Iter #50: [tensor([ 4.6949e-01, -1.2366e-01,  9.2846e-03,  7.7967e-03, -1.5208e-02,
         5.4041e-03, -1.5460e-02, -9.4877e-03,  5.9678e-04,  1.5835e-02,
         8.5449e-03,  8.0547e-03, -1.0303e-02, -2.3055e-02, -6.7685e-03,
         7.3008e-03,  1.1506e-02,  1.2609e-04,  8.4023e-03,  9.1056e-03,
         3.7319e-03, -3.8648e-03,  1.3359e-02,  3.9728e-03, -3.6880e-03,
         1.3359e-02,  3.9772e-03, -5.0910e-03,  1.9907e-03, -1.2174e-02,
         7.0955e-04, -2.7350e-02, -7.7394e-03, -9.7766e-03,  2.2529e-02,
        -5.2274e-03,  6.0934e-03, -3.3035e-03,  8.9377e-03, -1.2176e-02,
         1.3122e-02, -2.9118e-02,  1.9679e-03,  1.1749e-02,  6.9614e-03,
         2.6333e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1597, -0.3899, -0.0398, -0.0115, -0.0167,  0.0689, -0.0279, -0.0113,
        -0.0349,  0.0145, -0.0584,  0.0147,  0.0540, -0.0370, -0.0273, -0.0175,
        -0.0158,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3405e-01, -2.4578e-01,  7.2661e-03,  1.2227e-02,  1.4828e-02,
         3.1097e-02, -2.6330e-02, -9.0148e-03,  3.2537e-02,  1.1514e-02,
        -2.4836e-02,  6.6018e-03, -1.1164e-02, -6.4198e-03,  3.4373e-02,
         3.4064e-02,  2.2571e-04, -1.7793e-04,  1.0117e-02,  3.3110e-02,
         1.9809e-02,  3.9061e-04,  2.2676e-02, -1.7980e-02, -5.6662e-03,
        -6.2928e-03, -1.5745e-03,  6.3386e-03,  3.1789e-02, -1.2126e-02,
         4.0681e-04,  1.6520e-02, -8.8036e-03, -4.3183e-02, -1.7115e-02,
         3.6020e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1407, -0.4917, -0.0164, -0.0346, -0.0071,  0.0706,  0.0013, -0.0111,
        -0.0131,  0.0217, -0.0104,  0.0016,  0.0088, -0.0642, -0.0212, -0.0018,
         0.0084,  0.0142,  0.0143,  0.0468,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4341e-01, -2.6236e-01, -1.4784e-02, -9.7185e-03,  6.7230e-03,
        -1.9443e-02,  6.6888e-07, -3.3094e-03, -8.5341e-03,  1.0837e-02,
         3.9474e-03, -3.1287e-02,  3.3703e-02, -1.6179e-02,  1.0908e-02,
         1.2318e-02, -1.5020e-02,  1.2411e-02, -3.1345e-02, -1.7231e-02,
        -6.2716e-03, -9.0801e-04, -5.2611e-02,  2.9738e-03,  1.6772e-02,
        -1.2058e-02,  7.9079e-03, -1.5627e-03, -3.5461e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.5013, -0.1644, -0.0075, -0.0006, -0.0108, -0.0107,  0.0005, -0.0033,
        -0.0102, -0.0029, -0.0126,  0.0046, -0.0116,  0.0090,  0.0167,  0.0099,
         0.0028, -0.0291, -0.0309, -0.0033, -0.0006,  0.0338,  0.0054,  0.0173,
        -0.0039, -0.0182, -0.0113, -0.0017, -0.0025, -0.0103, -0.0077, -0.0165,
        -0.0139, -0.0029, -0.0109, -0.0005,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.3796, -0.3145, -0.0682,  0.0344,  0.0125, -0.0069,  0.0259,  0.0102,
        -0.0123,  0.0010, -0.0634, -0.0047,  0.0094,  0.0204,  0.0027, -0.0004,
         0.0007,  0.0052, -0.0055, -0.0220,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.1720e-01, -1.8453e-01, -2.3351e-02,  1.3099e-02,  1.4506e-02,
        -3.7919e-03,  1.7134e-02,  8.3757e-04, -2.0844e-02, -1.5327e-03,
         8.3258e-03, -8.3736e-03,  7.0008e-04, -3.1467e-03,  9.0070e-03,
         1.7438e-03, -2.9237e-02, -1.2964e-02, -1.5145e-02, -1.2327e-03,
         2.6314e-03,  6.3843e-03, -2.8711e-02,  4.8647e-03,  9.6243e-03,
         2.5958e-03, -1.1090e-02, -2.1669e-03,  2.3789e-03, -1.5563e-02,
        -5.6688e-03, -2.4595e-03, -7.7125e-03, -1.1499e-02,  7.9974e-05,
         1.8457e-02, -9.2159e-03, -6.7360e-03,  7.8742e-03, -1.0185e-02,
        -2.3998e-02,  1.2026e-02, -1.9868e-03, -9.3878e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.5525e-01, -1.2260e-01,  1.0359e-02, -9.7760e-03,  9.7287e-03,
        -3.4631e-03,  2.2813e-03,  9.0015e-03, -2.3088e-03,  6.4550e-03,
        -4.9622e-03,  6.3822e-03, -6.9067e-03, -2.0325e-02,  9.4859e-04,
        -3.2601e-04, -1.6736e-03,  5.7110e-04,  8.7275e-03,  2.3149e-03,
         1.8855e-02,  1.1001e-03,  2.0082e-03, -1.8142e-02,  1.7478e-02,
         2.0845e-02,  4.4390e-03, -2.3075e-02, -7.6801e-03, -1.3658e-02,
        -3.0450e-03, -2.8858e-03,  1.8514e-02, -1.1599e-02, -2.6592e-02,
         2.2574e-04,  2.0636e-02, -2.2520e-02,  8.7325e-03,  3.4529e-04,
         4.7337e-03, -8.3173e-03,  7.1607e-03,  1.7389e-04,  8.9363e-03,
         8.8130e-03,  3.6597e-04,  2.0710e-03,  1.5978e-02,  5.9800e-03,
         6.6862e-03,  3.4149e-03, -6.3692e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.2983, -0.2586,  0.0202, -0.0005, -0.0176, -0.0052, -0.0078,  0.0155,
         0.0183,  0.0058,  0.0150,  0.0300,  0.0391,  0.0034, -0.0017, -0.0167,
        -0.0043,  0.0095,  0.0124,  0.0212,  0.0033,  0.0247,  0.0102,  0.0139,
         0.0021,  0.0255,  0.0258,  0.0076, -0.0273, -0.0157,  0.0010,  0.0093,
         0.0046, -0.0050, -0.0229,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1615, -0.2160,  0.0093, -0.0092,  0.0170, -0.0514,  0.0486,  0.0217,
