03, -4.9294e-04, -2.0223e-01, -1.2980e-03,  4.6337e-04,
        -2.8177e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 1.3195e-03, -4.1497e-04,  2.0132e-04,  6.6164e-04,  2.5411e-05,
         5.9758e-04, -3.3867e-04, -2.0806e-04, -5.8051e-04,  5.5436e-04,
         5.7182e-04,  6.0331e-04, -5.5109e-04,  6.1979e-04, -1.1187e-03,
        -4.2078e-04,  1.7087e-04, -1.8345e-03,  9.8711e-01,  1.4562e-03,
         9.6224e-05,  5.4731e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3198e-03,  2.8569e-04, -1.0819e-04,  3.1819e-04,  2.1758e-04,
         1.3465e-04, -4.2383e-04, -5.8851e-05,  6.8177e-05,  3.6146e-04,
         4.0443e-05, -3.2905e-04, -3.0247e-04,  5.6155e-04, -2.6960e-04,
        -6.0844e-04, -3.0262e-04,  7.2481e-04, -2.6913e-04, -4.8849e-05,
        -4.1373e-01,  2.1408e-04, -4.3305e-04, -5.7661e-01, -3.2346e-04,
         5.0591e-04,  4.3678e-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([ 9.2950e-03, -1.3061e-03, -3.3857e-05, -6.5337e-04,  2.0029e-04,
        -1.8373e-03, -1.1976e-03, -2.5289e-04, -5.4410e-04, -1.0655e-03,
        -9.5552e-04,  4.1561e-04,  5.1350e-05, -4.7411e-01, -9.7808e-04,
        -9.5258e-04,  2.5237e-03, -5.0026e-01, -2.0743e-03, -8.5545e-04,
        -4.3151e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.1873e-04, -5.2134e-04, -8.0962e-04,  6.2691e-04,  4.5319e-04,
        -3.0642e-04,  3.0940e-04,  2.6365e-04,  1.8188e-04, -3.4527e-04,
        -7.2542e-04,  7.3863e-05, -5.7455e-04, -4.6797e-01, -3.6305e-05,
         2.5300e-04,  1.5030e-03, -5.2271e-01, -2.7965e-04, -1.2079e-03,
        -1.9412e-05, -1.0486e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.7144e-04,  6.6764e-04, -1.6186e-04, -2.2157e-04,  6.0594e-04,
        -6.2873e-04,  4.5005e-05,  1.2954e-04, -2.9894e-04,  4.9954e-04,
        -9.1526e-05,  9.2113e-05,  2.2338e-05,  2.8615e-01, -2.5998e-04,
        -3.0524e-04,  2.8416e-04,  7.0438e-01,  3.2924e-03, -5.7016e-04,
         4.2241e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3671e-03,  7.8127e-04, -3.6980e-04, -3.7714e-04, -3.3862e-04,
         1.1617e-05, -8.2986e-04, -5.9749e-04,  5.0537e-05,  1.3932e-04,
         5.7285e-05, -8.5902e-04,  7.1446e-04,  2.8296e-04,  2.1988e-04,
         9.8442e-01,  4.2653e-04, -2.5228e-04,  4.8556e-04, -7.8560e-04,
         4.3932e-03,  3.0609e-04, -7.3807e-05, -8.5707e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.0568e-03,  1.9094e-05, -1.7569e-04,  1.5962e-04,  2.6220e-04,
        -1.5460e-04, -4.6885e-05, -1.0538e-04,  1.6736e-04,  4.3879e-04,
        -6.2083e-04,  2.0676e-04, -6.5152e-04, -1.1198e-05,  2.6671e-04,
        -4.6908e-01,  3.5778e-04,  3.2642e-04, -2.1274e-05, -4.5355e-03,
        -5.1204e-01, -2.4705e-04,  1.0952e-04,  8.5946e-04, -8.6378e-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([ 2.5802e-04, -2.0717e-04,  1.1614e-04, -2.0955e-04,  7.8554e-04,
        -8.1138e-04, -3.5304e-04, -1.0115e-04,  8.3305e-04,  1.3203e-04,
        -2.2615e-04, -1.1112e-03,  3.9251e-04,  3.5782e-04,  4.9057e-04,
         9.6234e-01, -6.6839e-04,  8.8252e-04,  5.7322e-03, -1.4581e-04,
         2.1370e-02, -2.8789e-04,  2.5231e-04, -9.9285e-04, -9.3932e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7968e-04, -6.7547e-05,  6.9578e-05, -6.5150e-04,  3.2235e-06,
         3.7874e-04, -1.5320e-04,  3.2950e-04,  6.2259e-05,  1.6414e-04,
         2.6384e-04,  4.0602e-04,  4.3996e-04, -3.6107e-04,  8.6688e-05,
         4.2088e-01,  7.6962e-05,  4.7027e-06, -3.8579e-04, -4.8003e-04,
        -6.1400e-04,  5.0876e-03,  4.6128e-04,  3.4990e-03,  6.3659e-04,
         3.1479e-03, -1.2291e-04,  3.2358e-05,  5.5894e-01, -1.7425e-05,
         8.7248e-05, -5.8956e-04, -5.1231e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0965e-02, -6.1728e-04, -1.7257e-04,  2.6811e-04, -6.9312e-05,
        -1.2009e-04,  2.9865e-04,  4.4892e-04,  9.2243e-04, -2.0938e-04,
         1.7243e-04,  6.9248e-04, -4.6002e-05,  9.4099e-04, -2.5183e-04,
        -9.7009e-01, -2.9541e-03, -1.6776e-03, -5.2988e-03, -5.9198e-04,
         2.2212e-03, -9.6849e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5454e-03, -8.6370e-05, -1.6028e-03,  1.1408e-03, -9.2472e-04,
        -7.8668e-04, -7.2397e-04, -5.6665e-04, -5.3235e-04, -1.4666e-04,
        -3.6648e-04, -4.9386e-04,  3.5073e-05,  6.6929e-05, -3.7246e-05,
        -9.8258e-01,  2.0625e-04, -3.1675e-03, -3.1143e-03,  7.1659e-04,
        -1.6275e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7423e-03, -2.1497e-04, -9.2479e-05, -7.4944e-05, -3.5943e-04,
         1.2375e-04,  1.4968e-04,  5.8977e-05,  4.0107e-04,  5.2043e-04,
        -3.2557e-04, -2.4849e-04, -3.3647e-04, -7.3318e-05,  9.9641e-05,
        -2.8803e-04, -2.1678e-01, -4.2196e-04, -4.0211e-01, -1.0708e-03,
        -3.0717e-03, -2.5236e-01,  8.2114e-04,  4.1479e-04, -1.1606e-01,
        -1.2198e-03,  5.2157e-04,  4.1599e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-5.2300e-03,  1.4562e-05, -6.9221e-04, -1.1201e-04, -3.7769e-04,
        -6.5395e-05, -1.8906e-04,  1.2079e-04, -6.3139e-05, -1.2145e-04,
         1.5588e-05, -8.5511e-04, -5.3560e-06,  1.0533e-04, -6.9172e-04,
        -4.9369e-04,  6.4630e-05,  3.7662e-01,  1.8658e-03,  6.0424e-01,
         8.7187e-04,  5.9341e-03,  3.5262e-04, -6.2510e-04,  2.8022e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.8618e-04, -7.7256e-05, -7.5063e-05,  5.0017e-05,  2.4570e-04,
        -5.0891e-05,  2.9522e-04,  1.3441e-05, -1.1407e-04, -1.0653e-04,
        -1.0990e-04,  2.5903e-04,  2.0282e-04, -5.9175e-05,  8.1216e-05,
        -7.5134e-06, -1.7006e-04, -5.9412e-01,  3.6567e-04, -1.8434e-01,
        -9.9994e-05,  1.7743e-04, -3.2389e-03, -4.1981e-04, -2.5593e-03,
        -7.8459e-02, -1.7184e-04, -4.7948e-04,  1.0316e-04,  4.3014e-04,
         9.3477e-05,  2.8968e-04, -1.2638e-01,  1.8575e-04, -3.6408e-03,
        -9.8954e-04,  1.9520e-04,  3.1680e-04, -2.7808e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.6053e-04, -1.0423e-04, -2.1990e-04, -3.7874e-04, -6.3349e-04,
        -3.1930e-04, -1.1680e-04,  2.4611e-04,  6.4549e-04, -2.1622e-04,
        -8.0107e-05, -3.0915e-04, -2.6476e-04, -1.0085e-04,  1.3545e-04,
        -4.0648e-04, -9.3306e-02,  3.1453e-04, -2.9575e-03, -2.3646e-01,
         1.3831e-04,  4.5862e-04, -4.8614e-01,  5.8656e-04, -4.8962e-04,
        -6.9916e-04,  1.1114e-04, -1.7255e-01,  4.7698e-04, -7.2226e-05,
         3.0768e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2644e-03,  9.3911e-05,  9.6400e-05,  1.1989e-04, -3.5924e-04,
        -1.0656e-04,  1.3945e-04,  2.5649e-04,  3.7737e-04,  4.4094e-04,
        -1.6839e-04, -3.8327e-04, -1.4063e-04,  2.9530e-04,  2.6599e-04,
        -6.3975e-01,  1.6430e-05, -6.3004e-02, -1.2780e-03,  1.8454e-04,
        -6.6969e-04, -9.2986e-04, -2.7527e-01,  5.6655e-04, -1.1213e-02,
        -5.5676e-04,  6.0117e-05, -3.6202e-04,  5.4587e-04,  9.3163e-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([-2.7077e-03, -2.0766e-04,  2.0048e-04,  4.2220e-04, -5.3522e-04,
        -5.9603e-04, -1.0142e-04, -2.5669e-04, -1.4006e-04, -3.2042e-04,
        -6.4637e-05, -2.6470e-04, -7.7687e-05, -1.7976e-06,  3.4309e-04,
        -6.4315e-01,  4.4502e-04,  9.3253e-04,  5.1879e-04, -3.3029e-03,
        -3.4352e-01, -7.2587e-04,  5.7675e-04,  5.8730e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2933e-04, -3.6757e-04, -7.0788e-04,  7.3706e-05,  5.2886e-04,
        -1.5878e-04, -2.9997e-04, -1.6973e-05,  4.0752e-04,  2.1679e-04,
        -5.3441e-05, -1.8760e-05, -3.6447e-04,  1.0144e-03,  1.8722e-04,
        -8.2642e-05, -7.9642e-01, -6.6423e-04, -1.9454e-01, -3.9414e-04,
        -1.7096e-03, -1.3796e-04,  8.0726e-04,  2.3780e-04, -3.5938e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5725e-03, -8.6079e-04,  3.2119e-04,  6.4086e-04, -1.7268e-04,
        -7.1508e-04, -1.5578e-04, -9.1919e-04,  2.2327e-04, -6.2623e-04,
         4.8636e-04, -7.7551e-04, -2.5023e-04, -6.2665e-04,  1.4199e-04,
        -5.1936e-05, -1.7043e-01,  9.7951e-05, -8.1791e-01,  2.1385e-04,
         1.2526e-03,  5.6100e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0669e-04, -3.8311e-04,  9.0744e-05,  4.6308e-04,  3.2799e-04,
        -1.8864e-04,  4.2488e-05, -3.8276e-04, -4.5826e-05,  3.7991e-04,
        -3.9833e-04,  1.4315e-04, -2.1428e-04,  5.4179e-04, -3.6575e-05,
         2.5358e-04, -2.8238e-01,  1.6251e-04, -2.1853e-01, -2.9532e-04,
        -4.0875e-02, -4.5244e-01, -2.1118e-04,  2.8676e-04, -3.1763e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.0526e-04, -4.3660e-04, -2.4282e-04, -5.4576e-04,  1.9945e-04,
        -3.7843e-04,  2.5298e-04, -5.5866e-05, -3.7216e-04, -7.5273e-04,
        -7.7957e-04,  3.8947e-04,  5.7663e-05, -2.9024e-04,  7.3618e-05,
        -1.8248e-04,  5.6056e-01, -1.2525e-03,  4.3062e-01, -8.8831e-04,
         5.7007e-04, -4.8426e-04,  8.7027e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.0337e-03,  3.7596e-04,  3.7204e-04, -2.4194e-04, -7.0174e-05,
         8.9157e-05,  1.6135e-04,  1.8099e-04,  2.0953e-05, -1.3684e-04,
         8.7524e-05, -5.4037e-04,  5.0980e-04,  2.5806e-04, -1.5511e-04,
         4.3842e-04, -5.7402e-01,  6.3345e-04, -2.8616e-01, -8.4488e-04,
        -8.7626e-02,  1.2756e-04, -1.2585e-03, -3.9324e-02, -5.3672e-04,
         3.3765e-04,  4.6350e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1121e-03, -3.5523e-04, -7.8877e-04,  1.7072e-05, -1.1387e-04,
        -4.4008e-04,  1.6355e-04, -3.4157e-04, -4.0458e-04, -2.0240e-04,
        -5.0794e-04, -9.6370e-05, -1.3210e-04, -1.5537e-04, -1.3106e-04,
         2.3637e-04,  8.8658e-03, -4.5580e-04,  7.5355e-01,  2.1400e-05,
         2.2666e-01,  2.8587e-04,  2.4247e-05, -1.1605e-03, -6.6108e-04,
         6.6643e-04,  1.7732e-04, -8.0282e-05, -3.7832e-05,  1.4911e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2559e-03, -5.8506e-04, -3.5939e-04, -2.6908e-04, -9.3203e-04,
         4.8241e-04, -1.1697e-04, -2.1665e-04, -1.2004e-03, -5.7794e-04,
        -2.3705e-04,  9.5105e-04, -1.1220e-03,  4.0172e-04, -1.0130e-04,
        -8.1051e-04, -4.0931e-04,  5.3905e-04,  9.8693e-01,  1.7414e-04,
        -1.0761e-04, -6.0241e-04,  6.2263e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  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 #600: [tensor([-4.4749e-03, -6.1474e-04, -2.5217e-04, -1.8495e-04, -6.7017e-05,
        -1.9714e-04,  1.3290e-06, -1.0710e-04, -4.6785e-04, -2.9539e-04,
         4.9488e-05, -4.0389e-04, -1.6607e-04,  1.8434e-05, -3.2547e-01,
        -4.9708e-04, -2.4720e-04,  3.5360e-04, -3.4544e-01, -3.1807e-01,
        -1.2238e-03,  9.0612e-04,  5.0059e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4014e-03, -1.6635e-04, -1.2633e-03,  3.7004e-04, -6.7077e-04,
        -6.1017e-04, -3.6471e-04,  9.8899e-04,  9.2461e-04, -4.4507e-04,
         9.8932e-01,  1.1167e-03,  5.1713e-04,  1.5310e-04,  3.1472e-05,
         5.7440e-04,  9.2060e-04, -1.6042e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6475e-03,  6.7601e-05,  3.9921e-05, -4.6388e-04, -1.3727e-03,
        -1.1350e-04,  1.9744e-05, -1.3348e-04,  3.8275e-04, -4.2280e-04,
         9.9012e-01, -2.1360e-04,  1.9343e-04,  2.4268e-03, -3.8557e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6683e-04,  3.9780e-04,  5.5911e-04, -1.8691e-04,  6.6291e-05,
         2.5000e-04, -5.0045e-04, -5.9980e-04, -3.0063e-04,  3.3213e-04,
        -1.6544e-01,  2.6004e-04, -1.3853e-03,  1.1204e-03,  1.0397e-03,
        -8.2293e-01, -5.1500e-04, -1.3422e-03,  1.6838e-03,  3.4956e-04,
         5.7881e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1325e-03, -3.9540e-04,  2.3749e-04,  3.8465e-04, -3.0590e-04,
         1.2330e-04, -1.1501e-04,  1.2279e-04, -1.5291e-04,  1.6346e-05,
        -3.3587e-04, -4.1456e-01,  8.5668e-04,  4.0768e-04, -1.9321e-02,
        -5.5874e-01,  8.4424e-04, -9.9860e-04, -2.1551e-05,  4.4275e-04,
        -4.0570e-04, -7.2563e-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([-2.5952e-03, -3.8659e-04, -5.2700e-04, -6.4218e-04,  2.3230e-04,
        -1.6532e-04, -4.1298e-04,  6.5215e-04, -2.3096e-04, -7.8482e-05,
         1.6466e-04, -4.3439e-01,  8.0121e-04,  1.0737e-03, -4.4574e-02,
        -5.1206e-01,  5.0692e-05, -4.3197e-04,  5.2487e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8207e-03, -3.5764e-04, -3.9297e-05,  4.7006e-04, -6.7141e-04,
        -7.9718e-04, -3.0075e-04, -1.2842e-03,  2.2858e-04,  7.5902e-04,
        -6.5412e-04, -9.5661e-01,  1.5381e-03, -3.1018e-02,  2.2861e-03,
        -7.6253e-04,  4.0555e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2644e-03,  4.8390e-05, -4.8942e-04, -6.2983e-04,  2.1305e-04,
        -4.5817e-04, -3.2125e-04,  1.4883e-04,  4.8202e-04,  2.4788e-04,
        -2.2249e-04,  4.1120e-06,  1.4665e-04, -1.9815e-04,  1.3324e-04,
        -1.1396e-04,  2.5365e-04,  8.2243e-05,  8.0418e-01, -1.5807e-04,
         1.8190e-01, -4.5212e-04,  4.2426e-03,  1.5948e-04, -7.9517e-04,
        -3.6749e-04,  2.2115e-05, -4.1564e-04, -1.5765e-04,  2.1481e-04,
         3.6776e-04,  1.1382e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0549e-04,  1.8689e-04, -5.2049e-04, -4.0279e-05,  1.3368e-04,
         1.4394e-04,  2.0805e-05, -2.9343e-05, -1.4704e-04, -3.0838e-04,
         5.9621e-04,  1.7429e-04,  3.5374e-04,  1.7210e-04,  3.1905e-06,
        -7.3261e-04, -3.9706e-05,  1.8390e-05,  4.6383e-01, -4.5024e-04,
        -7.3211e-04, -7.9325e-05, -8.8457e-04, -5.5763e-04,  7.8845e-03,
         5.2130e-01, -3.5547e-04,  2.0873e-05, -7.8648e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7980e-03, -5.6184e-05,  3.9731e-06,  2.2454e-04, -4.1205e-05,
         2.1892e-04, -1.7912e-04, -4.7682e-05, -5.9295e-04,  7.6043e-04,
        -1.2349e-04, -1.1097e-04,  1.1280e-04,  5.8493e-05, -4.5443e-04,
        -2.1935e-05,  1.3893e-04,  1.4604e-04, -1.0034e-03, -8.2111e-01,
        -1.6857e-01,  3.8934e-04, -6.9251e-04,  8.8019e-06, -1.3483e-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([ 5.7206e-03, -4.5784e-04, -9.9870e-04, -2.3683e-04,  7.5057e-04,
         4.2201e-04, -4.3281e-04, -1.4066e-04,  4.4946e-04,  2.6140e-04,
        -6.0612e-04,  3.9012e-04,  1.1204e-04,  5.8817e-04, -2.7094e-04,
         7.7516e-02, -1.3329e-03,  9.0793e-01, -4.4455e-04, -3.4713e-04,
         1.8846e-04, -3.9872e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.3043e-04,  4.7138e-05,  2.3066e-04,  4.9348e-04,  1.6922e-04,
        -2.8492e-04,  1.6298e-04,  1.8599e-04, -4.9865e-04, -4.9476e-04,
        -5.7345e-05,  5.6954e-05, -6.8592e-04, -1.2800e-04, -1.5246e-04,
        -4.5671e-01,  1.2418e-04, -5.3717e-01,  2.7331e-05,  1.2671e-04,
        -7.6039e-04,  5.4074e-04,  2.6543e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  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 #650: [tensor([ 1.3423e-03, -2.9263e-04,  1.9948e-04, -1.6186e-04, -1.5490e-04,
        -5.7264e-04,  1.0919e-04, -2.9098e-04,  8.9978e-05, -1.6728e-04,
        -9.3956e-05, -9.2354e-04,  1.4995e-04, -3.6427e-04,  1.6203e-04,
         1.3936e-04, -5.1149e-04, -3.4899e-05, -1.3517e-04, -5.3947e-04,
         4.1424e-04, -1.0558e-04,  3.6614e-04,  1.0837e-04, -5.5365e-04,
         9.9022e-01, -7.1455e-05,  2.3127e-04, -9.2952e-04, -5.6118e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3478e-03,  1.3834e-04, -2.7991e-04, -7.1246e-05,  3.2723e-04,
         1.5066e-04,  3.7216e-04, -1.7313e-05,  4.7321e-04,  7.6993e-05,
        -9.7460e-05, -4.2464e-04,  5.5192e-04, -5.5274e-05,  6.4747e-05,
        -1.4177e-04,  1.8096e-04,  2.0388e-04,  4.9096e-04,  9.9156e-05,
        -4.7843e-04, -1.8978e-04, -7.4518e-05,  4.5705e-05, -2.0361e-04,
         9.9068e-01, -4.2657e-04, -6.0844e-04, -3.4463e-04, -9.0183e-04,
         2.7666e-04,  2.0052e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.2243e-03,  1.5582e-02, -5.8654e-05,  4.4039e-02,  4.2313e-02,
         1.8367e-03, -1.0826e-02,  6.4616e-02,  5.5599e-03,  7.7186e-02,
         1.8998e-02, -1.4137e-02,  7.0807e-02, -1.2834e-02,  9.3352e-02,
         5.6159e-03,  1.9200e-03,  2.3583e-02,  1.1662e-02,  7.0435e-03,
         7.9339e-02, -5.6459e-03, -1.4897e-02, -6.0722e-03,  8.2787e-03,
        -1.7680e-01, -7.9220e-02,  4.7880e-02, -5.3673e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.7931e-04,  1.7906e-05, -1.1506e-04, -1.3307e-05,  2.4541e-06,
        -3.4104e-04, -1.1381e-05, -9.9495e-05, -1.9929e-05, -1.8838e-04,
        -1.6536e-04, -2.6384e-05,  6.3305e-06,  5.7805e-05,  3.3600e-04,
         2.9987e-04, -2.5409e-04, -2.8596e-04, -1.7840e-04, -1.7498e-04,
         1.9413e-04,  1.0025e-04, -9.8195e-05, -2.4524e-04,  1.2267e-04,
        -5.3206e-01, -1.9042e-04, -3.4481e-01, -1.9882e-05,  3.2615e-05,
        -2.7486e-04, -1.8279e-04, -1.1751e-01,  3.5047e-04, -4.6280e-04,
         1.7314e-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([-3.8660e-03,  2.2964e-04,  2.8433e-04,  1.6806e-04,  2.2632e-05,
         1.0384e-04, -4.4423e-05,  1.6248e-05,  2.2739e-05,  7.6209e-06,
         5.7853e-05,  1.3448e-04,  1.2103e-05,  3.3999e-04, -7.2662e-06,
         4.0658e-05, -8.6601e-05, -2.0317e-04, -3.4225e-05,  4.0740e-05,
         1.8351e-04,  2.1353e-04, -2.6200e-04,  1.6606e-05, -9.3010e-05,
        -2.8614e-01,  1.4273e-04, -6.5921e-04, -6.9011e-02,  6.3227e-06,
        -1.6563e-01, -3.7329e-02, -1.6722e-04,  3.6596e-04, -2.2513e-01,
        -2.3098e-04, -2.0672e-01,  3.3439e-04,  7.5040e-04, -3.9788e-04,
        -2.0777e-04, -1.4062e-04, -1.4138e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8122e-03, -5.1976e-04,  4.5468e-04,  8.4557e-04,  4.7497e-04,
        -9.6599e-04, -4.6581e-04, -4.2691e-04, -2.0760e-04,  2.7259e-05,
        -2.5047e-04, -6.7908e-01,  1.3199e-04, -3.1026e-01, -1.4386e-03,
         1.4027e-03, -2.3063e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6987e-03, -3.8260e-06,  6.2587e-04,  8.3330e-04,  3.0915e-04,
         4.4294e-04,  4.1440e-04,  4.3688e-05, -4.2256e-05,  6.7749e-05,
         5.2108e-05, -1.0939e-01, -1.0921e-04, -8.8521e-01,  1.2793e-04,
        -2.4159e-04, -3.8332e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3207e-04,  1.2684e-04,  4.7167e-04, -1.3503e-04,  1.5540e-04,
         9.5809e-04,  2.2925e-04, -7.2309e-04, -2.5241e-04, -4.3573e-04,
        -4.0040e-04, -1.5454e-01,  9.1590e-04, -8.3089e-01, -3.6216e-04,
        -7.8013e-03,  3.1135e-04,  1.0608e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.5823e-05, -8.6365e-04,  3.4453e-04, -2.0527e-04,  6.2134e-04,
        -1.7600e-03, -3.8542e-04, -1.3275e-03,  1.6587e-04,  9.9074e-01,
         1.0275e-03,  1.8814e-04,  1.1814e-03, -1.1254e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3740e-04, -3.6237e-05,  1.4405e-04,  2.3644e-04, -8.6981e-04,
         3.4079e-04,  3.7299e-04, -1.1126e-06, -8.9350e-01,  9.5606e-04,
        -2.1992e-02,  3.1952e-05, -1.0391e-02, -6.9280e-02,  1.1555e-04,
        -4.9001e-04, -7.0259e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2339e-03, -2.2569e-04,  1.0692e-03,  1.7357e-04,  1.4705e-04,
        -2.7977e-05,  2.7907e-04, -1.3153e-04,  1.8551e-03,  9.5172e-01,
        -5.8014e-04,  9.8033e-03,  2.8831e-02,  1.3278e-03, -5.6563e-04,
         3.9630e-04, -6.3764e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8846e-04,  7.0002e-04, -5.2305e-04, -1.0010e-03,  4.4432e-04,
         4.4390e-05, -3.6321e-04, -4.4932e-04, -2.1067e-03, -5.5491e-01,
        -2.6898e-03, -4.3215e-01, -3.0812e-03, -9.0326e-06, -1.1347e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  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([ 1.7242e-03,  8.0137e-05,  7.2058e-05,  1.2303e-04, -1.0524e-04,
        -9.7453e-05,  2.6442e-04, -7.8346e-05, -2.9578e-04,  1.1520e-04,
         4.7653e-04,  8.8481e-05, -1.6316e-05, -6.3321e-01,  1.1389e-04,
        -2.7767e-02, -1.6017e-04, -3.3342e-01,  1.1726e-03,  2.1862e-04,
        -4.0240e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2019e-03,  1.0172e-04,  7.0957e-04, -1.8752e-04, -1.8418e-04,
         5.7597e-04,  1.6435e-04,  1.6893e-04, -3.7803e-04, -3.0576e-04,
         3.3874e-04,  3.7526e-04,  4.0447e-01,  1.7309e-04,  2.2459e-02,
         3.2965e-04,  3.2485e-03,  1.1300e-03,  1.6588e-03,  8.3139e-05,
        -2.6140e-04,  5.5839e-01, -1.1618e-03, -2.0779e-04,  4.4009e-04,
         5.6014e-06,  2.8989e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2204e-03,  1.9617e-04, -1.3944e-04,  1.3035e-04, -2.0303e-04,
        -5.2132e-04, -8.5058e-05, -3.4965e-04,  1.6240e-04, -7.6588e-04,
        -2.6399e-04,  8.9555e-05, -4.8197e-01, -9.5170e-05, -5.1171e-01,
        -1.9327e-04,  3.2421e-04,  8.8576e-05,  4.7026e-04,  2.1881e-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([ 1.1542e-03, -1.0460e-04,  2.7701e-04, -1.9405e-04, -8.4033e-04,
        -2.0244e-04,  7.7858e-04, -5.3694e-04,  9.6101e-04, -4.2338e-04,
        -8.4871e-04,  1.8585e-04, -9.8185e-01,  1.0755e-03,  5.2270e-04,
        -8.3582e-03, -3.1496e-04, -1.8081e-05,  7.6953e-04,  5.8553e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3147e-03, -2.9756e-05,  7.3692e-06, -2.8647e-04, -1.0738e-04,
        -2.1322e-04, -1.4402e-04,  2.6623e-04, -4.0843e-04,  1.6574e-04,
         2.7370e-04,  2.8008e-04,  2.1255e-04,  2.5739e-04,  2.4045e-04,
         1.4555e-04, -3.5206e-04,  4.1504e-05, -9.3406e-01, -2.8763e-06,
        -5.8364e-02, -8.1389e-05,  1.5460e-03, -8.4580e-05, -5.5981e-04,
         9.7086e-05,  2.1443e-04,  2.4486e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7677e-03,  1.1923e-04, -2.1808e-04, -2.8171e-04,  2.2789e-04,
        -2.4552e-04, -1.0555e-04,  1.2920e-04, -6.0774e-05,  2.7299e-05,
        -3.9459e-06,  2.3604e-04, -2.7177e-04, -4.1653e-04,  2.7125e-05,
         1.6032e-04,  3.2292e-04, -5.4687e-05,  7.4018e-01, -9.8607e-05,
         5.8457e-02,  5.0928e-04,  1.9489e-01, -4.0631e-04,  5.5816e-04,
        -2.2971e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.8599e-04,  2.5077e-04,  2.1306e-05, -1.6545e-04, -3.0448e-04,
         8.8755e-05, -1.6346e-04,  1.0418e-04, -1.8513e-04, -2.0419e-04,
        -1.6062e-04,  3.7268e-04,  1.4484e-04,  2.2200e-05, -4.2773e-04,
         3.4242e-05,  2.4518e-05,  8.7162e-05,  4.8049e-04, -3.9560e-04,
        -7.9765e-04,  1.1699e-02, -5.6004e-04,  8.4582e-01,  2.4259e-04,
         8.1364e-02,  5.4644e-02, -1.1516e-04,  3.0771e-05,  2.0551e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9713e-02,  1.7596e-03, -8.0875e-04,  3.0534e-03, -4.4457e-04,
         5.6243e-04, -9.9351e-04, -2.6575e-03,  2.5680e-03,  1.6379e-03,
         6.1122e-05,  4.5178e-04,  7.3634e-04,  3.2103e-03,  5.6501e-03,
        -7.4928e-04, -1.5085e-04, -1.9249e-04, -7.1058e-03, -8.3960e-03,
        -1.3990e-03,  1.6047e-03,  5.9294e-02,  8.1320e-01,  9.9955e-03,
         3.9732e-02, -2.4694e-03,  1.4052e-03,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9555e-04,  3.7697e-04, -2.7100e-04, -1.5374e-05, -9.2735e-05,
        -4.7247e-05, -4.3089e-04, -5.1247e-06,  1.3659e-04, -2.1194e-04,
        -2.0328e-04, -1.6765e-04, -2.0663e-04, -2.5929e-05, -4.2170e-04,
        -2.3003e-05,  8.3034e-05, -8.6194e-05,  3.4174e-04, -3.2291e-04,
         6.2191e-01,  6.6545e-04,  1.2768e-01,  2.4436e-01, -7.8638e-04,
         6.2378e-04, -1.1653e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4666e-04, -1.2136e-04, -1.4780e-04,  3.1440e-04,  3.9463e-04,
        -1.5198e-04,  3.6891e-05, -2.3267e-04, -1.3735e-04,  1.9712e-04,
        -2.4546e-05, -3.1855e-04, -1.3037e-04, -1.0679e-04, -1.4590e-04,
         1.7803e-07,  1.1619e-04,  1.1104e-04,  2.3033e-04,  1.8742e-04,
        -3.3764e-01, -1.3730e-04, -6.2044e-02, -5.9573e-01, -3.5240e-04,
         1.4478e-04,  4.9869e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.8308e-04, -1.2498e-04, -1.6459e-04, -8.1213e-05, -5.4383e-04,
        -9.9196e-05, -2.7862e-04, -1.0740e-04,  2.7511e-04,  1.9099e-04,
         2.2229e-04, -1.5566e-04, -7.8721e-06, -6.3052e-05,  2.5741e-04,
