Iter #50: [tensor([-0.3329, -0.1959, -0.0017,  0.0339,  0.0062, -0.0312, -0.0784,  0.0063,
        -0.0143, -0.0015, -0.0135, -0.0110, -0.0482, -0.0047, -0.0253, -0.0412,
        -0.1425, -0.0030,  0.0011, -0.0071,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.2345, -0.2617, -0.0153,  0.0355,  0.0084, -0.0356, -0.1022,  0.0179,
        -0.0237, -0.0276, -0.0086,  0.0022, -0.0537, -0.0260, -0.0467,  0.0027,
         0.0635, -0.0236, -0.0066,  0.0014, -0.0028,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4029e-01,  4.4692e-01, -1.3600e-02, -6.0774e-02,  2.3601e-03,
         5.5438e-02,  1.0591e-01,  6.8575e-03,  1.5151e-02, -9.5373e-03,
         3.3963e-02,  3.0916e-04,  3.2357e-02, -1.6423e-02, -7.9755e-04,
        -6.9398e-03,  8.4269e-03, -1.6827e-02, -1.2952e-02,  1.4164e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1255, -0.1543, -0.0108, -0.0142,  0.0171,  0.0123, -0.1513, -0.0027,
         0.0108, -0.0032,  0.0013, -0.1765, -0.0217, -0.0105, -0.0102, -0.0076,
         0.0123, -0.0238, -0.0258,  0.0025,  0.0068,  0.0186, -0.0015, -0.0208,
         0.0069, -0.0575, -0.0039, -0.0231, -0.0301, -0.0211, -0.0097,  0.0058,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1018, -0.1490, -0.0150, -0.0330,  0.0146,  0.0176, -0.1807, -0.0090,
         0.0122,  0.0094,  0.0018, -0.1532, -0.0099, -0.0161, -0.0173, -0.0148,
         0.0109, -0.0325, -0.0209,  0.0035,  0.0124,  0.0288, -0.0118, -0.0011,
        -0.0582,  0.0012, -0.0352, -0.0121,  0.0097,  0.0061,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9340e-01,  1.2882e-01,  1.3220e-02,  3.0332e-03, -1.7507e-03,
         5.1618e-03,  1.0317e-01,  1.0326e-02, -7.0596e-03, -8.6890e-03,
         3.3281e-03,  1.0546e-01,  1.5127e-03,  1.2798e-02,  6.9745e-03,
         6.2117e-03, -3.7451e-03,  1.9591e-02,  1.6983e-02,  5.2257e-03,
         1.6428e-04, -9.5831e-03, -1.3308e-02, -9.7407e-03, -1.0133e-02,
         9.8170e-02, -6.0598e-03, -5.8087e-03,  8.1154e-03,  1.6816e-02,
         6.2041e-03,  7.5430e-03, -4.5730e-04, -4.9022e-03,  1.6598e-02,
         1.2626e-02,  1.0838e-02, -6.4688e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.3248,  0.0420,  0.0258,  0.0500, -0.0102,  0.1192,  0.0660,  0.0008,
        -0.0207,  0.0323,  0.0842, -0.0101,  0.0284,  0.0482,  0.0513, -0.0418,
         0.0377,  0.0049,  0.0014, -0.0004,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.4674, -0.0119, -0.0559,  0.0076, -0.0539, -0.0600, -0.0314, -0.0159,
         0.0018,  0.0117, -0.0731, -0.0023, -0.0504,  0.0024,  0.0320, -0.0900,
        -0.0192, -0.0062,  0.0070,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0011, -0.0223, -0.0994,  0.0119, -0.0893, -0.0797, -0.0290, -0.0308,
         0.0139,  0.0021, -0.1524, -0.0049, -0.1017, -0.0697, -0.0762, -0.1081,
        -0.0375, -0.0530,  0.0100,  0.0070,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1106, -0.2129, -0.0146,  0.0069, -0.0245,  0.0129, -0.0109, -0.0165,
        -0.0166, -0.0816, -0.0092, -0.0406, -0.1684, -0.0035, -0.0149,  0.0010,
         0.0070, -0.0024, -0.0249, -0.1659,  0.0004, -0.0112,  0.0179, -0.0242,
         0.0009,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7398e-01, -1.2105e-01, -1.2616e-03,  1.9466e-03, -1.8604e-02,
         5.6649e-03, -5.2077e-03, -2.7728e-03, -6.9506e-03, -4.0357e-02,
        -1.4485e-02, -2.3758e-02, -8.9074e-02,  2.5664e-03, -1.8864e-03,
         6.6623e-03, -7.8672e-04,  1.4445e-02, -6.8603e-03, -1.0915e-01,
        -6.2893e-04, -7.7847e-03,  2.1137e-02, -1.1692e-01, -5.4065e-03,
         4.9457e-04,  1.5904e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.4648, -0.1365, -0.0084,  0.0099, -0.0137,  0.0102, -0.0057, -0.0056,
        -0.0128, -0.0354, -0.0104, -0.0317, -0.1314, -0.0039, -0.0055,  0.0005,
         0.0140, -0.0006, -0.0094, -0.0108, -0.0054,  0.0225, -0.0153, -0.0245,
        -0.0033,  0.0078,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #100: [tensor([-0.1717, -0.0232, -0.0106, -0.0995,  0.0272, -0.0131, -0.0069, -0.0580,
         0.0102, -0.0522,  0.0063,  0.0034, -0.1452,  0.0162,  0.0152,  0.0071,
        -0.1650, -0.0678, -0.0076,  0.0281, -0.0363,  0.0112, -0.0180,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1679, -0.0512, -0.0103, -0.1139,  0.0089, -0.0051,  0.0063, -0.0374,
         0.0226, -0.0630, -0.0025,  0.0061, -0.1846,  0.0209, -0.0051,  0.0109,
        -0.1889, -0.0612, -0.0153,  0.0180,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8333e-01, -1.7105e-01, -4.8402e-03, -3.0913e-03,  1.6956e-02,
        -1.8331e-01, -7.9325e-03,  3.7931e-03, -2.2476e-02,  5.2250e-03,
        -6.5569e-03,  5.3120e-03, -1.0313e-02, -7.3424e-03, -5.3360e-03,
         8.0597e-03, -8.5236e-03,  1.0682e-03,  2.3953e-03, -2.3906e-03,
         8.4675e-03, -2.0586e-02, -1.4105e-01, -5.9443e-03,  1.1665e-02,
        -1.4311e-01,  6.2529e-03, -1.5311e-04, -2.4395e-03, -1.0273e-03],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4346e-01, -2.1203e-01, -1.6088e-02, -2.4549e-04,  2.1266e-02,
        -2.0127e-01, -9.0671e-03, -3.3399e-03, -1.9250e-02,  6.6943e-03,
        -4.7066e-05,  3.1664e-03, -2.3119e-02, -8.6130e-03, -1.5880e-02,
         1.2718e-03, -2.0208e-02, -6.2310e-03, -2.4933e-03,  6.6203e-03,
         8.0636e-03, -8.8282e-03, -1.9534e-01, -9.0497e-03,  2.1044e-02,
         5.8718e-03, -1.5888e-02, -4.3901e-03, -8.8102e-03,  2.3488e-03],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.4023e-02, -2.4424e-01, -4.8133e-03,  5.0085e-05,  2.7562e-02,
        -2.0441e-01, -1.1842e-02,  1.6153e-03, -3.0005e-02,  8.3938e-03,
        -3.9806e-03, -1.9879e-03, -1.7936e-02, -6.2861e-03, -1.4485e-02,
         1.2263e-02, -2.2191e-02,  2.6758e-03, -5.6941e-03,  2.0263e-03,
        -2.0119e-01, -1.0343e-02, -1.8452e-02, -2.0830e-03, -6.7966e-02,
         1.5082e-03, -1.0216e-02, -1.7538e-03,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1975, -0.2935,  0.0123, -0.0035, -0.0035,  0.0137, -0.2870, -0.0055,
         0.0020,  0.0056, -0.0728,  0.0079, -0.0104,  0.0033, -0.0537,  0.0027,
         0.0096, -0.0079, -0.0015,  0.0061,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1200, -0.2226,  0.0090, -0.0013, -0.0003,  0.0130, -0.2067,  0.0021,
         0.0034,  0.0029, -0.1917,  0.0011, -0.0182, -0.0045, -0.1815, -0.0164,
         0.0017,  0.0013,  0.0023,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.5799e-02, -3.5933e-01,  1.3060e-02, -1.1718e-02, -5.7046e-03,
         1.6984e-02, -3.1747e-01, -7.5264e-03,  5.9364e-03,  2.7820e-04,
        -3.4217e-02, -4.3188e-03, -2.6930e-02, -2.9525e-03,  1.5736e-02,
        -6.3393e-03,  1.7453e-02, -3.6955e-02,  1.6547e-02,  4.1469e-03,
         5.9972e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1894,  0.3214, -0.0007,  0.0077,  0.0044, -0.0071,  0.3255,  0.0064,
        -0.0158, -0.0078,  0.0727,  0.0006, -0.0063,  0.0107, -0.0162, -0.0038,
        -0.0035,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0453, -0.3604,  0.0112, -0.0059, -0.0052,  0.0230, -0.2386,  0.0027,
        -0.0019,  0.0013, -0.2785,  0.0139, -0.0018, -0.0017,  0.0088,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1021, -0.2630,  0.0100,  0.0024,  0.0051,  0.0220, -0.2758,  0.0037,
         0.0040, -0.0046, -0.2826,  0.0044,  0.0012,  0.0036,  0.0097,  0.0057,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0541, -0.2328,  0.0116, -0.0030, -0.0019,  0.0131, -0.2290,  0.0008,
        -0.0005,  0.0059, -0.2025, -0.0069, -0.0180, -0.0234, -0.0320, -0.0021,
         0.0008,  0.0087, -0.0705, -0.0028,  0.0291, -0.0117,  0.0022,  0.0151,
        -0.0087,  0.0127,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([-2.7261e-02, -4.2716e-01,  6.4156e-03, -1.3332e-02,  1.7594e-03,
        -1.0262e-03, -2.8734e-02,  1.2364e-03,  1.1345e-02, -3.8309e-01,
        -3.0896e-03, -1.3905e-02, -3.1700e-03,  5.7572e-04,  4.1976e-03,
        -7.7819e-03, -6.4099e-04, -7.1103e-04, -8.2698e-03, -2.7839e-02,
         1.1287e-04,  2.1518e-02, -4.4772e-03,  2.3558e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1967e-02, -4.7495e-01,  6.6468e-03, -1.2521e-02,  3.6931e-03,
         3.1329e-03, -2.3333e-02, -5.4177e-04,  1.4355e-02, -3.6238e-01,
         2.3605e-03, -1.5388e-02, -4.0812e-03, -4.9922e-04, -3.0268e-03,
         8.7357e-03, -4.2295e-02,  2.5608e-03, -4.7748e-04, -6.3983e-03,
        -2.0416e-04, -4.5600e-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([-1.6039e-02, -4.2334e-01,  5.6985e-03, -1.1880e-02,  4.9091e-04,
         2.1198e-03, -2.9739e-02,  4.8045e-04,  1.4518e-02, -4.1696e-01,
        -3.9318e-03, -1.4603e-02, -4.0326e-03, -2.0944e-03, -1.9064e-03,
        -4.0500e-03, -5.7722e-03, -4.7726e-04, -1.2706e-04, -6.1811e-03,
        -3.0727e-02,  4.2196e-03,  6.0801e-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([-3.6517e-02, -5.9173e-03, -1.1534e-02, -2.8446e-03, -4.5416e-01,
        -7.4214e-04, -1.5711e-02, -2.7604e-02, -4.8596e-05, -4.3271e-02,
         1.2913e-02, -5.0303e-03, -1.3102e-01, -6.8964e-03,  2.0542e-03,
        -6.9020e-04, -1.5066e-01, -9.2148e-04,  8.7012e-04, -1.8680e-02,
        -1.6272e-02, -2.8078e-02,  8.0676e-03, -1.1061e-02, -6.3966e-03,
        -2.0434e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7791e-03, -2.3207e-03, -7.3161e-03, -4.2019e-03, -4.7187e-01,
        -1.3372e-03, -1.4685e-02, -1.8399e-02,  3.6282e-03, -2.9791e-02,
         7.4289e-03, -1.7098e-03, -1.0438e-01, -5.6872e-03,  1.5362e-03,
         1.2880e-03, -1.3139e-01, -4.0765e-04, -4.9228e-02,  1.3083e-02,
        -2.9651e-03, -1.1794e-01,  3.2030e-05, -1.3858e-05,  4.5967e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.1017e-02, -2.2179e-03, -1.2323e-02, -3.6688e-03, -3.6881e-01,
        -3.5659e-04, -1.4303e-02, -2.2050e-02,  1.2279e-03, -4.1609e-02,
         9.8040e-03, -2.3912e-03, -7.8067e-02, -5.0759e-03, -6.7376e-04,
         4.0327e-03, -1.2517e-01,  1.0918e-03, -3.6730e-02, -7.7965e-03,
        -2.6238e-02,  1.1226e-02, -1.0658e-01, -1.6454e-03, -2.4286e-02,
        -2.3942e-03, -2.1069e-03, -4.6407e-03,  2.4668e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0547,  0.0068, -0.0235, -0.0165, -0.0102, -0.3421, -0.0036, -0.0070,
        -0.0612, -0.0092, -0.0015, -0.0012, -0.0720,  0.0038, -0.0367,  0.0156,
        -0.3275, -0.0052, -0.0006, -0.0009,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0143,  0.0018, -0.0204, -0.0141, -0.0023, -0.2544,  0.0047, -0.0059,
        -0.0621, -0.0059, -0.0012,  0.0005, -0.0515,  0.0018, -0.0246, -0.0138,
         0.0029, -0.2346,  0.0053, -0.0157, -0.0154, -0.2434, -0.0009, -0.0016,
         0.0010,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.0512e-02,  2.2539e-04, -1.8992e-02, -2.0412e-02, -6.4718e-03,
        -2.3088e-01,  2.2590e-03, -3.6299e-03, -6.6066e-02, -1.1680e-03,
         4.4456e-04,  1.8413e-04,  1.1352e-02, -1.2531e-02, -2.8561e-01,
         1.5356e-03, -1.9048e-02, -2.3225e-04,  4.7944e-03, -2.1086e-01,
        -5.9792e-03,  3.9367e-03, -1.3118e-03,  1.5667e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0032, -0.7100,  0.0090, -0.0289,  0.0055, -0.0071, -0.0011, -0.0573,
        -0.1174, -0.0014, -0.0051, -0.0185, -0.0018, -0.0021,  0.0018, -0.0018,
        -0.0023, -0.0071, -0.0131,  0.0055,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0645e-03, -6.1609e-01,  1.0144e-02, -1.8616e-02, -5.2105e-03,
        -2.6592e-03, -7.0311e-04, -5.6660e-02, -7.5420e-02, -1.3381e-02,
        -1.3196e-02, -2.2829e-02, -5.6560e-03,  2.6303e-03, -1.1183e-01,
        -2.9985e-03, -6.3431e-04,  6.3088e-05, -3.2483e-02, -3.8882e-03,
         8.4034e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0433, -0.4738,  0.0020, -0.0149,  0.0006,  0.0047, -0.0051, -0.0465,
        -0.0920, -0.0030, -0.0059, -0.0125, -0.0022,  0.0051, -0.0906, -0.0046,
        -0.0023, -0.0044, -0.0318, -0.0400, -0.0841,  0.0010, -0.0011, -0.0097,
        -0.0159, -0.0030,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 9.2087e-04,  9.2176e-04, -1.1157e-02, -1.4153e-03, -2.0991e-03,
        -6.9773e-03, -4.6698e-01, -2.1754e-03, -1.0724e-03, -5.3089e-05,
         7.7580e-04,  2.9927e-03, -7.9980e-02, -6.0814e-04, -4.2654e-03,
        -4.1248e-01,  2.6781e-03,  1.5785e-03,  8.6921e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4651e-02, -4.7153e-01, -9.5112e-04, -4.9626e-03,  1.4588e-03,
        -1.2498e-03, -5.4886e-03, -3.7278e-04,  4.8828e-03, -6.0378e-02,
         1.2259e-03, -3.9426e-03, -2.5503e-01,  2.9685e-04,  4.3021e-04,
        -1.0766e-04, -6.4463e-02,  4.9839e-05, -3.4700e-03, -2.1658e-04,
        -2.7570e-03, -3.1679e-03,  2.8875e-04, -2.2092e-03,  2.0375e-03,
        -7.2535e-02,  1.5867e-03, -2.6703e-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([ 1.9213e-02, -3.8159e-01,  5.6054e-04, -5.3768e-03, -6.9392e-05,
        -1.9026e-03, -5.2714e-03,  5.2838e-04,  3.3345e-03, -5.1960e-02,
         1.6958e-04, -3.2368e-03, -2.8068e-01,  1.2381e-03, -1.3410e-03,
         1.7208e-04, -9.0211e-02, -1.9856e-03, -1.7254e-04,  1.0076e-03,
         5.1696e-04, -6.4220e-02,  6.5883e-04, -4.2500e-03, -7.7194e-02,
         2.9427e-03, -1.8930e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4940e-02, -4.2836e-01,  2.2679e-03, -3.8216e-03, -5.2570e-04,
         2.2356e-04, -3.8418e-03,  1.1083e-03,  3.6798e-03, -3.4199e-02,
         1.6180e-03, -2.4230e-03, -2.1173e-01,  2.1371e-03, -1.0899e-04,
         1.5618e-04, -1.6689e-01,  8.5562e-04, -3.9529e-03,  1.2713e-03,
        -3.0310e-03,  9.1809e-04, -3.7883e-02,  1.5780e-03, -4.1701e-03,
        -4.6671e-02,  8.5174e-04,  7.7808e-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([-9.5271e-03, -5.2267e-01, -1.8233e-03, -9.4927e-04, -8.4502e-03,
        -3.5136e-03, -4.0425e-03, -3.8639e-05,  7.8024e-04, -4.3373e-01,
        -6.9038e-03, -1.5614e-03, -3.3076e-03, -1.4968e-03, -9.3928e-04,
        -2.6294e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9756e-03, -3.7225e-01, -2.7432e-03,  7.2336e-04, -9.3973e-03,
        -1.3865e-03, -6.4666e-03, -2.8267e-04, -5.1384e-04, -2.7354e-01,
        -2.2398e-03, -1.4237e-04, -3.4896e-03, -6.0073e-03,  4.2741e-04,
        -3.0703e-01, -5.8240e-03, -3.5094e-03,  2.0812e-04,  8.3792e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8500e-03, -5.0903e-01, -2.8524e-03,  1.6127e-04, -7.1450e-03,
        -2.2244e-03, -9.0335e-03, -3.8156e-04, -3.9625e-04, -4.4187e-01,
        -3.3556e-03, -3.9596e-03, -7.3536e-03,  6.0493e-03,  1.7236e-03,
         1.4503e-03,  1.6603e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4698e-02, -5.6101e-01,  5.0834e-04,  1.3275e-03, -5.2126e-03,
        -1.8076e-03,  3.6120e-03,  1.5212e-02, -1.2142e-01, -3.4557e-03,
         1.5277e-03, -2.6044e-01,  1.2731e-03, -3.2382e-03, -2.5640e-03,
        -1.3294e-03,  1.3621e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4361e-03, -8.0915e-01, -1.4576e-03,  1.1518e-03, -8.5436e-04,
        -2.4009e-03,  3.0605e-03,  6.4494e-03, -8.0936e-02, -3.9386e-03,
        -7.1869e-04, -8.0403e-02, -1.4087e-04, -4.0319e-03, -8.4558e-04,
        -8.3647e-04, -1.8931e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.6385e-03, -8.1120e-01,  6.2462e-05,  1.2723e-03, -5.1347e-03,
        -3.0876e-03,  7.1407e-04,  7.4674e-03, -7.0125e-02, -1.2139e-03,
         1.8517e-04, -8.1116e-02, -8.1927e-04, -8.8785e-03,  2.1081e-03,
        -8.6818e-04,  1.0872e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4817e-02, -2.0854e-01,  1.9620e-04,  2.7528e-03, -4.9195e-04,
         1.9113e-03, -7.2326e-04,  1.1108e-03, -9.8513e-04,  3.5614e-03,
        -1.3606e-01,  7.7424e-04,  3.8262e-03,  6.9600e-04, -4.1615e-03,
        -1.7255e-01,  6.5130e-04, -1.9737e-04, -9.4930e-04, -1.8084e-04,
        -1.8151e-01,  2.4095e-04,  1.7953e-03, -6.2856e-04,  2.9499e-03,
         1.2756e-03,  1.5891e-03, -2.3923e-03,  2.4272e-03, -2.4849e-01,
         1.0306e-03,  1.5204e-04, -3.8354e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6829e-02, -2.8186e-01, -4.4631e-04,  1.2707e-03, -8.5508e-04,
         1.6856e-03, -3.7444e-04,  1.0151e-03, -6.6953e-04,  2.0290e-03,
        -1.4561e-01,  5.2384e-04,  2.2753e-03,  1.5998e-03, -1.8510e-03,
        -1.8211e-01,  1.2687e-04,  7.4034e-04,  3.2408e-04, -7.5725e-05,
        -1.3866e-01, -6.8345e-04,  4.4129e-04, -8.1672e-04, -7.9424e-04,
         2.9917e-03,  1.0275e-03, -4.9591e-03, -2.0270e-01, -4.2076e-04,
        -3.1715e-04, -6.8297e-04,  1.2504e-03,  8.8156e-04,  1.0975e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 1.5043e-03, -3.6526e-01,  2.2778e-04, -3.8375e-03,  1.5202e-04,
        -1.1941e-03, -7.5486e-03, -1.0199e-03,  1.3929e-03, -4.2864e-02,
        -1.8805e-03,  2.4634e-03, -8.0330e-04, -7.0289e-03, -2.5207e-02,
         4.6595e-04, -1.0021e-04, -3.1224e-04, -1.9437e-04, -6.6980e-03,
         4.7494e-03, -5.2341e-01, -1.3212e-03,  3.6420e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1175e-03, -3.4681e-02, -6.8415e-04, -2.2946e-03, -8.5022e-04,
        -2.9124e-01,  1.8139e-05, -6.3415e-04, -1.2517e-03, -7.3479e-04,
        -2.4040e-01,  1.3265e-03, -1.0612e-03, -1.0652e-03, -2.4502e-03,
        -2.6452e-04, -9.6964e-03, -2.3558e-01,  5.3838e-05, -1.2464e-03,
        -3.1907e-03, -1.4377e-02,  3.5216e-04, -5.0472e-04, -3.0208e-04,
        -3.6882e-02,  4.8189e-04, -4.2496e-03,  2.2548e-03, -1.0421e-01,
         1.0604e-03, -1.6278e-04,  3.1731e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1989e-03, -3.7423e-02,  1.8094e-05, -1.3641e-03, -7.0558e-04,
        -2.6227e-01, -8.9613e-04, -7.8781e-04, -7.6235e-04, -6.5928e-05,
        -3.6397e-01,  1.9699e-03, -9.1518e-04, -3.2639e-04, -2.5232e-03,
         1.3588e-04, -7.1487e-03, -2.0953e-01, -3.4278e-04, -7.1198e-04,
        -2.2728e-03, -3.3336e-02, -1.9092e-04, -1.4289e-04, -2.6641e-04,
        -6.0624e-02,  1.0722e-03, -1.4637e-03,  5.9107e-04, -7.7852e-04,
         2.5761e-03,  3.9762e-04,  8.5438e-05,  1.4551e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2867e-02, -3.0374e-02, -6.7465e-04, -2.5455e-03, -1.5272e-03,
        -2.6762e-01,  1.0887e-04, -2.2459e-04, -2.8199e-04, -2.4470e-04,
        -2.3549e-01,  1.1383e-03, -1.1173e-03, -1.7449e-05, -2.6702e-03,
        -5.5754e-04, -7.7900e-03, -1.9563e-01, -8.1453e-04, -1.5784e-03,
        -2.6914e-03, -2.2818e-02,  3.0670e-04, -4.5316e-04, -4.7679e-04,
        -7.9163e-02,  1.2639e-04, -2.9026e-03,  3.5581e-04, -1.2558e-01,
        -2.4571e-04, -6.9949e-04,  9.0462e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2556e-03, -3.4077e-01, -4.2128e-04,  2.7839e-03, -1.6229e-01,
        -7.5204e-05, -6.8878e-04, -5.7534e-04, -2.9361e-03, -3.3138e-01,
        -9.6114e-05,  1.6267e-03, -1.3947e-01, -5.1056e-04,  8.6971e-04,
        -3.5145e-04, -1.8770e-03, -7.3125e-04, -7.5390e-04,  5.4848e-05,
         1.2968e-05,  1.6224e-03, -5.5968e-04, -1.2745e-04, -1.1575e-03,
         6.5363e-04, -2.6098e-03, -7.4411e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.0105e-03, -3.0964e-01,  7.3973e-04,  1.8776e-03, -1.9431e-01,
         5.4655e-04, -4.1350e-04, -6.8048e-05, -2.0512e-03, -2.8885e-01,
        -8.2815e-05,  1.7189e-03, -1.8007e-01, -2.0159e-04,  1.5272e-03,
         1.1193e-04, -3.8931e-03, -4.6525e-04, -4.2752e-05,  1.6812e-04,
        -5.6628e-04,  1.2494e-03,  1.6406e-04,  4.2183e-04, -1.4139e-03,
        -1.9446e-04, -6.1962e-04, -5.8784e-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([-1.1614e-03, -2.5776e-01,  2.9175e-04,  2.2872e-03, -1.8555e-01,
        -9.6049e-04, -3.6477e-05, -7.5334e-04, -2.7648e-03, -3.4709e-01,
        -3.3521e-04,  1.3991e-03, -1.8722e-01,  7.1043e-04,  8.6383e-04,
        -5.1321e-04, -1.6283e-03, -6.5834e-04, -2.3219e-04,  4.6027e-04,
        -6.3759e-05,  4.0128e-04, -3.0763e-04,  7.6878e-05,  7.4129e-04,
        -3.4624e-04,  1.2154e-04,  2.0578e-03, -3.0896e-03,  1.0954e-05,
         1.0700e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.4275e-05, -3.0086e-01, -1.9698e-04, -3.0754e-03, -1.7298e-03,
        -9.9747e-04, -4.6653e-02, -3.0279e-04,  1.6415e-03, -3.8615e-01,
        -1.0747e-03, -2.2133e-03, -2.7940e-04, -7.2049e-04, -2.4977e-01,
         2.6741e-03, -5.0375e-04, -2.7611e-04,  7.8699e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4535e-02, -3.5323e-01,  7.7922e-04, -1.7444e-03, -1.1215e-03,
         1.3143e-04, -3.2072e-02, -3.8281e-04,  1.8414e-03, -2.4661e-01,
        -7.6818e-04, -1.0854e-03, -4.3873e-04, -3.2238e-04, -1.2060e-01,
         3.9071e-04, -2.0833e-04,  3.5032e-03, -2.0945e-01,  2.6824e-04,
        -5.1335e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.9179e-03, -3.8213e-01,  1.2111e-04, -1.8624e-03, -1.6190e-03,
        -6.9483e-04, -5.2785e-02, -1.2772e-03,  1.5897e-03, -2.2343e-01,
        -5.9773e-04, -9.8791e-04,  6.8859e-05, -6.5958e-04, -2.9150e-01,
        -5.4649e-06, -4.8820e-04,  6.0100e-04,  1.3965e-03, -2.6533e-03,
        -3.4253e-04,  1.2780e-03, -2.5805e-02, -1.0995e-03,  8.8750e-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([-1.7856e-02,  5.9074e-03,  3.7829e-03, -4.2027e-04, -1.6666e-03,
         7.5473e-01,  1.3816e-04,  5.7679e-04, -3.3261e-04,  1.5636e-03,
        -1.6076e-04,  2.0306e-01,  1.5739e-03,  5.0610e-03, -1.6756e-03,
        -8.7058e-04,  5.9930e-04,  1.7840e-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([-6.9244e-03, -4.0226e-03, -1.7663e-03, -5.7845e-04,  2.1753e-03,
        -5.1644e-01,  7.1244e-04, -7.3150e-04, -1.8790e-03, -1.3840e-03,
        -4.1762e-05, -9.5476e-02, -3.8443e-04,  1.0177e-03, -3.4487e-01,
         1.8074e-03, -1.0355e-02, -1.1320e-03,  4.2491e-03,  2.0474e-03,
         1.7033e-03,  6.7414e-05,  2.3988e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  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([-1.3131e-03, -6.3985e-04,  4.7077e-05, -8.2352e-03,  1.0659e-03,
         2.1990e-03, -6.3368e-01,  2.5852e-04, -2.1516e-04, -3.0256e-01,
        -2.9806e-03, -9.9689e-04,  1.6169e-04, -1.0083e-03,  6.7370e-04,
         1.6769e-04,  5.3319e-04,  1.8740e-04,  2.2592e-03, -3.9332e-02,
        -9.4067e-04, -4.3831e-04, -9.9408e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4616e-03,  1.6408e-04,  3.3937e-04, -6.4053e-03,  7.2648e-04,
         2.0525e-03, -3.0537e-01, -1.8050e-06, -9.2054e-04, -2.2635e-01,
        -1.3001e-03, -1.7810e-05, -4.1858e-04, -2.8351e-05,  9.3828e-04,
        -4.1871e-04, -2.1347e-01, -2.1197e-03,  8.6142e-04,  2.5602e-03,
        -2.3218e-01, -5.2263e-04,  3.6420e-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,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2201e-04,  1.8522e-04,  9.2765e-04,  4.1063e-04, -6.1209e-04,
        -7.7338e-01, -4.0260e-04,  3.5382e-04, -1.9291e-03,  1.9611e-03,
        -1.2521e-03, -7.2132e-02,  6.5614e-04, -2.3197e-04, -1.8833e-03,
        -1.2611e-01, -1.9443e-04, -1.3531e-02, -2.5011e-03,  2.3704e-04,
        -6.7882e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6768e-02, -2.2896e-03,  1.4856e-03, -1.5898e-04,  7.0022e-04,
         8.2814e-01,  1.0480e-03, -1.7234e-03,  3.2550e-03, -1.9117e-03,
         1.4480e-03,  9.2168e-02, -9.2746e-05,  9.9446e-04, -3.6925e-04,
        -1.0527e-03,  4.4770e-02, -6.9718e-04, -4.0731e-04, -5.1633e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.0346e-04,  4.5307e-04,  1.4918e-05,  5.0867e-04,  5.9150e-04,
        -7.4355e-01,  3.0458e-05,  1.7369e-03, -2.3571e-03,  9.7493e-04,
         7.7528e-05, -1.1288e-01, -2.0540e-04, -1.1926e-03, -6.0257e-04,
        -1.1337e-01, -6.7596e-04, -1.4452e-02,  2.6892e-03, -2.4806e-03,
         6.5687e-05, -6.7733e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7027e-03, -1.8447e-01, -9.1741e-05,  7.0248e-04,  1.5215e-03,
        -2.9360e-01, -2.6954e-04,  8.9962e-04,  6.3228e-04,  6.1462e-04,
        -9.3924e-04, -1.6761e-04,  9.2605e-04,  5.9981e-04,  1.8009e-04,
        -1.4572e-04,  1.1912e-03, -8.7276e-04, -3.8263e-04,  1.0050e-03,
         1.0677e-03, -2.0330e-01, -4.9107e-04,  1.9264e-04, -6.2265e-05,
        -2.4017e-02,  1.1750e-03, -1.3302e-03,  2.3507e-04, -2.7184e-01,
         2.8806e-04, -1.5275e-04,  7.7120e-04,  1.7071e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.5825e-03, -3.2883e-01,  3.8039e-04,  1.8166e-05,  1.0092e-03,
        -3.5140e-01,  6.1838e-04,  9.9011e-04, -3.6979e-04,  1.8127e-04,
        -1.2942e-03, -1.9926e-04,  7.5452e-04, -1.9565e-03, -2.9004e-04,
        -8.7914e-04,  8.1776e-04, -5.7754e-04, -3.1429e-04,  3.2529e-04,
         1.4039e-03, -1.4925e-01, -1.0230e-03, -3.0199e-04,  4.4383e-04,
        -4.0684e-02,  6.6671e-04, -1.2289e-03,  1.1419e-03, -1.0241e-01,
        -1.6113e-04,  3.1879e-04, -1.7355e-03,  6.5370e-04,  2.2673e-04,
        -5.2404e-04, -2.8835e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4860e-03, -2.9800e-01,  8.1350e-05,  3.4378e-04,  7.9267e-04,
        -2.9503e-01,  1.8197e-04,  5.1955e-04,  3.9316e-04, -1.4523e-04,
        -1.8181e-03, -4.2317e-04,  4.0598e-04, -1.2712e-03, -7.2661e-04,
        -2.0107e-03,  9.0078e-04, -4.6860e-04, -7.6341e-04,  4.0010e-04,
         8.1795e-04, -3.1157e-01, -8.2753e-04,  1.9458e-04, -2.6693e-04,
        -2.8951e-02,  3.4015e-04, -1.8962e-03, -4.4063e-04, -1.3834e-04,
        -5.4926e-04, -4.6421e-02,  8.3505e-05, -2.7254e-04,  6.8293e-05,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5114e-02,  1.4280e-01,  9.7052e-05,  1.1284e-02,  2.9844e-03,
         1.8265e-03,  2.3123e-03, -1.8420e-03,  7.8115e-01, -1.4421e-03,
         5.8552e-03,  1.7850e-03, -9.2718e-04,  9.0648e-04, -4.4576e-04,
        -6.8796e-04,  6.8789e-03,  3.9078e-03,  3.2539e-03,  3.9078e-04,
        -8.9181e-04,  7.9683e-04,  2.4163e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.5647e-04, -5.3576e-02, -1.3101e-03, -3.5599e-03, -2.3537e-03,
        -7.7940e-04, -7.8952e-04,  6.9403e-04, -6.3490e-01,  1.7080e-04,
        -2.3607e-03,  2.9516e-04,  7.6960e-04,  2.7611e-04, -1.1406e-05,
        -6.1961e-02, -1.3306e-03, -5.4612e-04, -2.2540e-03,  5.6884e-04,
        -2.2739e-04, -1.0449e-03, -1.3853e-03,  4.8551e-04, -3.8565e-02,
         2.5253e-04,  1.8800e-04, -2.6720e-03, -3.1799e-03, -1.8082e-01,
         5.6504e-04, -3.2421e-04,  4.4143e-04,  3.2356e-04,  2.6257e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1741e-02, -3.9409e-02, -6.0408e-04, -4.7965e-03, -2.1656e-03,
        -6.9787e-04, -1.4663e-03,  8.2263e-04, -6.8761e-01,  9.0847e-04,
        -2.5972e-03, -2.8712e-04,  4.5198e-04, -1.0097e-03, -1.6854e-04,
        -1.0757e-01, -8.5068e-04,  1.6321e-03, -1.6470e-03, -5.1988e-02,
        -9.0568e-04, -5.5738e-04, -1.6406e-03,  4.2592e-04, -7.7279e-02,
        -1.4790e-04,  4.8895e-04,  1.2211e-04,  8.1457e-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([ 6.5507e-03, -4.0547e-01, -9.9734e-04, -1.6017e-03, -2.4594e-01,
        -8.9989e-04, -2.9673e-04, -3.0379e-03, -3.0652e-02, -3.7014e-04,
        -5.2678e-05,  2.9796e-04, -1.2981e-01, -6.5973e-04, -2.4841e-03,
        -8.7462e-04, -1.6724e-01, -2.0073e-03,  3.9164e-04,  2.8124e-04,
         8.0649e-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>)]
Iter #350: [tensor([ 8.4735e-03, -1.5047e-01, -8.5245e-04,  2.1643e-03, -1.8514e-01,
        -3.2955e-05,  2.8448e-04, -8.1224e-04,  5.7436e-04,  1.3253e-03,
         7.3871e-04,  9.2695e-04, -1.1429e-03, -3.4774e-03, -1.5175e-03,
        -5.5945e-04,  5.7032e-04, -2.8841e-01, -2.8887e-04, -1.0654e-03,
         7.8462e-05, -3.4808e-01,  6.7450e-04,  6.6198e-04, -6.7936e-04,
         4.5684e-04,  5.4943e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0127e-02, -3.0519e-01,  2.3239e-04,  1.5071e-03, -1.1422e-01,
        -3.4147e-04,  5.9920e-04, -2.4077e-04, -4.9037e-04,  7.5331e-04,
         1.2917e-05,  2.8171e-04, -2.3628e-04, -7.7355e-04, -1.8544e-03,
         9.8852e-05,  6.0569e-04, -2.4436e-01,  5.1799e-04, -1.1084e-03,
        -6.9678e-04, -1.3323e-01,  4.1466e-04,  7.8501e-04, -3.0114e-04,
         9.8802e-04, -1.7895e-01,  3.8347e-04,  5.1639e-04,  4.9401e-05,
         1.4637e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.6057e-03, -4.6313e-05, -1.3319e-04, -3.8566e-04, -1.7868e-03,
        -1.5786e-03, -3.0732e-01, -3.4616e-06,  1.7660e-03, -3.9666e-01,
        -2.0157e-04,  2.3099e-04, -1.6332e-04, -4.5027e-04, -5.3826e-04,
         1.1146e-04, -1.9951e-03, -2.7972e-01,  2.2440e-04,  7.0319e-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([-1.5887e-03, -1.8269e-04,  7.9039e-04, -2.5674e-04, -1.4389e-03,
        -6.1349e-04, -5.6802e-01, -9.5238e-04,  9.4520e-04, -3.0022e-01,
        -7.5110e-04,  6.9622e-04,  9.3183e-04, -1.1514e-01, -1.3799e-03,
        -4.3224e-03, -1.6453e-03, -1.2948e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.9398e-03,  4.6580e-04,  8.9594e-04, -9.7459e-04, -1.1150e-03,
        -7.2743e-05, -6.4626e-01, -2.3269e-04,  2.7180e-05, -2.0380e-01,
        -4.7822e-04, -8.2259e-05,  9.3911e-05, -1.3073e-01, -1.3154e-03,
        -4.0711e-03, -1.3337e-03,  4.6892e-04, -7.2909e-04,  9.0888e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1572e-02, -8.5593e-04, -1.8427e-03, -8.9994e-04,  8.5937e-04,
        -4.1919e-01,  4.0200e-04,  1.9564e-04, -1.0932e-03, -3.8389e-01,
         1.1805e-03, -1.6935e-04,  2.7408e-04, -1.6720e-01,  1.1445e-04,
        -1.0837e-03, -8.1132e-03,  2.4485e-04,  5.2360e-04,  2.9221e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6794e-03,  2.3995e-04, -1.0337e-03, -7.0590e-04,  1.4524e-03,
        -2.0525e-01,  2.5549e-04,  1.1101e-03, -1.9031e-03, -5.7144e-01,
         5.1097e-04,  1.4349e-04,  5.6138e-05, -1.9779e-01, -4.9491e-04,
        -1.9771e-04, -1.1581e-02,  5.3654e-04,  3.9625e-04,  2.1996e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.3100e-03,  2.5913e-04, -1.2829e-03, -1.3839e-03,  7.1165e-05,
        -2.6313e-01,  2.5618e-04, -4.8304e-04, -8.7631e-04, -5.9183e-01,
         9.1864e-04, -3.9093e-04,  1.4341e-04, -1.1726e-01,  4.7938e-04,
        -1.0003e-03, -1.1231e-02, -5.2199e-04,  1.5968e-04, -3.1882e-04,
        -1.9204e-06,  6.8242e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3890e-03, -2.8905e-01, -3.0294e-04, -6.3126e-05, -2.9911e-04,
        -1.0634e-03, -7.2090e-04, -1.2341e-01, -8.1083e-04,  3.6182e-04,
        -8.9206e-02, -2.7220e-04, -7.0075e-04, -1.0614e-01, -3.9695e-04,
        -1.3832e-03, -8.9484e-05, -6.7965e-04, -6.0510e-04,  2.0121e-05,
        -3.1071e-01, -3.9906e-04, -2.4207e-04,  4.1073e-04,  4.8825e-04,
         1.5652e-04, -3.1178e-04, -6.6175e-04, -6.5311e-02, -2.2608e-04,
         1.2806e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1194e-02, -3.6042e-01, -4.0521e-04, -4.6407e-04, -9.1742e-04,
        -6.9569e-04, -4.2447e-05, -4.8229e-02, -7.4874e-05,  6.1347e-04,
        -1.0900e-01,  2.0509e-04, -9.6869e-04, -4.1784e-02, -3.4185e-04,
        -1.4305e-03,  7.4592e-04, -1.0452e-03,  2.1128e-04,  1.3716e-04,
        -3.4607e-01, -1.1335e-04,  6.3177e-04, -1.3425e-04,  7.9917e-05,
        -3.6105e-05, -1.6734e-04, -7.3605e-02,  1.2648e-04,  1.0152e-04,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.8842e-03, -4.2400e-01, -5.2847e-04, -2.9830e-04, -1.7150e-04,
         6.6717e-05, -5.9855e-05, -7.5610e-02, -6.9251e-04,  1.3567e-04,
        -1.5984e-01, -5.3067e-04, -1.2626e-03, -2.5535e-02, -7.4167e-04,
        -1.4435e-03,  2.4599e-04, -1.0604e-04, -2.2419e-04, -5.1958e-05,
        -1.7571e-01, -2.6184e-05, -4.2356e-04, -3.8798e-04,  6.1726e-05,
        -2.4917e-04, -1.2543e-01, -2.5902e-04,  3.0890e-05,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0617e-02, -8.2807e-04, -2.1638e-03,  2.2392e-03, -1.5995e-04,
        -3.1745e-04, -1.3022e-01,  8.0735e-04,  3.2002e-03, -5.6172e-02,
         2.2643e-03,  8.6774e-05,  5.1535e-04, -1.3984e-01, -9.9908e-04,
        -9.4086e-05, -6.3678e-01,  1.2407e-03, -5.6189e-04,  5.7355e-04,
         1.6450e-04, -1.5585e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-2.9874e-03,  4.6871e-04, -1.2408e-04,  1.9373e-04,  3.3235e-05,
        -4.7533e-05, -1.7023e-04, -3.5191e-01, -4.8666e-05, -4.1328e-04,
         5.1795e-04, -2.7606e-01, -1.0434e-04,  1.3239e-04, -2.6168e-03,
        -2.2064e-05, -1.5583e-04, -2.3755e-04,  4.8520e-04,  3.6313e-04,
         6.3502e-05,  5.4668e-06,  1.0482e-04,  1.0041e-05, -1.9992e-01,
        -3.8819e-04, -5.2643e-04, -1.4147e-04, -1.5914e-01,  5.0546e-05,
         4.8621e-04, -2.5829e-04, -2.4918e-04, -4.6952e-04, -2.2135e-04,
        -7.0522e-04,  1.4582e-04,  2.6897e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.8723e-03,  2.3131e-04,  6.7661e-04,  9.0707e-05,  2.6561e-04,
         7.6759e-04,  7.3414e-05, -5.9990e-01, -2.2573e-06, -6.4379e-04,
         1.0068e-03, -3.4150e-01,  1.8311e-04,  5.1247e-04, -1.6628e-03,
        -7.3671e-04,  3.3283e-04,  1.1902e-04,  3.3577e-04,  8.7588e-04,
         1.3957e-05,  8.9156e-04, -4.0497e-02,  2.4372e-04, -2.2174e-04,
         4.8045e-04,  3.7653e-04,  2.2862e-04, -2.5267e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+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-04,  1.6350e-04, -1.1224e-04, -1.5423e-04,  2.0056e-04,
        -3.9288e-04,  6.5874e-04, -8.3968e-01, -2.9594e-04, -5.5702e-04,
         2.2925e-04, -1.3873e-01, -5.3838e-04, -6.1479e-04, -6.3529e-03,
        -1.1128e-03, -7.7466e-05, -9.6942e-05,  1.0116e-05,  2.0357e-04,
        -2.6890e-04, -7.3558e-03, -1.4972e-03,  1.0191e-04, -4.6893e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9418e-02, -3.8704e-06,  3.1506e-03,  8.0785e-03, -1.7721e-04,
         4.1959e-04, -1.6692e-03, -1.3534e-04, -1.4504e-03,  8.7252e-01,
         1.6401e-03,  7.1808e-04,  7.7808e-03, -1.0639e-04,  2.2610e-02,
         6.7786e-04,  1.9596e-03,  4.0793e-03, -6.3828e-04,  3.2273e-04,
        -7.6877e-04,  1.9388e-03,  2.8531e-02,  5.4691e-04, -4.4451e-04,
         2.1804e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7639e-02,  5.9162e-04,  4.7956e-03,  2.3414e-02,  5.0250e-04,
         4.4924e-04, -2.5353e-03, -2.6409e-04, -1.4023e-03,  7.4822e-01,
         7.1907e-03,  1.5305e-03,  9.6179e-03,  2.3894e-03,  3.0498e-02,
        -4.5257e-04,  4.7792e-03,  2.6882e-03,  3.7929e-04,  2.7167e-04,
         8.2705e-04,  3.0024e-03,  6.7440e-02, -9.7291e-04,  9.6081e-03,
         3.9028e-04,  4.0390e-02,  4.6520e-03,  2.7776e-03, -1.1482e-04,
        -2.1965e-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([ 7.7315e-02, -2.2965e-03,  4.2059e-03,  7.8684e-03, -6.3917e-04,
        -1.8014e-03, -2.7511e-03, -3.0864e-03, -2.7114e-03,  7.7801e-01,
         3.2235e-03,  3.2228e-04,  6.6057e-03,  2.3123e-03,  1.5426e-02,
         8.6020e-05,  9.9897e-03,  1.9819e-03, -1.7355e-04, -6.5296e-04,
         3.8756e-04,  3.6079e-03,  6.7253e-02,  4.0606e-04,  1.7146e-03,
         3.2276e-03,  1.1658e-03,  7.7419e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0046,  0.0199,  0.0319, -0.0093,  0.0185,  0.0597, -0.0045,  0.1269,
         0.0185,  0.0006,  0.0630,  0.0300, -0.0153,  0.0095,  0.0392,  0.0017,
         0.2898,  0.0217,  0.1999,  0.0174, -0.0027,  0.0029,  0.0123,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1714,  0.0305,  0.0268, -0.0122, -0.0117,  0.0711,  0.0008,  0.1702,
         0.0523,  0.0238,  0.0730,  0.0250,  0.0057, -0.0097,  0.0126,  0.1081,
         0.0355,  0.0010,  0.0024,  0.0679,  0.0828, -0.0027,  0.0027,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1456,  0.0327,  0.0333,  0.0002,  0.0471,  0.1026, -0.0024,  0.0831,
         0.0312,  0.0476,  0.0788,  0.0833, -0.0175,  0.0087,  0.0895,  0.0562,
        -0.0190,  0.0121,  0.0382,  0.0666,  0.0017, -0.0027,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6925e-03, -5.5902e-02, -8.5416e-04, -1.1631e-03,  2.4172e-03,
        -6.0426e-02, -9.3754e-04,  8.4082e-05, -8.5900e-04,  7.9979e-04,
        -5.9444e-01, -3.0405e-04, -1.0426e-03,  2.7144e-03, -2.7492e-01,
        -4.1277e-04, -2.5261e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.5615e-03, -1.3724e-01,  6.3628e-04,  1.1610e-03,  1.6148e-03,
        -4.7192e-02, -1.4193e-03, -6.4491e-04, -6.0399e-04,  9.8560e-05,
        -7.9734e-01, -4.2508e-04,  2.1900e-04,  2.2397e-03, -6.0145e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3233e-03, -1.8116e-01, -8.4762e-04, -5.7457e-04,  3.6534e-03,
        -1.2816e-01, -7.2495e-04,  2.5638e-04, -1.0329e-03, -1.4620e-05,
        -6.7987e-01,  1.5830e-03,  5.6805e-05,  7.4845e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 3.7203e-03, -6.9169e-04, -1.2460e-03, -1.4512e-05,  1.0511e-03,
        -1.1749e-03, -4.8693e-01,  7.9261e-04, -1.4028e-03,  1.8090e-04,
        -7.2933e-05, -5.5345e-05, -7.6168e-04,  7.5413e-04, -6.5653e-04,
        -1.7405e-03, -6.9124e-04,  1.6023e-03, -4.9559e-01,  1.9118e-04,
         2.9751e-04,  3.8370e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9902e-03, -1.9757e-03, -6.4268e-04, -7.3400e-04,  4.1378e-04,
        -4.5396e-03, -9.7883e-01, -7.4320e-04, -1.9126e-03, -4.2987e-04,
         1.9787e-04, -1.6469e-04, -8.0612e-04,  1.8514e-04, -2.3769e-04,
        -1.2062e-03,  7.8207e-04, -8.4047e-05, -1.7545e-04, -1.0801e-03,
         9.2881e-05, -5.6903e-04, -1.5724e-03,  5.1042e-04, -2.1953e-05,
         5.4785e-05,  5.1683e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3313e-03, -1.7164e-03, -8.3680e-04, -9.4806e-05,  5.9699e-04,
        -3.6555e-03, -9.7786e-01, -4.5948e-04, -1.1570e-03, -1.2390e-04,
         3.7586e-04,  2.7396e-04, -1.3486e-03,  6.7507e-05,  4.8020e-04,
        -1.3815e-03, -1.0085e-04, -2.1179e-03,  8.3128e-04,  6.0350e-04,
         5.2514e-04,  9.3496e-04, -1.2449e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2500e-03, -3.9092e-01,  1.6594e-04, -3.2840e-04, -1.8391e-01,
         7.5770e-05, -3.5386e-04, -1.1551e-03, -1.4761e-03, -1.5875e-01,
        -4.8574e-04, -3.2722e-04, -6.2725e-05, -7.6027e-02,  2.2689e-04,
        -1.1287e-03, -5.4108e-04, -3.7531e-05, -1.8148e-01,  2.1006e-04,
         3.1076e-05, -6.0864e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5802e-04, -4.2269e-01,  2.4782e-04,  5.8600e-05, -1.1179e-01,
         4.9438e-04,  1.7138e-04, -1.1247e-03, -6.2691e-04, -3.3915e-01,
         3.6426e-04,  6.1721e-05,  1.3975e-04, -2.3890e-02,  1.4242e-04,
        -1.1137e-03, -2.1920e-04, -9.5887e-02, -2.7667e-04, -5.4175e-04,
        -2.5499e-05, -6.3473e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8089e-03, -3.9075e-01,  1.0570e-04,  1.7916e-04, -8.0652e-02,
         1.2021e-03,  2.6747e-04, -1.6466e-03, -6.5211e-04, -1.1791e-01,
         3.3921e-04, -2.3508e-04,  2.7304e-04, -3.1332e-01,  4.8165e-04,
        -1.3689e-03, -1.0951e-03, -8.4577e-02,  5.3193e-04, -6.0180e-04,
         4.9186e-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([ 6.0482e-03, -4.9122e-01,  3.6761e-04, -5.7208e-04, -3.3291e-04,
        -6.8717e-03,  1.2532e-04,  5.2301e-04, -2.0055e-01, -6.3254e-04,
         5.5198e-04, -4.4495e-05, -2.0530e-01, -1.1548e-03, -1.8003e-03,
         4.4036e-04,  1.1196e-03, -8.1408e-02,  2.7056e-04, -2.2909e-04,
        -4.4446e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.7138e-03, -5.0640e-01,  4.2594e-04, -3.6019e-03, -9.2942e-07,
        -2.4904e-02,  6.9256e-04,  1.3894e-03, -1.9545e-01, -4.0291e-04,
        -1.7915e-04,  6.7860e-04, -2.4160e-01, -2.7379e-04, -1.4252e-02,
         7.6814e-04,  9.3839e-05,  1.7458e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2398e-05, -5.5458e-01,  1.0723e-04, -3.0652e-03, -8.6595e-05,
        -4.2127e-03,  7.7737e-04,  1.9422e-03, -2.7436e-01, -4.3593e-04,
        -4.3591e-04,  5.2574e-04, -1.5657e-01, -1.3246e-04, -1.7279e-03,
        -1.5356e-04, -7.1477e-04, -8.5720e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0234e-03, -4.0004e-01, -1.4975e-04,  2.8786e-04, -1.0621e-03,
        -7.8045e-04, -5.4066e-04, -4.2619e-01,  6.1541e-04,  1.1972e-04,
         1.4461e-04, -4.9451e-02, -1.6088e-04, -3.3008e-04, -1.1522e-03,
        -7.6082e-04,  3.2882e-05, -1.1341e-01, -5.3459e-04, -5.0910e-04,
        -6.9861e-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([ 4.1156e-03, -2.8857e-01, -4.1418e-04, -5.9320e-04, -1.1482e-03,
        -7.7702e-04, -1.8689e-04, -5.6191e-01, -1.2363e-03,  1.2826e-04,
         1.2617e-04, -1.3805e-01, -3.9512e-04,  7.0990e-04, -1.0109e-03,
         1.4893e-04, -6.0100e-05,  5.1141e-06,  1.9324e-04, -2.1787e-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([-1.7745e-03, -2.7860e-01,  3.1182e-04,  9.2497e-05, -8.7419e-04,
        -5.8157e-04,  1.0074e-03, -5.8577e-01, -4.4126e-04,  2.6208e-04,
         4.4050e-04, -1.2753e-01,  2.4586e-04,  3.6657e-04, -8.4555e-04,
         7.4525e-05,  7.4957e-05, -6.1316e-04, -9.2273e-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 #500: [tensor([-4.9635e-03, -5.0905e-01,  1.9050e-04,  1.5489e-04,  3.0466e-04,