         0.0625,  0.0200,  0.0325, -0.0068,  0.0235,  0.0014, -0.0168,  0.0208,
        -0.0100,  0.0062,  0.0259,  0.0222,  0.0269,  0.0036,  0.0198, -0.0305,
         0.0056,  0.0464, -0.0054, -0.0074,  0.0270, -0.0090,  0.0051, -0.0301,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1977, -0.2352, -0.0215,  0.0095, -0.0221, -0.0063,  0.0097, -0.0272,
        -0.0004,  0.0475,  0.0123, -0.0085,  0.0054, -0.0483,  0.0206,  0.0154,
         0.0172, -0.0249, -0.0047, -0.0057,  0.0264,  0.0127,  0.0017, -0.0305,
        -0.0183, -0.0093, -0.0164, -0.0146, -0.0130, -0.0101, -0.0098,  0.0010,
        -0.0094, -0.0100, -0.0389,  0.0079,  0.0007, -0.0037, -0.0080,  0.0071,
        -0.0019,  0.0088,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #100: [tensor([-1.0608e-01,  8.3495e-01,  2.8223e-03,  4.5340e-03, -6.1087e-03,
        -2.0630e-03,  7.8655e-04, -2.2193e-03,  2.8801e-03, -2.1482e-03,
        -7.9456e-03, -7.9812e-03,  1.0045e-03, -1.1903e-02, -1.3004e-03,
         1.2555e-03, -4.0156e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.7656e-02, -8.4907e-01,  2.9448e-03, -3.2710e-03,  2.8784e-03,
         5.0787e-03,  3.5252e-04,  2.4459e-03,  5.6119e-03,  6.9329e-03,
         3.7865e-03, -2.9132e-03,  5.5580e-04,  2.9773e-03,  1.2785e-02,
         2.4928e-03, -3.1207e-03,  3.6272e-04, -1.0338e-03,  1.5841e-03,
         2.6335e-03,  3.1438e-03,  4.3360e-03,  6.7421e-03,  1.1018e-03,
         4.1878e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.3444e-02,  8.8233e-01, -5.6854e-03, -6.5037e-04,  2.1206e-03,
        -1.4465e-02,  6.8491e-03, -1.3968e-02,  2.8840e-03,  3.4542e-03,
         4.1473e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.4730e-02,  7.9490e-01,  2.2997e-03, -3.0121e-04,  1.9638e-02,
        -3.2993e-03, -6.6425e-03,  2.9460e-03, -7.9273e-03, -1.9964e-02,
         3.8889e-03, -2.4879e-03, -1.9037e-02,  9.6515e-03,  5.7208e-03,
         3.0321e-03,  1.1953e-04,  3.4112e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.2276e-02,  8.7569e-01, -1.1259e-02, -5.8430e-03, -1.6058e-03,
        -7.4109e-03, -2.2313e-03, -5.0680e-03, -9.0202e-04, -3.3006e-03,
         7.1578e-05, -4.0352e-03, -6.4953e-03,  2.3620e-03, -4.7606e-03,
        -2.7859e-03, -3.9063e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.3781e-02, -8.3056e-01,  3.9722e-03, -4.5345e-03, -9.9396e-04,
         1.9842e-03, -7.6972e-06,  1.3525e-03, -8.3172e-03, -1.3090e-03,
         2.5145e-04, -3.4769e-03, -5.0467e-03,  1.1438e-02,  1.8164e-03,
         5.7978e-03, -2.8712e-03,  6.6536e-03, -6.4299e-03, -1.4625e-03,
        -4.4141e-03,  3.2111e-03, -4.4777e-03, -1.5696e-03,  1.0854e-03,
        -3.3109e-03, -7.2297e-03, -6.8000e-03, -3.5678e-03, -2.0175e-03,
        -4.0227e-03,  5.6569e-03, -5.8182e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.2234e-02, -8.5342e-01,  2.1257e-03,  1.8058e-03,  6.3304e-03,
         4.6487e-03,  1.9673e-02, -1.7287e-04, -2.2371e-03, -4.7979e-03,
         4.4103e-04,  1.5968e-04,  7.7422e-03,  1.0578e-02, -5.8977e-03,
         1.8478e-02, -3.0103e-03,  5.0273e-03,  1.2182e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5907e-02,  9.3043e-01,  4.9351e-03, -3.8426e-03, -1.7512e-03,
        -9.9782e-03, -8.6905e-03,  8.0223e-04,  3.6602e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7442e-02, -8.6129e-01, -5.0582e-03, -7.2542e-04,  1.2574e-02,
        -1.0281e-02,  1.3708e-02,  2.3022e-03, -1.1086e-03,  1.5909e-02,
        -3.1892e-04,  4.7508e-03, -1.4106e-03, -2.1474e-03,  1.1136e-02,
         1.3540e-02,  2.2720e-03,  3.0589e-03,  1.8388e-03, -2.6125e-03,
         4.4864e-03, -6.1571e-04, -1.4084e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7274e-02, -8.2930e-01,  1.7556e-03, -1.5995e-03,  3.9826e-04,
         1.2583e-04,  1.0533e-03,  8.3201e-03, -2.0379e-03, -1.2907e-03,
         6.1299e-04,  2.6873e-03, -1.3862e-03, -3.7260e-03, -4.8404e-03,
         7.8255e-03,  2.8088e-03, -2.5352e-03, -5.2427e-03, -1.9315e-03,
        -1.0369e-02,  1.4921e-03, -1.1833e-02, -1.6447e-03,  3.6838e-03,
        -4.2603e-03, -2.7333e-03,  4.8913e-03,  8.2207e-04,  4.8757e-03,
         2.6067e-03,  6.7574e-03, -6.5802e-03, -5.9851e-03,  8.6302e-04,
         4.4613e-03,  2.6008e-03, -6.2970e-03,  4.2718e-03,  4.8643e-03,
        -4.8813e-03,  4.7641e-03, -1.2242e-03, -4.8127e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.3017e-02, -7.4936e-01,  5.1593e-03, -4.8818e-03,  2.3362e-02,
         5.9141e-03, -2.6331e-03,  2.5663e-03, -3.9565e-03, -4.0848e-03,
         6.6470e-04,  5.7517e-03,  1.0080e-03,  4.3491e-03,  5.2350e-03,
        -9.0635e-04,  5.6717e-04,  4.5699e-03,  1.4428e-02, -5.6022e-04,
        -9.5083e-04,  2.0110e-02, -5.3503e-03, -9.8937e-03,  1.0260e-02,
        -1.6750e-04,  5.5798e-03,  4.2873e-03, -2.4833e-03,  6.3738e-03,