         3.0371e-04, -9.7224e-05, -2.6362e-04, -2.3373e-01, -1.4358e-04,
        -2.9822e-02, -2.3563e-01, -4.9570e-01, -6.3146e-04, -4.4001e-06,
        -5.2379e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1672e-03,  2.4972e-05,  1.6436e-05, -1.7524e-04,  4.1627e-05,
        -3.9822e-04,  1.8890e-04,  4.9929e-05, -1.4868e-04,  5.6056e-05,
         1.4543e-04, -2.0729e-04,  2.0447e-04,  2.5408e-04,  1.4264e-04,
         3.8473e-04, -1.9877e-04,  4.4765e-04, -1.9682e-01,  3.8234e-05,
        -3.6671e-01,  6.2002e-04, -4.2900e-01, -2.5200e-04,  6.4669e-04,
         6.6069e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #750: [tensor([ 2.4332e-03,  9.0070e-04,  1.9177e-04,  1.0535e-04,  9.8894e-04,
         1.3757e-03,  7.5623e-04, -9.7171e-04, -9.3631e-02,  3.3646e-04,
        -2.9304e-03,  1.1934e-03, -8.8232e-01, -9.4161e-03, -9.0963e-04,
         1.1870e-03,  3.5021e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4758e-03, -3.2724e-05, -1.5636e-04, -9.7642e-04, -7.5809e-04,
        -6.1466e-05,  7.9903e-05,  5.0649e-03, -1.7264e-04,  2.5428e-01,
         1.8205e-01,  5.5308e-01, -1.0217e-03,  1.2482e-04, -5.0839e-04,
        -1.4941e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8581e-03, -1.0777e-03, -3.2401e-05, -1.0581e-03, -1.6046e-03,
        -3.7691e-04,  2.6934e-05,  1.6001e-03,  6.1356e-04, -3.6730e-04,
         9.8318e-01,  2.3836e-03,  1.2152e-03, -2.9238e-03,  6.8241e-04,
         6.0210e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1289e-03, -4.7254e-06, -2.6829e-05, -1.5372e-04, -2.2799e-04,
        -3.8235e-05,  2.3282e-06,  3.9714e-05,  2.5999e-04, -1.2090e-04,
        -6.7460e-05,  1.4784e-04, -8.2254e-05,  1.5396e-04,  6.3664e-05,
        -1.6786e-04,  9.2785e-05, -7.9519e-05,  1.8415e-04, -1.1567e-04,
         2.8308e-04,  3.9184e-04, -2.0127e-04,  2.0932e-04, -3.0684e-01,
         2.7379e-04, -5.8649e-01, -1.8180e-04, -2.2592e-05, -9.9835e-02,
         3.7461e-04, -4.2068e-04, -3.2143e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8341e-03,  1.5629e-04, -2.2974e-05,  1.2970e-04,  1.8012e-04,
         3.3472e-04, -2.5219e-04,  7.9073e-05, -2.3415e-04,  4.0748e-04,
         3.5658e-04,  4.0295e-04,  2.1628e-04,  2.9031e-04,  3.1669e-04,
         6.8720e-04, -5.2416e-05,  2.6745e-04, -2.9276e-04,  3.1552e-04,
        -9.9969e-05,  3.1322e-04,  2.9051e-04,  1.4263e-04,  9.5540e-01,
         4.3155e-04, -3.4617e-02,  4.8080e-04, -9.2874e-04,  4.6293e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8713e-03, -3.4989e-04, -6.4892e-04, -8.5490e-04, -6.1288e-04,
         1.2689e-05, -3.4635e-04, -2.2571e-04, -6.4661e-04, -4.8962e-04,
        -3.5761e-04, -1.5565e-04, -8.2638e-04, -2.1147e-04, -4.6787e-04,
        -1.1492e-04, -3.4691e-04, -3.1339e-04, -3.2098e-04, -5.7715e-04,
        -4.2219e-04, -2.6223e-04, -3.6135e-04, -7.6813e-04,  3.5834e-02,
         2.7460e-04,  9.4946e-01, -7.4451e-04,  5.7272e-04, -5.5289e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.3361e-04,  3.6967e-05,  3.6936e-04,  1.5887e-04,  1.3940e-04,
         2.2623e-04,  6.3713e-05, -2.9311e-05,  2.1786e-04, -4.8583e-06,
        -7.4181e-05, -8.6791e-06, -1.4583e-05, -2.0134e-04,  1.8922e-04,
         1.0666e-04, -3.8925e-05, -3.0752e-05,  1.1688e-04,  4.3846e-04,
        -3.6628e-04, -4.1617e-05, -8.5889e-05, -5.9654e-01,  6.1648e-04,
        -1.6305e-03, -5.0962e-02, -1.2182e-04, -2.2337e-02, -1.1440e-04,
         2.9933e-04, -1.7931e-03,  1.1714e-04,  2.1063e-05,  3.1449e-04,
        -2.9497e-04, -1.4176e-01, -6.4537e-05,  8.5626e-04, -1.7786e-01,
        -3.9658e-04,  3.0711e-05,  1.8117e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.5587e-04,  8.6501e-05,  1.3650e-04, -9.3017e-06,  1.6550e-04,
         1.9516e-04, -3.3998e-05,  4.5776e-04, -2.5119e-04,  2.4334e-04,
         2.8641e-04, -2.4535e-04,  6.1733e-05,  4.1930e-04,  1.4005e-04,
         3.2834e-04,  2.6082e-04, -1.5325e-04,  5.7560e-04,  1.9924e-05,
         3.3448e-05,  2.5496e-05,  1.0213e-04, -5.7966e-01, -6.4506e-05,
         9.0812e-06, -3.0123e-02,  3.0240e-04,  2.2000e-05, -3.8435e-01,
         1.3758e-04,  1.0564e-04, -2.3359e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6167e-03,  2.5205e-05,  1.0431e-04, -2.3389e-04, -1.5225e-04,
        -4.1277e-05, -8.2933e-05,  9.4372e-05, -1.0806e-04, -3.6624e-05,
         1.0071e-06, -1.9678e-04, -8.3925e-05, -4.4369e-04,  1.8925e-04,
        -2.4663e-05,  1.4893e-04, -2.5376e-04, -1.3259e-04, -1.6718e-04,
         1.3610e-04, -2.0532e-04,  1.9825e-04,  4.0046e-01, -7.0380e-05,
         2.9507e-03,  4.7177e-01, -7.8765e-05,  7.1231e-02,  3.9319e-02,
        -7.1657e-05,  7.9723e-06,  4.5664e-06,  4.0079e-03,  2.2627e-04,
        -1.3446e-04,  3.1556e-04,  5.9165e-06, -4.3404e-03,  1.6301e-04,
         2.5794e-05,  1.4210e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3376e-04, -8.6141e-05,  3.6333e-04,  2.1561e-04, -2.1962e-04,
        -4.5258e-05, -3.6467e-04,  7.1368e-04, -3.4553e-04,  2.5142e-04,
         3.5329e-04,  2.8604e-05,  3.3789e-04,  3.0335e-04,  3.8467e-04,
        -4.3754e-01,  7.3335e-05, -1.7101e-03, -3.1651e-04,  2.4007e-04,
        -5.4209e-01,  4.9710e-04, -7.4048e-03, -4.5319e-03, -8.1880e-04,
        -7.7100e-05,  3.4983e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5869e-05,  4.0738e-05, -4.6660e-04,  6.2179e-05, -3.6077e-04,
         6.2929e-04,  9.3083e-05,  6.8045e-05,  1.2789e-04,  7.9821e-04,
        -7.9826e-04,  1.2645e-04,  5.4682e-04,  3.3132e-04,  3.9243e-04,
        -5.1373e-01,  8.8861e-04, -4.7909e-01, -1.9083e-04, -4.4647e-04,
        -7.7675e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5722e-04,  9.6885e-05,  3.3266e-04,  1.7881e-04, -1.4192e-04,
        -1.6000e-04, -4.5313e-04, -1.9273e-04,  7.4139e-05,  4.2319e-04,
        -1.7131e-04,  2.9938e-04, -1.8107e-04,  3.5762e-04,  3.6794e-04,
        -7.0753e-01,  2.3442e-04, -9.0941e-02, -6.9696e-04, -1.9402e-01,
        -5.3479e-04, -6.1867e-04,  4.0169e-05,  3.7967e-04, -1.0012e-04,
        -4.0589e-04,  3.7330e-04,  5.2999e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  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 #800: [tensor([ 0.0328,  0.0337, -0.0016,  0.0009,  0.0094,  0.0282, -0.0084, -0.0064,
         0.0439,  0.0411,  0.0123,  0.0167,  0.0226, -0.0082,  0.0040,  0.0229,
        -0.0112,  0.0095,  0.0072,  0.0082, -0.0109,  0.0125,  0.3813,  0.2286,
        -0.0165, -0.0126,  0.0084,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1562e-03, -1.2163e-04,  4.1708e-04, -2.6863e-04,  1.7594e-05,
        -1.6341e-04,  7.1114e-05, -1.4344e-04,  3.4763e-04,  1.9756e-04,
         1.6256e-04, -4.1338e-04,  2.5397e-04, -1.4359e-04,  1.3003e-04,
         2.3621e-04,  1.6509e-04, -3.2423e-04, -5.8139e-05,  5.6501e-05,
         9.1247e-05, -6.7514e-06, -1.2779e-04, -1.9438e-02, -2.7174e-01,
         1.7170e-04, -1.1142e-05, -7.0901e-04,  1.7712e-04, -2.9876e-03,
        -6.9691e-01,  2.5841e-04, -7.6687e-05, -4.4555e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.4204e-04,  6.1554e-04, -2.2872e-05,  2.0023e-04,  6.1265e-05,
         7.9294e-04,  4.6780e-04,  1.6166e-04, -9.7321e-05,  5.1743e-04,
         8.2952e-05,  1.8821e-04,  8.2646e-05,  6.5049e-05, -3.1398e-04,
         1.6849e-04, -1.6233e-04,  4.7603e-04,  5.0697e-05,  4.2733e-04,
        -7.8105e-05, -1.9942e-04,  9.6262e-01,  3.0328e-04,  2.9188e-02,
        -6.8944e-04,  1.8447e-04,  8.3940e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1448e-03,  1.1201e-04, -3.8700e-04,  2.2105e-05,  2.7457e-04,
        -4.6796e-06, -7.0729e-05, -3.2906e-04, -1.8362e-04, -4.7039e-04,
        -1.0578e-04, -2.5738e-04,  3.3341e-04, -4.4829e-04, -1.1413e-04,
        -1.8668e-04, -1.7606e-04,  1.8689e-04, -1.6374e-04, -8.6933e-04,
         1.9919e-04, -2.5098e-04,  1.4115e-02, -4.7890e-04, -2.0134e-04,
         3.7796e-04,  9.7115e-01,  6.4464e-04,  4.0043e-04,  4.4560e-03,
        -2.0900e-04, -5.0732e-04, -1.0756e-04, -5.8443e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.4583e-04, -1.3021e-04,  2.8992e-04,  6.4722e-05, -7.7013e-06,
         3.1039e-05,  2.7645e-04, -6.5056e-05,  1.6812e-05, -1.5138e-04,
         7.5178e-05,  4.1988e-05, -5.7846e-05,  9.1282e-05, -3.5015e-05,
        -1.8831e-04,  1.6175e-04, -9.7420e-05, -9.9548e-05,  9.8384e-05,
         5.9967e-05,  6.4708e-05,  8.9177e-03, -4.9145e-04, -3.9055e-04,
        -8.1171e-05,  6.6883e-01, -6.0728e-04,  6.1700e-03,  6.4183e-05,
         1.4278e-04, -2.8775e-04,  3.1107e-01, -7.7952e-05,  2.8589e-04,
         1.3127e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6694e-03, -4.2279e-04,  4.4455e-04, -4.5521e-04,  2.8978e-04,
         5.3646e-04,  3.1192e-04, -4.5168e-04, -4.4770e-04, -9.3759e-04,
        -4.7960e-04, -6.1427e-04,  7.8256e-04,  3.8735e-04, -5.2043e-04,
        -6.1832e-04, -3.9748e-04,  2.3176e-04,  6.5452e-05,  8.3823e-04,
         6.6410e-04, -2.6794e-05,  9.8543e-01, -6.3149e-04,  1.2639e-04,
         1.8187e-03, -4.0151e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8739e-04, -1.3477e-04,  5.6702e-05,  1.7894e-04, -1.2651e-04,
         2.4908e-05,  3.3447e-04, -1.2935e-05, -3.4767e-04, -2.4333e-04,
         6.2050e-07, -9.0917e-05,  2.4683e-04, -7.3348e-05, -1.0520e-05,
         1.1506e-04,  1.6679e-04, -1.1482e-04,  2.2970e-04, -2.5131e-04,
         1.9782e-04, -1.9322e-04,  6.4294e-01, -4.3221e-04, -4.3050e-05,
         5.8139e-03,  2.2875e-01,  7.2747e-05,  1.2053e-03, -3.2546e-04,
         4.0912e-04,  2.8406e-04,  1.1554e-01,  2.0450e-04,  1.7945e-04,
         5.4421e-05, -2.9992e-04,  4.2634e-06], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4460e-03, -2.0990e-04, -1.3258e-05,  2.7498e-04, -1.9458e-04,
        -2.1918e-04,  4.1006e-05,  1.8414e-04, -4.8368e-04, -2.8233e-04,
        -1.7341e-04, -2.9501e-05, -1.2827e-04, -2.9434e-04, -3.4269e-04,
         5.0002e-04,  3.7165e-05, -1.4733e-04,  7.5366e-05,  2.5296e-05,
         3.2324e-04, -2.3238e-05,  7.4579e-01, -2.0386e-04, -3.8082e-07,
        -1.2943e-04,  2.4659e-01,  1.6383e-04,  1.0444e-03,  2.3818e-04,
         1.5617e-06,  2.9675e-04,  8.9677e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1652e-03,  8.3922e-05,  4.3513e-04,  3.5834e-04, -8.7335e-05,
        -5.0825e-05,  4.6206e-05,  2.6243e-04,  3.5113e-04, -2.0181e-04,
         2.0218e-04, -1.8701e-04, -8.0376e-05, -6.6665e-05,  7.6276e-05,
        -1.8509e-04, -1.5922e-05,  2.0583e-04, -1.6794e-04, -4.2122e-04,
         2.1035e-04, -5.9314e-05, -2.6878e-01,  7.7707e-05,  5.3328e-04,
        -1.7388e-02, -2.1659e-04,  1.3573e-04, -7.0703e-01,  7.6920e-04,
         1.3714e-04, -1.0591e-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([ 1.3997e-03,  5.6038e-05,  3.4615e-05,  2.1134e-04, -1.3296e-04,
        -1.8738e-05, -2.7122e-04,  9.0945e-05,  6.9634e-05,  2.6869e-04,
        -1.7839e-04,  1.1808e-04, -1.3476e-04, -1.8549e-04,  8.5298e-05,
        -4.6434e-05,  1.8913e-05,  5.6086e-05,  1.5668e-04, -4.8449e-04,
        -6.2354e-05,  2.7647e-04, -5.1246e-01,  2.0218e-04,  3.6473e-04,
         6.9088e-04, -3.7380e-01, -7.4396e-05,  1.6989e-04, -1.0706e-02,
        -9.6359e-02,  2.5295e-04,  2.9958e-04,  2.6038e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3030e-03, -4.6401e-04,  9.7600e-05,  1.1264e-04, -4.4690e-04,
        -7.0986e-04, -5.8531e-04, -1.1298e-04,  2.3528e-04,  3.8592e-04,
        -3.7812e-05, -1.8563e-04, -1.3130e-04, -1.1765e-04, -5.2727e-05,
        -3.2793e-04, -2.2346e-04, -7.0042e-04, -6.4439e-05, -3.8727e-04,
        -2.2542e-04,  4.2306e-04, -9.5247e-01,  5.4004e-04,  3.9537e-04,
        -2.5839e-03,  3.3190e-02,  4.9981e-04, -6.6835e-04, -3.2559e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5900e-03,  5.7359e-05, -2.7536e-04,  2.4015e-04,  1.6991e-04,
        -5.3984e-05, -2.2923e-05,  3.0755e-04,  1.7899e-05, -2.3065e-04,
        -1.6379e-04, -1.9677e-04,  1.5039e-04,  1.4428e-04, -3.1422e-05,
         1.9413e-04, -1.2950e-04,  2.7095e-04, -4.5886e-04, -1.5504e-04,
        -4.6864e-05, -9.0119e-05,  3.5300e-05,  2.4131e-01,  1.4142e-01,
        -6.3807e-05, -3.8649e-04, -1.7049e-04,  1.7330e-03,  6.0823e-01,
         1.5008e-04, -7.3875e-04,  4.3902e-04,  2.3898e-04, -2.4279e-05,
         5.9388e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #850: [tensor([ 2.0526e-03,  3.2199e-04,  1.0181e-03, -7.5398e-04,  4.3705e-04,
         2.6312e-04, -6.1752e-04,  8.8918e-04, -7.7996e-04,  2.0837e-04,
         1.8696e-04,  9.4298e-04,  7.5308e-04,  6.6296e-04, -5.6062e-04,
        -9.8732e-01, -7.9920e-04, -2.7574e-04,  8.4846e-04, -3.0908e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2070e-03, -1.5533e-03, -7.2943e-05, -3.5433e-04, -7.5785e-04,
         3.3758e-04, -2.0072e-03,  3.8052e-04, -1.0546e-04, -3.3852e-04,
        -4.9871e-04, -9.3503e-04,  6.7220e-05,  3.6920e-04,  2.1266e-04,
        -9.8306e-01, -1.2513e-03,  2.5895e-04, -1.4367e-03,  1.2998e-03,
         4.9551e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9104e-03,  4.5947e-04, -1.1463e-03, -1.0229e-04,  8.3947e-04,
         3.7189e-04, -9.3164e-04,  1.3447e-05, -5.9935e-06,  2.0127e-04,
         4.0428e-04,  6.9112e-04, -2.1687e-04,  4.3812e-04,  7.8177e-04,
         8.7277e-04,  9.8934e-01,  6.3044e-04,  6.4678e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0060e-03,  1.4661e-04, -3.1582e-05, -4.4272e-05, -2.9338e-05,
        -4.7593e-04,  7.7435e-04, -3.5820e-04, -2.6341e-04, -7.5571e-04,
         1.6343e-04, -8.6558e-04,  1.7103e-04, -5.2167e-04, -9.7275e-01,
        -1.3671e-05,  1.6396e-02,  6.0105e-04,  4.1257e-04, -9.2305e-05,
         6.8593e-04, -1.9533e-04,  1.3076e-03,  7.1596e-04, -2.2232e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3818e-03,  2.0413e-04, -2.3818e-04, -6.9416e-04, -5.0825e-05,
         4.1711e-04, -6.5634e-04,  8.5631e-05, -2.0919e-04, -6.0209e-04,
        -4.2105e-04,  2.2805e-04, -5.7375e-04, -5.0555e-04,  9.4255e-01,
         2.1231e-05,  4.8799e-02,  3.0083e-04,  1.2696e-03, -2.0653e-04,
        -5.8066e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1595e-03,  2.0052e-04,  2.6222e-04,  1.6075e-04, -2.0118e-04,
         1.8830e-04,  4.5809e-05,  2.6079e-04,  8.8639e-05, -1.8150e-04,
         1.8702e-04,  6.3965e-05,  3.6996e-05,  3.3767e-04, -2.2419e-04,
         1.3040e-04,  6.6090e-05, -2.2311e-05, -1.3122e-04, -1.2334e-04,
        -1.1649e-04,  7.9863e-05, -4.1592e-01, -8.7188e-05, -5.9987e-04,
         1.0399e-04, -5.7507e-01, -2.7497e-05,  3.0283e-04,  3.3800e-04,
         7.2471e-04, -4.2716e-04, -2.7807e-04, -1.8101e-04, -6.7357e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.4543e-02, -2.0378e-03,  3.0128e-03,  2.5744e-04, -4.7010e-03,
         5.4716e-04, -6.0568e-03,  7.9567e-04,  7.5659e-03,  4.6858e-03,
        -1.5134e-03, -5.1685e-03,  1.5367e-04, -5.9490e-03,  3.9398e-03,
         1.2906e-02, -1.9187e-03,  3.9795e-03, -4.7549e-03,  4.9307e-03,
         1.9945e-03, -6.0349e-03, -1.0219e-01, -3.4449e-03, -4.4121e-02,
         1.4237e-01, -6.2201e-03,  1.6397e-01, -1.8400e-01, -9.1742e-02,
        -9.4500e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3473e-03,  3.9156e-04,  3.2288e-04,  1.4999e-04,  4.6155e-05,
         6.7870e-04, -1.9167e-04,  2.8837e-04,  1.8619e-04,  1.0291e-04,
         2.9549e-05,  2.9922e-04,  2.1266e-04, -2.3743e-04,  1.1279e-05,
         2.8804e-04,  4.9341e-04,  4.0476e-04,  6.3122e-05,  3.6700e-04,
         2.6723e-04, -3.0257e-04, -2.1981e-03,  2.3321e-04, -1.4785e-04,
         9.0986e-04,  1.4663e-04, -9.8872e-01,  6.7123e-05,  3.9542e-04,
        -5.0401e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0436e-02,  4.0226e-03,  8.2439e-04,  3.0946e-04,  5.0221e-03,
        -3.6955e-04,  3.9363e-03,  5.1064e-03,  9.2374e-04,  9.9981e-03,
        -7.0783e-04,  1.7343e-03,  1.5947e-03, -4.5160e-03,  2.1794e-03,
        -1.7877e-03,  7.2619e-03, -4.4033e-04,  1.1913e-02,  5.5677e-03,
         4.0948e-03,  5.3090e-03,  2.3827e-03, -2.8751e-03,  1.6954e-02,
         2.6284e-03,  9.2231e-03,  4.7206e-04,  4.5100e-03,  1.1732e-04,
        -3.0255e-03, -2.6185e-04, -2.9286e-04,  1.1277e-01, -7.8130e-02,
         5.6145e-02,  6.1887e-01, -3.2856e-03,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1873e-02, -3.4867e-04, -3.2580e-03,  1.8506e-03,  1.0259e-03,
        -2.2250e-03,  2.0755e-03,  2.8277e-03,  2.1973e-03,  6.7358e-03,
        -4.4551e-06,  1.9768e-03, -2.1333e-03, -2.7959e-03,  8.9951e-04,
        -5.7664e-04,  1.9424e-03, -8.1741e-04,  2.1265e-03,  3.9102e-03,
         1.9032e-03,  3.1098e-03,  1.6788e-03,  2.3861e-03,  6.0991e-03,
         8.1465e-06,  4.4429e-03, -2.9224e-04,  1.6660e-03,  9.1342e-04,
        -4.0997e-03, -1.0743e-03, -8.8650e-04,  3.8863e-01, -6.2248e-02,
         2.5691e-01,  7.9292e-02,  1.0839e-01, -6.8902e-03, -7.4804e-03],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2839e-02,  1.4929e-02, -3.4109e-03,  5.3121e-03,  1.6427e-02,
        -8.2562e-04,  2.2227e-03,  1.0554e-02,  4.7656e-03,  1.9631e-02,
        -1.0466e-03,  3.6648e-03,  2.7288e-03, -3.2570e-03,  5.5317e-03,
         7.2230e-03,  1.7609e-02, -1.7079e-03,  1.4282e-02,  5.2692e-03,
         9.4693e-03,  1.4569e-02,  5.1429e-03,  1.6554e-03,  3.1019e-02,
        -3.2563e-03,  1.2251e-02, -1.2429e-04,  3.7773e-03,  1.7686e-04,
         9.7545e-04,  2.7998e-03,  3.5566e-03, -8.8514e-02, -2.1442e-01,
         2.0645e-01,  2.1601e-01, -1.2588e-02,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7964e-03,  6.7822e-05,  3.8686e-06,  4.3626e-04,  1.4849e-04,
         7.2180e-04,  2.3439e-04,  2.6006e-04,  6.8752e-04,  4.3469e-04,
        -5.7643e-01, -1.3056e-03, -4.1351e-01,  8.6231e-04,  1.0098e-03,
         9.6300e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #900: [tensor([-2.0343e-03, -1.9320e-04,  3.2322e-04,  5.3150e-04,  9.3573e-04,
        -1.6562e-05, -3.8310e-04,  6.3368e-04,  5.9534e-01,  6.9039e-04,
         8.9176e-04,  1.2840e-05,  3.9629e-01, -8.1578e-04,  9.0619e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7040e-03,  5.7090e-04, -5.6928e-05, -1.1261e-03, -5.8485e-04,
         9.6584e-05,  2.3271e-04, -1.5930e-04, -7.9925e-04,  1.9057e-04,
        -4.4026e-04, -8.2312e-04,  3.2756e-04, -1.0248e-03, -1.0474e-03,
        -5.7017e-04, -1.1534e-04,  9.6390e-01,  2.2179e-04,  2.4535e-02,
         1.0010e-03, -2.3935e-04, -2.3324e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7542e-03,  8.5789e-05,  1.1869e-04, -2.0148e-04,  2.3235e-04,
         4.7581e-05, -1.9082e-04,  2.5910e-04,  5.2297e-04,  1.7432e-05,
        -6.9002e-05, -5.8357e-04,  1.3548e-04,  5.5281e-05, -1.9101e-04,
        -1.1672e-04,  1.1915e-04,  4.8741e-01, -2.2401e-04,  2.5697e-01,
         1.2758e-04,  2.4793e-01, -9.7302e-04, -2.6931e-04,  1.7009e-04,
         2.2540e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.5492e-05,  1.8511e-04, -1.5127e-04, -2.1977e-04,  4.5340e-05,
         4.3123e-05,  7.2213e-05,  2.2721e-04,  1.1432e-04,  3.9187e-05,
         1.0158e-04,  2.4237e-04, -2.8247e-06,  1.2376e-05,  7.1531e-05,
         1.4239e-04, -8.9332e-06,  2.6406e-01, -3.5833e-04,  5.2799e-01,
        -2.2431e-04,  1.2802e-01,  1.3713e-04,  7.5989e-02, -4.1302e-04,
        -3.3766e-04,  1.4750e-05,  4.2352e-04, -6.0399e-05,  4.1467e-05,
        -1.6325e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6680e-04, -1.2297e-04, -1.8597e-04, -6.2006e-04, -3.5008e-04,
        -2.7176e-04, -5.7347e-04,  4.7470e-04,  9.4497e-04, -1.9846e-04,
        -2.8710e-04,  1.3316e-03, -1.2358e-04,  3.0708e-04, -3.8678e-04,
        -1.3100e-03,  2.5685e-04,  2.8212e-02,  9.6161e-01,  1.3745e-03,
         5.0209e-04, -1.8554e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.9620e-04, -3.6308e-04,  1.1860e-05,  5.3834e-04,  6.6298e-04,
         3.1715e-04, -4.6510e-04, -2.8321e-04,  4.1222e-04,  1.9189e-04,
        -1.2763e-04, -4.1433e-04,  3.6318e-04,  8.1421e-04, -4.5737e-01,
         2.9054e-04,  3.0853e-02, -1.0046e-03, -1.2252e-03,  2.5390e-04,
         9.4144e-03, -4.9373e-01,  6.0735e-05,  2.3554e-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([-1.9518e-03, -1.0435e-04,  1.1815e-04, -2.5436e-04, -2.4274e-04,
         7.1410e-05,  1.7121e-04,  3.4947e-04,  2.8263e-04,  7.4555e-05,
        -5.4242e-04,  1.5561e-04,  2.8451e-05,  3.0220e-04,  1.9531e-01,
         6.7457e-01, -1.3198e-04, -4.9730e-04, -3.2729e-05,  1.6207e-04,
        -5.4256e-04, -5.4413e-04, -3.6476e-04, -6.0073e-04,  1.2092e-01,
        -3.7430e-04, -4.6690e-04, -6.0393e-05,  4.7635e-04,  1.3747e-04,
        -1.5378e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0365e-03, -2.4610e-04, -4.5376e-04, -2.3693e-05,  3.8206e-04,
         7.0606e-04, -7.2836e-04, -4.8154e-04, -2.4558e-04, -2.1805e-04,
        -1.8670e-04, -1.5046e-03, -2.8438e-04, -3.7857e-04,  8.6245e-01,
        -3.1010e-05,  1.2968e-01, -8.8008e-04, -8.2861e-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([-4.8801e-04,  4.6759e-04,  2.4531e-04, -8.1624e-04, -4.8921e-04,
         4.5645e-04,  1.0282e-04,  3.8477e-04, -3.0574e-04, -7.8321e-04,
        -6.2542e-04, -8.1086e-05,  4.5646e-04, -8.8852e-05,  8.9902e-01,
         3.7423e-04,  9.4207e-02,  5.0407e-04, -1.0723e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5520e-03, -4.2585e-04,  1.4743e-04,  7.1593e-04,  4.3212e-04,
        -2.2204e-04,  7.5302e-05, -2.7124e-04,  3.4139e-04, -1.0033e-04,
         5.5174e-05,  7.4036e-04, -2.9863e-04, -1.4335e-04, -5.4454e-01,
         3.9677e-06, -4.4851e-01, -3.5226e-04,  7.1599e-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([-3.7214e-04,  4.8571e-04, -2.8969e-04,  1.0021e-05,  2.9047e-04,
         4.1206e-05,  2.8357e-04, -2.0275e-04, -1.5268e-04,  4.1294e-04,
        -4.7264e-04, -5.0409e-04,  2.6559e-04, -9.6145e-05,  2.2546e-04,
        -1.3190e-04, -2.6729e-04, -2.1630e-04, -7.5420e-04, -2.1025e-04,
         8.1265e-01, -1.9592e-04,  1.7964e-01, -5.8591e-06,  6.8677e-04,
        -9.1157e-04, -2.2777e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9898e-03, -3.1891e-06, -6.7155e-04,  3.7400e-05,  2.1118e-04,
        -5.1777e-04, -2.3455e-05, -4.3556e-04, -3.3983e-04, -3.3670e-04,
         4.3055e-04, -1.7309e-04, -5.6226e-04, -1.1581e-04, -7.4337e-04,
        -5.1831e-04, -3.6080e-05, -5.8616e-04,  2.4344e-04,  2.4119e-04,
         4.2128e-04,  8.0592e-01,  1.7081e-01, -4.1982e-04,  1.2321e-02,
         2.8901e-04, -4.2477e-04,  5.7012e-05,  5.8445e-05,  5.5296e-05,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #950: [tensor([-1.5437e-03, -2.6349e-04,  3.3124e-05,  1.6686e-04,  6.2393e-05,
        -1.9625e-04, -2.3581e-04, -3.7459e-04, -1.6276e-05, -1.2582e-04,
        -9.1346e-05, -4.0694e-04,  2.3288e-04, -1.4814e-04,  1.8708e-04,
         1.0762e-04, -4.6530e-04,  9.4812e-05, -9.1038e-05, -2.3327e-04,
         6.5050e-05, -2.1522e-04, -4.8907e-04, -1.2267e-03,  1.2221e-01,
         2.1414e-04,  6.8201e-05,  1.7621e-03,  1.5579e-01, -3.3177e-04,
         1.7177e-02, -8.1370e-05,  6.9474e-01, -2.0760e-04, -3.4144e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5915e-04,  1.6187e-04,  1.3917e-04,  7.0597e-05, -1.9883e-04,
        -4.5776e-05,  6.5018e-05, -1.1736e-04, -9.4902e-05, -4.5560e-05,
         3.0279e-04, -7.6413e-05, -1.1390e-04, -2.5498e-04, -1.7800e-04,
         9.6081e-05, -1.9051e-04, -4.0674e-05,  5.0981e-05,  6.5044e-05,
        -2.2875e-05,  3.7111e-04, -9.1884e-05, -1.1603e-04,  1.6730e-04,
         9.9374e-02,  3.7338e-04,  5.7216e-01,  1.8991e-04,  7.2859e-03,
         7.1620e-05,  1.0311e-02, -7.2163e-05,  1.7480e-02, -3.4953e-04,
        -5.6451e-04,  2.8770e-01,  1.3903e-04,  4.0352e-04,  4.2485e-05,
        -1.4301e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9682e-03,  2.6514e-04,  1.7069e-04, -7.8269e-05,  5.0556e-05,
        -6.3268e-06,  3.9286e-04, -3.7510e-04,  2.0391e-04,  3.3283e-04,
         4.1420e-04,  3.0577e-04,  1.2378e-04,  1.8510e-04,  2.7462e-04,