        -2.7343e-01,  3.0510e-04, -1.1993e-04, -5.6794e-04,  1.2901e-04,
         6.2149e-04, -6.2691e-02,  3.7258e-04, -7.7592e-05,  2.0721e-05,
        -8.1206e-02, -3.9464e-05, -2.6969e-04, -3.9469e-05,  2.1760e-04,
        -2.5155e-04,  1.9485e-04, -6.9459e-04, -8.1963e-05, -6.1778e-02,
         2.5432e-04, -2.0995e-04, -4.9473e-04, -5.5964e-04, -1.3159e-04,
        -1.2528e-04,  1.2804e-04, -2.3419e-04,  9.3967e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3789e-03, -2.5840e-01, -2.3923e-04,  4.2126e-04,  8.7089e-04,
        -3.3191e-01, -3.3452e-04,  6.8920e-05, -9.0606e-04, -8.0440e-05,
         9.5561e-04, -9.6082e-02,  6.6360e-04,  1.1852e-04, -2.1953e-04,
        -3.0181e-01,  3.2290e-05, -1.0038e-05, -2.7051e-04,  5.8729e-04,
        -5.5762e-04,  8.2907e-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([-5.2365e-03, -9.9380e-02,  2.6682e-04,  9.5394e-05,  2.0423e-03,
        -6.7154e-01,  6.1161e-04,  6.7114e-04, -5.8885e-04,  2.6019e-04,
         1.3920e-03, -1.0681e-01,  6.8126e-04,  4.9792e-04, -2.1937e-04,
        -1.0785e-01, -1.8848e-04, -1.6218e-04, -3.7279e-04, -5.0974e-04,
        -1.7372e-04,  3.5500e-04, -8.7903e-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([-8.5100e-03, -3.9669e-01,  1.1160e-05,  3.5425e-04, -2.6136e-04,
         9.4884e-04,  1.8104e-04, -1.4481e-01, -7.3549e-04,  1.1860e-04,
         4.1212e-04, -4.2294e-01, -6.9019e-04,  7.5166e-04,  7.4702e-04,
        -2.0371e-02, -5.7598e-04, -6.6935e-04,  2.2226e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0923e-03, -6.7273e-01, -3.6558e-04, -1.4967e-04,  2.9734e-04,
         1.6375e-05, -7.2088e-04, -2.6771e-02,  1.4269e-04,  1.6555e-04,
         1.1356e-04, -1.6388e-01, -1.0761e-03, -1.9205e-05, -9.4312e-04,
         1.3981e-03, -1.2897e-01,  7.1555e-04, -1.4260e-04,  2.8791e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.9408e-04, -4.7834e-01, -8.5215e-04, -5.1389e-04, -3.8099e-05,
         1.7936e-04, -1.0029e-03, -3.1736e-01, -7.2589e-04,  1.0114e-03,
         1.2476e-04, -1.9633e-01,  8.9343e-04,  6.2894e-04,  4.7830e-05,
         1.1589e-03,  9.4990e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9812e-04, -4.6839e-04, -4.3726e-04, -3.4157e-04,  4.2109e-04,
        -2.7604e-01, -6.7363e-04, -7.5111e-04, -1.7342e-04, -1.8379e-01,
         3.7901e-04,  1.4567e-04,  2.9669e-04, -3.3375e-01, -1.1502e-04,
        -1.2361e-03, -5.4912e-04, -1.9818e-01, -4.9113e-04, -4.9819e-04,
        -4.5908e-04,  2.0416e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5482e-03, -2.4732e-04, -3.1292e-04,  4.4290e-04,  8.9941e-04,
        -2.2976e-01, -7.5606e-04, -5.8743e-04,  2.1743e-06, -4.2261e-01,
         1.6508e-04,  3.5838e-04,  2.0651e-04, -2.6597e-01,  2.4545e-04,
        -1.1066e-03, -3.3807e-04, -7.1254e-02, -4.1291e-04, -7.7835e-04,
         3.8233e-04, -3.5144e-04, -2.6990e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3895e-03, -1.4438e-04, -4.6212e-04,  1.2100e-04,  6.3486e-04,
        -3.7661e-01, -5.4212e-05, -6.6023e-04,  8.4538e-05, -1.9005e-01,
         3.8542e-04,  1.6016e-04,  4.1929e-04, -2.2826e-01,  3.0597e-04,
        -9.8552e-04,  3.7473e-04, -1.9593e-01, -2.8474e-04, -7.1696e-04,
         3.0154e-04, -8.5066e-05, -6.9347e-04,  3.7950e-04,  2.8566e-04,
         2.1352e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.6058e-03,  1.7261e-03,  8.4156e-04, -3.4981e-04, -3.6733e-04,
        -1.5473e-03, -2.9087e-04,  1.1610e-03, -3.6517e-04,  6.1265e-04,
        -1.4891e-04,  9.7306e-01,  3.1369e-04, -5.6939e-04,  4.2986e-04,
         2.0687e-03,  9.2774e-04,  8.0415e-04,  2.7488e-03,  9.2312e-04,
         5.9035e-05, -6.3734e-04,  3.8025e-04, -6.2350e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8031e-03, -1.2854e-03, -1.4073e-03, -3.4543e-04,  7.2716e-05,
         5.3699e-04, -7.5880e-04, -2.3449e-03, -5.5153e-04, -1.4550e-03,
        -6.8343e-04, -6.9992e-01, -4.7455e-04, -6.5071e-05,  7.2390e-05,
        -1.4546e-03, -5.0354e-04,  4.3473e-04, -1.9377e-03, -1.4117e-04,
        -4.0838e-04, -2.8185e-01,  1.9861e-05,  2.4814e-04, -2.2414e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3617e-03, -1.8853e-03,  3.5483e-04,  5.6751e-04,  4.1656e-05,
         8.9783e-04,  1.0383e-04, -1.4259e-03,  6.0841e-05, -2.0053e-03,
        -1.5117e-04, -6.3077e-01,  7.5514e-04,  7.8786e-05,  6.9262e-04,
        -3.5423e-01, -8.5952e-05, -2.4979e-03,  2.9013e-04, -4.2292e-04,
        -3.0175e-04, -1.5109e-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>)]
Iter #550: [tensor([-8.3416e-04, -1.8921e-01,  2.8075e-04,  4.9915e-04,  4.7064e-04,
         4.6208e-04, -2.1137e-04,  3.7047e-04,  6.0016e-05, -2.3247e-01,
         6.0895e-06,  1.3262e-04, -5.0675e-04,  6.7159e-04,  4.7119e-04,
        -2.4725e-04, -3.3826e-04,  1.7894e-04, -1.5982e-01,  5.6931e-04,
        -8.5667e-04, -3.3084e-01, -8.7139e-05, -3.2141e-04, -7.9541e-02,
         7.3554e-05,  8.0648e-07, -4.6450e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2129e-03, -4.8095e-01, -8.1356e-05,  6.2212e-04,  8.1955e-04,
        -4.4475e-05, -7.0565e-04, -3.5059e-04, -1.1012e-04, -3.6069e-01,
        -1.0117e-04, -8.4982e-05,  1.6515e-04,  3.7263e-04,  2.0798e-04,
        -9.9514e-05,  1.6824e-04,  4.2129e-04, -1.4921e-01,  2.2097e-04,
         4.6342e-06, -4.4853e-04, -7.4961e-05, -3.4593e-04,  2.1117e-04,
         4.2147e-05,  2.4014e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.9335e-03, -2.3187e-04,  1.4047e-03, -2.2653e-03,  9.8114e-01,
        -8.5926e-04,  1.1097e-03,  7.0272e-04, -2.8487e-04,  1.2306e-03,
        -7.2906e-04, -1.3642e-03, -1.1797e-03,  5.6445e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9845e-03,  2.6595e-05, -3.1278e-04, -9.2652e-04,  9.8601e-01,
         7.5783e-04, -5.2368e-04, -5.8859e-04,  1.0558e-03,  8.4467e-04,
         1.2203e-03,  7.0001e-04,  9.6448e-04, -7.9711e-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([-5.0029e-03, -2.0202e-04,  2.2303e-03, -4.0395e-03,  9.7658e-01,
         4.4418e-03,  1.6660e-03, -9.5087e-04, -1.2821e-04, -4.4726e-04,
        -2.4197e-03, -8.4158e-04, -7.9757e-04,  9.2860e-05, -1.5966e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4548e-03, -5.2628e-01, -2.1060e-04, -3.5051e-04, -1.2881e-01,
        -6.0190e-05, -7.8748e-04, -3.7884e-04,  8.1110e-05, -1.6885e-01,
        -1.2317e-04, -1.0106e-03,  1.6578e-05, -1.7008e-01, -5.0060e-04,
         1.5323e-04, -6.5713e-04, -6.6795e-05,  1.3320e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7257e-03, -5.1133e-01,  6.3849e-04,  8.2601e-04, -3.2404e-01,
         5.5528e-04, -1.1101e-03, -2.0253e-04, -3.2202e-04, -1.5395e-01,
        -9.7311e-04, -7.8163e-04, -5.5860e-04,  4.8718e-04,  2.1730e-04,
         2.7542e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7156e-03, -3.3403e-01, -3.7407e-04,  1.2649e-03, -3.1774e-01,
         3.0659e-04, -1.0892e-03, -4.3737e-04, -5.9517e-05, -3.3801e-01,
        -6.1095e-04, -2.0163e-03, -1.4704e-03, -2.5344e-04,  1.1381e-05,
        -5.6522e-04, -3.6594e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8693e-03,  8.4474e-02, -6.8940e-04, -3.3967e-04, -2.8239e-04,
        -9.8401e-04, -3.9185e-04, -1.0152e-03,  1.4133e-03,  3.4211e-04,
        -4.4110e-04, -3.1694e-05,  3.7499e-04,  7.3065e-01, -3.8384e-05,
        -3.7761e-04, -2.6069e-03,  1.7010e-01, -1.3139e-03, -1.1263e-04,
         1.4999e-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([ 5.6738e-03,  6.8553e-01, -5.9981e-04, -1.6590e-04,  3.2026e-04,
        -7.2282e-04, -8.9139e-05, -3.7377e-06,  1.0904e-03, -2.0931e-05,
         2.8576e-06, -1.5165e-04, -2.6422e-04,  6.3733e-04, -3.1981e-04,
         4.0883e-04, -6.6678e-04, -5.0505e-05,  3.0119e-01, -1.1108e-04,
         2.2759e-04,  7.4442e-05,  1.8942e-04,  1.1310e-03,  4.9557e-05,
         4.8741e-05, -2.5325e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2877e-03, -2.7079e-01,  1.0259e-03,  3.3771e-04, -5.8165e-04,
         5.7584e-04,  6.4966e-05,  1.3620e-04, -1.7364e-03, -3.1634e-04,
         8.8146e-06, -1.1147e-04,  3.3945e-04, -3.0839e-01,  6.1839e-04,
        -5.9741e-04, -4.0969e-01,  9.8664e-04, -1.0968e-03,  1.4119e-04,
        -6.1730e-04,  5.4868e-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([ 3.4774e-03, -1.4712e-01, -5.1503e-04, -1.4536e-03,  5.7559e-04,
        -1.3274e-01,  7.1114e-04,  1.0005e-03, -8.2941e-04,  3.2739e-04,
        -1.9892e-01,  4.7164e-04, -1.0457e-04, -5.0971e-01,  9.5247e-04,
        -5.0922e-04,  5.8021e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  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([ 1.0912e-03,  4.2421e-04,  7.3070e-02,  3.7758e-04, -6.2473e-05,
         4.0531e-03, -4.0996e-04, -6.9244e-04,  2.3259e-04,  1.1437e-03,
         1.5619e-02, -3.3065e-04,  7.3226e-04, -4.0866e-04,  1.1668e-03,
        -3.3411e-04, -1.9396e-03,  2.8949e-04,  9.4219e-03,  8.8555e-01,
        -5.2122e-04, -5.4383e-06, -6.0290e-04,  2.0837e-05, -7.9675e-05,
         5.6299e-04,  2.6192e-04, -2.1931e-04, -3.2733e-04,  4.7643e-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([ 1.5664e-03, -3.6862e-05, -6.2631e-03, -6.8639e-04, -4.4277e-05,
        -2.2512e-04,  1.0977e-04,  3.5972e-04,  5.7175e-05, -1.8549e-04,
        -1.0816e-03, -2.4376e-04, -7.2238e-04, -3.2772e-04,  1.8358e-05,
        -1.8886e-05,  3.9852e-04, -1.6916e-04, -1.1545e-03, -4.5525e-01,
         2.1789e-04, -5.2533e-05,  1.2380e-04, -3.8823e-01, -3.7844e-04,
        -8.3098e-04, -4.2011e-05, -6.3148e-02,  2.8780e-04,  8.9603e-04,
        -7.6121e-02, -1.1509e-04,  5.1243e-04, -5.7164e-05,  6.6043e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5332e-03,  2.5557e-04, -3.8356e-03,  2.2492e-04,  1.9446e-04,
        -6.2080e-04, -2.7283e-04,  9.3464e-05, -6.8551e-05, -1.2168e-04,
        -5.8296e-04, -2.2204e-07,  1.8751e-05,  5.8146e-04,  4.2987e-04,
        -2.7682e-04,  7.1364e-04,  2.0500e-04, -7.7170e-04, -1.2066e-01,
         4.0269e-04, -7.7178e-06,  7.4683e-05, -4.3369e-04, -4.3262e-05,
         5.7878e-04, -3.1734e-01, -1.1878e-04,  1.5052e-04, -2.4600e-05,
        -2.1529e-01, -3.8231e-04,  2.5106e-04, -2.5976e-04,  3.6463e-04,
        -3.3193e-01,  3.4416e-04, -3.1864e-04, -2.2812e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8106e-03,  2.0303e-04,  1.6802e-02,  2.6762e-04,  7.0018e-04,
         7.7624e-04, -1.6714e-04, -7.6136e-04,  6.1572e-04,  3.3387e-04,
         4.6428e-03,  1.7937e-04,  6.3333e-04, -3.8737e-04, -2.1852e-05,
        -8.4099e-05, -4.9351e-04,  2.8339e-04,  4.0611e-03,  4.8952e-01,
        -1.9024e-05,  2.2851e-04,  1.5870e-04,  3.8944e-01,  1.5280e-04,
         3.2197e-03,  1.0181e-04,  8.2442e-02, -7.3310e-04, -3.6159e-05,
        -2.1265e-04,  1.7486e-04, -1.8966e-04,  7.6503e-05, -7.2938e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4826e-03, -7.1816e-04,  3.8102e-02,  1.0164e-03, -8.2849e-04,
         1.8325e-03, -6.6189e-05, -3.9289e-04, -1.9342e-04, -6.0129e-04,
         1.8214e-02, -6.9446e-04, -1.5387e-04, -1.3513e-03,  8.4938e-04,
         9.1299e-06, -1.2422e-03, -4.4370e-04,  6.0822e-03,  9.1893e-01,
        -4.5480e-04, -2.0747e-04, -5.0422e-04,  2.2688e-04,  9.3931e-04,
        -3.4539e-05,  1.0897e-03,  1.2807e-03,  2.2138e-04,  1.2811e-03,
        -5.5903e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4230e-03, -3.0764e-05,  5.2950e-02,  2.5073e-04,  2.2756e-04,
         1.2898e-03, -2.2661e-04, -1.0557e-03,  1.7193e-04,  2.0396e-04,
         3.9692e-03, -5.2586e-04,  7.2166e-04, -5.6468e-05,  6.4138e-04,
         5.2853e-04, -9.5520e-04,  6.9958e-04,  1.0895e-02,  9.1855e-01,
        -3.6543e-04,  2.7436e-04, -4.9523e-04,  6.0967e-04, -2.7347e-04,
        -1.5561e-04,  6.1173e-04, -2.7478e-04, -2.2437e-04,  2.8511e-04,
        -6.1224e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.1664e-05, -6.9854e-05,  1.8775e-02, -8.2250e-05,  2.9274e-04,
         4.7037e-04, -1.9924e-04, -3.8208e-05, -2.2182e-04, -6.5329e-05,
         9.3984e-04, -4.9655e-04,  5.1518e-04, -5.5718e-04,  2.5638e-04,
        -1.9096e-04, -9.2005e-04,  1.1238e-04,  2.5785e-03,  3.6931e-01,
        -2.4391e-04, -1.0321e-04,  3.4387e-04,  4.5627e-05, -5.2037e-04,
         2.4039e-04, -7.2397e-04,  6.0014e-01, -4.1287e-04, -5.5257e-04,
        -1.7767e-04, -3.5582e-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([ 2.1368e-03,  2.8210e-04,  5.2692e-03,  2.7220e-04,  2.4465e-04,
         3.5653e-07,  1.1783e-04, -1.0035e-03,  5.3268e-04,  9.9231e-04,
         2.4002e-03, -1.5592e-04,  1.0762e-03, -2.8347e-04, -7.9181e-04,
         7.4491e-04, -9.6703e-04,  3.4681e-04,  2.8891e-03,  8.9795e-02,
        -2.5198e-04,  3.2812e-04,  9.6937e-06, -4.0712e-04,  7.4616e-04,
        -1.1146e-04,  8.8498e-01, -8.5200e-04, -6.9675e-04, -4.7902e-04,
        -5.9498e-04,  1.9683e-04,  4.7531e-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([-5.0841e-03, -2.3473e-04,  8.2365e-02,  6.2629e-04, -2.9366e-04,
         5.8430e-03, -1.3040e-04, -4.4335e-04, -1.9529e-04,  3.5602e-04,
         3.3172e-03, -1.0215e-04,  6.5807e-04, -7.1739e-04,  8.4239e-04,
         8.8077e-05, -1.5319e-03,  4.1154e-04,  5.5435e-03,  7.6672e-01,
        -4.6376e-04,  5.1345e-04, -3.0369e-04, -5.9735e-04, -7.1546e-04,
         1.1929e-01, -3.8056e-04,  9.3257e-04, -8.9021e-04,  1.8462e-04,
         1.4111e-04, -8.6350e-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([ 4.3033e-03, -4.4833e-04, -1.5087e-02, -1.4373e-04, -1.9462e-04,
        -1.5566e-03, -2.3685e-04,  2.8032e-04, -1.2250e-04, -2.2043e-04,
        -3.1753e-03, -4.5837e-04, -3.7636e-04,  2.7970e-04,  1.5830e-06,
        -4.9222e-04,  5.5535e-04,  2.8891e-04, -2.0727e-03, -1.0298e-01,
         7.2927e-04,  3.3249e-04,  2.4552e-04, -1.9495e-01, -1.3613e-04,
        -5.0690e-04, -1.2236e-04, -2.3982e-04, -7.9999e-05,  5.1133e-04,
        -3.3881e-04, -3.8219e-04, -6.6704e-01,  5.3518e-04,  5.4852e-04,
        -2.3704e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7601e-03,  1.4216e-04,  2.5655e-02, -2.1872e-04,  2.7802e-04,
         1.3154e-03,  3.6430e-04, -3.1214e-04,  9.3406e-04,  3.8472e-04,
         6.2420e-03,  1.5748e-04,  9.5322e-05, -3.3994e-04, -3.1928e-04,
         3.8456e-04, -6.7811e-04, -6.2330e-05,  5.4135e-03,  9.3253e-01,
        -1.0573e-04,  2.5503e-04, -3.6786e-04,  1.9813e-04,  1.8552e-02,
         1.7800e-04, -7.6713e-04, -3.4078e-04,  1.3678e-04,  2.7654e-04,
        -2.3850e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7761e-03,  3.1275e-05,  1.9060e-02, -1.4898e-04, -7.9639e-05,
         1.6044e-03, -6.3816e-04,  1.9285e-04,  3.5810e-04,  4.9044e-04,
         1.8341e-03, -4.2317e-05,  7.4303e-05, -8.6608e-04, -6.5913e-04,
        -3.8719e-04, -9.2782e-04,  3.4025e-04,  1.4414e-03,  4.6323e-01,
        -2.5514e-04,  2.3846e-04, -4.0437e-04,  3.0367e-01, -2.3291e-05,
         3.1805e-03, -2.5738e-04,  3.7335e-02, -1.9273e-04,  2.5182e-02,
         8.5985e-02,  4.5756e-04,  1.9744e-04, -8.4310e-04,  4.6227e-02,
         1.9717e-04,  1.0522e-04, -7.0279e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #650: [tensor([ 1.8939e-03, -2.5978e-01,  1.1852e-04,  2.2650e-04, -1.1024e-03,
         4.5750e-04,  4.4568e-04, -3.1131e-04, -5.2519e-02, -3.9488e-04,
         2.3878e-04, -2.1247e-01, -1.1085e-04, -1.9740e-04,  2.5964e-04,
        -4.6771e-01, -1.2195e-04,  9.9366e-05, -3.8718e-04,  2.6809e-04,
        -2.8450e-04,  4.7717e-04,  7.9188e-05, -4.7328e-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([-6.3585e-04, -3.8527e-01, -1.0161e-04,  9.5604e-05, -9.1252e-04,
        -2.8403e-04,  2.5244e-04, -6.4000e-04, -5.1946e-02, -8.2654e-04,
        -4.5420e-04, -3.1039e-01, -4.0772e-04, -2.6609e-04, -2.6658e-04,
        -2.4449e-01, -6.9374e-04, -4.8845e-06, -7.9385e-04, -5.5017e-05,
        -5.5098e-04, -4.0087e-04,  8.9277e-05, -1.7021e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0353e-03, -3.0699e-01, -6.7918e-06,  3.1824e-04, -9.7021e-04,
        -1.7239e-04,  4.0312e-04, -2.8441e-05, -6.8033e-02, -6.5544e-04,
        -3.6362e-04, -3.2009e-01, -2.4526e-04, -5.5161e-05,  6.3693e-04,
        -2.9778e-01,  5.9898e-04, -2.5405e-05,  4.6622e-04, -6.8274e-05,
        -5.7186e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7075e-03, -5.5901e-05,  1.8088e-04, -1.2154e-04,  4.4501e-04,
         6.1372e-04,  6.7541e-04,  1.4970e-04, -2.2754e-04, -7.7039e-05,
        -3.5209e-03,  8.3304e-04,  5.5914e-04,  3.1757e-04, -1.4656e-04,
        -2.2545e-04, -3.6659e-04, -1.9916e-01,  1.7420e-04, -1.2861e-03,
        -9.2843e-05,  1.1501e-03,  6.3392e-04,  6.6381e-05,  1.7879e-03,
         1.0912e-03, -7.8309e-01,  5.5764e-04, -2.4900e-04, -4.3503e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2767e-03, -5.7999e-04,  5.1853e-04, -6.5228e-04,  2.3079e-04,
        -5.0260e-04,  2.7380e-02, -8.4512e-04, -5.7763e-04, -4.8642e-04,
         1.0337e-01,  7.3795e-04,  4.3964e-05, -1.0060e-03, -1.1695e-03,
         3.3383e-04,  9.6457e-05,  6.8818e-01,  2.4362e-04,  1.6454e-01,
        -3.9708e-04, -7.7888e-04, -1.3670e-03, -2.2857e-04, -8.1965e-04,
         4.0351e-04, -2.3120e-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([-2.4131e-02,  1.7052e-04, -2.2992e-04, -8.9596e-06, -2.8184e-04,
        -3.5072e-03,  9.1493e-02,  2.2780e-03, -1.0703e-03, -6.1258e-04,
         3.2625e-01,  6.9734e-03,  1.0904e-04, -6.3938e-03,  2.0260e-03,
        -6.3740e-04,  8.3512e-04, -2.7973e-03,  5.1995e-01,  1.6407e-04,
         9.2252e-03, -2.1601e-04,  6.3564e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.3560e-03, -2.5170e-04,  7.8307e-06, -9.0099e-04, -2.2391e-04,
        -1.5987e-04,  2.5206e-04,  2.9804e-04,  2.0902e-04,  3.4106e-01,
         3.7614e-05,  4.0762e-04, -1.5332e-04,  1.4726e-04,  1.6799e-04,
        -1.3658e-04,  9.5704e-05, -1.2351e-03, -2.4912e-05,  1.1080e-04,
         1.7523e-04, -1.3964e-04,  9.1019e-05,  5.2187e-01, -1.0450e-03,
         5.6202e-04, -6.7008e-04, -2.6061e-04,  1.6072e-04,  1.1970e-01,
        -6.8776e-04, -1.3210e-04, -1.9973e-04, -6.8568e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.2127e-03, -1.2820e-04,  3.2470e-04, -1.5217e-03, -1.6603e-04,
         6.0550e-04,  2.8788e-04,  4.0965e-04,  2.3226e-04,  4.1271e-01,
         3.4719e-04,  2.6000e-05, -3.9644e-04, -1.6663e-04,  1.3443e-04,
         1.0775e-04,  1.2799e-04, -8.0102e-04, -3.5535e-04,  2.6210e-04,
         2.4896e-04, -4.1538e-05,  3.3616e-04,  5.7310e-01,  6.6762e-04,
        -7.4532e-05, -6.1442e-04,  4.0223e-04,  1.9344e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9815e-03,  3.9998e-04, -6.9255e-04, -4.0982e-04, -6.7855e-04,
        -3.3448e-04,  2.8364e-05, -4.0011e-04,  5.4211e-04,  8.5488e-01,
        -1.3658e-04,  1.3489e-05, -8.0674e-04,  3.0933e-05,  2.9561e-04,
        -8.1474e-04, -2.6604e-04, -1.1328e-03,  1.2906e-04, -8.1393e-04,
        -1.6982e-04, -2.9417e-04,  1.3128e-01,  2.3750e-04,  7.9093e-04,
        -3.6991e-04,  1.6220e-03, -1.6162e-04, -2.8356e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.8869e-04,  2.0700e-01, -2.3299e-04, -2.3074e-04,  2.8606e-04,
        -9.5845e-04, -7.6295e-04,  6.7679e-01, -1.9445e-04, -3.2374e-04,
         3.3576e-04,  3.5710e-04,  3.0977e-04,  5.6899e-04, -2.2267e-04,
         5.3995e-02,  3.5107e-04, -1.5108e-04,  2.9818e-04,  5.2286e-02,
         9.6361e-05, -5.8509e-04,  6.6190e-04, -7.7444e-04,  6.3124e-04,
        -1.0624e-03, -1.4133e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.9686e-03, -5.5213e-01,  6.8813e-04, -1.3284e-06, -4.6263e-05,
        -6.6228e-05,  4.3693e-04, -2.0856e-01, -3.7867e-04,  5.2149e-05,
        -2.4906e-03,  2.3387e-04,  1.8613e-05, -1.2083e-04,  4.2122e-04,
        -1.9261e-01, -1.8013e-04,  6.1919e-05, -2.7610e-05, -3.5229e-02,
         2.4320e-04,  3.3657e-05,  2.0497e-04, -1.0773e-04, -5.6692e-04,
         1.1528e-04, -8.8207e-06,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9935e-03, -2.4656e-01,  1.8206e-04,  1.9870e-04, -7.0337e-04,
         1.4060e-04,  3.7234e-04, -3.3590e-01, -1.1785e-05,  1.5551e-04,
        -9.3621e-04, -1.1843e-04, -5.5194e-04, -3.9167e-04,  4.3027e-04,
        -8.7224e-02, -7.7811e-04,  4.0195e-04,  7.4728e-05, -1.5273e-01,
         5.5400e-04,  2.2237e-04, -1.3127e-04,  1.3360e-04, -5.5017e-04,
        -1.6632e-01,  1.2407e-04, -1.0876e-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 #700: [tensor([ 1.0517e-02, -1.0983e-03, -7.8470e-04,  9.1608e-06, -3.7301e-04,
         9.5148e-04, -2.6132e-04, -1.6292e-04,  4.9645e-05,  4.5814e-01,
        -1.5475e-04, -1.4448e-04, -1.4263e-03,  4.1639e-01, -5.1011e-04,
         1.0755e-01, -1.1630e-03,  1.9729e-04, -1.1615e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3616e-03, -1.8363e-04, -3.1746e-04, -1.1234e-03, -1.3964e-03,
        -1.4713e-03, -1.6825e-04, -7.2023e-04, -5.3920e-04, -2.4618e-04,
        -2.6253e-01, -2.3853e-04,  2.6371e-04,  1.0431e-03, -7.2377e-01,
         4.5789e-04,  8.4062e-04,  4.7910e-05, -2.8142e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9414e-03, -3.7653e-04, -4.4974e-06, -3.5566e-04, -8.1239e-04,
         8.5883e-04,  1.7892e-04,  1.2643e-03,  3.1518e-04,  3.3587e-01,
         3.4488e-04,  7.3431e-03, -6.0479e-04,  1.9210e-01, -1.0236e-03,
         2.4359e-04,  4.5267e-01, -5.6706e-04,  8.1127e-04,  3.1101e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8612e-03, -4.4124e-01, -7.3215e-05,  3.2048e-05,  2.0173e-04,
        -5.0771e-02,  1.7158e-04, -1.3705e-04, -7.0123e-04, -1.0350e-04,
         6.4243e-04, -2.6750e-01,  1.4007e-04, -3.6071e-03,  1.4570e-04,
         1.5928e-04,  4.7386e-04,  6.2571e-06,  1.2751e-04, -1.2007e-04,
        -5.2596e-05, -9.5784e-05, -4.5254e-02, -8.9307e-05, -1.0740e-03,
         1.4917e-04, -2.4234e-04,  2.8266e-04, -6.1285e-02,  9.1579e-05,
        -1.9433e-04, -1.2270e-01,  1.2351e-04,  9.5904e-05,  5.3845e-05,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.0406e-03, -1.6933e-01, -4.0598e-06,  1.0940e-04,  7.0010e-05,
        -1.4060e-01,  1.8736e-04,  1.8825e-05, -6.3149e-04, -2.0391e-04,
         5.5321e-04, -5.7385e-01, -3.2088e-04, -8.2964e-04,  5.0087e-04,
         2.0806e-04,  4.7704e-04,  2.6648e-04, -3.4441e-04, -7.0704e-05,
         1.6278e-04, -2.7788e-04, -1.0174e-01, -1.1218e-04,  8.0458e-05,
         1.2805e-04, -1.0909e-04, -3.8729e-04, -1.9362e-04, -2.0724e-03,
        -3.3863e-04, -1.1443e-04, -2.3729e-06, -2.5794e-04,  1.6022e-04,
        -2.5309e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.8221e-04, -7.6381e-02, -2.3319e-05,  3.8706e-05,  7.8163e-05,
        -1.0103e-01,  1.3612e-04,  5.1245e-05, -7.4755e-04,  1.8007e-04,
         4.2661e-04, -5.8901e-01, -1.3047e-04, -2.3377e-03,  3.7980e-04,
         3.6101e-04,  2.5697e-04,  2.1174e-04, -1.0556e-04,  4.1747e-04,
         7.2703e-05, -1.5947e-04, -4.2519e-03, -1.7435e-04,  1.7147e-04,
        -2.2320e-04, -6.3575e-02, -2.3014e-04,  8.9498e-04, -1.5648e-01,
         9.4008e-08, -6.5134e-05,  1.8997e-04, -1.7896e-04, -5.8436e-05,
        -8.8222e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.2796e-03, -4.6383e-04,  1.5724e-03,  7.5061e-04,  7.5231e-05,
        -3.4841e-04, -5.6743e-04, -1.0061e-03,  1.5788e-04,  1.0423e-01,
        -6.1675e-04,  2.7163e-04,  4.5511e-04,  8.6592e-01, -3.1456e-04,
         1.3032e-02, -7.3578e-04, -1.0711e-03, -4.8129e-04,  6.4884e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.9119e-03, -4.7132e-04,  1.6222e-03, -4.8809e-05, -1.3375e-04,
         1.4911e-05,  2.0619e-04, -5.0268e-04, -7.2318e-04,  3.6536e-01,
         3.6635e-04,  8.5394e-05, -3.6880e-04, -6.6973e-04,  2.7448e-03,
         5.5302e-04,  2.9790e-04, -9.0176e-04, -2.6748e-04, -3.9753e-04,
        -1.2519e-03,  6.1415e-01, -5.4919e-05, -2.3124e-06, -1.8974e-04,
         6.7891e-04, -3.3338e-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([ 2.1508e-03, -2.9618e-04,  2.7933e-03,  1.4891e-04,  9.8888e-05,
        -2.5227e-04, -1.5072e-04, -1.6272e-04,  4.5456e-04,  6.2045e-01,
         8.2354e-04,  8.4211e-05,  1.6760e-04, -9.2005e-04,  2.3265e-03,
        -9.3696e-05, -7.5898e-04, -5.1219e-04, -3.5880e-04, -2.4070e-04,
         3.0466e-04,  3.6566e-01, -6.9869e-06, -6.0027e-04,  1.6229e-04,
         1.9897e-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([-2.1082e-03, -6.3240e-04,  2.5746e-05, -2.3955e-04,  1.3649e-03,
        -5.1380e-01,  8.7693e-05,  5.6124e-05, -4.0340e-04,  5.3446e-04,
        -3.2852e-01, -8.6534e-04, -1.4996e-01,  1.0144e-03, -2.2542e-04,
        -1.5103e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6332e-05, -1.0420e-03, -6.5854e-04, -3.4064e-04, -4.4360e-04,
         4.6606e-01, -6.7244e-04, -2.2226e-04,  1.7158e-04, -2.1836e-04,
         3.1044e-01, -9.4753e-06,  6.7147e-02, -3.2688e-04,  1.5163e-01,
         1.8692e-04, -6.3826e-05, -3.3138e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7796e-04,  2.9387e-04,  8.3294e-04, -3.5703e-04, -7.4368e-05,
         9.9380e-01,  1.4599e-03,  4.8418e-04,  5.6188e-04, -7.3417e-04,
        -6.6829e-04,  3.0878e-04, -4.0902e-05, -3.2818e-06,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #750: [tensor([-0.1952,  0.0239,  0.0574, -0.0065, -0.0133,  0.0289, -0.0147,  0.0471,
         0.0148,  0.0109,  0.0664,  0.0082, -0.0024,  0.0804,  0.0360,  0.0040,
         0.0757, -0.0063,  0.0162, -0.0008, -0.0290,  0.0085,  0.1282,  0.0093,
        -0.0043,  0.0734, -0.0052,  0.0139,  0.0115,  0.0075], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.6940e-03,  5.0906e-04,  1.5624e-03,  5.5978e-04,  3.2210e-04,
         4.6138e-05, -1.7713e-03,  3.4195e-04,  7.5382e-05, -5.9978e-04,
         3.1588e-04,  9.6565e-05, -1.5418e-04,  2.3200e-03,  6.8770e-04,
         1.1599e-04,  1.2390e-03,  5.4977e-04, -1.0695e-04,  9.8109e-01,
         5.7099e-04,  6.1444e-04,  1.0753e-03,  1.2145e-04,  4.5642e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0924,  0.1989,  0.0493, -0.0085, -0.0046,  0.0463, -0.0187,  0.1548,
         0.0105,  0.0096,  0.0479, -0.0177, -0.0073,  0.0841,  0.0571, -0.0094,
         0.1585, -0.0100, -0.0107,  0.0016,  0.0014,  0.0005,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1079e-02,  2.7280e-04,  7.1050e-05, -2.9374e-04,  5.7068e-04,
         9.3934e-03,  5.5639e-04, -3.3001e-04,  1.2605e-01, -2.8962e-04,
         3.1971e-04,  3.3826e-02,  1.1534e-04,  1.0168e-04, -6.7617e-04,
         2.6358e-03, -3.3840e-04,  5.1960e-03,  2.7526e-04,  2.1670e-05,
        -5.6560e-04,  2.5032e-01,  7.6484e-05,  5.5538e-01, -6.0424e-04,
         6.2557e-04, -1.5727e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.6862e-03,  4.5854e-04,  2.6065e-04, -8.7167e-05,  2.6257e-04,
         2.4469e-03,  3.3397e-04,  3.6155e-04,  6.5102e-02, -5.6023e-04,
        -4.3289e-04,  2.1199e-02, -3.4720e-05,  1.8983e-04,  2.3625e-04,
         9.0757e-04,  6.4143e-04,  7.4620e-04, -1.2638e-04,  5.3038e-04,
         7.6417e-01, -4.6714e-04,  1.3321e-01,  3.4583e-04, -4.6398e-05,
         2.9648e-04, -7.6750e-05, -2.4120e-04, -2.1373e-04,  3.3163e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1797e-03,  1.4205e-06,  4.0204e-04,  1.7281e-04,  5.8625e-04,
        -2.3025e-04,  7.0214e-04,  2.7688e-05,  5.5027e-02, -4.3797e-04,
         2.2441e-04,  4.5443e-02,  5.0588e-06,  6.3528e-05, -1.6730e-04,
         1.2639e-03,  6.5526e-06,  9.7905e-04,  6.3749e-04,  5.7645e-04,
         1.6080e-01, -8.4991e-04,  1.5286e-02, -3.6490e-04, -2.1126e-04,
         7.1112e-01,  1.0876e-03,  3.0783e-04, -3.4425e-04,  4.9029e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9586e-03, -8.7844e-04, -5.5844e-04,  4.2346e-04,  3.6527e-05,
        -4.1674e-04, -2.5757e-03, -2.4354e-04, -1.8922e-04,  4.6281e-04,
         1.8152e-04, -8.0492e-01, -4.8605e-04,  6.1896e-04,  3.6902e-04,
        -1.7514e-04, -8.0029e-04, -2.5398e-02,  4.2348e-04, -9.3902e-05,
        -1.5587e-01, -4.5555e-04,  4.4529e-04,  1.3283e-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([-1.4624e-03, -3.8008e-04, -4.3150e-04,  1.3652e-04, -1.9982e-04,
        -4.0481e-04, -2.0008e-04, -1.8213e-04, -1.3379e-04, -2.1726e-06,
         4.5551e-04, -6.6966e-01,  1.7520e-05, -1.8114e-04, -1.0038e-04,
        -1.6950e-01, -2.4535e-04,  3.9573e-04, -1.4605e-03, -4.7629e-06,
        -4.6267e-04, -5.3181e-02, -9.9912e-05,  2.4258e-04, -9.9692e-02,
         1.9131e-04,  5.3300e-04, -1.3188e-05, -2.6440e-05,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8214e-04, -7.1355e-05,  2.1973e-04, -5.2018e-04, -2.9588e-04,
        -3.0730e-04, -6.1037e-04, -2.1924e-04, -1.4149e-05,  1.2697e-05,
        -5.0765e-04,  4.7134e-01, -4.8395e-04, -3.2106e-04, -1.4753e-04,
         2.8583e-01, -3.9349e-04, -3.6928e-04,  3.2407e-04,  2.0267e-04,
         5.6790e-04,  2.3574e-01, -7.4287e-04, -2.2407e-04,  5.5707e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6444e-03, -8.8919e-05,  4.2084e-04, -4.7898e-04, -7.0794e-06,
         6.8071e-04,  6.5906e-04, -1.7406e-04,  6.7525e-05,  6.9935e-04,
        -5.0576e-04,  8.9305e-01, -3.4263e-05, -1.4418e-04, -4.1989e-04,
         7.3182e-02,  5.3002e-05,  2.6683e-02,  3.7456e-04, -4.1745e-05,
         4.6636e-04,  4.7241e-07,  1.1998e-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([ 9.4402e-04, -4.1037e-04,  5.2572e-04, -3.0766e-04, -3.3317e-04,
         5.0869e-04, -2.9019e-04,  2.1959e-04,  8.3659e-04, -1.1553e-04,
        -2.3115e-04,  9.9154e-01, -4.8298e-04,  4.1242e-04, -2.3906e-04,
         8.3664e-04,  5.4653e-04, -3.2084e-05,  2.6314e-04,  1.0926e-04,
         5.0203e-04, -1.2305e-04, -1.8825e-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([ 5.2746e-03,  3.4045e-05, -1.2745e-04,  1.0947e-03, -7.7185e-06,
        -7.1170e-04, -5.0695e-03,  1.5249e-04, -6.7248e-04,  5.6756e-06,
         1.0141e-03, -5.5332e-01, -5.0545e-04, -1.6588e-04, -2.0245e-04,
        -1.3147e-03,  2.3001e-04, -2.0374e-01, -3.1664e-04, -2.2502e-01,
        -7.3368e-04,  1.4776e-04, -1.4079e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #800: [tensor([-2.0509e-03, -2.6022e-01, -7.2146e-04,  1.6155e-04, -4.6026e-04,
        -3.1416e-04, -2.9480e-05,  1.4747e-04, -1.3047e-01,  4.9286e-04,
         3.0743e-04, -9.3537e-02,  1.1759e-04,  1.9130e-04, -4.8407e-05,
        -4.9237e-01, -2.9964e-04, -3.8282e-04,  3.3683e-04,  4.4334e-04,
        -1.6370e-02, -1.7113e-04, -3.5943e-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([ 3.2938e-03, -4.4382e-01, -5.4964e-05,  2.9812e-04, -2.2827e-04,
        -6.0401e-04,  2.6635e-04,  2.5967e-04, -3.3881e-01, -2.3564e-04,
        -1.0291e-04, -7.5191e-02, -3.9531e-04,  1.2164e-04, -1.9595e-04,
        -8.8983e-02, -4.1053e-04,  3.0458e-04, -1.0026e-04, -7.1372e-05,
        -1.7871e-04, -4.5735e-02,  9.0948e-05, -2.5062e-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([ 2.6799e-03, -2.9117e-01,  2.0979e-04, -3.1515e-06,  1.3545e-04,
        -1.1107e-01,  4.9749e-05, -9.6803e-03, -1.1908e-01, -2.2679e-04,
        -2.0383e-04,  6.2139e-04, -5.2683e-02,  2.2413e-04,  5.8610e-05,
        -2.6521e-05,  9.4942e-05, -1.7081e-05, -2.6198e-01,  6.9234e-05,
        -1.4868e-01, -1.3833e-04,  2.8844e-04,  1.8862e-04, -4.2271e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9811e-03, -2.4108e-01, -8.5089e-05,  2.5095e-04,  5.1807e-04,
        -1.9096e-01, -1.6635e-04,  9.5926e-04, -3.7855e-01, -1.1430e-06,
        -1.6821e-04,  9.6556e-05, -6.0725e-03,  1.1289e-04, -1.6351e-04,
         6.9519e-05,  2.4430e-04,  8.2748e-05, -4.1866e-02,  1.8315e-04,
        -1.3407e-01,  1.3617e-04,  5.2022e-05,  1.2668e-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([ 2.4341e-03, -4.2663e-01,  1.3689e-04, -6.0241e-05,  2.7347e-04,
        -1.7634e-01, -2.1405e-04, -2.4051e-03, -1.1498e-01,  8.1566e-05,
        -1.2183e-04, -2.5881e-05, -3.6466e-02,  2.4325e-04,  2.4444e-05,
         4.8657e-04, -1.1132e-04, -1.2066e-04, -6.9090e-02,  1.2938e-04,
        -1.6921e-01,  6.9921e-05, -2.1923e-04,  1.3511e-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([-5.6474e-03, -4.2318e-02, -9.9486e-05,  1.5922e-04, -1.6494e-04,
         5.8427e-04, -5.7755e-01, -8.9578e-05, -2.3050e-04, -4.3768e-04,
        -3.0728e-02, -3.2509e-04, -2.9707e-05,  2.2248e-04, -4.2108e-05,
         1.7817e-04, -1.9702e-02, -2.7542e-04, -3.2669e-04,  9.0178e-04,
        -3.3025e-02, -1.4225e-04, -7.3571e-05, -2.3242e-02, -2.5311e-04,
        -8.0213e-04, -2.6153e-01, -3.6430e-04, -3.7257e-04, -1.9062e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3691e-04, -3.8930e-01,  3.0463e-04,  3.5997e-05, -3.4944e-05,
         7.2987e-04, -3.9461e-01,  3.4265e-04, -2.3732e-04, -4.9687e-05,
        -5.2863e-02, -6.1158e-05, -3.7008e-05, -3.0740e-04, -1.1392e-04,
         2.4511e-04, -1.2756e-01,  4.2363e-05, -2.9180e-04, -9.1078e-05,
        -3.1972e-02, -4.5052e-04, -3.1056e-05, -1.5102e-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([ 1.9433e-04, -6.8344e-01,  2.7587e-04, -3.3545e-04,  1.3159e-04,
         1.1879e-03, -1.1830e-01,  1.6459e-04,  1.2488e-04, -2.3588e-05,
        -3.0969e-02,  4.5866e-04,  1.2939e-04, -4.5742e-04,  1.3560e-05,
        -5.1384e-05, -1.1671e-01,  1.4419e-04,  9.7514e-04, -4.1585e-03,
        -1.4706e-04, -8.5623e-05, -4.0334e-02,  2.3821e-04, -1.4309e-04,
        -4.3712e-04,  2.9053e-04, -7.5760e-05,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8804e-04,  6.1766e-04, -4.4262e-04, -4.4300e-04, -1.4032e-04,
         3.6734e-04, -7.3607e-04, -2.8299e-04, -5.9757e-05,  2.8865e-04,
        -8.3516e-04,  2.2496e-01, -1.6464e-05, -5.7627e-04, -2.8798e-04,
         6.5289e-01,  1.0931e-03,  1.8443e-03,  1.6639e-05,  1.1346e-01,
        -2.4489e-04, -9.1052e-05, -1.2628e-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([-1.8501e-03, -5.2997e-04,  8.6217e-04,  4.0098e-04,  3.1124e-04,
        -2.8651e-05, -2.9728e-03, -7.4682e-05, -2.9421e-04,  6.1573e-05,
        -4.7367e-05, -1.8957e-01, -1.7170e-04, -7.5690e-05,  2.6650e-04,
        -5.0000e-01,  2.6073e-04, -1.8838e-03,  2.9550e-04, -2.9856e-01,
        -2.6056e-05, -4.1875e-04, -1.0781e-04, -3.0500e-04, -2.4276e-04,
        -3.7602e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3272e-03,  1.3289e-05,  1.4994e-04,  2.1331e-04,  4.9342e-04,
         6.2810e-05,  2.4073e-03, -2.0010e-04,  6.7526e-04,  1.9653e-04,
        -2.4760e-04,  5.8058e-01, -6.3538e-05, -6.5798e-05, -2.1887e-05,
         4.1135e-01,  3.6868e-04,  1.3322e-03, -6.5826e-05, -4.4040e-05,
         1.1522e-04, -7.7555e-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([-3.3953e-03,  8.7288e-01, -5.5299e-04,  1.0762e-04,  6.6966e-04,
        -8.4137e-05, -8.5833e-04,  1.3058e-03,  2.4392e-04, -1.9716e-04,
        -4.8078e-04, -1.1739e-04, -1.8556e-04, -1.7355e-04, -1.5569e-04,
         4.5060e-04, -1.1651e-04, -4.4612e-04,  1.1512e-01, -4.3140e-04,
         6.1945e-05, -4.5045e-04, -3.2042e-04,  1.6632e-04, -1.4044e-04,
        -6.0975e-04, -1.3429e-04,  1.3740e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #850: [tensor([ 3.3959e-05,  1.0809e-01, -2.1881e-04, -3.2856e-04,  7.8670e-02,
        -1.6838e-04, -1.1409e-04,  2.7194e-04,  3.2982e-04, -2.1247e-05,
         2.8645e-01, -7.2254e-05, -1.5478e-05, -1.8125e-04,  4.1864e-02,
        -5.1477e-04,  3.0545e-04,  1.7032e-04,  1.4031e-04, -1.3200e-04,
         4.7880e-01, -1.4927e-04,  2.8522e-04, -1.7379e-04, -6.1587e-04,
         2.1407e-04, -1.5943e-04,  7.9874e-04,  3.8737e-04, -3.2375e-04,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1114e-03, -2.4565e-01, -3.8120e-04,  1.4593e-04,  8.4744e-05,
         2.6175e-04, -3.0133e-01, -3.6572e-04, -1.1181e-04, -3.8486e-02,
         3.1607e-04, -1.7001e-04, -3.5076e-04,  4.5474e-04, -1.6101e-02,
         2.3910e-04,  2.0198e-05, -1.4405e-04, -3.9384e-01,  2.1445e-04,
        -2.1997e-04, -5.4484e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6735e-03, -4.2841e-01, -1.1664e-04,  2.7628e-04,  3.4501e-04,
         2.8628e-04, -1.5592e-01, -1.0050e-04,  1.4289e-05, -3.4221e-01,
        -3.5214e-04, -4.9666e-04,  1.1208e-04,  1.9956e-04, -9.0676e-03,
        -1.8334e-04, -1.0408e-04,  3.2174e-04, -5.8564e-02,  7.3714e-06,
         5.6393e-04,  6.9453e-05,  2.2340e-04,  1.5068e-04,  2.2834e-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([-2.5190e-05, -3.6488e-01,  1.6128e-04,  3.5442e-04, -2.7403e-05,
         5.1158e-04, -2.1732e-01, -2.0987e-04,  2.7185e-05, -2.0470e-01,
         2.3385e-05, -2.7692e-04,  3.1464e-04, -1.3915e-04, -3.2862e-02,
        -3.0318e-04,  2.4539e-04,  2.3461e-04, -1.7972e-04,  4.1345e-04,
        -1.2409e-01,  9.5155e-05, -1.7534e-05, -5.1853e-02, -3.1002e-04,
         3.1206e-04, -1.1864e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1202e-03, -5.3383e-04,  6.4217e-04,  4.3638e-04,  3.4124e-04,
         5.0130e-02,  2.0214e-04,  5.8461e-04,  6.3338e-01,  1.0390e-04,
        -2.5485e-04,  6.9287e-05, -9.1859e-05, -3.7437e-04, -8.8967e-06,
         2.6935e-04,  3.0724e-01, -7.1111e-04, -3.4961e-04, -5.8900e-04,
         1.0435e-04, -4.6281e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7408e-03, -2.6091e-05,  1.0127e-04,  3.1565e-04, -1.3586e-04,
        -1.2405e-01,  1.6580e-04, -1.3414e-03, -5.1369e-01,  8.6928e-05,
         1.3266e-04, -9.0347e-05,  1.4457e-04, -2.9818e-01, -2.9108e-04,
         1.7768e-04, -5.8031e-05, -5.8871e-02, -1.9827e-04, -1.6617e-04,
        -3.5838e-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([ 2.2779e-03,  7.1176e-05,  5.4682e-04,  6.4945e-05, -1.0495e-04,
        -2.6048e-01,  5.6743e-05,  4.5151e-04, -2.3468e-01, -3.0816e-05,
         4.8456e-04,  2.1815e-04,  1.3265e-04, -4.4994e-01,  3.3000e-04,
         1.7376e-04, -4.2346e-04, -4.6971e-02,  5.5139e-04,  4.4246e-04,
         1.7065e-04, -7.0243e-04,  3.8759e-04,  3.1209e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0898e-04, -8.9187e-04, -1.3444e-04,  2.2400e-04,  1.9877e-04,
        -5.4295e-05, -1.5488e-04, -1.1372e-04,  3.8250e-04,  9.4623e-01,
         1.2183e-03,  4.3117e-04,  4.6580e-04,  8.1020e-05,  2.5029e-04,
        -4.4282e-04,  2.3420e-04,  7.5017e-05, -5.8624e-04, -1.2061e-04,
         2.4348e-05,  4.5802e-02,  4.3873e-04,  7.0078e-05, -3.6878e-04,
         1.9312e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0125e-03, -5.5182e-04,  4.6866e-04, -2.4337e-05, -2.7335e-04,
         3.0292e-04, -4.8123e-05,  1.8612e-04,  2.0854e-04,  9.5575e-01,
         5.1807e-04, -2.6353e-04,  7.3904e-04,  8.3730e-05, -5.3946e-05,
         2.5189e-04,  2.2912e-05,  1.9392e-04, -3.0779e-04,  1.7728e-04,
        -9.9576e-05,  3.4190e-02, -3.9404e-04, -1.6677e-04,  8.2138e-04,
        -4.0226e-04, -1.1307e-04, -1.8621e-04, -1.8638e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9112e-03,  2.3886e-04, -2.4325e-05,  5.4421e-05,  3.5397e-04,
        -8.3594e-05,  2.8230e-04,  9.5941e-05,  4.0485e-04,  9.7834e-01,
        -4.1862e-04,  2.3284e-04, -5.1295e-04, -5.7142e-05, -7.8007e-04,
         7.1921e-04, -9.6848e-05,  4.2254e-04, -2.5343e-05, -4.6309e-04,
         1.1063e-02, -2.9216e-04, -2.1124e-04,  4.1427e-04,  1.6496e-04,
         3.3538e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3368e-03, -3.4872e-04,  1.3054e-04,  1.2612e-01, -9.4472e-05,
        -4.1350e-04, -1.0947e-06,  1.4539e-05, -5.6166e-04,  8.7365e-02,
         1.8362e-04,  6.0548e-03,  4.7842e-01, -3.7126e-04, -1.4512e-04,
        -1.1260e-04,  1.2002e-04,  1.5716e-01, -2.9220e-04,  1.7379e-05,
        -8.8919e-04,  1.3758e-01, -1.7801e-04, -2.4857e-04,  1.9592e-04,
        -6.9407e-04, -2.9543e-04,  7.6678e-05, -4.4300e-04,  1.1643e-04,
        -2.1289e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4581e-03, -9.3117e-04, -4.8445e-04,  5.9414e-02, -3.5229e-05,
        -7.3083e-04, -4.5465e-04, -2.8359e-04, -6.4487e-04,  3.0635e-01,
         1.4449e-04,  8.7244e-03,  1.5796e-01,  5.8827e-05, -1.8048e-04,
         2.3054e-04, -8.0950e-05,  4.3965e-01, -5.8824e-04,  1.8712e-02,
         1.0482e-06, -4.6889e-04, -1.9334e-04,  3.8373e-04, -6.3198e-04,
        -2.6925e-05, -1.7853e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #900: [tensor([-3.6307e-03, -3.9586e-01, -2.4810e-04, -3.5664e-04, -1.8516e-04,
        -4.2743e-01, -8.4607e-05, -3.9375e-04,  2.0692e-04, -3.6883e-04,
        -3.4231e-04,  8.2844e-04, -6.7299e-04,  2.7499e-04, -2.2501e-04,
        -1.6828e-01, -3.0022e-04, -2.9931e-04, -1.3386e-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([ 2.2498e-03, -3.1002e-01, -1.2268e-04, -7.4993e-04, -1.9680e-04,
        -3.2592e-02, -7.6469e-05,  3.4612e-04, -3.8621e-01, -6.2206e-05,
        -2.0631e-04, -2.0686e-04, -2.4160e-02, -1.8952e-04, -9.3298e-04,
        -2.2631e-04, -5.8098e-02, -3.6364e-04, -5.5267e-04, -1.8215e-01,