        -7.9347e-04,  7.7388e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.8906e-02, -7.1683e-01, -4.5910e-03, -1.0004e-02, -2.0230e-03,
         1.5886e-02, -5.1071e-03, -1.7716e-05, -1.0831e-02,  2.0482e-02,
        -1.5641e-03,  5.6016e-03,  2.5079e-03, -1.9847e-02,  3.8615e-05,
        -3.2488e-04,  7.5720e-03,  6.2659e-03, -6.2539e-03, -1.0036e-03,
         1.5708e-03,  1.4986e-03,  2.0204e-02, -3.0092e-03, -1.2007e-03,
        -2.1020e-03,  4.1556e-05, -6.0472e-03, -1.5719e-02, -5.7788e-03,
        -7.1727e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([-4.3148e-03,  9.5270e-01, -9.1650e-04, -1.7996e-03, -2.4931e-03,
        -3.5803e-04,  1.7416e-03, -5.6956e-03, -3.2917e-04, -3.8415e-03,
        -4.9095e-04, -1.7721e-03, -2.8549e-03, -1.7609e-03, -3.2466e-03,
        -5.1274e-03,  1.5806e-03, -1.8331e-03, -2.2150e-04, -2.9718e-03,
        -2.8039e-03, -1.1417e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8081e-03, -9.6206e-01,  7.5758e-04,  3.0574e-04,  2.2825e-04,
         4.6306e-04, -5.6101e-04, -1.6352e-04,  6.6824e-04, -3.2105e-04,
         1.3390e-03,  6.1151e-04,  2.4601e-05,  1.0600e-03,  1.1792e-03,
        -8.2218e-04,  5.7409e-04, -5.0909e-04,  1.4046e-03, -6.2704e-04,
         8.1072e-04, -4.0619e-04, -1.6553e-03,  1.2668e-03, -6.7132e-06,
         9.0611e-04,  3.2462e-04,  5.6471e-04,  2.2149e-03,  2.1617e-03,
        -1.0877e-03,  2.4155e-03, -6.0561e-04,  1.0537e-03,  2.0165e-04,
         5.5728e-04,  8.8729e-04, -1.5726e-04,  2.7323e-04,  2.4864e-04,
         8.3527e-05, -4.2489e-04,  3.5722e-04, -9.4151e-04,  1.9972e-03,
        -1.2460e-03, -8.4428e-04, -8.1589e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2124e-02,  9.5524e-01,  4.1016e-04, -1.6530e-03, -2.8377e-03,
         3.6244e-03, -8.1245e-04,  2.6138e-03,  2.0493e-03,  4.4407e-03,
         7.1742e-03,  2.3490e-03,  2.4814e-03,  1.2222e-03, -9.4191e-04,
        -2.1944e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7248e-03, -9.4744e-01,  2.9147e-03,  1.0729e-03,  1.4382e-03,
         4.5006e-04,  6.8025e-04,  1.6542e-03,  1.7915e-03, -6.3694e-04,
        -4.6203e-05,  8.1663e-05,  3.3830e-04, -7.1797e-04,  2.8598e-05,
         1.5684e-04, -1.0105e-04,  1.0653e-03,  1.9762e-04,  4.1412e-04,
         1.7328e-03,  4.4686e-04,  2.1090e-03,  2.4443e-04, -8.1088e-05,
         8.0631e-04,  1.1043e-03,  4.2029e-04,  3.2031e-04,  1.3433e-03,
         1.4399e-04,  3.6846e-04,  7.1789e-05,  8.6075e-04,  3.1453e-04,
         6.1366e-04,  4.0294e-04, -4.5260e-04,  9.2937e-04, -6.4397e-04,
         4.7538e-04,  1.3282e-03,  4.1654e-04,  9.4730e-04,  1.8925e-03,
         1.5791e-04,  3.0290e-03, -6.5624e-04,  6.9625e-04,  1.9651e-03,
         4.9907e-04,  2.0207e-04,  1.2127e-03,  1.0977e-04, -1.7115e-04,
         1.6992e-03,  1.0016e-04,  8.1153e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.3611e-03, -9.5413e-01,  1.4688e-03,  1.0331e-03, -7.9249e-05,
         4.0289e-04, -3.7842e-04,  1.4943e-03,  1.7517e-03,  3.6390e-04,
         9.0315e-04,  2.6311e-05,  7.6401e-04,  5.6749e-05,  3.9041e-04,
         6.3751e-04,  6.1732e-05,  1.8402e-03, -3.6831e-05,  3.7337e-03,
         3.6551e-04,  2.8249e-03,  2.1485e-04,  1.6765e-03,  7.4182e-04,
         1.3510e-03,  1.3091e-03, -4.4050e-05, -5.8614e-05, -1.2276e-04,
         2.5187e-05, -1.3693e-04,  2.3644e-03, -2.9612e-04, -2.6225e-04,
        -3.6870e-04, -2.4113e-04,  2.5479e-03, -8.7004e-04,  2.7853e-04,
         1.3591e-04,  4.9040e-04, -6.2230e-04,  1.9188e-03,  5.0681e-04,
        -1.4303e-04, -2.9643e-04,  6.4369e-04,  1.2297e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3796e-03,  9.5156e-01, -4.7008e-04, -1.4554e-03, -3.5626e-03,
        -2.1871e-03, -5.9070e-03, -4.2399e-03, -7.1020e-03, -5.1404e-03,
        -2.0629e-03, -7.3227e-03, -2.1594e-03, -2.4548e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6410e-02, -9.3442e-01,  2.9219e-03,  5.8736e-03,  5.6707e-03,
         5.3099e-03, -3.8943e-04,  3.9660e-03,  2.0776e-03,  3.4502e-03,
         2.4663e-03,  4.9554e-03, -1.1234e-04,  3.8133e-03,  4.7576e-04,
         2.2207e-03,  1.0339e-03,  2.1502e-03,  2.2832e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.7938e-03, -9.5649e-01,  5.2033e-04,  1.7121e-03,  4.0195e-03,
        -1.1009e-05,  5.0542e-04,  3.9922e-04,  2.7504e-04,  1.1895e-03,
         9.7725e-04,  3.4013e-03,  4.1319e-03, -1.4840e-04, -7.0739e-04,
         4.8292e-04,  2.2167e-03,  3.7422e-04,  3.7478e-03, -4.1708e-04,
        -6.1501e-04,  2.0651e-04,  1.5270e-03, -9.6665e-04,  1.3819e-03,
         1.5706e-03,  2.1196e-03,  1.2279e-03,  8.6791e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0231,  0.9510,  0.0063, -0.0089, -0.0057, -0.0050,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3471e-02, -9.3993e-01,  8.8026e-04,  8.9864e-05, -7.5424e-04,
         2.3677e-03,  7.6079e-04, -2.2591e-04, -3.2418e-05,  1.3238e-03,
        -9.6757e-04,  7.2030e-05,  1.5194e-04,  4.9701e-04,  8.3678e-04,