         1.3435e-04, -3.2176e-04, -5.3348e-01,  7.3298e-04, -1.7157e-01,
        -5.4306e-04, -2.8638e-01, -1.5942e-04, -1.4414e-03, -8.7922e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1401e-03,  9.4464e-05, -2.2606e-04,  1.1731e-04,  4.1116e-04,
        -4.0522e-05, -1.7829e-04,  2.2645e-04,  1.2061e-04, -3.3078e-04,
         4.9274e-04, -6.7102e-05,  1.8106e-04,  6.5513e-04, -2.7074e-04,
         2.8802e-04, -1.5120e-04, -6.9104e-01,  6.5855e-04, -3.0010e-01,
        -5.3752e-04, -2.2136e-03, -4.5743e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.7526e-03, -4.1544e-03, -2.4989e-03, -1.8362e-03, -2.7836e-03,
        -3.3088e-03,  8.8732e-03,  1.6888e-03,  1.1406e-03, -2.1338e-03,
        -6.7581e-04,  1.5441e-03,  1.1864e-03,  1.2521e-03,  1.4662e-03,
         1.4119e-04,  6.5729e-04,  1.8416e-02,  1.0430e-02,  8.3463e-01,
        -5.8656e-02,  3.6778e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8017e-04, -2.9490e-04, -2.1375e-04, -6.5669e-05, -7.7654e-05,
        -8.7676e-05, -1.8589e-04,  3.7836e-05,  5.8693e-04, -9.3404e-05,
         1.4105e-04,  4.1815e-04, -1.4818e-04,  1.5502e-04,  2.4047e-05,
        -1.7326e-04,  2.4315e-04,  2.9974e-04,  2.9168e-04,  1.4184e-04,
        -2.0101e-04, -5.1243e-05,  1.7975e-04, -4.5512e-05, -1.5435e-04,
        -1.0003e-04,  3.3977e-04,  5.8601e-05,  1.9656e-04,  1.7153e-04,
         3.3632e-04, -1.4278e-04, -8.2874e-05, -1.7234e-05,  5.1685e-04,
         3.1453e-04,  2.7222e-04,  2.2717e-04,  1.1925e-04, -7.6136e-05,
         2.7869e-04,  1.0795e-04,  7.3535e-05, -2.8752e-04, -1.1982e-04,
         7.2951e-04,  9.9049e-01, -2.9842e-04, -5.3268e-05,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1175e-05,  2.8229e-04, -1.2158e-05,  3.5963e-04, -1.3455e-04,
         7.6599e-06, -1.1648e-04, -2.5255e-04, -2.1572e-04,  1.1140e-04,
        -9.7964e-05, -3.9428e-04,  1.1791e-04, -5.1491e-05, -5.3245e-05,
         2.3854e-04, -5.1964e-04, -8.1023e-05, -1.1770e-04, -1.7636e-04,
        -3.0920e-04, -4.1385e-06, -1.9657e-05, -5.5045e-04, -2.8386e-04,
         7.3013e-04,  5.2277e-04, -5.6039e-04,  1.2958e-04, -6.8072e-04,
        -6.3695e-04,  4.7276e-06,  5.1135e-05, -1.9822e-04, -6.3786e-04,
        -3.3029e-04,  3.4997e-05, -3.9884e-04, -6.0482e-05, -6.8097e-06,
        -4.9220e-04, -1.3270e-04,  4.5438e-04,  9.4181e-06,  1.5846e-04,
        -9.4456e-04, -9.8675e-01, -1.1882e-03, -3.2580e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0374e-03, -6.9602e-05, -5.3842e-04, -2.4776e-05, -1.9803e-04,
        -3.4255e-04, -2.7188e-04,  1.7604e-04, -2.2997e-05, -1.9164e-04,
        -3.4850e-04,  1.6429e-04, -1.1310e-04, -1.6159e-04,  2.7318e-04,
        -2.5725e-04,  3.0127e-06,  5.5429e-04, -1.5343e-04,  2.0099e-04,
         3.9335e-04, -6.8622e-05,  3.7973e-04, -1.7779e-04,  2.1866e-04,
        -6.5383e-04, -1.0481e-04,  4.2310e-04,  2.8552e-05,  2.1112e-04,
         4.1332e-04,  8.4395e-05,  1.2388e-04,  2.7226e-04,  4.4437e-04,
         2.9395e-04, -2.8443e-04,  2.7616e-05,  1.7883e-04, -8.2351e-06,
         2.2099e-04, -7.0323e-05, -3.8723e-04,  4.0777e-05, -1.4408e-04,
        -1.9987e-04,  6.7483e-04,  9.8779e-01,  2.9104e-04,  2.8950e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4707e-03, -1.3280e-04,  2.1100e-04,  9.6906e-05, -2.4534e-04,
        -5.7208e-04,  1.3997e-04,  4.0657e-04,  3.4381e-04, -2.6426e-04,
         8.2022e-05,  4.7486e-04,  4.1515e-04, -1.4674e-04,  1.6180e-04,
         3.5551e-01, -5.4288e-04,  1.5044e-01, -1.2996e-04,  1.7926e-04,
         6.2003e-04,  4.8707e-01,  2.8201e-05,  8.5697e-05,  2.2497e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0859e-03, -2.2734e-04,  2.1508e-05,  1.1430e-04,  1.3959e-05,
        -3.8591e-04, -3.3642e-04,  1.9755e-04, -5.8736e-04, -1.5056e-04,
        -1.1824e-04,  1.6323e-04,  2.6125e-04,  6.9964e-05,  1.3291e-05,
         2.7565e-01, -2.2647e-04,  2.3771e-01,  6.9483e-04,  7.6004e-02,
         7.6100e-04,  4.0370e-01, -1.3805e-04,  3.1486e-04, -5.4555e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1934e-04,  9.9118e-06, -1.9588e-05,  5.7085e-05,  2.0213e-04,
        -1.7081e-05,  2.3778e-04, -5.7840e-05,  8.3203e-05, -2.7664e-04,
         3.5086e-04, -8.6675e-05, -2.3846e-04, -2.4038e-05,  2.6363e-04,
         4.5450e-01, -5.9187e-05,  3.3583e-04,  3.0523e-04,  6.5448e-02,
        -1.1762e-04, -3.5056e-05, -7.0984e-03, -4.2557e-04,  2.7364e-03,
         4.6500e-01,  3.0321e-04,  4.1589e-04,  1.6952e-04,  2.1075e-04,
        -2.5293e-05, -4.3230e-05,  2.2919e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.6377e-03,  3.5712e-03, -3.6648e-03,  1.2540e-02,  7.0562e-03,
        -9.2154e-04,  7.9038e-04, -2.0108e-04,  1.2940e-02,  1.6936e-03,
         4.2244e-02, -1.0245e-02,  2.4370e-03,  4.8206e-03,  2.5446e-03,
         3.8078e-03, -3.1202e-03, -1.3880e-02,  1.0061e-02,  5.9771e-01,
        -4.9017e-02, -7.4628e-02, -1.3647e-01,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1000: [tensor([ 4.4615e-03, -1.8448e-03, -4.9295e-04,  4.4819e-04, -1.6123e-03,
         1.0784e-03, -6.1061e-04, -1.3616e-03,  3.1376e-04, -4.8931e-04,
        -1.4765e-03,  4.1663e-04,  4.5862e-04, -1.8497e-03, -4.2509e-04,
         3.0680e-04,  8.1434e-04, -5.5450e-01,  5.3617e-04, -3.1454e-01,
         2.5317e-03, -1.0322e-01, -3.1640e-03,  1.0676e-03,  1.9853e-03],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5515e-03, -6.4578e-04, -4.3291e-04, -5.7001e-04, -1.5685e-05,
         5.2150e-05,  1.1264e-04,  1.1292e-05, -1.0686e-04, -1.5321e-04,
        -3.2268e-04,  1.3633e-04, -6.3799e-04, -5.2981e-04, -2.0376e-04,
         5.6042e-05, -1.1899e-04, -4.5648e-04,  9.9283e-01,  5.3105e-04,
         5.2326e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2384e-03, -4.4572e-05, -8.6913e-04,  4.6291e-05, -2.3147e-04,
         3.0637e-04,  3.4748e-04,  5.0512e-04, -6.1212e-05, -8.6567e-05,
         2.3099e-04,  1.7165e-04, -2.8246e-04, -3.8300e-05,  5.6093e-04,
         4.3384e-05,  3.8851e-04, -9.9000e-01, -1.1646e-03,  9.2177e-05,
         3.1561e-03,  1.3366e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6590e-03,  8.5053e-04, -3.8563e-05, -1.9126e-04, -2.9556e-04,
         1.0902e-03,  2.0552e-04,  4.7729e-06,  3.3535e-04,  2.7206e-04,
        -9.5784e-01,  5.4847e-04,  3.4885e-02, -1.1468e-03, -6.3379e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6203e-04, -1.0143e-04, -1.1794e-04,  4.2616e-04, -4.9397e-04,
         2.1477e-03,  3.2258e-04, -1.5237e-04,  9.0543e-04, -1.0119e-03,
         9.7992e-01, -5.9917e-04,  1.1921e-02, -3.5322e-05, -1.3869e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6036e-04, -9.0761e-05, -2.8834e-04, -7.7761e-04, -7.0000e-04,
        -3.8278e-04,  3.3696e-04,  2.2199e-04, -5.4014e-04, -2.2252e-04,
        -9.3330e-01, -2.0468e-04, -6.1774e-02,  3.1181e-04,  4.9242e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.2898e-04, -6.5675e-04, -3.3784e-04,  2.8771e-04, -1.0254e-04,
         1.3787e-04,  1.6457e-04, -8.2170e-05, -2.3327e-05, -4.5580e-05,
        -1.5264e-04,  3.3229e-05, -6.3588e-05,  8.8789e-01,  9.5439e-04,
         1.0787e-01,  6.8698e-05,  2.3934e-04, -2.5805e-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([-7.9933e-04, -6.1062e-04, -5.6012e-05, -2.3197e-04,  7.5883e-04,
        -2.7719e-04, -7.5156e-05, -5.6795e-07, -2.4980e-04, -7.4999e-04,
         6.3598e-05,  4.6248e-04, -1.4719e-04, -9.9042e-01, -6.8283e-05,
         2.9128e-04,  1.2452e-03, -2.0808e-04,  1.0214e-04, -1.8095e-04,
         1.3520e-03,  5.6197e-04,  1.0884e-03,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0212e-04,  4.8633e-04, -1.8303e-04, -2.3585e-04, -5.0798e-04,
        -1.2161e-04,  2.6553e-04,  2.0429e-04, -1.3577e-04,  4.1941e-04,
        -6.3147e-04, -1.5319e-04,  1.2742e-04,  7.9368e-01, -2.8910e-04,
         4.6612e-02,  8.2226e-04,  1.5404e-01, -3.2935e-05, -4.6980e-04,
        -8.2236e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3637e-04,  9.1604e-04, -1.7115e-04, -9.0815e-05,  4.3090e-04,
         1.1973e-04,  4.1933e-04, -1.2070e-04, -3.1539e-04,  2.3725e-04,
        -3.9919e-04, -1.8664e-04,  4.0597e-04,  1.6692e-04,  1.8262e-04,
         1.9488e-04,  2.6168e-04,  4.0464e-05,  6.7669e-01, -2.0276e-04,
         3.1774e-01, -1.9880e-04,  5.8340e-05,  3.1318e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8394e-03,  1.2365e-04,  4.1829e-04,  4.7414e-05,  3.3661e-04,
        -2.1607e-05, -5.5524e-04, -7.6464e-06, -1.0770e-04,  1.7957e-04,
        -7.2317e-04, -6.7877e-04,  6.5593e-04, -4.9186e-05, -3.4552e-04,
         5.1641e-04, -5.1341e-04, -6.3092e-04,  9.4919e-01, -8.0443e-04,
        -4.0393e-02, -6.7474e-04, -1.0466e-03, -1.4050e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4877e-04,  4.9363e-04,  2.7883e-04,  2.0770e-05,  5.2530e-04,
        -1.9199e-04, -3.2345e-04, -1.2075e-05, -2.0049e-04, -9.2993e-05,
        -1.5112e-04, -1.0431e-04,  1.3103e-04,  4.3223e-04, -2.0520e-05,
         2.7872e-04,  3.4861e-04, -2.7699e-04,  8.9934e-01,  5.5330e-05,
         9.4943e-02, -3.3177e-04, -1.0494e-03, -2.4564e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1050: [tensor([-2.1781e-03, -6.6872e-04, -2.8707e-04, -1.2192e-04,  2.5180e-04,
        -6.6234e-05, -4.1925e-05, -1.0094e-04, -1.5630e-04, -6.0298e-04,
        -2.3657e-04, -2.5337e-04, -6.3730e-04,  4.2293e-06, -3.0614e-04,
        -3.0773e-05, -2.9729e-04,  2.1536e-01, -7.7949e-04,  5.7747e-01,
         4.6085e-05,  1.9846e-01, -1.2264e-03, -4.1668e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.4102e-04, -3.9956e-05, -4.4902e-04, -3.0563e-04, -1.2114e-04,
         1.1441e-04, -2.0136e-04, -8.3635e-05,  1.5193e-04, -1.2178e-04,
         8.6347e-05, -2.5025e-04, -2.2119e-04,  1.2637e-05, -3.3412e-04,
        -3.5144e-04, -1.5242e-04,  4.7274e-01, -1.5345e-04,  7.7340e-04,
        -8.7903e-05,  5.1547e-01, -5.4219e-05, -2.2822e-04,  5.0642e-03,
         4.5800e-04, -2.3369e-04, -4.9012e-04, -4.0812e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2669e-03,  5.7347e-05, -4.8241e-04, -1.1511e-04, -3.8175e-04,
        -8.2224e-06,  5.4822e-04, -3.2048e-04, -3.7088e-05,  6.1882e-05,
        -9.8915e-06, -4.5683e-04, -6.6866e-06,  5.6238e-04, -2.0205e-04,
        -2.2479e-04, -3.6886e-05,  4.1694e-01, -2.2459e-04, -3.6134e-04,
        -5.9834e-04,  5.5525e-01,  2.0765e-02, -2.4706e-04, -5.0816e-04,
         3.2173e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.7626e-04, -6.4365e-04,  1.1056e-03,  1.1540e-04, -3.7751e-05,
         1.0610e-03, -3.2595e-04,  1.8732e-04,  4.1605e-04, -2.7724e-04,
        -3.5404e-05,  6.0359e-04, -1.5104e-04, -2.7463e-04, -2.4273e-03,
        -1.6546e-06, -4.6427e-04,  9.9025e-01, -6.6799e-04,  2.7835e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1869e-03, -6.4437e-05,  2.6036e-04, -9.2848e-04, -4.4058e-04,
        -2.9729e-04,  1.1749e-05, -1.6792e-04,  3.7539e-04, -7.8526e-04,
         3.7408e-04, -4.8711e-04,  7.5913e-05, -5.3965e-04,  4.4593e-04,
        -1.0892e-03, -5.9240e-02, -9.3308e-01,  6.3108e-05, -9.0147e-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([-2.6397e-03,  5.6942e-05, -3.0941e-03,  3.6826e-03,  2.3061e-03,
         2.8882e-03, -1.1780e-03,  2.3142e-03, -3.9760e-03, -1.2579e-03,
        -3.7964e-03, -7.4740e-04,  1.6777e-03,  8.0818e-04,  7.4704e-02,
         6.8243e-02,  2.5381e-02,  5.3182e-01,  2.2026e-01,  4.4849e-02,
         4.3194e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3362e-04,  2.7248e-04, -4.4274e-05,  1.8151e-04, -6.3447e-04,
         8.0332e-05,  9.6570e-05,  1.4498e-04, -1.1876e-04, -1.4191e-04,
         1.3204e-04, -3.3709e-04,  2.3419e-04, -2.2522e-04,  2.9733e-01,
         5.7080e-04, -5.4220e-04,  5.4005e-04,  1.5607e-04, -1.9107e-04,
         6.9732e-01,  2.5514e-04, -1.9950e-05,  9.3768e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.6943e-04,  1.2024e-04,  2.3884e-04,  3.2989e-04,  1.2672e-04,
         8.2973e-04,  2.6349e-04, -1.9534e-04, -4.5197e-05,  6.2043e-05,
        -4.6062e-04, -1.8374e-04, -1.4367e-04, -2.5993e-04,  9.9378e-01,
         2.2003e-04, -1.9494e-05, -8.3737e-04, -5.6309e-04,  6.4917e-05,
         5.8964e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5182e-03,  4.0388e-05,  2.1836e-04,  5.3345e-05,  4.8209e-05,
         2.2858e-04,  4.1640e-06, -9.1610e-05, -1.1354e-04, -5.9517e-05,
        -9.7924e-05, -2.1484e-04, -2.2404e-04,  5.8986e-05,  5.0947e-01,
        -3.5875e-05, -1.0979e-04,  4.8612e-01,  2.7762e-04, -1.9052e-04,
         6.7925e-06,  2.7551e-06, -4.2765e-04, -2.4313e-05, -3.6323e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3647e-04,  1.3772e-04, -4.0332e-04,  6.2514e-04,  3.1652e-04,
         4.5920e-04,  1.7340e-04,  1.1090e-03,  3.8546e-04, -6.1588e-04,
         2.7592e-04,  2.7750e-04,  5.7757e-04,  3.6605e-04, -2.0433e-04,
         8.9560e-06,  9.9152e-01, -5.6771e-05,  5.8601e-05,  6.3812e-04,
         2.9113e-04,  3.8005e-04,  5.7965e-05, -4.2167e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8489e-03,  4.4207e-04,  1.2457e-04,  2.3599e-04,  3.1018e-04,
         8.4012e-05,  4.3839e-04, -7.1507e-04,  1.4213e-04,  3.9527e-04,
        -1.1105e-04, -5.9265e-04,  2.9562e-04, -1.0158e-04, -3.4049e-04,
         2.8866e-04,  9.8465e-01,  3.7496e-04, -5.3231e-03, -2.6802e-04,
         2.4520e-04,  1.8466e-04, -4.4307e-04, -9.3292e-04,  1.1146e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5703e-03,  3.3208e-04, -2.7144e-06, -1.2198e-04,  1.5794e-04,
         4.6054e-04,  5.5645e-04,  5.0435e-04,  7.7660e-05,  4.1333e-04,
         2.4899e-04, -2.8745e-04, -2.6447e-04,  4.9996e-04,  3.3248e-05,
        -2.3795e-04,  9.8795e-01, -1.4761e-04, -2.8048e-03, -4.2128e-04,
        -2.4719e-04,  1.0489e-04, -6.9891e-05, -1.7196e-03, -1.8863e-04,
        -4.1447e-04, -1.6010e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1100: [tensor([-3.9602e-04, -1.0897e-04, -2.4538e-06,  4.2711e-04, -2.4538e-04,
        -2.4462e-04,  7.7993e-04,  1.2046e-04,  6.8917e-04, -2.2815e-05,
        -1.8623e-04, -2.8919e-04, -2.3263e-04,  1.3830e-04, -1.7618e-04,
        -5.5477e-04,  1.7817e-04, -3.1930e-04,  9.9184e-01,  2.2817e-04,
        -4.9256e-04, -1.0767e-03, -3.8697e-04, -8.1428e-04, -5.0447e-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([-1.9063e-03, -2.5283e-04, -3.1165e-04,  2.1462e-05,  1.6572e-04,
        -1.6019e-04, -1.0607e-04,  3.4363e-04,  4.7891e-04,  6.8315e-04,
         5.0911e-04,  6.8890e-04, -1.5670e-04,  1.1546e-05,  2.9073e-04,
         1.2126e-04, -3.9847e-01,  3.8710e-04, -7.9974e-04, -9.7086e-05,
        -5.9036e-01,  9.8662e-04, -7.1870e-06,  1.7458e-03, -1.9787e-04,
         4.7030e-04,  2.7069e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3079e-02,  4.1892e-04, -2.2853e-04, -1.5376e-03, -3.7290e-04,
        -2.9552e-03, -1.6204e-03, -9.2216e-04, -1.3354e-03,  1.0103e-03,
         6.7806e-04,  2.9580e-03,  2.7220e-04,  3.0698e-03,  3.4220e-04,
         1.2676e-03, -4.1509e-03,  8.5000e-01,  4.6236e-03, -4.5812e-03,
        -2.8048e-03, -3.3157e-03,  8.2141e-02,  8.5620e-04, -1.8719e-03,
         1.8394e-03,  1.6242e-04, -1.5799e-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([ 2.1072e-04,  1.2334e-04,  3.4347e-04,  4.3672e-04, -7.7474e-05,
         1.2819e-04,  9.0308e-05, -9.9581e-05, -7.4907e-05, -9.9693e-05,
        -2.4900e-04, -2.8981e-04, -3.6353e-05, -2.8391e-04, -9.8756e-07,
         1.1304e-04,  8.1217e-01, -9.2083e-05,  1.4115e-01, -4.9473e-03,
         3.6340e-02, -1.8595e-04,  1.8177e-04, -1.6205e-03, -5.2035e-04,
        -4.4201e-06,  1.3315e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8111e-04,  5.6903e-04,  2.7387e-04,  3.2413e-04, -3.0792e-04,
        -2.5155e-05,  4.1825e-04,  9.8542e-04, -6.9549e-05, -4.4857e-04,
         3.2817e-04,  3.6587e-04, -8.6970e-05, -3.0756e-04,  7.2517e-04,
         8.3851e-05, -1.8902e-04,  3.6694e-04,  9.9020e-01, -3.1000e-04,
         4.1624e-04,  1.1187e-03, -1.9656e-04,  8.5887e-04,  5.4214e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9529e-04,  3.9375e-04,  9.0607e-05, -2.2874e-04,  2.5752e-04,
        -8.5417e-05,  1.4317e-04, -3.1406e-04,  6.2871e-05,  1.0796e-04,
        -4.5946e-05, -1.4567e-04, -4.0436e-06, -4.1460e-04, -1.6631e-05,
        -3.3815e-04, -6.6346e-05, -2.8290e-05,  8.4046e-01, -1.7199e-04,
         2.1911e-03,  3.5156e-02, -3.8211e-04,  1.1715e-01, -2.3307e-04,
         1.2195e-04,  1.2031e-04,  2.7345e-04,  1.0087e-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([-3.2694e-04, -1.6356e-04, -3.3183e-05,  6.9140e-05,  2.5390e-04,
        -1.4521e-04, -4.7020e-04,  1.1561e-04, -5.6171e-05,  2.3993e-04,
        -4.0060e-05,  8.7878e-05,  9.6724e-04, -9.9078e-05,  7.6447e-05,
        -1.2507e-04, -4.2789e-03,  1.9508e-04,  9.8765e-01, -9.4947e-04,
         3.4402e-03, -5.1315e-05,  1.4309e-04, -2.4092e-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([ 7.1062e-04, -6.1295e-04, -1.7307e-04, -2.8910e-04,  7.3784e-05,
         5.3925e-05, -4.1483e-05,  3.5890e-04,  2.6293e-04, -3.0097e-04,
         3.3936e-05,  7.2938e-05, -1.2915e-04, -7.1541e-07,  1.1527e-05,
        -1.9393e-04,  4.0950e-01, -1.2432e-03, -7.3342e-05,  5.8501e-01,
         4.3735e-04,  2.3856e-04, -1.7704e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3042e-03,  9.5782e-05, -6.5426e-04,  2.1877e-04,  4.1471e-05,
         1.8108e-04,  4.3751e-04,  1.0103e-04, -3.4651e-04,  4.5710e-04,
         6.7147e-05, -1.0501e-04, -3.6387e-04, -4.4681e-05, -2.9839e-04,
        -2.7479e-05,  8.0392e-01, -4.7790e-04,  1.8909e-01, -8.6929e-05,
        -2.5220e-04,  6.2503e-04, -9.4597e-05, -2.3459e-04,  4.7599e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4020e-03,  1.3396e-03, -2.4300e-04, -8.0703e-05,  5.4568e-04,
         9.7135e-06, -1.4020e-04, -4.4732e-05,  2.3388e-04, -2.9479e-04,
        -8.1088e-04,  8.5954e-04,  4.1208e-04,  7.4906e-04,  1.3952e-04,
         6.5587e-05,  8.6847e-01,  7.3073e-04,  1.1994e-01, -3.6957e-04,
         1.1976e-03,  6.6725e-04,  2.5483e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.9237e-04,  2.3309e-05,  5.8211e-04,  1.0217e-03,  7.9151e-04,
         4.1638e-04,  7.5231e-04,  4.6349e-04,  8.3003e-05, -1.5722e-04,
        -6.0813e-04,  9.4480e-04,  4.3972e-04,  7.5449e-04,  3.0553e-04,
        -1.2296e-04,  4.8770e-04, -8.6628e-04,  9.8695e-01, -3.0440e-03,
         1.0692e-04,  8.9516e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6105e-03,  2.1644e-04,  4.0562e-04, -4.5012e-04,  6.9827e-04,
         3.4533e-04,  2.0056e-04, -3.4216e-04,  2.2007e-04, -2.6757e-04,
        -2.5833e-04,  3.2675e-04, -3.4144e-04, -3.5675e-06, -1.6235e-04,
         1.7515e-04, -1.8293e-04,  9.7650e-05,  2.4587e-04, -5.6802e-04,
         1.3788e-04,  3.3247e-04, -2.9508e-04, -2.3681e-04, -5.1744e-05,
        -6.4393e-02,  4.5309e-05,  6.1337e-05, -3.8153e-04, -3.2290e-04,
        -9.2291e-01,  5.7581e-04, -7.4393e-04,  1.3808e-05, -3.7804e-04],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1150: [tensor([ 5.0653e-02,  5.4841e-03,  1.3355e-03,  8.2848e-03,  4.4661e-03,
         2.4234e-03,  5.0921e-03,  6.9746e-05,  1.0836e-03,  3.5125e-03,
         2.0734e-03, -1.1442e-03,  8.2854e-03,  1.1776e-04,  3.5540e-03,
         2.4624e-03, -1.7080e-03, -7.9178e-05, -1.4691e-01, -4.0737e-01,
         2.3183e-01,  7.4281e-02, -1.0213e-02, -9.8164e-03, -1.7745e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.7843e-04, -2.2014e-04, -1.0624e-04,  3.2618e-04, -1.6547e-04,
         9.0029e-05, -2.4891e-04, -4.0239e-06,  1.4667e-04,  4.7345e-04,
        -2.6596e-04,  1.1881e-04,  1.2508e-04,  1.0169e-04,  3.6390e-04,
        -5.1578e-05,  8.1707e-05,  4.2245e-02,  4.4324e-04,  4.8540e-01,
         4.8702e-04,  4.6647e-01,  1.8259e-04,  2.9294e-04,  4.6628e-04,
         7.9644e-05,  4.6076e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6683e-03,  2.3597e-04,  4.2314e-04,  5.1328e-04,  1.7475e-04,
         1.0964e-04, -2.3086e-04,  2.3712e-04,  1.7039e-04,  1.5992e-04,
         2.8761e-06,  1.7458e-04, -5.6058e-05,  2.2121e-04,  2.7811e-04,
         7.4827e-01,  3.9897e-04,  3.7330e-02,  6.6382e-05, -4.8369e-05,
         4.5577e-03, -8.7562e-05,  4.2439e-05,  2.0387e-01,  8.1191e-05,
         1.7464e-04, -2.8356e-04,  1.2684e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3601e-03, -2.1269e-04,  3.7986e-04, -1.1423e-04,  4.7563e-04,
        -2.1279e-04,  4.6626e-05, -1.7730e-04, -3.1078e-04, -1.2259e-04,
        -1.6417e-04, -4.0148e-04, -2.5487e-04, -3.5747e-04,  1.8133e-04,
         6.0517e-02, -1.7356e-05,  5.1526e-01, -3.8019e-04,  9.9231e-02,
         2.9815e-01, -1.9273e-04, -1.5795e-04, -2.0019e-05,  2.0526e-02,
        -3.2568e-04, -3.1806e-04, -1.3073e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.1804e-04, -5.2110e-05,  3.4760e-04,  2.8860e-04,  2.7517e-04,
         3.8570e-05,  1.7925e-04, -3.3937e-05, -6.1221e-05, -5.1772e-05,
        -1.2202e-04, -4.9074e-04, -7.3735e-05,  2.0526e-04, -5.6529e-05,
         5.4536e-01,  1.3564e-04,  1.2154e-01,  2.3919e-04,  5.2677e-05,
         2.1385e-01,  3.7407e-04, -4.8023e-04,  1.0941e-01,  1.8335e-04,
         3.8513e-03, -6.8586e-04, -5.1180e-04, -4.0288e-05, -8.9195e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5896e-03, -7.0375e-05, -1.7809e-04, -4.8798e-04,  3.1911e-04,
        -5.9992e-04,  1.7639e-04, -1.4821e-04,  1.3232e-04, -3.7222e-04,
        -3.5530e-04, -2.3007e-01,  2.3987e-04, -7.6119e-01, -1.4666e-03,
        -2.9250e-04,  8.0781e-04,  5.0539e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.4772e-03,  7.2185e-05, -1.0471e-04,  3.0045e-04,  2.1063e-04,
         1.8506e-04, -1.6279e-04, -4.2418e-04,  8.2400e-04,  5.6462e-04,
        -1.2883e-04, -8.0211e-01,  1.5057e-03,  1.7358e-01, -4.8004e-05,
         1.0272e-02, -7.4592e-04,  1.4427e-03, -8.3756e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.7175e-04,  7.3713e-04,  7.6419e-04, -4.1467e-04,  2.7651e-04,
         5.5238e-04, -5.3724e-04,  2.1358e-04,  1.0235e-04, -2.7787e-04,
         4.9414e-05, -5.1849e-01, -3.6474e-04, -4.7488e-01, -7.2049e-04,
         1.0455e-05, -8.2539e-04, -1.1451e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6598e-03,  1.4152e-04,  3.6962e-04,  4.4543e-04,  5.2804e-04,
         5.4891e-04,  7.8493e-04,  4.7057e-05,  6.8766e-05,  7.4764e-04,
         3.4987e-04, -2.4502e-04,  1.8449e-04,  3.6968e-04,  2.3297e-04,
         8.1480e-01,  1.2710e-03,  2.4157e-04,  1.7626e-01, -2.1134e-04,
        -3.5365e-04,  1.4093e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4662e-03,  7.7356e-05, -2.9962e-05, -1.7485e-05,  1.1211e-04,
        -1.3844e-04,  1.1962e-05,  2.7675e-04, -2.3929e-04,  9.2077e-05,
         3.5817e-04, -2.7587e-04,  2.9397e-05,  1.7901e-04,  9.0219e-03,
        -4.9088e-04, -8.5340e-05, -8.6680e-01, -2.4458e-04, -9.2591e-05,
        -2.3139e-02,  4.5600e-04, -4.6599e-02, -2.9026e-04,  1.7471e-04,
        -4.9036e-02,  1.9397e-04,  7.0125e-05, -6.0485e-06,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7731e-03, -1.0045e-04,  4.2416e-04, -1.3382e-04, -3.8795e-05,
         1.2588e-04, -2.0427e-05,  4.2050e-04, -4.5915e-04,  2.5633e-04,
         5.1186e-05,  1.0659e-04, -1.8033e-05,  2.5054e-04, -3.1828e-01,
         6.3399e-04, -5.3735e-03, -1.8250e-04, -5.7080e-01,  4.0085e-04,
        -1.2745e-02, -8.7120e-02, -7.3886e-05,  1.5790e-04,  5.9271e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1800e-03, -1.2062e-04,  2.8332e-04, -6.4444e-04,  4.8480e-04,
        -1.5274e-04,  5.2360e-05,  1.8143e-04, -2.0861e-04,  2.3858e-04,
        -9.8147e-05, -3.5190e-04,  3.9047e-04,  2.5532e-04,  1.9668e-04,
         8.4068e-01, -1.1953e-04,  4.3076e-03,  5.0998e-05,  1.6803e-03,
        -6.9534e-04,  1.4684e-01,  6.4064e-06, -2.6436e-04,  5.1035e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1200: [tensor([-1.9684e-03, -3.0954e-04,  5.9891e-04, -5.9429e-04, -2.5390e-04,
         1.0200e-05,  5.2390e-06, -4.7629e-04,  3.1828e-04,  2.8999e-04,