        -5.9545e-05,  2.0772e-05,  2.0595e-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([ 3.4339e-03, -2.3335e-01, -6.8990e-04, -4.2463e-03, -8.1066e-05,
        -5.3010e-02, -2.4046e-04,  1.9969e-04, -1.8784e-01,  7.3864e-07,
        -8.1536e-05,  3.3771e-05, -2.8539e-01, -5.3956e-04, -1.8961e-03,
         9.6760e-05, -1.2919e-01, -9.2494e-05, -1.0407e-04, -1.6983e-04,
         2.8705e-04, -9.8634e-02, -6.2553e-05, -2.8128e-04,  4.6333e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6072e-04,  5.4615e-01, -6.1171e-05,  3.7463e-03, -2.1065e-04,
         1.4647e-02, -1.7577e-04, -1.9609e-04,  2.0172e-01, -4.6175e-05,
         8.1725e-05, -3.3069e-05,  8.6348e-02, -1.0505e-04, -3.5681e-04,
        -1.1215e-04, -1.0287e-04,  2.3820e-04, -2.7673e-04,  1.4268e-01,
        -6.3678e-06,  3.3888e-04,  2.9185e-05,  2.0755e-04, -2.7822e-04,
         1.6101e-04, -8.8679e-04, -1.0348e-04, -6.4421e-05, -7.9189e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1418e-04, -6.3299e-01,  5.9741e-05,  1.4881e-04, -9.0034e-05,
         1.6480e-04, -4.0117e-02,  5.2557e-05,  5.0334e-05, -2.1076e-04,
        -2.1308e-01, -1.0676e-04, -5.5254e-04, -7.4098e-05,  1.3511e-04,
        -1.1085e-01,  9.2121e-04, -6.8104e-05,  2.1919e-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([-1.2137e-03, -4.1930e-01, -6.9528e-05,  2.7437e-04,  6.5256e-05,
         7.7539e-04, -5.0833e-01,  2.7923e-04,  7.0724e-05,  7.4558e-05,
        -2.1145e-02,  2.7081e-04, -1.4227e-04, -2.5630e-04,  7.5878e-04,
         2.9743e-04, -4.5674e-02, -2.1886e-04, -4.8728e-04, -3.0207e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4546e-03, -6.5477e-01,  1.8736e-04, -1.5434e-04,  9.1117e-05,
         9.2958e-04, -5.2809e-02, -6.9345e-04,  1.2952e-04,  1.5301e-04,
        -2.8716e-01, -1.5762e-04, -8.8850e-04, -4.2302e-05,  7.2831e-05,
        -3.0247e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7717e-03, -2.1122e-01, -5.0148e-05,  5.0339e-04, -4.8370e-01,
        -2.8301e-04, -1.9100e-05,  3.8976e-04, -2.9800e-01, -1.9071e-04,
         3.2657e-04, -1.9948e-04, -3.4105e-04,  5.6337e-05,  1.9036e-04,
         2.7660e-04, -9.9897e-05,  1.3154e-04, -2.3724e-04,  4.7111e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3978e-03, -7.5827e-01,  1.2272e-04,  1.9343e-05, -1.0421e-01,
         4.4006e-04, -8.1704e-05, -1.7469e-04, -1.3131e-01, -1.7732e-04,
         1.1004e-04,  2.2583e-04, -2.5087e-04,  1.3451e-04,  3.3392e-04,
         1.3980e-04, -2.5933e-05,  1.4008e-04,  1.1162e-04,  4.1418e-04,
         3.6492e-04,  4.3981e-05,  1.5829e-04,  1.6411e-04, -8.0838e-04,
         1.1533e-05,  3.6746e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7681e-04, -6.4121e-01,  3.1346e-04, -5.4636e-04, -3.9889e-02,
        -6.2010e-04, -2.1320e-04, -3.5493e-04, -3.1401e-01,  3.9386e-05,
         3.7150e-04, -1.1423e-04,  2.5035e-04, -2.6791e-04,  2.4175e-04,
        -1.9443e-04, -2.5810e-05, -4.8568e-04,  2.8136e-06, -4.0385e-04,
         9.6934e-05,  1.6929e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6368e-03, -2.2383e-04, -2.9395e-04, -5.4525e-05, -3.1530e-04,
         5.5242e-04,  4.0829e-04, -5.0787e-04, -5.9545e-04, -8.4647e-04,
        -1.8233e-04,  4.8527e-04, -2.1276e-01,  3.7277e-04, -1.2700e-03,
         2.5419e-04,  1.0755e-03,  4.8607e-04, -7.7432e-01, -3.9065e-04,
         2.2733e-04,  7.4700e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.8424e-03,  6.6080e-04,  6.0220e-05,  3.1906e-05,  5.6648e-04,
        -1.3995e-04, -3.2136e-04,  5.9112e-04, -5.1396e-04, -4.7321e-04,
         3.5169e-04, -4.6268e-05,  8.1974e-01, -2.4355e-04,  6.3056e-04,
         6.5229e-05, -7.3810e-04, -2.3888e-04, -3.7217e-04,  1.6469e-01,
        -9.6029e-04, -6.8100e-05, -6.4895e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #950: [tensor([-1.0581e-03, -3.9127e-01, -1.3222e-04, -3.1667e-05, -1.1354e-02,
         6.1308e-05, -1.0454e-03, -7.6558e-02, -2.9896e-04, -1.5669e-05,
        -4.6932e-05, -2.5564e-05,  3.8725e-05, -2.6693e-04, -3.0781e-04,
        -9.1485e-02, -2.9471e-04, -5.8478e-03, -1.3072e-06, -2.4703e-04,
        -7.2783e-02, -3.8156e-04, -8.8164e-05, -1.3964e-04, -1.8943e-05,
        -2.9735e-01, -2.1642e-04,  2.0764e-05, -2.4064e-04, -7.6484e-05,
        -3.5317e-05,  2.1887e-04, -2.8650e-04,  2.0326e-04, -4.6894e-02,
        -1.0916e-04,  3.3199e-04, -1.3624e-04,  7.9477e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.6731e-04,  1.3767e-01,  1.6715e-04,  5.8104e-04, -1.3889e-04,
         8.4508e-05,  4.7021e-01,  2.6389e-04, -3.1451e-05, -8.9809e-05,
         1.3985e-01,  8.9055e-05, -5.4528e-05,  5.3073e-05, -2.6664e-05,
         2.9677e-04,  1.3600e-04, -1.6820e-04,  1.3695e-04,  6.4333e-05,
         2.2503e-01, -1.3776e-05,  1.7468e-04,  2.3212e-02,  1.5633e-05,
        -9.4055e-05, -1.4310e-04, -1.6613e-04,  4.8239e-05, -7.3145e-05,
        -1.1375e-04,  2.7709e-05, -2.0437e-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([ 1.5269e-04, -5.9553e-01, -4.5752e-05, -9.4540e-04,  1.8691e-04,
         2.8761e-04, -1.6532e-01, -3.7645e-04, -1.1289e-04, -1.1953e-04,
        -1.0156e-01,  4.0479e-05, -8.7646e-05,  1.4939e-04, -6.8748e-05,
        -2.3971e-04,  7.0493e-05,  1.4482e-04,  2.5003e-05,  1.8174e-04,
        -1.0345e-01, -6.9506e-04,  4.5736e-05, -2.9341e-02,  3.1888e-04,
        -2.1962e-04, -8.1215e-05, -2.0438e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9155e-03, -1.3635e-01, -1.7489e-04,  2.9153e-04,  4.9057e-05,
         2.5643e-05, -3.4498e-01,  5.0960e-05,  2.1798e-04, -2.6621e-04,
        -1.0123e-01, -2.1411e-05,  5.2900e-05, -1.1632e-04,  1.5196e-04,
         1.4934e-04,  7.1819e-05,  2.9231e-04, -1.5339e-05,  3.9598e-05,
        -2.4805e-01, -3.6383e-05, -3.7188e-04,  1.6106e-04,  2.4767e-04,
        -1.6217e-01,  1.9380e-04,  1.0554e-04, -7.3964e-05, -1.2522e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.0640e-04,  4.1526e-01,  3.0689e-04, -1.5788e-05,  1.1547e-04,
        -7.7197e-05,  6.3959e-04,  7.4761e-06,  3.7322e-01,  4.2414e-04,
         4.4511e-05,  4.5104e-04, -1.4302e-04,  1.8215e-04,  2.0063e-02,
         5.3809e-04,  2.0983e-04,  1.8899e-04, -1.5697e-05, -1.3912e-04,
        -8.2280e-05,  3.5441e-05,  1.4257e-04, -1.7620e-05,  5.0140e-05,
         6.6698e-04,  6.5786e-05,  1.8574e-01,  1.2878e-04,  9.5366e-05,
        -1.2822e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1039e-04,  1.1991e-01, -1.1363e-04, -3.2425e-04, -4.2921e-04,
         2.9231e-04,  8.9068e-04, -1.2211e-04,  1.8117e-02, -4.3524e-04,
        -2.3497e-04, -1.0469e-04, -4.6492e-04,  1.0855e-05,  8.4156e-02,
        -2.3194e-04,  8.3905e-06,  1.7886e-04, -8.6463e-06, -1.2364e-04,
         1.9366e-04,  3.6332e-05,  3.7768e-05, -1.6773e-05,  1.1958e-01,
        -3.3149e-05,  9.3731e-05,  6.5269e-01, -4.7059e-04, -2.0152e-04,
        -1.8041e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7686e-03, -1.4812e-01,  2.5127e-04,  3.0903e-04,  3.0159e-05,
         3.7919e-04, -8.6339e-05, -4.4299e-06, -1.4048e-01, -1.4258e-04,
         3.0813e-04,  2.6862e-04,  9.2869e-05, -2.9505e-04, -2.8986e-01,
         8.0670e-05,  1.8132e-04,  8.8437e-05,  5.4143e-04, -1.6646e-04,
        -1.0417e-07,  4.8297e-04, -2.5248e-05,  3.5232e-04, -3.6173e-01,
        -1.5626e-04, -3.5857e-05,  2.6690e-04, -5.1036e-02,  1.5444e-04,
        -3.0744e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8646e-03, -3.4524e-01, -2.5008e-04,  5.1199e-04, -1.7273e-01,
         2.5506e-05,  6.2229e-05,  2.1111e-04,  4.3454e-05, -2.4700e-04,
         1.3511e-05, -2.2196e-01, -3.9944e-05, -3.6938e-04, -4.2565e-05,
        -2.2901e-01, -1.0442e-04, -1.0618e-04,  4.3671e-04, -2.6255e-02,
         2.3851e-04, -2.3742e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3335e-03, -4.4210e-01, -5.8843e-04,  1.5314e-04, -5.9983e-02,
         3.7971e-04,  9.0964e-04, -2.3490e-04,  2.3366e-04,  4.8172e-04,
         6.7944e-04, -4.8942e-01, -3.6796e-04, -4.6700e-04,  6.3797e-05,
         2.4053e-04,  1.2015e-04, -8.0059e-04, -5.8682e-05, -3.8097e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.8236e-04,  2.2650e-01, -1.6264e-04, -5.9991e-04,  5.9753e-01,
        -1.9695e-04, -2.9583e-04,  2.7361e-04, -1.4536e-04, -2.6547e-04,
         1.0975e-04,  1.7032e-01, -7.7168e-05,  5.0683e-04,  2.2088e-04,
         2.7181e-04,  5.7020e-06,  1.3229e-04, -1.0818e-04, -2.2977e-04,
        -5.1466e-04,  5.1306e-04,  2.3511e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4238e-03, -2.0989e-01, -1.2035e-04,  4.1517e-04, -1.9006e-01,
        -1.9142e-04, -1.6666e-04,  6.3862e-05, -3.2829e-04, -6.2763e-02,
         4.2513e-05,  1.2584e-04,  3.8761e-05,  2.6825e-04, -8.5831e-05,
         2.3583e-04,  4.2605e-04, -2.5906e-04, -1.7751e-04, -1.3143e-01,
         2.4984e-04, -2.6566e-04, -1.2842e-02, -2.2032e-04,  1.7174e-04,
         4.4705e-04, -3.8511e-01, -2.4235e-04,  3.6184e-04,  1.0422e-04,
        -4.8184e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1508e-03, -1.9562e-01,  2.6799e-04,  5.4271e-04, -1.9794e-01,
         1.7082e-04, -1.0574e-03,  8.1769e-05,  1.7783e-05, -4.6828e-02,
         1.0149e-04,  3.1526e-05, -2.1258e-05, -1.7727e-04, -5.3838e-05,
        -2.2841e-05,  6.3711e-05,  1.1383e-05, -7.0964e-05, -2.8749e-01,
        -2.3178e-04, -1.2779e-03, -4.6370e-04, -2.6363e-01,  5.6894e-05,
         8.4482e-06, -2.6715e-05,  1.8638e-04, -2.6923e-04,  1.2538e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  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([-9.2386e-04, -5.2989e-01,  1.2305e-04,  2.4846e-04,  2.2753e-04,
        -2.7659e-01, -1.3443e-04,  2.1551e-04,  2.7046e-04,  1.8659e-04,
        -7.8995e-05,  2.5026e-04,  3.8653e-04, -1.3561e-01, -1.1910e-04,
        -3.1950e-05,  1.6212e-04,  6.2150e-05, -1.1377e-02, -8.2977e-06,
        -2.8340e-04, -3.6906e-04, -3.2589e-04, -2.8947e-04,  8.9974e-05,
        -4.1531e-02,  3.6573e-05, -6.0413e-06, -1.7642e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6424e-04,  3.4033e-01, -2.1780e-04, -2.3214e-04, -1.9717e-04,
        -5.8552e-05,  4.2001e-01,  2.3843e-05,  2.6962e-06,  2.1067e-04,
         1.3252e-01,  8.5260e-05, -8.6511e-05, -3.9800e-04,  4.4011e-05,
         2.2280e-05, -2.7880e-04, -1.8889e-04, -1.9029e-04,  1.0176e-01,
        -2.3311e-04,  1.5693e-04,  2.5995e-04, -5.8242e-04,  7.4770e-05,
         4.1164e-04, -4.6197e-04, -1.2540e-04,  1.3207e-04, -1.4164e-04,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8412e-04, -8.7756e-02,  3.0675e-04,  1.9851e-04, -3.3037e-04,
         8.3122e-05, -3.9003e-01,  6.5535e-04,  1.5229e-04, -9.5399e-05,
        -5.0617e-01,  5.8025e-05,  2.9522e-04, -3.3549e-04, -2.5465e-04,
        -5.7221e-04,  4.9971e-05, -3.6060e-05,  3.9528e-04, -8.7935e-03,
         5.1387e-04, -1.7429e-03,  1.9574e-04,  2.3551e-04,  3.2884e-05,
        -1.4645e-04,  2.7847e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4554e-03, -1.0243e-01,  7.1261e-05,  3.2600e-04, -1.5016e-05,
         4.0723e-04, -2.1279e-01,  4.6128e-04,  1.0656e-04, -1.3692e-04,
        -6.7110e-01, -1.6079e-04,  7.5763e-05, -3.4201e-04,  1.7840e-04,
        -5.1398e-04,  1.2395e-04,  9.7214e-05,  4.6693e-04, -5.8487e-03,
        -2.3013e-04,  3.3812e-04, -4.5085e-04, -5.1806e-05,  1.3097e-03,
         3.2055e-04,  7.9786e-05, -1.0590e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5824e-04, -4.9795e-01,  8.3122e-06, -6.8038e-04, -2.7802e-04,
         2.0612e-05, -5.3556e-04, -1.4898e-01,  1.3909e-04, -1.0908e-03,
         1.0368e-04, -1.6153e-04, -6.8013e-05, -9.7280e-05, -2.4206e-04,
         3.1495e-04,  2.9544e-04, -1.7298e-04,  4.5064e-05, -1.4112e-01,
        -8.1322e-05, -1.0419e-03, -2.1342e-04,  1.3840e-04, -2.0563e-01,
         6.7099e-06, -3.9318e-05,  7.9431e-05,  1.0666e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3544e-03, -5.8904e-01,  1.4553e-04, -1.7839e-03,  1.6235e-04,
        -2.1108e-04, -1.5946e-05, -1.9682e-01, -2.0919e-05, -1.3629e-04,
         4.0163e-04, -1.5501e-04, -1.9261e-04, -1.2322e-04,  5.1775e-05,
         1.4355e-04,  2.3308e-04,  8.5842e-05, -1.0326e-04, -1.5779e-01,
         1.3194e-04, -5.3807e-04, -2.5141e-04, -2.7144e-04, -4.8166e-02,
        -2.4639e-04, -8.3695e-06, -2.4563e-04, -1.6518e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0729e-03, -4.1983e-01, -9.2659e-05, -8.2674e-04,  2.7031e-04,
         6.7696e-05, -1.8827e-03, -2.5300e-01,  2.7925e-04, -1.8419e-03,
         2.5315e-04,  2.8420e-04,  4.6190e-04, -5.9722e-04, -2.7567e-04,
        -3.9206e-04, -6.3928e-07,  7.9528e-06, -5.3321e-05, -3.1640e-01,
         8.9827e-05, -1.4448e-03,  2.0910e-05, -6.7980e-05,  1.1815e-05,
        -9.9176e-05,  3.8063e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8113e-03,  4.5653e-01, -2.0054e-04,  6.6411e-04, -1.8109e-04,
        -1.2648e-05,  3.4446e-02, -5.6920e-04, -1.0889e-04, -4.3985e-04,
        -2.8996e-04, -3.7145e-05, -9.0682e-04, -2.7139e-04, -5.7364e-04,
        -1.8413e-04, -9.7947e-04, -2.4707e-04,  1.0339e-01, -6.9018e-05,
        -6.6942e-05, -1.5309e-04, -4.1753e-04,  3.1572e-04, -1.1257e-04,
         3.9529e-01, -2.2837e-04, -2.3158e-04, -2.6965e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.6325e-04, -1.2006e-01,  3.5623e-05, -1.0828e-02,  4.4602e-05,
        -4.3119e-04, -1.0520e-01, -6.3391e-04, -3.9201e-04,  1.4232e-04,
        -4.2952e-03, -9.6417e-05, -5.6916e-04,  1.8737e-04,  3.6976e-04,
         1.0480e-04, -1.8441e-03,  1.5226e-04, -3.3597e-01, -1.6837e-04,
         2.3462e-04, -4.8422e-05,  1.5615e-04, -1.0527e-03, -4.0235e-05,
         1.7000e-04, -1.6074e-05, -4.1547e-01,  1.1616e-04,  1.2604e-04,
         1.7565e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0132e-03, -8.8540e-02, -2.3944e-04, -1.1835e-03, -1.7213e-05,
        -2.8753e-04, -3.9143e-02, -3.8749e-04, -4.5247e-04,  1.5262e-04,
        -1.9586e-03, -3.1841e-04,  2.7245e-05, -1.0075e-04,  1.2606e-05,
         1.3508e-04,  9.6268e-04,  2.7247e-04, -4.8623e-01, -4.0030e-04,
         2.0532e-04,  1.5326e-04, -3.3447e-04, -9.5019e-04,  1.3531e-04,
        -3.7404e-01,  1.9022e-04,  2.3868e-05,  1.3613e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2841e-03,  2.2949e-05,  4.1855e-04,  9.5752e-05, -2.9512e-04,
        -1.9591e-01,  5.0152e-04, -1.7629e-04, -3.1237e-04, -3.4221e-05,
         9.3368e-05,  2.2057e-04, -2.0606e-01, -6.6628e-05, -4.2757e-04,
        -1.3483e-01, -1.3896e-05, -6.9713e-05, -2.3903e-01, -3.7007e-05,
         1.6832e-04, -1.9162e-04, -2.1529e-01, -3.5626e-04, -1.6897e-03,
        -9.9987e-05, -2.5767e-04,  4.0858e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5109e-03, -4.9388e-04,  1.0214e-03, -7.4420e-04,  3.4772e-03,
         9.7890e-01, -7.8846e-04, -2.4157e-04,  3.1466e-04, -1.0423e-04,
        -2.4126e-04, -1.2909e-03, -2.5165e-04,  6.5583e-03, -3.8506e-04,
         2.0953e-03,  5.2412e-04, -6.0759e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1050: [tensor([-1.0936e-03, -7.9080e-01, -2.3452e-04, -3.5379e-04, -1.3853e-04,
        -9.7249e-02, -5.0330e-04,  6.5799e-05, -3.7341e-04, -3.7102e-04,
        -1.0230e-04, -2.6033e-04, -3.9305e-02,  3.1669e-04, -8.5544e-05,
         5.9178e-04, -1.2884e-03,  3.1060e-04,  7.2595e-05,  6.3550e-05,
        -2.0752e-03, -2.9364e-02, -3.4126e-04, -5.4632e-04,  1.8013e-04,
        -3.3583e-02, -1.4942e-04, -5.8160e-06,  1.7711e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.9518e-04, -5.3449e-04,  4.7745e-04,  3.8801e-04, -1.3116e-04,
         1.7082e-02, -2.4812e-04, -2.8013e-04,  1.1888e-04,  4.2105e-04,
        -2.1484e-04,  3.1348e-01, -7.1750e-04, -5.8529e-04, -1.0828e-03,
         6.6264e-01,  1.0779e-04, -1.6446e-04,  3.2657e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.1875e-03,  2.8268e-04, -2.7493e-04, -1.1869e-03,  2.0628e-04,
        -9.9723e-04,  9.3201e-04,  1.4099e-05,  2.2450e-04,  5.5879e-04,
        -7.2607e-01, -3.5210e-04, -4.8486e-04, -5.8282e-04,  3.4703e-04,
         1.4226e-03, -2.6052e-01,  2.1136e-05, -2.0558e-04,  7.3276e-05,
         5.4534e-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([ 1.2628e-03, -1.0258e-03,  5.7736e-03,  4.6419e-03, -2.2843e-04,
         9.4691e-03,  3.9735e-04, -3.2254e-04, -2.9446e-04, -3.8285e-04,
         5.9513e-01, -1.6695e-04, -6.4619e-04,  3.7892e-01,  7.8630e-04,
        -4.1231e-04,  1.4520e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6558e-03, -8.1599e-04, -1.5624e-03, -1.8719e-03, -2.7963e-03,
        -9.8681e-04, -4.2990e-04, -3.0259e-04, -2.1957e-03,  1.2519e-03,
         9.8181e-01,  1.5580e-04, -4.9760e-04, -2.6677e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.3735e-02,  1.5323e-01, -9.2521e-03,  1.0158e-02, -6.0987e-02,
         1.9354e-01, -1.5385e-02, -1.3107e-02, -1.4406e-03,  4.1050e-02,
         1.7626e-01,  7.9094e-03, -4.6945e-02,  2.0029e-01, -1.1471e-02,
         5.2059e-03, -4.1217e-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([-2.3079e-02,  1.4395e-02, -2.3210e-03, -3.3131e-04, -4.1275e-03,
         8.7031e-03, -2.5565e-04,  9.8810e-04,  3.8107e-05,  3.2327e-02,
        -1.5791e-04,  8.7112e-01,  2.8305e-03, -2.5457e-05,  3.7017e-02,
        -4.6161e-04,  1.5286e-03,  2.8807e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1743e-03,  8.3569e-07,  9.5925e-01,  1.5564e-03, -1.1674e-03,
        -4.0820e-04,  1.3867e-03,  1.7712e-03, -5.4154e-04,  2.4518e-02,
        -1.8168e-03, -9.1427e-04,  3.8776e-03,  9.5840e-04,  6.6197e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4082e-03, -5.1883e-04,  7.8830e-03, -1.7463e-03, -1.0745e-03,
         5.6487e-04,  6.6151e-05, -1.7161e-04,  9.2182e-01,  5.4492e-04,
         6.1719e-02,  1.8641e-04, -4.7295e-04, -8.5543e-04,  6.4004e-04,
         3.2463e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9365e-03,  1.6747e-04,  4.1616e-02, -1.2021e-04, -4.5166e-05,
         3.5814e-04,  2.7741e-04,  7.1274e-04,  8.8207e-01,  6.9904e-04,
         6.7397e-02,  6.1268e-04, -4.1235e-04, -8.7257e-04,  2.2690e-04,
         4.7699e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1657e-03,  7.3797e-01,  1.7199e-04, -2.0819e-04,  4.5757e-04,
        -2.3135e-04, -8.1283e-05, -4.6068e-04, -6.1714e-04,  1.1808e-01,
         6.7828e-04, -2.2138e-04,  9.8053e-05,  1.3615e-01, -3.2970e-04,
         7.3407e-04,  2.0138e-05, -3.2710e-04, -1.2979e-04, -6.1210e-04,
        -1.5727e-04, -1.0704e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.4525e-04, -1.3712e-01, -2.2184e-05,  4.3287e-04, -8.9382e-05,
         2.6204e-04, -6.9953e-04,  1.1681e-04, -2.0036e-04, -3.5893e-01,
         5.5650e-06,  3.0439e-04, -2.4928e-04, -3.1256e-01, -6.1830e-04,
         3.2678e-04, -3.7480e-05, -2.8353e-04, -1.8662e-01, -4.6232e-05,
        -2.3986e-04,  0.0000e+00,  0.0000e+00,  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 #1100: [tensor([-1.9985e-03,  5.4991e-02,  2.1396e-04,  5.2852e-03, -5.1787e-04,
        -5.2005e-04,  1.1403e-05,  8.9177e-01, -9.9046e-04,  6.5822e-03,
         2.2410e-04,  5.0266e-06, -5.8087e-05, -3.1783e-03, -2.9914e-04,
        -1.6858e-04, -2.5269e-03, -3.9590e-04, -7.4964e-04,  4.7979e-05,
         4.7043e-04,  2.8232e-02,  7.6319e-05, -1.1803e-05,  5.0394e-05,
         8.6920e-05, -1.7963e-04, -2.4008e-04,  1.1942e-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([-2.0618e-03,  4.1141e-01,  1.5028e-04,  2.5460e-02, -2.0753e-04,
        -3.9076e-04,  9.7710e-05,  2.8393e-01, -2.5432e-04,  5.5859e-02,
         1.7572e-04,  1.8110e-04, -1.2620e-04,  3.8358e-03, -2.2317e-04,
        -1.3646e-04,  2.8788e-02, -1.1735e-04, -1.8790e-04,  9.9073e-05,
         2.5669e-04,  1.6809e-01,  3.9308e-06,  1.6923e-02, -9.4169e-05,
         8.3817e-05, -1.9693e-04,  2.8115e-04,  1.2473e-04,  1.9593e-06,
         2.4862e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0662e-03, -9.1697e-05, -1.6573e-04, -1.3350e-04,  1.2051e-04,
         7.0840e-04,  3.1020e-04,  1.1409e-02, -2.5169e-04, -5.7856e-05,
        -2.5884e-05, -1.7032e-04, -8.6937e-05, -1.0974e-04,  5.2633e-04,
         6.3508e-01,  2.2801e-04,  5.7547e-05, -1.8747e-05,  2.4209e-04,
        -1.9376e-04, -1.7376e-04, -1.2754e-04, -2.9095e-05,  7.4843e-04,
         3.4369e-05, -3.1329e-04,  1.8911e-04,  3.4569e-01,  1.9977e-04,
        -1.8142e-04, -4.0311e-05, -2.2003e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0049e-03, -6.9986e-04, -2.1004e-04,  4.1851e-04, -2.5373e-04,
         9.2800e-04,  9.1424e-04,  5.2314e-02, -2.1998e-05,  4.6513e-04,
         3.8868e-04, -7.3533e-05, -2.2009e-04,  3.7350e-04,  4.2845e-04,
         9.3841e-01,  5.2716e-04,  3.4945e-04,  1.1189e-05,  6.8463e-04,
         7.8961e-05, -2.7120e-04, -1.4972e-04,  1.1052e-04,  1.1926e-04,
         4.9242e-04, -7.8473e-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([-2.7168e-03, -5.0623e-04, -2.8329e-04,  5.9541e-04,  5.8877e-04,
         1.5009e-04,  5.8351e-04,  7.2273e-02, -2.2697e-05, -8.2197e-06,
        -1.8409e-04,  5.9118e-05,  1.3679e-04,  1.4427e-04,  5.6593e-04,
         9.1861e-01, -8.7706e-05,  3.8531e-04,  3.4782e-04,  2.6307e-04,
         4.2185e-05, -5.6270e-04, -3.1084e-04, -2.9956e-05,  6.4671e-05,
        -2.3901e-04, -2.3643e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3050e-03, -7.5351e-05, -4.7021e-01,  1.1034e-04,  1.1719e-04,
         2.9345e-04, -3.7260e-04,  1.1520e-04, -4.7228e-04, -6.7254e-04,
        -4.5814e-04, -3.8891e-02,  1.6558e-04,  4.7722e-04, -4.2556e-01,
        -6.2397e-04, -3.1167e-04,  2.1894e-04,  5.0708e-04, -4.1692e-02,
        -1.6740e-05,  1.4059e-03, -1.2264e-02, -6.4367e-05,  3.3811e-04,
         3.9862e-05,  2.0571e-04,  1.6876e-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([-1.8816e-04, -8.0917e-07, -3.7714e-01,  2.6056e-04, -7.9241e-05,
        -4.5307e-05, -1.0142e-03,  1.4227e-04, -2.1815e-04, -7.3865e-03,
         8.1444e-05, -1.5238e-02, -1.8222e-06, -1.9845e-03, -9.0943e-02,
        -3.6365e-04,  2.9197e-04, -1.0158e-04, -8.6212e-05, -3.6667e-02,
        -6.6404e-05,  1.5387e-04, -2.7930e-05, -1.4221e-03, -4.5243e-05,
        -2.2420e-04, -6.2807e-06, -8.8179e-04,  9.2180e-05, -4.5437e-02,
        -3.5017e-04,  1.3327e-03, -4.1750e-01, -2.1010e-04, -1.5827e-05,
        -6.6374e-06], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8850e-03, -4.0518e-05, -1.2004e-01, -2.6553e-04, -1.0665e-05,
        -1.1713e-05, -3.2962e-03, -2.7641e-04, -2.6147e-04, -5.7988e-04,
        -1.2518e-04, -3.8065e-02,  2.2159e-04, -2.5608e-03, -3.8212e-01,
        -1.2797e-04, -3.0739e-05,  1.4597e-04, -4.4747e-04, -9.8893e-05,
        -1.4665e-04, -3.9534e-03, -1.4333e-04, -4.3350e-04, -9.9781e-05,
        -1.2593e-04, -2.2863e-04, -3.1265e-01,  9.1399e-05,  1.7604e-04,
        -1.2990e-01,  8.3677e-05,  2.0641e-04,  1.4287e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6759e-03,  4.0361e-04, -4.5790e-04, -2.8759e-04, -2.9606e-04,
        -1.9029e-03, -1.6745e-04,  4.9799e-05, -2.5092e-01, -1.9703e-04,
         7.9499e-04,  1.3997e-04, -4.6992e-04, -4.4517e-01, -3.0132e-04,
        -1.3842e-01,  2.0951e-05, -1.5635e-01, -5.2058e-04,  1.7719e-04,
         1.6542e-04, -1.1753e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3605e-03, -3.0788e-04,  4.7328e-04, -2.4359e-04, -1.9355e-04,
        -4.2899e-04,  1.2454e-04,  1.6985e-04,  3.1368e-01, -4.5871e-05,
        -2.3225e-04,  5.0915e-06, -2.1547e-04,  2.7039e-01, -4.7818e-04,
         4.0663e-01, -3.0287e-04, -3.6896e-04,  1.9319e-04, -5.4359e-04,
         1.0072e-04, -1.6965e-04,  1.3006e-04, -2.0950e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5032e-04, -7.1677e-04,  8.5641e-04, -9.6326e-04,  6.5212e-04,
        -1.0487e-04,  3.3522e-05, -2.1946e-04,  8.9127e-01,  4.5467e-04,
        -2.3982e-04,  8.6268e-04,  1.1756e-04, -7.6784e-04, -1.2532e-03,
         9.9358e-02,  1.6730e-03,  2.0421e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8790e-03,  3.3993e-04, -6.7841e-04,  8.9624e-05, -3.2302e-04,
         2.1672e-04,  2.6196e-04,  3.4261e-04, -1.0539e-04, -1.8606e-05,
         1.7639e-04,  6.7308e-01,  3.9607e-04,  8.0380e-06,  6.0632e-04,
         1.2543e-04,  1.3496e-04, -9.1086e-04,  3.1932e-01, -8.0214e-04,
        -1.0210e-04,  8.0540e-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>)]
Iter #1150: [tensor([ 1.4686e-03,  5.8654e-04, -2.7810e-06,  6.0662e-05,  3.1787e-04,
        -5.7463e-01, -1.0883e-03, -4.9575e-04, -1.3536e-01, -6.1451e-07,
         1.3614e-04,  1.1439e-04, -1.8312e-04,  4.1424e-04,  1.9401e-06,
        -2.1583e-02, -1.7203e-04, -2.6249e-01,  3.4831e-04, -5.4055e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.8506e-04,  4.7621e-05, -1.9975e-04, -7.2991e-05, -3.5180e-04,
        -5.7831e-01, -4.8369e-04, -2.6101e-04, -3.6752e-01, -5.4458e-04,
         3.2427e-04, -3.4651e-04,  7.7277e-04,  5.3022e-04, -7.3140e-04,
        -4.7839e-02,  4.1388e-04,  3.6179e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2964e-04,  9.9385e-06, -2.8198e-04, -3.4070e-05,  5.8910e-01,
        -1.2172e-04,  5.5467e-04, -2.1635e-04,  3.4285e-01, -4.8096e-04,
        -1.6089e-04,  1.8731e-04,  2.4903e-04,  6.4088e-02, -3.2834e-04,
         4.6691e-04, -1.0572e-04,  5.3461e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4697e-03, -1.9198e-04,  3.7881e-04, -5.6251e-04,  8.6101e-01,
         3.7616e-04,  6.8985e-05, -4.8185e-04, -1.3439e-03, -1.3783e-04,
         2.5572e-04, -5.2811e-04,  8.9179e-04,  5.7613e-02,  5.9001e-04,
        -2.2657e-04,  7.2643e-02,  3.2762e-04, -8.6930e-04,  3.4092e-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,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2777e-03,  3.3157e-04, -6.5186e-04, -6.5857e-05, -4.0409e-02,
        -6.9569e-04, -4.1652e-04, -4.1829e-04, -4.3022e-01, -2.1378e-04,
         9.6821e-05, -1.8860e-04,  7.1723e-05, -5.2274e-01,  5.5086e-04,
        -6.9040e-04,  4.5012e-04,  5.1161e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2882e-03, -6.6365e-04, -9.4878e-04, -1.4058e-05, -5.5658e-04,
        -8.3409e-04, -6.2940e-04, -9.0831e-01, -1.6975e-04, -1.8592e-04,
         1.2874e-03, -8.6630e-04,  5.0666e-04, -5.7954e-05, -3.2596e-04,
        -5.3187e-02, -7.4507e-06, -8.5024e-05,  3.6876e-04, -2.8112e-02,
        -1.2556e-03, -3.4354e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1102e-04, -3.4230e-04,  1.0642e-03, -1.1888e-03,  2.8775e-04,
        -2.3248e-04,  7.0715e-04, -9.3685e-01,  5.6869e-04,  3.7428e-04,
         8.5723e-04, -4.6812e-05,  5.3145e-04,  1.7038e-04,  1.2575e-04,
        -5.3803e-02,  1.4861e-04, -1.4842e-03,  7.0239e-04, -4.0222e-04,
        -5.7680e-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([ 5.7774e-04, -3.8901e-04,  5.1670e-04, -6.3249e-04, -1.1325e-04,
        -4.5227e-04, -8.4301e-04,  8.6318e-01, -6.0075e-04, -1.5718e-04,
        -3.8800e-03, -6.6705e-04, -7.2712e-04,  3.7807e-04,  5.5599e-04,
         1.2508e-01,  6.9648e-05,  2.2682e-04,  4.3225e-04,  5.2179e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0604e-03, -1.4458e-01,  2.1887e-04,  4.8403e-04, -7.2647e-01,
        -2.1366e-04,  3.9563e-05,  1.0975e-04, -2.5038e-05,  3.5208e-04,
        -3.5241e-04, -3.2727e-04, -2.1554e-02, -6.5942e-04,  2.0361e-04,
        -2.3301e-04, -1.5809e-04,  1.1045e-04, -1.4168e-02,  3.3981e-05,
         5.2373e-04, -8.5518e-02, -4.2595e-04,  3.0892e-04, -7.2808e-04,
        -3.7184e-04, -1.0824e-04, -3.7152e-04, -4.2499e-05, -2.4308e-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([-1.5404e-03,  3.3940e-01,  2.8119e-05, -8.2503e-05,  3.5107e-02,
         1.2014e-04, -2.7434e-05, -4.9609e-06, -9.2786e-08,  1.2099e-04,
         1.3694e-05, -2.8093e-04,  3.3817e-01, -4.7199e-05, -4.6776e-05,
         3.5282e-05, -1.1619e-05, -8.4991e-06,  2.2405e-01, -1.6984e-04,
        -2.7045e-04,  1.5528e-03, -6.0830e-05,  3.3445e-05, -9.3625e-05,
         1.8121e-03,  5.3165e-02,  1.1271e-04, -1.6587e-04,  5.6796e-04,
         2.0290e-03,  9.0540e-05, -3.2512e-04, -3.5129e-04,  1.8911e-05,
        -4.5596e-05, -4.2968e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8919e-03, -1.4613e-01,  2.2465e-04,  5.8911e-06, -4.1616e-01,
        -7.5447e-05, -2.1543e-04,  9.6066e-05,  2.0653e-05, -7.1503e-04,
         2.9605e-06,  1.5834e-04, -3.0716e-01, -7.3754e-05,  2.5067e-04,
        -3.2311e-04, -1.2336e-04, -1.1993e-04, -1.2216e-01,  7.7637e-04,
        -1.8671e-05,  3.5420e-04,  8.4511e-06,  1.9695e-04,  4.0008e-06,
         2.7942e-04, -2.4333e-04, -6.3366e-05, -1.6338e-05, -4.6228e-05,
         1.9707e-05,  7.7110e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.8955e-04, -2.8086e-01,  1.2283e-04,  1.3046e-04, -3.0393e-01,
        -4.4875e-05, -3.3480e-04, -1.4298e-03, -1.7301e-04,  8.1969e-05,
        -7.6505e-04,  4.3603e-07, -8.9200e-02, -1.3999e-04, -3.7351e-05,
         1.7921e-04, -4.6223e-05,  1.3123e-04, -1.5180e-04,  2.6927e-04,
        -7.8355e-02,  2.3969e-04,  4.8833e-05, -3.0649e-05, -1.5640e-01,
        -4.9656e-05, -1.0603e-03, -1.3612e-04, -8.4077e-02, -1.4253e-04,
         3.2092e-05, -1.5124e-06, -1.0734e-04, -7.6005e-05, -5.7927e-05,
         1.6901e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1200: [tensor([ 2.1839e-03,  3.9036e-02, -3.2235e-04,  9.8696e-04,  1.0525e-04,
         6.6320e-01,  6.5769e-05, -3.9482e-04, -6.1337e-04, -1.0219e-04,
        -2.3645e-04,  2.0870e-04, -4.6031e-05,  2.7364e-04, -1.5234e-04,
        -1.9376e-04,  3.9562e-04,  2.7016e-01, -3.4274e-04,  2.5518e-06,
         1.9357e-02, -1.0020e-04,  3.2100e-04, -2.7933e-04,  2.9618e-04,
         1.6825e-04,  4.5364e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3793e-03, -1.5432e-02, -2.5003e-04, -7.2277e-04, -5.5407e-04,
        -2.8386e-01, -5.8898e-05, -3.9114e-04, -2.1105e-04,  3.6173e-04,
         4.1790e-04, -6.9417e-05, -4.9202e-04, -1.1278e-04, -1.5808e-04,
        -6.3514e-01, -2.0515e-04, -9.7167e-04, -4.6034e-04,  1.0785e-04,
        -4.6162e-05, -3.8465e-02,  2.9606e-04, -1.6505e-04,  7.1974e-05,
         2.9135e-04, -1.6876e-02, -1.4157e-04,  1.7498e-04, -1.1909e-04,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0829e-03,  8.3550e-02, -6.6499e-05,  2.8618e-04, -1.3455e-04,
         7.5266e-01, -3.1892e-04, -7.0409e-04, -4.3187e-04,  9.4000e-05,
        -2.4027e-04, -8.3532e-05, -3.2574e-04,  2.6705e-04, -3.2695e-04,
        -2.6302e-04,  1.1205e-04, -2.7513e-04,  1.1043e-04, -6.7484e-05,
        -4.3838e-05, -2.4982e-04, -1.2989e-04, -1.6212e-04, -3.4930e-04,
         8.1003e-05, -2.2034e-04,  1.5550e-01, -3.3535e-04,  1.9015e-04,
        -3.3570e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.3579e-04, -3.2936e-01, -1.7817e-04,  1.1284e-04,  3.1163e-05,
        -2.5516e-01, -2.1596e-04, -1.1094e-04, -5.9805e-03,  2.0435e-04,
        -4.2529e-04, -2.6513e-01, -6.2490e-05, -3.7022e-04, -1.2946e-04,
        -3.4977e-02, -6.0999e-05, -2.3090e-04,  1.3735e-05, -7.7726e-03,
        -1.6017e-04, -4.5202e-04,  1.1863e-04,  1.4298e-04, -9.6938e-02,
        -2.5111e-04, -3.6679e-04,  1.1028e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.2570e-04, -1.1562e-01, -1.6231e-04, -3.0372e-05,  1.1177e-04,
        -7.5212e-02, -5.6359e-04,  1.7495e-04, -2.4927e-01, -1.5562e-04,
        -3.5713e-03, -5.2343e-01,  2.1509e-04,  2.1214e-04, -9.1882e-05,
        -1.2686e-02, -2.4322e-04, -9.4738e-05, -1.3917e-04, -1.4753e-02,
         1.1276e-04, -9.6242e-04, -1.1681e-05,  2.4763e-04,  4.1520e-04,
         2.9426e-04, -1.0989e-04, -1.4183e-04,  7.6017e-05, -6.9646e-05,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7589e-04, -1.1113e-01, -6.2340e-04,  7.2121e-05,  1.9518e-04,
        -5.9074e-01, -1.7274e-04,  1.7589e-04, -8.0603e-02, -1.0088e-04,
        -1.0693e-04, -6.7901e-02, -5.0470e-05, -5.4758e-05,  4.2716e-04,
        -8.7439e-02, -4.4149e-04, -7.6285e-05,  2.4806e-04, -5.8463e-02,
        -2.1132e-05, -5.3761e-04,  5.5663e-05,  3.3878e-05,  6.0763e-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([ 2.5704e-03,  2.1728e-01,  1.1943e-04,  4.9379e-05,  4.5072e-01,
         2.2727e-04,  7.4688e-05, -5.3362e-04, -1.6979e-04, -1.2278e-06,
        -2.8873e-05, -7.5317e-06, -1.6330e-04,  3.2672e-01, -8.3360e-05,
         3.7942e-05, -1.6229e-04, -1.3266e-04, -4.0694e-05,  1.0658e-04,
         2.4684e-04,  2.7797e-05, -1.1701e-04, -3.7412e-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([ 6.5208e-03,  8.5168e-01, -5.3019e-04, -2.5015e-04,  9.0386e-02,
        -1.8873e-04,  9.3694e-06, -1.1555e-03,  5.2976e-04,  7.6496e-04,
         2.1743e-05, -6.5060e-04, -6.4051e-04,  4.4797e-02, -4.3565e-04,
        -5.5565e-05,  2.3910e-04, -3.9328e-05, -1.6977e-04,  2.4518e-04,
         5.5715e-05, -1.9264e-04,  6.3045e-05, -2.6449e-04, -1.1635e-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([-5.8525e-04, -4.0293e-02, -3.3123e-04, -1.8171e-04, -8.0963e-01,
        -8.5402e-04,  2.6469e-05, -1.8224e-03, -2.6495e-04, -1.0042e-03,
        -2.3423e-04, -4.3388e-04,  1.0697e-04, -1.4105e-01, -3.7275e-05,
         5.7119e-04, -2.9010e-04,  4.0089e-04,  9.7511e-05,  5.1658e-04,
        -4.5577e-04, -4.4548e-04,  2.6669e-04,  3.1647e-05,  6.9766e-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([-1.5327e-03, -5.4342e-01,  5.7777e-05, -1.6410e-04,  2.8885e-04,
        -2.1963e-01, -1.7057e-04,  7.9461e-05, -1.2091e-04, -1.7134e-03,
        -1.5844e-04, -9.9248e-05,  2.8273e-04, -1.1225e-03, -5.6423e-05,
        -1.1038e-04,  1.3153e-04,  4.8598e-04, -1.3810e-01,  2.6591e-04,
        -1.6427e-05,  1.3903e-04, -9.0479e-02, -5.1828e-04, -4.7223e-04,
        -6.1308e-05, -1.4736e-04,  5.2290e-05,  1.2901e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4868e-04,  8.3537e-02, -3.0246e-04, -8.8464e-05, -7.8729e-05,
         6.5445e-01, -2.1780e-04,  5.0328e-04,  1.1526e-04,  1.0722e-02,
         3.4537e-04,  5.4052e-04,  5.2243e-05,  1.4052e-03,  1.4007e-04,
         1.6030e-04, -7.3742e-05, -3.1850e-04,  2.0809e-01,  2.4825e-05,
        -4.4692e-04,  2.3252e-04,  3.7091e-02, -2.4625e-04,  2.8716e-05,
         2.8177e-04,  3.1866e-05,  1.3400e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0343e-04, -3.3484e-01, -3.4943e-04,  7.8954e-05, -1.3715e-04,
        -5.7547e-01, -8.4161e-06,  2.0953e-05,  1.0212e-04, -2.7357e-05,
         2.8611e-04, -2.3276e-04,  3.7770e-04, -2.2252e-04, -1.0584e-04,
         1.5471e-05, -4.6044e-04,  4.7128e-04, -1.3762e-02,  1.8499e-04,
         1.8840e-06,  1.1080e-04, -7.0777e-02,  7.3406e-05, -3.3749e-04,
        -2.9242e-04, -1.2679e-04, -5.5813e-05, -1.9426e-04,  6.9169e-05,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1250: [tensor([ 1.2299e-03, -4.2882e-04, -1.9705e-05, -6.0511e-05,  2.3407e-05,
        -2.0406e-05, -9.7338e-05, -1.5858e-05, -5.3612e-01, -1.7134e-04,
        -5.2898e-05, -9.8428e-05,  1.3463e-04,  2.5616e-06, -9.2089e-03,
        -1.4313e-04,  1.8608e-05, -6.1598e-05, -5.0397e-04, -3.0941e-01,
        -1.5722e-04, -1.8071e-04,  3.1829e-04, -4.1907e-04, -1.4057e-01,
        -4.3984e-04, -6.8931e-05, -2.8933e-05,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1330e-04, -1.7861e-01, -2.0779e-05, -4.0875e-04, -2.5119e-06,
        -4.9738e-05, -1.4796e-01, -5.0809e-05, -7.8962e-04, -3.2697e-04,
        -3.9992e-04, -8.3796e-05, -6.3958e-04, -1.6815e-04, -4.6346e-04,
        -2.2538e-04, -3.8770e-04, -2.9143e-05, -1.3858e-04, -3.3596e-04,
        -2.3409e-04, -3.8031e-01, -8.5317e-05, -1.5621e-03,  7.7599e-05,
        -2.9250e-04, -2.8509e-01, -4.8559e-04, -3.0331e-04, -1.5767e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4208e-03, -2.9445e-01, -2.7819e-04,  3.2593e-05,  1.2114e-04,
         5.8464e-05, -3.7965e-01, -2.7292e-04, -1.2471e-03, -3.4045e-04,
        -2.4359e-04, -1.8836e-04, -7.3700e-04, -7.2933e-05, -1.7542e-04,
        -2.2285e-05, -1.5252e-04,  1.8259e-04,  3.3429e-04, -2.2694e-01,
        -1.6063e-04, -5.7931e-04, -9.1777e-02,  3.6549e-04, -2.0993e-05,
        -1.7858e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.9031e-04,  1.5449e-01,  1.5678e-04, -8.5146e-04, -4.1257e-04,
         1.7142e-04,  7.5526e-01, -1.6693e-04,  3.3478e-03, -2.4001e-04,
        -1.6089e-04, -3.2126e-04,  1.1032e-03, -1.7919e-04, -3.4574e-04,
        -1.1185e-05, -1.6731e-04,  3.5452e-04, -3.2637e-04, -6.8914e-05,
        -2.6239e-04,  7.5133e-02, -3.5709e-04,  4.9814e-03, -5.9749e-05,
         9.4700e-05, -1.9033e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.0587e-04, -2.8339e-04, -5.9594e-05,  1.6030e-04, -5.0198e-02,
        -9.6920e-05, -8.8922e-04, -1.0716e-04, -3.5854e-03, -2.7704e-04,
         1.7637e-06, -4.3316e-04, -3.7600e-04, -7.1667e-05, -3.6337e-04,
        -6.3139e-01, -5.8501e-04, -1.1072e-03, -8.2393e-04, -5.6604e-05,
        -3.0708e-01, -2.0593e-04, -6.5578e-04, -2.9245e-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([-3.8821e-03,  6.9310e-04,  3.4967e-04,  3.4312e-04,  9.2142e-02,
         3.0254e-04,  9.5546e-04,  6.3056e-04,  3.1997e-03,  1.2118e-03,
        -4.6124e-04,  8.5455e-04, -2.6168e-04,  3.8914e-04,  3.0825e-05,
         8.9225e-01,  3.3649e-04,  2.1165e-04,  7.7546e-04, -2.8164e-04,
         4.3980e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7054e-03, -4.6857e-04,  4.5043e-04, -2.6527e-04,  2.5637e-01,
         7.1391e-04,  1.3747e-04,  6.7088e-04,  5.1624e-03,  1.0046e-03,
        -4.1984e-04, -1.5199e-05, -4.8003e-04,  7.5210e-04, -2.1361e-04,
         7.2684e-01,  4.9679e-04,  2.5337e-04, -8.4703e-04, -6.2626e-04,
        -1.0733e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7397e-03, -8.4188e-01, -6.2898e-04, -1.3980e-03, -2.8313e-04,
        -8.1887e-04,  1.0063e-04,  6.8883e-05, -1.9178e-04, -2.7161e-04,
        -3.5697e-04,  5.5349e-04,  5.2063e-05, -1.4956e-01,  2.6395e-04,
        -1.0564e-03, -5.4737e-04,  3.0850e-05, -1.9323e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3418e-04,  9.6073e-01, -4.3547e-04,  3.3006e-02,  8.1050e-04,
        -1.7418e-04, -3.2235e-04,  5.7395e-04, -5.6188e-04, -3.0325e-04,
         1.0170e-03,  4.1460e-04, -1.4168e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3739e-04, -8.4706e-01, -1.1584e-04,  1.6288e-03, -3.7157e-04,
        -1.4292e-04, -4.0819e-05, -2.5132e-04, -4.6810e-05,  1.5577e-04,
        -1.3960e-04, -7.0331e-04, -1.4857e-01,  7.4774e-06, -3.6784e-04,
        -6.5793e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.9199e-04,  1.5965e-01,  3.5214e-04, -1.3914e-04, -1.7177e-04,
        -3.1450e-04, -3.3961e-04,  1.6765e-04, -3.1755e-05, -7.9457e-04,
        -3.5438e-04, -4.5021e-04,  5.7758e-04,  1.9215e-04,  8.3171e-01,
         2.0717e-04, -2.6742e-04,  1.1875e-03,  4.7944e-04,  5.2292e-05,
         5.2523e-04, -7.0878e-04,  3.8183e-04, -1.4657e-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([-2.2040e-03, -4.2916e-01,  1.7925e-05, -2.9963e-04, -8.9950e-05,
        -2.2231e-04,  3.7298e-04,  2.3801e-04,  6.8186e-05,  1.0190e-05,
        -5.1209e-04, -9.4124e-05, -1.4119e-04,  1.2124e-04, -4.2060e-01,
         5.0500e-04, -4.7766e-04,  1.1579e-04,  2.8410e-04,  3.7452e-04,
        -1.4350e-01, -7.6674e-05, -3.3201e-04,  1.1032e-04,  8.0091e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1300: [tensor([-1.3621e-03,  5.8942e-01,  2.4329e-04, -5.7308e-05,  3.7050e-01,
        -4.9111e-05, -1.6847e-04, -2.7545e-04,  3.6946e-04,  3.0684e-02,
         8.2452e-05, -7.7199e-05, -9.9093e-05, -1.3536e-04, -1.2237e-04,
         3.2410e-04,  7.9329e-05,  5.3428e-05, -1.7932e-04, -2.2092e-04,
        -1.8552e-04,  1.8008e-05,  4.7598e-03,  7.1731e-05, -6.9294e-05,
         3.4845e-04,  3.8676e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3169e-03,  3.3936e-01,  1.1158e-05,  2.6883e-04,  2.0048e-01,
        -5.6359e-05,  1.2327e-06, -1.5903e-04,  1.2779e-04,  1.0468e-01,
        -5.0235e-05,  9.4441e-06,  1.2527e-04,  1.9904e-04, -8.9748e-05,
         5.9583e-05, -1.0804e-05, -1.5456e-04,  1.8404e-04,  2.7876e-05,
         3.4777e-01,  2.4601e-04,  2.6305e-04,  1.0300e-04,  5.3982e-05,
         2.5434e-04, -1.3794e-04,  2.2571e-05, -6.5562e-05,  1.9067e-03,
         2.0697e-04,  1.8071e-04,  3.8157e-04, -3.6065e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7897e-05, -2.4288e-01, -1.6652e-04, -3.7792e-04, -3.8931e-02,
         1.7508e-04, -3.8356e-05, -4.2781e-04, -3.1394e-04, -2.8111e-01,
         2.4088e-04,  9.9807e-05, -4.1085e-04,  7.6351e-05, -1.0205e-04,
        -1.8110e-05, -9.3485e-05,  2.1845e-04, -6.5208e-05, -2.1377e-04,
        -3.2933e-02, -1.6536e-04, -2.2204e-04, -2.9640e-04, -3.9999e-01,
         6.1559e-05, -1.9240e-04, -1.4280e-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([-6.4988e-04,  1.4682e-01,  8.3255e-05,  5.6250e-05,  4.1950e-02,
         4.8805e-05, -9.2785e-05,  3.0003e-04,  1.8305e-04,  3.3464e-02,
         3.2250e-05,  1.7487e-04, -4.0926e-04,  2.0880e-04, -1.7634e-04,