        -2.7229e-04, -1.2681e-03,  1.0951e-03,  1.3311e-03,  8.9871e-04,
         6.7464e-04, -9.7234e-04, -3.5388e-04, -5.9763e-04, -3.8223e-04,
         2.2872e-03,  9.0713e-04,  1.1826e-03,  1.6020e-03,  4.6157e-04,
         2.3567e-04,  1.2539e-03,  1.0958e-04, -8.1000e-05,  2.2008e-05,
        -2.0738e-05, -7.3213e-04,  8.9896e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2080e-03,  9.6433e-01,  1.8685e-04, -1.3374e-03, -2.8493e-03,
        -3.4253e-04, -8.9402e-04, -2.2260e-04, -1.2282e-03, -1.2199e-03,
        -7.3348e-04,  6.6915e-04, -3.0987e-03, -1.1012e-03, -1.8605e-03,
        -3.6885e-04,  2.8452e-04, -1.6893e-03, -2.5474e-03, -7.3833e-04,
        -5.1406e-04, -4.7814e-04, -5.8666e-04, -9.1168e-04, -3.1268e-03,
        -2.0135e-03, -4.1070e-03, -1.2136e-03, -1.3974e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0766e-03,  9.5993e-01,  7.6613e-04, -1.7754e-04,  2.2871e-03,
        -9.5473e-04, -4.5150e-04, -7.4184e-04, -8.4417e-04,  1.7626e-03,
        -6.8887e-04,  2.8183e-03, -7.3445e-04,  1.9872e-03, -4.4304e-03,
         9.2823e-04, -8.1953e-04, -5.2620e-03, -2.9130e-04, -2.2898e-03,
        -4.9027e-03,  1.9827e-03, -1.4725e-04, -7.2148e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 6.3917e-03, -9.7271e-01, -6.6234e-04, -1.7243e-04,  3.6790e-04,
        -9.3111e-05, -7.0131e-05,  1.5191e-04, -8.5945e-04, -1.3940e-04,
        -1.0558e-03,  5.1938e-04, -9.9289e-05, -1.6206e-03, -1.5868e-03,
        -3.7145e-04, -1.5670e-03,  1.3275e-03,  2.1508e-04, -6.2676e-04,
        -3.5062e-04,  7.2051e-05,  6.7076e-04, -4.2997e-05,  7.8222e-04,
        -7.5399e-04,  6.8197e-04, -3.1260e-04,  3.0635e-05, -1.1783e-03,
         5.2255e-04,  5.3221e-04,  1.2903e-03,  2.7285e-04, -1.3022e-03,
        -5.9363e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5637e-02,  9.1865e-01,  3.0607e-04, -5.8429e-03,  1.4872e-03,
         1.1960e-03, -7.0434e-04,  6.9068e-04,  2.1538e-05,  7.1065e-04,
         1.6640e-03, -2.2478e-03,  1.7332e-03, -2.1968e-04,  8.5436e-05,
        -5.2867e-05,  1.6381e-03, -6.7081e-04,  1.0940e-03,  3.4304e-04,
        -7.9723e-04,  1.2090e-03,  3.6410e-03, -1.4453e-03, -3.1579e-04,
         1.8808e-03,  2.2111e-03,  3.0778e-03,  2.0960e-03, -7.0907e-04,
         4.1594e-04, -3.2358e-03,  9.1984e-04,  7.5849e-04,  2.1531e-04,
         2.0795e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.2840e-03,  9.6707e-01, -4.2604e-03, -1.3043e-03,  2.2444e-03,
        -1.1596e-03,  1.2394e-03,  5.6757e-04,  1.0385e-03, -3.0641e-03,
        -1.0102e-03,  4.0712e-05,  1.5475e-03, -1.3720e-03, -1.2815e-04,
        -3.7165e-03, -8.3499e-04,  4.2609e-04, -8.3664e-04,  8.5524e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3925e-02,  9.5910e-01,  2.2698e-04, -6.8787e-04, -2.3755e-03,
         2.0720e-03,  4.4513e-04,  9.9684e-04,  1.6030e-03, -1.0165e-03,
         1.5889e-03, -1.2409e-03, -1.0654e-03,  3.2991e-04, -1.5094e-04,
        -9.4155e-04, -6.0494e-04, -1.5528e-03,  8.5676e-04,  4.1319e-04,
         3.0381e-04,  5.9269e-05, -1.3584e-03, -1.1108e-03,  3.5202e-04,
        -7.0524e-04, -2.9741e-04, -4.2513e-04, -1.9395e-04,  1.7321e-04,
        -1.0321e-03,  5.1195e-04,  5.7515e-04,  5.2553e-04,  8.8875e-05,
        -5.6510e-04, -4.1579e-04, -1.1437e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4997e-03,  9.7124e-01, -5.9237e-04,  9.9805e-04, -7.6325e-04,
        -2.1827e-05, -5.9957e-05, -3.0044e-05, -8.9499e-05, -9.3039e-04,
        -4.8448e-04, -6.4685e-04, -2.6125e-04,  1.3640e-04, -1.3797e-03,
         6.9005e-04, -1.5789e-03, -1.2873e-03, -1.3617e-04, -3.1534e-03,
        -4.9109e-04, -1.0860e-03, -1.2534e-03, -1.2127e-04, -4.8783e-04,
         4.5261e-04, -5.1817e-04,  9.9216e-04,  7.5864e-05, -1.0338e-04,
        -8.5869e-04, -1.2204e-03, -6.2113e-04, -7.2061e-04, -1.1341e-03,
         1.3139e-04,  5.6390e-04,  4.5710e-04, -6.2505e-04,  6.9462e-05,
         3.6014e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4910e-03, -9.6447e-01,  9.6241e-04,  2.3619e-03,  1.7651e-03,
         1.6909e-03,  2.6060e-03,  4.2512e-04,  9.7069e-04,  2.0547e-03,
         2.0274e-03,  2.3714e-03,  1.1808e-03, -6.2654e-04,  1.0775e-03,
        -1.1833e-03,  1.2754e-03,  1.3824e-03,  1.5143e-03,  3.4235e-03,
         7.7517e-04, -3.6250e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6535e-02,  9.4678e-01, -2.4847e-03,  4.7869e-04, -3.1785e-03,
         1.6053e-03,  9.8606e-04, -3.4302e-03, -2.8865e-04, -2.1022e-03,
        -1.0892e-03, -6.5543e-04, -1.6788e-04,  2.5089e-03, -2.2042e-03,
        -6.5511e-04, -1.0624e-03,  1.3548e-03, -4.5229e-04,  1.4651e-04,
        -5.1075e-04, -2.4696e-04, -1.4651e-05, -1.6543e-03,  4.8768e-04,
        -7.2131e-04, -4.5720e-05,  6.9781e-05,  3.0446e-04,  1.0892e-03,
         5.0647e-04, -6.1522e-04, -1.3844e-03, -2.3723e-03, -1.8136e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3141e-03, -9.6720e-01,  1.1399e-04,  1.4861e-04,  1.1319e-03,