        -5.7829e-06,  1.6503e-04,  2.9270e-04, -1.0006e-04, -9.8784e-01,
        -9.1226e-05,  7.1302e-04, -5.0519e-03, -2.1885e-04,  6.9428e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5957e-03, -4.9248e-04, -5.8137e-06, -6.9951e-04, -4.0844e-04,
         8.1114e-04,  1.6260e-04, -8.7909e-04, -5.8387e-06, -8.6285e-04,
        -6.3521e-04, -3.1476e-04,  2.5172e-04, -8.0279e-08,  4.9317e-04,
        -1.8424e-01, -9.1363e-04, -2.4439e-01,  7.9715e-04, -2.3976e-02,
         5.3701e-01, -4.1878e-04,  4.3370e-04,  1.9937e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3362e-04,  2.7797e-05,  6.4498e-05, -3.9651e-05, -6.0741e-05,
        -8.6272e-04,  9.0878e-05,  4.1914e-05,  4.6062e-04, -1.4749e-04,
         3.8322e-04,  6.5347e-05,  1.8013e-04,  2.8027e-04, -2.6975e-05,
        -1.4662e-02,  3.8090e-04,  2.3422e-02, -4.7883e-04,  3.9655e-03,
         9.5346e-01, -1.9110e-04, -2.5203e-04,  3.1683e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.1813e-04,  2.2573e-04,  1.8487e-04, -1.7756e-04,  1.0565e-04,
         1.4893e-04,  2.4244e-04,  1.7428e-04, -3.5215e-04,  3.0157e-04,
         2.1448e-04,  3.8011e-04,  1.3942e-04, -9.6004e-05,  8.4590e-05,
         6.7472e-01, -5.2837e-04,  2.5765e-01, -3.0859e-04,  6.2699e-02,
         5.6522e-05, -3.6520e-04,  1.2475e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.4232e-04,  3.0749e-05,  4.1504e-05,  1.9901e-04,  1.1252e-05,
         6.0579e-05,  8.3528e-05, -5.6541e-06,  3.6851e-04,  2.7300e-05,
        -1.3034e-04, -2.6845e-05,  1.0664e-04,  9.8814e-05, -3.7484e-05,
        -6.4149e-05,  1.9063e-05, -3.0123e-05, -1.8731e-04,  2.6268e-05,
         1.3029e-01, -2.7830e-04,  3.0757e-04,  3.6496e-02,  4.3815e-05,
         2.7429e-03,  2.3954e-01, -4.3762e-04, -4.1408e-04, -1.3204e-04,
         5.5661e-01,  1.2074e-04,  2.8487e-02, -2.2207e-04,  5.7785e-05,
        -9.7873e-05,  9.2412e-04, -4.2928e-04,  1.7577e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.2861e-04,  3.1461e-05, -1.1022e-04,  2.4465e-04, -2.1655e-04,
        -1.8483e-04,  1.3840e-05,  6.6096e-05, -5.5374e-05, -1.6189e-04,
        -2.5671e-06, -9.7667e-06,  1.2637e-04,  9.8510e-06,  1.0951e-04,
        -8.7673e-05, -3.5626e-05,  1.0120e-04, -1.5162e-04,  1.4750e-04,
         5.1092e-01,  9.5570e-05,  4.6360e-06,  3.6871e-03, -2.9426e-04,
         2.6062e-04,  1.9057e-01, -1.0751e-04, -2.7075e-04, -1.3987e-04,
         2.8310e-01,  2.1790e-04,  5.7105e-03, -1.0124e-04, -1.8336e-04,
         1.2498e-04,  1.1380e-03, -2.9985e-04,  1.7905e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0796e-03, -1.0854e-04, -4.8827e-05, -1.1365e-04,  1.6165e-05,
        -2.4686e-04, -2.9859e-04,  6.8825e-05, -1.5928e-05, -2.3791e-04,
        -2.6622e-04,  1.1945e-04,  1.5912e-05, -1.5571e-04, -2.0197e-04,
        -3.7253e-05, -8.5185e-05, -2.6356e-04, -3.4767e-05, -1.2218e-04,
        -6.3706e-02, -2.1444e-04,  3.6640e-04, -6.2345e-04, -5.7596e-02,
        -6.1964e-05, -2.7964e-04, -3.3651e-01,  2.6301e-05,  1.0770e-04,
        -2.1581e-04, -5.2090e-01, -4.3346e-04, -1.4019e-02,  8.1669e-05,
         2.5105e-04, -3.0174e-04, -2.2435e-04,  3.3611e-04, -2.0232e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7873e-03, -1.8766e-04, -2.6156e-05, -8.3146e-05, -1.4283e-04,
        -1.2602e-04, -1.3512e-05,  3.3871e-05, -1.7190e-04, -1.8308e-05,
        -5.4391e-05, -1.9239e-04,  1.8995e-04, -1.3218e-04,  1.8452e-04,
         1.5616e-01, -4.0624e-04,  2.4230e-04,  4.3756e-01, -4.3648e-04,
         1.7001e-01,  1.7939e-03, -1.1575e-04,  3.1708e-04,  2.2798e-01,
         1.2142e-03,  2.0726e-04, -6.1129e-05,  1.6420e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.9880e-04,  5.2961e-05, -7.9829e-04,  1.6881e-04, -5.3442e-04,
        -7.0372e-05, -5.2790e-05, -2.7611e-04,  4.4600e-04,  2.3447e-04,
         3.0785e-05, -9.4163e-05,  4.3194e-05, -9.9435e-05,  2.4317e-04,
         4.0787e-01, -4.4687e-04,  2.9525e-01, -2.1490e-04,  2.7442e-05,
         2.9136e-01, -2.7274e-04,  6.1600e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8408e-03,  2.5661e-04,  6.8774e-04,  2.0996e-04,  6.8596e-04,
         8.8652e-04,  4.0650e-04, -4.6882e-04, -1.6474e-04,  1.3632e-03,
         6.0898e-04,  4.3088e-04, -1.7970e-04,  7.0532e-05,  1.9517e-04,
        -1.6276e-02,  1.6461e-04, -9.4756e-01, -3.3087e-03,  2.2991e-02,
         6.7534e-04,  1.8068e-04,  2.1134e-04,  1.7172e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5217e-03,  7.0841e-05, -9.1266e-05,  3.1540e-04,  1.8864e-04,
        -3.8874e-05,  5.7319e-05,  5.8356e-05,  5.6835e-04,  2.1967e-05,
         2.3167e-04, -1.6026e-04,  2.3387e-04, -3.5178e-04, -5.4122e-04,
         2.7901e-04,  8.0628e-05, -2.7907e-04,  8.7944e-04,  4.9485e-04,
         9.0412e-04, -9.9099e-01, -3.6504e-04, -2.7947e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6331e-03,  6.1099e-04,  4.7741e-04,  8.0643e-04,  8.3412e-04,
         3.7091e-03,  2.3648e-03, -8.0343e-07, -2.6536e-03, -8.6891e-04,
         7.3401e-05, -2.0039e-03, -1.6616e-03, -2.6750e-04,  2.3748e-03,
         6.3516e-01,  1.0832e-01,  2.3430e-01,  8.7105e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  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 #1250: [tensor([ 1.0277e-03,  1.7577e-05,  4.5553e-04, -1.4407e-05,  2.5518e-04,
         2.8290e-05, -2.0691e-04,  6.1080e-04, -6.4809e-05,  9.0028e-04,
        -8.5362e-04, -3.8344e-04,  2.7769e-04,  2.1913e-04, -9.8630e-01,
        -2.2177e-05,  5.5054e-03, -4.3441e-04, -9.4029e-04, -6.3006e-04,
        -4.4687e-04, -4.0452e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3857e-02,  5.4285e-02,  5.6214e-03,  5.8624e-02, -1.0419e-02,
         1.7917e-02,  3.5402e-02, -1.9994e-02,  2.1312e-04,  1.8152e-05,
         4.0027e-02, -9.7178e-03,  1.9717e-02,  8.5249e-03, -2.7658e-03,
        -1.2900e-01, -3.8028e-01, -3.1595e-03,  9.0527e-03, -1.4141e-01,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.8884e-05, -9.4910e-05, -1.5936e-04,  3.1082e-05, -2.1813e-04,
         2.3175e-04, -2.9108e-05, -2.6650e-04, -3.1429e-04,  9.0702e-05,
         1.2562e-04, -5.2338e-05,  4.2528e-04, -1.5089e-04,  4.8372e-04,
        -2.8909e-04, -1.7010e-04, -3.2088e-04, -1.9353e-04,  2.3291e-04,
         3.0419e-04, -3.8896e-04, -7.2456e-05, -2.8916e-04,  1.6553e-04,
         9.6163e-01,  2.7882e-04,  2.7782e-02, -1.5276e-04,  3.2371e-03,
        -9.9804e-04,  4.9240e-06, -1.6013e-04,  5.7245e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5279e-04,  3.2031e-04,  1.0219e-04, -2.2521e-04, -1.2116e-04,
        -3.3905e-04, -9.1487e-05,  4.6792e-06,  4.5340e-05, -6.0426e-05,
        -1.4880e-04, -1.9858e-04, -1.3441e-04, -2.2347e-04, -1.1307e-04,
         1.1246e-04,  9.2568e-05,  6.1666e-05,  2.3879e-04, -1.5440e-06,
        -6.7432e-05,  5.3393e-05,  1.2435e-04, -2.1290e-04, -9.3745e-05,
         9.7761e-01, -2.8984e-04,  4.4672e-03, -4.3076e-05,  2.9173e-04,
        -1.3343e-02,  1.0346e-04, -1.0893e-04, -2.1258e-05,  2.7957e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7047e-04, -1.5867e-04,  7.6707e-04,  1.8325e-04, -2.1489e-04,
        -7.5178e-05, -3.9302e-04, -2.1714e-04,  1.1293e-04, -1.7269e-04,
         2.2864e-04, -5.3974e-04,  2.5949e-05, -2.5357e-04, -3.4248e-05,
         6.9476e-05,  3.4276e-05,  2.1930e-04,  7.8960e-04, -1.8423e-04,
        -1.7813e-04,  1.1096e-04,  6.2786e-04, -8.2561e-05, -1.0583e-04,
        -9.2639e-01,  9.2956e-04, -6.3709e-02, -1.1109e-03,  1.0788e-03,
        -4.4169e-04, -1.4596e-05, -3.7998e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0812e-04, -1.9745e-04, -2.0884e-05, -8.3811e-05,  1.1749e-04,
        -2.6717e-04, -3.4450e-05, -1.3201e-04, -5.0944e-05,  1.5737e-04,
        -1.2065e-04,  1.1894e-04,  3.5485e-04, -6.1611e-05, -1.0088e-04,
        -7.5081e-05,  9.3286e-05,  6.8757e-05, -2.1662e-04,  4.9548e-05,
        -3.5960e-05, -1.6699e-04, -1.7321e-04,  9.6989e-05,  1.7729e-04,
         4.8312e-01,  9.2364e-05,  1.6074e-02, -1.0295e-04,  4.9596e-01,
         1.3823e-04, -3.8384e-04, -4.0676e-04,  1.4210e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.4508e-04, -7.8891e-05,  2.0357e-05, -4.1898e-05, -6.1876e-05,
        -1.7536e-04, -1.4162e-04, -4.0646e-04,  1.4296e-05,  1.1964e-04,
         1.6765e-04, -5.2340e-05,  2.6957e-04,  2.4407e-04,  2.4558e-04,
        -1.5332e-04,  1.4795e-04,  4.7461e-05, -1.3921e-04, -2.3870e-04,
        -2.8783e-06, -9.3028e-05, -2.2165e-04,  1.5319e-04,  1.7539e-04,
         7.2563e-04,  3.9493e-04,  7.0612e-01, -3.6375e-04,  2.8069e-01,
        -3.1429e-06, -1.6464e-04, -1.6512e-05,  6.8056e-03,  1.3110e-04,
         1.5498e-04,  1.0084e-04,  1.6995e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6641e-05,  9.6395e-05, -4.1334e-04, -3.2629e-04, -5.3752e-05,
         1.0964e-04, -1.4032e-04,  3.7079e-04,  2.5169e-04,  2.9022e-04,
        -1.1647e-05, -5.3179e-04,  4.7335e-04, -1.9629e-04,  2.2095e-04,
         5.1314e-05, -6.5629e-05, -2.5126e-04,  3.0825e-04, -4.3913e-05,
        -2.4781e-04,  3.9803e-04,  2.2127e-04, -3.1065e-05, -2.3645e-04,
        -1.4964e-04,  1.5906e-03, -2.1675e-04,  9.8746e-01, -1.0184e-03,
        -2.2078e-03,  1.6128e-03, -8.8200e-05, -3.0248e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.8665e-04, -2.6898e-04, -4.8084e-04,  1.6243e-05,  2.1011e-04,
        -2.1040e-04,  2.9881e-04,  1.7153e-04, -2.9818e-04,  2.0214e-06,
        -1.8379e-04, -1.1862e-04,  1.7760e-04, -3.3628e-04, -3.9883e-04,
        -1.2789e-04,  4.4959e-04,  1.4480e-01, -3.6266e-05,  8.5050e-01,
        -1.3348e-04, -3.5221e-04,  3.7361e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7946e-04, -4.5461e-04, -3.9411e-04, -2.9609e-04,  1.3400e-04,
        -9.6165e-05,  1.8763e-04,  3.1461e-04, -4.5130e-04,  1.4888e-04,
        -5.3558e-05,  1.1498e-04, -6.2666e-05, -6.4148e-05,  1.4018e-04,
        -4.7449e-04,  5.5957e-04,  6.8719e-01, -3.7590e-04,  1.6833e-01,
         5.8546e-04,  1.2901e-01, -9.5122e-03,  4.9743e-05,  2.2205e-05,
         9.6945e-05,  6.0200e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.0332e-04, -2.1866e-04, -7.9218e-05,  4.0142e-04,  5.6662e-05,
        -3.1681e-04,  4.3018e-04, -5.1327e-04, -2.9679e-04,  2.2489e-04,
         1.2858e-04,  1.6641e-04,  2.5195e-04, -3.2934e-05,  7.2041e-05,
         3.0245e-04,  2.0203e-04,  5.3240e-01, -6.2335e-04,  1.3283e-01,
        -1.7438e-04,  8.8589e-04,  3.2815e-01, -1.9739e-04, -3.6199e-04,
         8.1210e-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([-8.3713e-04,  2.4479e-04, -4.9069e-05,  6.3756e-04, -4.3470e-04,
         4.9635e-04, -1.1606e-04, -1.7377e-04,  2.9194e-04, -3.8441e-04,
         1.4028e-04,  4.5248e-04,  9.9954e-05,  1.9695e-04,  3.8415e-04,
        -9.2248e-01,  7.1844e-02,  3.1571e-04,  4.1623e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  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 #1300: [tensor([ 5.2512e-03,  2.5730e-05, -1.6336e-05, -6.9036e-05, -1.4426e-04,
         3.2591e-04, -1.4206e-04,  3.0111e-04, -1.1488e-05, -9.7883e-06,
        -9.1109e-05,  7.7138e-05,  3.0989e-05, -1.9597e-04, -1.1167e-04,
         2.2526e-04,  3.5659e-06, -2.5511e-04, -2.4003e-05,  1.9812e-04,
        -2.5775e-04,  1.6240e-04, -1.0675e-04,  1.0760e-04,  2.2285e-04,
         8.0120e-03,  1.4373e-05,  1.3948e-04,  1.3475e-04, -3.8741e-02,
         3.6136e-04,  1.6814e-04,  6.9376e-04,  2.0288e-04,  8.3998e-02,
        -1.6584e-04, -1.7695e-04,  1.0135e-03,  7.2159e-04, -8.5599e-01,
         1.4015e-04,  4.0215e-04,  3.8592e-04,  1.7518e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.8906e-04,  1.5711e-04,  1.4292e-04,  1.5715e-05, -1.0253e-04,
        -5.4280e-05,  1.7015e-04,  2.2904e-04, -6.5579e-05,  2.6901e-04,
         2.8929e-04, -1.6746e-05,  1.6778e-04, -1.4003e-04, -2.1862e-04,
         7.5988e-05, -6.7227e-05, -2.7714e-04, -1.7255e-04, -1.7994e-04,
         2.3793e-04,  2.9090e-05, -1.2466e-04, -2.3824e-01,  1.7009e-04,
        -7.9454e-03, -2.9566e-05,  1.2659e-03,  1.9124e-04, -3.4538e-01,
         1.9333e-04,  4.0796e-04,  3.7287e-06, -4.0184e-01,  1.6004e-04,
        -1.0947e-04, -1.5717e-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([ 7.1867e-04,  4.4135e-04,  2.1339e-04, -3.6037e-06, -2.7754e-04,
        -2.0524e-04, -6.5447e-05,  1.6676e-05, -5.9911e-05,  3.8799e-04,
         8.2399e-05, -4.5396e-05,  1.0651e-04, -2.9652e-04,  1.8251e-04,
         5.2622e-04,  2.9253e-04, -1.8171e-04,  4.7950e-05, -2.2065e-05,
         1.8491e-04, -3.1873e-05,  6.3852e-05, -1.7407e-01,  9.5174e-05,
        -5.1867e-04, -2.6782e-04, -1.3612e-04, -8.1977e-01,  5.6764e-04,
        -1.0909e-04,  1.6475e-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([-2.9794e-04,  5.2557e-05,  2.8400e-04,  1.4573e-04,  3.5575e-05,
         7.7143e-05,  5.0708e-05,  1.7939e-04,  9.8228e-05,  1.0099e-04,
         3.9123e-05,  1.6040e-05,  1.2400e-04, -1.5396e-04, -1.2692e-04,
        -1.2207e-04, -1.6160e-04, -2.6138e-04, -9.3826e-06,  5.8333e-05,
         1.6007e-04,  2.6038e-04, -9.3602e-05,  1.0715e-04,  4.8686e-04,
        -7.9509e-02,  1.8561e-04,  1.8530e-04,  8.0628e-05, -5.9138e-01,
         8.6542e-05,  2.3908e-05,  1.4023e-03,  5.4193e-05, -3.0844e-01,
        -7.0358e-04, -1.9771e-04,  4.9535e-05,  1.3623e-02,  1.3015e-04,
         1.8611e-04,  2.6079e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.5804e-04,  2.3439e-04,  9.0397e-05,  7.3711e-05,  1.0861e-04,
        -4.3064e-05, -9.8685e-05, -2.1048e-05,  1.0268e-04, -3.7798e-07,
        -4.2458e-04,  1.1635e-04,  2.2278e-05,  7.4171e-05, -2.6776e-05,
         7.3587e-05, -1.3782e-04,  2.3202e-04, -2.4220e-06, -1.1634e-04,
        -4.5500e-04,  5.8850e-05,  8.7572e-05,  3.3881e-04, -7.3273e-05,
         3.9689e-01, -2.5883e-04, -1.6369e-03,  2.2131e-04, -4.8040e-02,
        -1.8172e-05,  7.7153e-04, -1.8505e-03,  5.4602e-01,  1.4100e-04,
         1.3455e-04,  1.4973e-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([ 1.1969e-04, -3.4338e-05, -1.8725e-04, -1.7936e-04, -3.3827e-04,
        -3.9977e-05, -2.2653e-05, -4.7865e-04,  1.9240e-04, -4.5577e-04,
         2.1728e-04, -1.4214e-04, -2.0328e-04,  9.2079e-05, -1.9919e-04,
        -3.4894e-04,  7.5403e-05,  3.1133e-05,  3.5952e-05, -7.4100e-05,
         1.9729e-04, -8.5624e-05, -1.6194e-05,  7.2084e-01, -2.9628e-04,
         2.6115e-02,  9.7840e-05,  2.9268e-03,  2.4544e-01,  3.0282e-04,
        -6.0549e-05, -1.4864e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9941e-04,  2.8764e-04,  2.0472e-04, -2.4663e-04, -3.5787e-04,
        -1.2137e-04, -4.7929e-04, -1.5893e-04,  3.9037e-04, -1.8586e-04,
         1.7747e-04, -1.0304e-04, -1.3439e-04, -2.7468e-04,  7.6829e-05,
        -2.0182e-04,  2.1017e-04, -1.0675e-04,  1.3807e-04,  7.4232e-05,
         2.8952e-04,  1.1872e-04,  1.5748e-05,  4.3159e-04,  2.4805e-04,
         5.9174e-02, -7.5238e-04, -3.1019e-04,  7.8342e-01, -3.6461e-04,
         1.5059e-01, -4.5018e-05, -4.9807e-07,  1.0702e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6290e-03, -4.8459e-05,  8.1881e-05,  1.7261e-04,  1.0626e-04,
         1.2875e-04,  7.2676e-05,  1.8141e-04, -1.7106e-04,  2.5537e-05,
        -1.8360e-04,  3.6810e-05,  2.1931e-05, -1.0911e-04,  2.8144e-04,
         2.0970e-04, -1.4472e-04, -5.3398e-05, -6.7163e-05, -9.1308e-05,
        -3.3959e-05, -2.7273e-04,  1.4447e-05, -2.1384e-04, -1.8052e-01,
        -4.0315e-01,  1.3083e-04, -9.8918e-02, -3.1256e-01, -5.1376e-05,
         2.8050e-04,  3.8151e-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([ 1.3869e-03, -2.1898e-04, -1.4997e-05, -2.3751e-04, -1.5873e-04,
         7.7551e-05,  1.3987e-04, -6.1078e-06, -1.5371e-04, -1.9016e-04,
        -2.0123e-04, -3.9246e-04, -3.1647e-04,  2.4094e-06, -1.9658e-04,
         2.1698e-04, -1.3786e-04,  6.4872e-05, -3.4066e-04, -2.4185e-04,
         8.8371e-06, -1.1283e-04,  8.9131e-05, -8.9287e-02, -2.5090e-04,
        -1.6473e-05,  2.6929e-04, -4.5894e-02,  8.5802e-01,  7.2567e-04,
        -9.0045e-05, -5.4332e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2564e-04, -2.4072e-04, -2.3308e-04, -6.8695e-05,  7.3536e-05,
         1.1006e-04,  1.3020e-04,  5.0329e-05,  3.8626e-04, -9.3663e-05,
        -2.9717e-04,  1.1691e-04, -3.7172e-04,  4.5809e-06,  1.1676e-04,
         4.9080e-05, -2.4528e-04,  7.2453e-05, -1.5030e-04,  4.9154e-04,
         4.7897e-04,  3.5239e-04,  5.1373e-04,  7.7829e-03, -9.8594e-01,
         1.6989e-04, -5.5776e-04,  6.2553e-04, -5.5147e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.1312e-04,  1.3156e-04,  3.0705e-04, -7.5523e-05, -2.0592e-04,
         3.2581e-04, -7.8098e-05, -1.5534e-04, -2.7681e-04,  1.4543e-04,
        -8.5256e-06, -3.1149e-05, -2.7801e-05, -2.5702e-05,  3.5752e-04,
         1.1160e-04, -4.9424e-05,  1.5226e-04,  3.3383e-05, -3.2036e-04,
         1.8022e-04,  3.2746e-04, -5.2585e-02,  6.0561e-04, -1.6660e-04,
        -1.0780e-04,  8.9133e-05, -1.4096e-03,  9.3960e-01,  3.1078e-04,
         8.4274e-04,  1.4912e-04,  8.8590e-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([ 5.8928e-04,  2.0067e-04,  1.4792e-04,  2.0877e-04,  2.9609e-05,
         2.3746e-04,  5.3187e-04, -1.0554e-05,  1.5611e-04,  1.0838e-04,
         4.0040e-05,  1.9833e-04,  2.1400e-04,  6.5701e-05,  1.3302e-04,
         6.2970e-05,  7.6829e-04,  5.2415e-04,  2.6974e-04,  2.7572e-04,
        -2.6867e-04,  1.8344e-05,  8.0303e-01,  2.8830e-04,  1.8486e-01,
         3.8332e-04,  3.9590e-05, -1.2532e-03,  2.8467e-03, -1.8448e-03,
         7.0533e-06, -3.8929e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  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([ 3.9273e-05,  8.0827e-05,  9.6957e-05,  1.8553e-04,  4.5115e-04,
        -1.5073e-04,  8.1117e-06,  3.2117e-05, -1.8303e-04, -1.2458e-04,
        -2.2717e-04,  9.9274e-05,  2.1361e-04,  2.9151e-05, -5.0931e-04,
        -6.5660e-04, -4.0204e-04,  7.4258e-04, -9.2148e-05, -2.1424e-04,
         5.2896e-04, -2.1190e-04,  1.4985e-04, -9.9289e-01,  1.2363e-03,
        -9.1812e-05,  3.5169e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4326e-04,  2.0502e-04, -1.9016e-04, -3.0951e-05,  1.9824e-04,
         3.7687e-05,  2.1808e-04, -2.7913e-06,  5.9263e-05,  2.4342e-04,
        -5.9139e-05,  7.4689e-05,  1.1395e-04, -1.0571e-04,  9.9676e-05,
        -1.4875e-04,  1.9866e-04,  2.6338e-04, -3.8659e-06,  7.1640e-05,
        -1.1806e-04, -3.4698e-04, -1.2122e-04,  1.9377e-01, -4.6264e-05,
         6.9934e-01,  2.5847e-04,  2.5547e-04,  9.7475e-02, -4.2296e-03,
         5.1633e-04,  4.9758e-04, -9.7415e-05, -6.0784e-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([ 7.3067e-04, -1.8340e-04, -1.3973e-04, -3.0593e-04, -1.6145e-04,
        -5.6572e-04,  4.8613e-05, -4.4175e-04,  1.2275e-04,  3.1202e-05,
        -3.4261e-04, -3.6136e-04, -1.4627e-04,  9.2103e-05,  2.1241e-04,
        -3.7966e-04, -2.4652e-04, -2.6956e-04, -2.8606e-04,  1.7282e-04,
         4.1958e-04,  8.1340e-04,  7.8872e-04,  8.5330e-01,  1.3790e-01,
         1.2896e-04,  1.0555e-03, -3.3790e-05, -2.0824e-04, -1.1069e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.5544e-04,  8.9078e-05,  3.9624e-05,  5.6216e-05, -2.6835e-04,
        -2.0598e-04, -3.0898e-04,  2.6769e-04,  1.7956e-04,  1.1274e-05,
         4.0091e-04,  1.9560e-04,  2.8703e-04,  1.3228e-05, -2.0205e-04,
         3.5872e-04,  2.5782e-04,  3.8845e-05,  2.0011e-04,  6.7682e-05,
        -1.6985e-04, -9.5724e-05,  3.4242e-05, -5.7794e-01, -1.6943e-04,
        -4.1658e-01, -2.0767e-05, -1.0411e-04,  3.6839e-04, -3.0703e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4412e-04, -4.1611e-04, -2.4422e-04, -3.2581e-04, -1.7170e-04,
        -3.9901e-05, -3.8723e-04,  2.9184e-04, -1.8751e-04, -6.0888e-05,
         2.4245e-04, -1.0816e-04, -4.9467e-04, -1.9289e-04,  8.9747e-05,
         1.7592e-04,  8.4004e-05, -1.8114e-04, -4.5111e-05, -1.0561e-04,
        -1.1252e-04, -3.4047e-04, -4.9443e-04,  1.5981e-01,  2.4739e-03,
         8.3097e-01,  8.8726e-04, -5.9896e-04,  8.5998e-05,  2.3635e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1445e-03,  3.9232e-04,  4.7363e-05,  5.5719e-05, -2.2579e-04,
         7.0565e-05, -3.5582e-04,  2.1175e-04,  2.4803e-04,  6.9675e-04,
         1.3687e-04,  1.1794e-04,  6.8908e-05,  1.2682e-04,  6.4940e-04,
         6.3495e-05, -1.3365e-04, -3.1040e-05, -2.2935e-06, -1.4373e-04,
         2.0842e-04, -3.1323e-04,  5.2315e-05,  9.6031e-01, -2.2677e-04,
        -3.2209e-02,  3.4687e-04, -8.6432e-04, -5.4658e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.2630e-04,  3.5170e-04,  1.7127e-04,  9.5223e-05, -2.9215e-05,
         6.0223e-04,  1.6182e-04,  3.6794e-04,  1.9851e-04, -1.3770e-04,
         4.1181e-04,  4.4498e-04,  4.8286e-04,  3.2123e-04, -2.5378e-05,
         1.3792e-04, -1.0743e-04,  3.8089e-04, -1.6075e-04,  1.9974e-04,
         9.9191e-01, -3.3622e-05, -4.9861e-04, -7.1886e-04,  5.0055e-04,
         6.0815e-04, -1.8102e-05, -5.3278e-05, -1.4449e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3141e-05,  7.8999e-05, -8.1123e-05,  2.7829e-04,  3.0057e-04,
        -9.0581e-05,  2.9110e-04, -7.9775e-05,  2.2293e-04,  8.3214e-05,
         3.8704e-04,  1.4076e-04,  2.1395e-04, -3.4630e-04,  1.7631e-04,
         1.1429e-04,  2.1875e-04,  3.6516e-04,  3.6331e-04,  1.6442e-04,
        -1.3658e-04, -8.3811e-05,  9.4953e-01, -3.2095e-04, -6.4733e-04,
         2.0386e-02, -2.3478e-02,  2.9536e-04, -4.3758e-04, -1.1724e-04,
         1.0609e-04,  2.7650e-04,  6.8131e-05, -1.1086e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3325e-03,  3.7521e-04, -3.4784e-05,  1.1403e-04, -6.5051e-04,
         6.0411e-04, -1.4708e-04,  3.2293e-04, -2.5211e-04, -3.1140e-05,
        -1.7235e-04,  6.9116e-05, -4.8406e-04, -7.1340e-05, -3.4589e-04,
         4.4722e-04, -3.0516e-04,  6.8039e-05,  5.1863e-04,  3.5189e-04,
         4.5938e-04,  1.9675e-05, -2.2021e-04, -3.9090e-04,  1.0576e-02,
        -4.2496e-04, -3.9328e-04, -9.7998e-01,  3.9450e-04, -3.9345e-04,
         5.2402e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9915e-04,  9.9885e-05,  2.3827e-04, -1.5906e-04, -1.3019e-04,
        -5.4409e-04, -1.0359e-03, -7.4766e-05, -8.3081e-05, -1.2314e-04,
        -1.4972e-04, -3.3750e-05, -3.6778e-04, -1.9911e-04, -2.7383e-04,
        -6.2844e-04,  2.8340e-05, -1.6553e-04,  8.4842e-04,  2.0494e-05,
         5.2716e-05,  9.3622e-04,  4.6279e-04, -9.9022e-01, -2.4372e-04,
         4.4480e-04, -5.7239e-04, -8.7751e-04, -4.2278e-04, -6.2480e-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([ 2.4977e-04, -1.0135e-04,  2.3892e-04,  2.8174e-04,  3.9226e-04,
         7.4897e-05,  4.9150e-04,  2.2281e-06,  3.7957e-04, -1.9607e-04,
         1.0176e-05,  1.6262e-04, -2.5270e-04,  2.0160e-04,  9.2367e-06,
        -5.7680e-05,  2.6997e-04, -2.2384e-04, -1.1298e-05,  1.6058e-04,
        -7.5591e-05, -4.0280e-04,  3.2768e-04,  3.3145e-02, -1.8356e-04,
         9.6051e-01, -4.8395e-04, -4.6509e-04, -6.4143e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8279e-05,  1.0680e-04,  1.9363e-04, -8.8891e-05,  7.4801e-06,
        -4.4287e-05,  1.4891e-04, -3.5218e-04, -5.9268e-05, -1.2122e-04,
         5.9929e-06,  6.9362e-05,  2.3757e-05, -5.9427e-04,  8.2395e-06,
        -8.4115e-05,  1.2070e-04, -9.9807e-05, -3.2969e-04,  7.7291e-05,
        -2.1108e-04,  2.5061e-04, -7.3323e-05, -1.4814e-04,  3.9493e-04,
        -2.6315e-04,  1.6083e-04, -6.9597e-05, -6.7012e-06,  4.1584e-05,
         1.5184e-04,  9.9842e-05, -2.4588e-04, -8.6574e-05,  1.0817e-04,
        -3.5306e-05,  1.1036e-05, -5.3133e-05, -2.8628e-04,  2.3653e-04,
        -3.1922e-03,  9.6118e-01,  2.7257e-02, -6.7535e-04, -1.3841e-04,
        -4.1908e-04,  5.5135e-04, -4.6895e-04,  1.0197e-04, -5.1108e-04],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1400: [tensor([-3.5728e-04, -1.5040e-04,  2.3712e-04,  2.7461e-04, -1.1281e-04,
         2.3745e-04, -1.4835e-04, -1.7890e-04,  1.6906e-04,  2.6566e-04,