         2.3009e-04, -4.0851e-05, -2.2276e-04, -1.2249e-04, -5.9916e-05,
         4.4741e-01,  3.9523e-05,  8.4177e-04,  1.4134e-04,  3.2562e-01,
        -2.0125e-05,  4.9766e-04,  9.2553e-05, -7.7251e-06,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.3552e-04,  2.8215e-01,  1.4416e-04, -1.2899e-04,  2.0511e-01,
         2.0153e-04, -4.7766e-04, -1.6996e-05,  2.3505e-04,  8.7683e-02,
         8.7420e-05,  3.8575e-05, -2.2065e-04, -4.9100e-05,  5.9035e-05,
        -3.8644e-05,  2.7825e-05, -2.4769e-04,  2.4283e-04,  1.9335e-05,
         2.4793e-01,  1.5304e-04,  1.9709e-04,  1.7330e-01, -4.5228e-05,
         1.4908e-04, -6.2342e-05,  1.9710e-04, -5.3702e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1075e-03,  2.2548e-02, -2.8742e-04, -3.0014e-06,  1.2791e-01,
        -1.7035e-05, -2.6104e-04, -2.9910e-04, -1.3455e-04,  1.2365e-01,
        -1.0336e-05,  1.8092e-04,  4.1999e-04, -1.1817e-04, -7.3053e-04,
        -4.2587e-05, -7.3001e-05, -4.0625e-05, -1.0414e-04, -4.2456e-05,
         7.1967e-01, -2.6843e-04, -2.2782e-04,  1.2207e-04, -4.6507e-05,
         1.3296e-04,  4.7607e-04,  7.4866e-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([ 2.3370e-03, -2.0707e-01,  1.2670e-04, -1.3358e-04, -5.1446e-01,
         9.7184e-05, -1.2826e-03, -1.4553e-04, -2.9926e-04, -5.8013e-02,
        -3.1578e-04, -8.0382e-05, -2.4132e-04, -2.0189e-05, -9.1061e-05,
        -1.3353e-04,  4.0740e-06, -2.6816e-04, -1.8169e-05,  2.1825e-04,
         4.3324e-05, -1.3380e-04, -2.1373e-01,  3.7117e-05, -4.3296e-04,
         2.3003e-04, -3.5522e-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([-1.8752e-03,  3.0196e-02, -9.6492e-05,  1.0155e-04,  3.4958e-01,
        -6.5178e-05, -6.6610e-04, -6.0028e-05,  1.6578e-04,  5.3357e-03,
        -2.0692e-04,  1.4100e-04, -1.8904e-04, -8.4482e-05, -2.8571e-04,
         3.3145e-04, -1.4411e-05, -2.0212e-04,  1.8568e-04, -2.1128e-04,
         1.2665e-03,  2.4619e-04,  2.1411e-04,  2.4120e-04,  1.4733e-04,
         6.0625e-01, -3.2609e-04, -7.6533e-04, -3.1873e-04,  1.3993e-04,
        -8.4080e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2537e-03, -7.3349e-02,  3.5991e-04,  3.1946e-05, -4.7753e-01,
        -3.4550e-05, -1.0699e-04, -7.6283e-04, -3.0485e-04, -2.4530e-04,
         2.2855e-04,  1.3716e-04,  2.3846e-05,  1.6211e-04,  9.9234e-05,
        -3.8879e-01, -2.4588e-04, -2.0431e-03, -5.2843e-02, -1.5069e-04,
        -6.1857e-04,  2.4165e-04, -4.4344e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8718e-03,  5.9774e-02, -4.2327e-04,  4.6339e-04,  2.9098e-01,
         7.4746e-05, -1.5269e-04, -6.0767e-04,  1.5906e-04, -9.8314e-05,
        -2.2468e-04,  1.8317e-04,  4.9065e-04,  1.0732e-04,  1.3024e-04,
         6.4122e-01, -5.4007e-04, -3.6863e-04, -3.8220e-04,  3.4534e-04,
         2.6209e-04, -8.5149e-05,  5.3084e-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([ 1.4370e-03, -2.6331e-01, -3.9858e-04, -2.8358e-04, -2.7707e-02,
         1.5746e-04, -1.9391e-04, -7.3613e-04, -2.2481e-05, -1.6564e-04,
        -1.1621e-03,  4.1192e-04,  5.7035e-05,  7.2779e-05,  3.7024e-04,
        -6.9460e-01,  3.7819e-04,  2.6662e-04, -7.4027e-03, -5.0306e-05,
        -5.8229e-04,  2.3227e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9664e-03,  1.2916e-04,  1.7100e-02,  6.8810e-05,  2.1709e-05,
         5.9510e-02,  1.1770e-04, -2.5404e-04,  3.2925e-04, -7.8959e-05,
         7.4606e-02,  3.4195e-04,  1.0457e-04,  1.2576e-04, -1.4843e-04,
        -2.2095e-04, -2.3070e-04, -5.6639e-05, -2.3195e-04, -4.1735e-04,
         8.1066e-01,  2.3400e-04,  4.5063e-03,  2.6847e-02,  1.6070e-05,
         4.1473e-04,  1.1602e-04,  5.3079e-05,  9.4980e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1350: [tensor([-2.5823e-03, -2.7159e-01,  4.5599e-06,  1.2775e-05, -2.4668e-01,
        -1.0714e-04,  2.8527e-04, -2.0284e-01,  7.2052e-05,  5.6656e-05,
         4.8513e-05,  2.7206e-04, -1.6801e-01,  2.9225e-05,  4.0075e-05,
        -2.3455e-05, -1.4250e-03, -8.9818e-02, -4.1820e-05, -2.8920e-05,
         6.6431e-05,  3.0580e-04,  1.7252e-04, -7.9884e-04, -2.5168e-06,
        -1.1105e-04, -9.8226e-05, -1.8630e-04, -3.6620e-04, -1.2988e-02,
        -9.6321e-05, -2.4007e-04, -2.4793e-04, -1.9005e-04,  1.6000e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6711e-03, -5.7321e-01, -4.0028e-04,  1.0237e-04, -2.7037e-02,
        -4.1271e-05,  5.8381e-05, -9.7745e-02, -6.8415e-05, -1.8051e-04,
        -2.3263e-04, -1.5869e-04, -4.7581e-02, -2.5097e-05,  1.2848e-04,
        -4.3893e-05,  6.9114e-04, -2.5334e-02, -1.7558e-04, -1.6892e-04,
         3.0225e-05, -3.1811e-04, -1.0427e-01,  6.2329e-05,  1.3729e-03,
        -2.9798e-05, -1.1859e-01, -6.2902e-06,  1.1454e-04,  6.9023e-05,
         3.1318e-05,  5.4047e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0379e-03, -5.4884e-02, -4.7105e-05, -1.7609e-04, -2.6918e-04,
        -1.6571e-04, -2.7441e-04, -2.2404e-01, -2.4882e-05, -4.9133e-05,
        -1.5789e-05,  5.1884e-04,  6.4192e-05, -2.7279e-04,  1.0727e-04,
         7.7411e-05, -3.1297e-02,  9.7864e-05,  7.9453e-05, -3.2343e-01,
        -6.4224e-05, -5.0405e-05,  3.0927e-04, -1.0857e-04, -5.5930e-05,
        -2.5558e-04,  1.3563e-04,  3.1806e-05,  4.9360e-06,  5.8678e-05,
        -2.8041e-01, -3.5558e-04, -3.6018e-05, -3.4475e-05, -8.0080e-02,
         6.1856e-05,  1.7328e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.3574e-04, -3.9045e-01, -2.6886e-04, -1.8616e-04, -3.7946e-04,
        -1.0308e-04, -2.4422e-04, -1.1632e-01, -2.1390e-04, -5.7112e-05,
        -1.3845e-04,  9.9869e-04, -1.5639e-04, -2.1834e-04, -5.1114e-05,
         1.0557e-04, -7.9642e-02,  6.4916e-05, -1.3285e-04, -3.4696e-01,
         1.6907e-05, -1.3167e-04, -4.8142e-05,  6.4697e-06, -1.3610e-04,
        -1.4501e-04,  9.6191e-05, -1.1217e-05, -1.1149e-04,  8.6587e-05,
        -5.9812e-02, -1.3656e-04, -1.0339e-04,  1.3825e-03, -5.0553e-05,
        -7.5446e-05,  1.2700e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2120e-04, -2.9999e-01, -4.1255e-06,  2.3590e-04,  7.0161e-06,
         2.4810e-05, -1.6725e-04, -3.9953e-01, -1.6455e-05, -1.2313e-04,
         1.2102e-05,  4.7174e-04,  1.5209e-04, -6.1504e-05,  3.8337e-04,
        -1.5677e-04, -1.5046e-01, -1.8960e-05, -8.5523e-06, -1.3319e-01,
        -1.5856e-04,  2.6599e-05,  4.3321e-05,  1.7428e-04,  1.5360e-04,
        -1.4212e-04,  1.2914e-04,  9.9022e-05, -1.2211e-04, -2.9307e-05,
        -1.2747e-02, -3.0347e-04,  1.4589e-04,  2.6255e-04, -1.8131e-05,
        -9.7851e-05,  6.4315e-06], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5042e-03, -1.4579e-01, -4.5311e-04, -1.9968e-04,  2.0240e-03,
        -2.4458e-04, -1.5645e-04, -2.5127e-04, -3.3961e-03, -4.2389e-04,
         5.6255e-04, -3.5587e-04,  2.0807e-04,  4.0003e-06, -3.2484e-01,
        -1.3827e-04, -2.1836e-04,  2.3618e-04, -5.1571e-01,  3.3826e-04,
        -2.2850e-04,  3.8258e-04,  3.3909e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5133e-03, -6.1041e-01,  3.5214e-05,  1.8383e-04,  3.5904e-04,
        -4.5191e-04, -7.6169e-05,  1.9872e-04, -1.9581e-01,  8.7993e-04,
         2.2264e-04,  2.6592e-04,  8.0835e-05, -1.3245e-04, -1.8839e-01,
        -1.2340e-04, -2.8422e-04,  2.8764e-05,  2.2955e-04,  3.2863e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.7680e-04, -2.0985e-01, -4.0777e-06,  3.1685e-05,  2.0516e-03,
         4.7395e-04,  7.2346e-05,  1.5631e-04, -2.4654e-01,  1.8848e-04,
        -6.3251e-06, -2.6060e-04,  6.8058e-05,  4.8576e-05, -3.7089e-01,
         1.2578e-04, -2.3418e-05, -2.3025e-04,  2.9887e-05, -1.6755e-01,
         1.5861e-04, -1.1392e-04, -1.0454e-04, -2.4373e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4808e-03, -1.8831e-02,  5.6895e-04,  4.7579e-04, -3.0171e-01,
         1.0735e-03, -4.2383e-05, -5.9847e-02, -6.2399e-05,  5.5571e-04,
         2.3913e-05, -6.1083e-01,  4.5481e-04, -6.7777e-05, -5.2053e-04,
         1.3656e-04,  8.5508e-05,  2.4104e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0625e-04, -2.6763e-01,  5.3349e-04,  6.2824e-05, -3.5886e-02,
        -1.7497e-05,  2.7381e-04, -1.4832e-01,  6.9252e-04,  2.8145e-05,
         3.7241e-05, -5.4310e-01,  1.6580e-04,  1.2667e-04,  2.7450e-04,
         1.1806e-03,  8.9649e-04,  2.0837e-05,  2.4751e-04, -4.0258e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.6840e-04, -3.1233e-01, -1.9955e-04,  7.4104e-05, -5.4564e-02,
        -1.2430e-04,  9.5319e-05, -6.5587e-02,  8.5660e-05, -1.0117e-04,
         3.6834e-05, -2.8332e-01, -1.1706e-04, -1.2040e-04, -1.6627e-01,
        -1.1874e-04, -1.3075e-04, -7.0104e-04, -1.1425e-01, -2.5118e-04,
        -4.0680e-04, -1.4022e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7601e-03, -3.1599e-01,  3.9784e-04, -3.1086e-04,  2.4536e-04,
         8.1754e-04,  1.1006e-04, -5.6028e-01, -1.5283e-04, -3.7865e-04,
        -1.1265e-04, -1.1777e-01,  3.7207e-04,  2.0084e-04,  3.1754e-04,
         3.5800e-05, -2.4709e-04, -5.0042e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1400: [tensor([ 8.4349e-04, -1.5277e-01, -4.2595e-04, -3.3341e-04, -3.1841e-01,
        -2.1254e-04, -2.6870e-04, -9.6177e-05, -4.8393e-04,  1.3579e-04,
        -4.2738e-04, -1.1598e-04, -5.2263e-01, -3.9612e-04, -2.3458e-04,
        -7.1177e-04, -1.2786e-04, -2.7299e-05,  2.2009e-04,  1.3569e-04,
        -3.9643e-04, -3.5481e-04,  1.8121e-04, -6.7317e-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([ 1.0619e-04, -3.7989e-01, -3.6204e-05, -3.6637e-05, -3.3173e-01,
        -2.4235e-05,  7.3628e-05, -5.2743e-04, -3.2027e-04,  3.2593e-04,
        -9.6507e-05,  7.1950e-05, -6.9080e-02,  1.2404e-04,  2.7954e-04,
         6.7702e-05,  2.7484e-04,  2.2919e-04, -6.4151e-05, -1.3189e-02,
        -1.3624e-04, -2.1985e-03, -2.0049e-01,  2.0699e-04, -5.9076e-05,
        -3.6148e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8089e-03, -1.0241e-01,  3.0104e-04, -6.8952e-05, -5.3403e-01,
        -8.6998e-05, -3.5684e-04,  1.1664e-04, -6.8909e-05,  5.8159e-04,
        -7.5352e-05,  5.9577e-05, -6.0478e-02, -2.3555e-04,  1.4453e-06,
        -4.9071e-04,  1.6021e-04, -4.8293e-05, -2.7063e-04, -2.9524e-01,
         4.2412e-04, -1.4708e-03, -2.6472e-05,  1.9533e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7280e-03, -4.0232e-01,  1.3131e-04, -5.0129e-04, -1.5604e-01,
        -7.2909e-05, -4.5807e-05,  1.0821e-04, -8.9869e-05,  5.2782e-05,
         3.3581e-05,  1.8697e-04, -1.9660e-04,  1.1436e-04, -1.9782e-01,
         3.6862e-05, -1.0365e-04, -3.3007e-04, -5.5009e-04,  8.6242e-05,
         8.9362e-05,  9.9557e-05, -1.2164e-01, -8.1277e-05, -2.1632e-04,
         4.4347e-05, -3.0743e-05, -1.0747e-04, -1.8541e-05,  1.1481e-04,
         1.4765e-04, -1.2972e-05, -4.9129e-05,  1.0367e-04, -1.2981e-04,
         2.2956e-05, -3.8383e-05, -2.0721e-04, -1.4064e-05, -1.5014e-05,
        -6.0687e-06, -1.1562e-01,  3.1542e-05, -6.0568e-05, -5.9922e-05,
        -8.3102e-05, -3.1102e-04,  5.2253e-05, -3.9950e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5149e-03, -9.7872e-02, -1.6685e-04, -5.2416e-04, -4.2801e-01,
        -5.9480e-06,  5.6095e-05, -2.6142e-04,  1.5233e-04, -5.1719e-05,
         4.0726e-05,  4.1150e-04,  1.3682e-05, -7.5182e-05, -1.3624e-01,
        -4.6297e-05, -3.1429e-04,  3.8546e-04, -1.4099e-04,  1.4480e-04,
        -1.8368e-04, -5.6031e-05, -8.9228e-02, -6.3370e-05, -8.7807e-05,
        -1.4927e-04,  2.8522e-04,  1.7726e-04,  2.8072e-05,  9.2723e-05,
        -3.7473e-05, -2.4279e-01,  8.5131e-05, -1.6011e-04,  1.5037e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8904e-03, -1.0834e-01,  1.1799e-04, -4.0260e-04, -1.7685e-01,
        -1.5907e-04,  1.8661e-04, -6.4474e-06, -1.6079e-04,  1.3170e-04,
         5.0693e-05,  4.9105e-04,  3.9616e-04, -4.1566e-05, -3.7030e-01,
        -1.0448e-04,  1.2860e-04,  8.5406e-05,  3.5660e-04, -1.6395e-05,
        -1.7856e-04,  3.8737e-05, -4.7247e-02,  3.6406e-04,  4.8174e-04,
         5.9163e-06,  1.0526e-04, -2.9093e-01, -2.2479e-04, -8.9511e-05,
        -2.2612e-05, -9.1667e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0903e-03,  6.7955e-01,  1.5410e-04, -2.9701e-04,  2.8639e-04,
         1.7991e-04,  4.5870e-04,  3.6099e-05, -2.7855e-05,  1.0908e-04,
         7.8695e-03,  1.4775e-04,  3.0603e-03,  3.7491e-04,  1.9055e-02,
         4.0823e-04,  3.6388e-04, -9.2120e-06,  7.8936e-03, -2.1153e-05,
        -6.5851e-05, -3.6530e-04,  2.6462e-01,  2.8811e-04,  8.4369e-06,
         5.5560e-04, -3.3848e-04, -1.1554e-03,  4.2150e-04, -5.4956e-04,
         3.6867e-04, -1.0624e-04,  2.8925e-03, -5.3599e-04,  2.9290e-04,
         6.7080e-04,  3.3592e-04,  3.1999e-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([ 2.7326e-05, -6.1548e-02,  2.3597e-04,  1.1362e-04,  3.9317e-04,
        -3.6790e-04, -2.1446e-04, -2.0186e-04,  1.8600e-05,  4.1244e-04,
        -3.9554e-04, -3.9659e-04, -8.5069e-03, -8.2470e-05, -1.1348e-03,
        -3.7386e-04, -1.3222e-04,  2.7914e-04, -3.2361e-03, -1.5865e-04,
         3.4207e-04, -3.0121e-07, -9.1655e-01,  2.4082e-04,  1.0278e-04,
        -1.1988e-03, -1.5248e-03, -4.8819e-05,  2.2809e-04, -4.1035e-05,
        -1.4757e-03, -1.4998e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4390e-03,  7.0774e-01, -1.6968e-04, -1.6921e-04, -4.5290e-04,
        -1.7782e-05,  3.0967e-04, -3.2881e-04, -6.5726e-05, -2.3853e-05,
         2.5286e-03, -3.1093e-04,  7.3533e-04, -2.3719e-04,  9.2649e-03,
         2.0763e-04, -2.0169e-04, -3.9692e-04,  7.8455e-03,  2.9545e-04,
        -1.0045e-05, -8.6895e-06,  4.8647e-02, -1.1867e-04,  9.6575e-04,
        -1.6760e-04,  8.3654e-05, -1.8597e-04,  8.6899e-05, -3.9416e-04,
         2.1573e-01, -1.1958e-04, -5.6447e-04, -1.6922e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8559e-03, -3.7373e-04,  4.8362e-05,  7.1405e-06,  4.7096e-04,
         3.5846e-03,  3.2501e-01, -8.1700e-05,  3.1427e-03, -1.7837e-04,
         5.2263e-02, -5.0257e-04, -1.3377e-03,  1.4833e-01,  4.1653e-04,
        -1.2063e-04, -4.8124e-04,  1.7301e-02, -8.4450e-05,  5.6587e-05,
        -1.8551e-04,  1.7645e-02, -1.3095e-05,  1.6589e-06, -1.7442e-04,
         4.2346e-01, -2.1591e-04, -1.5629e-04, -1.0904e-04,  1.2882e-04,
         2.2525e-04, -4.1220e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.0074e-04, -1.4147e-04, -1.9913e-04, -7.6218e-05,  4.8272e-05,
        -4.5310e-03, -2.7657e-01,  5.9323e-05, -5.9505e-04, -1.9935e-04,
        -3.0837e-01, -1.7074e-04, -4.7645e-04, -2.5335e-01,  8.7109e-05,
        -4.2167e-05,  8.0487e-05, -3.0873e-02,  1.1444e-04,  2.8058e-04,
        -3.2383e-05, -5.8934e-02, -3.0941e-05,  6.7499e-04,  2.1497e-04,
        -1.5969e-04, -7.3654e-05, -2.3722e-02, -1.1654e-04,  8.0134e-05,
        -3.8596e-02,  3.1110e-04,  1.4130e-04, -4.9230e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8484e-04, -3.9734e-05,  8.6906e-05, -7.7094e-05,  3.7662e-05,
         5.5105e-05, -6.0532e-01, -1.6578e-04, -4.3658e-03, -2.2258e-04,
        -2.7882e-02, -7.5674e-05,  1.2926e-03, -2.4632e-02,  1.9926e-04,
        -1.0942e-04,  9.5298e-05, -8.6112e-03, -1.5792e-04, -4.4469e-05,
         4.2351e-05, -1.8979e-01, -1.7127e-04, -1.3526e-03, -1.6901e-04,
         2.0296e-05, -2.5024e-04,  1.5395e-03, -3.0569e-05, -9.7097e-06,
        -1.3955e-04,  4.0809e-05,  2.8521e-04, -1.3219e-01,  1.3044e-04,
         2.1497e-05,  5.0769e-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>)]
Iter #1450: [tensor([-2.6651e-03, -2.8170e-01,  5.6490e-04, -5.1400e-04,  5.6390e-04,
        -3.9935e-04, -7.0010e-03, -3.6076e-01, -1.9958e-04,  5.0024e-04,
         6.1561e-04, -1.4014e-04, -3.4107e-01,  4.7739e-04, -2.3513e-03,
        -1.8518e-04, -2.8789e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4312e-03, -5.2208e-02, -3.8046e-04, -6.4093e-04,  4.2264e-04,
         6.4694e-04,  4.6156e-03, -9.3426e-01, -8.9280e-05,  1.1885e-03,
        -6.5582e-04,  2.8376e-04, -3.2335e-04, -8.1244e-04, -4.7231e-04,
        -1.3749e-04,  4.2851e-04,  2.6322e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4950e-03, -6.2177e-05,  2.3344e-04,  1.8032e-04, -2.6845e-01,
        -5.7298e-05,  2.3660e-04,  2.5023e-04,  7.7403e-05,  1.1435e-04,
         1.8161e-04, -6.8008e-01, -2.6537e-04, -3.1551e-05, -4.2278e-03,
        -4.2822e-05, -2.3815e-04,  1.5914e-04, -3.8003e-02, -3.4159e-05,
        -2.4234e-04, -5.7823e-04, -2.1072e-05, -3.7770e-04,  2.5140e-05,
        -2.7213e-04, -6.3717e-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([ 2.7588e-04, -2.4028e-04,  3.9014e-04, -6.1474e-05, -1.4804e-02,
         3.3228e-04, -3.4458e-06,  1.1218e-03, -2.8758e-04, -1.2441e-05,
         1.5413e-05, -1.6302e-01,  1.2014e-04,  8.8484e-05,  6.9620e-03,
         2.3332e-05,  9.3918e-05, -2.2152e-04, -2.8682e-01,  4.8282e-04,
         7.5588e-04,  5.2240e-04, -2.4624e-04,  9.0440e-05, -5.2192e-01,
         8.5051e-05, -2.1929e-04,  3.7157e-04,  2.5088e-04,  1.6092e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9147e-04, -1.2861e-04, -2.5991e-04, -4.6402e-04, -2.4209e-01,
        -7.5727e-05,  1.7554e-05,  2.2350e-04, -6.0454e-04,  2.1752e-05,
        -5.0231e-04, -2.9311e-01, -2.1068e-04, -2.1597e-04,  1.6903e-03,
        -1.4258e-04,  8.5806e-06, -2.6074e-04, -2.6820e-01,  5.5091e-05,
         7.7024e-05,  8.8427e-05, -4.7957e-04, -2.2416e-04, -1.9435e-04,
        -4.2788e-04, -1.8897e-01, -3.4722e-04, -2.4329e-04,  6.2636e-05,
        -1.2024e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0624e-03, -1.9627e-01, -3.2876e-05, -2.3857e-04, -2.7691e-04,
        -1.3520e-04, -2.7740e-05, -8.1974e-02, -1.6089e-04, -2.8695e-05,
         3.3893e-05, -1.3511e-04, -2.7493e-02, -3.6706e-05,  9.2315e-05,
        -3.2962e-01, -3.0448e-04, -3.9419e-04, -2.7010e-05, -1.6928e-01,
        -7.5237e-05,  2.3816e-04,  1.0732e-04, -3.6035e-02, -5.0041e-05,
        -1.9906e-04,  1.7637e-04, -3.6898e-04,  6.4912e-04, -1.5419e-01,
        -2.2999e-04, -5.2883e-05, -9.6656e-07,  8.6910e-06,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0329e-03, -4.4123e-01, -1.8541e-05, -4.0545e-04, -2.8665e-04,
        -1.9964e-04,  1.5862e-04, -1.2795e-01,  3.5906e-05,  6.6456e-05,
         1.5888e-04,  9.0110e-06, -1.2440e-01, -3.9979e-05,  9.0556e-04,
        -4.0763e-02, -5.4988e-04, -6.3409e-05,  1.3997e-04, -8.2183e-02,
        -1.1447e-04, -3.6394e-05,  1.0693e-04, -1.7615e-01,  8.2735e-07,
         2.1567e-03, -6.6726e-05, -2.0266e-04, -3.1891e-05, -4.6838e-04,
         7.1954e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1267e-05, -2.3754e-01, -4.6190e-05, -2.5157e-04, -2.8331e-04,
        -7.3432e-05, -1.5951e-04, -1.5462e-01,  1.0164e-05, -3.6368e-05,
         4.0126e-04, -6.2034e-05, -1.6263e-01, -3.2010e-04, -4.9548e-02,
        -5.8432e-02, -3.3095e-04, -4.1190e-05, -2.1643e-04, -9.0196e-02,
        -2.2550e-04,  1.5551e-04, -5.6360e-05, -2.4151e-01, -2.0224e-04,
         1.7354e-03,  1.2293e-04,  1.4384e-04, -3.1035e-04, -3.5664e-05,
         6.4986e-05, -2.1922e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1452e-03, -1.4864e-01,  3.2020e-05,  4.2571e-05, -1.2085e-04,
        -2.6326e-04, -1.2993e-05, -7.0059e-02,  3.4008e-04,  7.2572e-05,
         4.0215e-05, -1.4905e-04, -9.6944e-02, -6.1907e-05,  2.0328e-03,
        -7.1207e-02, -3.9650e-04,  1.5051e-04,  1.1227e-04, -3.2435e-01,
        -1.5710e-04,  1.0244e-04,  9.1121e-05, -1.0751e-01, -1.3466e-04,
         1.0632e-05, -8.2802e-05, -1.7564e-01, -3.8443e-05,  3.3642e-05,
         2.1478e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.3197e-04, -3.3558e-01, -1.6048e-04,  3.0602e-04, -1.3071e-04,
         1.8927e-04,  1.9009e-04, -2.3676e-01,  4.1674e-05, -3.3137e-04,
         6.1829e-04,  1.2440e-04, -3.0026e-01,  3.2032e-04, -5.8563e-03,
        -5.5447e-02,  7.9751e-05,  3.0185e-06,  1.3851e-05, -3.2172e-02,
        -1.4078e-04,  1.4616e-04,  6.8738e-05,  6.0594e-05, -2.3671e-05,
        -2.9889e-02,  1.4116e-04, -4.9871e-06, -1.0605e-04, -1.8287e-04,
        -1.5705e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7296e-04, -1.5563e-01, -1.6999e-05,  4.4601e-05, -1.4934e-05,
        -1.9716e-04,  3.2023e-05, -3.2009e-02,  2.7772e-05,  7.8645e-05,
        -9.5019e-05, -1.4170e-04, -3.8046e-01, -3.0800e-05, -2.0675e-03,
        -9.4252e-02, -5.4216e-04,  8.2309e-05, -1.2201e-05, -1.4155e-01,
         6.9796e-05,  1.4117e-04,  1.3708e-04, -1.8162e-01, -1.6952e-05,
         2.5423e-05, -1.0454e-02,  2.6191e-06, -7.5399e-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([ 2.9556e-04, -3.6296e-01,  1.4104e-04, -1.0456e-04,  2.3581e-05,
         1.7765e-05, -2.4525e-04, -1.0089e-01, -5.0673e-05, -1.3600e-04,
         3.1218e-04,  7.6950e-05, -1.2963e-01,  1.1671e-04,  3.6674e-04,
        -1.4287e-01, -1.9950e-04, -1.9878e-04, -3.8964e-04, -2.9702e-02,
        -3.7441e-05, -1.1001e-04,  1.2208e-04, -2.2942e-02, -1.4949e-04,
        -9.8151e-05, -1.3386e-04,  1.0489e-03, -1.8790e-04, -2.0206e-04,
         4.1193e-05, -2.4006e-04, -2.0569e-01, -2.6005e-04, -5.3287e-06],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1500: [tensor([-2.6457e-03, -1.2639e-03,  9.1921e-04,  1.9900e-04, -1.2668e-04,
        -4.9072e-01, -4.2301e-04,  2.5914e-04, -2.3048e-04,  4.1331e-04,
         4.3318e-04, -2.6355e-01, -5.2516e-02, -6.5657e-04, -1.7509e-01,
        -4.0778e-04, -2.0410e-04, -1.2870e-03,  8.0576e-04,  4.7400e-03,
        -1.3384e-03, -4.4977e-04,  1.0957e-03,  2.2809e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2163e-03, -7.1053e-01,  4.4576e-04, -4.9694e-04, -4.4760e-04,
         2.4897e-03,  3.3520e-06, -4.6013e-05,  1.8448e-04, -7.3482e-05,
        -6.7041e-03,  7.6121e-05, -2.9576e-04, -1.8589e-04, -3.4158e-04,
         2.9673e-03, -9.6903e-05, -3.2901e-04, -2.5131e-04,  7.3277e-05,
        -5.0045e-02, -1.4862e-05, -1.2646e-04, -1.6407e-04,  1.7868e-04,
         1.8298e-04, -3.2994e-04,  5.3526e-05, -2.1894e-01, -2.0208e-04,
         7.0714e-05,  8.7068e-05, -2.2779e-04, -1.2126e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3578e-04, -4.0087e-01, -1.4663e-04, -4.6526e-04,  1.6592e-04,
        -5.2448e-02,  6.4520e-05, -4.5215e-04,  1.4534e-04, -2.6517e-04,
        -8.8940e-02, -8.9305e-06,  2.1997e-04, -4.3277e-04,  4.7119e-05,
        -1.2897e-03,  5.9931e-05, -2.9057e-04, -2.9087e-04,  3.8520e-05,
        -1.2179e-01, -3.2645e-05,  2.1411e-04, -7.5660e-05, -1.6550e-04,
        -1.5583e-04,  2.4404e-06,  8.3688e-05, -3.2893e-01, -2.3537e-04,
         2.4416e-05, -1.2100e-03, -7.5763e-05, -2.0993e-04,  9.0654e-06],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0459e-03, -3.2458e-01, -6.7710e-05, -3.9779e-04, -2.0468e-04,
        -5.7157e-02, -2.3253e-04, -2.4260e-04,  1.1390e-04, -1.9709e-04,
        -2.4277e-01, -1.1011e-04, -3.6119e-05, -2.1089e-04,  1.5197e-05,
         1.8879e-03, -2.0056e-04, -5.4461e-05, -6.8916e-04, -5.8272e-05,
        -9.3615e-02, -1.2026e-04,  1.1469e-04,  1.4154e-05, -8.0895e-06,
         1.4440e-04, -1.6467e-04, -1.7633e-04, -2.7425e-01,  2.9791e-04,
        -2.8783e-04,  1.8349e-04,  3.4542e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2859e-04, -4.3936e-04, -5.9739e-03,  2.2300e-04, -3.0313e-04,
         3.4396e-04, -6.0316e-05,  5.4622e-04,  3.1553e-03, -6.4890e-01,
         3.3895e-04, -6.7270e-04, -1.1919e-04, -1.2822e-04, -3.3637e-01,
         4.3503e-04,  1.6708e-04,  8.6238e-04, -2.8806e-04,  4.4846e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.6275e-04, -3.0534e-04, -3.9409e-02, -1.8116e-04,  4.5150e-05,
         1.1399e-04,  1.8483e-04,  6.5567e-04,  8.7986e-04, -5.6134e-01,
        -4.7687e-05, -1.5905e-04,  8.3808e-04, -1.7904e-04, -2.3804e-01,
         1.9154e-04, -4.3690e-04, -3.1843e-04, -1.5577e-01, -1.1899e-04,
         1.0519e-04,  2.0250e-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([ 3.4700e-03,  2.9508e-05, -1.4175e-01,  3.0525e-05, -3.9349e-04,
        -5.0918e-06, -2.8641e-04, -1.2424e-04,  1.0239e-02, -4.8210e-01,
         3.0069e-04, -3.1199e-04,  8.6171e-04, -2.3306e-02,  6.2736e-04,
        -3.3462e-01, -1.0222e-04,  9.2699e-04, -2.4596e-05, -6.0278e-05,
         4.3378e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.2032e-04, -4.4679e-01,  1.4232e-04, -1.0355e-03, -1.0327e-04,
        -3.2504e-04, -2.9829e-01, -6.1439e-05,  7.1351e-04,  1.3779e-04,
        -2.1825e-01, -6.1338e-04, -3.4604e-04, -3.7690e-04, -4.0300e-04,
        -1.1532e-04, -3.0655e-02,  3.4735e-05, -1.3986e-04, -6.7946e-04,
        -7.8511e-05,  8.3003e-06, -1.7893e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4814e-04, -2.5051e-01,  6.2954e-04, -4.3929e-04, -1.4401e-04,
        -2.2347e-04, -2.7642e-01,  1.6274e-05,  2.7129e-04,  3.8396e-06,
        -3.1440e-01, -1.8105e-04, -4.0745e-04, -1.0290e-04, -9.3526e-05,
         1.9633e-04, -3.5476e-02,  2.8648e-04, -2.5390e-04,  1.0083e-03,
        -3.1560e-05, -1.1723e-01, -2.1900e-04,  4.6298e-04,  5.4625e-05,
        -1.5028e-04, -1.1922e-04, -1.7079e-04, -2.3485e-05, -1.3462e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0035e-03, -3.3143e-01,  4.0801e-04, -2.9578e-04, -1.4828e-04,
        -2.1851e-05, -2.4545e-01,  4.2351e-05, -1.0334e-03,  3.7647e-04,
        -1.1305e-01, -2.1750e-04, -5.3618e-04, -5.3980e-04,  1.7017e-05,
        -1.9122e-05, -1.4657e-01,  3.6598e-04, -4.6567e-04, -7.2598e-03,
        -1.3380e-04, -1.5034e-01,  3.2084e-05, -5.7750e-05,  1.8808e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5577e-03, -5.9866e-04,  1.5026e-04, -3.5507e-04,  5.1008e-05,
        -2.7017e-04, -9.9056e-01,  4.4970e-04, -2.5722e-04, -5.8731e-04,
        -3.8893e-04,  9.6104e-05,  3.6908e-05, -8.5103e-04,  5.0656e-04,
         3.3440e-04,  3.5288e-04, -3.0304e-04,  3.4184e-05,  2.5635e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1339e-03,  4.5629e-04, -6.1156e-05, -2.3298e-04,  3.5054e-04,
        -4.2003e-04, -7.7215e-01, -6.1207e-04, -1.4255e-04, -8.6330e-05,
        -1.8304e-02,  2.7395e-04,  3.1467e-04, -2.0203e-01, -1.0110e-03,
        -6.3970e-04,  2.5162e-04, -5.3466e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1550: [tensor([-5.1543e-04,  3.1465e-04, -1.0085e-03, -2.2810e-04,  2.8795e-04,
        -9.8015e-05,  1.0618e-06,  5.4339e-01,  2.4473e-04, -2.4244e-04,
        -3.0380e-04, -1.3544e-04,  3.3664e-01,  2.4454e-04,  1.7852e-03,
        -3.1903e-04,  7.2360e-05,  1.1369e-01, -5.4999e-05, -3.4467e-04,
        -8.2016e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6628e-04, -2.5546e-04,  1.9602e-04, -4.2474e-04, -2.5148e-04,
        -4.5039e-04, -3.0047e-04, -2.9144e-01,  1.0607e-05, -5.3300e-05,
        -9.4436e-05, -9.5218e-05, -2.1050e-01,  2.1496e-04,  5.8675e-05,
        -2.6836e-01, -1.3088e-05,  1.4374e-04, -4.5360e-04, -7.2861e-05,
        -2.2579e-01, -9.6843e-05, -2.1558e-04,  2.4525e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8715e-03, -2.6951e-05,  1.5835e-04, -1.8525e-04, -5.0581e-02,
         2.2988e-06, -4.4119e-01,  1.7930e-04, -3.7131e-01, -3.0352e-04,
        -1.0755e-04, -1.4430e-04, -3.5264e-04,  7.1257e-05, -2.4170e-05,
        -5.7847e-05, -1.6544e-04, -4.9239e-04, -5.9344e-05,  3.4616e-04,
         3.6782e-04, -1.3152e-01,  2.0789e-04,  2.4518e-04,  2.5618e-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([ 2.6860e-03, -6.9993e-05, -8.6062e-04,  2.6342e-04,  3.7569e-01,
         2.5224e-04,  1.2409e-01,  4.4563e-04,  4.7678e-01,  4.0504e-04,
         1.1794e-04,  2.6761e-04,  5.5435e-04, -1.7547e-04,  3.5805e-04,
         1.1823e-04,  4.0544e-05, -3.6243e-03, -3.4352e-04,  5.0995e-04,
         1.1891e-02, -1.5104e-04,  3.8924e-05,  2.6283e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1056e-04, -7.6044e-05, -8.2810e-06, -3.1238e-04, -4.5867e-01,
        -1.2336e-04, -1.4432e-01, -2.1026e-04, -2.8744e-01, -3.1316e-04,
        -5.1693e-04, -2.7248e-04, -3.4295e-04, -8.5762e-04, -5.2997e-04,
        -2.8357e-05, -5.4502e-04, -9.4881e-02, -8.6745e-05, -3.2299e-04,
        -3.2770e-04, -9.0627e-03,  8.2350e-05,  2.5041e-04, -1.0030e-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,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.4842e-04, -8.5295e-02,  6.6262e-05, -4.1035e-02, -3.9876e-01,
        -6.1259e-05, -2.1689e-04, -2.7703e-04, -1.3274e-04, -3.5148e-04,
         2.7882e-04, -1.2850e-04,  2.7908e-04,  9.8977e-05, -3.3200e-01,
        -2.4381e-04, -1.9509e-04, -1.1172e-04, -1.1515e-01,  1.2933e-04,
        -2.9667e-04,  4.1163e-05, -2.3812e-04, -3.3557e-05, -3.1853e-04,
         2.6270e-04,  1.2624e-04,  1.6148e-04, -2.2861e-02, -1.6374e-04,
         1.1137e-05, -1.1821e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9882e-03, -3.1857e-01,  1.3345e-04, -1.1305e-01, -2.4293e-01,
        -4.6159e-05, -1.2900e-04, -3.4596e-04, -1.9224e-04, -1.7613e-04,
        -1.2370e-04, -3.2552e-05,  3.0604e-04, -1.8641e-05, -1.6390e-01,
         1.6452e-04, -2.3978e-04, -5.7672e-05,  1.1084e-04,  1.5761e-03,
        -2.8805e-03,  1.3389e-05,  1.8369e-04, -1.5238e-01,  2.2283e-04,
        -1.4192e-04, -8.6502e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.9100e-04, -1.9003e-01,  2.1724e-04, -3.4851e-01, -4.9867e-02,
        -1.2787e-04,  1.9747e-04, -9.6568e-05,  1.2170e-04, -6.5952e-04,
         4.6972e-04, -2.7984e-04, -3.7402e-04, -3.2869e-05, -3.7196e-01,
        -6.4238e-05, -1.6486e-04,  1.3449e-04,  5.4774e-04,  1.9754e-04,
        -1.8622e-05, -4.5246e-04,  1.1497e-05, -1.1451e-04,  4.7234e-04,
        -3.1076e-04, -3.3535e-02,  3.0491e-04,  1.8903e-05,  2.2103e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.9164e-05, -4.7786e-03, -1.3234e-04, -7.6350e-02, -1.7504e-04,
         1.0833e-04, -8.0780e-03, -5.2487e-01, -3.2960e-04,  1.4595e-04,
        -3.6318e-04, -1.8580e-04,  1.2385e-04, -2.1055e-04, -3.3136e-04,
         3.9132e-05,  3.0559e-04, -6.6542e-03,  6.5631e-05,  1.0145e-04,
         4.1441e-04, -5.0982e-02,  1.5167e-04, -1.9299e-04, -7.7551e-05,
        -2.8564e-01,  1.7727e-04,  2.0676e-05,  1.0297e-04, -1.9802e-04,
        -8.6536e-03,  4.4432e-05,  4.4171e-04, -6.3429e-05,  6.1217e-05,
        -1.0684e-04, -2.7313e-02, -4.1030e-05,  1.9459e-04, -1.5472e-05,
        -1.7170e-04,  1.1671e-03, -1.2019e-04, -1.3193e-04,  5.8333e-07,
        -8.5767e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2018e-03, -2.0361e-02,  2.3999e-05, -1.0534e-01, -3.8826e-05,
         2.5882e-05, -1.1192e-01, -2.9932e-01,  1.2325e-04, -5.5256e-05,
        -7.2384e-05, -6.6851e-05, -9.0716e-05, -2.7825e-04, -1.1888e-04,
         9.0054e-05,  1.6424e-05, -2.7493e-02, -6.7044e-05, -3.2498e-04,
         6.4087e-04, -1.8238e-01,  1.2036e-04,  1.2048e-04, -1.4749e-04,
        -4.5313e-02,  1.2551e-04, -5.2390e-05, -2.6668e-06, -1.6284e-04,
         8.3682e-04, -9.0323e-05,  4.5608e-04, -1.2875e-04, -6.1443e-02,
         1.3996e-04,  3.9369e-05,  5.0850e-05,  3.2131e-05,  5.3299e-05,
         7.0776e-05, -6.1909e-02, -7.5335e-02, -7.8573e-05, -4.4086e-05,
        -1.6859e-04, -2.6275e-05, -8.0158e-06], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.0316e-04, -1.9116e-01, -7.5890e-05, -1.9226e-02, -1.8306e-05,
        -8.3915e-05, -8.5358e-02, -5.5710e-02, -6.9834e-05,  3.2218e-05,
         6.5473e-05, -7.2199e-05, -4.4420e-05, -1.6854e-04, -2.3923e-04,
         1.2507e-04, -5.5063e-05, -5.5179e-03, -2.9098e-05,  1.1522e-04,
        -1.5511e-03, -2.5242e-01,  6.3929e-05,  7.3494e-05,  5.2742e-05,
        -3.9419e-03,  3.8246e-05,  8.0936e-05, -5.7096e-05, -5.5245e-05,
        -1.3561e-03,  6.8745e-05,  2.0830e-03, -1.3826e-01, -8.9901e-05,
        -2.3729e-01,  4.4220e-05,  2.1349e-03, -1.1254e-03, -1.9637e-05,
        -5.0341e-05, -2.0445e-04, -1.6465e-05, -1.2656e-04, -1.2471e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2914e-03, -9.9754e-03, -1.7616e-04, -1.7132e-01,  8.0745e-05,
         5.4581e-05,  4.7827e-03, -3.7994e-01, -1.3492e-04, -2.0302e-04,
         1.3111e-05,  1.7767e-04, -6.5642e-05,  2.1094e-05, -2.5606e-04,
         3.2673e-04,  1.3311e-05, -2.3418e-04, -9.9368e-05, -8.5266e-02,
         9.2647e-05, -9.6300e-05, -4.4355e-04, -8.8529e-05, -4.0320e-05,
        -8.4498e-05, -3.4014e-01, -1.9269e-04, -2.2769e-03, -1.3218e-04,
        -1.9887e-05, -1.4896e-04, -1.7596e-05, -1.8676e-04, -9.3439e-05,
         1.1705e-04,  1.3512e-06, -1.3801e-04, -5.1067e-05,  1.1991e-04,
        -9.2448e-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>)]
Iter #1600: [tensor([-1.5443e-04, -2.7350e-04, -3.0091e-05, -6.8738e-03, -2.7747e-01,
         3.0635e-04, -1.5212e-04, -4.4590e-04,  9.3267e-05, -5.8597e-01,
         1.3352e-04,  9.2574e-05, -2.6555e-04, -1.9836e-04, -1.3416e-05,
        -2.1684e-04, -4.1352e-04,  8.5088e-05, -5.6782e-05,  4.5695e-05,
        -4.9245e-03, -3.4946e-04,  2.5512e-04, -8.0174e-05, -2.9017e-06,
         6.5589e-05, -2.6604e-02, -6.0343e-05, -2.5531e-04, -3.5435e-04,
        -8.6064e-05, -3.2752e-02, -8.2371e-07, -4.3196e-04,  2.5074e-04,
        -5.9844e-05, -5.9938e-02,  1.6062e-04, -3.5858e-05,  4.2137e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5897e-03,  5.8471e-04,  3.3251e-04,  6.0232e-04, -1.1241e-03,
        -2.9477e-04, -1.1476e-03, -6.4319e-04,  1.2323e-01, -7.0196e-04,
         5.2975e-05,  4.5983e-04,  1.6417e-04, -3.2317e-04,  7.6146e-04,
        -2.8845e-04,  8.6169e-01, -1.1542e-03,  6.9564e-04,  3.3566e-05,
        -1.2318e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0890e-03,  8.2651e-05, -4.3210e-04,  7.8340e-04,  5.2852e-04,
         2.1041e-04,  9.7451e-04, -9.4833e-05,  6.3346e-01,  7.0937e-06,
        -1.6527e-04,  2.3564e-04,  6.4779e-05,  3.5992e-01,  3.8279e-04,
         2.6369e-04,  8.7013e-04, -2.1218e-04,  2.1666e-04, -1.0396e-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([ 8.5670e-03, -4.2910e-05,  2.3205e-04,  5.3269e-04,  4.6174e-04,
        -2.6038e-04,  1.1567e-03, -1.0336e-03,  9.7634e-01, -1.0007e-03,
        -7.3880e-04,  1.3753e-05,  9.5730e-04,  1.2255e-04, -7.5609e-04,
        -4.2589e-03, -1.1044e-03,  5.0487e-04,  5.8057e-04, -8.6795e-04,
         4.6499e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.3065e-04, -4.9695e-05, -2.8238e-04,  1.2085e-03, -3.4910e-04,
        -9.4807e-05,  9.4819e-04,  1.9572e-04,  6.8042e-01, -4.0340e-05,
         4.5564e-04,  4.2157e-05, -2.2321e-04,  2.8323e-04, -4.9391e-04,
         3.1370e-01, -2.2276e-04,  1.6471e-04, -1.9422e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.5296e-04,  2.5516e-04,  2.6608e-04,  1.8458e-05, -3.0745e-04,
         2.8843e-04,  1.4430e-04,  1.2073e-04,  1.9362e-01, -2.3147e-04,
        -2.0223e-04,  1.9594e-04,  4.5708e-05,  1.9790e-01,  5.6524e-04,
         8.7040e-02, -1.1322e-04, -6.4240e-04,  5.1644e-01,  3.9648e-04,
         1.0541e-04,  1.4733e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2313e-03,  4.9735e-04, -4.2130e-05, -6.9578e-04,  3.1122e-04,
         7.5153e-05, -4.2854e-04, -9.0765e-04, -6.0558e-01,  1.0381e-04,
        -6.2326e-04,  4.5271e-04,  2.2776e-04, -1.8863e-01,  3.0313e-04,
        -1.9891e-01,  3.9595e-04,  1.0087e-04,  4.8314e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5323e-03, -6.1638e-04, -3.8905e-04, -4.1248e-04,  8.7443e-04,
         1.4157e-03, -2.8652e-06, -1.0927e-04,  1.4433e-01, -1.9680e-04,
        -8.5346e-05,  9.9973e-05, -2.2940e-04,  8.4373e-01, -4.7857e-04,
         6.3191e-04, -1.7389e-04, -2.0922e-04,  3.4032e-04, -1.4610e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0973e-04, -8.1108e-05, -2.1829e-04,  6.5835e-04, -4.1844e-04,
         3.6275e-04,  6.4940e-04, -1.0125e-03,  8.7444e-01,  2.1212e-04,
        -5.7406e-04,  7.0178e-04,  1.2242e-04,  1.1924e-01,  2.0359e-04,
         2.7160e-04,  2.5850e-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([-5.7263e-04, -8.2669e-05, -2.2119e-04,  8.7294e-04,  1.0533e-03,
         4.9459e-04, -4.0835e-04, -9.0323e-04,  1.8625e-01, -1.6238e-04,
        -1.2260e-04,  2.1947e-06, -1.1646e-03,  8.0682e-01,  2.7321e-04,
        -5.5155e-04,  4.5008e-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([ 0.1864,  0.0212, -0.0394,  0.0882,  0.0583,  0.1402, -0.0328, -0.0394,
         0.0512, -0.0013,  0.0658,  0.0335, -0.0280,  0.1184,  0.0588,  0.0359,
         0.0014,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.5385e-05, -1.1542e-03,  2.7244e-04,  2.7380e-04, -2.8638e-04,
        -4.0283e-04, -4.9200e-04,  8.9462e-04,  1.4236e-01, -2.0823e-04,
        -2.7858e-04, -7.9807e-04, -8.7188e-04, -3.0404e-04, -5.8846e-03,
        -2.7488e-04,  8.4389e-01, -5.1815e-04, -1.2583e-04,  6.1538e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  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([-4.7382e-04,  8.8072e-04,  2.0245e-04, -8.7339e-04, -2.0471e-04,
        -4.9200e-04,  3.6418e-04,  3.6476e-04,  3.6150e-04, -1.2055e-03,
         9.9253e-01, -8.8590e-04,  8.7029e-04, -2.8872e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7830e-03, -1.9577e-04, -2.6092e-04, -2.6370e-04,  2.5037e-04,
         7.1506e-01,  4.5697e-04,  4.2087e-04,  6.8017e-04,  1.8007e-04,
         4.6395e-04, -3.3191e-04, -1.8387e-04, -9.7408e-03,  2.6682e-01,
         5.6994e-05,  7.1533e-04,  1.3040e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0082e-04, -5.2000e-04, -3.7756e-04, -3.1009e-04, -6.2577e-05,
         6.8008e-01, -2.3079e-04,  4.6002e-04,  7.8923e-06,  2.7193e-04,
         2.2431e-01, -3.3055e-04, -1.6775e-03, -8.7548e-05, -4.0264e-04,
        -9.2151e-04,  8.8892e-02, -3.7980e-04,  5.1817e-05, -4.2149e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7701e-05, -1.0868e-04, -1.6429e-04,  3.4768e-04, -1.7858e-04,
         7.3207e-01,  1.1219e-05,  4.3878e-05, -1.1710e-04,  4.9100e-04,
         2.5984e-01, -4.2228e-05, -1.6115e-03,  1.0057e-04,  3.1757e-04,
        -1.3489e-03, -1.2085e-03, -2.9588e-04,  7.8610e-05, -1.1788e-04,
         2.8695e-04, -1.5347e-04, -5.8067e-04, -7.1684e-05, -1.7219e-04,
        -2.0134e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2059e-03, -2.3548e-01,  2.5402e-04,  1.2014e-02, -1.2136e-04,
         6.5438e-04,  1.8970e-04, -6.8334e-01,  8.9766e-04,  3.4831e-04,
        -2.9602e-05, -9.7552e-06,  5.9176e-04,  3.9585e-04,  5.7798e-02,
         2.9485e-05, -4.0719e-04, -4.4634e-03,  4.6533e-04, -3.0571e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.4902e-03, -7.5450e-01, -1.2333e-05, -9.3851e-02, -2.2067e-04,
         3.9702e-05,  5.0159e-04, -1.1676e-01,  8.8631e-04,  1.5769e-04,
         6.1677e-04, -9.2464e-04,  4.5282e-04,  8.5783e-04,  4.5004e-04,
         3.3737e-04,  1.8073e-04,  2.2730e-04,  2.3591e-02, -6.0700e-05,
        -1.3889e-03, -4.8820e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0936e-03, -6.3357e-01,  2.1468e-04, -7.7214e-03, -1.1671e-04,
         3.2158e-04, -1.5060e-04, -2.8425e-01,  5.2695e-04, -4.6035e-04,
        -2.1761e-04,  4.7983e-04,  2.4444e-04,  1.5326e-04, -2.8388e-02,
         2.7953e-04, -2.4888e-05, -3.9888e-02, -4.1037e-04,  4.9459e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7317e-03, -3.8298e-01, -1.7614e-04,  2.7568e-04, -1.4693e-01,
         9.6677e-05,  2.2092e-03,  4.2141e-04, -1.4235e-04,  2.8752e-04,
        -1.8500e-04,  2.0813e-04,  2.1435e-04, -3.1079e-01,  2.7111e-04,
         3.9202e-04,  1.3190e-04,  3.0255e-04, -1.5029e-02,  1.3593e-04,
         1.6062e-03, -6.4029e-05, -1.3352e-01,  2.1495e-04,  4.3100e-05,
         2.4270e-04, -2.1516e-04,  1.4722e-04, -3.0152e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0802e-03, -4.2329e-01,  1.0771e-04,  6.4143e-03, -2.1400e-01,
         2.2569e-04,  5.1082e-03,  2.4846e-04,  9.4804e-06,  2.0293e-04,
        -5.3209e-05, -3.4763e-04,  3.4727e-04, -2.6561e-01,  1.3239e-04,
         5.4218e-04,  1.1290e-04,  5.3879e-05, -6.4255e-02,  1.0169e-04,
         5.3994e-03, -1.0388e-04,  1.8699e-04,  1.1574e-02,  3.5093e-04,
        -1.1320e-04,  3.1629e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8137e-03, -8.3657e-02, -2.9063e-04, -3.5400e-02, -2.1493e-01,
        -2.2132e-04,  2.2824e-03, -4.3241e-05, -2.5223e-04,  4.3138e-04,
        -8.1199e-05,  6.5214e-05,  1.1576e-04, -2.3622e-01, -1.0830e-04,
         1.7829e-04, -2.7370e-05,  1.7005e-04, -7.0575e-02,  2.3576e-04,
         4.2471e-03, -4.8242e-05, -1.9650e-01, -3.8711e-05,  1.2011e-03,
        -1.4953e-01, -1.1870e-04,  3.3750e-05,  1.8405e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.9151e-04, -8.8824e-01, -1.5688e-04,  1.2037e-03,  1.3568e-04,
         5.2549e-05,  5.6670e-03, -2.7107e-04, -1.0283e-04, -4.7137e-05,
        -4.0061e-04, -3.0321e-03,  1.6767e-04,  4.1713e-04,  1.4663e-03,
         2.2007e-04,  1.4480e-04, -3.0379e-05,  8.3851e-05,  4.8362e-05,
        -1.6759e-04,  7.0497e-05,  5.3260e-05, -9.6613e-02,  1.7213e-04,