        -1.0921e-03, -1.0390e-03,  3.8191e-04,  8.3763e-04,  1.9212e-03,
        -2.7461e-04,  1.3751e-04,  7.9752e-04, -7.5940e-04, -1.1353e-03,
         2.0125e-03,  1.6389e-03, -7.9919e-04,  1.2024e-03, -1.4145e-03,
         1.2730e-03, -1.0745e-03,  1.6911e-03,  1.7167e-03,  1.7006e-03,
         2.6533e-03,  9.9544e-04,  1.7596e-04, -1.3687e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.1381e-03, -9.6934e-01,  1.6542e-03,  4.8937e-04,  1.5496e-03,
        -2.4409e-04, -1.2280e-03,  1.3048e-04,  2.1374e-03,  1.9694e-03,
         6.7486e-04,  1.2490e-03, -1.0671e-05,  8.2799e-04,  2.3073e-03,
        -6.4791e-04,  1.9596e-03,  7.9747e-04,  5.6092e-04,  8.5900e-04,
        -1.7065e-04,  2.0801e-04,  1.1336e-03, -1.6619e-03, -5.4029e-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([ 8.8156e-03,  9.6126e-01, -1.0587e-03, -6.3205e-04, -9.1614e-04,
        -2.0737e-03, -6.3407e-04, -6.6773e-04,  3.4616e-04, -3.3075e-04,
        -1.7541e-03, -6.3916e-04, -4.6646e-03, -4.8963e-03, -1.3335e-03,
        -1.4008e-03, -1.0927e-03, -7.8999e-05, -2.7545e-03, -7.0984e-04,
         9.2841e-05, -5.8827e-04, -3.2582e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.6406e-03, -9.7230e-01, -2.5769e-04, -5.3405e-04,  3.4467e-04,
         6.4853e-04,  4.8126e-04, -2.0446e-04,  3.0741e-04,  1.2827e-03,
        -9.8190e-04,  2.3767e-04, -4.2042e-04,  5.6902e-04,  1.5371e-03,
         1.3107e-03,  6.8763e-04, -1.2086e-04, -2.0599e-04, -8.7584e-04,
        -1.4221e-04,  6.5666e-04,  1.1102e-04,  5.2955e-04,  8.6919e-04,
         6.0023e-04,  1.9541e-04,  5.5382e-04,  8.5676e-04,  7.1721e-04,
         3.0154e-04,  1.0019e-03,  1.5178e-04, -2.3256e-04,  2.5878e-04,
        -2.9431e-04, -1.3178e-03, -2.8297e-05, -2.3034e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1558e-03,  9.7187e-01,  2.7895e-03,  2.4234e-04,  1.4288e-03,
         2.8176e-03,  1.5302e-03,  2.6550e-03,  3.5820e-03, -8.2072e-05,
         5.4152e-04,  6.3013e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-1.4585e-02, -9.7097e-01,  2.4693e-04, -2.5658e-04, -1.8952e-04,
         6.3107e-04,  5.5435e-04,  4.3159e-04, -2.9241e-04, -4.1986e-05,
         2.8899e-04, -1.0748e-03, -1.2578e-04, -5.6085e-04, -8.5625e-04,
         4.3574e-04,  6.0565e-05,  1.2963e-03,  1.1675e-03,  2.6391e-04,
         4.4955e-04,  6.2013e-04,  2.4409e-04,  7.9386e-04,  2.4002e-04,
        -8.4767e-04,  1.1612e-03,  7.7999e-04,  3.7461e-04, -1.5604e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3111e-02,  9.6622e-01,  1.3170e-03, -6.4203e-04, -3.5007e-04,
         2.8114e-04, -5.1053e-04,  1.3400e-03, -9.3388e-04, -1.8819e-03,
        -6.8146e-05, -1.3384e-04,  3.9088e-04, -7.0823e-05,  2.4286e-04,
        -7.3559e-05,  4.1691e-04, -3.0626e-04, -6.5131e-04, -5.4043e-04,
         3.5234e-04, -2.3553e-04, -1.1969e-03, -4.9019e-04, -1.5625e-03,
         4.5833e-06, -1.9435e-04, -7.8164e-04, -7.4633e-04, -6.5059e-06,
         3.8754e-04, -1.0872e-03,  4.7567e-04, -8.1455e-04,  2.6324e-04,
        -8.2754e-04, -7.4746e-04,  3.4312e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.7690e-03,  9.6071e-01,  5.5943e-04,  3.4020e-04,  5.7645e-04,
         5.4378e-04,  8.4625e-04,  1.1155e-03,  1.4584e-03,  1.5720e-03,
        -6.6533e-05, -5.6600e-05, -3.3278e-04,  1.3884e-03,  7.9006e-05,
         2.9592e-03,  1.4374e-03,  1.5249e-03,  3.6291e-03, -9.5882e-04,
         6.6965e-04,  1.7223e-03,  2.6621e-03,  9.3945e-04,  4.1753e-04,
         2.5551e-03,  9.4903e-04,  3.4236e-04,  1.4294e-03, -3.8962e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1551e-03,  9.8246e-01, -5.4777e-04, -2.7724e-03,  7.4408e-04,
         4.8742e-04, -6.1402e-04, -1.2054e-03, -2.6200e-03, -1.2112e-03,
        -6.9896e-04,  5.8377e-04,  7.1554e-04, -3.8592e-05, -1.2771e-03,
         3.2195e-06,  1.1066e-03,  9.3546e-04,  8.2509e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8660e-02,  9.5774e-01,  8.5319e-04, -2.2759e-04, -8.3589e-04,
        -1.3867e-03,  8.9554e-04,  3.2710e-04, -1.6273e-03,  1.6129e-04,
         6.0596e-04,  8.4979e-04, -1.3114e-03, -1.3723e-03, -3.6042e-03,
         3.6111e-04, -7.4054e-04,  1.9528e-04,  1.0226e-03,  1.0509e-03,
         6.3728e-04, -5.5295e-04, -1.1473e-03, -1.3979e-03, -1.0408e-03,
         1.3949e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7284e-03, -9.7346e-01,  4.2509e-04,  3.0707e-04,  3.0614e-03,
        -3.3154e-04,  1.1659e-03,  1.9211e-03,  3.4549e-04,  1.0980e-03,
         2.9550e-05,  8.5100e-04,  5.2893e-04,  1.2203e-04,  7.8798e-04,
        -3.9312e-04, -7.0713e-04,  1.4601e-03,  2.6575e-03,  1.1359e-03,
         2.6631e-03, -7.0527e-04, -1.1201e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4521e-04,  9.7487e-01, -1.3009e-03,  1.2061e-03,  1.7379e-03,