        -6.5204e-04,  3.9942e-04,  7.1407e-01, -3.4606e-04,  2.8130e-01,
         2.3946e-04, -7.4577e-04, -1.1590e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2096e-04,  9.8635e-06, -8.6778e-05, -2.6101e-04, -6.1547e-05,
        -1.8344e-04, -2.9106e-05,  1.4778e-04, -6.5921e-05,  1.7947e-04,
         1.6878e-04,  2.8650e-04,  5.3185e-01,  1.8245e-04,  7.1708e-02,
         8.2482e-05, -1.0234e-02, -9.1544e-05, -6.6648e-03,  3.7732e-01,
        -1.6316e-05, -1.6180e-04,  7.8903e-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([-1.6707e-03,  4.1834e-04,  5.0783e-04, -2.5970e-04, -1.7485e-04,
         4.4438e-04,  1.5677e-04, -4.1839e-04,  9.6041e-05,  2.6952e-05,
         2.7195e-04, -2.9891e-04, -5.4852e-01,  6.7448e-05, -4.4452e-01,
         1.4429e-04,  2.8203e-04, -5.6703e-05,  7.2472e-05, -5.2780e-04,
        -5.5245e-04, -5.1418e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.3540e-04,  1.7074e-04,  2.1975e-04, -1.8833e-05, -3.2822e-04,
         1.9717e-04, -2.1912e-04,  1.7119e-04, -1.3872e-04, -1.2280e-04,
         8.3026e-05,  1.1209e-04,  7.9520e-01,  2.6734e-04, -2.2445e-04,
         1.9045e-01, -7.0866e-05, -5.3480e-03, -1.2449e-04, -3.6039e-03,
        -2.0083e-03, -4.2335e-05, -1.2642e-06, -4.0512e-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([-2.0838e-04,  1.7245e-04,  1.3184e-04,  4.0496e-04, -1.3520e-04,
        -2.3616e-05, -3.1300e-05,  1.0363e-05,  4.1752e-04,  7.6605e-04,
         2.6952e-04,  2.5394e-04,  6.5398e-01,  7.0227e-04,  2.9364e-01,
         4.3736e-04,  4.6889e-02,  4.1275e-04, -1.7272e-04,  1.3128e-04,
        -1.7153e-04,  6.3302e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6586e-04,  5.9996e-04,  1.5956e-04, -7.8853e-04,  5.3825e-04,
         1.5935e-04, -2.1269e-04, -2.4902e-04,  1.6224e-04,  2.7046e-04,
        -3.4453e-04, -6.2984e-05, -9.8334e-01, -4.0404e-04, -2.8503e-04,
         4.5100e-04,  2.9139e-03,  7.4137e-03, -8.5623e-04, -3.9874e-04,
         2.2551e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.3794e-04, -8.4574e-05,  1.5724e-04,  9.3439e-06,  6.2630e-04,
         9.6231e-04,  1.8215e-04,  2.9757e-05, -4.0414e-04,  1.3487e-04,
         1.5279e-04,  6.7345e-04,  4.4750e-04,  3.4486e-02, -7.7732e-04,
        -9.5684e-01, -1.2647e-03, -8.0660e-04, -1.3237e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.4981e-04, -2.1873e-04, -2.4973e-04,  1.7736e-04,  1.0225e-05,
        -2.7808e-04,  2.4744e-04, -4.0265e-04,  9.3151e-05, -1.6987e-04,
        -3.9462e-04, -6.5841e-05,  2.0311e-04, -4.9060e-01, -9.0648e-05,
        -2.1292e-01,  1.5653e-05, -2.9225e-01, -2.0435e-04, -3.1136e-06,
         5.5955e-05,  5.9530e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9399e-04,  1.2724e-04,  4.2735e-04, -2.1787e-04,  7.6616e-05,
        -2.9464e-04, -3.0214e-05, -4.9033e-04,  3.8613e-04, -9.1253e-05,
         7.0507e-04, -6.9981e-05,  2.8836e-04, -9.1662e-01,  7.4446e-04,
        -7.6756e-02,  1.4358e-04, -1.0801e-03, -3.5517e-04,  8.0103e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1993e-04,  1.5183e-04,  3.0571e-04, -2.2154e-04, -1.9068e-04,
        -7.7660e-05, -3.5171e-04, -3.1383e-05,  6.9330e-05, -1.8891e-04,
        -2.8166e-04,  2.4212e-04, -1.0491e-04, -2.6796e-04,  2.7703e-05,
        -3.2874e-04,  1.1406e-04, -4.7877e-04, -6.2432e-05,  2.4743e-04,
        -5.7475e-04, -4.5866e-01,  5.4864e-04,  7.4805e-04, -5.3421e-01,
        -1.9065e-04,  3.7968e-04, -3.2497e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.3477e-04,  2.7264e-04,  1.9663e-04,  7.6224e-05, -4.9741e-05,
         2.2909e-04, -1.8285e-04,  1.4603e-04,  6.4747e-05,  1.7046e-04,
        -3.5835e-04,  2.8187e-04,  2.8012e-04, -4.0869e-05, -1.6468e-04,
        -1.8477e-04, -1.7279e-04,  1.0329e-04,  2.9095e-05,  2.7502e-04,
        -1.5283e-04,  1.2332e-02,  1.3355e-04, -9.2643e-05, -9.8215e-01,
        -1.3813e-04, -7.3234e-04,  2.1769e-04,  1.4809e-04, -1.8515e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.2718e-04, -1.1866e-04, -4.0527e-04,  3.8808e-05, -2.6095e-04,
         1.5550e-04,  5.4447e-04,  2.7433e-04, -1.4990e-04,  1.2852e-04,
         2.6664e-04,  7.1775e-05,  1.9294e-04, -1.3441e-04, -4.5923e-04,
        -1.3855e-04, -3.7604e-04, -1.1857e-04, -6.4987e-04, -4.1164e-04,
         3.1316e-04,  9.6995e-01,  1.2188e-04, -4.9234e-04,  2.3532e-02,
        -6.4717e-05, -6.5391e-05,  3.6090e-05,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1450: [tensor([ 2.3342e-04, -2.0875e-04,  2.4438e-04,  5.5465e-05, -6.0611e-05,
        -3.7140e-04, -1.3624e-04, -2.1960e-04, -1.7374e-04, -1.3727e-05,
         6.4094e-05, -4.2123e-02,  1.9486e-04, -1.6765e-01, -1.4210e-01,
        -2.5620e-01,  3.8010e-04, -3.5800e-02, -3.5294e-01,  6.7510e-05,
         5.5861e-04,  1.1966e-04,  8.5375e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0433e-03,  3.2481e-05,  3.9009e-04, -2.0931e-04,  2.3034e-04,
         1.3167e-04, -1.5540e-04, -3.5127e-05,  3.4352e-04,  9.8148e-05,
         1.5708e-04, -4.2909e-01,  3.6716e-04, -5.6380e-01, -6.5449e-04,
        -2.1399e-03,  4.0876e-05, -3.4774e-04,  1.6105e-04, -5.7526e-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([ 4.4085e-04, -6.8605e-05,  2.2753e-04,  3.9683e-04, -5.7065e-05,
         1.7202e-04,  2.2238e-04,  3.6777e-04,  4.7156e-04,  2.1668e-04,
         4.6483e-04, -1.5313e-01, -3.3619e-04, -8.4151e-01, -1.1960e-04,
         4.6898e-05, -2.7352e-04,  7.4864e-04,  3.7938e-04,  9.0563e-05,
        -2.6159e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3310e-04, -3.8025e-04, -5.4921e-04, -1.5287e-05,  8.0772e-04,
        -7.3758e-05,  7.2393e-04, -2.3568e-04,  3.3827e-04,  5.2352e-05,
        -3.0407e-05, -7.3443e-01, -9.9908e-04, -2.5021e-01,  4.7322e-04,
         9.9340e-03,  6.6275e-05, -3.4611e-04,  1.9830e-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([-2.9457e-03, -3.4531e-04, -5.9101e-04,  3.0403e-04,  3.7439e-04,
        -2.5281e-04, -8.5152e-05, -1.1921e-05,  4.9110e-04, -2.6313e-04,
         4.2911e-04, -1.7143e-01, -2.8952e-04, -8.2115e-01,  1.8855e-04,
         3.2546e-04,  5.1932e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0481e-03, -2.9251e-04, -2.8546e-04,  6.7771e-04, -2.7384e-04,
        -1.6337e-04, -5.9635e-04, -1.6052e-04,  3.9567e-04,  1.5991e-05,
        -3.1011e-05, -6.7025e-01,  1.2489e-04, -3.1734e-01,  6.4241e-04,
        -2.0063e-03, -5.2367e-03, -1.8105e-04, -7.4710e-05,  2.0948e-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([ 7.6772e-04,  6.8028e-05,  5.1714e-04,  2.0179e-05, -2.9214e-04,
         4.3575e-04,  3.8615e-04,  4.2928e-04,  3.7671e-04,  2.8984e-04,
         4.5014e-04, -8.4699e-05, -4.5779e-05,  3.2370e-04, -3.9142e-01,
        -3.3485e-04, -6.0158e-01, -1.5331e-03,  3.5789e-04,  2.8761e-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([-6.3296e-04,  1.9867e-04,  4.0732e-04,  5.9654e-04,  6.7814e-04,
         1.8876e-04,  3.7611e-05, -1.2063e-04,  1.6541e-04, -3.2385e-04,
         1.8407e-04,  2.2045e-04,  1.6529e-04,  1.5748e-04,  1.8160e-02,
         6.4457e-04, -8.0166e-01,  6.2646e-05,  1.7720e-03, -1.7281e-01,
        -4.6872e-04, -1.4441e-04,  1.9739e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5745e-04, -1.1939e-04,  2.3430e-04,  1.4344e-04,  3.3788e-04,
         8.7382e-06,  1.5781e-04,  2.5446e-04,  2.9650e-04,  1.9641e-04,
        -3.6884e-05,  9.3074e-05,  3.2412e-05,  2.7614e-04, -3.6420e-01,
         2.5433e-04, -4.8414e-01, -6.8894e-04, -1.4705e-01, -1.9352e-04,
         4.4325e-05,  2.3097e-04,  3.7307e-04, -2.7816e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.5290e-04, -2.5741e-04,  1.2459e-04,  3.3240e-05,  3.9083e-04,
        -3.0522e-04,  1.9613e-04,  1.3760e-04, -9.2259e-05, -3.1065e-04,
         6.7547e-05,  8.3529e-05,  6.0550e-05,  4.6490e-04, -9.7445e-01,
        -7.1483e-04, -1.2042e-02,  1.6159e-05,  3.2450e-04,  5.7014e-03,
         2.6638e-03, -1.9599e-04,  2.1850e-04,  2.9190e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9458e-04,  2.8695e-06,  2.6172e-04, -8.3378e-06, -1.6581e-04,
         1.8433e-04,  2.3657e-04, -3.6156e-04, -1.4375e-04,  1.6989e-04,
        -4.5803e-04,  1.6634e-05, -6.2348e-04,  3.0092e-05,  2.9294e-01,
         8.8220e-04,  3.7083e-02,  6.6426e-01,  3.3287e-04, -1.6062e-04,
        -2.3761e-04, -6.9643e-04,  3.4978e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4325e-04,  1.2481e-04,  1.5638e-04, -5.1393e-05,  4.1673e-05,
         7.7941e-05, -1.8507e-04, -1.9347e-04,  1.4714e-04,  7.4379e-05,
         3.0317e-04, -2.3720e-04,  1.2827e-05,  8.3234e-05, -6.8349e-01,
         1.2906e-04, -3.0601e-01, -1.7736e-04, -8.1248e-04, -1.5708e-04,
        -1.2351e-04, -5.2782e-03,  1.1706e-03,  1.4608e-04,  2.9437e-04,
        -8.4198e-05], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1500: [tensor([-1.3635e-05,  1.0906e-03, -5.7166e-05, -2.2550e-04, -1.6619e-04,
        -4.2160e-04, -2.8106e-05, -5.2512e-04,  6.9338e-04,  9.9434e-01,
        -8.4637e-04, -1.0506e-06, -2.1638e-04,  5.6809e-04,  3.8296e-05,
         7.6700e-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([ 1.4986e-03, -3.6489e-06, -2.1109e-04,  2.2054e-04, -1.8230e-03,
         5.3880e-04,  1.6913e-03, -5.1005e-04, -3.3935e-04,  8.3884e-01,
         9.4618e-03,  1.3917e-01,  1.6126e-03,  2.9690e-03,  9.3444e-04,
         1.6825e-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([ 2.0495e-04,  6.5306e-05, -7.4612e-04,  6.6788e-04, -2.0447e-04,
         2.1492e-05, -4.5326e-05,  3.9120e-04,  1.0582e-04,  6.0252e-01,
         2.9712e-04,  3.8678e-01, -4.6930e-04,  6.9662e-03, -3.8370e-04,
         1.3516e-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([-9.2295e-04, -5.6326e-05, -2.0403e-04,  7.3255e-05,  1.4896e-04,
        -4.4194e-04, -1.4668e-03, -2.3119e-04,  2.2234e-04, -2.0546e-03,
        -9.9219e-01,  8.8775e-04, -7.5676e-04, -3.4348e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4680e-03, -2.0864e-05,  4.8690e-04, -1.8871e-04, -2.5233e-04,
         1.8279e-04, -9.9368e-05,  1.3925e-04,  4.3939e-04, -7.6678e-01,
        -1.0371e-03, -2.2174e-01, -2.2675e-04,  3.3915e-04, -2.0249e-04,
         2.2050e-04, -3.4910e-03, -1.0932e-03, -2.1904e-04, -3.7153e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4780e-03,  4.1259e-04, -1.6238e-04,  1.3753e-04,  6.0911e-04,
         1.4882e-05, -8.3161e-04, -1.1671e-04,  2.2686e-04,  4.5382e-04,
        -7.0272e-01, -5.0559e-04, -2.8955e-01,  1.1007e-03,  3.1508e-04,
         3.6571e-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([-6.2003e-04,  1.1858e-04,  3.5276e-05, -4.2190e-04,  2.0458e-05,
        -1.2213e-04,  4.3576e-04, -2.1128e-05,  5.0399e-05, -4.9835e-01,
         1.8805e-05, -4.8855e-01, -5.7000e-04,  8.9619e-03, -1.7793e-05,
        -4.4405e-04,  2.9533e-04,  2.6578e-05,  9.9510e-05,  3.9895e-04,
        -2.0259e-04, -2.1738e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.9910e-05, -1.8314e-04, -5.3639e-04, -3.5371e-04, -2.9648e-06,
        -1.4084e-04,  4.5717e-05,  1.4432e-04,  3.1543e-04, -5.7276e-01,
        -6.3797e-04, -1.4774e-01, -1.0253e-03,  7.7028e-03, -2.6773e-01,
         3.1380e-04, -2.4565e-04,  2.5119e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9003e-03,  1.2028e-03, -5.3385e-04,  1.4721e-03, -1.2551e-05,
        -1.0989e-03, -5.1622e-04,  5.0721e-04,  7.6739e-04,  5.6015e-04,
         9.8885e-01,  1.6372e-04,  2.1071e-03,  1.2337e-05, -2.9378e-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([-2.9395e-04, -6.5814e-05, -6.9017e-04,  1.7405e-04, -5.1396e-04,
         2.6464e-04,  8.8958e-06,  5.3192e-05, -1.7552e-04,  5.3140e-01,
        -1.1511e-04,  4.6437e-01,  5.0086e-04,  7.2028e-04,  6.4696e-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([-1.3278e-03,  1.5534e-04, -7.1822e-04,  1.6474e-04,  4.2543e-04,
         1.3386e-04,  9.0521e-04,  6.4851e-04, -5.4007e-04, -6.4275e-01,
        -6.5124e-04, -3.4137e-01, -1.2912e-03,  6.2859e-03, -1.8375e-03,
         2.1814e-04, -5.7402e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5053e-04, -5.2595e-04, -6.9565e-04,  3.9520e-04,  3.8014e-04,
        -4.0611e-05,  2.7484e-04,  1.6076e-04,  2.2291e-04,  8.5997e-01,
        -1.5397e-03,  1.3395e-01, -5.6484e-04,  5.0671e-04, -1.8242e-04,
         2.4435e-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 #1550: [tensor([-9.4406e-05, -3.5042e-04, -1.1483e-04, -2.5570e-05, -4.2477e-04,
         8.8253e-05, -2.6233e-04,  4.3447e-05, -2.7681e-06, -2.0108e-04,
        -2.5952e-04,  9.3969e-05,  1.4760e-04,  2.5759e-04,  7.4883e-05,
        -2.2064e-04,  1.1129e-04, -1.9279e-04,  4.3596e-04,  1.7778e-04,
        -1.8551e-04, -2.1135e-04,  1.1249e-04, -6.1903e-04, -6.5148e-05,
         1.5193e-04,  7.0655e-06,  4.6909e-05, -1.6607e-03, -2.3079e-04,
        -3.3388e-04,  6.6930e-01,  3.2289e-01,  1.7084e-04, -4.0334e-04,
         2.8930e-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([ 1.1988e-03,  4.2117e-04, -1.5702e-04,  2.7334e-04, -6.5228e-05,
        -2.2203e-04, -1.9614e-05, -3.9250e-04,  2.3449e-05,  2.4743e-04,
         8.1114e-05,  1.3453e-04, -1.6792e-06,  3.9598e-05,  3.5813e-04,
         3.1717e-05, -1.4203e-04, -2.0165e-04,  6.3684e-05, -1.1208e-04,
         2.9498e-04,  2.0150e-04, -1.3706e-04,  1.0030e-04,  3.6867e-05,
         1.1823e-04, -3.0307e-04,  2.7637e-04, -1.4316e-01, -1.0799e-04,
         4.9285e-04, -1.8054e-04, -3.4378e-04, -8.4701e-01,  5.0140e-04,
         1.8987e-03, -1.6994e-04, -1.7542e-04,  8.0112e-05, -2.2238e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1791e-03,  2.2631e-04,  1.4417e-04, -8.7987e-05, -3.9761e-04,
         2.1009e-04, -2.8689e-05, -1.3784e-04,  1.2882e-04,  2.8117e-04,
         9.8761e-06, -1.9134e-04,  2.4358e-04,  1.8017e-04, -1.7444e-04,
        -6.6183e-05, -1.1378e-04,  8.4282e-05, -2.1657e-05,  7.0204e-05,
         2.6683e-04,  8.7585e-05,  9.8735e-05, -1.2174e-04,  1.9708e-04,
        -1.9414e-04,  2.4205e-04,  9.6950e-01, -7.8928e-04, -2.2935e-02,
        -8.0972e-04, -5.4882e-04,  1.4610e-04, -8.4896e-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([ 6.4072e-04, -2.6045e-04,  1.1823e-04,  3.4990e-05, -2.8339e-04,
         2.0583e-04,  1.3929e-04, -3.5060e-04,  3.1480e-04,  2.3112e-04,
        -2.8461e-04,  3.2223e-04,  1.7905e-05, -7.5094e-05,  1.9026e-04,
         5.8095e-05, -1.6368e-05, -1.3179e-04, -3.0100e-04, -2.2243e-04,
         4.2523e-05,  3.2907e-04, -1.3756e-04,  2.9535e-05,  1.7270e-04,
         1.3368e-04,  1.2207e-04, -1.9432e-04, -9.9278e-01, -8.2094e-04,
        -3.6136e-04,  3.6150e-04,  3.1535e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2811e-04, -1.2555e-05, -1.8636e-05,  4.7299e-05,  4.0464e-05,
        -1.0384e-05,  2.5947e-05, -1.0245e-04, -5.0763e-05,  7.1983e-05,
        -2.5746e-04,  2.3126e-05,  1.7261e-04,  3.2939e-05,  9.4521e-05,
         7.5014e-05, -3.8140e-05,  1.7105e-04, -6.2600e-05, -1.1308e-04,
        -7.1700e-05, -5.3680e-05,  4.7522e-05,  1.1448e-04,  1.3512e-04,
         1.0108e-04, -1.0191e-06, -1.7745e-01, -1.4262e-04,  1.2471e-04,
        -2.9767e-01, -2.4473e-04, -5.7729e-04, -2.7510e-01, -2.4583e-01,
         1.3428e-04, -1.4033e-04, -2.1570e-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([-5.6181e-05,  2.0542e-04,  3.7860e-04,  1.3113e-05, -2.2684e-04,
        -2.1461e-04, -1.4754e-04,  2.1762e-04, -3.4644e-05, -8.3480e-05,
         6.4669e-05,  6.7966e-05, -2.1692e-04, -2.2636e-04, -1.5557e-04,
        -6.6734e-05, -8.0472e-05,  2.1932e-04, -2.7886e-05,  1.1425e-04,
         3.4892e-05, -1.5549e-04,  3.9672e-04, -2.1881e-04, -1.0520e-04,
        -1.8876e-04,  1.8620e-04,  9.1897e-01, -2.9341e-04, -1.8614e-04,
         1.6387e-04, -5.9568e-02,  1.2833e-04,  7.0800e-04, -1.3749e-03,
         1.7761e-04,  3.8259e-05, -1.3756e-02, -2.1230e-06, -8.8572e-05,
        -3.4552e-04, -9.1898e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.2350e-04,  2.9915e-04, -8.6212e-05,  1.8867e-04,  1.0912e-04,
        -2.0565e-05,  9.8139e-05, -9.8935e-05, -1.1281e-04,  1.4222e-04,
         4.7125e-05, -1.0421e-04,  1.8025e-04,  3.7426e-04,  2.0599e-04,
        -2.3685e-04,  1.7710e-04,  1.0151e-04,  7.4399e-05,  3.5559e-04,
         2.0851e-05,  2.2875e-04,  7.9363e-05,  1.4026e-04, -7.6791e-05,
         1.7928e-04,  2.7674e-05, -5.7280e-04, -1.6484e-04, -2.9390e-04,
         2.3884e-04,  1.2754e-04,  1.4093e-04,  3.6196e-02,  9.0067e-01,
        -4.7120e-05, -6.5513e-05,  4.1765e-04, -3.5361e-04,  4.6722e-02,
        -1.9335e-04, -8.9838e-03,  1.2202e-04, -2.0045e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1282e-03,  5.4093e-04, -4.0572e-03,  4.6640e-03, -8.5344e-03,
         9.7943e-04, -4.6706e-04, -2.4314e-03, -5.4118e-03, -1.3557e-03,
         8.9891e-03, -2.7235e-03,  2.3257e-03,  1.4601e-03,  2.6428e-03,
         1.1837e-03, -2.1950e-03,  1.7156e-03, -2.6608e-03, -5.7930e-03,
         6.4174e-03,  2.8003e-05,  2.6149e-03,  9.2229e-04, -1.6642e-03,
         1.2028e-03,  2.2258e-04, -2.0690e-03,  8.9929e-01,  1.3502e-02,
        -5.8100e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.8980e-04, -5.7350e-05, -5.2525e-06, -2.2430e-04,  1.1419e-04,
        -1.5376e-05, -3.0990e-04, -8.5759e-05,  4.0160e-04,  3.6928e-04,
         3.0992e-04, -3.4310e-04,  2.9427e-04,  5.0101e-05, -7.5850e-05,
         1.6489e-04,  7.8754e-05, -2.4751e-04, -5.7588e-04, -2.5753e-05,
         4.5170e-04, -1.6979e-04, -4.9334e-04,  9.8720e-01, -7.4311e-05,
         3.4605e-03, -5.3480e-04, -4.1337e-04,  1.5569e-03,  4.6968e-04,
        -4.0436e-04,  1.2827e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5149e-04,  2.4300e-04, -2.0423e-04, -1.8856e-05,  1.6505e-04,
        -7.0414e-05, -3.3158e-04, -1.9047e-04,  2.3214e-04, -1.6878e-04,
         3.9906e-04,  2.4923e-04,  1.3707e-04,  2.4730e-04,  3.4988e-04,
         3.9993e-05, -3.5247e-04,  1.0476e-04,  1.9364e-05, -1.8647e-05,
        -1.0429e-04,  4.5103e-04, -1.1902e-04,  1.1072e-04,  7.9266e-02,
         3.9351e-04,  9.1499e-01,  1.3051e-04, -2.0088e-04,  3.4350e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.0390e-04, -3.4244e-04,  2.7169e-04, -4.5766e-04, -5.8734e-04,
         4.4206e-05, -2.9409e-04,  3.5700e-05, -1.7204e-04, -6.7062e-04,
        -1.3791e-04, -2.3514e-04, -2.7217e-04, -2.7705e-05, -1.7450e-04,
        -6.5463e-04, -4.4904e-04,  1.4280e-04,  8.9291e-05,  1.0568e-04,
        -2.8093e-04, -5.9321e-04,  5.9456e-04,  2.6154e-04, -9.9148e-01,
        -7.1728e-04,  1.6049e-04, -3.4695e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1064e-03,  7.3969e-05, -4.9648e-04,  9.2967e-07,  2.2289e-04,
         2.2229e-04,  2.1875e-04, -3.0351e-04,  1.4679e-04,  1.1633e-04,
         7.3995e-05,  1.0419e-04, -2.9500e-04,  9.4414e-05,  2.7474e-04,
         1.8594e-05, -2.7274e-04, -5.2941e-05, -1.4624e-04,  3.7345e-04,
        -4.7008e-04, -3.3905e-04,  3.8811e-04,  2.8712e-04, -1.1535e-04,
         4.2661e-02,  1.5624e-04,  9.4929e-01,  3.1702e-04,  3.9900e-04,
         3.6600e-04, -9.0968e-05, -1.1908e-05,  3.5637e-04, -1.4193e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  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 #1600: [tensor([ 6.0512e-04, -3.8116e-04, -8.5344e-05,  4.5954e-04,  2.3638e-04,
         3.8236e-04,  4.3772e-06,  9.0915e-05,  9.2354e-05, -1.7644e-04,
         3.9322e-04,  1.8815e-04, -6.4404e-01, -3.7216e-04, -3.5114e-01,
         2.3796e-04,  6.6121e-04, -4.5920e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0893e-04,  3.7057e-04,  3.5349e-04,  4.8349e-04,  2.3920e-04,
         3.9396e-05,  1.6502e-04,  5.1112e-04,  3.6284e-04,  5.1686e-04,
        -1.1420e-04,  1.1976e-04, -3.0292e-01,  3.2603e-04, -3.8248e-01,
        -4.3934e-04, -3.0911e-01,  3.8502e-04, -6.2802e-04, -3.2427e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3668e-04,  1.6222e-04,  4.3028e-05,  4.0960e-04,  1.0258e-04,
        -3.6763e-04,  7.7321e-04, -3.6342e-05, -1.3035e-03,  2.6073e-05,
        -9.9132e-01,  1.3079e-03,  1.9784e-03,  1.0402e-03, -5.9082e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7538e-04, -2.3665e-04,  3.0909e-04,  3.1313e-04, -1.3539e-04,
        -1.4158e-04, -2.2536e-04, -5.1065e-04,  2.1563e-05,  1.8893e-04,
        -9.9490e-01, -1.2713e-04,  6.3285e-04,  1.7865e-04,  1.1025e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9380e-04,  6.7060e-04, -3.7663e-04, -2.6015e-04,  1.6564e-04,
         1.8686e-04,  8.2450e-04,  5.2995e-04,  5.6613e-04, -1.7839e-04,
        -9.3708e-01,  1.1612e-03, -2.6734e-03, -5.4375e-02,  2.6329e-04,
        -9.8629e-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([-6.2726e-05, -4.5946e-05,  4.8271e-04,  9.6507e-05,  7.7575e-05,
         5.8506e-04, -3.0223e-04,  6.4668e-05, -4.9104e-05,  1.6018e-04,
        -2.8375e-04,  1.1427e-04,  1.1278e-04,  1.0959e-04,  4.4954e-05,
         4.1118e-01, -1.3097e-04,  2.9656e-04,  4.7349e-02,  1.1468e-04,
        -4.7876e-05, -2.1973e-03,  2.3995e-04,  1.4624e-03,  5.3367e-01,
         4.3255e-04, -2.0000e-04,  9.0659e-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([ 1.4719e-03, -1.3112e-05,  1.8659e-04,  2.0746e-04,  1.6831e-04,
         7.1072e-07,  2.2403e-04,  1.9611e-04,  2.8241e-04,  1.1715e-04,
        -2.2772e-04,  3.5525e-05,  8.2968e-05, -6.6808e-05,  2.2107e-04,
        -2.3831e-05,  2.3450e-04, -4.9257e-05,  2.8577e-04, -4.4733e-01,
        -3.2546e-05,  1.9937e-04, -1.5138e-02,  3.6585e-04, -1.1895e-01,
         2.6305e-04, -2.1137e-01,  2.7996e-04, -2.3946e-03, -1.9870e-01,
         3.4966e-04,  1.5720e-04, -2.4202e-04,  1.3403e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1131e-04,  5.5205e-05, -8.1125e-07, -2.1493e-04, -4.6051e-06,
        -2.6246e-04,  6.8565e-05,  4.3532e-05, -1.7574e-04, -9.2307e-05,
         2.6186e-04,  5.4290e-07, -1.7925e-04,  5.1642e-04, -3.1299e-05,
        -5.4863e-01, -3.0822e-05, -5.3315e-02, -1.9100e-04,  2.2894e-04,
        -2.9801e-01,  1.0695e-04, -9.6897e-02,  2.0596e-04,  1.3078e-04,
         1.2898e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4892e-03,  5.3774e-04,  3.9728e-04, -5.3471e-04,  2.5568e-04,
         3.7084e-04,  2.0447e-04,  2.9797e-04, -2.5255e-04, -6.3402e-01,
        -5.6853e-05, -3.4864e-01, -2.2669e-04, -1.1772e-02, -6.8390e-04,
        -4.6102e-05,  1.1195e-04,  1.0483e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7444e-03,  9.2168e-04, -5.4811e-05,  3.4344e-04, -1.1659e-04,
        -6.6216e-04,  6.1839e-04,  8.7572e-04, -8.3370e-05, -1.0140e-03,
        -9.8938e-01, -9.8830e-04,  1.2535e-03,  1.5072e-03,  4.4096e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4089e-03,  5.4777e-04, -1.8852e-04,  2.6400e-04,  7.0661e-05,
         5.2207e-05,  2.1002e-04,  1.1506e-04, -2.0851e-04,  8.8826e-02,
         4.8514e-04, -6.1794e-01,  6.4775e-04, -2.8860e-01,  2.0101e-04,
         7.7379e-05,  1.5773e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.8809e-03, -1.9157e-04, -8.4074e-04,  1.0169e-04, -3.2205e-04,
         6.0994e-04, -1.8954e-04,  7.8522e-05, -1.3152e-04, -3.2281e-04,
        -3.3984e-02, -6.2987e-04, -3.7758e-04, -5.8465e-04, -9.5696e-01,
        -1.3604e-04,  3.5405e-04,  3.0483e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  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 #1650: [tensor([ 1.0127e-03, -3.2528e-05, -1.4741e-04,  4.0439e-04,  1.6583e-04,
        -1.8291e-04,  1.7632e-04,  5.8879e-04, -3.2003e-04,  1.0533e-04,
         7.8329e-05, -5.7829e-01,  1.9928e-04, -4.3116e-05,  9.1594e-05,
         4.6954e-04, -4.1705e-01,  2.8153e-04,  2.9146e-04, -7.4301e-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([ 6.9395e-04,  1.4607e-04, -8.9284e-04, -1.7917e-04, -4.9647e-04,
        -1.9808e-04, -1.7941e-04,  2.5346e-04,  3.1982e-04, -1.4742e-04,
        -5.3522e-05, -3.9439e-05,  6.4359e-04, -2.1996e-04, -1.1417e-04,
         2.2296e-04,  9.9349e-01, -1.0085e-03,  6.9720e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.5195e-05, -6.5754e-04,  2.7736e-04, -2.5038e-04,  3.2391e-04,