         1.9553e-04, -4.3827e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5988e-03,  9.3055e-01,  7.0967e-04,  2.1246e-02, -3.5130e-04,
         3.0658e-05, -4.5375e-03,  5.5685e-04, -1.6122e-04,  3.5405e-04,
         1.2131e-04,  7.7462e-03,  2.8607e-04, -2.8486e-02, -2.6626e-04,
        -2.3350e-05,  8.4141e-05, -3.3610e-04, -6.3003e-04, -7.2327e-04,
         1.8183e-04,  2.0361e-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>)]
Iter #1700: [tensor([ 1.0741e-03, -5.3860e-01, -2.5072e-04, -5.1233e-04, -1.6779e-04,
        -5.2508e-04, -5.8709e-04,  1.8287e-04, -2.3930e-04,  2.5598e-02,
         3.4938e-04,  3.2489e-04, -6.3046e-04,  2.5799e-04, -4.3043e-01,
        -1.3424e-04, -2.9136e-05,  1.0238e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0962e-03, -6.7713e-01, -5.2373e-04, -1.3045e-04, -2.3513e-04,
        -1.6339e-05, -3.8988e-04,  3.8005e-04,  1.3940e-04,  4.0745e-02,
         1.1137e-04, -9.9178e-05, -9.0823e-05, -8.8744e-05, -2.7657e-01,
        -3.9736e-05, -1.5232e-03, -8.1698e-05, -6.7949e-05, -3.5125e-04,
         1.9236e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7396e-04,  2.8395e-06, -1.3970e-04,  3.7308e-05, -2.2320e-04,
        -6.0759e-01, -1.6897e-04, -2.2139e-02, -5.5733e-05,  3.2832e-03,
         1.5256e-04,  1.6530e-04, -5.7211e-06,  9.5583e-05, -2.2687e-01,
        -3.9299e-04,  2.5527e-05,  1.6006e-04, -9.3964e-05,  1.4386e-03,
        -7.4461e-05,  1.0152e-05, -1.2502e-01, -9.2159e-05, -1.9424e-04,
         2.4326e-03, -6.9629e-03, -3.2604e-04, -2.2573e-04, -5.4521e-04,
        -1.3544e-04, -3.7817e-04, -2.4178e-04,  1.2623e-05, -1.3754e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1623e-03,  2.3022e-04,  4.5375e-05,  1.6123e-04,  8.5834e-04,
        -7.0966e-01, -2.7346e-05,  2.3424e-02, -1.9478e-04,  2.3770e-02,
        -2.5763e-04,  3.4920e-04, -1.6890e-04, -1.0455e-04, -2.1706e-01,
        -2.2311e-04, -3.5637e-04,  8.0633e-05, -6.1290e-04,  8.2791e-03,
        -7.5894e-04,  3.5050e-04, -8.6808e-05,  5.6208e-05, -5.3514e-04,
         1.2810e-04, -4.6965e-04, -1.4113e-05,  8.1112e-03,  5.8993e-05,
        -2.9899e-04, -1.1213e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.0921e-04, -3.5616e-04, -4.3812e-04,  4.6610e-04, -1.4627e-04,
        -6.3092e-01, -2.2997e-04,  8.0317e-03, -2.2915e-05, -4.4446e-03,
         9.6076e-05,  2.0952e-04, -5.1000e-05, -1.7592e-04, -2.7672e-01,
        -1.6853e-04,  1.1175e-04, -2.4364e-05, -7.1257e-05,  2.2728e-04,
         2.9910e-05,  5.3842e-03, -2.9804e-04, -1.2973e-04, -4.2103e-05,
        -4.7709e-04,  3.5110e-03, -1.8850e-04,  5.3556e-05, -2.9821e-04,
        -6.5674e-02, -2.5752e-04, -2.1997e-04, -8.5201e-06,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1584e-03, -9.6600e-02, -3.6214e-04, -4.6527e-04, -1.6967e-04,
         2.3565e-04, -1.5143e-01, -5.6659e-05,  6.5389e-04,  6.9736e-05,
        -1.1101e-04, -7.2046e-01, -7.7257e-04, -1.2792e-03,  2.7338e-04,
         2.4710e-04,  2.4595e-02,  1.4570e-04, -6.9472e-04, -9.3165e-05,
        -1.3249e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7210e-03, -4.9593e-01,  1.9651e-04, -1.4486e-04, -3.2838e-04,
         1.4784e-04, -3.9468e-02, -4.2289e-05, -3.7256e-04, -2.8157e-04,
        -2.7199e-05, -4.5855e-01, -2.9652e-04, -3.3789e-04, -4.2320e-04,
         1.2950e-04, -4.9534e-04, -6.1786e-04,  3.7901e-04,  1.1051e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4400e-03, -5.7814e-01, -3.1605e-04, -5.8266e-04,  3.9526e-05,
        -1.9634e-04,  2.9062e-02, -2.9098e-04,  6.9297e-04, -7.1119e-04,
        -1.5326e-04, -3.7776e-01, -8.2772e-05, -7.7891e-04, -2.5247e-04,
         2.2596e-04,  8.6886e-03, -1.0356e-04,  1.6151e-04,  3.1516e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2719e-03, -1.2018e-04,  5.2747e-05, -4.8987e-04, -2.2749e-04,
         2.5489e-05, -1.6706e-04, -9.5549e-05,  4.5656e-04, -2.3732e-05,
        -2.2554e-04,  2.3440e-04,  9.5023e-01, -1.2178e-04, -3.6284e-04,
        -1.3015e-04, -2.5340e-04,  2.4129e-03, -6.3232e-04, -2.9316e-04,
         3.4370e-02, -1.6165e-04, -2.7981e-05, -6.7105e-05, -6.0217e-07,
         5.6719e-04,  4.8590e-03, -2.2463e-04, -3.1654e-04,  1.0956e-04,
         2.5015e-04, -5.1609e-05,  1.0739e-04, -5.5615e-05,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.2541e-04,  2.5202e-04, -1.3990e-04,  6.1801e-04, -3.3783e-05,
         1.7199e-04,  7.6626e-07,  1.4877e-04, -1.6165e-04, -2.8019e-05,
         8.4424e-05,  4.4668e-01, -6.6216e-04,  8.9609e-05, -2.0686e-04,
         8.6281e-05, -3.7575e-04, -1.9049e-04, -1.0283e-04,  2.6294e-04,
        -5.0719e-05,  3.7778e-01, -9.4538e-05, -1.4191e-03,  1.6918e-01,
         4.7129e-05, -1.2052e-05,  2.0213e-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([ 7.8420e-03,  1.0399e-03, -6.4680e-04,  1.5402e-03, -1.4848e-04,
        -1.3499e-04, -7.0041e-05,  9.6924e-06, -1.2658e-03, -6.6714e-04,
        -8.0344e-04, -9.8310e-01, -5.5174e-04, -6.4734e-05,  3.2357e-04,
        -9.3001e-04, -6.2638e-04, -2.3515e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4432e-03,  2.8798e-04, -4.9574e-04, -2.4107e-04,  4.0688e-04,
         5.9624e-04,  1.0784e-03, -9.5606e-01,  2.4600e-04,  1.9314e-05,
         5.1488e-04, -7.1644e-05, -5.7237e-04,  3.0402e-02,  6.9071e-05,
         2.1149e-04,  9.1313e-04,  5.0263e-03,  9.9835e-05, -1.5033e-04,
        -9.6203e-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>)]
Iter #1750: [tensor([-6.0177e-04, -2.9608e-01,  6.7921e-05,  5.2137e-04, -2.9874e-01,
        -8.6941e-05,  3.0471e-04, -6.3893e-05, -5.3728e-05,  7.5917e-05,
         1.3546e-03, -7.0350e-05, -3.5910e-01,  3.6740e-05, -3.9268e-04,
        -1.5705e-04,  8.4573e-03,  3.7952e-04, -3.0086e-04, -2.5174e-04,
        -1.7390e-04,  3.2337e-02, -6.6320e-05, -3.0688e-04,  1.9040e-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([-2.6793e-03, -3.1140e-01, -2.2135e-05,  3.3613e-04, -2.9579e-01,
        -3.6984e-04,  2.7813e-04,  1.7254e-04, -3.0901e-04, -5.6337e-06,
         7.3683e-05, -1.4440e-04, -3.5555e-01, -1.6793e-04, -3.3540e-04,
        -4.2230e-04,  3.0934e-02,  9.2137e-05,  4.7450e-05, -1.1440e-04,
        -2.9196e-04,  8.7193e-05,  6.1850e-05,  3.2129e-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([-5.5062e-03,  3.6262e-04,  1.2393e-04, -2.9482e-04, -2.1966e-02,
         6.5834e-04, -1.0583e-01, -3.0819e-01,  3.8482e-04,  3.4133e-04,
        -9.8242e-02, -9.0825e-02,  5.1900e-04, -3.4807e-04, -1.8077e-04,
         2.3258e-04,  1.8917e-04, -3.7831e-05,  2.9258e-04,  3.5333e-02,
         3.7131e-02,  2.5349e-04,  1.0647e-01, -1.4855e-04,  1.7163e-04,
         4.2975e-02, -1.4186e-01,  3.8218e-04,  3.6417e-04, -1.0821e-04,
        -2.7765e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3107e-03, -1.0547e-04, -1.0365e-04,  9.3351e-05,  1.9611e-02,
         1.7917e-04,  1.2625e-02, -2.5607e-01,  4.6125e-04,  2.0894e-04,
         3.1654e-02, -5.8862e-01,  6.2540e-05, -2.4083e-04, -3.6454e-04,
        -3.0073e-04,  3.5413e-04,  4.7970e-03, -1.2714e-03,  3.0121e-04,
         8.0197e-02,  4.5680e-04, -1.3214e-04, -2.0504e-04,  2.6117e-04,
        -1.6860e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1454e-04, -3.5746e-04,  4.6528e-04, -4.9216e-06,  1.1283e-02,
         6.4427e-05,  4.4907e-02, -3.7379e-01,  1.6662e-04, -4.6152e-05,
         7.9706e-03, -2.1599e-01,  4.8337e-04,  2.0192e-04, -5.9552e-05,
        -9.4542e-06, -2.3491e-05,  1.9651e-03,  1.0701e-04,  3.6590e-02,
        -3.0481e-01,  1.3168e-04,  4.0545e-05,  1.2803e-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([ 7.2902e-03, -3.0471e-01,  2.1363e-04,  3.6751e-05,  3.0664e-02,
         4.1056e-05,  3.9072e-04,  1.7553e-04,  2.6234e-02, -2.5177e-04,
         8.4854e-05, -3.2398e-04, -3.6639e-01,  4.8220e-04,  1.2168e-04,
        -2.9996e-04, -3.1415e-02,  2.6128e-04,  2.2896e-02, -6.3751e-04,
         4.5591e-04, -1.5924e-04, -2.0568e-01,  3.1771e-04,  3.1657e-04,
        -1.5082e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2776e-04, -4.3950e-01, -1.1272e-05, -2.6996e-05, -9.5628e-03,
        -2.7917e-04,  4.0715e-04, -1.0020e-04, -1.2427e-02, -6.8798e-04,
         4.8697e-04, -3.1844e-04, -4.7367e-01, -1.3869e-04, -5.4748e-04,
        -3.2651e-04, -5.8091e-02,  1.6991e-04,  1.2179e-03,  7.4229e-05,
        -7.4488e-04, -5.7130e-04, -2.1970e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.8878e-04, -5.8267e-01,  1.8225e-04,  3.2372e-04,  2.1747e-02,
        -3.1837e-04,  9.6256e-05, -7.5434e-05,  1.3857e-02, -2.9703e-04,
         2.3858e-04, -4.2262e-04, -2.1732e-01,  2.6966e-04, -2.2904e-04,
        -1.3096e-04,  3.0052e-02, -5.5922e-04,  6.9113e-03, -3.9587e-04,
         2.7328e-03,  1.1970e-01, -1.9521e-04,  4.8358e-04, -2.0154e-04,
        -8.7482e-06,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5205e-04, -2.8376e-04, -1.0539e-02, -7.6060e-05,  3.2733e-04,
        -1.6019e-04, -5.4671e-02,  4.9496e-04, -1.6454e-04, -1.5576e-04,
        -2.8966e-04, -5.1535e-01,  6.8094e-06, -1.3681e-04,  2.4407e-04,
         1.3357e-05,  7.9603e-03,  4.3799e-05,  2.5070e-05, -4.3861e-05,
        -4.0856e-01, -7.8098e-05, -3.8376e-05,  8.5334e-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([ 1.1311e-03, -6.5737e-04, -7.3959e-02, -3.5719e-05,  4.1640e-04,
        -5.0314e-05, -1.2943e-01,  9.9932e-04,  2.6325e-04,  4.9936e-04,
         4.4239e-04, -7.6349e-01, -8.6606e-05, -6.2156e-05, -1.8831e-04,
        -8.5991e-04,  1.6337e-03, -6.2849e-04, -2.3265e-02,  7.2599e-04,
         3.9021e-04, -1.1202e-04,  6.7494e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0389e-03, -3.5118e-04,  8.1080e-03, -1.0235e-04,  1.5329e-04,
        -5.1169e-04,  1.9698e-02, -1.2721e-04, -9.0977e-05, -2.6574e-05,
        -1.4207e-04, -5.5866e-01, -3.4977e-04,  2.9631e-05, -8.6542e-05,
         4.9113e-04, -4.4246e-05,  1.8540e-02, -6.8341e-04, -4.9423e-05,
        -3.9005e-01,  6.3035e-05,  4.0512e-04, -1.8809e-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([ 7.7780e-04,  1.2424e-04,  6.9341e-05, -5.2083e-04,  3.4361e-02,
         9.8900e-05,  5.4487e-05, -1.8394e-04, -1.5523e-04, -1.8213e-04,
        -2.7226e-01, -3.2449e-04, -6.5535e-04, -3.8411e-01, -3.0160e-05,
         2.2531e-05, -2.9407e-04,  5.6587e-05, -2.6206e-01, -2.8303e-04,
         4.3214e-02,  1.4484e-04, -1.4115e-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 #1800: [tensor([-8.8385e-04, -4.9005e-01,  1.2734e-04,  3.5144e-05,  2.5588e-04,
         1.7637e-04,  8.5704e-05,  3.3999e-04,  1.7024e-04,  3.9238e-04,
         2.1701e-02,  8.8940e-05,  1.1421e-04,  8.2343e-05,  3.9964e-04,
         1.0486e-03,  2.3399e-04,  2.3020e-02, -1.3840e-01, -1.6265e-04,
        -1.4455e-04,  1.3808e-04, -3.2081e-01, -2.2816e-04, -2.3278e-05,
         1.2072e-04,  1.1959e-04,  4.4400e-05,  4.9637e-04,  3.5407e-05,
         8.2912e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7643e-04, -3.3668e-01,  4.7022e-05,  3.7322e-04, -3.9080e-05,
         5.5046e-04,  4.3607e-05,  2.4079e-04,  4.3720e-04,  6.8039e-05,
         9.2263e-03, -1.7211e-04,  2.6412e-04, -9.5291e-07,  5.7841e-04,
         1.0145e-03,  3.1286e-04,  8.5181e-03, -1.3854e-01,  2.1068e-04,
        -2.0585e-04,  1.0392e-05, -5.0028e-01, -2.7061e-04,  3.4918e-04,
        -9.3250e-05,  2.1155e-04, -2.2397e-04, -2.5832e-04,  2.4409e-04,
         2.4974e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3416e-04, -4.4715e-01,  2.7350e-04,  1.3020e-04,  1.8022e-04,
         5.0878e-05,  2.7922e-04,  6.5643e-05,  2.1028e-02,  2.5043e-04,
        -2.1499e-04,  1.3673e-04,  6.3092e-05, -3.0464e-01,  1.8826e-06,
        -9.3233e-05, -1.1062e-04, -1.8530e-01, -1.0490e-04, -5.6841e-04,
         1.4434e-04, -3.2441e-04, -1.4477e-04, -3.5452e-02, -8.6484e-05,
         1.2736e-03, -3.0768e-04,  1.0204e-04,  4.1630e-04,  1.6644e-04,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.4130e-04, -4.4260e-01,  6.4564e-04, -1.1873e-04, -9.6301e-05,
         9.4965e-05,  1.4742e-05,  1.2242e-05, -2.1953e-01,  2.7177e-04,
         1.9278e-04,  5.9472e-05,  5.7713e-04, -2.1784e-01,  6.1408e-05,
         2.1680e-04, -1.9538e-04, -1.1412e-01,  4.6648e-04, -1.3229e-03,
         3.6062e-04, -2.6078e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1346e-03, -2.3855e-01,  2.9210e-05, -1.7138e-04, -1.9993e-04,
         2.9963e-04,  4.4063e-05,  2.1826e-04, -3.2342e-01,  2.8919e-04,
         2.0504e-04, -1.4989e-04,  1.8594e-04, -2.6813e-01,  4.4341e-04,
         5.1605e-04,  4.0041e-04, -1.4566e-01, -2.3972e-04,  1.8355e-02,
        -1.2778e-03, -8.2911e-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([ 1.3218e-04,  3.6996e-04, -1.2312e-04,  3.1941e-05,  4.8457e-03,
         9.7697e-04,  4.6951e-04, -2.8971e-04,  2.5575e-04, -3.6257e-01,
         1.2283e-04,  3.3445e-04,  2.4390e-04,  2.0162e-04,  3.0157e-05,
         1.1958e-02, -1.7110e-04, -1.3110e-04, -4.1825e-01, -3.0706e-04,
        -2.2952e-04, -5.4561e-05,  8.2575e-03, -1.8917e-01, -1.1985e-04,
         2.0950e-04, -4.2555e-05, -1.0518e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8963e-03, -1.8606e-04,  1.3573e-04, -1.1168e-04, -1.0716e-01,
         5.6924e-03, -2.9294e-04,  4.6957e-04, -7.8869e-04,  8.2331e-01,
        -2.8133e-04, -2.0319e-04, -2.2578e-04,  6.7893e-04, -2.0368e-04,
        -5.6527e-02,  4.5335e-04, -2.2653e-05,  3.5961e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2017e-04, -7.4733e-05, -7.2034e-05, -2.0820e-04, -1.3410e-03,
         3.0450e-03, -1.0502e-04,  4.2048e-04, -8.3260e-04,  9.9151e-01,
        -6.1173e-05, -2.9387e-04,  1.0264e-05, -3.8219e-04, -5.5207e-04,
         2.3287e-04, -2.6477e-04, -3.7406e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1552e-04,  9.5404e-04,  9.4645e-04, -4.1091e-04, -1.3293e-04,
        -9.0023e-04,  6.6101e-04, -2.4819e-01,  4.6583e-04,  5.8048e-04,
        -3.7826e-04,  3.1000e-04,  2.8134e-04, -4.9391e-05,  3.0013e-02,
         6.6298e-04,  5.6164e-04, -7.1407e-01, -7.2967e-05,  4.1517e-05,
        -2.0247e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4692e-03,  5.1429e-04,  6.2165e-04, -3.8217e-05, -1.0668e-04,
        -7.5036e-04,  2.0772e-04, -6.1180e-01, -1.1804e-04,  1.7948e-04,
         4.7892e-04,  2.6198e-04,  4.0052e-04, -1.1119e-04,  1.4062e-02,
        -1.3553e-04,  2.8086e-02, -3.3927e-01,  3.1082e-06,  2.8271e-04,
        -1.0818e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0552e-03,  7.8413e-04,  4.9010e-04,  6.4049e-04, -3.7412e-04,
        -8.1927e-04,  7.1769e-04, -3.7153e-01,  9.0291e-04,  6.5011e-04,
        -2.6789e-04, -2.7711e-04,  1.0987e-03,  7.4770e-04,  1.2188e-01,
         8.1366e-04,  6.6803e-04, -4.5836e-01,  2.9565e-04, -4.0777e-04,
        -3.4859e-02,  4.7716e-04,  2.4343e-04, -4.0324e-04,  2.3572e-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([-1.6563e-04, -6.2650e-01,  1.4626e-05,  2.7190e-04, -4.6389e-04,
         5.7376e-04,  3.6987e-04,  1.0147e-03, -3.6947e-01, -1.4842e-04,
        -7.3304e-04,  1.5952e-04, -1.1474e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         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 #1850: [tensor([-1.2585e-04, -7.4365e-01,  1.4058e-04,  3.0682e-05,  1.5085e-04,
         5.0676e-03,  3.9664e-04,  1.7286e-05,  4.1396e-05,  1.3519e-04,
         1.7292e-03,  4.8523e-04,  1.9004e-04,  8.3211e-05,  2.9758e-04,
         1.0938e-04, -2.1895e-05,  3.7136e-04, -2.4002e-01,  1.2018e-04,
         7.7138e-05,  6.1870e-03, -2.2087e-05,  1.9215e-04, -1.1265e-04,
        -1.9509e-04, -3.5602e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.9148e-04,  4.3897e-01, -4.2973e-04, -1.5065e-04,  4.1031e-05,
         6.3419e-03,  6.8746e-04,  1.2035e-04, -9.3553e-04, -8.8191e-05,
         1.0212e-04,  2.6642e-05, -8.8952e-04,  5.1253e-01, -1.2837e-04,
        -2.0996e-05, -2.3558e-05,  8.0698e-05, -3.6139e-02, -5.8024e-04,
         7.9058e-05, -4.8223e-04, -1.3478e-04, -4.1801e-05,  1.1166e-04,
         1.6563e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5051e-03,  3.5629e-02,  1.6089e-04,  7.7053e-04, -3.1594e-04,
         8.6113e-04, -9.8926e-05, -3.8317e-05,  1.3402e-03,  9.5758e-05,
        -3.3387e-04,  9.5079e-05, -7.3206e-05,  2.1384e-02,  2.4531e-04,
        -4.5755e-06,  7.7179e-05,  3.4044e-04, -5.8671e-01,  1.5504e-04,
        -4.7736e-05, -3.4919e-01, -2.6613e-04, -9.6247e-05,  1.7060e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2010e-05, -6.2080e-03,  1.5447e-04,  6.5292e-05, -4.7685e-04,
        -1.5026e-03, -4.6003e-04,  1.4240e-04,  4.8451e-04,  2.4969e-04,
         3.6170e-04,  3.5414e-04, -3.2720e-04, -7.2602e-01,  2.5798e-04,
        -1.8343e-04,  3.3216e-04,  7.5832e-04, -2.6053e-01, -1.8798e-04,
        -6.7241e-05, -1.4163e-04,  3.9154e-04,  1.3054e-04,  1.8031e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9429e-04, -6.1555e-01,  1.3026e-04, -2.4368e-04, -2.8639e-01,
        -8.8845e-05,  1.2813e-04,  1.9812e-04,  4.4939e-04,  5.7223e-04,
         4.5569e-04,  3.3958e-02,  1.8336e-04, -4.8156e-05,  1.7318e-04,
         8.6843e-06, -3.5759e-02,  7.8315e-05, -1.0463e-04,  4.4047e-04,
        -3.7238e-04,  2.4327e-02, -7.7524e-05,  4.9053e-06,  6.8778e-05,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9858e-05, -4.6769e-01,  2.0200e-04, -3.3185e-04, -1.7398e-01,
        -2.4872e-04,  1.4190e-04,  9.6845e-06,  3.5543e-04,  9.6287e-05,
         3.2266e-04, -9.1198e-02,  1.3086e-04, -2.0387e-04,  8.0519e-05,
         1.8510e-04, -1.2419e-01,  1.3226e-04,  9.5645e-05,  5.0101e-04,
         1.5184e-02, -1.2401e-01, -1.5219e-04, -3.5755e-04, -1.9121e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.4571e-04, -3.2174e-01, -3.2559e-04, -2.5269e-05, -1.0616e-01,
        -1.1206e-04,  3.8949e-04,  2.4675e-05,  4.1832e-04,  3.2719e-04,
         8.9497e-06, -2.4870e-01, -3.5891e-05, -1.3420e-04,  1.5749e-05,
        -1.4553e-04, -3.1882e-01,  7.6472e-05, -2.6657e-04, -1.6808e-04,
        -8.0896e-05,  6.4022e-04, -7.4388e-05,  2.6587e-04, -1.9867e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.7123e-04, -5.0286e-01,  3.0329e-04,  2.5206e-04,  4.3847e-05,
         3.0451e-04, -3.4451e-02,  3.5149e-04, -2.8609e-04, -1.6275e-04,
        -4.9269e-05,  1.1387e-05, -1.7679e-01,  6.0813e-05,  5.4845e-04,
         5.9150e-05, -2.8231e-01,  2.3646e-04, -3.5168e-04, -3.9787e-05,
        -5.4473e-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([ 1.6427e-03,  5.9330e-01, -1.6619e-04, -1.7335e-04,  1.1164e-03,
         5.5521e-05,  4.0168e-01, -1.4485e-04,  3.8943e-05,  5.0073e-04,
        -2.5204e-04,  1.7252e-04,  9.5226e-05,  7.4953e-06, -2.4695e-04,
         4.5393e-05, -2.9713e-04, -6.9137e-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([ 2.9835e-04,  2.7351e-01,  5.0099e-05, -5.0214e-04, -2.4156e-04,
        -1.6087e-04,  3.1661e-01, -4.1616e-04,  4.1104e-04,  4.4991e-04,
        -1.2206e-04, -1.6007e-04,  2.0816e-01,  2.9122e-04, -4.2695e-04,
        -8.2553e-05,  2.7680e-04,  1.9587e-01,  5.7220e-04, -5.1394e-04,
        -9.4808e-05,  5.6355e-04,  1.2384e-04,  9.4227e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.9686e-04,  8.1449e-01,  2.7096e-05, -2.0434e-04,  4.2582e-04,
        -6.8351e-04,  1.8035e-01,  6.2769e-04, -1.6419e-04, -1.8434e-04,
        -1.6665e-05,  3.0303e-05, -9.0313e-04, -6.4441e-04, -4.5836e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9621e-03, -4.4981e-01,  1.3952e-04, -1.0327e-04, -2.7973e-05,
         4.0120e-04,  1.6157e-01,  8.8741e-05, -2.3551e-04,  5.0621e-04,
        -3.8340e-01,  9.6105e-05,  5.4236e-04,  1.1060e-03,  4.7108e-06,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1900: [tensor([-1.7239e-03,  2.1170e-04,  1.0945e-04, -2.0528e-04, -4.2651e-04,
         4.0989e-04, -2.6181e-04,  4.1407e-01,  1.0756e-04, -1.7831e-04,
         8.2099e-05,  4.2799e-01, -5.3054e-04, -5.7802e-04,  1.4499e-04,
         1.5203e-01,  2.4260e-04,  2.2665e-04, -2.7329e-04,  1.9848e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2075e-05, -4.6574e-01, -4.2172e-05, -3.6544e-04, -5.0733e-05,
         8.5738e-03, -3.3419e-04, -2.1602e-04,  2.2422e-02,  3.7148e-04,
        -2.3938e-04, -2.0826e-04, -2.8567e-01,  2.1204e-04, -2.2040e-04,
        -2.1441e-01,  2.3106e-04, -1.8472e-04,  4.2527e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6552e-03, -2.1910e-01,  3.8942e-05,  2.9669e-05, -1.4599e-04,
         1.3906e-03,  1.1648e-04, -2.7275e-04, -3.3582e-01,  1.2699e-04,
        -4.4198e-05,  1.2029e-04, -1.9331e-01,  8.3081e-05, -4.5102e-04,
        -2.4699e-01, -1.8006e-04, -7.8233e-05,  4.1834e-06,  4.1332e-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([-2.9964e-04,  3.8336e-01, -6.7608e-05,  4.7111e-05, -9.7765e-05,
         1.3339e-03, -1.1354e-04, -1.2788e-04,  2.6118e-01, -1.5260e-05,
        -1.4464e-04, -2.9548e-04,  9.8504e-02, -4.2384e-05,  3.2793e-04,
         9.3929e-02,  6.0591e-05,  1.9153e-04,  1.5947e-01, -5.8184e-05,
        -2.8858e-04,  4.3955e-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([ 3.3474e-03, -9.4454e-05, -4.9708e-04,  1.2486e-04,  6.5185e-01,
         5.8112e-05, -2.0314e-04, -1.8230e-04,  1.7255e-04, -3.8856e-04,
         6.4417e-04,  3.4005e-01,  2.1671e-04,  1.1043e-03, -9.9533e-05,
        -3.1361e-04,  6.6185e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2431e-04,  5.1659e-05, -3.7877e-04, -7.5449e-04,  4.1179e-01,
        -5.3932e-05, -1.6332e-04,  2.0784e-04, -1.7816e-04, -1.5911e-04,
         3.7322e-05, -5.9465e-02,  1.2468e-03, -2.9607e-04,  4.9429e-04,
        -1.3974e-04,  5.2346e-01, -8.3890e-04, -2.2428e-05, -3.3820e-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([ 2.1642e-03, -2.1617e-04, -4.9730e-05,  6.6640e-05, -3.6568e-01,
        -4.2253e-04, -1.4568e-04, -2.5318e-04, -4.5523e-05, -4.1087e-01,
         5.3171e-04, -1.2674e-04,  2.9116e-04,  2.0623e-04, -2.1851e-01,
        -2.5936e-04, -3.7181e-05,  7.9514e-05, -4.3914e-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([-5.5140e-03, -3.4255e-04,  7.4164e-05, -2.5508e-04, -7.9354e-03,
         5.9432e-04, -2.1424e-04,  6.7798e-01, -4.3256e-04,  3.6838e-05,
        -2.0655e-04,  3.2355e-04,  9.8008e-02, -4.7394e-04, -1.2255e-01,
         2.7014e-04,  1.0762e-04, -8.3148e-02, -2.1183e-04, -1.7319e-04,
        -6.6564e-04,  4.8152e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5700e-04,  3.9545e-04,  3.6929e-04,  4.1384e-04,  3.6002e-03,
        -1.0031e-04, -2.1617e-04, -3.5231e-01, -7.3783e-05, -4.8046e-06,
         1.6826e-04, -1.2171e-04, -3.7296e-01, -8.8100e-05,  2.8544e-04,
         7.8382e-04, -1.6394e-04, -2.6663e-01, -3.4889e-04,  3.0826e-04,
         2.6645e-04, -1.2712e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.4000e-04, -9.8427e-05, -7.0681e-04, -6.5588e-04, -1.5970e-03,
         5.4591e-04, -7.9108e-04,  4.3483e-01, -1.8877e-04,  1.2562e-04,
        -1.0778e-04,  2.4484e-04,  4.6621e-01, -4.8912e-04, -2.4420e-04,
        -7.8854e-04,  9.1220e-02, -2.1355e-04, -2.1660e-04, -1.3685e-04,
         5.1343e-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([-4.6605e-04,  6.0470e-01,  1.2680e-04, -1.5394e-04,  3.4214e-03,
         1.2508e-04,  3.4713e-04,  4.0616e-04,  1.0729e-04, -7.8971e-05,
        -3.2026e-05, -3.6908e-05,  5.2410e-06, -1.0052e-04, -1.3277e-04,
         3.0806e-02,  9.6813e-05, -2.1776e-04,  6.3229e-04, -1.4221e-05,
         3.5744e-01,  3.0908e-04, -2.4091e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9994e-04,  4.3846e-01, -1.6207e-04, -1.3794e-04,  4.3135e-01,
         7.0979e-05, -4.1617e-05,  1.4551e-04,  3.3101e-04, -8.2135e-04,
        -6.3680e-06,  1.9649e-04,  1.7180e-04,  4.9084e-05,  1.0294e-04,
        -8.1711e-03, -1.9571e-04, -5.4149e-05,  1.4071e-04,  2.4911e-04,
         1.1742e-01, -7.8665e-05, -2.6343e-04, -8.8538e-05, -8.1336e-05,
        -7.7663e-05, -4.1852e-04,  1.4988e-04,  7.2919e-05,  7.0459e-05,
         1.1037e-04,  1.8855e-05], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1950: [tensor([-3.4826e-04, -2.6874e-01, -2.2768e-04,  9.0982e-05, -1.6074e-04,
         7.7864e-05, -2.0473e-04, -1.4851e-04, -6.3507e-02, -1.9205e-04,
        -2.6055e-05,  1.8786e-04, -1.2277e-04, -2.6236e-01, -3.9061e-04,
         3.3714e-04, -2.8922e-04,  5.4646e-05, -4.0220e-01, -3.7184e-05,
         2.0704e-04,  8.5150e-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([ 1.0567e-03, -2.3721e-01,  2.2181e-04,  5.8246e-05, -2.6110e-01,
         3.2137e-04,  6.5494e-05, -1.4939e-05,  1.3428e-05, -5.2510e-06,
         6.2293e-04,  1.3181e-04,  9.7360e-05,  3.3328e-05, -9.1584e-06,
        -1.9231e-01,  3.7050e-04,  5.9256e-04,  1.1487e-04,  8.0455e-05,
         2.7658e-04,  1.4113e-02, -2.1546e-04,  2.0145e-04, -2.8809e-01,
         3.1300e-04,  1.7902e-03, -9.6412e-05,  2.5661e-04,  8.1859e-05,
         1.3921e-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([ 2.4239e-03,  5.0821e-01, -1.9938e-04, -8.6954e-04,  4.5466e-01,
        -5.6375e-04,  5.7935e-05, -2.0138e-04, -5.0700e-05,  2.6460e-04,
        -1.6149e-03,  4.0787e-04,  1.9269e-04, -2.8602e-05,  1.3853e-04,
        -1.3348e-04, -2.2801e-04,  1.3003e-05, -2.9244e-02, -1.0078e-04,
        -1.5781e-04,  2.3443e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7459e-04, -5.6048e-01,  5.0497e-05,  5.5270e-04,  1.3122e-02,
         3.2650e-04, -3.2312e-05,  5.9457e-05,  7.1090e-06, -4.9387e-05,
         9.4159e-04,  7.4462e-05,  1.6932e-04,  1.7240e-05,  1.6149e-05,
        -1.6647e-01,  1.9163e-04,  2.7110e-04,  2.7386e-04,  4.0603e-03,
         1.4228e-05,  1.7593e-04, -2.4842e-01,  4.4855e-05,  2.5676e-03,
        -3.8952e-05,  2.3840e-04, -3.5412e-06,  1.6959e-04,  4.7824e-05,
         2.8419e-05,  2.8581e-04, -6.0010e-05,  1.1035e-04, -1.6817e-04,
         1.1284e-04, -7.7755e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5205e-04, -6.2400e-01,  1.8658e-04,  1.7795e-04,  3.5863e-04,
         9.2411e-04, -9.2300e-02, -3.4714e-04,  3.4031e-04,  2.6016e-04,
        -1.3095e-04,  3.1228e-04,  1.0499e-04, -6.7970e-05,  3.9459e-04,
         8.0779e-05,  3.4125e-05, -2.7805e-01,  1.8524e-05,  5.3843e-06,
         6.6504e-04, -7.4207e-05,  2.7032e-04,  4.1774e-04,  2.6153e-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([-1.1585e-03, -5.7478e-01,  2.2219e-04, -1.0686e-05, -3.5374e-04,
         9.2265e-03, -1.5496e-01, -1.3692e-04,  7.1071e-05, -9.9894e-05,
        -2.6550e-05, -1.1623e-05,  4.6406e-05, -2.3987e-04,  9.9115e-05,
        -9.3937e-05, -6.6384e-04, -2.5658e-01, -5.3799e-07, -2.3339e-04,
         1.7063e-04,  1.2413e-04,  2.3167e-04, -4.6166e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9622e-03,  3.7686e-01,  8.8393e-05,  2.9455e-04, -2.1161e-04,
        -3.1165e-03,  4.3507e-01, -7.1361e-05,  1.9123e-04,  6.7526e-05,
        -2.5859e-04, -1.3792e-04,  6.1032e-05,  2.9444e-04,  1.0776e-04,
         3.2872e-04,  5.0439e-05,  1.7989e-01,  1.3498e-04, -6.0855e-05,
        -5.3864e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.5921e-04,  4.1235e-02,  1.0185e-04, -1.9661e-04, -4.6378e-04,
         9.4167e-03, -7.7345e-05, -3.7543e-04,  5.2365e-04, -4.8455e-04,
        -1.4384e-04,  3.2275e-05, -3.4206e-04,  1.0930e-04,  3.6289e-07,
        -1.5206e-05, -6.9450e-03, -9.3738e-01, -5.1092e-05,  5.8501e-04,
         5.0267e-04,  1.5465e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.8465e-05,  6.5910e-01, -9.9605e-05,  2.1567e-04, -4.7339e-05,
        -2.2538e-03, -9.9761e-06,  1.3110e-04, -3.6747e-04,  1.5410e-04,
         3.3344e-04,  1.8509e-04,  1.2287e-04, -3.4099e-04, -1.5551e-04,
         1.4525e-04, -3.7101e-04,  3.8322e-04, -4.3497e-03,  3.3054e-01,
         2.1829e-04,  1.8703e-04,  2.2438e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.0134e-04, -7.4761e-01,  4.3851e-05,  1.2767e-04,  2.7444e-04,
         5.7442e-03, -7.7977e-05, -3.4316e-04,  3.5570e-04, -4.1195e-05,
        -6.8017e-05, -1.8112e-04, -1.8134e-04,  2.5825e-04,  1.5605e-05,
        -1.7252e-04, -2.2315e-04, -2.4291e-01,  3.0750e-04,  1.3988e-04,
        -1.3902e-04,  8.4197e-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([ 2.5187e-04,  4.6186e-02,  3.4047e-05,  1.4873e-04, -1.0568e-01,
        -2.4732e-04,  2.4122e-02, -1.8555e-04,  4.4936e-04, -5.6026e-01,
         3.3936e-04,  9.6864e-05, -1.3026e-04, -2.1557e-01,  4.5209e-05,
         3.3078e-04, -3.1780e-04,  7.1473e-04,  4.3902e-02,  3.8189e-04,
         2.2322e-04, -3.8219e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+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-04, -4.3278e-01, -8.4187e-05,  4.0911e-05, -2.9659e-01,
        -2.5598e-05,  9.4576e-04, -2.7120e-04, -2.4862e-05, -1.5623e-01,
         2.4579e-04, -7.0816e-05, -4.0664e-05, -1.1124e-01,  9.3465e-05,
         7.9421e-04, -1.5309e-05, -5.7822e-05,  2.8626e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  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 #2000: [tensor([ 2.5637e-03,  6.4051e-05,  4.7553e-04, -9.3245e-05,  1.5122e-04,
        -2.6233e-04, -4.5346e-04,  1.5020e-05, -4.3077e-03,  5.9043e-01,
         2.9749e-04, -1.6725e-04, -1.9978e-04, -6.5977e-05, -2.3880e-05,
        -3.0402e-04,  4.7212e-04,  1.7835e-04,  1.0767e-05, -2.0355e-04,
         3.9732e-01, -3.7330e-04,  6.2185e-05, -2.5816e-04,  1.8923e-04,
        -7.8997e-04, -4.3165e-05, -2.2659e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9178e-05,  4.5764e-05,  3.8457e-06,  1.2479e-04, -1.3211e-04,
        -9.8445e-05, -2.0203e-04,  6.3881e-05, -2.6549e-03,  4.0301e-01,
         1.4102e-04, -4.8397e-04,  6.7613e-05, -1.8814e-04, -2.5640e-04,
        -1.2653e-04,  2.4201e-05, -6.9043e-05, -2.2121e-05, -1.8433e-06,
         5.8833e-01, -1.7586e-04, -3.0614e-03, -3.7999e-04, -2.4728e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9275e-03,  9.9540e-05,  6.0863e-04, -4.3035e-03, -2.9224e-04,
        -1.1386e-03,  3.3736e-04, -3.7173e-01, -3.4155e-03,  4.8048e-04,
        -2.5829e-04,  6.2610e-05,  5.1187e-04, -4.1162e-04, -1.1751e-03,
        -4.2412e-04, -2.2026e-04, -1.6011e-04, -6.0650e-01, -2.2423e-04,
        -2.2878e-04, -9.2234e-05,  4.7726e-04,  3.2137e-03,  4.2670e-04,
        -2.0577e-04, -7.9850e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2938e-03, -4.2760e-04, -8.9122e-05, -9.5151e-04,  1.6186e-05,
        -8.0216e-04, -4.5614e-04, -4.9774e-03, -1.6632e-03, -9.9922e-05,
         5.4555e-05, -2.8861e-04, -3.9732e-04,  1.2226e-04, -2.0878e-03,
        -1.9777e-04, -4.7027e-05,  1.2307e-04,  2.7388e-02,  1.4122e-04,
         1.6446e-04, -8.1121e-05, -4.0727e-04, -1.4060e-02, -5.2255e-04,
         9.4248e-01, -3.5355e-04, -8.3839e-05, -2.1805e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9933e-04,  1.6648e-04,  1.2080e-04,  3.0984e-04, -5.8225e-05,
        -3.1274e-04,  1.5901e-04, -4.5164e-02,  6.0332e-04,  7.4999e-04,
        -3.3088e-04,  1.9200e-04,  4.0398e-04, -1.8261e-04,  5.8669e-04,
         3.3951e-05,  8.7421e-05, -3.8821e-04, -2.3148e-02,  3.9679e-05,
        -4.5478e-05,  7.7027e-05,  6.1171e-04,  1.3802e-03, -5.1480e-05,
        -2.3598e-05, -9.2427e-01, -1.1822e-04,  7.8096e-05,  8.1633e-06],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9834e-03, -1.5951e-02, -1.2972e-04,  9.7541e-05,  2.8351e-04,
         1.8990e-04,  9.2142e-05, -2.4327e-01,  2.4694e-04, -1.1980e-05,
        -7.3901e-05, -1.5293e-04,  3.9279e-04, -2.5194e-01,  9.9939e-06,
         1.1002e-04, -2.4230e-04, -1.2861e-01,  6.7090e-05, -2.4127e-05,
        -5.9365e-05, -7.7436e-05, -1.3584e-01,  1.2679e-04,  2.0281e-04,
        -1.3543e-05, -2.1722e-01, -1.1198e-04,  1.0654e-04, -3.5122e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6636e-03,  4.3140e-02, -5.8998e-05,  1.2888e-05, -3.9192e-04,
        -2.3791e-04, -4.0659e-04,  7.0090e-02,  9.0520e-05, -1.8576e-04,
        -4.1488e-04, -3.1576e-04,  2.3892e-04, -5.2436e-01, -3.4244e-05,
         2.4235e-04, -2.1255e-04,  6.3409e-02, -2.2606e-04,  1.5584e-04,
        -5.8654e-05,  4.8850e-05, -1.6428e-05, -2.9152e-01, -1.4918e-04,
         2.2837e-04,  8.9884e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6809e-03,  2.7591e-02, -1.5770e-04,  3.1591e-04,  1.1591e-03,
         3.0803e-04,  2.2235e-04, -2.5848e-01,  1.2901e-04,  1.7212e-05,
        -2.7350e-05,  4.6123e-05,  6.1957e-05, -3.5523e-01, -1.2461e-04,
        -1.7655e-05, -1.0807e-04, -1.3562e-01, -4.0520e-04,  3.5444e-04,
         1.7766e-04,  2.5569e-04, -2.1698e-01, -1.3605e-04, -3.0822e-04,
        -7.9840e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6482e-03,  1.0919e-03, -1.9278e-04,  1.2433e-03,  4.7482e-02,
        -9.6806e-04,  6.7818e-04, -9.4012e-01,  9.2412e-04,  7.2212e-04,
        -1.1197e-04, -2.8555e-05, -8.1698e-04,  8.3937e-04,  1.4140e-03,
         5.5581e-04, -1.1627e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2693e-03,  2.5543e-04,  9.6691e-05,  5.7593e-04, -5.4297e-01,
        -3.6377e-04,  5.4563e-04, -4.4304e-01,  7.0812e-04,  4.3537e-04,
         5.2132e-06,  4.9703e-04,  2.4963e-04,  7.2436e-03,  8.3419e-04,
         3.7555e-04,  5.3362e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7841e-04, -5.1647e-05, -4.6306e-04,  5.4537e-05, -9.4421e-02,
         2.0296e-04,  6.9573e-04, -8.9855e-01,  4.2536e-04,  6.5636e-05,
         2.8301e-04,  1.1388e-03,  4.3110e-04,  3.4388e-04,  3.6836e-04,
         5.6145e-04,  1.3689e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9221e-04, -3.5697e-02, -5.2830e-05, -2.1839e-05,  4.3847e-03,
         1.0827e-04,  5.0428e-05, -2.3528e-04, -4.1835e-04,  1.9100e-04,
         1.0644e-02,  1.0174e-04, -3.2625e-04, -3.0927e-04, -4.0179e-01,
         1.8189e-04,  1.5935e-04, -4.0825e-04, -7.6393e-05, -1.7470e-04,
        -2.8657e-01,  3.7033e-05,  7.4597e-05, -2.5644e-01, -6.9595e-05,
        -3.5805e-04, -2.1602e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #2050: [tensor([ 1.6658e-03, -8.1376e-01,  8.3340e-04,  2.0102e-02, -2.5798e-05,
         3.9754e-04,  7.5741e-05, -3.8630e-05,  5.4250e-04,  4.4902e-05,
         5.6132e-05,  3.3154e-02,  3.6601e-04,  4.5197e-02, -8.8070e-05,
         8.2527e-02, -7.2435e-04, -1.0424e-05,  3.9304e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6794e-04, -5.8953e-01, -2.2788e-04, -3.0192e-04,  1.2844e-05,
         1.7524e-04, -3.9048e-04, -1.2565e-04,  1.3819e-02, -2.5684e-04,
         3.2581e-04, -2.0078e-04, -1.9906e-01,  8.8858e-04, -2.5420e-04,
         2.1329e-04,  2.2058e-03, -3.7530e-04,  1.1837e-05, -1.9037e-01,
        -3.7142e-04,  2.8343e-04,  2.3773e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4672e-03,  3.9925e-01, -2.7153e-04, -7.4864e-04,  4.4806e-05,
        -3.9015e-06,  2.0493e-04,  1.2684e-04, -1.5118e-02,  5.1195e-04,
         3.6211e-04, -4.4673e-05,  5.5333e-01, -6.7234e-04, -6.5068e-04,
         4.8603e-04, -9.6167e-04, -1.6985e-04,  5.3694e-05, -2.4578e-02,
         3.3521e-04, -2.5399e-04,  3.1925e-04, -3.6980e-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([ 1.7240e-03,  9.0581e-01, -3.8894e-04, -7.4499e-04,  7.0152e-05,
         7.4047e-05, -2.0049e-04, -7.5832e-05,  4.7491e-02, -7.9498e-05,
         3.0575e-04, -4.1210e-04, -1.3934e-02, -2.4214e-04, -9.0937e-04,
        -2.8050e-05, -3.2168e-03, -4.1411e-04, -1.5547e-04, -2.3058e-02,
        -3.3466e-04, -3.5699e-05, -2.9837e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3483e-04, -2.1699e-01, -1.1293e-04, -1.1561e-04, -1.4301e-02,
        -3.0732e-04,  1.5226e-04,  1.1062e-04,  7.0272e-05, -2.1032e-01,
        -5.0389e-05,  1.5165e-04,  4.9125e-05, -1.8602e-05,  6.5366e-04,
        -2.9591e-01, -3.1667e-04,  2.3452e-04,  1.2139e-02,  9.1188e-05,
         7.6385e-05, -1.1668e-04, -1.2768e-01, -1.3674e-04,  6.1173e-03,
        -1.6225e-04, -1.1303e-01, -9.6791e-05, -2.5593e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0755e-03, -2.2849e-01,  8.1560e-06, -6.8362e-05, -2.1965e-01,
        -1.3144e-04,  7.5561e-05, -1.4526e-04, -1.3943e-04, -1.4101e-01,
         3.7775e-05,  5.2547e-05, -2.4745e-04, -1.4994e-04,  7.3836e-04,
        -1.4713e-01, -1.6200e-04, -1.3889e-04, -9.9461e-02, -6.7728e-05,
        -8.4134e-05,  2.4536e-04, -8.7132e-02,  1.1251e-04,  8.9384e-07,
        -1.7901e-04, -7.2638e-02, -3.8360e-05, -2.1756e-04, -1.7073e-04,
         2.0871e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6686e-03, -2.3429e-01, -3.8173e-04, -1.2882e-04, -2.2132e-01,
        -3.7191e-05,  4.1020e-06, -2.7479e-04, -1.9244e-04, -1.8558e-01,
        -1.8293e-04,  2.4447e-06, -1.2230e-04, -1.6268e-04,  2.5475e-03,
        -2.6395e-01, -5.0416e-04,  9.8513e-05,  1.7805e-02, -1.8087e-04,
         3.3104e-05, -1.6150e-04, -6.9493e-02, -8.7081e-05, -5.8828e-04,
        -1.7435e-04, -3.6179e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5453e-03, -3.0257e-01,  2.3806e-05, -3.2896e-05, -2.0760e-04,
         1.4421e-03, -5.2438e-05, -1.3826e-05, -1.8502e-01, -1.9477e-01,
        -1.3204e-04, -1.2561e-04, -1.9154e-04, -1.0842e-04,  3.5232e-02,
        -1.6314e-04,  6.4456e-05, -1.0133e-04, -2.7676e-01,  3.7780e-04,
        -5.4713e-04,  7.6707e-05, -1.6097e-04, -6.5727e-05,  2.1465e-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([-9.2643e-04, -4.2033e-01,  1.3364e-04, -2.8054e-04,  2.1442e-04,
         3.5220e-03, -1.4550e-04,  1.1407e-04, -1.4626e-01, -3.1408e-01,
        -1.6006e-04, -2.2933e-04,  1.7274e-04,  2.2248e-04, -1.0093e-01,
         5.2897e-05, -4.3842e-05,  1.1857e-04,  1.1182e-02, -2.2870e-04,
         7.2671e-05,  1.7076e-04, -2.1495e-04,  1.8835e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.5391e-04, -4.5318e-01, -2.6049e-05, -1.3336e-04, -1.5500e-04,
         9.2123e-03,  1.9799e-04, -1.8626e-04,  1.8940e-02, -1.1499e-01,
         3.1735e-05, -1.9080e-04,  3.9475e-04,  1.3117e-04, -2.3286e-01,
        -2.2257e-04,  1.2149e-04,  1.1416e-04, -1.3975e-01, -2.0974e-04,
         8.6830e-05, -2.7277e-02, -1.6285e-04,  1.8699e-04,  3.7205e-04,
         1.2329e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.8094e-04, -3.6500e-01, -2.8860e-04,  2.6966e-05, -7.0287e-05,
        -9.4266e-04, -2.0380e-04, -7.7197e-05,  3.9220e-04, -8.1766e-05,
        -4.7489e-05, -1.6730e-01, -2.0900e-04, -1.4764e-04,  5.5278e-05,
        -1.9945e-01,  6.2539e-05, -2.5519e-04, -2.2505e-04,  2.3283e-06,
        -1.6712e-01,  4.5212e-05,  4.5952e-05, -9.6825e-02, -1.4321e-05,
        -1.6591e-04, -1.5491e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2693e-03, -5.4776e-01,  1.6375e-04, -5.9004e-04,  9.4878e-05,
         2.3011e-02, -3.6551e-04, -3.6734e-05,  2.8121e-04,  1.1307e-04,
        -2.1983e-04,  6.1485e-03, -3.0606e-04, -1.2349e-04, -1.7066e-04,
        -1.5020e-01,  9.3486e-05, -2.8206e-04, -3.4605e-04, -1.9550e-04,
        -2.6606e-01,  1.2704e-04,  1.6536e-05,  3.2836e-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>)]
Iter #2100: [tensor([ 5.2077e-05, -1.3493e-03,  3.0380e-05,  3.8956e-04, -6.5461e-05,
        -6.2012e-03,  8.3053e-05, -1.9445e-04,  1.2516e-04, -3.9689e-04,
         9.8737e-02, -3.3058e-04, -3.1793e-03,  5.6829e-01,  4.4743e-05,
        -1.5454e-04, -6.0132e-05, -3.7241e-04, -3.3675e-04,  2.2548e-04,
        -3.6125e-04, -7.1240e-05, -3.9616e-03,  5.2530e-02, -1.3130e-04,
        -7.5990e-03, -4.4591e-04,  2.4152e-04, -8.6639e-05,  2.5338e-01,
        -6.3757e-05, -1.9403e-07, -9.5808e-05,  1.8561e-04,  1.2862e-04,
        -9.7466e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.7988e-04, -5.3983e-05,  3.6928e-04,  8.1437e-05,  5.9805e-04,
         1.5251e-04, -2.8315e-04,  1.2692e-03, -5.7656e-01,  3.4415e-04,
         4.7160e-05, -5.1260e-04, -1.2512e-04, -2.0564e-04, -3.4858e-01,
         3.3992e-04,  3.2050e-03, -4.7136e-05,  6.5826e-02,  2.3123e-04,
        -7.9084e-05,  2.0296e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.6420e-04,  4.0081e-04,  3.3006e-04, -4.5876e-04, -6.6068e-03,
         1.4810e-05, -3.7717e-05, -1.1218e-03,  9.8671e-01,  5.8890e-04,
         6.3793e-04,  6.2219e-04,  7.0760e-05,  2.2357e-04,  3.0275e-04,
         3.1271e-04,  5.9402e-04, -2.6914e-05, -1.0473e-05,  6.8286e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.6439e-04,  5.5254e-04,  5.5392e-04,  1.0175e-03,  1.0936e-03,
         6.0191e-04, -4.9947e-05,  1.1379e-03, -9.9034e-01,  5.7195e-05,
        -2.9576e-04, -7.5300e-04, -4.5516e-05,  4.1219e-04, -1.1484e-04,
         4.7375e-04, -2.9758e-04,  2.3110e-04, -2.9915e-04,  1.1054e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3253e-05, -2.2241e-01, -1.1515e-04, -8.2361e-06, -2.6400e-01,
         1.2599e-05,  4.5063e-04,  2.0574e-04,  1.0620e-04, -3.0407e-01,
         9.4152e-06,  1.8996e-04, -5.2045e-05,  3.0956e-04,  8.6137e-05,
         1.2844e-04, -6.0687e-02, -6.4228e-05,  2.5280e-04, -1.4595e-01,
        -3.6527e-05,  1.8088e-04, -2.3798e-04, -1.4009e-04, -1.3380e-05,