        -3.7196e-03, -2.9968e-04,  1.7703e-03, -1.8549e-03, -1.1297e-03,
        -2.3611e-03, -1.2860e-03, -3.1519e-03, -3.0984e-03, -1.7716e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8334e-02, -9.6299e-01, -1.3105e-03, -1.9535e-04,  4.7693e-04,
         3.1027e-04, -8.1603e-04, -3.8837e-05,  4.0307e-05, -1.3702e-03,
         2.7024e-03, -2.4040e-03,  2.3046e-04, -1.9725e-03, -4.5097e-03,
        -2.2009e-03,  9.3570e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.4504e-05,  9.8651e-01, -1.1292e-04,  6.6548e-04, -5.8008e-05,
         1.9249e-04,  1.0165e-03, -1.0192e-03, -1.7039e-03,  1.0527e-03,
        -1.9560e-03, -1.3576e-03, -1.0833e-03,  8.8345e-04,  8.2580e-04,
        -6.2468e-04, -1.2100e-04,  7.4704e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8130e-03,  9.8683e-01, -1.0641e-03, -1.0954e-04, -4.3426e-04,
         1.4893e-04,  9.9408e-04,  2.3252e-04,  3.3545e-04,  7.6690e-04,
         5.4555e-04, -6.8699e-04, -5.1417e-04, -3.6704e-04, -6.2776e-04,
         4.7447e-04, -7.6694e-04, -6.5760e-05, -8.7846e-04,  5.4158e-05,
         1.3971e-04, -3.3212e-04, -4.1223e-05,  1.2773e-04, -2.3400e-04,
         4.1943e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5496e-04, -9.8037e-01,  5.3301e-04,  7.7206e-04,  2.6991e-04,
        -1.7464e-04,  2.0249e-05,  3.3363e-04, -5.5029e-04,  9.1656e-04,
        -1.0212e-03,  2.9266e-04,  7.8195e-04, -1.0887e-03,  9.2973e-05,
        -1.2633e-04,  3.9467e-04,  8.1608e-04,  1.2447e-04, -1.1639e-03,
         3.0926e-04,  4.6306e-04, -1.0159e-04,  4.9165e-04,  2.8977e-04,
        -3.0114e-04,  4.2468e-05,  5.6499e-04, -1.7973e-03,  4.3684e-04,
        -4.3453e-04,  1.0602e-03, -8.8506e-04,  2.0058e-04,  6.3525e-04,
         7.2717e-04,  3.3046e-04,  1.6810e-05, -7.1058e-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([-9.3054e-03, -9.7365e-01, -7.8432e-04, -5.0907e-05, -2.7385e-04,
        -9.5795e-04, -8.2679e-04,  4.1196e-04, -2.1027e-04, -2.8180e-04,
        -2.3399e-04, -6.5360e-05, -3.2504e-04, -2.6906e-04,  1.1895e-03,
         2.1689e-04, -2.0901e-04, -3.0056e-04, -6.5831e-04, -5.5698e-04,
         3.5600e-04,  7.1409e-05,  2.0030e-04,  1.4858e-04,  6.8310e-04,
         7.5286e-05,  2.2164e-04,  3.9585e-04, -3.3089e-04, -3.7174e-05,
        -5.0775e-04,  5.1766e-04, -1.6303e-04, -3.1786e-05, -4.2921e-04,
         1.1245e-04,  1.8948e-04, -1.0123e-03,  2.2604e-04, -1.0167e-03,
        -9.5358e-04,  3.5104e-04,  2.6446e-04, -5.7524e-04, -4.9161e-05,
         3.0064e-04], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #300: [tensor([-1.3357e-02, -9.6583e-01, -9.9164e-04, -8.9847e-04, -4.5603e-04,
        -1.1915e-03,  9.5321e-04,  9.0944e-04, -1.7677e-05, -2.1987e-04,
        -4.5952e-04, -2.0666e-05, -8.0746e-04, -7.0936e-04, -1.1770e-03,
        -4.2839e-04, -2.3053e-04,  5.6122e-04, -9.2263e-04, -1.2611e-04,
        -8.3933e-04,  1.0446e-03, -1.8938e-04, -1.0335e-03, -7.5293e-04,
        -1.7786e-04,  3.7378e-05,  2.2272e-04,  7.0125e-04,  6.6264e-04,
        -8.1060e-04, -7.5892e-04, -4.7936e-04, -3.4447e-04, -3.6327e-04,
         4.4665e-04, -5.5430e-04, -2.7237e-05,  2.9079e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.6568e-03, -9.7150e-01,  4.8898e-04,  2.6966e-04,  1.8366e-03,
         5.5289e-04,  4.5502e-04,  4.2260e-04,  4.4782e-04,  3.5531e-03,
        -1.6848e-03, -5.6710e-05,  5.7889e-04, -1.9666e-04,  4.6502e-04,
         5.2919e-04, -3.7721e-05, -7.3331e-04,  8.8470e-04, -1.2870e-05,
         5.6148e-05,  8.5362e-04, -8.7989e-06,  1.9874e-03,  2.4840e-04,
         1.7046e-04,  2.6543e-05, -1.4988e-03,  7.9028e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.7426e-03, -9.7696e-01,  2.1722e-05,  1.8000e-03,  1.5891e-03,
        -3.5638e-03, -5.6035e-04, -1.3830e-03, -1.9082e-03, -1.8954e-04,
         9.1473e-04,  2.2495e-04, -7.0771e-04,  1.7456e-03, -4.9628e-06,
         6.8806e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.0045e-03, -9.7537e-01,  1.5183e-03,  2.5927e-04,  4.1889e-04,
         6.2699e-04,  8.8389e-04,  3.6206e-04,  1.0481e-03, -1.3889e-03,
        -6.7685e-04,  9.3831e-04, -2.4254e-04, -1.1039e-03, -1.6215e-03,
         6.0408e-04,  5.6378e-04,  7.4460e-04,  1.6492e-03,  4.8807e-04,
         6.2480e-04, -1.1994e-03, -3.3241e-04, -3.2495e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1586e-02,  9.7319e-01,  2.7061e-03, -3.0020e-04,  9.4760e-04,
         1.5124e-03,  5.9598e-04,  1.6819e-03,  2.8634e-03,  1.1739e-03,
        -5.6734e-04,  4.2747e-04,  2.4521e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.5403e-03,  9.7044e-01, -1.4202e-03,  7.3061e-04,  6.5426e-04,
         1.6512e-03, -6.1079e-04, -9.3276e-05, -6.0269e-04,  6.4773e-04,
         1.1290e-03,  2.3198e-05, -7.9216e-04,  7.5515e-04,  5.3842e-04,
         1.3325e-03, -6.9535e-05,  9.9495e-04, -1.3795e-04, -1.0151e-03,
         2.0227e-04, -1.7004e-03, -3.4793e-04,  6.0030e-04,  7.6000e-04,