        -4.2520e-04, -2.7629e-04, -2.2893e-04,  1.7788e-04, -3.0594e-04,
        -7.7759e-05, -3.2002e-04, -5.8556e-04,  1.1697e-04,  6.0576e-04,
         9.9387e-01,  4.4688e-04,  3.0326e-06,  4.7241e-04,  4.8886e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8208e-04,  1.9930e-05, -6.9121e-05, -1.6964e-04,  8.0332e-04,
         1.3827e-04,  3.1708e-05, -3.6918e-05,  8.3572e-04, -4.3347e-04,
         4.5567e-04,  3.8795e-05, -1.0904e-04, -5.8796e-04,  3.7153e-04,
         9.9412e-01,  6.1227e-04,  3.0970e-04,  5.7111e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.6172e-04, -6.0211e-05, -3.2286e-06,  9.9195e-05, -1.6903e-04,
        -4.5544e-05,  2.0057e-04, -1.2717e-04,  2.6279e-04, -1.9538e-04,
        -3.5779e-04, -3.9771e-04, -1.2203e-04, -2.4368e-04,  1.3202e-04,
        -4.2183e-04, -3.1976e-04,  3.2949e-04, -1.1045e-04,  8.9045e-01,
         7.0251e-05, -2.7149e-02, -2.5495e-04, -5.9361e-04, -4.6049e-04,
        -7.5764e-02,  1.2636e-04,  3.6792e-04, -4.9898e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.9111e-04, -1.1778e-04,  2.7619e-04,  6.7449e-04, -3.1062e-04,
        -8.8724e-05,  4.1307e-04,  6.7592e-06,  2.9074e-04,  5.8163e-05,
         4.9199e-04,  2.7050e-05,  2.0809e-04,  1.0482e-03, -1.9673e-04,
         3.1111e-04, -7.3124e-05,  3.6235e-04,  5.4582e-04,  9.3023e-01,
         6.4811e-04,  6.2157e-02, -4.7710e-04, -5.0511e-05, -3.4951e-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([ 4.4898e-04, -3.3175e-04,  6.4999e-05,  1.0488e-05, -5.5912e-05,
        -5.9075e-05, -1.0805e-05, -3.5052e-04,  2.0906e-04, -2.9653e-04,
        -1.6244e-04,  4.0818e-06,  1.2920e-04, -1.9391e-04, -4.9166e-05,
        -1.5713e-04, -1.6236e-04, -9.6281e-05, -1.4122e-04,  6.1865e-01,
         7.8839e-06,  3.6938e-01,  6.4696e-05, -1.9725e-03,  2.9076e-04,
        -4.4440e-05, -5.3948e-03,  2.1958e-04, -7.4093e-04, -1.3331e-04,
        -1.7555e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3693e-04,  3.6301e-04,  5.9750e-04,  1.3495e-04, -9.9811e-07,
         2.1112e-04, -3.0824e-05,  4.9335e-04,  2.2575e-04,  1.0638e-04,
         5.8975e-04,  4.7966e-04,  2.2462e-04,  1.1337e-04,  8.2913e-01,
        -1.9275e-04,  4.0790e-04, -9.3791e-05,  1.6570e-01, -3.7095e-04,
        -2.9509e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4655e-03,  2.7974e-04, -5.9425e-04,  8.0105e-04,  5.0252e-04,
        -8.0047e-05,  2.2949e-04,  6.5407e-04,  9.1654e-04,  2.1677e-04,
        -1.8633e-04, -3.8352e-04,  4.4959e-04, -4.8538e-04,  9.9026e-01,
        -4.3215e-04,  3.8472e-04,  1.0431e-03, -2.3653e-04,  3.9829e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4724e-03,  2.7582e-04,  5.0592e-04, -5.3751e-04, -6.8663e-04,
         2.6323e-04,  1.8524e-04, -3.9075e-04,  6.4450e-05, -3.8594e-04,
         4.9409e-04,  1.4622e-04, -9.7309e-05,  1.5496e-04, -9.9037e-01,
         9.7883e-04, -5.0302e-04, -5.7438e-04, -1.7977e-04, -1.4532e-04,
        -5.8876e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2208e-04,  4.3146e-05, -4.5310e-04,  1.0757e-04, -3.3259e-05,
         3.1737e-05,  2.5460e-04,  2.4103e-06, -9.4018e-05,  7.2578e-05,
         2.9371e-04, -7.1380e-05,  4.3195e-04,  9.5579e-01,  6.3148e-04,
        -1.1407e-02,  7.2287e-05, -3.1250e-04,  2.8888e-02, -1.0761e-04,
        -2.0875e-04,  4.7412e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4930e-03,  2.3953e-04,  5.0352e-04, -2.0474e-04,  1.6136e-04,
         1.2911e-04, -9.1606e-05,  3.0377e-04,  1.7726e-04, -1.7120e-04,
         2.4053e-04,  9.2783e-05, -1.9063e-04, -5.3109e-01,  4.4978e-05,
        -5.1063e-02,  1.1628e-04, -4.1275e-01,  5.2143e-05, -4.8486e-05,
        -8.2862e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         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 #1700: [tensor([-8.1394e-04, -1.3894e-04, -8.2885e-05, -8.6027e-05, -2.7876e-04,
         1.8621e-04,  1.9504e-04, -1.6776e-04, -1.5187e-04, -3.7048e-04,
        -1.0440e-05, -4.6355e-01,  1.1568e-04,  3.0871e-04, -3.1594e-01,
         6.0800e-06,  1.4996e-02,  2.6851e-04, -4.1113e-03, -1.9722e-01,
        -4.1194e-04,  2.4725e-04,  3.4712e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7475e-04, -2.0338e-04, -4.2848e-04, -5.4466e-05,  3.5498e-05,
         4.1732e-04,  2.3300e-05, -5.4380e-04,  4.0864e-04, -1.3910e-05,
        -2.1233e-06,  2.5239e-04, -3.2510e-03,  3.4558e-04,  4.5862e-01,
         3.1782e-04,  5.3404e-01, -3.9206e-04, -2.4702e-04, -3.0914e-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([-4.7457e-04, -2.5694e-04, -8.6374e-05,  3.4244e-04,  2.9701e-04,
         3.3349e-04, -1.6593e-04, -7.2352e-05,  2.5767e-04,  7.4389e-04,
        -2.6295e-04,  1.8244e-05, -1.5771e-04,  9.9542e-01, -3.5352e-04,
         3.6146e-04,  3.9859e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3574e-04,  6.7992e-05, -1.0988e-04,  2.5074e-04, -1.8827e-04,
        -1.4574e-05,  7.1833e-05,  1.7613e-04, -3.0336e-04, -5.1427e-05,
         6.6371e-05,  3.9708e-04,  4.5047e-04, -1.8024e-03, -3.4895e-01,
         4.0114e-04, -6.7121e-04, -6.4457e-01,  3.3951e-04,  6.3793e-05,
         5.2197e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9694e-04,  3.0083e-04,  1.7862e-04, -1.2065e-04,  2.6356e-04,
        -1.2444e-06,  1.3574e-05,  7.8169e-05,  8.0127e-05,  2.3192e-04,
        -4.9820e-04,  1.0910e-03,  4.9660e-04,  1.6327e-04,  2.1621e-02,
         2.4048e-04,  9.7214e-01,  1.1741e-03, -2.8781e-04,  3.6878e-04,
         2.5124e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2350e-04, -3.7501e-04, -1.4351e-04,  4.0018e-04,  1.7041e-04,
         7.7632e-05, -4.9037e-04,  7.9762e-05, -3.2133e-05, -3.5133e-05,
         1.1258e-04, -4.0138e-05,  3.0205e-04,  8.0195e-02,  9.1637e-01,
        -1.1897e-04,  5.1425e-04, -2.2443e-04, -1.9807e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.2432e-04, -2.7243e-04,  1.8244e-05,  1.5414e-04, -3.0052e-04,
        -7.9814e-05, -1.4467e-04,  6.3740e-05, -9.9237e-05, -9.6976e-04,
         9.9351e-05, -1.3517e-03,  2.7372e-04, -8.9471e-02, -9.0383e-01,
        -9.1576e-04, -6.4840e-04, -4.8666e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0231e-06, -4.6381e-05, -2.4796e-04,  1.0165e-03, -1.1455e-03,
         3.4057e-05,  4.7161e-04, -3.0332e-04, -6.7776e-05, -1.7475e-04,
        -2.9122e-05, -7.0311e-04, -2.9404e-04, -7.8012e-04, -6.4769e-04,
        -2.3790e-04,  1.7157e-04,  4.7515e-04, -9.9256e-01, -1.5164e-04,
        -6.9730e-05,  3.6000e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0182e-04,  7.8947e-05, -5.9959e-04,  6.2642e-04,  4.6778e-04,
         3.9143e-04,  1.1589e-04,  1.2604e-04,  1.9669e-04,  7.8525e-05,
        -1.2131e-04,  1.7519e-04, -1.3537e-04,  5.9658e-04, -4.7485e-05,
        -1.4199e-04,  3.5272e-04, -5.1435e-04,  1.8277e-01, -9.8427e-05,
         8.1117e-01, -3.1015e-04,  3.6829e-05,  5.4410e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.5282e-04,  6.5992e-04, -2.8156e-04,  5.6897e-04, -1.8437e-04,
         1.4600e-04, -1.5026e-05,  4.5093e-04, -1.8850e-05, -5.4420e-06,
         2.3038e-04,  3.1065e-04, -2.8182e-04,  4.1342e-05,  3.8076e-04,
        -1.1020e-04,  1.4556e-04,  1.4765e-04,  4.1563e-01,  4.3837e-06,
         5.7872e-01, -2.2770e-04, -4.9649e-04, -9.6112e-05,  1.9690e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6411e-04, -1.1195e-04, -5.5407e-05, -4.8052e-05,  1.0144e-04,
         8.7856e-06, -1.0132e-04,  3.3659e-05, -8.0052e-05, -1.8899e-05,
         7.8526e-06, -1.8191e-04, -8.3711e-05, -2.3538e-05,  2.9914e-05,
         4.9435e-05, -8.3521e-05, -1.3224e-04, -5.9569e-05, -7.5262e-05,
        -9.3321e-05,  1.1970e-04, -1.6295e-04, -2.1142e-04, -2.6252e-04,
         1.2510e-04, -3.2506e-05,  1.1535e-04, -2.1623e-05,  1.9877e-02,
        -8.5204e-05,  2.9439e-04,  1.4149e-04,  6.3581e-03,  4.9235e-04,
        -1.8459e-04,  3.0692e-04,  1.5032e-04, -9.5700e-01,  3.3871e-04,
        -8.6033e-05,  1.2488e-05,  3.2789e-04,  3.4992e-04,  3.1070e-04,
         3.0152e-05,  1.0479e-02,  3.1080e-05,  3.4008e-04,  6.3983e-05,
        -3.9832e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.0119e-04, -1.2886e-04, -8.6205e-05,  2.5543e-06, -2.4326e-05,
        -2.6582e-05,  9.8687e-05,  4.7230e-05,  3.4149e-05, -2.6239e-05,
        -8.8356e-05, -1.3920e-05,  2.5689e-05,  1.3923e-04, -1.3670e-04,
        -7.6130e-05, -6.9652e-05,  2.4899e-04, -1.2861e-04,  7.2964e-05,
        -6.9195e-05, -1.7670e-04, -1.9393e-04, -2.1037e-04,  1.2350e-05,
        -9.2205e-05, -1.9978e-04,  8.5498e-05, -1.7461e-04,  8.9235e-01,
        -2.0438e-04, -7.4944e-04,  3.8124e-04, -9.0514e-05,  3.1015e-02,
        -1.6666e-03,  4.4803e-04, -8.6993e-06,  6.6627e-03,  2.5331e-04,
         5.1062e-06, -1.6243e-03,  8.4181e-05, -8.7338e-06,  6.0228e-02,
         3.5935e-04,  1.0408e-04,  1.1548e-05, -1.4945e-04,  2.0178e-04,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1750: [tensor([ 6.9027e-05, -3.8698e-05,  2.0808e-04,  1.8363e-04,  1.6647e-04,
        -1.5660e-04, -1.4593e-04,  2.4936e-04,  3.0922e-04, -1.4162e-04,
        -1.4855e-04, -1.2910e-04, -1.8970e-04,  2.8177e-04,  1.4508e-05,
         2.4318e-04,  1.2954e-04,  2.5427e-04, -2.3094e-04, -1.1254e-04,
         2.4034e-04,  4.6683e-01, -1.8654e-04, -5.3405e-03,  5.2334e-01,
        -5.4140e-04, -1.2237e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2588e-04,  2.3724e-04,  2.0038e-04,  3.2343e-05,  2.9957e-05,
         3.1075e-04,  1.2778e-04, -2.2705e-04, -1.6410e-04, -5.0540e-04,
        -9.2942e-05,  6.2932e-05,  1.7944e-04,  2.2246e-04, -6.8771e-05,
         2.7336e-04, -2.3278e-04,  1.8855e-04, -3.5911e-05,  9.2179e-05,
         8.6590e-05,  2.5959e-01,  1.7991e-04, -2.9411e-04,  7.2870e-04,
         2.4970e-04,  7.3398e-01, -5.1193e-04,  3.9171e-04, -1.7848e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.6177e-04,  1.0781e-04,  2.3055e-04,  8.4721e-05,  6.6803e-06,
        -5.9391e-05, -8.0930e-05,  2.1343e-04,  2.1144e-04, -4.8753e-04,
        -1.6873e-04, -2.5804e-04, -1.5531e-04,  9.4874e-06,  5.1727e-06,
         1.0859e-04,  3.3396e-04, -3.3903e-05,  2.7562e-04,  8.0062e-05,
         9.7797e-05,  8.4551e-02,  3.9035e-04, -4.4215e-05,  8.6829e-04,
         9.0993e-01,  3.1034e-04,  3.8511e-05,  9.8164e-05,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5344e-04, -1.3077e-04, -2.0698e-04,  3.5027e-05,  2.0327e-04,
        -2.5530e-04,  6.8886e-05,  1.0992e-04, -5.9404e-05, -4.7058e-04,
        -3.5927e-04, -1.1984e-04,  4.4263e-04,  1.7637e-04,  4.4028e-05,
         2.7027e-04,  2.1215e-04,  1.9507e-04,  2.4470e-04, -1.0793e-04,
         1.6001e-04,  8.6149e-01, -5.0368e-04, -8.3830e-03,  4.6969e-05,
         1.2455e-01,  5.9766e-04,  2.2861e-04,  1.8275e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2981e-04, -1.7012e-04, -4.4336e-06, -2.1616e-04,  1.0264e-04,
         3.2974e-06, -3.9704e-05,  6.0926e-05, -1.3424e-04, -2.3597e-04,
        -1.9595e-04, -2.3669e-04, -2.5439e-04, -1.1724e-04, -1.1122e-04,
         7.2439e-05,  1.3042e-04, -1.8938e-04,  5.3748e-05,  2.9752e-04,
        -3.4522e-04, -2.4474e-02, -1.4351e-04, -5.5668e-01, -2.3473e-04,
         4.1428e-05, -4.1503e-01, -5.5971e-05,  2.3600e-04,  5.9364e-06],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0005e-04,  5.7401e-05,  4.6721e-05, -2.4756e-04, -5.2319e-05,
         3.0531e-04, -2.4602e-04,  4.3769e-05,  5.8747e-05, -2.2697e-04,
        -1.1821e-05,  1.1706e-04,  1.6014e-04,  1.6707e-04,  3.8858e-04,
         1.6861e-05, -2.7409e-04,  8.7437e-05, -2.0899e-04,  1.9939e-04,
         1.8103e-04,  7.4116e-01,  4.0503e-04,  9.2315e-03, -5.4294e-05,
         2.4406e-01,  3.1569e-04,  6.9362e-04,  4.8215e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1189e-04, -1.0907e-04, -2.7246e-04,  2.3876e-04, -7.3392e-04,
         4.3289e-06,  5.4692e-05, -4.9750e-04,  1.6263e-04, -5.2054e-04,
         3.9373e-04,  4.3733e-04, -5.5672e-05,  2.1977e-03,  4.6682e-04,
         9.8767e-01,  2.6034e-04,  3.1122e-04,  2.7018e-03,  8.5203e-04,
         1.1292e-03,  3.1653e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.9192e-04,  3.8612e-04,  3.2415e-04, -8.3578e-04, -8.3063e-05,
        -5.2352e-04, -1.9945e-04,  3.2196e-04, -4.3692e-04,  7.8329e-05,
        -7.6596e-04, -1.7589e-04, -4.9192e-05,  1.6657e-02, -1.2947e-03,
        -9.7608e-01, -1.3410e-04, -3.6692e-04,  4.1634e-04, -2.7988e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3310e-04, -3.3873e-04, -1.5799e-04, -1.3081e-04,  2.1896e-04,
        -4.8744e-04,  1.7442e-04,  3.4304e-05, -1.7948e-04, -2.2322e-04,
        -2.0077e-04,  1.3547e-04, -7.7195e-05, -2.6926e-04,  2.4301e-04,
        -4.2981e-05,  3.8478e-02,  1.1276e-01, -5.2964e-04,  2.8288e-02,
         8.1628e-01, -4.4604e-05, -2.6412e-04,  1.1448e-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([-9.7562e-05,  4.5601e-04, -1.2236e-04,  3.1970e-04,  3.5661e-04,
        -5.5669e-04,  4.9477e-04,  8.1949e-05,  1.1245e-04,  8.2603e-05,
         3.5086e-04,  8.1785e-01,  4.3506e-04,  3.0058e-04,  1.4880e-01,
         4.8224e-04,  3.0975e-04,  7.7718e-03,  1.2623e-04, -1.9607e-02,
         6.1713e-04,  3.5009e-04,  3.2018e-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([ 5.7791e-04, -1.0397e-04,  2.4494e-04,  4.8688e-04, -8.3867e-05,
        -2.0478e-04,  5.7045e-05, -1.5346e-04, -2.6093e-04, -2.0319e-04,
         7.7830e-05,  2.8000e-01,  1.9374e-04,  9.0662e-04,  2.1886e-04,
        -4.4297e-04,  7.1541e-01, -1.3015e-04,  8.4560e-05, -1.5293e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1408e-03, -2.6824e-05,  1.1395e-04, -6.7538e-04,  2.3681e-04,
         3.3275e-04,  3.4492e-04, -1.2033e-05, -4.2709e-04,  6.2544e-05,
        -4.4089e-04, -9.4411e-01, -3.3478e-04, -3.3107e-04,  4.2882e-04,
         4.5488e-02, -2.1630e-04,  8.2400e-04, -8.9870e-04, -1.9140e-04,
        -3.6693e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  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 #1800: [tensor([ 6.0212e-04, -7.4663e-05,  6.8040e-05,  7.5326e-05,  4.9843e-04,
        -5.3246e-05, -9.0205e-05, -5.5872e-04, -1.8354e-04, -4.3710e-05,
        -2.4824e-04,  2.2255e-04,  5.9032e-02,  1.2141e-04, -3.4710e-05,
         1.4572e-04, -3.8519e-03, -1.3781e-04, -1.0694e-03,  9.3253e-01,
        -2.3343e-05, -1.9424e-04,  1.4185e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1309e-03,  2.3928e-05, -3.0432e-04,  4.4434e-04, -3.0037e-04,
        -5.3118e-04, -3.7191e-04, -1.2145e-04,  6.0253e-05, -1.7202e-04,
        -1.2521e-04, -1.8104e-04, -9.7353e-01,  9.5952e-05, -1.1262e-04,
         2.7644e-04,  3.6031e-04, -2.5319e-03, -5.3462e-04, -3.3981e-04,
        -1.6428e-02,  4.5602e-04, -4.2691e-04,  1.4217e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3296e-04, -1.5676e-04,  3.3892e-05, -1.3530e-04,  1.1503e-04,
        -3.8734e-04, -2.2146e-04,  3.2526e-04, -2.2319e-04, -2.1392e-04,
        -1.4397e-04, -9.4118e-05, -6.2074e-01,  3.5268e-04,  1.5135e-05,
         3.3646e-04,  8.4420e-04,  2.6654e-04, -1.6418e-02, -3.5774e-01,
         4.1485e-04,  4.4785e-04, -1.4607e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6471e-03, -4.6476e-05, -1.8525e-04, -1.0184e-04,  8.5867e-05,
        -8.8971e-05,  1.7314e-04,  1.4878e-05,  2.6317e-04, -2.1108e-05,
         2.1446e-04,  1.1152e-04, -4.0662e-04,  3.1676e-04,  1.1112e-04,
        -1.6195e-04, -2.5034e-04, -4.8108e-01, -1.2029e-04, -5.0242e-01,
        -4.3759e-04, -1.9930e-04, -1.2746e-04,  1.0130e-02,  1.1533e-04,
        -3.2392e-04, -3.3986e-04, -5.0756e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1350e-04, -3.1622e-04, -1.3964e-04, -1.0443e-04, -1.8058e-04,
        -4.3627e-04, -2.4171e-04, -1.6795e-04, -2.5295e-05,  1.0120e-04,
         1.7423e-04, -4.1375e-04, -8.1560e-05,  7.3499e-05,  1.2514e-04,
         2.1455e-04,  1.5994e-04,  6.0062e-01,  1.1200e-05,  3.6807e-01,
         4.5216e-04, -2.7061e-02, -1.8860e-04, -1.5194e-05,  4.1314e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1561e-04,  2.9213e-04, -7.3177e-05,  3.7439e-04,  2.5047e-04,
        -6.0844e-06,  2.7079e-04, -1.6967e-04,  2.3326e-04, -1.1625e-04,
        -2.2088e-04,  5.5452e-04, -1.7867e-05,  2.3320e-04,  1.4025e-04,
         2.0542e-04, -8.7466e-05, -5.5765e-01,  4.0420e-06, -4.3803e-01,
         3.4949e-04, -1.6040e-04,  1.9705e-05,  1.8739e-04, -2.4236e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4911e-04, -2.4661e-04, -2.0479e-04,  5.5796e-04, -3.2454e-04,
        -3.8784e-04, -2.9905e-04,  5.2184e-05, -3.5879e-04, -4.9702e-04,
        -4.9741e-04, -9.9499e-01, -4.0987e-04, -9.6384e-05, -1.4854e-04,
         5.3708e-05, -6.2306e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9153e-05, -1.9959e-04, -3.6156e-04,  1.1461e-03, -5.8236e-04,
         4.2966e-04, -3.9101e-04,  1.8653e-04, -3.4740e-05, -1.3640e-03,
        -1.1192e-03,  2.8931e-03, -9.9079e-01, -9.3698e-05, -2.4477e-04,
        -1.1017e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0071e-02, -7.3304e-03, -4.6268e-03, -5.6030e-04,  3.9102e-03,
        -5.9080e-03, -1.0370e-02,  3.7975e-03, -4.3590e-03, -8.7155e-03,
         8.9144e-01,  5.2261e-03, -2.3681e-02,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4413e-04,  8.5466e-05, -8.2902e-05,  1.6639e-04,  5.0762e-05,
        -6.7250e-05,  5.7662e-05,  2.1684e-04,  8.8149e-05, -1.6852e-04,
         2.4289e-04, -4.7258e-05, -6.1778e-05, -2.6640e-04, -3.1601e-05,
         4.4362e-05, -2.1363e-05,  1.4459e-04, -1.2405e-04,  1.1656e-04,
         2.6834e-05,  2.5024e-05,  1.1921e-04,  1.6540e-04,  4.4373e-06,
        -4.6053e-05, -6.9701e-05, -1.1686e-04,  6.2237e-05,  4.1449e-05,
        -1.8137e-04,  1.3350e-04, -9.0063e-05,  3.2924e-04,  1.0220e-04,
        -1.9065e-05,  2.3653e-05, -1.4823e-01,  1.8514e-05,  2.5967e-04,
        -2.6840e-01,  1.2346e-04, -5.4162e-06,  2.4654e-02, -5.5405e-01,
         2.1812e-04,  1.6321e-04, -9.2959e-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([-5.6133e-04,  1.7767e-05, -1.1232e-04,  7.4513e-05,  5.5861e-05,
         4.6566e-05,  3.9141e-06,  1.1286e-04, -7.2851e-05,  2.0689e-04,
        -1.4450e-04, -2.1862e-04,  1.5695e-04, -1.6419e-04,  1.7451e-04,
         5.8998e-05, -4.7122e-05,  9.4286e-05, -1.4417e-04, -1.7576e-04,
        -5.1385e-05, -2.2138e-04, -1.7496e-04,  1.5050e-04, -1.0626e-06,
        -3.5382e-05, -1.2219e-04,  1.8974e-04,  2.0010e-05,  9.4749e-05,
         1.3299e-04, -3.1910e-05,  1.4484e-04, -1.7234e-04,  1.1283e-04,
         1.1872e-04, -2.6326e-02,  3.1610e-04,  3.1311e-01, -6.6596e-04,
         6.5418e-01,  1.2070e-04,  3.0741e-04,  3.4208e-04,  2.1590e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.5730e-04, -1.1646e-04, -2.2616e-05, -5.3753e-05,  2.4454e-04,
         2.0932e-04, -3.5345e-05, -2.0117e-05, -7.3733e-05, -8.2666e-05,
        -2.9325e-04,  9.3242e-05,  7.4038e-05, -3.5207e-05, -3.9335e-05,
         4.3765e-05, -8.0622e-05, -1.2890e-04,  3.1382e-05,  1.2054e-05,
        -3.3273e-04,  3.5566e-05, -2.1550e-04,  2.4928e-04, -4.0189e-05,
         1.7820e-04,  6.6539e-05,  1.9692e-05, -1.1436e-05, -5.9987e-05,
         6.0190e-06,  7.7569e-06, -6.3462e-05, -2.0522e-04,  1.2454e-04,
        -1.4505e-05,  9.0290e-01, -1.8733e-04, -1.1298e-03,  8.0510e-05,
         1.6086e-04,  2.4302e-04, -3.0922e-02, -1.4262e-04,  6.8320e-05,
        -1.4803e-02, -1.6533e-03, -2.8236e-05,  1.5283e-04, -5.1893e-04,
         2.7127e-03, -3.9645e-02,  1.7211e-04,  1.3022e-04, -1.7332e-04],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1850: [tensor([-1.6696e-03, -1.0029e-04,  2.7779e-05,  1.1515e-04,  2.0718e-04,
         9.9916e-05, -5.2863e-04,  4.0239e-04, -4.0050e-04, -1.1343e-04,
         2.3442e-04, -2.3964e-04, -2.2294e-01, -5.0006e-04,  2.0827e-04,
        -7.6981e-01,  3.1443e-04, -1.2826e-04, -2.3045e-04,  6.4443e-04,
        -8.4331e-04,  1.0274e-04,  9.1567e-05,  5.2807e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9619e-06, -4.6824e-04, -2.7817e-04,  2.5008e-04, -1.9131e-04,
         1.3210e-04,  8.7835e-05, -5.1651e-04,  6.1932e-05, -8.3533e-04,
        -1.5357e-04, -1.0272e-04, -8.4837e-01, -3.4903e-04, -1.2547e-01,
        -4.9331e-04,  2.0930e-02, -9.0983e-04,  3.1145e-04, -8.0002e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3792e-03, -2.5450e-05,  3.6812e-04, -1.4849e-04,  5.2749e-05,
        -6.9398e-05, -1.8862e-04, -2.5437e-04,  1.4106e-04,  3.5528e-04,
        -2.2802e-04,  2.5095e-05, -9.6032e-05,  2.3979e-04,  3.7620e-04,
         1.3220e-04,  8.4983e-01,  5.8650e-04, -2.0082e-04, -1.2416e-03,
         1.4323e-01,  2.3726e-04,  1.3955e-04,  4.5902e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.3376e-04, -3.0623e-05, -1.2088e-04,  2.4091e-04,  4.6678e-04,
         3.1760e-04, -2.3370e-04,  8.4449e-06,  1.0406e-05, -4.0937e-06,
        -2.3534e-04,  5.0718e-05, -2.8276e-05, -6.8566e-05,  4.1438e-04,
         3.6727e-05, -1.6724e-01,  1.8095e-04, -6.9003e-04, -3.3335e-03,
        -1.9843e-04, -4.4426e-03,  3.7468e-04, -1.7534e-04, -1.9290e-04,
        -6.1714e-02,  7.5819e-01, -1.6771e-04,  2.8928e-04, -1.6121e-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([-4.6570e-04, -2.4108e-04, -3.2473e-04,  3.6735e-04,  6.6748e-05,
         1.5040e-04,  2.8805e-05, -9.8009e-05, -1.4593e-04, -9.0801e-05,
        -1.0097e-04, -1.5341e-04, -1.9472e-04,  3.1742e-04,  2.8705e-05,
         4.1425e-05,  4.1445e-01, -1.2800e-04, -2.7150e-04, -2.3011e-05,
         2.7992e-04,  5.8164e-01,  2.5447e-04,  4.8035e-05, -9.8375e-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([-1.5634e-03,  4.0838e-07,  5.1924e-04, -6.7224e-04,  4.6309e-04,
         3.3792e-04,  1.9868e-04, -3.0561e-04,  1.2210e-04,  4.6059e-04,
         5.6133e-04,  1.6996e-04,  4.6041e-04,  4.2565e-04, -6.0189e-04,
         3.1621e-04,  5.8971e-04, -9.8736e-01,  2.8633e-03,  9.6526e-04,
         8.6787e-04, -4.3566e-05,  1.3615e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2377e-04, -1.7104e-04, -1.3854e-04, -3.8001e-04, -4.0785e-04,
        -3.1592e-04,  2.0290e-05, -4.8578e-04, -2.7814e-04, -2.6174e-04,
         2.7928e-04,  3.7570e-04, -1.1004e-04,  8.6722e-05, -2.8279e-05,
        -4.4155e-04, -2.5068e-04, -4.5991e-04,  9.9435e-01,  6.6936e-05,
        -3.0674e-04, -5.5780e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6867e-04,  3.3138e-04,  1.0977e-04, -3.4179e-04,  3.0969e-04,
        -1.4704e-04, -4.7123e-04, -3.8587e-04,  5.6760e-04,  2.9971e-04,
        -6.3674e-04, -2.6013e-04, -2.7902e-04, -2.7224e-04,  3.8572e-05,
         4.2137e-05, -2.3046e-04,  9.9277e-01, -1.5673e-04,  5.0588e-04,
         1.0261e-03,  5.5300e-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,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6995e-03, -2.0159e-04, -1.6633e-04,  8.2805e-06, -6.0687e-05,
        -9.2440e-05, -1.4744e-05,  1.7080e-05,  1.5763e-04, -6.7136e-05,
        -1.3351e-04,  3.8261e-05,  4.5778e-05,  1.9282e-04,  4.7996e-05,
        -1.4592e-05, -4.0246e-05,  1.3850e-04, -1.6121e-05,  5.9331e-05,
         6.6211e-05,  7.8950e-05,  1.4566e-04,  6.6951e-05, -8.6857e-05,
         6.6119e-01, -1.4169e-04,  1.0152e-01, -3.1547e-05, -3.6127e-03,
         2.2355e-01, -3.3761e-04,  1.8719e-03,  8.1336e-05, -2.4079e-03,
        -7.2502e-04, -5.2803e-04,  2.1888e-04, -1.3048e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1323e-04,  1.1977e-04, -5.7158e-05, -6.4345e-05, -1.3625e-05,
        -3.0968e-04,  3.4843e-05, -3.3239e-04, -3.0059e-04, -1.3781e-05,
        -2.3486e-04,  1.4131e-04, -7.4252e-05, -4.2463e-04, -2.1737e-05,
        -2.5099e-04, -3.7178e-04, -3.6530e-05, -2.2458e-04, -1.0889e-04,
         1.7060e-04, -8.1885e-05,  1.3053e-04, -4.3774e-04,  1.6982e-04,
         7.3594e-02,  3.6598e-04,  1.9716e-01,  8.6943e-04, -4.1142e-01,
        -2.4083e-01, -3.9595e-04,  5.1503e-04,  6.2833e-02,  6.7464e-03,