        -1.9869e-04,  1.4919e-05, -5.4802e-06, -3.0462e-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([ 8.6785e-04,  7.8579e-02,  4.0516e-04,  5.1932e-04, -1.5775e-04,
         4.3392e-05, -6.5011e-05, -4.3509e-03,  1.6794e-04,  1.1450e-01,
         4.4034e-05,  1.7881e-04,  2.0254e-04, -5.9004e-04, -1.3018e-05,
        -5.6262e-05,  4.4915e-02,  7.9493e-06,  5.2938e-04,  3.9992e-01,
         1.5059e-04, -7.4464e-05, -2.2145e-04,  3.2471e-05,  3.8417e-05,
         4.3147e-04,  3.5228e-01,  1.6606e-04,  2.7992e-04, -1.4484e-04,
         6.4768e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4519e-03, -8.6499e-02,  1.0094e-04,  7.9392e-06, -2.2469e-01,
        -2.3777e-04, -1.7431e-05,  1.3677e-03,  1.7472e-04, -2.6051e-01,
         5.2852e-05,  1.8790e-04,  8.9261e-05,  2.7332e-04,  2.5538e-04,
         1.8164e-05,  5.9366e-02,  1.8443e-04,  3.8922e-04,  1.2948e-04,
        -3.6361e-01, -9.1498e-05, -2.9099e-05,  2.4454e-05, -2.3172e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.0484e-04,  4.7945e-04,  3.7459e-04,  9.9098e-01,  4.3948e-04,
         2.6564e-04, -5.3710e-04,  3.9601e-04,  2.6143e-04, -6.3258e-04,
         1.0775e-03,  1.5545e-04, -4.6540e-04,  3.0067e-04,  1.0865e-03,
         5.7897e-04,  4.0936e-04,  2.9380e-04,  2.7393e-04,  8.8788e-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([-6.2514e-03,  5.6219e-04,  4.4952e-04,  9.8414e-01,  2.5126e-04,
        -2.3528e-04,  1.1149e-04,  8.6126e-04, -7.6121e-04,  8.1275e-04,
         3.2878e-04,  5.7754e-04,  4.5718e-04,  4.8170e-04, -2.0832e-04,
         5.1747e-04, -3.3638e-04, -6.3928e-04, -2.2538e-04, -5.9934e-04,
        -8.0731e-04,  3.8639e-04,  1.1740e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3391e-03,  5.7043e-05,  1.7146e-04,  9.8703e-01,  6.7150e-04,
        -8.4438e-04,  3.1805e-04, -7.0482e-04, -3.6986e-05,  3.6476e-04,
         1.0531e-03,  1.8240e-04, -2.9606e-04,  4.3735e-04,  1.0972e-03,
         2.5199e-04,  1.3128e-03,  1.8339e-04,  4.1049e-04,  1.1260e-03,
         2.4159e-05, -2.9008e-04, -1.0659e-04,  6.9490e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.1530e-04,  7.4545e-02, -1.6621e-04,  3.9166e-04,  1.4525e-04,
         4.3336e-04,  1.5746e-04, -4.9942e-01, -2.1202e-05,  2.0389e-04,
         1.4885e-04,  4.2447e-05, -8.0916e-05,  5.5525e-04, -3.9600e-01,
        -2.7186e-05,  2.4784e-04,  3.1157e-04,  1.5072e-04, -7.0457e-04,
        -2.4681e-02,  9.2065e-06, -4.5869e-04,  2.9048e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9738e-04, -1.6519e-01,  7.8553e-05,  3.6925e-04,  9.0487e-05,
        -1.4203e-04, -1.2291e-04, -2.2382e-01,  1.0314e-05,  4.9928e-04,
         2.6476e-04,  8.0508e-06, -3.0247e-04,  9.5263e-05, -6.0653e-01,
         1.0847e-04,  2.4013e-04,  5.0020e-04,  3.8084e-05,  3.0143e-05,
        -2.7649e-04,  4.6670e-04, -1.2063e-04, -4.0589e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         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 #2150: [tensor([ 6.7170e-04, -3.9712e-01,  6.4592e-05, -1.0584e-04, -2.9922e-01,
        -2.4442e-05, -3.8344e-04, -1.2610e-01, -4.5375e-04, -4.3369e-04,
         1.8972e-03, -2.8106e-04, -1.5981e-04, -2.9625e-04, -1.7077e-01,
        -1.1167e-05, -3.0070e-04,  4.1807e-05,  1.2471e-04,  1.0945e-04,
        -3.7112e-04,  1.9263e-05, -7.0798e-05, -3.4322e-04, -4.1385e-04,
         3.8745e-05,  1.7286e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1213e-03,  7.8097e-05,  4.8494e-04,  2.9957e-04,  1.9674e-04,
         1.2533e-04, -1.9728e-04,  3.3623e-04,  3.5665e-04,  4.0772e-04,
        -8.7098e-05,  2.6989e-04,  1.9303e-04,  2.4115e-04,  1.4053e-04,
         8.4990e-05, -2.4950e-05,  3.6085e-04,  9.9469e-01,  1.7296e-05,
        -9.6697e-05, -1.8597e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3583e-04,  5.7837e-04,  6.4474e-04,  7.6469e-04, -6.6568e-05,
        -9.0084e-05,  2.9679e-04,  8.3909e-04, -1.3563e-04, -1.4272e-04,
         7.1759e-04, -2.9646e-05,  2.8790e-04,  4.5328e-04,  1.4409e-04,
         5.7910e-01, -5.1326e-04, -3.7491e-04,  4.6589e-04,  3.2204e-04,
         6.2643e-04, -1.7305e-04, -2.6285e-04,  4.1228e-01, -2.3065e-04,
         3.0843e-04,  1.0957e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3140e-04,  5.3606e-05,  3.3120e-04,  4.1005e-04, -7.3382e-05,
         2.5604e-04,  3.3263e-04,  1.7527e-04,  3.0545e-04,  2.4847e-04,
         5.9643e-04, -1.4688e-04,  3.4274e-04, -1.6758e-04,  2.0965e-04,
         3.9956e-01, -2.4138e-04,  1.7810e-04,  5.9481e-01,  3.6144e-04,
        -5.5935e-04,  2.1637e-04, -1.4396e-04, -2.0394e-05, -1.3015e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4272e-03,  3.8876e-04, -8.9854e-05, -4.2300e-05, -4.4748e-06,
         1.5494e-04, -2.8386e-01,  9.1919e-05,  1.1793e-04,  2.7786e-04,
        -2.8296e-01,  1.5096e-04,  3.2361e-04, -1.2027e-01,  7.0759e-04,
         1.2470e-04,  4.3546e-05,  9.6104e-06,  5.7453e-03, -2.1894e-05,
        -9.9833e-05,  1.3193e-04,  1.3612e-04,  3.0596e-04, -1.6435e-05,
         5.4978e-06,  2.3827e-04,  3.1377e-05,  1.0507e-05, -1.7536e-01,
         3.9056e-05, -2.8249e-05,  5.5382e-06, -1.2527e-01, -1.6540e-04,
         9.7837e-04,  1.8329e-05, -2.1088e-04,  1.2677e-04,  1.1540e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6479e-03,  1.7275e-04, -9.9404e-05,  7.3483e-05, -1.0487e-05,
         9.2915e-05, -3.7093e-01, -4.2118e-05,  4.1298e-05,  1.1708e-04,
         7.0813e-02, -1.7973e-04,  2.4363e-04, -2.1246e-01, -3.8301e-04,
         4.1831e-05,  5.2636e-05, -5.2568e-05,  1.8485e-02, -2.4858e-05,
         2.8080e-04,  9.4798e-05,  5.6954e-05, -3.1337e-04, -1.5442e-04,
         1.0522e-04,  6.0185e-06,  1.6477e-04, -2.6925e-05, -1.7271e-01,
        -7.3441e-05, -2.8082e-04, -5.8478e-05, -1.4746e-01,  1.3969e-05,
        -7.6191e-05, -4.3958e-05,  7.3056e-05,  4.3725e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0375e-03, -2.3668e-04, -1.4672e-04, -1.8113e-05,  2.3075e-05,
         7.8784e-05, -2.2531e-01, -1.0819e-04, -4.4980e-05, -1.8297e-04,
        -2.9671e-01, -4.7708e-06,  1.2111e-04, -1.9451e-01,  1.2102e-04,
        -1.9569e-05, -2.8114e-04, -1.1426e-04,  2.8938e-03, -1.9483e-05,
        -5.6123e-05,  1.0785e-04,  1.0279e-04, -2.6907e-04,  1.2383e-04,
         3.2490e-05, -5.5187e-06, -2.2668e-05, -1.0899e-04, -1.8774e-01,
         6.8532e-05, -1.7906e-04,  1.1511e-04,  1.7404e-02, -7.1790e-05,
        -1.0251e-04, -2.2741e-04, -1.0056e-04,  1.6791e-04, -7.0906e-02,
        -8.8034e-05, -1.1738e-05, -8.2050e-06], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1362e-03,  4.4537e-01,  4.9595e-05, -8.8182e-05, -9.2001e-05,
        -1.4760e-03,  7.1928e-05,  2.4895e-04, -2.0157e-05, -3.9856e-05,
         1.6135e-01, -6.6014e-05, -3.0038e-04,  7.6005e-02,  1.4252e-04,
        -1.3963e-04, -4.6376e-05, -2.5986e-04,  2.1366e-01,  1.4838e-04,
        -6.0267e-05, -1.2775e-04,  9.8365e-02, -1.2742e-04,  3.0800e-04,
         1.0592e-04, -2.0633e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.2697e-04, -3.8450e-01,  1.0985e-04, -2.8446e-04, -2.5560e-04,
         3.0943e-04,  1.9014e-05, -3.9265e-04,  1.2483e-04,  2.1470e-04,
        -3.2699e-01, -1.5361e-04,  3.8577e-04, -1.0142e-01, -2.9932e-04,
         1.1177e-04, -3.9332e-05,  2.9882e-05, -1.8251e-01, -6.2250e-04,
         1.1203e-05,  1.9409e-04, -1.0615e-04,  8.7129e-05,  1.0810e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.0786e-04, -3.9675e-01,  2.2995e-05, -1.4519e-04,  9.6133e-05,
         7.0931e-04, -1.5981e-04, -2.4813e-04, -1.8106e-04, -3.0847e-04,
        -2.4899e-01, -7.7784e-05,  1.4265e-04, -7.1573e-02, -1.6250e-04,
         4.9556e-06,  1.9709e-05,  3.8059e-05, -2.7891e-01,  4.8990e-06,
        -4.7027e-05,  2.0029e-05, -4.3158e-04,  2.2173e-04,  3.5864e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3728e-03,  6.0586e-04, -1.5881e-04, -1.1528e-04,  2.3277e-04,
         8.9448e-05,  5.6977e-04, -4.9673e-01,  1.7188e-04,  3.1284e-04,
        -3.7725e-04, -1.0849e-04,  1.2276e-04, -1.8269e-04,  3.3245e-04,
        -5.2432e-04, -4.9749e-01, -1.0895e-05, -3.2458e-04, -1.6637e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.1784e-03, -2.1888e-04,  3.6449e-04,  1.1109e-04, -5.0006e-04,
         1.9223e-04,  6.9764e-04, -7.0517e-01,  3.2189e-04, -9.8992e-06,
        -3.7797e-04,  4.3169e-04,  1.1517e-04,  4.2522e-04,  1.6951e-03,
        -2.8424e-01,  1.8894e-04,  5.1051e-05,  7.0822e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  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([-1.3042e-03,  5.9911e-04,  3.2294e-04, -3.4138e-01, -7.8843e-04,
         8.2956e-05, -2.0391e-04, -3.3168e-04, -1.4370e-01,  1.4310e-04,
         3.4592e-02, -4.7478e-01, -3.4326e-05,  5.8160e-04, -1.0718e-04,
         1.5485e-04, -2.9871e-04,  6.9813e-05,  4.4781e-05, -1.8262e-04,
         1.7694e-04, -7.8589e-05,  3.9074e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.6823e-04,  6.2915e-06,  1.2867e-04, -2.8853e-01, -4.0014e-04,
         2.9640e-04, -2.2850e-04, -8.8109e-05, -1.1402e-01,  2.1378e-04,
         2.0384e-02, -3.7297e-01,  2.0871e-04,  1.8692e-03, -1.8470e-04,
         7.8638e-05,  1.5379e-04,  1.3302e-04, -1.7494e-04, -1.9894e-01,
         1.6768e-05, -2.9007e-05, -9.7783e-05, -7.7670e-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([-8.4611e-04,  1.4911e-04,  9.9387e-05,  2.1523e-01,  2.3222e-04,
        -1.0585e-04,  6.3110e-05, -6.0526e-04,  3.5686e-01, -2.7868e-05,
        -2.1921e-02,  2.9350e-01,  1.2561e-05,  1.8573e-06,  1.8143e-04,
         1.0124e-04,  2.6072e-04,  1.8845e-04,  2.0130e-04,  1.0900e-01,
         1.4591e-04, -1.8211e-04,  8.1318e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0097e-03,  1.4223e-04, -1.0193e-04,  4.5003e-05,  1.1015e-03,
         3.5398e-05,  2.4107e-04, -9.7300e-05, -2.1831e-05,  1.3299e-04,
         2.5932e-05, -8.2542e-06,  1.0275e-04,  6.9396e-01,  2.2573e-04,
         2.1864e-04,  1.4485e-01,  2.0163e-04, -2.3495e-04, -5.5466e-05,
        -7.6956e-05, -1.8196e-04, -3.6302e-05, -4.6547e-05, -1.0502e-05,
         9.5416e-05,  1.0953e-04, -7.9379e-05,  9.2606e-05, -1.3291e-05,
        -2.0484e-04,  2.7079e-04, -1.6348e-04,  1.5561e-01, -1.0350e-05,
        -6.7956e-05, -2.0629e-05,  8.9802e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.9921e-04,  6.1102e-05, -1.6098e-04, -3.2446e-06, -8.9275e-02,
         1.9734e-05, -5.2650e-05,  3.9382e-05, -8.6460e-05, -6.1987e-05,
        -1.5888e-05,  7.3770e-05, -1.1399e-04, -2.9167e-01,  3.6216e-05,
        -1.4792e-04, -5.5945e-02, -1.5416e-05,  2.1713e-05,  5.0656e-05,
        -5.6221e-05,  5.6522e-05, -9.5633e-05, -1.3840e-05, -5.8683e-06,
         1.8809e-04, -6.5925e-05,  8.6050e-05, -6.6017e-05,  1.3591e-05,
        -1.1320e-01, -9.1326e-05, -1.5310e-01,  1.1949e-05,  3.0768e-04,
         8.9417e-04,  4.4672e-05,  8.1692e-05, -2.9282e-01,  1.4525e-04,
        -3.9124e-05, -5.8886e-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([ 9.3393e-05, -1.1055e-04, -1.8657e-04, -1.5201e-04, -1.6351e-02,
        -1.5183e-04, -1.9075e-05, -2.3409e-05, -3.9601e-04, -1.4501e-05,
        -9.4262e-05, -1.5776e-04, -4.9383e-04,  4.1489e-01, -2.9338e-04,
        -1.8370e-05,  1.0965e-01, -4.6417e-05, -4.1944e-05, -2.2025e-04,
        -7.4218e-05, -7.7559e-05, -1.0699e-04,  7.5581e-05,  2.4303e-04,
        -2.5847e-04,  1.5091e-04, -5.0653e-05, -1.9988e-04, -1.6861e-04,
         6.6040e-05,  1.0207e-04,  5.3118e-05,  1.1168e-01, -2.3083e-04,
        -3.9486e-04,  3.4218e-01,  2.2959e-05, -2.3305e-04,  1.9126e-04,
         3.6476e-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([ 4.0624e-04,  1.7083e-04,  1.1048e-01, -5.3339e-05,  4.7301e-05,
         4.6485e-05, -5.4623e-01,  5.2339e-05, -4.5198e-05,  9.0108e-04,
        -2.6800e-04,  3.9293e-04,  1.2390e-04, -4.4077e-05,  2.9305e-05,
         1.5701e-03, -1.0409e-05, -7.6428e-05,  2.9078e-05, -1.0161e-04,
         3.1225e-04,  5.8826e-02, -1.7699e-04, -1.5946e-04,  1.1942e-04,
        -2.6220e-04, -1.7395e-04, -9.9698e-05, -8.6265e-06, -8.4464e-05,
        -2.7723e-01,  3.1963e-04,  5.2152e-04, -6.3502e-05,  1.1685e-04,
         4.8960e-05, -1.3977e-04, -1.1192e-04, -7.7398e-05,  6.5417e-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([ 1.1565e-03,  1.4430e-04, -1.5805e-01,  4.1589e-06,  4.5841e-05,
         4.5102e-05, -4.0676e-01,  1.5655e-06,  1.7434e-05,  3.3529e-04,
         3.8657e-05, -8.1417e-05,  2.4168e-04, -3.9821e-05, -4.0842e-05,
         2.8679e-03, -9.2566e-05,  4.4623e-05,  8.2234e-05,  1.5803e-04,
        -1.8059e-05,  1.6137e-02, -3.8134e-05,  7.5276e-05,  1.3402e-04,
        -9.5325e-05,  5.9417e-05, -4.0123e-05,  5.6727e-05,  3.9303e-05,
        -1.8982e-01, -4.8513e-05,  6.2653e-04,  1.8553e-04, -2.9664e-05,
        -8.2292e-05, -8.0510e-05,  2.8357e-03, -6.0545e-05, -2.7853e-05,
         1.5883e-04, -1.8734e-05,  6.0712e-06, -3.7205e-05,  1.2045e-04,
        -8.2281e-02, -8.0104e-05,  7.2818e-05, -8.1556e-05, -1.3601e-01,
         3.8984e-05, -1.3625e-04,  1.3256e-04, -4.1544e-05, -4.4012e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.3624e-04,  2.9822e-04, -9.8955e-02,  1.4172e-04,  1.8945e-04,
         6.9786e-07,  3.4562e-01, -1.1007e-04,  8.3356e-05, -2.7449e-04,
         1.4302e-04, -4.3287e-04, -1.1200e-04,  4.8428e-05,  1.7553e-05,
        -5.0178e-04,  2.2497e-04,  1.3582e-05,  8.6560e-05, -2.2398e-04,
        -1.4402e-04,  1.0480e-01,  1.9334e-05,  3.2106e-05, -1.5893e-04,
        -1.0470e-04, -1.7751e-04, -1.5293e-04,  4.4173e-05,  1.6484e-04,
         2.8544e-01,  1.0536e-04, -8.4707e-04, -1.1193e-04,  3.7561e-04,
         1.5913e-01, -4.9601e-05, -2.8174e-06, -3.8504e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.0466e-04,  4.6816e-01,  7.6961e-06,  2.3890e-05,  5.2243e-05,
        -3.1379e-03,  4.1957e-05,  3.2719e-05,  2.7698e-05, -3.5494e-04,
        -4.9503e-05,  1.0778e-04,  1.5568e-01,  8.3659e-05,  2.1433e-05,
         3.9525e-07,  9.1623e-06, -8.7525e-05,  7.6385e-05,  3.8707e-02,
        -5.2360e-06, -1.3966e-04,  1.1294e-05, -1.8172e-05, -6.6675e-06,
         6.1718e-05,  1.9304e-04, -4.7492e-06, -1.6717e-04, -1.9238e-04,
        -4.1493e-05, -1.0727e-05,  1.5293e-01, -4.7786e-05,  1.1420e-05,
        -6.1791e-04,  1.5856e-05,  1.5266e-04,  3.4988e-02,  6.8698e-05,
        -4.3459e-05,  1.0397e-04,  4.7149e-05,  7.8140e-02,  6.4412e-02,
         5.8895e-05,  7.1688e-05, -2.0609e-04,  2.8428e-06, -8.4939e-05,
        -3.8004e-06,  2.6500e-05, -3.2608e-05, -1.9196e-05,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.5174e-04, -1.7313e-01,  2.7697e-05, -5.6711e-05,  2.7471e-05,
         6.9649e-03,  7.3697e-06, -3.1061e-04, -1.2007e-04,  4.5384e-04,
         3.8485e-05, -1.1035e-04, -9.7576e-02, -5.7228e-05,  6.9753e-05,
         2.3743e-05, -3.4430e-05, -1.3233e-06,  2.4205e-04, -3.6047e-01,
         1.6631e-05,  1.7422e-04,  2.1354e-05, -1.8793e-05,  1.8702e-04,
         2.1187e-05,  9.3252e-05, -1.0938e-04,  1.5423e-04,  9.9050e-05,
         9.3028e-05,  9.3391e-05, -2.1747e-05,  5.2442e-05,  1.0780e-05,
        -2.5122e-01, -7.4983e-05,  1.2735e-04, -1.0681e-01,  9.7233e-06,
        -1.0164e-05, -1.0547e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7077e-03, -7.6759e-03,  6.0281e-05,  6.6944e-05,  5.9090e-05,
        -2.0173e-03, -8.2342e-05,  5.4652e-05,  2.3243e-04, -4.8674e-04,
        -6.1023e-05,  2.4161e-04,  1.2166e-01,  7.6801e-05, -1.5586e-05,
         1.8938e-05,  8.7098e-05,  3.5364e-05,  2.3704e-04,  4.3768e-01,
         1.9690e-04, -6.7654e-05,  9.2285e-05, -2.1866e-04,  1.2850e-04,
         1.0520e-06,  2.7796e-04, -6.1249e-05,  1.4710e-04, -1.0339e-04,
         6.6650e-05,  1.2531e-04,  6.9322e-02, -2.2345e-04, -1.0276e-04,
        -1.0775e-03, -1.0160e-04, -1.8752e-04,  3.2792e-01,  5.4605e-05,
        -5.1149e-06, -1.2364e-04,  5.2361e-06, -5.4850e-05,  1.0958e-04,
        -2.6420e-02,  3.2028e-05, -1.9345e-04, -1.5006e-05, -7.1756e-06,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #2250: [tensor([ 1.2684e-04, -3.5419e-01, -5.4399e-05, -1.3037e-01,  3.1028e-05,
         6.8594e-05, -1.9216e-01, -3.6943e-05,  2.3751e-05,  4.1360e-05,
         1.9031e-04,  1.1839e-04, -1.8501e-04,  4.3442e-05,  1.2499e-04,
        -1.1537e-04,  5.5106e-04, -1.4810e-02,  1.0529e-04,  2.6263e-05,
         9.8356e-05,  1.1184e-06, -7.5770e-02,  6.8448e-06,  6.8854e-05,
         5.8852e-05,  6.6784e-06,  6.3915e-05, -1.6022e-01, -1.9985e-05,
         9.9731e-05,  5.1694e-05, -1.4629e-04,  2.2979e-03, -3.9748e-05,
        -1.4406e-05,  3.5648e-05, -6.7543e-02, -3.6567e-05, -4.6553e-05,
        -8.0976e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0854e-03, -1.8982e-01, -6.1384e-05, -1.3675e-01, -2.0363e-05,
         7.1903e-05, -1.7039e-01,  1.1516e-04,  4.2250e-05,  3.6953e-04,
         3.0664e-05,  1.2944e-04,  1.2484e-04,  5.0030e-05,  2.8384e-05,
         4.2373e-05,  3.6622e-04, -1.5288e-01,  1.7140e-05,  2.4041e-05,
         1.6608e-04,  7.8982e-05, -7.8794e-02, -4.1947e-05,  1.2702e-05,
        -2.3969e-05, -8.6898e-05,  6.8290e-05, -1.7107e-01, -2.1728e-04,
        -2.2873e-05, -7.9408e-05, -3.0054e-05, -5.2671e-05,  1.0254e-04,
        -9.4820e-02,  1.3022e-04,  5.8337e-04,  3.8188e-05, -4.9308e-05,
         1.0436e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.2972e-04, -6.9396e-02,  3.5973e-05, -1.1274e-01,  9.9241e-06,
        -1.1930e-05, -1.9890e-01, -3.6309e-05,  8.8508e-06,  1.6958e-04,
         4.3798e-05,  4.6572e-05,  6.2033e-06, -2.2258e-05,  4.5802e-05,
         9.5537e-05,  4.3210e-05, -1.6839e-01,  5.2097e-05,  4.2279e-05,
         4.9142e-05,  3.8457e-05, -1.5307e-01,  1.2460e-04, -8.8625e-06,
        -4.6421e-05,  2.9336e-05,  6.0306e-05, -1.0215e-01, -2.1502e-06,
         1.7942e-05,  3.2499e-05,  1.2900e-05,  3.0423e-05, -1.3930e-04,
        -3.7345e-05, -5.4729e-05,  2.3846e-05, -2.3252e-02,  1.3657e-05,
        -1.0169e-04,  1.0718e-04,  6.1878e-05, -8.7456e-05, -9.4696e-02,
         4.3435e-05,  1.7510e-04,  3.8528e-05,  4.0747e-05,  4.6277e-05,
        -7.5789e-05, -7.9913e-05, -3.7941e-05, -7.4175e-02, -1.0439e-04,
         3.6285e-05,  6.8503e-05,  3.4584e-06], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.6966e-04,  1.8329e-01,  2.8939e-05, -2.2870e-02, -8.4687e-05,
        -9.9131e-05,  2.3665e-01,  1.8471e-04,  7.2119e-05, -3.9871e-04,
        -1.3737e-04, -6.8322e-05, -9.3681e-05,  6.1035e-06, -1.7768e-05,
        -1.0990e-04, -1.6215e-03,  1.1358e-01,  7.3468e-05,  1.7192e-05,
        -1.4002e-05, -8.8894e-05,  2.7171e-01,  4.5288e-05, -1.6673e-04,
        -6.4267e-05,  3.2793e-05,  1.2070e-04,  1.6700e-01,  8.2485e-05,
        -1.8008e-04,  1.0536e-05, -1.4026e-04, -8.6226e-05, -6.1400e-05,
        -5.3558e-06, -1.0902e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8372e-03, -2.2553e-01, -4.3321e-05, -1.0370e-01, -1.8780e-04,
        -1.2973e-05, -1.7055e-01, -1.5799e-04, -3.4508e-05,  1.6046e-04,
        -1.1523e-04,  5.6677e-05,  2.9903e-06, -1.8225e-04, -9.5853e-05,
        -1.8041e-04, -1.7143e-04, -2.2402e-01,  3.9265e-05, -4.7721e-05,
         7.5936e-07, -1.1648e-05, -1.2257e-01, -1.5181e-05,  6.1117e-06,
         1.0339e-05,  3.0363e-05, -3.5766e-05, -4.7417e-02, -7.7344e-05,
        -8.5924e-06,  1.2693e-02, -1.5064e-04, -5.8241e-05,  8.6549e-06,
         8.1125e-05, -8.0404e-02, -5.7515e-05,  3.2378e-05, -7.1140e-06,
         9.0461e-03, -7.8440e-05,  5.5399e-05,  1.4515e-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([-1.3191e-05, -1.2936e-01, -1.2630e-04, -9.9033e-02, -2.3345e-06,
         1.1486e-05, -1.9633e-01, -1.3676e-04,  2.2334e-05,  2.1896e-04,
        -1.3436e-04,  1.8148e-04, -1.4382e-04, -4.1846e-05,  6.6135e-05,
        -2.0916e-04,  7.6702e-05, -2.9766e-01,  7.7993e-05, -6.7688e-06,
         6.2418e-05,  3.9310e-05,  4.6952e-02,  5.7676e-05, -1.4623e-07,
        -1.9474e-04, -7.6651e-05, -7.0881e-05, -9.4701e-02,  1.1371e-04,
         8.5819e-05,  1.0358e-04, -3.8569e-05, -1.3342e-01, -7.9430e-05,
         1.4903e-05, -1.0759e-04, -1.5079e-05,  1.5387e-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([-7.8530e-04, -2.7831e-01, -1.7115e-04,  6.2279e-02, -2.6102e-04,
        -1.0057e-05, -1.5612e-01, -9.7221e-05, -3.5405e-05, -1.9078e-04,
         1.4834e-04,  2.1328e-04, -1.7670e-04,  4.5671e-05,  9.2582e-05,
        -3.2065e-04, -1.8153e-04, -1.9032e-01,  1.2975e-04, -2.2440e-04,
        -2.3263e-04,  6.1543e-05, -1.4923e-01,  1.7271e-04, -2.0296e-05,
        -3.2667e-05,  4.6206e-05, -6.3417e-05, -1.5953e-01, -1.5927e-04,
         1.0426e-04, -9.0421e-05,  3.8595e-05, -1.0066e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.7323e-04, -1.5816e-01, -3.3989e-05,  1.5454e-02, -5.9823e-06,
         7.5618e-05, -1.7796e-01, -4.2224e-05, -5.8338e-05,  4.4997e-05,
         7.6066e-05,  1.3450e-04, -3.3073e-05,  1.2612e-05, -3.2233e-05,
        -1.5347e-04,  3.2999e-04, -2.1014e-01,  8.6845e-05, -1.4553e-05,
         3.7063e-05, -1.5592e-04, -1.0799e-01,  3.8130e-05,  6.8711e-05,
         9.4816e-05, -4.5169e-05, -2.4149e-05, -1.8533e-01, -1.5746e-04,
         6.9891e-04, -1.2233e-04, -1.4078e-01,  4.1239e-05,  6.3385e-07,
         4.5237e-05,  3.8256e-05, -8.5194e-05, -2.6976e-06, -1.4842e-05,
         7.7956e-07,  1.0241e-04, -7.7326e-05,  1.3735e-04, -6.7785e-05,
        -1.6101e-04,  1.1733e-05,  5.4437e-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([-1.3192e-03, -9.2731e-02, -2.1788e-04, -4.6162e-02,  9.4133e-05,
         2.4098e-05, -2.8600e-01, -5.3915e-05,  2.7848e-05,  7.7140e-06,
        -1.8255e-04,  4.8015e-05, -1.3876e-05, -2.6878e-05, -3.3172e-05,
        -5.0936e-05,  2.7385e-04, -1.7767e-01, -1.3857e-04,  5.7511e-05,
         2.2251e-04,  1.6043e-05, -2.1337e-01,  3.6112e-05, -8.9019e-05,
         6.0108e-05,  6.0386e-05,  5.0326e-05, -1.7971e-01, -2.1984e-04,
         4.4826e-04, -1.5033e-05, -1.3946e-04,  7.6976e-05, -1.3087e-04,
        -1.6778e-05,  1.3317e-04, -6.8261e-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([-1.8571e-03,  7.4006e-05, -4.2740e-05,  9.3993e-04,  4.1328e-04,
        -4.3066e-05,  7.7713e-04,  1.0579e-05,  2.0582e-03, -7.8262e-02,
         6.2368e-05, -2.8395e-05, -4.5040e-06, -2.9166e-01, -1.9756e-04,
        -1.8555e-04, -8.1512e-06, -2.6747e-01,  1.5626e-04,  3.5815e-04,
        -3.5519e-01,  9.3625e-05, -5.5689e-06,  9.4999e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.0970e-04, -1.4776e-04, -8.6103e-05,  1.0885e-02,  2.1561e-05,
        -2.9164e-04,  9.2563e-04,  2.3003e-04,  2.3498e-02, -3.7513e-01,
         2.5469e-04,  8.7616e-05, -1.9715e-04, -3.5815e-01, -6.2830e-05,
         2.5218e-04, -2.2833e-01, -1.5405e-04, -3.2177e-04,  2.6832e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.9747e-04, -1.1096e-04, -3.0276e-05,  1.0205e-03, -2.6499e-04,
        -1.0986e-04,  7.3133e-04,  2.0940e-04,  1.2746e-02, -4.6501e-01,
        -1.1607e-04,  3.9736e-05,  8.6780e-05, -1.7514e-01,  7.9431e-05,
         1.3448e-04, -3.2166e-01, -6.9730e-06, -5.5294e-05,  7.8970e-05,
        -2.1253e-02,  1.5909e-05, -1.5133e-04, -5.1616e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #2300: [tensor([-2.1204e-03, -2.0090e-04, -5.6741e-04, -6.6050e-04, -2.6497e-04,
        -9.7619e-04, -4.0645e-04,  8.3214e-05, -3.1780e-02, -4.7773e-04,
         2.2770e-05, -2.1375e-01, -8.3641e-04,  5.0053e-06, -2.1949e-01,
         4.4843e-04,  9.3943e-05, -2.4257e-01,  2.6189e-06, -1.3694e-04,
        -1.4139e-03, -6.8803e-04, -1.9565e-04, -5.5973e-04, -3.3911e-04,
        -2.7980e-01, -1.6153e-03, -2.7219e-04,  2.2216e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.0404e-03, -3.0131e-04, -9.2931e-05,  2.1300e-04, -3.3037e-04,
        -6.6706e-04, -6.4243e-04, -2.4017e-04, -2.9483e-02, -4.1992e-04,
         2.2585e-05, -2.1587e-01, -7.2250e-04,  2.6612e-04, -3.2368e-01,
         1.8935e-04, -3.3749e-04, -4.1535e-01, -8.1292e-04, -2.9541e-04,
        -6.9378e-04, -5.8698e-04, -2.3806e-04,  2.6168e-04, -2.2394e-04,
         1.1359e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.3481e-04, -2.4990e-01, -2.8242e-04, -1.3992e-04,  7.2391e-04,
        -6.4235e-05, -1.7852e-04, -6.1773e-05,  5.9563e-04, -3.7693e-01,
         2.6813e-04,  4.7933e-03, -2.0541e-04,  1.2967e-04,  2.2924e-05,
        -1.3494e-01, -2.4080e-04,  3.1430e-04,  2.3303e-04, -2.2863e-01,
         3.9440e-04, -1.4780e-04, -7.1642e-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([-2.7036e-03, -1.6956e-01, -6.4003e-04, -1.1069e-04,  1.5552e-04,
         1.6382e-04, -8.3893e-05,  2.0715e-05,  2.2517e-03, -5.7963e-01,
         1.4162e-04,  1.6096e-02, -6.0753e-05,  1.5513e-04,  6.1624e-05,
        -2.2656e-01,  3.0942e-04, -5.8405e-04,  1.1121e-07,  1.5634e-04,
         7.1029e-05,  2.2481e-04,  2.6461e-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([-2.1280e-03, -4.7120e-01, -1.3849e-04, -1.1953e-05,  4.4781e-04,
        -4.9928e-06, -4.7648e-04, -2.1505e-04, -3.3426e-04, -2.4217e-01,
        -2.6115e-05,  1.3091e-02, -8.8152e-05, -2.6950e-04,  1.1068e-04,
        -2.6455e-01,  7.5010e-05, -3.3781e-04, -4.0511e-04,  3.1149e-03,
         4.5131e-04, -3.0768e-05,  2.6573e-04,  5.7969e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.8338e-03, -6.3765e-01, -3.1436e-04,  1.3122e-04,  2.1844e-05,
        -1.9010e-03,  8.3122e-05, -8.6069e-04, -7.3745e-02, -3.6810e-04,
         4.9406e-04, -9.9539e-05, -2.7278e-01,  2.7719e-04, -9.6962e-04,
        -1.2461e-04,  2.0471e-04, -1.3585e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1504e-03, -4.1119e-01,  1.2478e-04,  9.9532e-05, -1.7802e-06,
        -1.6745e-03,  1.6659e-04, -6.2129e-04, -6.5835e-02, -1.3932e-04,
         7.9221e-06,  1.4465e-04,  8.2160e-02,  4.2252e-05,  1.9875e-03,
         4.7659e-05, -1.5229e-01,  1.7146e-04, -1.6538e-04, -2.8013e-01,
        -4.6856e-04,  4.2542e-05,  3.3016e-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([-1.2002e-03, -4.9105e-01,  7.1423e-05,  7.8703e-05,  1.5836e-04,
        -1.1215e-03, -4.4940e-04, -6.5413e-04, -1.1552e-01, -4.5988e-04,
         2.8573e-04,  3.4769e-04, -3.2499e-01,  1.8225e-04, -4.5430e-04,
        -6.5707e-05,  2.4475e-05, -6.1249e-02,  3.9629e-04, -4.7509e-04,
         5.1285e-04,  2.5725e-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([ 2.4460e-03, -3.1986e-01, -3.4837e-05,  1.3741e-03, -3.1777e-04,
        -1.9396e-04, -2.4583e-04, -1.2091e-01, -2.9569e-04, -9.2978e-06,
        -1.9564e-01,  2.2104e-03, -2.3375e-04, -1.5179e-04, -5.2276e-02,
         2.6075e-04, -1.2223e-03,  9.1657e-05,  3.0332e-05, -3.0191e-01,
        -2.1949e-04,  6.3339e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1247e-03, -3.1480e-01, -2.0083e-04,  9.8317e-03,  9.5715e-05,
        -1.2430e-05, -1.0207e-03, -1.6144e-01,  6.2680e-04,  1.7185e-04,
        -1.6111e-01, -7.3660e-04,  2.4058e-04, -7.7675e-05,  2.0208e-01,
         1.5822e-05, -3.0387e-03, -1.1917e-03,  1.4019e-01, -3.8702e-04,
         2.6108e-04, -3.3607e-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([-7.2639e-04, -4.7822e-01,  3.4505e-04,  1.9605e-02,  6.7944e-05,
        -1.1338e-04, -3.8407e-04, -2.5389e-01,  4.5985e-04,  2.3087e-04,
        -8.5981e-02,  3.0057e-04,  9.7475e-05,  2.5208e-04,  1.5528e-01,
        -2.2151e-04, -2.3531e-03,  1.9932e-04, -1.0767e-03,  1.9953e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.1968e-04, -6.0552e-01, -4.9184e-04, -1.8401e-04,  1.5133e-03,
        -5.8766e-05,  1.0329e-05, -1.7701e-01,  4.5182e-03,  1.4225e-05,
        -2.1297e-04, -3.3441e-02, -2.3907e-04,  4.0466e-04,  3.0124e-04,
         6.9329e-04, -1.7382e-01,  2.4261e-04,  2.8243e-04,  3.1412e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  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 #2350: [tensor([-9.6258e-04, -2.3143e-01,  1.3263e-04, -4.5673e-02, -1.7721e-01,
        -2.4836e-04,  4.6752e-05,  3.0985e-04,  1.0088e-04, -1.2555e-04,
        -1.1245e-01,  1.1610e-04,  1.9635e-02, -7.8731e-02,  3.7788e-04,
         6.4676e-05,  1.3058e-04, -1.7912e-01,  2.1864e-04,  4.8146e-03,
        -7.5699e-05, -5.6877e-05,  7.1343e-05,  1.7760e-04, -6.1444e-05,
         2.5894e-02,  4.6925e-05,  5.0862e-05, -1.2119e-01, -1.7810e-04,
         1.8041e-04,  1.1897e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3377e-03,  3.1412e-01, -3.7479e-05,  9.5425e-02,  6.0700e-02,
         1.9201e-04,  4.2068e-05, -1.4242e-03, -6.6670e-05, -6.6805e-06,
         1.3454e-01, -4.1382e-05,  6.2707e-02,  1.7493e-01,  1.5441e-04,
        -9.6608e-05,  7.0803e-05,  8.0292e-02, -1.5341e-04, -1.3885e-02,
        -1.9888e-05, -7.0945e-05, -2.4395e-05, -6.9657e-05, -8.2728e-06,
        -2.3204e-02, -1.6478e-04, -3.1084e-05, -3.5451e-02, -1.4719e-04,
        -1.5182e-04, -2.3274e-04, -6.4528e-05, -1.3800e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.0778e-04, -2.8938e-01, -4.7938e-05, -1.2825e-01, -5.3814e-02,
        -1.5453e-04,  9.6056e-05,  8.3093e-04, -2.3388e-05, -1.2185e-04,
        -1.7508e-01,  3.8245e-06,  3.8862e-02, -7.2849e-02, -4.7383e-05,
        -6.7459e-06, -9.0228e-05, -8.0712e-02,  3.9436e-04,  9.7698e-04,
         2.1011e-04,  2.3669e-05,  2.4099e-04, -2.0798e-04,  1.4184e-04,
         2.4340e-02, -3.0068e-06,  4.5374e-06, -2.7429e-04,  2.1424e-05,
        -4.0112e-05, -1.6333e-04, -1.3170e-01,  2.7089e-05,  1.6761e-05,
        -1.4021e-04, -4.2650e-06], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3208e-05, -7.9571e-02, -3.6414e-05,  1.4989e-02, -3.6046e-02,
        -1.4439e-04, -2.1888e-05,  1.9146e-03,  6.3355e-05, -8.7012e-05,
        -1.6692e-01,  9.7926e-05, -2.8093e-02, -1.8322e-01, -1.0892e-04,
        -1.3305e-04,  1.6121e-05, -2.5992e-01,  3.6193e-04,  1.4055e-03,
         8.0622e-05, -7.9072e-05,  6.0026e-05, -2.2530e-01, -1.8052e-05,
         1.3705e-04,  2.2855e-04, -1.2745e-04,  1.7067e-04,  1.2012e-04,
        -1.0577e-04,  7.3328e-05,  1.8763e-04,  1.0483e-05, -8.8287e-05,
         3.8034e-05,  7.6131e-06], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2507e-03,  5.4562e-01, -1.2697e-04,  5.6739e-05,  3.4598e-04,
         1.9291e-04, -5.5732e-05,  3.3344e-04,  5.9900e-05, -1.4453e-04,
         2.8114e-01,  2.1002e-04,  4.0756e-04,  2.0279e-04, -4.1615e-04,
         1.6454e-01, -1.9052e-04, -2.9413e-03, -4.6178e-04, -1.3311e-04,
         1.3226e-04, -1.9547e-05, -1.9676e-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([ 6.8864e-04, -6.3247e-01,  1.5547e-04,  2.6040e-04, -5.6830e-05,
        -3.1740e-05,  1.4188e-04, -1.1199e-05, -1.3902e-04,  5.3852e-05,
        -1.8731e-02, -2.3344e-04, -7.6077e-05, -3.0564e-04,  3.7867e-06,
        -3.4548e-01,  2.0862e-04,  1.8396e-04,  3.2676e-04, -7.0628e-06,
        -2.1501e-04,  8.9927e-05, -1.3229e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2876e-04,  5.1647e-01,  6.4757e-05, -1.4770e-05,  3.8308e-04,
         3.2481e-04, -1.1741e-04,  1.4659e-04, -3.4868e-04, -4.0616e-04,
         4.1622e-02, -2.2024e-04,  2.1041e-04,  2.8277e-05, -2.5879e-04,
         4.3673e-01,  2.7621e-04, -1.2850e-03,  5.2779e-04, -9.1059e-05,
         4.2612e-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([ 6.1830e-04, -3.8272e-01, -2.7949e-04, -3.5703e-05, -2.3596e-05,
        -2.3934e-04, -1.4052e-04, -5.7370e-04, -3.5171e-04, -1.4059e-04,
        -4.0789e-02,  1.7993e-04, -3.9700e-04, -3.3321e-04,  5.1029e-05,
        -1.8664e-05,  3.0761e-04,  4.9121e-04,  2.4750e-04, -5.7156e-01,
         1.8392e-04,  9.9259e-06,  3.1026e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0417e-04, -6.6919e-01, -1.3761e-04,  1.2984e-04, -2.3961e-05,
        -2.5591e-04,  2.1889e-04, -3.1519e-04,  3.8351e-04, -1.7827e-05,
        -7.3417e-02, -2.7894e-04, -4.0012e-04,  1.5331e-04,  7.0057e-05,
         5.9042e-05, -2.3755e-04,  4.2472e-04, -2.5328e-01, -3.2785e-04,
         3.3224e-04,  1.4455e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4370e-03,  3.3276e-01,  2.3742e-04,  1.1423e-04,  7.9609e-04,
         1.6861e-04, -2.1513e-04, -3.7855e-04,  2.7038e-04,  8.2717e-05,
         1.9833e-01,  2.1146e-04, -2.9614e-04, -1.1461e-04, -1.8156e-04,
         4.5967e-01, -8.8183e-05, -3.2886e-03, -6.8269e-05, -7.1327e-04,
        -1.6414e-04, -4.2238e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7171e-05,  3.2954e-01,  2.2845e-04, -2.4982e-04,  4.6677e-04,
         1.6084e-04, -3.1046e-04,  7.2990e-04,  8.7527e-04, -2.6807e-04,
         2.8631e-01,  2.9317e-04,  3.4979e-04,  1.2528e-04, -8.5083e-05,
        -6.9239e-02,  2.7572e-04,  1.9141e-04, -7.2325e-05, -3.2702e-04,
         3.0970e-01,  1.4825e-04, -1.6232e-05, -1.7988e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9128e-03,  4.1207e-01,  2.0740e-04,  1.9773e-04,  5.1485e-04,
         2.7820e-04, -1.8286e-04, -3.4598e-05,  2.4687e-04,  3.1766e-04,
         1.7033e-01, -1.1955e-04, -1.5546e-04, -2.2832e-04, -9.3995e-05,
        -1.2142e-01,  1.7018e-04,  1.4150e-04,  1.4936e-04,  2.2589e-04,
         3.3636e-04, -1.4062e-04, -1.7732e-04,  1.9852e-04,  2.2350e-04,
         2.0618e-04,  5.7143e-04,  2.8831e-01,  1.7450e-04,  7.7462e-05,
         8.0999e-05,  1.6256e-04, -2.1144e-04,  1.3218e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #2400: [tensor([ 4.1900e-04,  5.3359e-02, -3.8884e-04, -5.8862e-04, -9.0401e-05,
         1.4175e-05, -3.6563e-04,  9.1625e-03, -2.6743e-01, -8.6484e-04,
         3.4826e-04, -3.7165e-01, -5.4476e-04, -1.1022e-04, -6.7619e-04,
        -9.8380e-02, -1.3560e-03, -5.0788e-04, -4.0081e-04, -1.9116e-01,
        -1.0549e-04, -2.8817e-04, -6.0991e-05,  1.2774e-03, -3.1072e-04,
        -1.5098e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.9885e-03, -4.5432e-04, -8.9869e-04, -1.2959e-03, -4.7269e-04,
         4.7952e-05, -1.1177e-03,  2.2009e-02, -1.7429e-01, -1.2424e-03,
        -2.7435e-04, -5.4062e-01, -7.2500e-04, -2.6852e-04, -1.3894e-04,
        -2.4354e-01, -1.5885e-03, -7.6074e-04,  1.8619e-04, -4.2410e-04,
         6.2580e-05, -1.1932e-04, -1.6556e-04, -4.3044e-04, -8.8571e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6562e-03,  6.4269e-03,  1.6895e-03, -7.4446e-04, -1.1238e-03,
        -4.6896e-04, -1.4318e-03,  1.1824e-02, -2.3646e-01, -4.6749e-04,
         2.9824e-04, -2.9992e-01, -1.3749e-04, -7.4573e-04, -5.7501e-04,
        -1.2836e-01, -1.3027e-03,  5.4834e-04,  6.3320e-04, -3.0019e-01,
         3.0612e-04,  7.5310e-05, -1.2544e-03,  3.0324e-04, -7.7145e-04,
         1.2835e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5944e-03, -1.5360e-03, -1.0095e-03, -6.4788e-04, -4.3928e-04,
        -6.2294e-04,  4.8763e-04,  4.8800e-02, -1.2294e-01, -5.3595e-04,
        -3.9264e-04, -4.3994e-01, -8.2816e-04, -2.3189e-04,  2.8762e-04,
        -6.8651e-02,  2.0876e-04, -5.2459e-04, -3.2138e-04, -3.0894e-01,
        -7.0290e-04, -8.5834e-06,  4.2671e-05, -2.9930e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.6344e-03, -3.7739e-04,  1.6144e-04, -3.8309e-04, -2.4107e-04,
         2.3993e-04, -3.8902e-04,  1.5403e-03, -1.2509e-01, -4.1595e-04,
        -9.2809e-05, -5.9325e-01, -7.3257e-04, -1.0162e-04,  1.1090e-04,
        -8.6567e-02, -3.0916e-04,  8.0510e-02,  7.4154e-05,  4.5885e-04,
        -3.0807e-04,  8.4485e-05,  1.0028e-01, -1.3765e-03, -1.5139e-04,
        -8.5477e-05,  3.9219e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.1607e-03,  1.2027e-02, -4.9078e-04, -5.2420e-04, -7.7667e-04,
        -3.1821e-04, -6.3740e-04,  3.0017e-02, -2.4211e-01, -9.4516e-04,
         2.4709e-04, -4.6865e-01, -1.0684e-03, -5.0979e-04, -5.3897e-04,
        -1.0777e-01, -2.2587e-04,  1.2519e-01, -9.0113e-04, -1.4385e-03,
         2.0989e-04, -2.3485e-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([ 5.6435e-03, -8.5763e-04, -4.2704e-05, -6.1673e-04,  1.9286e-04,
        -2.8143e-04,  7.8759e-05,  7.3076e-03, -1.3148e-01, -6.9612e-04,
         1.5216e-04, -3.5662e-01, -1.6316e-04, -1.7809e-04,  2.5354e-05,
        -3.4574e-02, -4.8419e-04, -4.9055e-04, -4.5678e-01,  3.2491e-04,
        -1.1656e-03,  9.4209e-04, -3.2164e-04, -3.4591e-04, -2.3657e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.7955e-03, -2.3955e-03, -5.9383e-04, -8.7778e-04, -5.5203e-04,
        -5.7094e-04, -3.0970e-04,  2.4656e-02, -2.5419e-01, -1.6317e-03,
         2.8332e-04, -5.1735e-01, -5.5278e-04, -1.0641e-04, -8.0459e-05,
        -1.3115e-01, -1.4151e-03, -6.4895e-04, -2.8437e-04, -1.5164e-04,
         5.6258e-04, -1.1986e-03,  4.0940e-02,  9.3882e-03, -4.4645e-04,
        -1.6303e-03, -1.4378e-04,  9.0918e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.5806e-04, -9.0211e-04, -5.4008e-04, -1.9109e-04, -3.1410e-04,
        -6.9031e-04, -5.2453e-04,  9.2874e-02, -7.8729e-02, -4.9386e-04,
         1.8692e-05, -1.8741e-01, -3.9232e-04, -3.6855e-04, -3.6559e-04,
         3.8324e-02, -1.7700e-03, -6.7350e-04,  1.6238e-03, -2.6110e-01,
        -1.6317e-04, -2.1166e-04, -7.7397e-02, -2.8077e-04,  2.2964e-04,
        -2.5295e-01, -4.0120e-04, -5.5078e-04,  5.9417e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.8892e-03,  8.9085e-03,  1.2834e-04, -4.4594e-04, -3.9576e-04,
        -2.8838e-04, -3.7270e-04,  1.1376e-02, -2.4449e-01, -5.4840e-04,
        -3.2730e-04, -3.0969e-01, -2.2386e-04, -2.9402e-04, -5.2475e-04,
        -6.4479e-02, -3.5897e-03, -3.9780e-04,  5.6179e-02, -3.8878e-04,
        -2.9149e-01, -9.8679e-05,  8.5844e-06, -6.5256e-05, -4.0172e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3732e-03,  1.0035e-02,  9.0898e-03, -8.2614e-04, -8.1138e-04,
        -2.9654e-04, -6.8933e-04,  6.5768e-02, -2.5529e-01, -9.0738e-04,
         1.6638e-04, -4.0717e-01, -8.2156e-04, -3.3763e-04, -1.9632e-04,
        -9.0186e-02, -6.9907e-04, -5.2806e-04, -1.9771e-05, -3.7527e-04,
         8.5574e-05, -6.7426e-04, -2.2312e-03, -9.5882e-04,  1.4765e-01,
         2.8316e-04, -1.0091e-03,  9.2370e-05,  4.3026e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4252e-04,  6.7052e-04, -1.1203e-03, -3.0049e-04, -2.9934e-04,
        -6.0637e-05, -2.5817e-04,  1.5843e-02, -1.4748e-01, -4.9355e-04,
         3.1568e-04, -2.2015e-01,  4.3493e-05,  2.1636e-05,  9.8078e-05,
        -1.1101e-03, -1.2401e-03, -1.5982e-04,  6.2180e-03, -4.1675e-01,
        -4.3196e-04, -1.8603e-01, -3.5923e-04, -3.0751e-05, -6.6654e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #2450: [tensor([ 6.7962e-04, -5.7035e-01,  3.7902e-05,  3.2740e-04,  2.7490e-05,
        -4.6699e-04,  3.4097e-04, -3.5913e-01,  2.6402e-04,  6.5851e-02,
        -3.1830e-06,  1.6637e-04,  6.0343e-06, -3.1169e-04, -6.6297e-04,
         9.4721e-05,  6.5465e-05, -4.4885e-04, -4.6456e-04, -2.9705e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0036e-03, -3.8029e-01,  1.5221e-04,  5.3722e-04,  7.9176e-05,
        -1.1039e-04, -1.2709e-04, -2.4670e-01,  4.9646e-04,  2.4119e-02,
         1.3239e-04, -1.6236e-05, -5.0549e-05, -1.5235e-05, -1.5782e-01,
        -4.2901e-05,  2.3371e-05,  2.4839e-03,  3.8703e-05, -5.6200e-04,
        -2.2955e-04, -1.8461e-01,  1.6261e-04,  2.1529e-05, -1.7651e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1246e-03,  2.4972e-04,  2.3486e-05,  1.6818e-04, -2.0848e-01,
        -2.9295e-05,  2.4128e-05,  4.7104e-04, -9.7527e-02,  1.3851e-04,
         4.0059e-05,  6.1245e-05,  5.3318e-05, -1.2108e-04, -3.7878e-01,
        -2.5676e-04, -1.5840e-04,  3.3787e-04, -4.0580e-05,  6.1510e-06,
        -2.1492e-05,  2.3818e-04,  1.7707e-03, -1.2415e-04, -9.4898e-05,
         5.9713e-05, -6.4488e-05, -1.5149e-04, -6.4301e-02, -1.1429e-04,
         2.3844e-04,  1.6122e-04, -2.4313e-01, -3.4154e-04, -2.3901e-06,
        -1.0275e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3250e-03, -8.6881e-05, -1.2174e-04,  4.5151e-04,  7.3277e-02,
        -1.9032e-04,  3.0392e-04,  5.8448e-04, -2.9341e-01,  3.0125e-04,
         2.0837e-04, -1.9346e-04, -6.4009e-05,  4.8370e-04, -3.2209e-01,
         1.8125e-04,  4.3650e-05, -3.3157e-05, -3.4412e-04,  1.3559e-04,
         2.0274e-05, -2.3600e-04,  1.5232e-03,  9.1519e-05,  1.1274e-04,
         2.6609e-04, -3.1023e-04, -2.9011e-01, -3.6149e-04,  5.1876e-03,
        -4.2892e-04,  6.7610e-03,  2.8473e-04,  2.8898e-04,  1.3356e-04,
         6.3272e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.8768e-04,  3.3052e-04,  2.7019e-05, -2.7119e-04, -3.3302e-01,
        -1.4086e-04,  1.7712e-04,  2.5066e-04, -1.2229e-01,  8.3880e-05,
         3.3451e-05,  2.6096e-04,  8.6172e-05,  6.4226e-04, -2.4274e-01,
         5.8298e-04,  2.3597e-05,  1.1713e-04, -1.2853e-04, -7.3938e-05,
         2.8611e-05, -1.3118e-04,  1.6735e-03,  3.4923e-04, -2.1590e-04,
         3.0276e-04, -8.6479e-05,  9.6721e-05, -3.7947e-04, -1.1906e-01,
        -1.8138e-04, -5.3465e-05,  2.9783e-02, -1.4539e-01, -1.6638e-04,