        -1.2090e-04, -5.2915e-04,  1.3344e-04,  1.9992e-04,  9.9116e-05,
         1.9405e-04,  5.2844e-04,  1.7334e-03,  2.8946e-04,  1.2063e-03,
         4.3539e-04,  4.5242e-04,  1.5922e-03,  2.8858e-04,  1.6868e-04,
        -2.3511e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5117e-03, -9.8617e-01,  3.6886e-05, -5.5554e-04,  7.2009e-05,
        -6.7870e-05,  3.9863e-05, -1.8280e-04, -4.0365e-04,  6.5669e-04,
        -3.2767e-04,  3.6110e-05, -4.8782e-04,  7.8546e-05,  4.6292e-04,
        -1.7319e-04, -7.3117e-04, -2.3841e-04, -2.9053e-04,  1.3655e-04,
        -5.0703e-05, -6.3026e-05, -7.1020e-04,  2.8218e-07,  1.3110e-04,
        -2.3128e-04, -6.4618e-04,  3.6848e-04,  5.3904e-04, -2.0112e-04,
         1.2714e-04,  2.0112e-04, -2.1094e-04,  4.1066e-04, -1.1831e-03,
        -1.1213e-04, -6.5699e-04, -1.4092e-04, -4.7190e-05, -8.0961e-05,
         9.0519e-06, -5.7057e-05,  1.6148e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.3486e-03, -9.7573e-01, -5.2045e-04,  7.4553e-05,  1.4365e-03,
        -7.1132e-04, -2.3288e-05,  6.2602e-05,  1.2839e-03,  1.2424e-03,
         9.1136e-04,  1.8475e-03,  7.1027e-04,  4.8016e-04,  1.6104e-03,
         1.0043e-03,  2.1015e-03,  9.3326e-04, -2.6598e-04,  8.2917e-04,
         3.4122e-04, -4.9179e-04,  3.4470e-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([-1.6418e-02,  9.6056e-01,  1.7829e-03,  4.7806e-04,  2.3947e-03,
         6.6122e-03,  5.8186e-03, -4.2093e-03,  1.7211e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.2099e-04,  9.7689e-01, -5.7865e-05, -1.2305e-03, -1.0428e-03,
         1.3572e-04, -1.6694e-04,  7.3969e-04, -3.1852e-04, -1.8376e-03,
        -1.1064e-03, -8.9548e-04, -9.3454e-04, -3.2766e-04, -3.7813e-04,
        -4.4723e-04,  1.3635e-05,  1.9555e-04,  1.0580e-03,  1.9221e-04,
        -3.6674e-04, -1.5449e-04, -4.8342e-04, -5.3824e-04, -6.2881e-04,
        -1.6313e-04,  3.2700e-04, -5.8599e-04, -1.5779e-03, -7.5392e-04,
         3.8927e-05, -2.0264e-03,  1.0788e-03,  1.0839e-04, -4.2370e-04,
        -1.3725e-04, -4.2878e-04, -1.1779e-03, -1.5109e-04,  1.5998e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.6620e-04,  9.8680e-01, -9.6007e-04,  3.4271e-04, -9.7716e-04,
        -3.0495e-03, -2.0827e-03,  1.4191e-04, -2.2375e-03,  6.7274e-04,
        -1.6780e-04,  6.8517e-04, -1.1117e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0032, -0.9798,  0.0012,  0.0041,  0.0058,  0.0034, -0.0010,  0.0016,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([ 0.0026,  0.9691, -0.0019, -0.0049, -0.0053, -0.0091,  0.0021,  0.0048,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.2891e-03, -9.7204e-01,  1.5837e-03,  9.0321e-04,  5.7104e-04,
         2.4745e-04, -3.3125e-04, -3.7155e-04, -5.3044e-04, -6.3299e-04,
        -6.2504e-04,  8.4837e-04, -3.9706e-04,  6.7396e-04,  9.9698e-04,
         1.3425e-04, -1.5716e-05,  1.2013e-03, -6.3691e-04,  8.9207e-04,
         4.6194e-04,  4.8374e-04, -5.8298e-04, -4.6556e-04, -1.4079e-04,
        -2.9767e-04,  8.0032e-04,  1.1005e-03,  5.0165e-04,  1.1351e-03,
        -4.2079e-05,  2.5181e-04,  8.2775e-05, -1.6940e-06, -7.0830e-04,
         5.7987e-05, -3.4270e-04, -5.6185e-04,  2.0114e-04, -2.7231e-04,
         5.8521e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.1163e-03, -9.8145e-01,  2.4490e-04,  5.1740e-04, -2.2253e-05,
        -2.8371e-04,  5.1674e-04, -1.0815e-04,  3.4466e-04,  2.8664e-04,
        -5.4730e-04, -5.9188e-04, -1.8707e-04, -5.1167e-04,  2.3492e-04,
        -3.0151e-04, -1.8373e-03,  4.4213e-04,  3.9559e-04,  4.7797e-04,
         1.5963e-04, -2.3309e-04,  2.9560e-04,  6.4516e-04,  1.0009e-03,
        -6.3601e-04, -6.2509e-04,  9.4595e-04, -2.0673e-04, -8.3313e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7469e-04,  9.8555e-01, -1.1735e-03, -2.7148e-03,  6.9431e-04,
        -4.1312e-04, -6.2633e-04, -2.5940e-04, -5.6655e-04, -8.2543e-04,
        -1.2997e-03, -1.6828e-04,  5.0669e-04, -7.1362e-04, -1.8383e-05,
        -1.3933e-03, -1.8594e-03,  9.4289e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.9007e-03, -9.7776e-01,  5.5372e-04, -8.8102e-05, -1.0372e-04,
        -6.6222e-05, -8.6523e-05,  5.0625e-04, -1.9443e-04,  1.0556e-05,
         2.2314e-04, -5.0298e-05,  6.4361e-05,  6.2450e-04,  1.2536e-03,
        -1.4187e-04, -1.2672e-04, -1.2917e-04,  3.3339e-04,  1.4260e-04,
        -1.9967e-04,  4.7901e-04,  6.5819e-05,  1.3599e-03,  9.8585e-05,
         1.4167e-04,  1.6927e-05,  3.1177e-04, -2.1661e-05,  2.0110e-04,
         3.7056e-04,  6.6542e-04,  5.5175e-04,  1.4838e-05, -4.3139e-04,
         5.5174e-04,  9.0264e-05,  4.8829e-04,  5.8314e-04,  4.3916e-04,
        -5.1504e-04,  5.9804e-04, -1.6459e-04,  5.6628e-04,  4.7998e-04,
        -3.5160e-04, -5.2603e-04,  3.8462e-04,  3.4476e-04,  4.0628e-04,
         3.4031e-04,  5.0848e-04,  8.3312e-04, -5.3440e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.8314e-03,  9.7846e-01, -1.4448e-03,  4.5506e-04, -1.6523e-03,