         3.8355e-04, -2.6372e-04, -3.8735e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1053e-03,  1.8013e-04,  3.0940e-04,  1.2876e-04, -4.3497e-06,
        -8.1164e-05,  4.2328e-05,  4.3561e-05,  2.0668e-04,  7.1549e-05,
        -1.9310e-04, -3.5012e-05, -6.4487e-05,  2.6256e-04,  3.3976e-04,
        -1.7926e-04,  1.5055e-04,  1.1255e-04,  4.2649e-04,  5.8744e-05,
        -1.0960e-04,  2.0954e-04,  2.2080e-04, -1.8563e-04, -7.3819e-05,
        -1.5994e-02,  3.3574e-04, -1.1709e-01, -1.8117e-04, -8.5939e-01,
         4.1872e-04,  3.3128e-04, -3.3062e-04, -1.3713e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.9312e-04, -5.2063e-06,  2.6102e-05, -1.1553e-04,  2.5691e-05,
         2.2371e-04,  3.1073e-05, -3.8863e-05,  1.2900e-04,  3.1416e-01,
         6.8770e-05,  1.9989e-04, -2.0107e-03,  6.8064e-01,  2.4717e-04,
        -1.0079e-04, -8.2878e-04, -1.7622e-04,  4.6698e-04, -1.4342e-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 #1900: [tensor([ 1.0868e-04, -6.6061e-04, -3.5856e-04, -3.3800e-04, -2.9616e-04,
        -1.1240e-04, -3.3058e-04, -1.6378e-04, -2.0230e-04,  3.8078e-04,
        -1.0886e-04,  2.8657e-04, -3.6930e-04, -5.3654e-04, -1.1983e-04,
         3.8382e-04, -1.0794e-03,  9.9309e-01, -3.7499e-04,  1.5370e-04,
        -4.1482e-04,  1.3266e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0250, -0.0055, -0.0025,  0.0020,  0.0076,  0.0094,  0.0156, -0.0056,
         0.0067,  0.0083, -0.0047,  0.0018,  0.0014,  0.1459,  0.0157,  0.1233,
         0.3191,  0.2718,  0.0281,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9813e-03, -1.1457e-04,  7.1058e-05, -5.4635e-06, -2.2491e-04,
        -7.5981e-05, -5.9151e-05, -2.2081e-04,  1.0431e-04, -6.0402e-05,
         3.0739e-05,  2.0315e-04, -3.2773e-05, -4.9149e-01,  2.7770e-05,
        -2.6345e-03,  5.9290e-02, -2.4779e-04, -3.1602e-04, -4.4234e-01,
        -1.6141e-04,  2.1817e-04, -9.1054e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8529e-04,  3.4262e-04, -2.3862e-04, -7.3464e-05, -4.9566e-04,
        -1.7214e-04,  2.4443e-04, -5.7097e-04, -1.1067e-04, -3.6327e-04,
        -3.0787e-04, -6.1417e-04, -1.2952e-04,  6.2670e-04, -2.9925e-04,
         1.0152e-04,  5.1842e-04, -2.9103e-04,  1.0462e-03, -4.0298e-02,
        -9.5197e-01, -1.7036e-04, -6.4743e-04,  7.9810e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.2412e-04, -2.3829e-04,  5.6710e-05,  6.8195e-04, -4.9575e-04,
        -5.2360e-04,  2.4505e-04, -6.3271e-05,  4.9193e-04, -1.6791e-04,
         1.3424e-04, -3.3759e-04, -3.4707e-04, -9.9501e-01, -4.1692e-04,
         2.6956e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2394e-05,  2.2932e-05, -2.3798e-04, -2.8492e-04, -5.6370e-04,
        -3.2086e-04, -2.9465e-04, -4.6028e-04, -6.0042e-05,  9.2493e-05,
        -4.5811e-04, -6.6324e-05,  4.8176e-04,  9.0584e-01, -5.0022e-04,
        -1.0690e-02, -7.9164e-02, -2.5566e-04, -1.6280e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2275e-04,  2.5624e-04,  1.4140e-04,  1.6436e-04, -8.5795e-06,
         8.3399e-05, -3.6565e-04, -1.0357e-04, -2.9950e-04,  1.7821e-04,
         1.2858e-04,  3.2143e-04, -1.2995e-03, -9.1963e-01,  2.2828e-04,
         4.1187e-03, -7.1962e-02, -4.7432e-04, -1.1811e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0026e-04,  1.4801e-04,  3.6521e-04,  4.0561e-04,  4.4577e-04,
        -7.0074e-05,  4.4552e-04,  1.3998e-04,  4.6648e-04, -4.0006e-05,
         3.8363e-04,  5.0478e-05, -1.2515e-04, -1.4126e-04, -8.4713e-01,
        -1.5859e-04,  1.6086e-02, -5.7284e-04,  3.0276e-04, -1.3072e-01,
         4.3613e-04, -3.3817e-04, -1.1433e-04, -5.0790e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2182e-03, -1.9583e-04,  2.3022e-04,  3.0641e-04, -1.7521e-06,
        -9.6141e-05, -4.2555e-04, -7.1714e-05,  5.7396e-04,  5.8183e-05,
         3.8146e-04, -3.2149e-04, -5.1110e-04,  3.5896e-04, -9.7400e-01,
        -5.5540e-05,  2.0723e-02,  2.1750e-04, -1.0481e-04, -1.5062e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9526e-04,  4.4689e-04,  3.0004e-05,  8.0546e-06,  4.6452e-04,
         2.9606e-04, -5.6077e-06,  1.7120e-04, -1.2398e-04,  7.6762e-05,
         1.7432e-04, -1.2731e-04, -4.9716e-05,  6.4104e-04, -4.2879e-01,
         2.0601e-04,  2.7580e-04, -8.5513e-04,  2.0456e-04, -5.6553e-01,
         3.6717e-04,  4.5056e-04,  1.6710e-04,  2.4040e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.6913e-06, -1.7428e-04,  1.4796e-04,  1.7844e-04,  1.3702e-04,
        -1.3888e-04,  8.3210e-05, -2.8570e-04, -4.4037e-05, -8.3210e-05,
        -8.8632e-05,  1.6792e-04,  4.7717e-01, -2.4755e-05,  4.6858e-01,
         1.0928e-04,  5.2222e-02, -2.6301e-05, -4.8053e-05,  2.8049e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4357e-03,  1.6381e-04, -7.6163e-05, -2.0952e-04, -1.1038e-04,
         1.8025e-04,  1.0373e-06, -1.6770e-04, -5.2639e-05,  4.7842e-04,
        -1.9805e-04, -3.2662e-04, -3.9111e-01,  1.2922e-04, -7.7349e-02,
         4.2321e-04, -1.5722e-01,  2.5893e-04, -1.2358e-04, -3.6834e-01,
        -5.2599e-04,  3.0182e-05,  8.1891e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1950: [tensor([-4.5453e-04,  3.4388e-04, -1.2586e-04,  3.3215e-05, -1.2522e-05,
         1.9636e-04,  3.2432e-04,  1.3877e-06,  9.2250e-05,  1.9396e-04,
         1.6959e-04,  1.9778e-04, -1.0024e-04, -1.5447e-01, -4.7128e-04,
        -1.9416e-02, -3.7148e-04,  1.0750e-02, -7.6043e-05, -8.1168e-01,
         1.0097e-04,  1.5282e-04, -2.9613e-05, -1.7576e-04,  6.4621e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.5458e-04, -5.0053e-05, -2.3966e-04, -3.0945e-04,  4.7748e-04,
        -2.0414e-04, -9.0691e-05,  2.7665e-04, -4.0511e-04,  2.5740e-04,
         6.8004e-05, -3.3923e-04,  1.8358e-04, -1.2754e-01, -6.8905e-04,
         8.6527e-01, -3.4342e-04, -6.2029e-04, -1.3594e-03,  4.1522e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2105e-03, -2.9550e-04,  3.4332e-04, -6.8115e-05, -7.2769e-05,
         1.5595e-04, -5.4103e-05,  1.2296e-04, -6.5478e-04,  3.0668e-05,
         1.6400e-04, -7.8274e-01, -8.1036e-05, -8.2802e-05, -2.1369e-01,
        -1.3438e-05, -2.8283e-05,  1.8991e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5305e-03, -1.2823e-04, -4.3607e-05,  8.9123e-05,  1.3596e-05,
         4.0781e-04, -1.2228e-04,  3.4825e-04, -3.3114e-04,  1.0411e-04,
         2.1867e-04, -3.7579e-01,  2.7635e-04, -4.7346e-02,  9.0693e-04,
        -2.1961e-01, -3.5167e-01, -1.4552e-04,  4.0642e-04,  5.0891e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.8944e-04, -2.7599e-04,  5.8722e-04, -4.2421e-04,  2.3396e-04,
        -1.0455e-03,  5.9745e-05,  8.8378e-05,  1.6866e-04, -2.1230e-04,
         3.6133e-04, -6.9705e-02,  5.6585e-04,  9.2245e-01, -1.3050e-03,
         7.1267e-04,  1.0110e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8592e-03, -6.5807e-05, -3.0877e-04, -4.1452e-04, -1.3651e-04,
         4.3060e-05,  2.2664e-04,  1.8948e-04, -6.3460e-04,  7.2325e-05,
        -7.3785e-05, -1.7490e-01,  5.6116e-04,  1.9649e-04, -5.1595e-01,
        -3.0399e-01, -3.5135e-04, -2.4370e-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([-4.9181e-04,  1.4341e-04,  3.2964e-05, -9.8772e-05, -4.8060e-04,
        -5.0726e-05,  3.2539e-04,  1.8251e-05,  2.2820e-04, -4.2897e-05,
        -4.1665e-05,  4.7924e-01, -1.2077e-04, -2.4299e-04,  5.5206e-05,
         8.2703e-04,  5.1720e-01, -3.8466e-05,  3.2902e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2342e-04, -2.2330e-04, -1.3078e-04,  2.9755e-04, -9.7239e-05,
         6.6612e-05,  1.2963e-04, -4.9156e-04, -2.1657e-04, -4.6534e-04,
        -2.9361e-04, -5.6786e-01,  8.1170e-04,  8.8819e-04, -2.2543e-01,
         2.0141e-01, -6.8693e-04, -3.9036e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.8697e-04,  8.6632e-06,  1.9506e-04,  6.3724e-05,  1.7629e-04,
         5.2320e-05, -5.7674e-06,  2.7023e-04,  1.2435e-04, -2.3545e-01,
        -3.4294e-01, -5.0205e-05, -1.7987e-01, -2.3896e-01, -2.3119e-04,
        -4.3586e-04,  2.7434e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5251e-03,  3.7795e-04, -6.6473e-04, -3.1495e-05, -2.9130e-04,
         4.8696e-04, -3.2869e-04, -8.1875e-05,  9.9286e-01, -5.6981e-04,
         7.2715e-04,  8.3527e-05, -1.3841e-03, -2.6638e-04, -3.2583e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.3897e-04,  4.6166e-05,  1.0902e-03, -2.5453e-04,  2.2690e-04,
         1.3805e-04, -2.3859e-04, -1.6781e-04, -3.3274e-01,  2.1121e-03,
        -1.4885e-01, -1.8334e-01, -3.2956e-01,  4.8524e-05, -7.2362e-04,
         2.9017e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6610e-03,  8.7047e-05, -5.5757e-05, -2.8085e-04, -6.2280e-05,
         3.4724e-04, -6.7081e-05, -2.2744e-05, -1.3468e-04, -1.0755e-04,
         2.8225e-04,  1.0402e-04,  3.0621e-04,  9.1835e-06,  7.4880e-05,
         3.3936e-04,  3.8923e-05, -1.0253e-04,  1.8168e-02, -2.2886e-04,
        -1.4419e-02, -1.1241e-04,  5.6718e-05, -3.0786e-01,  1.8027e-02,
        -6.3559e-01,  9.3428e-05, -2.4300e-04, -1.2299e-04], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #2000: [tensor([-5.8041e-04, -1.4028e-04,  1.8758e-04,  1.8201e-05,  9.0579e-05,
        -1.4449e-04,  1.2825e-04,  2.7112e-04,  3.3660e-04, -1.3509e-04,
         7.1517e-05,  3.4957e-04,  2.3685e-04,  9.0212e-05,  2.4941e-05,
        -2.0897e-04,  7.8318e-05, -4.3634e-05,  1.0453e-04, -1.0412e-04,
         6.9803e-05,  4.9425e-01,  1.4610e-04, -5.6152e-05,  6.9853e-04,
         4.9938e-01, -2.7281e-04,  1.4218e-03, -2.5887e-04, -9.7937e-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([-1.5242e-03,  1.2683e-04,  1.0743e-05,  1.3889e-04,  1.1781e-04,
        -1.3891e-04,  4.7784e-05, -2.1569e-05,  7.9762e-05,  8.0377e-06,
        -1.0093e-05, -5.9902e-05,  5.2090e-05, -1.3776e-04,  2.4080e-04,
         1.2607e-06, -1.3177e-04, -7.7112e-05, -7.9119e-05, -9.4750e-02,
        -7.1261e-06, -1.7496e-01,  1.7860e-04, -6.4364e-03, -3.2213e-01,
         9.0779e-05,  1.0413e-04,  2.4576e-05, -2.5457e-01, -7.2580e-06,
        -5.9738e-02, -8.3716e-02, -1.4901e-05, -1.9637e-04, -6.4633e-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([-6.5455e-04,  2.4358e-04,  2.0983e-04,  3.0477e-05,  2.5647e-05,
         1.5930e-04,  4.9046e-05, -5.0118e-05, -1.9296e-04,  1.7583e-04,
         3.3836e-04,  4.6598e-04,  3.6057e-04,  2.5185e-04, -2.6281e-05,
        -1.9511e-04, -4.2007e-04, -2.8917e-04,  4.8274e-05, -8.7826e-05,
         2.7006e-04, -5.4317e-04,  4.4974e-02,  4.1469e-04, -9.4766e-01,
        -2.8355e-05, -8.6649e-04,  2.4002e-04, -5.7806e-04, -1.5306e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.2377e-04,  2.3445e-04, -3.4029e-06,  2.6670e-04,  1.9950e-04,
         2.8400e-04, -1.3483e-04, -1.1690e-04, -2.2510e-04,  5.7159e-04,
        -1.3140e-04, -1.4277e-04, -5.0802e-05, -7.6139e-05,  2.5213e-04,
        -1.2347e-04,  2.0694e-04, -3.3285e-04, -1.2788e-04,  5.3860e-01,
         2.3856e-04, -1.9509e-02, -3.4163e-04,  2.1990e-04,  4.3617e-01,
        -1.7099e-04,  2.2801e-04,  1.6428e-04, -1.4826e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0817e-04, -8.0765e-05, -1.8234e-04, -1.1689e-04,  2.1738e-05,
        -3.8350e-05, -1.9365e-04,  5.8251e-05, -1.3575e-05, -7.0014e-05,
         1.0105e-04, -2.3448e-04, -3.6325e-04, -8.6248e-05,  1.0297e-04,
        -1.4711e-04, -6.0699e-05,  1.6168e-04,  1.9360e-05, -1.6137e-04,
         2.7098e-01,  2.5957e-04, -1.2035e-04, -4.8715e-05,  1.9091e-04,
        -1.1828e-03, -2.7450e-04,  3.3703e-01,  2.9273e-04,  3.8206e-02,
        -1.6792e-04, -3.2786e-03, -2.5755e-05,  3.4415e-01,  3.1446e-04,
         1.9889e-04,  4.6115e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2348e-04,  7.0478e-05,  1.1970e-04, -2.8331e-05,  2.8251e-05,
         6.5388e-05, -1.1352e-04,  1.2095e-04,  1.9793e-04,  5.9188e-05,
         3.9850e-05, -2.3797e-04,  1.1149e-04,  3.8490e-05, -7.9031e-05,
         1.4840e-04,  2.0521e-04,  1.2634e-04, -4.2502e-05, -1.7609e-04,
         5.2844e-02,  3.1704e-04,  5.0764e-05, -7.7966e-03,  6.5618e-01,
        -3.1346e-04,  3.2739e-05, -8.6022e-05, -4.2026e-04, -4.8925e-02,
        -2.1996e-04, -9.5889e-05, -1.3585e-04, -5.9387e-05,  2.2994e-01,
         9.8318e-05,  2.9858e-04,  5.3852e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4570e-04, -4.0338e-05, -1.1692e-04,  1.3040e-04, -9.8131e-05,
         1.5819e-04,  1.8284e-04, -1.5370e-04,  2.5414e-04, -8.7709e-05,
        -4.1711e-05, -3.2271e-05,  8.4440e-05,  1.0484e-04,  3.3671e-04,
        -5.3180e-05, -3.1015e-05, -3.5316e-05,  9.5477e-05, -1.0541e-01,
        -1.5263e-04,  5.6165e-04, -4.4014e-01,  2.7137e-05,  9.1111e-05,
        -1.7838e-04,  7.6270e-05, -4.5048e-01, -1.1615e-04, -2.6851e-04,
         1.7232e-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([-7.6718e-04, -4.1591e-05, -2.6480e-04,  3.1664e-04,  1.8246e-04,
        -1.8381e-05,  7.8954e-05, -1.6289e-04, -3.1614e-04, -1.5072e-04,
         1.6023e-04,  1.5529e-04,  1.3164e-05,  1.0355e-04,  8.9852e-05,
         8.6989e-05, -4.0713e-04,  1.5694e-04,  2.2202e-04,  2.8201e-06,
         3.3463e-04,  1.1581e-04, -1.1760e-04,  2.9368e-03, -5.3675e-04,
         8.3859e-04, -9.9010e-01, -5.5603e-04,  1.5968e-04, -5.1457e-04,
        -9.3782e-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([ 3.0417e-04,  1.4298e-04, -2.3017e-04, -1.1044e-04,  2.9893e-04,
        -3.6511e-05, -5.6483e-05, -2.1144e-04, -1.2728e-04,  2.6087e-04,
         7.7711e-05, -1.2585e-04, -4.8175e-05, -1.0813e-04,  1.0330e-05,
         3.7151e-05,  2.2014e-04,  1.7365e-04, -4.3439e-05,  4.3537e-01,
        -1.5790e-04, -7.1615e-03, -1.2927e-05,  5.4510e-01, -1.8347e-04,
        -7.7608e-03, -3.3872e-04,  1.1181e-03, -1.1953e-04, -4.8881e-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([ 1.2867e-03, -3.3634e-06,  1.2770e-04, -5.7651e-05, -2.0062e-04,
        -3.8440e-05, -6.0296e-05, -9.5829e-05,  3.2911e-04, -1.0606e-06,
         1.4829e-04,  1.4068e-04,  6.0944e-06,  4.1910e-06, -2.9282e-05,
        -6.8857e-05, -2.5577e-04,  1.2550e-04,  1.5127e-04, -3.0953e-01,
        -2.3925e-04, -2.0340e-04, -2.3978e-01, -1.8777e-04,  1.5552e-04,
        -4.4561e-01, -2.8085e-04, -6.6877e-04, -2.0897e-04,  3.7561e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1166e-04, -1.1644e-04, -5.3995e-05,  7.6925e-05, -1.8593e-05,
        -6.2204e-05, -7.3351e-05, -1.1465e-04,  3.4138e-05, -1.1124e-04,
         8.6967e-05,  1.7485e-04, -6.7640e-05, -2.6110e-05,  1.0831e-04,
        -3.7151e-05, -1.8950e-04,  1.4741e-04, -4.3764e-05, -8.6821e-02,
        -2.0350e-04, -1.4146e-01,  4.7855e-05,  4.4428e-03, -2.0993e-01,
        -1.7202e-04,  2.0554e-03, -2.8802e-01,  8.0018e-05,  1.8253e-03,
        -1.9903e-01, -8.5333e-05,  3.2257e-04, -6.2200e-02,  7.2296e-05,
         6.3947e-04, -2.4476e-04, -2.8648e-04,  1.0018e-04, -9.9927e-05,
        -9.7579e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.4005e-04,  1.8957e-05, -3.5279e-06, -1.1320e-04, -5.3852e-05,
         1.3273e-04, -9.4350e-05, -7.5205e-05, -2.8084e-04, -1.4933e-05,
         4.5278e-05,  8.8110e-05, -1.2529e-04, -5.5364e-05, -6.0037e-05,
        -2.4011e-04, -3.0320e-04, -1.5627e-05,  4.6889e-05, -2.7922e-01,
         1.4995e-04, -6.7904e-02, -2.1139e-01,  1.0883e-03, -1.8546e-04,
         4.5420e-05, -4.2769e-01, -2.7785e-04, -5.9077e-05,  9.3770e-03,
         9.2741e-05,  4.2870e-05, -7.3665e-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 #2050: [tensor([-1.0343e-04,  9.6257e-05,  9.2647e-05,  1.4373e-04,  1.4328e-04,
        -2.4394e-04, -6.0017e-05,  9.9332e-05,  8.6499e-05,  4.5439e-05,
        -3.8108e-05, -1.6536e-04,  3.5780e-04, -5.1773e-05,  1.0881e-04,
        -1.9406e-04, -3.4965e-06,  8.5005e-05,  1.3869e-04,  2.3907e-07,
         7.0047e-05,  2.1231e-04,  5.3434e-02,  9.7194e-06, -2.6224e-04,
        -3.7542e-01, -3.1888e-04, -5.6673e-01, -5.4524e-05, -1.8305e-04,
        -7.5991e-04, -2.3032e-04, -5.7333e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6417e-04,  4.1832e-06,  2.1879e-04,  3.9337e-05,  8.8966e-05,
        -6.7006e-05, -2.7662e-05, -5.5203e-05, -2.3651e-05,  8.9085e-06,
        -1.4633e-04, -1.4624e-04,  1.7646e-04,  1.5887e-04, -3.8993e-05,
         1.1579e-04, -7.7847e-06,  5.2184e-05,  4.3765e-05, -5.0848e-05,
         7.7737e-05, -6.6179e-05, -3.5588e-01, -2.4578e-04, -2.9857e-01,
         3.1933e-04,  5.6371e-04,  8.9321e-04, -3.4133e-01,  1.0830e-04,
        -1.6633e-04, -5.3401e-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([-4.5659e-04,  6.1698e-05,  1.3886e-04, -2.1994e-05, -1.6907e-04,
        -8.3911e-06,  5.2683e-05, -6.3778e-05, -5.8773e-05, -4.8445e-05,
        -5.4706e-05,  1.6140e-05,  1.0413e-04,  6.3295e-01,  9.4418e-05,
         3.1611e-01,  1.4923e-05, -1.6275e-03, -7.6181e-04,  2.1620e-02,
         1.9372e-04, -2.3552e-02, -2.2241e-04, -8.9068e-05,  1.1632e-04,
        -3.6172e-04,  3.5507e-04,  3.2975e-04,  1.9616e-04,  1.4673e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7599e-04,  1.5526e-04,  2.1629e-04,  1.7901e-04, -9.9921e-05,
         3.7252e-04, -2.2422e-04, -1.1832e-04,  3.4993e-05,  2.2396e-06,
         1.9754e-04, -1.9982e-04,  4.3096e-05,  5.5155e-01,  4.6656e-04,
         4.4353e-01, -1.9577e-03, -3.4944e-04, -1.3166e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.6624e-04, -1.0301e-04, -2.3281e-04,  3.7605e-04, -2.7285e-04,
         3.2594e-04,  8.8918e-07,  1.8790e-04,  6.7489e-05, -5.0868e-05,
        -2.1600e-04,  1.5747e-04, -1.8593e-04,  9.6223e-01, -4.0341e-04,
        -3.3992e-02, -6.6435e-05, -6.2370e-05,  3.9992e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7844e-04, -2.4114e-05, -8.6314e-05, -6.3742e-05,  2.3403e-04,
        -1.3089e-04,  5.6633e-05, -6.6106e-05,  3.3841e-04,  1.5560e-04,
        -6.6321e-05,  1.5831e-04,  1.4569e-04,  4.2243e-04,  4.4550e-05,
         9.6703e-05, -1.5694e-04,  1.1398e-05,  1.7696e-04,  1.2546e-04,
         1.0407e-04, -3.4163e-04,  1.4319e-04, -2.1487e-04,  3.1163e-04,
        -6.6165e-05,  2.0537e-04,  1.1211e-04,  2.8606e-04,  3.4821e-05,
         9.1954e-01, -1.4385e-04, -1.0380e-04,  6.3298e-02, -5.0148e-04,
        -2.4046e-03,  1.6145e-04, -7.3518e-03, -3.3019e-04,  1.4292e-05,
        -1.1558e-03, -1.7024e-04, -1.0110e-04, -1.6769e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.0017e-04,  1.9348e-04, -8.4174e-06,  1.3727e-04, -4.4346e-05,
        -3.5171e-05, -7.0336e-05, -1.0100e-04,  2.5551e-04,  7.8740e-05,
         1.9839e-04, -2.3345e-05,  1.1839e-04,  5.7961e-06,  1.7354e-04,
        -2.4174e-04, -4.2643e-05,  2.0312e-04, -8.0514e-05, -8.2293e-05,
        -2.2800e-04,  2.7756e-04,  1.1444e-04, -6.9257e-06,  3.6878e-05,
         1.9844e-04, -1.1645e-04, -9.5521e-05, -1.4934e-04, -8.5060e-05,
         7.7701e-02,  2.5165e-04,  2.9535e-04, -9.0576e-01, -2.8300e-04,
         5.5189e-05,  1.5823e-04,  1.2356e-03,  2.3062e-04,  9.8572e-05,
        -3.5012e-04,  8.6805e-03,  1.7743e-05, -6.5560e-04, -4.0333e-05,
         8.1231e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2540e-02,  4.4386e-03, -2.9545e-03,  3.6431e-03, -1.0184e-03,
         4.0202e-03,  4.4866e-04,  6.8400e-03, -2.2963e-03,  3.9060e-03,
         5.6248e-03,  2.9032e-03,  2.5912e-03,  4.5621e-03, -1.0737e-03,
         1.3204e-03, -2.2193e-03,  1.0269e-03, -4.4936e-03,  3.1593e-03,
        -2.1589e-04, -3.1843e-03,  3.3404e-03,  7.3506e-03, -2.7649e-03,
         5.5107e-03,  5.2537e-03,  3.1550e-03, -1.0009e-03, -4.3741e-03,
        -1.9948e-02,  8.3303e-01, -5.5185e-03, -3.4274e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3129e-04, -2.1412e-04,  2.3100e-05, -1.4785e-05, -4.1607e-04,
        -3.3714e-04,  1.3384e-04,  1.1413e-04, -4.6834e-04,  3.3582e-05,
        -3.6019e-04, -3.4391e-04,  1.7207e-04, -2.9663e-04, -1.4242e-04,
         2.0788e-04, -8.6304e-02,  2.9351e-04,  2.5995e-04,  8.5890e-01,
         1.2541e-04, -2.1713e-05, -1.6857e-02, -5.2716e-04, -3.1223e-02,
        -8.6909e-04,  6.1053e-06, -2.4440e-04, -3.0881e-04,  2.8029e-04,
         3.6851e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8625e-04,  3.5748e-04, -1.5584e-04,  1.6703e-04, -8.4496e-05,
         2.6940e-05,  1.0425e-04,  4.0392e-04,  9.9231e-05, -1.0351e-04,
        -3.2229e-05,  4.2531e-05, -1.0664e-04,  2.2543e-04, -1.2529e-05,
         1.9421e-04,  6.0997e-01,  2.5038e-04,  3.8693e-01,  3.4916e-04,
        -2.3014e-05, -1.6687e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0880e-04,  1.8329e-05, -1.0923e-04,  1.1698e-04, -1.1304e-04,
         6.1226e-05,  1.6402e-04,  1.1557e-04,  3.2196e-04, -5.0623e-05,
         2.3610e-04, -5.5303e-05, -4.1532e-05, -1.6082e-04,  7.5949e-05,
         5.8444e-05,  2.7305e-02,  1.6460e-04, -6.2724e-04, -9.4688e-01,
        -2.0886e-04,  2.8279e-05,  1.3254e-04,  6.3097e-03, -4.9289e-05,
         1.4699e-02,  2.6722e-05,  2.7973e-04,  2.3113e-04, -2.1490e-04,
        -2.8333e-05,  3.1671e-04, -2.5375e-04,  1.4992e-04,  8.4376e-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([ 1.0047e-03,  7.6318e-06, -1.9034e-04, -3.2650e-04,  5.4222e-05,
        -9.4523e-05,  4.6135e-05, -1.7723e-04, -5.6023e-05,  4.3832e-04,
        -1.6737e-04, -3.7698e-04, -2.1721e-04, -3.1185e-04,  2.6613e-05,
         1.5143e-04, -6.1281e-05, -5.1261e-01,  2.4339e-04, -4.5733e-01,
         5.4460e-04,  1.1771e-04,  2.4906e-02,  5.3634e-04,  3.8507e-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], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #2100: [tensor([ 1.5427e-05,  1.7768e-04, -2.6754e-04,  2.7631e-04,  4.1704e-06,
         1.3968e-04, -3.8800e-04,  3.8269e-04,  3.5164e-04, -1.6650e-04,
         2.5788e-05,  9.6361e-06,  2.4134e-04,  2.5423e-05,  5.0039e-04,
        -3.1971e-01,  2.9049e-04,  4.4931e-04,  6.7778e-04, -2.3727e-01,
        -4.3818e-01, -2.5839e-04,  1.1178e-04,  7.2669e-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([ 9.6068e-05, -4.5007e-04, -4.0804e-04,  3.2766e-04,  1.9553e-04,
        -1.3856e-04, -3.7229e-04,  1.7493e-04, -1.0709e-04, -2.1735e-04,
        -2.3090e-04, -4.1568e-04,  6.2756e-04, -2.3334e-04,  2.6339e-05,
         9.9341e-01, -4.1174e-04,  4.5256e-04, -3.6739e-04,  1.3941e-04,
        -1.2589e-04, -4.4055e-04, -6.3289e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2479e-04, -1.7375e-05, -5.8323e-04, -9.3416e-05, -4.3220e-04,
        -4.4218e-04, -4.8449e-05,  1.2491e-04, -3.4213e-04,  3.5295e-04,
        -1.3171e-05,  7.4367e-05, -1.3504e-04, -4.0685e-04,  2.0772e-04,
         2.0143e-02, -1.1751e-04,  2.8060e-04,  1.9492e-04,  4.0855e-04,
        -1.4873e-04,  2.1198e-04, -9.7433e-01, -1.1599e-04, -1.6752e-04,
        -2.3502e-04,  1.3889e-04, -9.8281e-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([-2.5682e-04,  1.4472e-04,  1.7725e-04,  1.1061e-05,  7.5688e-05,
        -3.2379e-05, -2.3735e-04, -1.4424e-04, -2.0583e-04, -8.8572e-05,
        -1.7143e-04, -1.0292e-04,  5.2633e-05,  1.2832e-05,  1.8681e-04,
         9.4553e-05,  1.0629e-04, -5.5308e-05,  9.3782e-01,  4.6355e-04,
         3.6140e-04, -6.2274e-04, -5.8156e-02, -6.5586e-05, -2.6004e-04,
         9.3027e-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([ 5.2704e-04,  2.8743e-04,  3.0143e-04,  1.6942e-04, -4.1843e-04,
        -1.6922e-04, -1.1662e-06,  1.3271e-05, -4.0464e-05,  2.8338e-04,
        -1.2014e-04, -1.4923e-04,  5.4499e-04, -1.1903e-04, -1.2920e-04,
         1.1332e-04, -1.3331e-06, -2.3904e-02, -1.3659e-05, -3.9302e-04,
        -9.7145e-01, -5.7787e-04, -1.1884e-04, -9.3794e-05, -6.0089e-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([ 1.3619e-04, -1.4975e-04,  5.1613e-05,  4.1396e-05, -4.2434e-04,
         1.2487e-04, -3.5284e-04, -1.0437e-04, -1.6975e-04, -2.7528e-04,
         1.6386e-04, -1.8753e-04,  2.3300e-05,  1.6760e-04,  5.6660e-05,
        -1.9938e-04, -6.9123e-05, -5.9461e-02,  2.1299e-04,  4.4582e-04,
         1.4091e-02, -2.1247e-03,  2.1478e-01,  7.0557e-01, -2.2939e-04,