         3.1581e-04, -1.1012e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3345e-03, -2.4920e-01,  4.7445e-05, -2.6028e-05, -1.0708e-01,
         1.2885e-04, -1.6538e-05,  1.7280e-04, -1.5196e-04, -7.6288e-05,
         9.1529e-04,  2.1709e-05,  3.2658e-02, -1.7407e-01,  9.7398e-05,
         1.1444e-04,  4.4794e-05,  2.6877e-04,  1.0874e-04, -2.5841e-01,
         1.4190e-04,  8.6085e-04, -1.2997e-04, -1.1325e-04, -8.4158e-05,
        -5.9346e-07, -1.5748e-01,  9.7132e-05,  1.4376e-02,  4.0615e-05,
        -1.2631e-04, -3.2007e-05,  1.6670e-04,  2.1661e-04, -1.5998e-04,
         3.0250e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.1739e-04, -3.4687e-01, -1.8992e-04, -6.1095e-05,  3.9074e-02,
         8.3533e-05, -1.5100e-05,  2.3724e-05, -1.7133e-04, -1.5849e-04,
         3.7641e-04,  1.9709e-04,  2.6125e-03, -1.4550e-01,  1.9003e-04,
        -4.5757e-05,  7.0271e-05, -3.0721e-05,  5.6793e-04, -2.2517e-01,
        -1.9386e-04,  8.0894e-05,  2.7147e-04, -8.2780e-05,  5.0023e-05,
        -2.0797e-04, -1.6021e-01,  1.0785e-04,  1.9372e-02,  3.3502e-05,
        -5.7095e-02, -6.4052e-05,  1.1977e-04, -3.9348e-05, -1.4219e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.5064e-04, -7.3811e-02,  1.2452e-04,  8.1770e-05,  3.1993e-02,
         1.4390e-04, -2.1413e-05,  1.6233e-04, -1.3492e-04, -4.7198e-05,
         3.1803e-04,  1.3991e-04,  3.7835e-02, -2.8855e-01,  1.1855e-05,
         2.2602e-04,  2.1496e-05, -6.7797e-07,  3.8418e-04, -3.0794e-01,
         4.7211e-05,  2.0017e-03,  2.2691e-04, -2.3815e-04, -8.2974e-05,
        -2.9860e-05, -2.4382e-01, -7.7688e-07,  1.0230e-02,  4.3659e-05,
         1.3978e-04, -1.3637e-04, -1.3444e-04, -3.0614e-05,  2.0669e-04,
        -9.2374e-05, -1.9587e-05,  1.5997e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.9378e-04, -1.3295e-04,  3.9628e-04, -2.6250e-01, -1.5119e-04,
         9.2160e-05, -2.8845e-04,  8.6990e-05, -4.3502e-05,  3.6941e-02,
         4.2691e-06, -1.2940e-04, -2.6160e-04, -7.9017e-05, -2.7133e-01,
        -2.8175e-04,  4.9170e-05,  2.9537e-06, -3.5573e-04,  1.0503e-04,
        -4.2441e-01,  1.1614e-04, -2.0424e-04,  5.1623e-05,  1.6039e-04,
         2.8851e-04,  2.9528e-04,  1.4239e-04, -4.3811e-05,  2.3681e-05,
         3.3577e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5673e-03,  9.6643e-07,  2.0465e-04,  8.8088e-02,  6.6122e-06,
         1.5478e-05, -7.4699e-05, -3.8040e-05, -8.1936e-05, -4.3152e-01,
         8.3668e-05, -7.8394e-05, -1.4402e-04,  2.7429e-04,  5.9190e-02,
         1.2025e-05, -1.3497e-04, -1.3401e-04, -1.3882e-04, -2.1245e-04,
        -2.7307e-01, -2.8104e-04, -1.1169e-04, -1.3040e-04, -1.6500e-04,
         1.9561e-04, -1.8001e-04, -1.3814e-01, -8.3994e-05,  2.4012e-03,
         1.0852e-03,  4.6361e-05,  4.8491e-05, -5.6982e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.2500e-04, -6.8972e-05, -1.9854e-04, -3.4518e-01, -5.6912e-05,
        -6.7526e-05, -1.1254e-05,  9.8638e-05, -9.7244e-06, -2.0992e-02,
         1.1960e-05, -1.2429e-05,  8.9787e-05,  1.6674e-04,  2.3827e-01,
         3.1574e-05,  4.1819e-05,  2.4795e-05, -2.3808e-04,  1.8471e-04,
        -1.9747e-01, -1.2267e-04, -7.2089e-05, -1.3790e-04, -1.9400e-01,
        -1.3778e-04, -9.0523e-05, -4.9828e-05, -1.7927e-04, -1.8505e-04,
         7.1903e-05, -1.6263e-04,  3.1784e-04, -5.6868e-05,  4.7285e-05,
        -7.3648e-05,  3.1821e-04,  1.1891e-04, -1.0502e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5495e-03, -1.0714e-01, -1.5279e-04,  1.1808e-04,  8.7151e-05,
         3.0948e-03,  1.2133e-04,  2.2725e-04, -7.7780e-05,  5.8852e-05,
        -7.4410e-05, -1.0793e-04, -1.2198e-04,  6.0137e-05, -2.2579e-04,
        -2.9640e-05, -9.3388e-05, -3.7812e-01,  8.3997e-05,  1.7728e-04,
         7.4179e-05, -2.1468e-05, -5.8291e-05, -2.7799e-01,  2.4447e-05,
        -1.4195e-04,  1.8491e-04,  1.7423e-04,  1.8652e-02,  1.0919e-04,
         1.8445e-04, -2.1000e-01, -2.5465e-05, -1.8234e-04,  1.3608e-04,
         9.5608e-05, -2.2144e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #2500: [tensor([-3.4430e-03,  6.7576e-04, -5.0905e-05, -1.8974e-04,  2.0449e-03,
         7.6082e-02, -6.4057e-04,  2.5083e-04,  6.1355e-05,  2.7982e-04,
        -1.8210e-04,  6.9813e-06, -6.6630e-01,  5.5905e-04, -1.5414e-04,
        -3.4507e-04,  6.4849e-04, -2.4740e-01, -4.8728e-04, -8.3419e-05,
         1.1442e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1941e-03,  2.1689e-04, -4.1525e-04,  8.1595e-05,  6.3588e-03,
         8.2360e-02,  4.2290e-05,  1.4123e-04,  3.3815e-04, -4.6383e-04,
        -2.1927e-04,  1.9853e-05, -3.4727e-01,  1.1689e-04,  7.0066e-05,
        -6.3500e-04,  1.2663e-01, -2.5160e-04, -2.0960e-05, -9.8336e-04,
         8.0302e-03, -4.2194e-01,  3.8203e-04, -4.5592e-04,  3.6538e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5027e-03, -1.7036e-01, -5.6319e-05,  5.4894e-05, -3.3616e-02,
        -5.6844e-05,  2.6735e-04,  1.3789e-04, -1.4525e-04,  6.5808e-04,
         4.1318e-04, -7.0946e-05,  2.5681e-04, -8.0267e-05,  1.3195e-04,
        -8.9882e-05,  1.1201e-03, -3.8914e-01,  2.0787e-04, -2.2963e-01,
         1.0556e-04,  3.9482e-03,  1.4769e-04, -9.6069e-05, -2.7161e-04,
        -1.5375e-01,  1.3487e-04,  1.3031e-02, -7.7117e-06, -3.9681e-04,
         7.8185e-05, -3.6834e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3221e-03, -1.2660e-01, -4.5613e-05, -3.4407e-04, -3.8492e-01,
        -2.1708e-04, -6.2913e-05,  1.0376e-04, -5.0072e-05,  2.7519e-04,
         2.0827e-04,  1.6235e-05, -1.7044e-04, -6.5402e-05, -8.0460e-05,
         8.8862e-05,  1.2474e-02, -3.3866e-01,  1.3767e-04,  1.4706e-02,
         5.5650e-05,  3.6762e-04, -3.5904e-05, -4.5978e-05,  5.5826e-06,
        -7.5470e-02,  2.2037e-04,  2.3595e-02, -1.2760e-04,  1.9180e-02,
         3.1502e-04, -1.3837e-05, -1.1566e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4778e-03, -4.8779e-01, -1.0574e-04, -6.3994e-06,  2.6254e-02,
        -2.5404e-04,  1.9508e-04, -2.3147e-05,  1.3798e-04,  9.3146e-04,
         9.0575e-05, -1.2136e-04, -5.6238e-06, -8.6382e-05, -2.7890e-04,
         9.6597e-06,  7.5343e-04, -1.1499e-01, -8.5080e-05,  2.8771e-02,
         1.4205e-04,  1.2508e-03,  2.4516e-04, -5.2371e-05, -5.3551e-05,
        -1.2569e-01, -1.6465e-04,  2.1379e-03,  5.3194e-06, -2.0758e-01,
         2.4206e-04, -2.3910e-05, -4.9992e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.3438e-03,  1.8390e-02, -2.0432e-04, -9.9996e-05,  5.8692e-02,
        -1.6716e-04, -2.9107e-04, -2.2851e-04,  6.7559e-05,  6.0785e-05,
        -4.7741e-04, -1.2571e-05,  1.5512e-04, -1.4122e-05, -1.2245e-04,
        -2.6534e-01, -9.4127e-05, -2.9256e-04, -1.4576e-04, -1.0738e-05,
        -4.2524e-01, -1.0434e-04,  4.5852e-02, -1.8913e-04,  4.8456e-03,
         2.0772e-04, -1.7467e-01, -3.3558e-04, -1.2122e-04,  2.1607e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5737e-03, -3.4470e-01, -1.2156e-05,  3.8234e-05,  4.0689e-02,
        -2.5699e-04,  4.0893e-05,  1.8946e-04, -2.7561e-05,  1.0098e-05,
         2.1292e-04,  2.2679e-04,  1.7936e-03, -1.0818e-04,  3.0995e-04,
        -1.3037e-01, -3.3143e-04, -1.9673e-04,  1.6882e-04, -1.5328e-07,
        -1.6307e-01,  1.3250e-04, -3.1438e-04,  3.2233e-04,  2.0856e-02,
         8.6729e-05, -2.9315e-01, -2.4384e-04, -2.1407e-04,  4.3875e-05,
         3.1123e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3663e-03, -4.6826e-01,  3.8508e-04,  1.9163e-04, -3.1601e-01,
        -6.6546e-05, -1.0178e-04, -1.6576e-04, -1.5425e-04, -3.8472e-05,
        -1.5831e-04,  3.2051e-04,  9.6533e-04, -7.1312e-05,  1.7873e-04,
        -1.1334e-01, -2.1307e-04, -1.2628e-04,  9.8061e-05, -5.4353e-05,
        -5.5259e-02,  2.9532e-04,  4.1617e-02,  1.7049e-05,  1.2559e-04,
         1.3903e-04, -8.1426e-05,  1.3103e-04, -6.0535e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3885e-04,  4.1574e-01,  6.5273e-06,  4.0758e-04,  5.6777e-01,
        -9.3754e-05, -4.0891e-04, -1.9008e-04,  9.7211e-05,  2.6074e-04,
        -4.7894e-04, -1.9142e-04, -4.8404e-03,  1.0771e-04, -8.0856e-06,
         2.5001e-06, -6.1110e-04,  2.5376e-04, -5.9726e-05, -1.1269e-04,
        -1.3290e-03, -1.5069e-04,  1.3146e-04, -5.2695e-03,  1.4543e-04,
         4.1464e-04, -4.1287e-04, -6.5303e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.1433e-04, -7.0169e-01,  2.3596e-04, -3.4052e-04, -2.9173e-02,
        -2.6047e-05,  6.3188e-04,  1.6849e-04, -5.3453e-05, -3.9486e-04,
        -1.7538e-04,  5.3979e-04,  1.3299e-03, -5.2098e-05,  2.9503e-04,
         3.4003e-05,  8.0555e-04, -1.0504e-04, -1.9555e-04, -2.6757e-04,
        -2.5355e-01,  5.0080e-04,  7.9868e-03,  1.9459e-04,  3.1947e-04,
         2.8724e-04, -6.6915e-06,  2.9374e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.5027e-04,  4.2088e-01,  1.0432e-04,  6.5209e-04,  2.9895e-01,
        -9.7928e-05,  3.4367e-05, -3.9027e-04, -4.3302e-05,  3.4117e-04,
        -7.4966e-04, -2.3358e-04, -2.7098e-03, -1.0619e-04,  3.9194e-04,
         2.2280e-05, -8.3021e-04, -2.2465e-04,  2.0635e-04,  1.7818e-04,
         2.3705e-01, -3.4788e-05, -3.4510e-02, -8.8501e-05,  1.5029e-04,
         3.6798e-04,  5.0138e-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([ 1.2178e-02, -4.0908e-04, -1.7241e-03,  7.2901e-04,  1.3047e-03,
        -3.2677e-04, -1.0260e-03, -1.5175e-03, -6.7202e-04, -3.5491e-03,
         2.1208e-01,  2.1850e-04, -5.5312e-04,  4.9476e-04,  1.6574e-04,
        -5.2512e-01, -1.2181e-03,  2.3540e-01,  1.7922e-04,  8.3903e-05,
        -6.1446e-04,  4.3183e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  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 #2550: [tensor([-2.3211e-03, -5.9593e-01,  1.8778e-04,  1.4097e-03,  8.3787e-04,
         6.9520e-05, -3.5015e-04, -2.2391e-04, -2.9820e-01, -1.5857e-04,
         9.7690e-02,  1.6768e-03,  6.0144e-04,  7.6198e-05,  2.2305e-04,
         3.6792e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7485e-03, -4.7957e-01, -1.0124e-04,  1.4022e-04,  1.3422e-04,
        -5.8181e-04,  2.9097e-05,  7.1114e-05, -1.1942e-01,  2.8510e-05,
         2.8388e-03, -7.1560e-04, -2.6884e-04, -3.9122e-01,  3.2066e-04,
         4.0444e-04, -4.1238e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.9092e-06, -7.1153e-01, -1.4384e-04,  6.7383e-03, -3.5238e-04,
        -2.7932e-04,  6.4155e-04, -2.4702e-04,  1.0104e-04, -3.5922e-04,
        -2.4469e-01, -4.8939e-04,  3.3527e-02, -3.9774e-04, -4.9939e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4436e-03, -3.6725e-01, -1.2566e-04,  1.4467e-03,  4.0388e-04,
        -5.8669e-04,  1.3442e-04, -1.7542e-04, -1.8415e-05,  6.0601e-04,
        -1.5600e-01,  2.1971e-04, -4.5351e-04, -3.8001e-04, -4.7028e-01,
         1.9362e-04, -2.8362e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6732e-03, -5.7712e-01, -5.9707e-04,  3.8879e-03, -5.7775e-04,
        -2.0380e-04,  4.6451e-04,  5.3223e-04,  2.4514e-05,  6.0263e-05,
        -4.0212e-01, -9.8739e-05,  9.0097e-03, -2.0604e-04,  3.6475e-04,
         6.2535e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2843e-04, -3.1360e-01,  7.9068e-05, -6.7104e-05, -2.3421e-04,
         2.4632e-02,  2.9074e-05,  7.1919e-06,  1.2260e-04,  6.1490e-05,
         1.6649e-04,  1.1492e-04,  5.8111e-05,  1.3891e-04, -8.2547e-05,
        -8.3833e-05, -5.4797e-05,  1.0060e-04, -1.6823e-01,  4.5434e-05,
        -5.3290e-05,  1.1599e-04,  7.3035e-05,  5.1136e-05, -3.6537e-05,
        -1.4199e-04, -1.7060e-04,  4.9553e-05, -2.0418e-01,  8.8012e-05,
         2.1132e-04,  8.5413e-03, -1.2075e-05, -1.0430e-05,  1.5702e-06,
        -1.4182e-05, -2.2625e-06,  7.3388e-06, -8.5593e-02,  3.7691e-05,
        -4.8087e-02, -5.8543e-05, -4.4666e-05, -1.4391e-01,  1.7674e-05,
        -2.3587e-05,  1.3889e-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([ 2.1247e-04, -3.2477e-01,  6.3064e-05, -1.2870e-04,  1.5280e-05,
         9.6236e-04,  1.0619e-04,  1.6175e-04, -9.5688e-05,  9.1144e-06,
         3.2374e-04, -4.1840e-05,  1.1164e-04,  2.3335e-04, -4.6676e-05,
        -2.9075e-05,  1.1956e-04,  7.6609e-05, -1.8222e-01,  8.9034e-05,
         3.3965e-04,  6.9470e-05,  5.8510e-05,  1.5628e-04,  1.9088e-05,
         1.9784e-05,  1.2579e-04,  1.6045e-04, -1.8237e-01, -1.9051e-05,
         2.2417e-05, -3.8577e-02, -4.1413e-05,  3.7859e-04,  3.8706e-05,
        -1.3129e-04,  1.0363e-04, -4.6291e-05, -1.3844e-01, -1.7126e-05,
        -4.3511e-02,  1.0813e-04, -3.9813e-03, -8.0606e-02,  1.9396e-04,
         3.2587e-04,  1.9382e-04,  5.4525e-05,  3.5525e-05,  2.1889e-06,
        -2.8275e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.8083e-04, -4.5811e-01,  8.7982e-05,  1.0647e-04, -9.5995e-05,
         7.3447e-03,  5.8378e-05,  6.9743e-05,  1.3126e-05,  7.3656e-05,
         2.0288e-04, -5.0519e-05,  8.0575e-05,  5.7741e-04, -7.9426e-06,
        -8.5519e-05,  1.7436e-05, -2.4295e-05, -2.0851e-01, -7.8389e-06,
         1.9449e-04,  8.6572e-05,  6.3419e-06,  1.6306e-04,  1.0368e-04,
         9.7968e-05, -1.6211e-05, -9.2386e-06, -3.1175e-04,  3.4357e-05,
         1.3201e-04, -3.5549e-02, -1.4249e-04,  4.4732e-04,  1.1880e-05,
         1.3536e-04,  1.2050e-04, -2.4250e-05, -1.2924e-01, -6.6153e-05,
         1.4456e-02, -5.8459e-06, -1.5141e-04, -1.3686e-01, -2.3896e-05,
         5.5484e-05, -8.3781e-05,  5.1041e-03, -8.0140e-05, -1.8260e-04,
        -5.0164e-05,  8.3103e-05, -6.9975e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9350e-03, -1.2810e-01, -2.4412e-05,  2.5512e-03,  1.5413e-06,
         1.9133e-03, -3.9799e-01,  5.7336e-04, -4.7422e-04, -2.6529e-04,
         2.0288e-04,  1.7005e-04,  2.0506e-05, -1.7729e-01, -2.6989e-04,
         2.2868e-02, -1.8880e-04, -2.6430e-04, -2.6379e-01,  5.1905e-04,
         5.2306e-05,  3.8717e-04, -1.4569e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3043e-04, -1.0176e-01,  9.0743e-05,  5.1258e-04,  8.9849e-05,
         3.9927e-03, -3.3430e-01,  3.2869e-04, -2.9634e-04, -3.2733e-04,
         4.9664e-05, -1.0269e-04,  1.3954e-04, -9.5235e-02, -3.4020e-05,
        -4.8629e-02,  1.3449e-05, -1.8512e-01, -2.8022e-04, -8.8366e-06,
        -2.2739e-01,  4.9368e-04, -2.2185e-04, -4.1121e-04,  1.4691e-05,
        -2.0925e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1229e-03, -3.3761e-01, -2.3856e-04,  8.0013e-04,  2.1453e-04,
         2.0577e-02, -3.0990e-01,  2.4801e-04, -4.7432e-05, -1.7263e-04,
         8.7437e-05, -3.3316e-06, -4.6897e-05, -2.2814e-01,  3.0547e-04,
        -9.9164e-02,  3.1733e-04,  3.4919e-04, -4.3820e-04,  2.1063e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.6836e-04, -4.3527e-01,  2.4902e-05,  3.8040e-04,  8.0421e-06,
         4.6381e-05, -4.0671e-05,  5.0549e-05, -2.3558e-04, -1.6046e-01,
        -2.1140e-04,  1.2105e-05, -7.9102e-05, -2.7265e-01, -1.9058e-04,
        -2.0656e-04, -1.2911e-04, -1.2820e-01, -1.7846e-04,  1.2790e-04,
         4.2387e-04, -5.4044e-05,  5.8738e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #2600: [tensor([-1.9278e-03, -4.3310e-01, -1.0892e-04,  1.4593e-04,  4.5273e-05,
        -3.0221e-04, -9.8281e-05, -2.5501e-01,  1.2964e-04, -7.1376e-05,
        -1.2874e-05, -1.2015e-04,  4.2821e-05,  2.1510e-02,  2.4427e-04,
         1.4518e-04, -1.0569e-04, -2.5058e-02, -4.1905e-04, -3.7499e-04,
        -7.6781e-05, -2.6631e-05, -2.6054e-01, -2.2339e-05,  2.8951e-04,
         6.7105e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7763e-05, -3.7383e-01, -4.8323e-05, -3.8744e-05, -1.5492e-04,
        -1.6033e-04,  4.2701e-05, -2.3969e-01,  1.4429e-05, -1.5506e-04,
        -9.4683e-05, -6.1962e-06,  1.9543e-04,  3.2465e-03,  8.9244e-05,
         1.3839e-04, -4.1903e-05, -2.1799e-01,  1.8721e-06, -8.4323e-05,
        -3.4154e-05, -1.1665e-06, -1.6341e-01, -7.3595e-05, -2.8403e-04,
         1.7860e-05, -5.6898e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7448e-04, -4.0403e-01,  6.0430e-05, -2.3790e-05, -1.0844e-01,
         1.0841e-04,  2.7282e-04,  2.2881e-04, -1.4661e-05,  7.4379e-05,
        -5.3190e-05,  1.0989e-04,  9.7941e-06, -8.5076e-05,  9.8560e-06,
        -2.5350e-01,  1.9577e-05,  1.4007e-04,  8.6318e-05, -1.5018e-01,
        -1.0110e-04,  3.7278e-05,  1.1691e-04,  1.2858e-04, -8.1545e-02,
        -3.0959e-05,  1.5401e-04,  1.5576e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.9547e-04, -1.9996e-01,  1.3751e-04, -2.2618e-04, -3.4493e-01,
         4.7254e-05,  1.4105e-04,  1.3622e-03, -1.8888e-04, -2.2226e-04,
        -2.1682e-04,  1.5263e-04,  2.0747e-04,  6.5233e-05, -1.3678e-04,
        -1.7021e-01,  1.0943e-04,  6.2268e-05,  1.2112e-04, -1.5826e-01,
         2.1983e-04,  5.9252e-04, -9.6372e-05, -1.2125e-01,  5.7231e-05,
         1.3556e-04,  6.9650e-05,  1.3113e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9666e-03, -3.9193e-01,  9.2475e-07,  1.0387e-04, -2.9236e-01,
        -2.0902e-04,  2.4469e-04,  5.1032e-03,  5.5229e-04, -3.3728e-04,
        -2.0814e-06,  1.2467e-04,  7.4064e-05,  7.6861e-05,  2.9695e-04,
        -1.2934e-01,  1.0457e-04,  1.4767e-04,  8.5856e-05, -1.5083e-01,
         1.5979e-04,  3.1001e-03, -2.3327e-04,  2.2293e-02, -5.9829e-05,
        -2.4987e-04,  1.2182e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.9972e-05, -2.2700e-01, -4.5466e-05, -1.2182e-01, -1.5127e-04,
        -1.2711e-04,  3.7286e-04,  2.6187e-04, -3.4394e-04,  1.0103e-05,
         9.6191e-02,  2.3077e-04,  2.3930e-04,  2.2816e-04, -2.0925e-05,
        -1.5326e-04, -2.5685e-01, -1.8369e-05,  1.5984e-03,  4.3289e-04,
        -2.9330e-01, -1.4735e-04,  2.9587e-04, -6.6646e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3196e-03,  4.9206e-01,  1.8188e-04, -1.8230e-02, -9.4527e-06,
         4.3387e-05, -2.3466e-04,  2.5351e-04, -2.5770e-05,  4.1390e-04,
         2.3781e-01, -2.9759e-04,  4.7234e-06, -2.0779e-04, -1.5869e-04,
        -2.0961e-04,  5.1552e-02,  9.0382e-05, -1.4072e-02,  8.8414e-04,
         3.0966e-04,  1.8014e-01,  2.6343e-04, -1.4900e-04, -7.7560e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4482e-05, -4.4943e-01,  2.7749e-04, -1.8304e-01, -1.9893e-04,
         1.7649e-05, -6.7344e-05,  3.6388e-03,  8.8174e-05, -4.2587e-04,
        -1.8955e-01, -1.7379e-04,  3.3192e-04,  4.3948e-05,  1.4636e-04,
         1.2489e-05, -1.7146e-01,  1.7597e-04,  4.9066e-04, -1.1758e-04,
        -1.3028e-04,  1.1534e-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([-5.4807e-03,  1.5508e-02,  4.8488e-04, -3.5828e-04, -1.8946e-04,
         4.7004e-01,  3.8955e-04, -3.3560e-04, -5.0117e-01, -3.1889e-04,
         1.9723e-04, -3.1659e-04, -3.0843e-04, -3.1671e-04, -5.8416e-04,
         1.0923e-03, -7.6255e-04, -1.8938e-04,  9.1955e-04,  4.9894e-04,
        -5.3756e-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([-2.1454e-02,  8.0154e-03,  1.2108e-03,  5.7905e-04,  1.2770e-04,
         6.6904e-01, -1.6006e-03, -1.4527e-03,  2.8849e-01,  4.0078e-05,
         3.4827e-04, -6.0133e-04,  7.9360e-04,  7.3369e-04,  1.0022e-03,
        -2.6697e-03, -5.7591e-04,  6.1977e-04,  6.4969e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2557e-02,  3.9068e-03,  7.7824e-05,  1.1046e-03,  1.1119e-03,
         7.4999e-01,  1.6424e-04, -3.9041e-04,  2.2320e-01, -9.6114e-04,
        -9.9250e-04,  7.1242e-04,  6.6123e-04, -6.2788e-05,  1.3791e-03,
        -1.7338e-03,  8.3013e-04,  1.3997e-04,  1.7276e-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([ 2.0149e-03,  3.1035e-04, -1.6207e-04,  4.8670e-04, -6.8544e-04,
        -6.6695e-04,  6.2802e-04, -1.9187e-04, -5.0214e-04,  3.6522e-04,
        -5.4411e-04,  8.2054e-01,  5.5980e-04, -1.6774e-01,  7.1269e-04,
         1.5131e-03, -1.5200e-03,  6.0019e-04, -2.5465e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #2650: [tensor([-9.4187e-04, -4.9368e-01, -1.3380e-04,  5.5847e-05,  2.5501e-04,
        -2.2700e-04,  4.2546e-05, -5.8239e-05,  1.2055e-04, -4.2813e-04,
         1.3724e-04, -3.4044e-04, -4.8624e-01, -6.2584e-05,  1.6684e-02,
         4.1016e-04, -3.9893e-05, -1.4579e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9072e-04, -7.0625e-01, -4.2442e-04, -2.1881e-04, -3.3635e-05,
         2.2616e-04, -2.8712e-04, -2.9156e-04, -9.6196e-05, -1.4070e-04,
        -7.9525e-05,  7.3035e-05, -2.8226e-01,  1.2301e-04,  3.6426e-03,
        -1.4270e-04, -7.1988e-05,  3.8312e-03, -4.1249e-05, -4.3699e-04,
        -2.3989e-04,  1.9523e-04, -6.4866e-05,  2.2790e-04,  1.0430e-04,
         1.0843e-04,  2.4450e-06,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.8466e-04, -5.1336e-05, -1.9861e-04, -6.5211e-01, -9.7911e-05,
         1.3697e-03, -1.7503e-04, -6.0013e-04, -3.7588e-04, -1.5524e-04,
         8.8843e-03, -1.0452e-04,  1.3259e-04, -1.3183e-04, -3.6360e-04,
         1.3483e-04,  1.6260e-01,  1.8624e-04, -1.0059e-04, -9.8012e-05,
         4.2632e-06,  2.0783e-03, -1.4502e-04,  5.7905e-05, -1.6878e-01,
        -1.7767e-04,  7.3470e-05,  1.2500e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0063e-03,  1.9993e-04,  7.2286e-05, -7.0701e-01,  5.5325e-05,
        -2.3327e-05, -1.7443e-04, -5.9193e-04, -3.4619e-04, -3.4304e-04,
         3.2174e-04, -3.2752e-04, -2.9671e-04, -3.1032e-04, -1.4095e-04,
        -2.0449e-04, -2.2458e-01, -6.0518e-04, -2.7940e-05, -1.9192e-04,
         4.3798e-04,  9.9226e-03,  1.2956e-04, -5.0114e-02, -3.0153e-04,
         7.5621e-06,  2.5299e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.2300e-04,  2.4795e-04,  1.3782e-05, -2.3570e-01, -1.7905e-04,
         9.3178e-04, -1.3849e-06,  4.0948e-05, -1.4382e-04,  7.2315e-05,
         3.0339e-04, -1.8871e-04, -2.6823e-04, -4.1232e-04,  3.7260e-04,
         1.2898e-04, -2.9097e-01, -3.9850e-04, -2.0149e-05,  1.1641e-04,
         3.0658e-04, -2.4456e-01, -1.3131e-01,  9.1914e-05,  3.0075e-02,
        -7.7809e-05,  4.2175e-05, -6.1927e-02,  2.0955e-05, -1.0152e-04,
         4.1482e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7560e-03,  7.5859e-06, -5.7548e-04,  6.2127e-04,  1.2425e-03,
         1.9467e-01, -1.6856e-04, -8.3887e-05, -5.0614e-04,  6.2850e-04,
        -5.1247e-04,  4.8240e-04,  7.8850e-01, -1.5718e-04,  4.9354e-04,
        -4.3070e-04, -6.2577e-03,  1.3355e-03,  1.6201e-05, -2.8277e-04,
        -6.4970e-05,  1.7208e-04,  3.6327e-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([-4.4450e-05,  6.2278e-04,  2.2359e-04, -2.2853e-04,  1.1977e-04,
         5.3227e-01, -7.0090e-04, -1.7783e-04,  3.9031e-05,  1.0055e-04,
        -4.6440e-01, -5.1563e-04, -1.1297e-07,  1.2206e-04,  4.3152e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2548e-04, -2.4588e-04,  1.4789e-04, -4.9814e-04,  5.2935e-04,
         3.3257e-01, -3.4026e-04, -7.2465e-05,  1.1907e-04,  1.1798e-04,
        -2.8525e-01,  2.6258e-05, -2.9280e-01, -5.5205e-05,  4.1883e-05,
        -8.6498e-02,  4.1590e-05, -3.5125e-05,  1.9395e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3252e-03, -5.7189e-01,  1.0102e-04, -1.7724e-04,  7.8822e-05,
         8.2849e-04,  2.0834e-05,  1.0873e-03, -5.4344e-05,  5.1248e-04,
         8.7552e-03,  2.4296e-04, -4.1892e-05,  1.4026e-04, -1.1571e-04,
         2.8119e-04, -1.6550e-01, -9.0777e-05,  7.6309e-05,  3.0168e-04,
        -4.9961e-05, -2.4102e-01, -8.1097e-05,  5.8970e-03,  2.3711e-04,
        -4.2691e-05,  4.9364e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2515e-03, -4.6266e-01, -4.0259e-04,  2.7872e-04,  8.2005e-05,
        -1.9525e-04,  1.7242e-05,  2.4678e-03, -1.6783e-04, -2.3246e-04,
        -2.3652e-01, -2.3732e-04, -2.2389e-04, -3.7687e-04, -2.7362e-04,
        -2.0892e-05, -2.9233e-01, -1.7642e-05, -4.6292e-06, -1.7545e-04,
        -3.3812e-04, -3.3836e-04, -1.0219e-04, -1.2391e-04, -1.6705e-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([ 1.5353e-04, -4.2868e-01, -1.2837e-04, -3.5680e-05,  6.7931e-05,
         1.8202e-04,  1.3829e-05,  1.7543e-03,  1.7255e-04,  2.0446e-04,
        -2.2688e-01,  2.4005e-04,  1.0428e-04, -8.2831e-05, -7.0490e-05,
        -9.0473e-05, -1.9445e-01,  2.2708e-05,  3.0569e-04, -1.4468e-01,
         5.6176e-05,  3.7573e-04,  5.7053e-05,  8.8725e-04,  1.7123e-04,
         7.9427e-05,  5.0990e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1242e-03,  2.1652e-04,  8.4826e-05, -7.1369e-01,  3.1493e-05,
         2.4839e-04, -8.3815e-05,  6.3414e-05, -4.2994e-04, -2.1707e-04,
        -1.6270e-04,  3.5471e-05, -2.8066e-01,  5.9536e-04,  2.9427e-04,
         2.9318e-05, -2.2480e-04,  2.3642e-06,  7.5619e-05,  4.7788e-04,
        -1.6667e-04, -8.3749e-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>)]
Iter #2700: [tensor([ 6.6785e-04, -1.3391e-04,  9.8224e-05, -2.2467e-04,  6.3035e-04,
        -3.0641e-04, -2.7627e-04, -2.6415e-05, -1.1496e-04,  2.3309e-04,
        -1.9614e-01, -1.9865e-05, -9.0757e-05, -2.0186e-04, -2.7076e-05,
         7.3303e-05, -8.2643e-05,  1.3672e-04, -4.1522e-01, -7.3270e-05,
         5.7176e-04, -2.0125e-04,  3.6513e-05, -3.8398e-01,  5.2551e-05,
         1.5836e-04,  1.0277e-04,  1.1956e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4891e-03,  1.0747e-01, -7.5006e-05, -4.1015e-05, -3.0878e-01,
        -2.9920e-04,  1.6414e-03, -4.8240e-06, -1.7746e-01,  5.2631e-05,
        -2.7625e-04, -1.4704e-04,  1.0973e-04, -2.5471e-01, -1.5282e-06,
         4.2092e-02, -4.4376e-04, -7.8948e-05,  3.3971e-04,  1.6864e-03,
        -1.0061e-01,  3.1749e-04, -6.4299e-04, -1.2714e-04, -1.0869e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.6077e-05, -3.6607e-01,  2.9149e-04,  1.1326e-04, -2.4868e-01,
         7.2232e-05,  8.7608e-04, -1.0727e-04, -1.0960e-01,  8.3373e-05,
        -4.5624e-05, -8.3645e-05, -4.9428e-06, -7.7940e-02,  1.8324e-04,
        -1.4150e-04, -1.9533e-01, -9.3822e-06,  1.2159e-06,  2.3074e-05,
        -4.5778e-05, -6.6909e-05, -1.5156e-04, -8.2294e-06,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3225e-03, -1.6293e-01, -1.5370e-05,  7.7061e-05, -2.0437e-01,
        -1.6437e-04,  6.4338e-04, -4.6902e-06, -7.6966e-02,  1.7261e-04,
        -1.1771e-04,  2.0888e-04,  1.2906e-04, -1.8953e-01, -9.1006e-05,
         1.2467e-04, -1.2015e-04, -1.6282e-01,  2.6232e-04, -3.1901e-02,
        -1.6712e-01,  2.9861e-05, -5.7316e-04,  1.8875e-04,  1.0444e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7922e-03,  2.8604e-04,  2.2575e-04, -1.2357e-03,  6.4766e-04,
        -7.4722e-04,  4.1944e-05,  6.9010e-04,  9.8946e-01,  1.7122e-03,
        -8.3579e-04, -7.9928e-04,  5.2263e-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>), tensor([-5.0034e-03, -2.1315e-04,  3.4235e-04,  4.5638e-05, -2.3870e-04,
         3.1013e-05,  4.4032e-05, -1.0225e-04, -1.6356e-01, -4.0606e-01,
        -3.7047e-04,  1.2318e-01,  1.5237e-04,  2.7343e-04, -2.9976e-01,
        -4.2995e-05,  2.5240e-04,  3.3219e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0563e-04,  4.1319e-04,  2.6680e-04,  6.0122e-05,  9.9910e-04,
        -1.0439e-04,  3.0133e-04, -6.4081e-04,  8.9059e-05, -5.8419e-01,
         1.3681e-04, -4.1139e-01,  5.0935e-04, -7.3716e-04, -5.1088e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8769e-03,  2.4343e-01, -6.4110e-04, -6.4156e-04, -3.2629e-04,
        -4.2206e-01,  2.9316e-04,  1.3388e-04,  6.3380e-04,  7.9390e-05,
        -4.4624e-04,  6.2598e-05,  4.5817e-04, -2.7668e-04,  3.0764e-01,
         5.6809e-05,  4.3704e-04,  9.4468e-04, -2.3078e-04, -1.1860e-02,
        -1.8094e-04, -2.7175e-04, -3.0326e-03,  3.4602e-04,  1.8899e-04,
        -1.7348e-05, -7.9300e-04,  4.5981e-04, -1.8613e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.8666e-04,  8.6340e-02, -2.0275e-04,  4.5297e-05, -1.2616e-04,
         1.9590e-01, -1.8790e-04,  2.1052e-04, -2.6622e-04,  3.0969e-05,
        -6.2341e-05, -8.5046e-06,  1.2558e-04,  1.9956e-04, -3.8885e-01,
        -2.8292e-06, -3.8665e-05,  2.8269e-02, -5.9436e-05, -2.9753e-01,
        -4.6280e-04,  1.4705e-04,  1.4711e-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([-1.4993e-04,  6.9996e-01, -3.6785e-04,  1.0525e-04, -2.8442e-04,
        -2.1833e-02, -2.3688e-04, -5.2994e-04, -1.1716e-04,  2.7282e-04,
         1.5051e-04, -4.3456e-04, -3.3535e-04, -1.6048e-04,  2.7373e-01,
         2.2678e-05, -3.6370e-04, -2.1600e-04, -3.9924e-04,  1.9652e-04,
         1.3895e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1735e-03, -3.7332e-01,  6.8890e-05, -1.6599e-04,  1.2927e-01,
        -3.0203e-04, -2.1728e-04,  6.9078e-05, -8.6475e-02, -6.8406e-05,
         1.0311e-05,  7.5767e-05,  7.6030e-04, -6.7196e-05, -2.4536e-01,
         3.2926e-04, -1.0660e-04, -1.6086e-01,  1.5078e-04, -3.3799e-05,
         1.1769e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5263e-03, -2.1109e-01,  3.1104e-04, -1.8772e-04,  6.9272e-02,
        -1.5803e-04,  2.4041e-04,  4.4341e-04, -1.2081e-01, -4.3582e-05,
         1.4779e-04,  1.8237e-04, -1.7535e-01,  1.4452e-04, -2.1853e-01,
        -1.1757e-04,  2.9828e-04, -1.9987e-01,  9.1318e-05,  1.7490e-05,
         1.6988e-04,  0.0000e+00,  0.0000e+00,  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 #2750: [tensor([ 7.6734e-04, -2.9412e-01,  1.7959e-04,  2.8681e-04,  1.1869e-04,
         2.0052e-04, -4.0698e-01, -1.8699e-04, -7.4528e-06, -5.1355e-05,
        -9.3536e-05,  9.6879e-05, -9.3172e-05, -7.5031e-06,  4.6301e-04,
         9.3308e-05,  1.6074e-04,  1.7061e-04,  9.6354e-06, -2.6629e-04,
        -3.6842e-05,  8.8748e-05, -1.8235e-04,  4.2388e-05, -2.4758e-01,
         1.0255e-06,  1.0539e-04, -3.6078e-05,  5.0491e-05, -4.7031e-02,
        -4.9240e-05,  1.3680e-04, -1.5279e-04, -7.2217e-05,  8.1856e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5675e-04, -4.6356e-01, -1.5107e-05,  3.0271e-04,  1.2158e-04,
        -2.3200e-05, -1.7585e-01,  1.7260e-05,  6.5657e-05,  2.3541e-04,
         8.1759e-05, -2.2954e-05,  5.8863e-05,  5.5608e-05,  9.1394e-05,
         1.0144e-04, -3.6977e-05, -1.9260e-05,  1.3498e-04, -5.9668e-06,
        -1.4826e-04,  1.2989e-04,  5.3745e-05, -9.3393e-06,  8.6487e-05,
         1.3574e-04, -2.9893e-07,  2.1802e-05, -1.7421e-01, -1.6888e-06,
         2.6836e-03,  2.0791e-05,  1.2241e-05, -1.3784e-01,  1.6330e-04,
        -2.2891e-05, -4.8331e-05, -4.5176e-05,  1.7334e-05, -1.9381e-05,
        -4.2830e-02,  4.2020e-05, -3.0361e-05, -1.0731e-04,  7.7949e-05,
        -9.5200e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8306e-04, -4.0146e-01, -3.9930e-05, -2.7569e-05,  6.5029e-05,
        -1.0203e-04, -2.7951e-01,  2.7055e-05,  1.5771e-04,  1.1490e-04,
         7.8220e-05, -3.2492e-05, -3.1414e-05,  3.1199e-05, -1.9750e-05,
         7.4136e-05,  9.1712e-06,  7.4712e-05, -6.7250e-05,  1.2673e-04,
        -4.1203e-05,  9.2495e-05, -1.2324e-04, -1.2156e-04, -2.7706e-01,
        -9.0565e-05, -1.2394e-04,  1.0658e-04, -3.8480e-02,  1.2907e-05,
         8.8151e-05,  1.7830e-04, -1.2420e-04,  1.5712e-04,  3.7266e-05,
        -1.2992e-04,  1.2770e-04, -5.9938e-05,  1.2828e-04, -1.2297e-04,
        -1.6298e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3557e-04, -2.4623e-01,  4.9845e-05,  1.1643e-03, -1.5104e-04,
        -1.1572e-04,  5.1609e-02, -8.2491e-05,  1.9432e-05,  2.3708e-04,
         9.2617e-05, -2.8960e-05, -4.1640e-05, -9.5627e-06,  7.4754e-05,
         4.6421e-05,  1.6281e-04, -1.5951e-05,  3.7774e-06, -2.3085e-05,
        -1.9056e-04, -1.5889e-05, -2.4556e-05, -8.2456e-05, -2.2536e-01,
        -1.7535e-05, -1.5373e-04, -6.0110e-05, -2.3118e-01,  3.8788e-05,
         9.7910e-05, -1.4160e-01,  1.0756e-04, -1.0042e-01,  1.2817e-05,
        -1.2837e-04,  2.4968e-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([-4.8222e-04, -3.1555e-01,  8.6218e-05,  9.4595e-04,  2.6342e-04,
        -2.5783e-04, -1.4325e-01,  2.7567e-06,  5.8666e-05, -2.0587e-05,
        -5.2193e-05, -1.4125e-05, -1.8673e-05,  1.2782e-04,  1.1972e-04,
        -2.4850e-05,  2.0107e-04,  9.7709e-05, -9.5617e-05, -1.4830e-04,
        -8.4194e-05,  5.1026e-05,  1.3210e-04, -1.7718e-05, -2.6794e-01,
         5.4830e-05, -6.4239e-05,  1.4874e-05, -1.6012e-01,  7.6276e-05,
         9.6912e-05, -3.3361e-05, -1.2260e-04,  1.1575e-05, -1.0916e-01,
         1.4630e-04,  6.2332e-05,  7.0550e-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([-1.0693e-04, -5.4625e-01,  7.7158e-05,  6.6205e-04,  4.0269e-05,
         1.2556e-04,  2.9067e-04, -2.3991e-04,  4.4200e-05, -1.0334e-01,
        -7.6712e-05, -1.6728e-04,  7.0962e-05,  1.7739e-05,  3.5101e-02,
         1.4319e-04,  2.1555e-04,  8.0257e-05, -3.8565e-05, -1.3494e-04,
         4.0511e-03,  1.2664e-04,  8.4680e-05,  4.9959e-05,  2.3503e-04,
        -1.2647e-01, -4.3096e-05, -4.1153e-06, -1.6903e-04, -1.8138e-01,
        -8.8318e-06, -5.5841e-05,  9.9528e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.0158e-04, -3.6683e-01,  1.8459e-05,  1.9638e-03,  4.4127e-04,
         1.1200e-04, -1.0395e-04, -1.4159e-04, -2.7126e-05, -2.1013e-01,
        -6.6601e-05,  1.4253e-04,  8.7769e-05, -1.3716e-04, -1.8896e-01,
        -8.4061e-05, -1.4628e-04,  3.7067e-05,  1.7331e-05, -7.4372e-06,
        -2.2887e-01, -6.0139e-05, -6.7506e-05,  7.9260e-05, -1.5742e-04,
        -1.4465e-05, -5.2494e-05,  6.7459e-05, -2.2377e-04, -3.3398e-05,
         1.7441e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1470e-03, -3.6717e-01, -5.6553e-05,  4.7732e-04,  2.0926e-04,
        -2.3301e-04, -1.4552e-04, -1.0095e-04, -1.1972e-04, -2.4120e-01,
         7.5979e-05,  1.2050e-04, -1.0239e-04,  5.1371e-05,  3.0299e-02,
        -1.0402e-05, -7.2494e-05, -6.6006e-05,  5.4639e-05, -4.9716e-06,
        -2.2917e-05,  1.2676e-04, -4.0508e-05, -2.3142e-01,  1.8768e-04,
        -1.5311e-04, -1.2616e-01,  5.8372e-05, -4.3859e-05,  7.4247e-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([-2.1211e-04, -7.5252e-05, -3.8981e-05,  8.4297e-05,  1.0862e-01,
        -1.6669e-04, -2.4387e-04, -1.2280e-02, -2.6367e-05, -1.4976e-04,
        -6.1779e-05,  1.0490e-05, -3.2099e-04,  1.0477e-04,  1.5535e-05,
        -1.5997e-05, -1.9048e-05,  4.3803e-02, -1.0408e-04, -1.0204e-04,
        -4.4797e-05, -5.1901e-04, -1.1894e-05, -2.3177e-05,  1.5473e-01,
        -2.0474e-04, -1.4222e-04, -2.4873e-05,  1.8295e-04,  3.2398e-01,
        -9.2009e-05, -6.3638e-05, -4.2618e-05,  7.1287e-05,  9.3253e-06,
        -8.6947e-05, -3.9574e-03,  1.3865e-05,  1.1141e-01, -5.4399e-05,
        -8.6912e-05, -5.7396e-05,  2.3708e-01, -2.1395e-04, -2.7704e-05,
        -1.2844e-04,  6.3612e-05, -1.3642e-04,  8.3230e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9081e-04,  3.2644e-05,  1.2427e-05,  1.1673e-05,  2.5460e-02,
        -5.3121e-05, -8.2194e-05,  4.9278e-02,  1.3467e-04, -2.7303e-05,
         2.7473e-06,  1.6330e-04, -6.9055e-05,  1.0086e-04, -2.6844e-04,
        -3.3304e-05,  2.1458e-04,  4.8606e-02, -1.9576e-04, -2.4130e-04,
        -4.3268e-05, -2.1054e-04, -3.2931e-05,  6.8460e-05,  2.3243e-01,
        -1.7659e-05, -2.0461e-04, -3.4807e-05,  8.7773e-05,  2.7868e-01,
         8.3683e-05, -1.2541e-04,  1.5833e-04, -7.2948e-05,  1.2948e-04,
        -1.1587e-04, -5.8574e-05,  3.6180e-01,  1.6896e-04,  1.9111e-05,
         1.0972e-05,  5.9384e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.7243e-04, -5.1891e-05, -1.3388e-04, -1.9144e-04,  1.2631e-02,
        -2.2352e-05,  5.4480e-05,  3.7760e-02, -3.4416e-05, -2.2753e-04,
         9.1904e-05,  1.2957e-05, -1.8878e-04,  1.4341e-04, -1.5501e-04,
        -3.6177e-05,  4.1661e-05,  3.4930e-02,  5.8140e-05,  3.1691e-05,
        -9.7747e-05,  1.0157e-05, -2.1609e-05,  3.3796e-05,  4.2168e-02,
         2.9649e-05,  8.2253e-06, -1.7060e-04,  3.6932e-05, -3.7771e-02,
         3.4629e-05,  7.5650e-05,  5.4715e-05, -3.0485e-01, -3.4411e-05,
        -5.2759e-05, -4.9049e-01,  8.5015e-05,  1.1971e-04,  3.5830e-02,
        -1.0021e-04,  6.5196e-05, -3.6444e-05, -5.0791e-05,  8.1827e-06,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2645e-03, -4.1817e-05,  8.1041e-05, -1.6075e-05,  3.3214e-01,
         3.3923e-05, -5.6610e-05,  2.7789e-02,  8.5706e-05,  3.1132e-04,
         1.2723e-04,  8.1474e-05, -2.9964e-04,  1.6677e-04,  1.3699e-04,
         9.3738e-05,  9.2495e-05,  6.1108e-02,  3.6254e-05,  2.8773e-05,
         9.0251e-05, -1.8302e-05,  1.3337e-04,  7.5691e-06, -8.7782e-02,
        -1.7034e-04,  1.1229e-04,  2.8718e-04,  8.5789e-05, -3.2186e-02,
         9.6630e-05,  4.1088e-05,  9.9655e-05, -1.5394e-04, -4.9535e-05,
         4.4354e-01,  3.9125e-04,  2.9525e-05,  2.0097e-04,  6.6252e-05,
         6.2402e-05, -9.6939e-04, -3.1575e-07, -1.1194e-04, -8.1593e-03,
        -4.2398e-05, -3.5866e-05,  9.3776e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #2800: [tensor([-8.8684e-04, -2.8174e-01,  2.7753e-04,  7.7648e-04, -9.8372e-05,
        -9.8387e-06, -2.7349e-01, -6.1151e-05, -2.0421e-04,  5.3029e-04,
        -2.4712e-01, -8.3658e-06,  1.0059e-04, -2.3124e-05, -1.0199e-04,
        -1.6141e-01,  4.7545e-04,  1.3186e-04, -1.7934e-04, -1.0882e-04,
         3.0837e-02, -1.0740e-04,  9.6176e-04, -4.9085e-05,  1.6324e-04,
         1.5022e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4673e-03, -3.2599e-01,  6.0779e-04,  5.2404e-04,  1.6248e-04,
         2.5990e-04,  1.7443e-03, -1.8719e-01,  2.0318e-05,  3.0287e-05,
        -1.4918e-01, -2.1691e-01,  8.8308e-06,  1.9557e-02, -5.7926e-05,
        -9.6009e-02,  1.8788e-04,  7.0788e-05, -2.2550e-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([-1.3882e-03, -4.5194e-01, -1.7772e-04, -2.1232e-05,  2.0895e-04,
         5.3238e-05, -3.8431e-04, -2.4038e-01,  2.5554e-04,  1.9794e-04,
         2.1762e-04,  1.5657e-02, -2.5929e-01, -3.2915e-04, -2.1690e-04,
         6.8320e-04, -2.8328e-02,  1.8208e-04, -3.9727e-05,  5.0435e-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([-8.3987e-04,  1.6694e-01,  2.6926e-04, -5.0880e-04,  1.6202e-04,
        -6.3099e-05, -1.9072e-03, -5.8928e-01, -1.0218e-04,  4.8036e-04,
        -2.3570e-01, -2.4824e-04,  2.5773e-03, -3.6109e-04,  3.3584e-04,
         2.2350e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.2851e-04, -3.8032e-01,  8.9013e-05,  9.1797e-06,  5.2488e-05,
         1.1206e-01,  3.0641e-05,  1.6178e-04,  2.0444e-04, -6.7541e-05,
         8.2648e-04,  4.5627e-05,  6.2278e-02, -1.5651e-01, -4.1096e-05,
        -8.3442e-05,  2.0350e-04,  1.6455e-04,  1.2109e-04,  4.5449e-05,
        -1.0867e-01, -1.9212e-04,  6.4300e-03, -3.3258e-05, -1.7013e-01,
        -1.7188e-04, -3.6506e-05,  6.0000e-05,  3.9693e-05,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.6332e-05,  2.1792e-01, -1.9143e-04, -1.4673e-04, -9.7261e-04,
         2.4254e-01,  1.2629e-04,  4.7069e-04,  3.5407e-05,  1.9071e-04,
        -5.4020e-04,  9.4284e-06, -2.0077e-03,  3.0858e-01, -4.3069e-04,
         3.6677e-04, -4.3769e-05, -1.4042e-04, -1.2531e-04,  5.7520e-04,
         1.6174e-01, -2.6965e-04,  9.7381e-03, -1.5292e-04, -2.7714e-04,
         2.6265e-02,  2.5858e-02, -1.0397e-04,  4.3909e-05,  4.6949e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4821e-04, -4.5935e-01,  1.5606e-04, -1.3885e-04, -7.4543e-06,
         6.2075e-02, -2.3586e-04, -1.5899e-04, -1.3999e-04, -2.8410e-04,
         3.5095e-04,  1.2712e-04,  2.1504e-02, -2.7422e-01, -9.2936e-06,
        -2.9154e-05, -1.0786e-04, -2.5001e-04, -3.2208e-05, -1.1835e-05,
        -2.4158e-02, -7.7295e-05, -4.5648e-05,  2.1528e-04, -1.5589e-01,
        -5.2465e-05, -1.5792e-05, -4.3594e-06,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7008e-04,  3.0853e-01, -1.4478e-04,  6.7071e-06, -2.5032e-04,
        -4.2666e-05,  7.1929e-04, -4.2106e-04,  2.5217e-04,  7.1948e-05,
        -1.3646e-05, -1.0811e-04, -3.4536e-04, -8.1172e-05,  8.4303e-06,
        -2.2781e-01, -1.0913e-04,  8.5184e-02,  1.2099e-04, -7.2665e-04,
        -3.7458e-01, -1.7074e-06,  8.9139e-05,  1.3071e-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([ 1.8248e-03,  6.4479e-01, -3.0446e-04, -3.3660e-04, -4.2169e-04,
         2.2666e-04, -7.9303e-04,  6.1058e-05, -2.6896e-04, -1.0444e-04,
        -1.2346e-04, -4.1024e-04, -3.2253e-05, -2.7739e-04,  3.2971e-05,
         8.4490e-02,  5.6021e-06, -2.2882e-04, -2.2920e-03, -4.2863e-04,
         2.6178e-01, -1.2449e-04,  4.1689e-04,  3.7691e-05, -1.8756e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2661e-03, -2.8336e-01,  1.5938e-04,  1.3684e-04,  1.1578e-04,
         3.8626e-05,  3.1283e-04,  3.3238e-05, -3.6541e-04,  2.2771e-04,
         9.2765e-05,  1.0067e-04,  3.5747e-04,  2.0655e-04, -2.0636e-05,
        -1.7886e-01, -1.3190e-04, -4.7146e-05,  9.5699e-03,  2.1523e-05,
        -5.2224e-01, -2.3464e-04, -1.5758e-04,  2.0269e-04,  1.6196e-04,
         2.4112e-04, -9.6990e-05,  8.9662e-05, -1.5117e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.2664e-04, -2.3831e-01,  1.7205e-04,  1.5249e-05,  5.6880e-05,
        -8.8336e-06,  2.3837e-04, -2.2150e-01,  8.8865e-05,  2.0298e-04,
        -1.0398e-04, -3.3826e-01,  5.3379e-05, -3.0372e-04,  1.5855e-04,
         1.8889e-04, -1.3194e-01, -1.1604e-04,  6.7043e-05,  1.2546e-02,
         2.8579e-04, -5.4607e-02,  7.6870e-05, -6.7463e-05, -5.6447e-06,
         9.7439e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.9390e-06, -4.4093e-01, -9.5033e-05, -3.7853e-04,  1.0948e-04,
        -3.9827e-06,  8.1946e-05, -1.7252e-01,  4.9518e-05,  7.6304e-06,
        -1.0594e-04, -1.7818e-01, -7.6428e-05, -2.4617e-04,  5.4108e-05,
         7.2715e-05, -2.0659e-01, -1.6860e-04, -7.2670e-05,  2.3767e-05,
         1.4674e-04,  2.7466e-06, -8.1660e-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>)]
Iter #2850: [tensor([ 2.8086e-03, -1.9507e-04,  5.6868e-04,  1.8733e-02,  1.3181e-03,
         5.0110e-04, -6.7871e-04, -3.7798e-04, -6.8771e-01, -4.5076e-04,
         2.4107e-01,  1.0519e-04,  4.3827e-02,  3.0563e-04, -8.2780e-05,
        -1.5789e-04,  1.1172e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.9751e-04, -3.3322e-05, -3.3625e-04,  1.9727e-02,  1.8469e-04,
        -2.1677e-04, -1.3934e-04,  4.7582e-05, -2.8940e-01,  1.5522e-04,
         3.0390e-02, -1.2252e-05, -1.2098e-04, -2.9506e-01,  2.1074e-05,
         2.9820e-04,  1.1636e-04,  5.9564e-03,  2.1590e-04, -3.5589e-01,
         6.1812e-04,  4.9055e-05,  1.1227e-05,  1.9097e-04,  2.2153e-05,
        -1.9282e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7913e-04, -5.0533e-04,  1.4200e-04,  3.7633e-03, -6.2144e-04,
        -5.0568e-04,  3.8574e-04, -4.1771e-04,  4.1667e-04,  2.3460e-03,
        -9.6987e-01, -9.7241e-05,  5.0555e-04,  3.7012e-04, -1.7092e-02,
        -5.7929e-04, -4.3789e-04,  6.5432e-04, -6.8275e-04,  3.2779e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8220e-03, -4.5944e-04, -2.9003e-04,  2.1602e-03, -2.0006e-05,
         4.8025e-04,  6.8555e-04,  3.6144e-04, -1.8140e-01,  1.6154e-04,
         2.8162e-02, -7.8382e-01, -1.7393e-04, -5.3046e-06,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1646e-03,  9.4219e-04,  1.2414e-03,  2.4048e-02,  4.4062e-04,
         6.0335e-04,  7.4359e-04,  4.8109e-05, -9.6685e-01,  9.9124e-04,
         6.2291e-04,  1.1620e-03,  4.5819e-04,  3.0292e-05,  6.5198e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.1461e-03, -6.1378e-04, -6.7102e-04,  3.0450e-02, -9.3127e-05,
         3.6645e-04, -3.2294e-04,  9.0119e-05, -3.1791e-01, -4.4361e-04,
         4.5127e-03, -2.5520e-04,  1.0392e-03, -6.3416e-01,  4.1834e-04,
         1.8980e-04, -3.2492e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1624e-04,  1.4803e-03, -2.1411e-04,  6.2588e-02,  1.7980e-03,
         9.4701e-04,  2.5835e-04,  2.3271e-03,  9.1160e-01, -1.2512e-03,
         1.5531e-02,  3.5885e-04, -3.7573e-06,  1.1294e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.3206e-03, -2.2891e-04, -6.3166e-04,  1.3777e-02,  5.9274e-04,
         2.3230e-04,  1.2466e-04, -6.9042e-05, -7.2447e-01,  5.6599e-04,
         2.4816e-01,  8.9899e-04,  6.3709e-04, -3.9372e-04,  8.9732e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3047e-03, -3.3405e-01, -3.8047e-05,  2.4143e-04,  1.9195e-03,
        -4.3249e-05, -2.6614e-01,  8.2192e-05,  1.4596e-04,  4.8013e-05,
         1.9354e-03,  1.3390e-04, -7.0127e-06, -1.5519e-05, -5.3093e-05,
        -9.2426e-02, -1.2284e-04, -1.0006e-01, -3.8524e-05, -2.0036e-01,
        -3.3315e-04, -2.0557e-04, -2.2796e-04,  6.7455e-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([ 4.4355e-04, -3.7939e-01, -2.0607e-04,  1.6025e-04,  1.8301e-03,
         2.0658e-05, -3.0756e-01,  6.3049e-05,  1.4826e-04, -5.4183e-07,
         1.3727e-03,  1.8698e-05,  6.6794e-05, -8.3523e-05, -1.6491e-04,
        -5.4291e-02,  4.5840e-05, -2.5056e-04, -2.3687e-01, -2.8010e-05,
         1.8949e-04,  8.5765e-04, -2.4902e-04,  2.3531e-04,  8.2882e-04,
        -5.7154e-05,  1.4434e-02, -8.0621e-05, -5.7058e-05, -3.4650e-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([-2.0758e-03, -1.8242e-01,  1.8253e-05, -4.0576e-05,  2.3908e-03,
        -4.0219e-06,  9.1283e-02,  3.6040e-04,  9.5879e-05,  1.2344e-05,
         2.1195e-03, -1.1097e-05,  2.6200e-04, -1.3285e-04, -3.7538e-05,
        -1.3696e-01,  8.6315e-05, -1.6453e-04, -1.2846e-04,  1.5375e-03,
         5.5489e-05, -1.0834e-04,  3.0810e-04,  2.6087e-04,  1.0056e-02,
         8.9191e-06, -2.4346e-04,  5.2377e-03,  5.7356e-05, -5.6254e-01,
         4.3205e-04,  3.0469e-04, -4.8552e-06,  2.6520e-05, -9.6950e-05,
         1.2193e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1126e-03, -3.4027e-01, -1.0748e-04, -8.8964e-05, -1.8067e-01,
         7.4448e-05, -1.5849e-04,  2.3291e-05,  6.3512e-05, -2.2417e-01,
        -3.3030e-05,  2.1543e-04,  6.0956e-03,  1.3884e-04, -2.4698e-04,
         1.7927e-04, -2.3143e-04,  2.6516e-05, -9.7567e-05, -6.1429e-05,
         1.0720e-03, -2.4693e-04,  6.0066e-05,  9.5891e-05, -1.4588e-04,
        -7.6616e-05, -1.8829e-01, -4.3380e-05,  6.3419e-05, -2.5331e-05,
         1.1973e-04, -8.8003e-05, -5.5430e-02, -1.3889e-04,  2.7237e-05,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #2900: [tensor([ 1.3773e-03, -3.4404e-01, -7.5786e-06, -1.0888e-01, -3.1437e-01,
        -3.7999e-04, -4.5046e-05, -9.1539e-05, -9.1168e-05, -9.8356e-05,
        -2.5059e-04, -4.2111e-04,  2.2523e-05, -1.4976e-04,  6.0641e-04,
        -1.0530e-01,  9.2310e-05,  1.4538e-05, -3.0811e-05, -9.9299e-05,
        -9.8611e-02, -5.8231e-06,  1.2239e-04,  9.1527e-03,  1.5414e-02,
        -2.5253e-04,  2.2526e-05,  4.8715e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5104e-03, -5.8314e-01,  5.4941e-04, -1.7711e-04, -4.3336e-06,
         5.4979e-02, -4.0187e-05,  7.4716e-04, -3.2211e-01,  4.4271e-04,
         2.4916e-04,  2.2854e-04,  3.4674e-05,  5.1685e-04,  1.3839e-02,
        -2.6347e-04, -1.1127e-03, -8.0306e-05,  1.9422e-02, -9.1045e-05,
         2.4395e-04,  2.1772e-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([-1.7473e-04, -6.4570e-01,  6.5423e-05,  1.6526e-04, -3.3413e-04,
         3.6650e-02, -8.6915e-05,  6.2135e-05, -8.4746e-02, -1.1736e-05,
         2.5267e-04, -1.1866e-04,  3.5149e-05, -1.0925e-04,  7.7113e-02,
        -2.1554e-04,  2.9951e-04,  4.2015e-02,  6.0458e-07,  2.1575e-02,
        -5.2627e-04, -8.9058e-02, -2.0018e-04,  1.5680e-04, -3.2734e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3097e-03, -3.2650e-01, -1.1286e-04,  2.2358e-04,  1.2974e-04,
         2.5560e-01,  7.5922e-05,  3.5285e-04, -2.8568e-01,  9.5067e-07,
         4.6496e-04,  2.3560e-05,  7.6608e-05,  7.4861e-05,  5.7187e-02,
        -1.5069e-04,  8.4840e-04,  5.6531e-05,  9.0794e-03, -6.1777e-02,
        -1.1851e-04, -3.6569e-06,  1.5747e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3017e-04, -3.2821e-02, -6.7058e-05,  1.5683e-05, -1.8016e-01,
         4.0260e-04,  9.2522e-04, -1.5798e-01,  1.2770e-04,  1.1045e-04,
        -3.0342e-01,  2.7571e-04, -1.5830e-05, -5.0313e-05, -2.1899e-01,
         8.5430e-05,  8.2023e-04, -3.4826e-06,  3.8594e-04, -1.0268e-01,
         1.9770e-04,  1.4226e-04,  1.0049e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8054e-04, -2.2499e-01, -1.6200e-04,  4.0377e-06, -2.5572e-01,
        -7.7992e-05,  3.6390e-03, -1.7750e-01,  2.6677e-05,  5.2128e-05,
         2.9620e-03, -7.7214e-05, -8.2121e-06,  3.1960e-05, -1.2899e-01,
         1.0823e-04,  1.7158e-05, -2.0416e-01, -2.3718e-05,  9.4847e-05,
         2.8003e-04,  3.9033e-04,  2.6827e-04, -2.1621e-04,  1.2569e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0018, -0.0052, -0.0072, -0.1842,  0.1954,  0.3257,  0.0576, -0.0263,
        -0.0872,  0.0260,  0.0195,  0.0556, -0.0083,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.3590e-03,  6.6606e-04,  2.7627e-05,  2.0135e-03, -3.7229e-05,
        -3.4680e-03,  1.9095e-03,  6.9024e-05,  2.0427e-04,  8.0144e-05,
        -5.6091e-04, -4.5114e-04,  9.8276e-01,  1.3513e-03,  8.4251e-05,
         9.5585e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0106,  0.0113, -0.0274, -0.0525,  0.1109,  0.2368,  0.0243,  0.0029,
        -0.0354,  0.0271,  0.2197,  0.2059,  0.0015,  0.0311, -0.0025,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.9106e-04, -5.9922e-01,  1.5585e-07,  4.6298e-05,  5.9594e-04,
        -9.7649e-05, -2.3947e-04, -2.4884e-04,  1.0396e-02, -3.0085e-02,
         4.7414e-05, -4.1378e-05,  3.8298e-04, -5.2193e-05,  8.8847e-05,
        -1.9813e-04,  7.8983e-05, -1.6454e-01,  2.0152e-05,  3.7657e-04,
        -1.9194e-01,  2.9578e-05,  2.0253e-04,  2.7849e-05, -2.4733e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4767e-04,  9.5346e-01,  5.8138e-04,  1.3479e-04, -1.2948e-03,
         1.1235e-04,  1.8140e-04,  1.9731e-04, -1.4820e-02, -1.2714e-02,
        -5.5033e-05, -5.3604e-05,  2.2744e-04,  1.6776e-04,  1.6652e-05,
        -9.5726e-05, -6.1350e-05, -1.4662e-02,  3.1449e-04,  2.0129e-04,
         2.2682e-04,  1.6885e-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([ 1.6561e-04,  6.9630e-01,  2.8428e-04, -7.8576e-06, -1.7234e-03,
         2.9173e-06, -4.0187e-04, -3.0371e-04, -1.9116e-02, -3.5602e-03,
         2.7547e-04,  4.3085e-05,  1.8476e-04,  2.1481e-04, -3.5673e-05,
         3.7078e-05,  2.6141e-01,  2.3859e-04, -1.4820e-02,  1.8450e-04,
        -2.7482e-04,  1.5623e-04, -2.2191e-04, -2.7555e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #2950: [tensor([ 0.0632,  0.0113,  0.1118,  0.4966, -0.0481, -0.0069,  0.0595,  0.0114,
        -0.0060,  0.0097, -0.0295,  0.0987, -0.0314,  0.0133,  0.0025,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3707e-04, -2.4160e-04, -6.7732e-05,  6.4299e-04, -3.9792e-04,
         2.9228e-04, -7.0517e-04, -1.9912e-04, -1.4987e-04, -5.8932e-01,
        -3.3753e-04,  2.5629e-02, -1.5316e-05, -3.8162e-01,  1.1173e-05,
         2.3617e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5692e-03,  3.4348e-04,  7.4480e-04,  1.9427e-03,  1.2359e-03,
         4.9950e-05, -4.5960e-03,  1.0279e-03, -1.7948e-04,  7.8483e-01,
        -8.6999e-05, -5.1187e-02, -4.9618e-04,  1.5212e-03,  1.1588e-03,
        -1.7494e-02,  2.7553e-04, -1.2892e-01, -2.5183e-04,  1.5569e-03,
        -5.3164e-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([-1.1545e-03, -3.3321e-04,  1.4927e-04,  6.7586e-04,  5.7501e-04,
        -5.1037e-05,  5.5918e-04,  1.6517e-04, -4.3965e-04, -5.4425e-01,
         2.4103e-04,  1.4138e-04, -4.5069e-01,  1.6042e-04, -4.1295e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3192e-03, -4.3094e-01, -1.9152e-04,  1.2420e-03, -2.9973e-01,
         1.7122e-05, -2.2409e-04,  3.3924e-04, -6.5884e-06,  1.5114e-04,
         1.5297e-05,  3.0432e-04, -1.3838e-04,  2.4568e-04,  1.9010e-04,
        -1.2377e-01,  8.0296e-05, -1.8544e-05, -8.1044e-05,  7.8710e-05,
         1.3487e-04, -1.3954e-01, -1.7322e-04,  1.0464e-04,  5.1706e-04,
        -1.1761e-04,  3.3165e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2937e-03,  2.3985e-01,  5.4223e-05,  7.5400e-04, -3.5852e-01,
         2.0402e-04, -5.7644e-05,  3.2075e-04,  9.4915e-06, -5.3934e-05,
        -9.6085e-05,  1.1337e-05,  3.1008e-06, -1.1743e-04,  8.7515e-05,
        -1.8198e-01, -2.4606e-04,  6.4364e-05,  1.7738e-04, -1.4818e-01,
        -3.8980e-04,  4.8433e-05, -1.2834e-04,  6.6778e-02, -1.9644e-04,
         2.6266e-04,  1.9253e-05, -9.4268e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4512e-03, -4.5202e-01, -1.3137e-04,  7.2916e-04, -6.4128e-02,
        -1.9207e-04,  7.3152e-05,  7.8959e-05,  1.0820e-04,  1.3073e-04,
        -9.0526e-05,  1.2023e-04,  7.1568e-06,  7.8419e-05,  4.6872e-05,
        -1.5164e-01, -7.6732e-05,  1.3167e-04, -7.9299e-06, -9.4668e-02,
         8.2508e-05, -2.3185e-01, -1.4186e-04,  1.7925e-03,  1.0728e-06,
         1.4995e-04, -7.3695e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4554e-03, -1.3098e-04,  5.1646e-06, -5.4285e-01, -1.6203e-04,
        -1.8445e-05, -6.5198e-05, -6.3130e-04,  4.4884e-04, -4.3318e-04,
         9.8987e-05,  3.4501e-05, -2.3530e-04,  1.6309e-04, -2.4562e-04,
        -1.4144e-04, -6.4972e-04, -4.4869e-01, -3.6612e-04,  6.8783e-04,
         3.3017e-04, -1.0585e-04,  5.0341e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2167e-04, -3.6247e-05,  2.9403e-04, -3.8408e-01, -9.3883e-05,
        -4.0071e-04,  2.8521e-04, -7.7766e-04, -4.6705e-04, -2.0630e-04,
        -4.8268e-04, -4.2171e-04,  9.5952e-06,  9.3105e-05,  2.3092e-04,
        -4.6449e-05, -2.2567e-04, -6.1086e-01, -5.7308e-04, -6.5163e-05,
        -7.9197e-05, -4.5207e-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([-2.7155e-03, -1.0586e-04, -1.2853e-04, -3.5665e-01, -3.0318e-04,
        -3.0163e-04,  1.0380e-04, -4.0118e-04, -8.7362e-05,  1.9647e-04,
         1.4256e-04,  1.7516e-04, -7.4227e-04, -4.3783e-05, -3.1353e-04,
        -6.7177e-05,  1.2806e-04, -6.3618e-01, -5.2091e-04,  1.7235e-04,
         3.3104e-04,  6.6432e-05, -1.1957e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5464e-04,  9.3273e-05,  1.6406e-05, -2.5666e-04, -4.5192e-01,
        -3.1346e-04, -2.4211e-04, -3.1327e-01, -3.4008e-04, -1.3711e-04,
         2.9855e-04, -2.3236e-01, -2.2370e-04,  3.1244e-04, -6.7908e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7072e-03, -1.1039e-04,  5.2972e-05,  6.5499e-04, -3.9761e-01,
        -2.9443e-04, -2.0644e-04, -2.6200e-01,  9.6562e-05, -2.1041e-04,
         3.0041e-04, -3.3598e-01,  2.1386e-04, -5.6598e-05, -4.9741e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  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 #3000: [tensor([ 2.5635e-04,  5.7085e-02,  3.0777e-05,  6.6925e-05,  5.8102e-04,
        -4.9945e-05, -2.7927e-04,  2.3989e-04,  1.6837e-04, -9.3752e-01,
         1.0075e-04,  1.7528e-04,  6.3498e-05, -1.4662e-04, -1.2544e-04,
        -9.8239e-05,  2.4181e-04,  1.4048e-04,  4.8809e-04,  6.3370e-04,
        -4.0391e-04, -2.6045e-04,  4.2288e-04, -1.1587e-04,  3.0858e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8254e-03,  6.4591e-01, -1.5274e-05,  5.1855e-04,  3.2211e-05,
         9.7324e-04, -8.2217e-05,  7.2314e-04, -5.0785e-04, -5.9523e-04,
        -5.4953e-05,  1.8969e-05,  2.3270e-01,  4.0446e-04,  4.0875e-04,
         2.7024e-03,  1.1228e-01, -5.0742e-05, -1.9170e-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([-2.8622e-04, -4.7643e-01,  2.3539e-05,  1.3265e-04, -1.6054e-04,
        -3.2235e-04, -4.5977e-04,  2.9424e-06,  4.3243e-03, -1.9976e-05,
         2.7090e-04, -5.7836e-07, -2.5563e-01, -4.0137e-04,  1.4382e-02,
        -7.8995e-05, -4.0671e-04,  5.3032e-05, -2.4653e-01,  6.5195e-05,
         1.4147e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2144e-04,  5.8586e-01,  7.4919e-05,  1.9491e-04, -1.2627e-04,
        -8.3433e-06, -5.4397e-04,  1.4842e-04, -1.3456e-02, -4.8745e-04,
         3.1806e-04,  6.8956e-05,  2.3761e-01, -2.8088e-05, -4.5417e-04,
         1.5909e-01,  2.4526e-06,  2.7488e-04,  5.6467e-04,  2.7047e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3179e-04, -3.2400e-01,  1.1867e-04,  1.2926e-02,  1.3056e-04,
        -7.4703e-02, -3.8413e-05, -2.9351e-05, -2.0915e-01,  2.4383e-06,
         1.7498e-04, -5.0389e-05,  1.0036e-04,  2.0228e-04, -6.2739e-02,
         3.0253e-05, -1.0655e-01,  2.5201e-04, -2.0816e-01,  1.4434e-04,
         7.7143e-05,  4.5721e-05,  5.5677e-05,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1310e-03,  2.1468e-02,  3.8960e-04, -1.7994e-04,  2.4093e-04,
        -3.2271e-01,  1.0271e-04,  2.3420e-04, -2.5225e-01,  2.3158e-04,
         2.0866e-04,  1.5428e-05, -9.8201e-05, -8.6853e-05, -2.2748e-01,
         1.2549e-04, -1.7235e-01, -2.2555e-04,  2.8241e-04, -2.4050e-05,
        -1.6663e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7529e-04, -6.5246e-01, -9.1605e-05,  2.2364e-02,  4.0190e-06,
         8.1726e-02,  2.9522e-05,  4.5704e-05, -1.7227e-01,  2.1715e-04,
        -6.9159e-05, -2.4707e-05, -1.5080e-04, -6.7862e-05, -2.1877e-04,
         2.0662e-04, -5.6947e-02, -5.2754e-04,  1.2201e-02,  6.7939e-05,
        -5.0285e-05,  7.9532e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8782e-04,  5.7823e-01, -3.4792e-06, -5.1551e-04, -1.6253e-04,
        -1.9703e-04,  4.7018e-04,  3.7345e-01, -2.1485e-05,  2.5696e-04,
         1.2923e-04, -2.4472e-04,  1.2264e-04, -8.9400e-05, -4.5365e-02,
        -1.7656e-04,  1.6831e-05,  3.3644e-05,  3.2698e-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([ 2.4311e-04, -7.0535e-01, -2.4052e-05,  9.9603e-04,  2.6779e-05,
         4.6017e-05, -1.6902e-04, -2.6709e-01,  2.6900e-04, -2.9717e-04,
        -2.3732e-04,  1.9874e-04,  2.0662e-05,  1.0707e-04, -2.3090e-02,
        -1.9788e-04,  4.3801e-05, -1.0348e-04,  8.4422e-04,  4.1127e-04,
         2.3157e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.6631e-04, -6.1768e-01, -3.9281e-04, -1.4448e-05, -8.2966e-05,
         2.0591e-04, -2.2789e-04, -2.9864e-01,  1.6444e-04, -2.0763e-04,
        -2.7617e-04,  1.6691e-04, -1.3984e-04, -1.9566e-04,  7.9047e-02,
        -6.2745e-05, -9.5836e-04, -3.3899e-04, -5.0632e-05, -2.9054e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1864e-04, -6.4807e-01, -4.1843e-05,  1.4629e-03, -3.6690e-04,
         6.1795e-04, -1.1494e-04, -3.1474e-01, -1.4515e-04, -2.1797e-04,
        -1.5254e-04,  1.4178e-04, -1.1064e-04,  2.7714e-04,  3.1284e-02,
         3.8658e-05,  1.3279e-03,  1.9916e-04,  1.4926e-04,  3.6129e-04,
        -1.7621e-05, -4.1054e-05, -5.1639e-06,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1436e-04, -4.4087e-01,  1.5556e-04,  2.5047e-03,  2.3809e-04,
         8.8356e-04, -1.4294e-04, -1.9208e-01,  7.4803e-05,  1.9822e-05,
         3.7413e-05, -6.3919e-06, -1.1392e-05, -2.5674e-05, -2.1756e-01,
        -4.2485e-05,  1.8346e-03, -1.3710e-04, -2.6421e-04,  1.5526e-03,
        -1.4066e-01, -1.9363e-04, -1.8860e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #3050: [tensor([-2.3398e-03,  4.4517e-04,  8.5510e-04,  4.5919e-04, -9.8609e-04,
        -2.1103e-04, -2.2150e-04,  1.9721e-05,  1.5652e-04, -3.2286e-04,
         1.5573e-03,  9.9035e-01,  9.2224e-04, -6.8124e-05, -1.2290e-04,
         9.6700e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3726e-03, -1.0659e-04,  2.3600e-04, -7.7564e-04,  1.9414e-03,
         5.5323e-04,  3.1211e-03, -5.7261e-04,  1.5760e-04, -2.3232e-03,
         9.7875e-01, -3.5616e-03, -7.5505e-04, -1.6971e-03,  2.0784e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7512e-04, -3.2034e-01,  1.5775e-04,  3.6013e-04,  2.7504e-04,
         1.3128e-04,  8.3920e-02,  3.9225e-04, -4.5278e-05,  2.2354e-04,
        -3.8316e-01,  1.2381e-04,  7.1920e-05, -2.4396e-06,  1.8354e-04,
        -2.0853e-01,  4.9257e-05,  1.1974e-04,  6.0310e-04,  4.0244e-04,
         1.8942e-04, -1.4498e-04, -1.5729e-06,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.8336e-04, -4.9554e-01,  2.4421e-04,  8.8035e-06,  2.8791e-04,
        -9.6384e-05, -1.7519e-01,  1.3838e-05, -3.1316e-04, -1.4167e-04,
         1.4755e-01, -2.8244e-04, -7.6742e-06, -3.7580e-05, -2.1493e-04,
        -1.4524e-01,  2.7153e-04, -1.1027e-04,  3.2850e-02,  2.0851e-04,
         1.8578e-04,  2.1548e-04, -1.8804e-04,  1.7759e-04, -1.3760e-04,
        -1.3432e-05,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2719e-03, -2.4418e-01, -3.1555e-05, -2.5375e-04, -3.0661e-05,
         1.0139e-04,  1.2523e-01, -2.3594e-04, -6.8072e-05,  6.6147e-05,
        -3.1124e-01, -1.7759e-04,  1.7083e-05, -1.7335e-04, -1.3433e-04,
        -9.4760e-02, -5.9730e-05, -1.1140e-04,  6.4342e-05, -1.8854e-04,
        -1.5944e-04,  4.0402e-03, -4.2020e-04, -2.1675e-01, -1.8166e-04,
        -1.7578e-05, -4.8230e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6458e-04, -4.8784e-04, -1.5701e-04, -2.3894e-04,  1.3972e-01,
         8.4844e-04, -4.5505e-04, -8.3572e-04,  1.3671e-04,  6.5118e-04,
        -1.2583e-05, -8.5571e-01,  2.2479e-05,  3.6200e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.4842e-04,  4.3151e-04,  6.5461e-05,  1.3283e-03, -4.8961e-01,
         5.4594e-04,  2.3137e-04,  1.9881e-04, -2.1736e-04,  1.4439e-04,
        -4.4922e-04,  4.3120e-04, -5.0539e-01, -1.4093e-04, -1.6849e-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,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0887e-03,  2.1144e-04, -3.2197e-04,  8.1159e-05, -7.4322e-01,
         2.2578e-04,  2.3093e-04, -1.2589e-04, -1.1707e-05, -6.1990e-04,
         3.4493e-04,  1.4658e-04,  3.2982e-04, -2.5248e-01,  3.4842e-04,
        -2.0807e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.6172e-04, -6.7742e-01, -1.6166e-04,  2.4256e-02,  1.3242e-05,
        -8.5167e-02,  3.2102e-04,  2.0723e-04,  1.6301e-02,  1.0401e-04,
         1.2619e-04,  1.9235e-04,  1.3551e-04,  6.1848e-05,  2.7697e-04,
        -1.9300e-01,  1.4033e-04,  5.0130e-04,  3.5118e-04,  3.1788e-04,
         3.4872e-05, -1.5423e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4649e-04,  2.1325e-01, -4.7683e-05, -2.2396e-02,  1.2938e-04,
         3.0330e-01, -1.4022e-04,  4.4354e-04,  3.2892e-04, -1.2679e-05,
         1.9208e-04, -1.2748e-05, -2.6448e-04,  3.1984e-04,  2.3674e-01,
         1.8870e-04, -6.0416e-03,  2.1467e-04,  2.1303e-01, -1.4377e-03,
         3.4718e-04,  4.3122e-04, -8.6301e-05, -2.0005e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5853e-03, -2.9543e-01,  2.7775e-04,  1.4479e-02, -6.6740e-05,
         7.0687e-02, -3.1516e-04,  3.9090e-04,  1.4588e-02, -3.4659e-04,
         8.9035e-05,  1.0159e-04,  3.4171e-05, -2.4542e-04,  2.9509e-04,
        -4.5033e-01, -3.3948e-04, -1.0327e-04,  1.4915e-01, -2.0293e-04,
        -3.9513e-04, -1.2004e-04,  2.0434e-04, -1.5664e-05,  6.4026e-05,
         1.3730e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.6283e-04, -4.3649e-01,  3.6106e-05, -1.3155e-04, -1.7331e-01,
        -1.7933e-04, -5.3290e-04, -5.0033e-05,  2.8926e-04, -2.8949e-05,
        -2.9707e-05, -2.2475e-04, -2.9701e-01, -2.3306e-04, -1.5687e-04,
         8.6096e-05, -8.9632e-02, -6.4725e-05,  3.0555e-04, -7.8640e-05,
        -2.0703e-04, -6.0340e-05, -1.0127e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #3100: [tensor([-6.0544e-04, -3.0411e-01,  9.4156e-05,  7.9095e-05,  8.0103e-05,
         1.3210e-04, -3.6452e-01, -1.2275e-04,  9.4595e-05, -1.4279e-04,
         1.2983e-02,  5.8347e-03, -6.6656e-05,  2.4250e-05, -9.7509e-05,
         7.2679e-05,  9.5389e-02,  4.8164e-05,  1.4744e-04, -4.4833e-06,
        -1.5776e-01,  1.2057e-04, -1.4387e-04, -5.7235e-02,  5.8812e-05,
         7.3289e-06, -2.4561e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5115e-04, -4.0476e-01,  1.1380e-04,  5.9922e-06,  5.5340e-04,
        -1.6704e-05, -1.8126e-04,  1.6344e-04, -6.4849e-02,  3.3451e-05,
         2.2990e-04,  1.6378e-04, -3.0673e-01,  8.5680e-05,  1.0479e-04,
         2.9456e-05, -9.9808e-02,  2.4432e-04,  1.1716e-03, -1.9540e-04,
        -1.2004e-01,  1.2558e-04,  1.0299e-04, -3.7374e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3205e-04, -3.9633e-01, -9.7698e-05, -2.3132e-04,  2.7346e-04,
         2.7061e-04, -1.6089e-04, -1.0521e-04, -7.8597e-02,  6.2561e-05,
         7.7043e-05, -2.0996e-04, -1.2361e-01,  1.0205e-04, -9.9878e-05,
        -1.1846e-05, -1.0662e-01, -3.1888e-04,  3.4042e-04,  1.5528e-04,
        -1.6276e-01, -1.0374e-04, -2.4397e-04, -1.2852e-01,  3.6279e-05,
         1.1524e-04, -1.1005e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1276e-04, -3.6345e-01, -8.0837e-05, -8.4948e-06,  1.1742e-03,
        -6.1543e-05, -1.6696e-04,  1.5981e-04, -1.1939e-01, -7.9009e-05,
         1.7062e-04,  6.5816e-05, -8.9861e-02,  1.4267e-04, -4.9595e-06,
        -6.1400e-05, -1.6536e-01, -8.7846e-05,  6.2547e-04,  7.1692e-05,
        -2.4777e-01, -1.4633e-05, -1.7120e-04, -1.0603e-02,  1.4675e-05,
        -9.0001e-05, -9.0746e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.3489e-04,  5.6036e-04, -1.6002e-04, -1.4726e-04,  3.6466e-04,
        -6.4435e-01,  6.7903e-04,  2.8048e-04, -1.2679e-04, -3.3805e-01,
         3.3281e-05,  1.3894e-02, -2.7509e-04,  1.4490e-04, -2.9676e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5405e-04, -4.0164e-04, -1.1412e-04, -3.8132e-04,  7.2446e-04,
        -4.8993e-01, -1.0916e-04,  6.4566e-07, -5.0954e-05, -4.9913e-01,
        -1.2826e-04,  8.3264e-03, -1.3087e-04, -1.8267e-04,  2.3297e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5082e-04,  2.9158e-04, -1.4305e-04,  1.9266e-04,  2.9453e-04,
        -3.4842e-01,  2.1563e-05, -2.4122e-04,  1.1283e-05, -3.6464e-01,
         5.1161e-05,  2.7787e-02,  4.1855e-04,  1.8154e-05, -2.5659e-01,
         2.3594e-04, -1.2688e-04, -7.9401e-06,  6.3421e-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([-8.4236e-04, -5.0426e-01, -1.7478e-04,  9.0577e-05,  1.0777e-03,
        -8.9244e-05,  2.4281e-02,  1.4597e-04,  7.8348e-04,  2.0114e-04,
        -2.3373e-01,  4.1832e-04,  9.2594e-05,  7.8792e-05, -2.3203e-01,
        -9.9905e-05,  8.1360e-04, -1.4348e-04, -2.7429e-04,  2.3845e-04,
        -1.2373e-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([ 1.0113e-04, -4.9407e-01,  2.8684e-04,  1.2223e-04,  8.3808e-04,
         1.7210e-05,  2.9851e-03, -2.6872e-04, -2.7252e-04,  7.9077e-05,
         1.5073e-01, -2.9841e-04,  1.7105e-04,  2.4003e-04, -5.2391e-04,
        -2.6311e-02, -1.3932e-01,  3.4322e-05,  7.8045e-03, -1.7528e-01,
        -1.6002e-05,  1.4189e-04,  8.0981e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2760e-05, -5.9920e-01,  1.6335e-04,  1.0600e-04,  7.3710e-04,
        -2.2077e-05,  3.6176e-03,  1.4071e-04,  5.3867e-04,  6.4767e-04,
        -2.8621e-01, -2.2913e-04, -1.2711e-04, -1.0579e-06, -1.0488e-01,
        -3.8443e-05,  2.7502e-03, -1.0727e-04, -2.6833e-04, -6.3032e-05,
         2.4781e-05, -1.1185e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4528e-04, -7.1481e-01,  4.0754e-05,  5.5165e-05, -1.4407e-04,
         2.6640e-04,  1.0484e-04,  2.8522e-04, -2.6497e-05,  1.2376e-02,
        -1.0405e-04,  2.4849e-04,  6.8498e-06, -2.6415e-01, -1.5433e-04,
        -2.0663e-04,  1.5501e-03, -1.6013e-05,  1.4298e-04,  4.2808e-03,
        -2.5065e-04, -2.3584e-04,  1.2687e-04,  1.7026e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.3071e-04, -3.2964e-01, -1.2905e-04,  7.7769e-05, -1.3767e-04,
        -1.0822e-04, -1.7648e-04, -2.6727e-04, -2.6090e-04,  6.7480e-03,
        -7.4077e-05,  9.3648e-05, -9.9868e-05, -2.0972e-01, -4.9226e-05,
        -4.8677e-02, -2.5491e-05, -2.1644e-01, -1.1531e-04,  4.5202e-03,
        -2.5787e-04, -1.8078e-01, -2.4711e-04,  4.3261e-04,  1.8948e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #3150: [tensor([ 1.2357e-03, -2.0584e-01,  2.9658e-05,  3.9297e-04, -2.1140e-01,
         1.3922e-04,  1.6423e-02, -2.4599e-01,  4.0811e-05,  2.0477e-04,
        -1.9483e-04,  9.9279e-05, -2.0177e-01,  3.4662e-04,  1.5695e-02,
        -8.3833e-02,  5.1780e-05, -4.7358e-05,  1.1536e-04,  1.2839e-04,
         1.5678e-02,  1.3209e-04,  5.2540e-05, -8.6002e-06, -9.0644e-05,
        -5.4705e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9038e-04, -2.7232e-01,  3.4379e-06,  9.6871e-05,  5.8176e-02,
         1.2529e-04, -8.4571e-02, -1.4479e-01,  1.7848e-05,  5.4846e-05,
        -1.1068e-04, -4.2286e-05, -9.0743e-02, -1.4710e-04,  6.4800e-03,
        -1.8133e-01,  1.2678e-04, -3.7443e-05, -1.0889e-04,  6.2927e-05,
         1.1478e-03,  7.2941e-05, -1.3402e-04, -6.4269e-05, -1.1949e-04,
        -1.5857e-01,  1.2695e-04,  1.4618e-04,  4.0549e-05,  3.7615e-05,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2185e-03,  1.9925e-01, -8.0009e-05, -3.0920e-04, -1.0919e-01,
        -9.8411e-05,  2.2636e-03, -1.5805e-01, -3.6586e-05, -1.2728e-04,
        -1.8797e-06, -5.5187e-05,  8.6801e-02,  3.0274e-05,  4.2161e-03,
        -1.2100e-01, -1.9949e-04, -8.1861e-05,  3.5140e-05, -1.1069e-01,
        -1.7332e-04, -2.0358e-04,  3.5406e-02, -3.1623e-04, -1.1057e-01,
        -1.0131e-04, -4.3751e-05, -5.9363e-02, -4.3733e-06, -5.9455e-05,
         2.6146e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.4606e-04, -6.6753e-05, -2.0748e-04,  5.0750e-04,  2.7680e-04,
        -2.0790e-05,  1.4847e-04, -2.0434e-01,  1.3804e-04,  3.3553e-04,
        -3.6178e-01,  2.2007e-04,  1.1204e-04,  1.8967e-04, -2.3335e-01,
        -6.5503e-05,  2.1173e-02, -1.7564e-01,  1.7368e-05, -6.1706e-05,
        -1.3101e-04,  1.1884e-04,  1.5993e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8087e-04, -1.3339e-04,  8.8038e-05, -9.3281e-04, -1.8915e-04,
         2.6726e-04,  1.6865e-04,  1.8377e-01,  1.0766e-04,  4.7643e-04,
         3.1823e-01, -2.9729e-04, -2.4588e-04, -6.4023e-05,  1.6072e-01,
         4.5722e-04, -3.2667e-04, -1.9863e-02,  1.5904e-01,  2.0803e-04,
        -3.4687e-04,  1.5344e-01,  2.4270e-04,  3.7524e-06,  2.7766e-05,
         1.7667e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9446e-03,  3.9791e-05,  3.0538e-04,  1.1115e-03,  4.8546e-04,
        -1.2225e-04,  4.9139e-04, -2.5095e-01, -4.2700e-04, -3.2721e-04,
        -6.0289e-01, -4.1737e-05, -1.5701e-04,  2.6837e-04, -1.3967e-01,
        -2.9974e-04, -4.9233e-05,  3.1682e-05, -1.8838e-04, -1.7514e-04,
        -2.0860e-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([-2.7232e-04, -9.4372e-02,  4.0109e-04, -1.1166e-02,  3.1695e-01,
         1.5283e-04,  1.9586e-04, -2.9848e-05, -2.7656e-03,  3.3024e-01,
        -7.9313e-05,  3.6801e-05,  2.3861e-04,  5.2799e-05,  2.4208e-01,
         2.9620e-04, -5.3756e-04, -1.1660e-04, -1.1588e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1953e-04, -5.7364e-01, -1.7588e-04,  1.5893e-02, -5.8271e-02,
         1.8908e-05,  2.3638e-04,  4.8090e-04, -1.5686e-02,  1.5381e-02,
         1.2024e-04, -6.9529e-05,  1.4058e-06,  7.2987e-05, -4.7531e-02,
        -2.3706e-04,  3.3097e-04, -2.7098e-01,  1.5793e-04, -9.8412e-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([-1.3166e-03,  2.6606e-01,  2.6378e-05,  1.2282e-01,  3.1539e-01,
        -2.1332e-04,  1.2263e-04, -2.7302e-04, -3.8681e-03,  2.1186e-01,
        -2.3722e-04, -1.3373e-04,  2.7143e-06, -3.3709e-05,  7.5786e-02,
        -2.1457e-05, -5.8139e-04,  2.9219e-04,  4.6347e-04,  1.6799e-04,
         2.1196e-04,  1.1848e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.7833e-04,  2.1178e-05,  3.6102e-04, -1.4489e-05, -1.5340e-04,
         6.2072e-02,  9.2182e-05, -1.0175e-04, -3.2999e-04, -2.1814e-04,
        -3.0214e-01,  3.8581e-04, -2.9554e-01,  3.5729e-04, -3.3722e-01,
        -3.3928e-04, -3.7167e-08,  7.2379e-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([ 2.4518e-04, -4.8725e-04, -2.1350e-04,  1.0252e-04, -3.2168e-04,
        -1.5436e-02, -6.5810e-04,  1.0266e-03,  4.6041e-06, -4.2933e-04,
        -4.2214e-04,  1.2064e-04,  9.7961e-01,  4.8482e-04,  1.0234e-06,
        -4.3688e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4838e-03,  9.3968e-05,  3.2426e-05,  1.7527e-04,  1.4456e-04,
         4.4428e-02, -8.0180e-06, -2.1604e-04,  5.7652e-05,  3.9635e-04,
        -5.0052e-01, -3.0319e-04,  4.9426e-04,  2.9594e-04, -2.2464e-04,
        -2.1723e-01, -1.3487e-04, -2.3224e-01, -3.7177e-04,  1.2201e-04,
         3.0900e-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>)]
Iter #3200: [tensor([ 2.3038e-04, -5.0584e-01,  8.7389e-05,  3.1021e-02,  1.3169e-04,
         8.5303e-05, -9.5587e-05,  2.4327e-04, -1.4200e-01,  8.6752e-05,
        -1.7946e-04, -5.3211e-05, -3.1209e-01, -2.7959e-04,  6.0294e-03,
        -3.5098e-04, -2.0539e-04, -1.1426e-04,  5.1468e-04, -1.1662e-04,
        -1.2150e-04,  1.2306e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.2914e-04, -5.2772e-01, -1.3030e-04,  1.0375e-02,  1.3797e-04,
         1.4092e-04, -3.1850e-04,  2.9189e-04, -1.4338e-01, -3.4456e-04,
        -1.0901e-04,  2.8091e-05, -3.0063e-01,  4.3495e-05,  2.4325e-03,
        -2.5980e-04,  1.3570e-04, -1.7392e-04,  1.7670e-04,  1.2093e-02,
         3.0228e-05,  1.4526e-04, -1.7689e-05, -2.2291e-04,  3.8691e-05,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2362e-04, -3.7024e-01, -6.7429e-05,  1.8513e-02,  1.5564e-05,
         1.7735e-04, -1.3857e-04,  3.2376e-05, -1.4627e-01,  2.5218e-04,
        -1.5834e-04, -1.8385e-04, -8.8402e-02,  1.2552e-04,  2.8966e-03,
        -1.9730e-04, -1.4653e-01, -3.3417e-04, -2.2474e-01,  1.4217e-04,
        -1.3329e-04, -2.8522e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4391e-04, -4.1719e-01, -4.8083e-04, -1.1060e-04,  7.8762e-03,
         3.7104e-05, -1.5651e-01, -1.7544e-04,  1.2414e-02, -1.3620e-04,
         1.7831e-04,  1.3418e-04,  3.3170e-04, -1.9710e-01, -4.7791e-05,
        -6.5732e-02,  3.8966e-05, -1.2625e-04, -1.4068e-01,  2.6922e-05,
         1.0004e-04,  3.2111e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6710e-04, -4.0839e-01, -2.4360e-04,  2.0337e-06,  3.1287e-03,
        -1.8945e-04, -1.4481e-01, -9.7021e-05, -1.4458e-01,  8.5335e-05,
         6.5690e-05, -4.6866e-05,  9.1859e-05,  8.8955e-05, -1.4108e-05,
        -4.0799e-02, -1.9674e-04,  3.7675e-03, -2.4208e-01,  3.0376e-04,
         1.0427e-02,  4.8685e-05,  1.6724e-04,  1.0781e-05, -3.2844e-05,
        -7.3864e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.5337e-05, -4.9592e-01, -1.1041e-04, -2.5435e-04,  3.7414e-03,
         5.9083e-05, -1.7183e-01,  2.6584e-04,  2.7568e-02,  9.8908e-05,
         1.8456e-04,  8.1050e-05,  2.0708e-04, -1.3435e-01, -1.1759e-04,
         1.8778e-04, -1.6669e-04,  1.2387e-04,  1.0028e-05,  6.7704e-03,
        -1.5757e-01, -2.4008e-04, -3.1111e-05,  9.0947e-06,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.2771e-04,  6.6757e-01,  3.2472e-05, -3.1001e-03,  3.0701e-04,
        -7.0897e-02, -9.4507e-05, -5.1748e-03, -1.2874e-04,  2.5572e-04,
        -3.8506e-04,  5.9470e-05, -4.3880e-05,  1.7995e-01, -2.7688e-04,
         6.7537e-02,  2.1500e-04, -1.5225e-04, -2.1835e-03, -4.9317e-04,
        -3.6190e-04, -2.9175e-05,  2.5954e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3216e-03,  4.5285e-01,  4.9179e-04, -3.8448e-03,  1.9328e-04,
         1.2384e-01,  6.1411e-05, -1.8873e-02, -2.6319e-04,  3.4576e-04,
         1.4509e-04,  9.8693e-06, -7.4707e-05,  3.3194e-01,  4.6089e-04,
         2.8157e-02,  6.6693e-04, -6.6555e-05,  3.4729e-02,  2.2817e-04,
         1.3257e-04,  1.1680e-03,  8.6226e-05,  5.6293e-05,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4743e-04, -3.7150e-01,  1.6035e-04,  2.1939e-03,  2.0104e-05,
        -3.6513e-01, -4.1908e-05,  1.4970e-02,  5.1854e-04,  6.8292e-05,
         7.7465e-05, -1.5310e-04,  8.3251e-05, -1.8478e-01,  1.6982e-04,
        -4.1438e-02,  1.1633e-04,  1.7947e-02,  1.1793e-05,  2.0772e-05,
         2.5082e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2704e-03, -6.2163e-05,  2.4343e-05, -2.5019e-03,  3.3666e-04,
        -2.2876e-04, -5.5962e-04, -1.0271e-02, -1.2600e-04,  7.5252e-05,
         2.3463e-04,  4.3868e-01, -1.9497e-04, -1.9334e-04, -2.3411e-04,
         1.4291e-05, -3.8358e-04,  5.7651e-05,  5.4416e-01, -2.8917e-04,
        -9.9063e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7984e-04, -5.2444e-06,  2.7617e-05,  7.8068e-04, -8.9446e-05,
        -9.7514e-06, -1.3754e-04,  5.4246e-03, -7.4959e-05, -3.0163e-04,
         3.8567e-05, -2.8018e-01, -1.2500e-04, -2.0068e-04,  6.6874e-05,
        -1.3979e-01, -2.0380e-04, -2.3900e-01, -4.8428e-04,  7.1315e-05,
        -3.3223e-01, -2.7214e-04,  1.0730e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.9806e-04,  4.1270e-04,  4.5888e-04,  5.1037e-04,  5.8336e-04,
        -4.9701e-04,  2.2276e-04,  4.0521e-03, -2.7581e-05, -4.5517e-05,
         3.8920e-04,  1.5336e-01,  3.2290e-04,  3.3799e-04, -5.6026e-05,
        -4.2214e-01,  5.4551e-04, -4.1431e-01,  6.8142e-04,  1.4350e-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 #3250: [tensor([ 1.3635e-03, -4.0397e-04,  2.8253e-05,  6.8441e-03, -7.2022e-05,
        -2.4831e-04, -1.2735e-04, -1.7419e-04, -6.5926e-02, -2.2883e-05,
         1.3448e-04, -6.0366e-01,  1.3566e-04, -1.8369e-04, -3.8865e-06,
        -1.3994e-01,  2.2978e-04,  4.4679e-03, -1.6499e-04, -1.7454e-01,
         4.0881e-04,  2.0948e-04, -3.2275e-05,  2.5448e-04, -3.0967e-05,
        -1.6854e-04,  2.3306e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4778e-03, -1.6757e-04,  1.7026e-04,  1.2941e-03,  2.3585e-04,
        -1.1848e-04,  3.6457e-04,  2.9770e-04, -2.5129e-01,  1.8409e-05,
         2.1839e-04,  1.3427e-01,  3.1861e-04, -1.1451e-04,  1.4188e-04,
         1.7684e-02, -2.3344e-01,  2.8813e-04, -9.1177e-05, -3.5794e-01,
         2.8357e-05,  2.6065e-05,  4.0602e-06,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.6593e-04,  5.2296e-04,  4.7713e-06, -9.1527e-04, -2.2171e-04,
         5.3629e-05,  6.7961e-05,  4.0473e-04,  1.3948e-01,  1.7114e-04,
         3.5825e-05,  6.1475e-01,  6.6035e-05,  3.3672e-05, -1.6359e-04,
        -4.2545e-05,  2.4181e-01, -2.6814e-05, -3.4236e-04,  1.3640e-04,
        -2.8723e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6345e-03,  2.0386e-04,  2.1323e-03, -1.6061e-04,  1.5638e-04,
        -2.2298e-04, -2.8344e-04,  1.6672e-05, -9.7162e-01, -2.5639e-04,
        -1.3515e-04, -2.7875e-04,  1.5079e-04,  8.2560e-03,  8.1342e-05,
         1.6678e-03, -8.6951e-05, -3.0155e-04,  2.0427e-04, -1.0670e-04,
         9.5745e-05,  5.9452e-05, -2.2040e-04, -2.0697e-04, -3.3708e-04,
         7.1761e-05, -2.8887e-04,  9.9802e-03,  4.2433e-06, -4.8240e-04,
        -2.2567e-04,  6.5666e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0070e-03, -9.1781e-05, -6.1951e-04, -2.8386e-04,  1.1154e-04,
        -3.2528e-05, -1.2705e-04, -4.2186e-05,  5.0096e-01, -1.0192e-04,
        -2.0401e-04, -3.9896e-05, -5.7638e-05, -2.3759e-02,  2.1492e-05,
        -2.6664e-04, -3.4296e-05,  1.9663e-04,  6.3092e-05,  6.5867e-06,
         6.9659e-05, -1.5282e-04,  7.8281e-05, -7.1624e-05, -2.6562e-05,
        -2.5708e-05, -3.5157e-05,  2.8198e-05, -1.1288e-04, -8.9948e-05,
        -6.0602e-05,  4.4438e-01,  1.4795e-05, -1.3466e-04, -5.0274e-05,
        -3.3393e-04, -2.3789e-05,  8.2639e-05, -1.5503e-04, -2.1778e-02,
        -2.4771e-04,  1.9812e-04, -1.4236e-03, -9.0505e-05, -6.4972e-04,
        -1.1434e-04,  1.2369e-04,  9.4441e-05, -1.0531e-07,  1.3499e-04,
        -7.0844e-05,  1.2785e-04, -1.6305e-07], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5986e-03, -9.5328e-05,  4.7031e-04,  2.3152e-04, -2.6315e-04,
         5.5498e-05,  2.2061e-04,  5.6354e-04, -2.7146e-01, -1.8533e-04,
        -6.2853e-05,  3.9056e-04,  5.5362e-04,  4.6391e-03,  4.2975e-05,
         5.4592e-03,  2.6071e-05, -7.7293e-05,  4.6027e-04,  2.8246e-04,
         9.0554e-05,  1.9378e-04,  7.5149e-05, -6.8564e-05,  2.9215e-04,
         2.1042e-04, -2.4048e-04,  2.2997e-04,  4.1970e-04, -7.0687e-01,
         6.6603e-05,  4.7115e-04,  4.6869e-04,  1.1856e-04,  1.3782e-04,
         1.7677e-04,  1.5936e-04,  1.0125e-03,  3.2397e-04,  6.4169e-05,
        -1.7504e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.8779e-05, -3.9666e-01,  1.4077e-04,  8.1206e-03,  2.6829e-05,
        -1.3433e-04,  2.4961e-05,  6.3497e-05,  4.2819e-05, -2.6915e-01,
         1.5676e-04,  6.6016e-02,  1.2701e-04, -1.2286e-01, -4.3602e-05,
        -1.3609e-01,  1.4265e-04, -1.5629e-05,  1.3533e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.8862e-04, -4.0970e-01, -9.5526e-06,  2.0514e-02,  9.5055e-05,
         1.2483e-04,  9.8905e-06, -4.0970e-05,  1.4026e-04, -1.4554e-01,
         9.4589e-05,  3.1061e-02, -7.6717e-05, -1.0234e-01, -8.7651e-05,
        -4.7763e-02,  2.1130e-05,  9.0267e-05, -2.4144e-01,  7.1113e-05,
        -5.6440e-05,  2.3916e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0482e-05, -2.8129e-01, -1.1225e-04,  1.5566e-02, -2.9571e-04,
        -1.6709e-04, -1.0985e-04,  1.8400e-04, -3.1317e-04, -4.0837e-01,
         2.4360e-05,  2.4806e-02,  1.4958e-04, -2.6787e-01,  4.3589e-04,
        -1.8522e-04, -1.0925e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2983e-04, -5.7901e-05, -1.1540e-04,  1.3898e-03, -3.9910e-01,
        -2.5822e-05,  1.8727e-05, -4.4358e-05, -1.2439e-04, -3.0848e-01,
        -2.1470e-04,  1.1521e-04,  1.2544e-04, -1.4244e-04, -7.7906e-05,
         5.1205e-03, -1.0122e-01,  2.5574e-05, -1.6619e-01,  1.0448e-04,
        -4.5481e-05, -1.5628e-04,  1.6066e-02, -1.1033e-04, -2.5641e-04,
        -4.1341e-04,  1.8798e-05, -1.1023e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3942e-05,  2.0216e-04,  5.5707e-05,  7.7907e-04, -5.2843e-01,
        -2.1425e-04, -1.9738e-04, -1.4741e-04, -2.2726e-04, -3.3517e-01,
        -1.7511e-04,  1.4222e-04, -3.0111e-04,  9.6488e-05,  1.2507e-04,
        -1.3321e-01, -7.2999e-05, -1.7829e-04,  2.0851e-05, -1.9210e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4468e-03,  1.9531e-04, -2.8532e-04, -1.6178e-03, -2.3960e-01,
         3.0042e-04,  2.3834e-04,  2.4017e-04,  9.4778e-05,  6.4059e-01,
         1.6793e-04,  4.2761e-05,  2.0250e-04, -3.6950e-04,  6.8880e-04,
         1.1304e-01,  4.7006e-04,  1.6515e-04,  2.4719e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