         1.0020e-04, -2.7878e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9831e-05, -1.0370e-04, -9.0441e-04, -1.9349e-04,  1.4343e-04,
        -5.9210e-04, -4.8102e-05, -2.6974e-04,  1.6966e-04, -1.6663e-04,
        -2.3453e-04, -2.9331e-05,  5.2031e-05, -7.6706e-04,  5.4978e-05,
        -1.1330e-05, -1.1098e-04,  9.5873e-01,  2.6381e-04, -3.1437e-03,
        -3.2746e-02,  1.3529e-04,  3.3770e-04, -1.7448e-04, -5.9558e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.8902e-05, -1.2523e-04,  4.4004e-05, -2.4396e-06,  2.0571e-04,
         1.7305e-04, -2.0378e-06,  3.6021e-05, -1.2370e-04,  9.9359e-05,
        -2.0172e-04, -5.5764e-04, -9.1765e-05,  5.7174e-04,  6.3544e-05,
        -1.0665e-04,  1.0837e-04, -1.6446e-04, -6.9476e-01, -9.8943e-05,
         1.0854e-02, -6.7601e-02, -2.2294e-01,  4.5670e-04, -3.5586e-04,
        -7.5494e-05, -8.9028e-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([ 4.0268e-05,  6.4803e-05,  3.5022e-04, -1.3016e-05, -1.7797e-04,
        -8.9079e-05,  1.8919e-04, -9.7545e-05,  2.4531e-04,  8.1084e-06,
         1.1348e-04, -1.0045e-04,  8.0810e-05, -2.5558e-04,  2.1320e-05,
         6.1583e-05, -2.2911e-04,  3.0486e-04,  6.0846e-02, -3.5743e-08,
        -6.9025e-03,  8.3274e-01, -1.5124e-04,  9.5581e-02, -9.7985e-05,
        -8.1704e-04, -3.9513e-04, -2.7553e-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([-4.0336e-04, -2.0809e-04, -9.7084e-06, -8.2358e-05, -7.2238e-05,
        -6.9548e-05, -3.4310e-04, -1.2753e-05,  1.1019e-04,  1.4186e-04,
         3.2766e-05,  1.2114e-04,  7.9088e-05,  5.2689e-05,  7.7687e-05,
         1.7218e-05, -1.9842e-05,  1.8337e-04,  5.0723e-01, -6.0823e-05,
         2.9123e-01, -3.2292e-04,  8.6498e-05,  1.9676e-01,  3.5566e-04,
         1.1692e-04, -1.4262e-03, -7.0041e-05, -2.4194e-05,  2.8298e-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([-1.3088e-03, -1.9957e-04, -8.5782e-05,  2.4469e-04,  7.9209e-05,
        -6.0414e-05,  1.4866e-04,  9.1459e-06,  9.4046e-05, -9.0131e-06,
         1.1119e-04, -4.0637e-05, -4.2098e-06, -2.5390e-05, -4.8596e-05,
        -1.6102e-04,  9.9900e-06, -3.0590e-05,  1.0713e-05, -4.4988e-05,
         8.0797e-05, -8.8756e-05,  1.1651e-04, -9.6880e-05,  3.3190e-01,
         3.3804e-04,  8.6209e-03,  1.2673e-04, -2.8544e-01, -9.5741e-05,
        -3.6935e-01, -1.5420e-04, -2.8968e-04, -4.8251e-04,  9.7063e-05,
         7.8157e-06], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6517e-03, -1.5969e-04,  3.4938e-04, -5.8029e-04, -2.3527e-04,
        -6.2152e-04,  1.9499e-04, -8.2361e-05, -3.7209e-04,  1.1885e-04,
        -2.7736e-05,  8.6874e-05,  1.0247e-04, -8.2072e-05,  7.8620e-05,
        -3.6784e-04,  1.6696e-04, -3.8414e-04, -6.3769e-05, -2.7007e-04,
        -1.2368e-04,  2.8219e-04,  1.2814e-04,  8.2338e-06, -1.2258e-04,
         4.2851e-05, -1.6682e-01,  4.8957e-04, -8.2509e-01,  5.5290e-04,
        -2.6398e-04,  7.8954e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #2150: [tensor([-6.5914e-05,  4.3608e-05,  8.3114e-05,  2.3084e-04, -1.1789e-04,
         1.6866e-04, -5.1068e-05, -1.9354e-04, -2.0804e-04, -2.0087e-04,
        -9.8586e-05,  1.9004e-04,  2.2655e-04, -1.1258e-04, -1.8796e-04,
         4.7028e-01,  4.3377e-07,  5.2672e-01, -1.8620e-04, -1.6774e-04,
         3.0904e-04, -1.5981e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2976e-04,  3.9576e-05,  2.7697e-04,  7.6758e-05,  2.2160e-04,
        -3.7485e-04, -7.5626e-05,  1.7474e-05,  1.0156e-04,  3.1020e-04,
         4.3584e-05, -9.1320e-05,  1.6862e-04, -1.5905e-04,  2.2053e-04,
         2.6362e-04,  2.7506e-05, -2.1848e-01, -3.5370e-04, -3.5145e-01,
        -6.6708e-05, -2.3373e-04,  6.7135e-05,  1.5538e-02, -4.1012e-01,
         1.4032e-04,  4.0714e-04,  2.4186e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8299e-04,  1.6281e-04,  2.4436e-04,  2.1431e-04, -2.4502e-04,
         3.2380e-04, -1.7584e-04,  1.7267e-04,  1.5100e-04,  1.0241e-04,
        -2.0242e-04,  1.5038e-04,  1.2285e-04, -8.1005e-05,  3.9630e-04,
         3.2281e-04,  1.8860e-01,  4.2670e-04,  2.3640e-04,  1.3517e-01,
         6.7141e-01,  3.7686e-04, -6.5967e-05, -1.6467e-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([ 1.2750e-04,  4.4666e-04,  1.2202e-04, -4.8618e-04, -5.8688e-04,
        -1.7568e-04, -7.3731e-04, -6.2047e-04, -3.5144e-04,  2.3358e-04,
        -3.1480e-04, -4.7180e-04,  6.5079e-04,  5.5371e-04,  3.3315e-04,
         9.9244e-01,  2.4816e-04, -2.6800e-04,  1.8717e-04,  6.4763e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.6072e-05, -1.2655e-04, -7.3608e-05, -4.5694e-04,  5.9145e-05,
        -1.9239e-04, -2.2657e-04,  1.4484e-04, -1.2430e-04,  3.2794e-04,
        -1.4109e-04, -2.1456e-04, -2.1296e-05, -5.8983e-04, -1.1739e-04,
        -3.8778e-04,  2.0467e-04, -1.1823e-03, -5.4476e-02, -9.3991e-01,
         7.6146e-05, -3.0075e-04, -5.7625e-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([ 7.2112e-04, -2.4647e-04, -3.3291e-05, -4.6465e-05,  6.1461e-05,
        -5.6451e-05,  3.5926e-04,  1.3983e-04,  3.9765e-05,  1.3763e-04,
         4.7265e-05,  3.1460e-05,  9.4397e-05, -8.7336e-05, -2.0184e-04,
         1.8086e-04,  9.6236e-05, -4.4187e-01, -1.1227e-04, -3.7854e-05,
        -1.4662e-03,  4.1476e-04, -2.4102e-01, -3.0935e-01,  3.4872e-04,
        -1.0542e-03, -1.2727e-03, -6.0233e-05,  1.3179e-04,  2.7521e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2112e-04, -4.7103e-05,  1.4386e-04, -1.1338e-04,  4.4171e-05,
         3.3609e-04,  8.4844e-05,  5.8056e-05, -5.2409e-05, -1.4433e-04,
        -7.9891e-05,  6.3968e-05,  2.3850e-05, -6.4929e-01, -1.6357e-04,
         2.9808e-04, -5.4417e-06, -4.1796e-04, -8.7205e-02, -2.5674e-01,
        -2.3767e-04, -3.5131e-03,  1.5728e-04, -3.7672e-06, -3.5217e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.1031e-04, -2.7310e-04,  6.6783e-05, -8.5707e-05, -4.6625e-06,
         3.5378e-05, -4.1622e-05,  1.4559e-04,  4.9666e-05, -6.9754e-05,
        -1.5048e-04,  4.3337e-05, -4.9121e-01, -5.1266e-04,  2.9236e-04,
         6.1583e-04, -5.8055e-04, -6.7380e-03, -6.2456e-05, -4.9734e-01,
         2.2972e-04, -5.1086e-04, -1.0109e-04,  3.0338e-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([-2.7456e-04, -2.8591e-07,  1.5855e-04,  8.5317e-05, -3.6893e-04,
         2.2377e-04, -1.4210e-05,  2.8560e-04,  1.5799e-04, -2.4927e-04,
        -3.2103e-05, -1.0477e-04,  9.9632e-02,  8.2072e-05,  2.4708e-04,
        -1.6034e-04, -5.0710e-01,  1.1622e-04, -3.8950e-01,  3.3049e-05,
        -7.1264e-04,  9.6812e-05,  2.1794e-04, -1.4532e-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([ 3.3674e-03, -4.9177e-04, -4.2455e-04, -2.5304e-04, -4.0429e-05,
        -1.2372e-04,  6.4001e-04, -6.2585e-04, -4.1975e-05, -7.6257e-05,
         1.0260e-04, -1.3844e-04,  4.0738e-05,  1.3271e-01, -4.4139e-04,
         3.6740e-03,  8.9104e-04,  4.0981e-04,  8.5349e-01,  1.4211e-04,
         7.3508e-04, -5.4050e-04, -6.0133e-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([-7.4315e-04,  2.3200e-04, -1.6286e-04,  6.4273e-04,  5.8217e-04,
         8.7353e-05,  4.4043e-04,  1.3550e-04,  1.9031e-04,  3.1561e-04,
        -7.8288e-05, -2.7041e-04, -1.2464e-04, -1.0619e-03,  4.1869e-04,
         4.9435e-02,  9.4147e-01,  6.3216e-04,  6.0414e-04,  1.5405e-03,
        -5.7376e-04,  2.5890e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2095e-03,  5.6396e-04,  2.0309e-03,  1.0546e-03, -3.2659e-03,
        -3.9083e-04, -1.3921e-03, -8.7494e-04,  3.1604e-03,  4.0234e-03,
         1.4639e-03, -3.0033e-04, -4.0349e-03, -1.4833e-02,  7.4059e-03,
         3.2835e-01,  6.1697e-01,  4.8579e-03,  8.1964e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  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 #2200: [tensor([ 5.6709e-04,  3.1945e-04,  2.5596e-04, -4.8227e-04, -4.4171e-04,
         2.7398e-05,  3.0287e-05,  1.3960e-04,  1.1625e-04, -9.9431e-05,
         6.8457e-05,  2.0522e-05,  6.5702e-06, -3.3602e-04, -9.5153e-05,
        -2.7584e-04,  1.0943e-04, -3.6799e-04, -3.7444e-04, -2.4771e-04,
         9.2989e-05, -9.7373e-01,  9.6323e-05, -2.0080e-02,  3.0179e-04,
        -6.5058e-04, -5.0182e-04, -1.6303e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.2286e-04,  4.3727e-04,  2.7454e-04,  3.1977e-05,  1.4188e-03,
        -6.1846e-04,  2.4424e-04,  7.3805e-05,  2.9767e-04,  1.2455e-03,
         7.0428e-04,  2.4139e-04,  2.1687e-04, -4.3164e-05, -4.7087e-04,
         2.1110e-04,  1.7604e-04,  2.3917e-03, -3.3410e-03, -6.2613e-04,
         8.5650e-01,  1.1925e-01,  8.5697e-05, -9.3563e-03,  1.0191e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7502e-05,  9.9270e-06,  1.9886e-04, -1.7975e-04,  2.6832e-05,
        -1.1189e-04,  3.1491e-04,  1.8062e-04, -1.2603e-04, -2.1643e-04,
        -4.5310e-04,  4.0642e-04,  1.9724e-04, -1.4901e-04,  1.1284e-04,
        -2.1967e-04,  1.4037e-04,  1.7179e-04,  4.2703e-04,  8.8689e-01,
         5.4125e-04,  4.3236e-04,  1.0777e-01,  1.8989e-04,  8.5932e-05,
         4.2901e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.7356e-05, -2.1435e-04,  1.9188e-04, -3.1319e-04,  1.1861e-04,
        -1.6544e-04,  3.4690e-04,  3.0644e-05, -5.0051e-05, -3.3381e-04,
         5.7514e-04,  8.4210e-05,  8.8511e-05, -1.8691e-01, -4.8399e-04,
        -8.0687e-01, -2.0829e-04, -2.5449e-03, -2.8270e-04, -1.2397e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2482e-04,  6.8286e-05, -2.0096e-04,  1.7562e-04, -6.9965e-04,
        -3.7092e-04, -1.2632e-04,  2.8197e-06,  2.4539e-04, -8.5219e-04,
         9.6198e-04,  7.7951e-04, -1.6767e-04,  1.7258e-02, -1.1925e-03,
        -9.7326e-01, -7.4269e-05,  7.9435e-04,  2.0145e-04, -9.3811e-04,
         8.8138e-04, -5.2415e-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([ 1.4423e-03, -5.1156e-04,  9.5371e-05,  2.0395e-04,  3.8798e-04,
        -5.7617e-04,  4.1884e-04, -1.9215e-04,  2.1390e-05,  1.2237e-04,
        -1.5247e-04,  6.3909e-05, -5.2029e-04,  3.1157e-02,  3.0419e-04,
        -9.6007e-01,  3.7057e-04,  8.7642e-05,  6.9330e-04, -1.1016e-03,
        -2.9124e-04, -3.6414e-04, -8.5685e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.3102e-04,  3.0128e-04, -1.7341e-04,  1.6014e-04,  2.8630e-04,
         2.8810e-04, -6.8968e-04, -1.9246e-04,  1.1844e-04,  4.8505e-04,
         5.4243e-04, -3.6370e-05,  6.5575e-05, -9.9368e-01, -4.8876e-04,
         1.7617e-04,  4.5763e-04, -2.4305e-04, -4.8829e-04, -4.0030e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.3833e-04, -1.2881e-04, -3.5970e-04,  1.1074e-04, -7.5761e-04,
        -1.7325e-04, -1.6247e-04,  2.3138e-04, -1.6732e-04, -2.4112e-05,
        -2.4136e-04, -2.0425e-04, -6.3773e-04,  7.3030e-01, -1.0184e-04,
         1.0748e-04, -2.7683e-05,  2.6508e-01, -1.9296e-05, -1.8151e-04,
        -4.2583e-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([-1.8811e-03,  4.5967e-05, -5.4290e-05,  1.4481e-04,  3.2637e-04,
        -2.7887e-05, -1.0810e-04, -1.7914e-04, -4.3579e-04, -1.4207e-04,
        -3.9086e-05, -2.9146e-04, -1.5972e-04, -9.9412e-01,  3.9675e-04,
        -8.9806e-05, -6.6772e-04, -7.5987e-04, -1.2683e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.2036e-04,  1.7175e-04,  6.4382e-04,  1.3583e-04, -1.8588e-04,
         3.5367e-04, -7.2065e-05, -1.4003e-04,  1.4255e-04,  5.2796e-05,
        -3.7627e-06,  9.8572e-06,  1.3824e-04, -3.5047e-04,  5.8881e-04,
        -6.8408e-01,  4.1793e-04,  3.9354e-04, -1.1071e-04, -2.4330e-04,
        -3.0540e-01,  6.2771e-04, -4.2366e-04, -3.8688e-03,  2.9833e-04,
        -1.1617e-05, -3.0895e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.2531e-04, -2.0260e-04,  4.8369e-04,  9.3281e-05, -1.0207e-04,
        -1.9825e-04,  2.3793e-04,  4.2364e-04,  1.0181e-03,  2.1475e-05,
        -5.6608e-05,  2.1533e-04, -2.6433e-04, -3.9939e-04,  6.1606e-04,
        -8.8753e-01, -1.0646e-01, -1.5290e-04,  2.5174e-04, -4.4979e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.5659e-06,  1.9072e-04,  3.0625e-04, -3.3591e-04,  1.8448e-05,
        -3.5253e-04, -2.0641e-04, -2.1966e-05,  1.2286e-04, -3.1955e-04,
         4.4567e-05, -3.4236e-04,  1.7865e-05,  2.9590e-04, -3.6884e-04,
        -1.1977e-02,  1.0740e-04, -9.5520e-04,  9.7354e-05, -2.7868e-04,
         9.5030e-01, -3.1681e-02, -7.1561e-05, -2.7928e-04,  4.5581e-04,
        -8.2060e-04,  1.7093e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #2250: [tensor([ 6.2211e-04,  8.6842e-06, -1.7611e-04, -3.4977e-04, -2.2559e-04,
        -4.3215e-04,  2.7006e-04,  1.1618e-04,  5.2006e-04,  1.5326e-04,
        -1.5754e-04, -8.2756e-05,  1.4204e-04, -9.7745e-05, -2.1503e-04,
         6.3912e-04, -9.9423e-01, -3.3148e-04, -5.4860e-05,  7.5254e-04,
        -4.2535e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7008e-04,  1.0424e-04, -1.9328e-04,  8.3031e-06,  1.6151e-04,
        -1.3816e-04,  2.8789e-04,  3.9081e-04,  4.2952e-04,  3.2092e-04,
         2.6713e-04,  1.8763e-05, -2.2704e-04, -9.0570e-05, -2.9842e-04,
        -9.0805e-01, -6.8979e-04, -4.3648e-02, -9.0004e-04, -4.2998e-02,
         1.6766e-04, -1.7055e-05, -4.2417e-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([-2.3019e-05,  2.6211e-04, -4.6356e-05,  2.6211e-04,  1.2915e-04,
         2.0086e-04,  8.7065e-05,  2.1828e-05, -4.7799e-05, -7.2180e-05,
         2.6511e-04,  3.3200e-04,  4.2919e-05,  2.0282e-05,  2.9210e-04,
        -5.8859e-05,  9.1993e-05, -1.3868e-01,  3.1897e-05, -4.3471e-01,
        -4.2412e-01, -8.6501e-05,  5.1701e-05, -5.6093e-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([ 2.2258e-04,  2.3376e-04, -3.1466e-04, -1.7117e-04, -4.8453e-05,
        -1.8643e-04, -1.0351e-04,  4.2039e-04,  1.5163e-04,  1.6427e-04,
        -2.0620e-04,  2.9965e-05,  1.6072e-04,  2.6675e-05, -2.1829e-04,
        -1.5515e-04,  9.7561e-05, -6.0401e-01,  1.1133e-04, -2.7502e-01,
         2.8595e-04,  1.1468e-01,  1.0827e-03, -1.7351e-04, -8.1333e-04,
         7.6567e-04, -1.5417e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4675e-04,  1.6766e-04,  8.8892e-05,  1.4204e-04,  3.4277e-04,
        -8.8612e-05,  9.7326e-05,  1.9434e-04,  2.0606e-04,  1.9198e-04,
        -2.7550e-04, -6.9583e-06, -1.4412e-04,  3.3559e-04,  1.0351e-04,
         2.3244e-04, -8.1330e-05, -5.7128e-01,  4.7875e-04, -1.6149e-01,
        -5.0414e-04, -2.6248e-01, -9.1333e-05,  3.5805e-04, -3.6791e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4131e-04,  2.2394e-04,  2.8073e-04, -2.9424e-05,  2.7962e-04,
         1.3331e-04,  1.4383e-04,  8.0224e-05,  2.5400e-04,  4.4957e-05,
         3.3630e-05, -4.8182e-05,  5.9927e-06, -1.9099e-05, -7.8064e-05,
         4.3179e-05,  1.2566e-04,  1.0449e-04,  8.7778e-03, -5.2560e-05,
        -5.5182e-04, -8.6348e-01, -2.5850e-04, -1.2094e-01, -2.0143e-03,
        -4.6469e-04, -5.8043e-04, -3.9627e-04, -5.7542e-05, -3.5649e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.6175e-04,  1.2617e-04,  4.1713e-04,  1.3452e-04,  1.1430e-04,
        -3.7816e-05,  2.3117e-04,  1.6797e-04,  9.3764e-05,  1.3683e-04,
         3.8296e-04, -1.7763e-04,  2.7699e-04,  1.6931e-04,  1.0124e-04,
         8.5224e-05,  5.3676e-05,  2.7153e-05, -3.9855e-01,  3.8464e-04,
         1.7376e-04, -4.5035e-02, -2.7673e-05, -5.5163e-01, -6.4862e-05,
         2.8562e-05, -4.0557e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.1138e-04,  5.0903e-04,  4.7198e-05, -9.8066e-06,  3.4665e-04,
        -5.7221e-04,  3.0724e-04,  1.0159e-04, -5.0093e-06,  1.2988e-04,
        -4.2143e-05, -2.7769e-04, -2.4145e-04, -1.2519e-04,  4.0973e-04,
         6.2204e-05, -2.6545e-04,  2.0788e-05, -9.4262e-01,  1.5670e-04,
        -2.0861e-04, -5.1089e-02, -6.4817e-04,  1.5140e-04, -2.0701e-04,
        -5.3284e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4869e-05,  1.4940e-04,  1.0048e-05,  1.4728e-04,  7.6731e-05,
         3.8240e-04,  1.7174e-04,  2.0187e-04,  2.6494e-04,  1.6067e-04,
        -2.8793e-04, -6.9259e-05,  2.1861e-04,  1.8914e-04, -1.9275e-04,
         1.2559e-04, -4.8019e-05, -3.0648e-01, -3.1784e-04, -1.6588e-01,
         1.8158e-04,  6.4539e-03, -8.4139e-05, -1.0973e-04,  4.6323e-03,
        -5.1303e-01, -7.9810e-06, -7.1797e-05, -6.0151e-06,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9199e-04,  1.6885e-04,  3.1210e-04, -4.2917e-04, -2.1210e-05,
         1.4158e-05, -5.3412e-06,  1.7699e-04, -1.0726e-05,  2.1706e-04,
        -1.5318e-04, -2.1194e-05,  1.3090e-04,  2.4631e-04, -1.7637e-04,
        -6.4493e-05, -3.5474e-05, -5.4704e-01,  5.8286e-04, -4.3783e-01,
        -2.6897e-04, -6.1925e-04,  1.0419e-02,  1.3969e-04, -4.9385e-04,
         1.2839e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4002e-03,  2.6511e-04,  1.7890e-04,  2.8251e-04,  1.4525e-04,
         1.4211e-04,  1.7162e-04,  3.1676e-04,  4.9473e-04,  1.9834e-04,
         9.9154e-05,  2.4503e-04,  4.5932e-04, -1.2512e-04, -5.3201e-05,
        -1.2894e-04,  1.5777e-04, -4.5588e-01,  2.7504e-05, -5.3824e-01,
        -9.4262e-05, -5.7690e-04, -3.1050e-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([-1.3204e-04, -2.5322e-04,  8.4415e-04, -1.3377e-04,  2.2441e-04,
        -4.1361e-04,  2.4477e-04, -5.1248e-04, -5.4785e-04,  2.3954e-04,
        -9.3553e-05,  9.6784e-05, -1.1160e-04, -1.0445e-04, -9.9356e-05,
        -9.9407e-01, -5.0699e-05, -3.5398e-04, -1.6315e-04, -3.6641e-04,
        -8.8175e-05, -1.1764e-04,  4.5351e-05, -2.7909e-04, -3.1194e-04,
         1.0389e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #2300: [tensor([-1.3305e-03, -1.0940e-04, -1.8358e-05,  4.6072e-04, -1.7685e-04,
        -7.3390e-06, -3.5736e-04,  1.2893e-04, -3.4164e-04, -8.5960e-05,
         1.4845e-04, -9.7403e-01,  5.3367e-05, -2.6386e-04, -2.1167e-02,
        -5.1481e-04,  5.9704e-04,  2.0584e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8870e-03,  3.7402e-05,  9.8470e-05, -1.1973e-04, -3.2957e-05,
         7.1387e-05,  2.9693e-06,  1.6980e-04,  7.7029e-05,  3.9926e-05,
         1.3324e-06, -4.7519e-05,  3.7020e-05,  1.8701e-04,  6.6458e-05,
         6.4906e-05, -2.8012e-05, -2.2485e-06,  2.1818e-04, -4.8608e-05,
         6.7003e-05,  1.1997e-05, -8.0577e-05,  5.3466e-05, -5.6389e-05,
         1.0167e-04,  2.7802e-05,  1.6679e-04, -1.5695e-05, -9.6647e-05,
         7.7319e-05, -1.7814e-05, -9.1655e-05, -2.9589e-05,  5.6043e-05,
        -8.8842e-06,  1.6423e-04, -3.5829e-01,  7.4604e-05, -1.9102e-03,
        -3.1445e-01,  9.8597e-05,  4.2055e-03, -2.2015e-04, -3.1608e-01,
         1.1999e-04, -1.6932e-04,  1.8195e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.5443e-04, -1.0657e-05,  1.0546e-05, -9.8709e-05, -5.9005e-05,
         3.5756e-05,  8.0564e-05,  5.2330e-05,  8.8847e-05,  1.3930e-04,
        -1.1758e-05,  2.6547e-04, -1.4030e-04, -7.1953e-05, -4.8967e-05,
        -1.0531e-04,  6.1149e-05,  1.9634e-05, -1.5098e-05, -8.8608e-05,
        -1.5283e-05,  4.9892e-05, -7.5046e-05,  7.0755e-05,  2.8201e-04,
         7.2721e-05,  3.3553e-04, -5.0419e-06,  5.2188e-05, -7.8805e-05,
         1.2377e-05, -1.0615e-04, -1.4204e-04, -7.6532e-05, -2.5525e-04,
         5.6944e-05,  4.6885e-05,  1.4366e-01,  1.3216e-04,  3.2550e-01,
        -9.6657e-05,  5.3698e-03,  1.4983e-01, -2.7091e-06,  4.2842e-05,
         3.7095e-01, -2.7559e-04,  7.6318e-05, -1.7084e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.4227e-05, -1.8612e-05, -1.4254e-04, -1.0462e-04, -1.2578e-04,
         1.4920e-04, -1.1576e-04, -6.6927e-06, -4.0748e-05, -1.7410e-05,
         1.7718e-06,  4.5122e-05, -2.4917e-05,  9.4507e-05, -4.3523e-05,
        -7.7322e-05,  6.8421e-05, -5.2506e-05, -2.2048e-04, -3.9701e-05,
        -1.4029e-04,  8.8310e-05, -1.4506e-04,  8.7760e-06, -1.4296e-04,
        -1.1530e-04,  1.0000e-04, -1.0338e-05, -1.4934e-04, -5.4839e-05,
         5.5478e-05,  4.7140e-05, -6.4331e-06, -7.2230e-05, -1.6287e-04,
         3.7979e-05,  6.2992e-05,  2.2321e-01,  1.2913e-04,  1.5781e-02,
        -6.9844e-01, -1.6482e-05,  9.4322e-03,  1.5181e-04,  4.9257e-02,
        -2.4845e-04,  1.0859e-04,  3.4022e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2024e-04,  1.3065e-04,  6.2224e-05, -1.2957e-04, -5.5201e-05,
        -2.2044e-04, -1.5775e-04,  4.3559e-05, -1.5411e-04, -3.2561e-04,
        -1.3342e-04,  2.1481e-04, -1.6346e-04,  1.5987e-04, -5.6172e-05,
        -2.6960e-04, -3.7355e-04, -8.2852e-05, -2.1710e-04, -7.9494e-05,
        -7.0630e-06, -3.3964e-05, -1.0287e-04, -1.0634e-04,  8.8813e-05,
         3.6504e-06, -9.6824e-05,  1.2941e-02, -2.5922e-04,  9.4827e-01,
        -5.7241e-05,  2.0812e-04, -1.9751e-04,  3.4845e-04, -3.3385e-02,
        -4.5836e-04,  2.3983e-05, -1.6005e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.4096e-04,  1.6090e-05, -1.1721e-04,  2.6761e-05, -4.4041e-05,
        -6.9991e-05, -3.5337e-04, -2.6171e-06,  1.4127e-04, -2.1076e-04,
         1.2597e-04, -1.0912e-05, -2.6680e-05,  1.2662e-05, -6.5205e-05,
        -3.1995e-06, -4.6795e-04,  7.2555e-06, -1.5164e-04,  4.0115e-04,
         1.4247e-04,  2.6234e-04,  1.9233e-04, -4.4890e-05,  2.3344e-04,
        -1.9838e-05,  6.6534e-05,  2.2713e-04, -9.7418e-01, -3.3540e-04,
        -1.0944e-03, -1.7914e-04,  1.0658e-02,  7.8547e-03,  4.0696e-04,
        -1.4496e-04, -7.5855e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0152e-04,  1.8394e-05,  1.0989e-04,  7.3901e-05, -1.0693e-04,
         2.4692e-04, -7.4147e-04,  2.9566e-04, -5.2342e-05, -5.1401e-04,
         1.9407e-04, -4.6665e-04, -6.9888e-05, -1.0634e-05, -4.7333e-05,
         3.1003e-05,  6.5905e-05,  4.4249e-04,  7.3896e-06, -4.9906e-04,
        -2.2262e-04,  3.5425e-05, -3.4734e-04, -2.0864e-04,  1.3085e-04,
         1.9633e-04, -2.4864e-04,  1.4716e-04,  9.6827e-01,  2.5097e-04,
        -1.5450e-02, -9.3263e-03,  4.3364e-04, -8.3223e-06,  5.3011e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2647e-03,  1.5196e-04, -5.8869e-05,  8.4302e-05,  3.7328e-05,
         3.1166e-04,  1.8175e-04,  1.5809e-04, -1.5982e-04, -9.6633e-05,
         3.5650e-05, -1.1541e-04,  6.7437e-05,  1.4734e-04, -7.9651e-05,
        -1.5212e-04,  6.0501e-05, -1.0653e-04, -1.6567e-04,  5.1605e-01,
        -1.4384e-04,  4.7455e-01,  3.0924e-04,  8.8313e-06, -4.7499e-03,
         3.6934e-04, -1.7295e-04, -2.0969e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1728e-04, -4.0599e-06, -2.0321e-04, -2.1589e-04, -1.0737e-04,
        -4.4914e-05,  3.3115e-04,  6.5298e-05, -2.0808e-04, -4.1566e-04,
        -2.0562e-04,  2.4934e-05, -1.2257e-04,  3.5422e-04,  4.5664e-05,
         9.3448e-05,  1.1294e-04,  1.8233e-04, -1.4338e-04,  5.8146e-01,
        -1.2445e-04, -2.0153e-04,  1.7406e-04,  9.5686e-05,  2.1125e-04,
         4.1447e-01, -1.5341e-05,  6.5189e-05, -7.7831e-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([-1.0025e-04,  1.1462e-04, -1.9088e-04, -5.5437e-05,  1.8226e-05,
        -2.0232e-04, -1.5472e-04, -1.0248e-04, -8.1780e-06, -1.7985e-04,
         1.6481e-04,  2.8977e-04,  3.7484e-05, -2.2759e-04,  3.6418e-05,
         1.4706e-04, -1.6445e-04, -2.6957e-05, -3.5308e-06,  8.4638e-01,
        -1.8768e-04, -2.5050e-04, -1.6409e-02,  8.3472e-05,  3.9378e-04,
        -4.5897e-04, -1.7658e-03,  1.3148e-01,  1.7242e-04,  3.8821e-05,
         1.5987e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2765e-04, -1.9698e-04, -5.9109e-04, -5.9780e-04,  2.1556e-04,
        -7.9065e-05,  8.7762e-05,  9.6472e-05, -3.5203e-04, -6.1227e-05,
         6.3437e-05, -2.4545e-04,  1.7467e-04,  4.1505e-04, -4.9661e-04,
         1.7012e-04,  1.7921e-04, -1.8923e-05,  5.7626e-04,  9.5696e-01,
         3.4932e-02, -1.4112e-04,  1.7947e-03, -1.0144e-03, -1.3448e-04,
        -7.3295e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2530e-04, -3.6654e-05,  8.9623e-06,  1.1726e-05, -3.8762e-04,
         4.8915e-04, -4.6484e-04, -2.5095e-04,  1.4204e-04, -3.0725e-04,
        -3.8522e-04,  1.4113e-04, -5.6433e-04, -5.6384e-04, -1.0809e-04,
         2.4225e-04,  9.0969e-05,  1.3848e-04, -1.0458e-03, -9.9404e-01,
        -1.2142e-04,  5.6157e-06, -2.8059e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