        -1.5777e-04,  7.3182e-03, -1.3048e-03,  6.4899e-04, -1.0998e-03,
        -6.2435e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.2127e-04, -9.8820e-01,  6.5632e-04,  1.1965e-04, -1.4096e-04,
         6.4225e-04,  7.0595e-04, -1.9818e-05,  3.6574e-04,  7.8574e-04,
         4.9168e-04, -1.6391e-04, -3.1588e-04,  2.4374e-04,  1.2754e-03,
        -5.3544e-04,  9.0753e-05, -4.9317e-05,  5.9390e-04,  7.4418e-04,
        -5.0588e-04, -1.3111e-03,  1.9853e-04,  1.1449e-04,  2.0453e-04,
        -4.4776e-05,  5.3492e-04, -8.0350e-05,  1.4453e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8710e-03,  9.8206e-01,  1.0339e-03,  1.5168e-03, -1.5026e-03,
         1.3564e-03, -6.0470e-04,  5.6710e-04, -1.7929e-05, -1.4532e-03,
        -6.0103e-04,  2.8331e-03,  9.5192e-04, -9.9648e-04,  1.6318e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.7777e-03,  9.6131e-01,  1.6863e-03, -8.0894e-05,  4.4191e-05,
         7.7082e-04,  6.6447e-04,  6.6108e-04,  1.8096e-03,  1.5869e-03,
        -4.1848e-04,  1.3378e-03, -7.1362e-05,  1.6194e-03,  7.5060e-04,
         1.6063e-03, -8.3565e-04,  1.2677e-03,  3.1705e-03,  5.4572e-04,
         1.3877e-03,  5.5638e-04,  1.0052e-03,  1.7400e-03,  1.1730e-03,
         9.3615e-04,  2.0679e-03,  1.0333e-03, -1.4236e-04, -5.0914e-04,
         1.2344e-03,  9.9193e-04, -3.5559e-04, -8.5280e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0946e-02, -9.6908e-01, -1.4372e-03, -2.6490e-04, -3.7075e-04,
        -6.3795e-04,  7.4369e-04, -1.6721e-04, -2.0713e-04, -1.2109e-03,
         2.7747e-04, -5.1308e-04, -7.1287e-04, -3.2739e-04,  1.0874e-04,
        -1.3939e-04, -3.3701e-04, -7.7091e-04, -9.7736e-04,  8.0293e-04,
         2.6814e-04,  3.6442e-04,  2.6719e-04, -5.5694e-04,  2.9995e-04,
         1.4506e-04,  4.1282e-04, -5.2850e-04, -1.9811e-04,  3.4993e-04,
         1.1411e-03, -7.5511e-04, -7.1798e-04,  2.2620e-04, -1.3246e-03,
        -5.4191e-04, -6.6342e-04, -7.6684e-04, -4.3681e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.7874e-03,  9.6688e-01, -1.3514e-05,  1.8847e-03,  1.7344e-03,
         1.3582e-03,  1.3655e-04,  2.2468e-03,  2.3033e-04,  9.6339e-04,
         2.3095e-04, -1.8559e-03,  1.9207e-04,  5.4685e-04,  4.8349e-04,
         1.7276e-03,  2.4910e-03,  1.5103e-04,  3.0797e-03,  4.5338e-04,
         1.3989e-03,  6.4895e-04,  7.1662e-04, -1.1755e-04,  6.1290e-04,
         6.0811e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.7664e-03, -9.7016e-01, -1.8048e-03,  1.1639e-03,  1.9958e-03,
        -1.2448e-03, -1.7340e-04, -7.3889e-05, -1.1809e-03, -5.2298e-04,
         8.8830e-04, -8.3867e-04,  6.3398e-04, -1.8448e-03, -4.6143e-04,
         1.4886e-04, -9.2579e-04, -1.3996e-03, -2.2920e-03, -3.5320e-04,
        -2.0887e-04, -8.4794e-04, -3.0072e-03,  4.4237e-04,  6.2146e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+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([ 2.4244e-03,  9.8979e-01, -1.2185e-03,  8.0161e-05, -1.4045e-03,
         2.5314e-03, -9.6542e-05, -1.6540e-03,  2.7997e-04, -4.6562e-06,
        -5.1365e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9784e-04, -9.8673e-01, -4.0197e-04, -1.9786e-04, -3.7384e-04,
        -4.8633e-04, -3.1861e-04,  7.0232e-05, -3.5988e-05, -3.5259e-04,
         4.1344e-04, -2.4295e-04, -3.2626e-05,  1.9259e-04, -5.0386e-04,
        -9.0456e-05, -6.3377e-04,  6.8063e-04, -7.6877e-04, -4.5844e-05,
        -1.8750e-04,  7.0969e-05, -2.0979e-04, -2.9870e-04,  1.6440e-04,
        -6.7145e-05, -3.9516e-04, -1.8453e-04, -3.9533e-04, -4.0581e-04,
         1.1995e-04, -5.3881e-04,  7.0083e-05, -4.3544e-05,  1.2195e-04,
        -1.4868e-04, -1.0776e-04, -1.8229e-04,  1.3806e-04, -1.8487e-04,
        -1.2068e-05, -4.8762e-04, -1.5061e-04, -1.0611e-04, -2.8094e-05,
         4.2828e-04, -6.2937e-04, -1.8904e-04, -1.0708e-04, -1.6852e-04,
         5.8539e-04, -2.0315e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8473e-03, -9.7944e-01, -9.6238e-04, -4.5033e-05,  3.7135e-04,
        -1.1748e-04, -1.5440e-04, -1.0777e-03, -1.3015e-03,  1.1237e-04,
        -2.2422e-04, -6.2008e-04,  2.4829e-05,  1.2110e-03, -1.1406e-03,
         1.7292e-03, -3.1783e-04, -1.8692e-04,  1.4771e-03, -6.1855e-04,
        -3.0200e-04,  5.0489e-04, -1.3436e-03, -1.5951e-03,  2.7060e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2763e-03,  9.8267e-01,  1.7284e-05,  6.4645e-04, -6.8178e-04,
         5.9957e-04, -9.9796e-04, -5.1440e-04,  1.6198e-03, -1.7775e-03,
         4.6780e-04,  1.1055e-03,  8.0940e-04, -3.8326e-04,  2.1534e-04,
         1.4021e-03, -4.2174e-04, -1.5649e-03, -1.0696e-03,  6.5809e-04,
         6.2741e-04, -4.7027e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4353e-03,  9.7199e-01, -6.1999e-04,  1.3672e-04,  1.9809e-03,
         6.8257e-04, -2.6122e-04,  2.0468e-03,  3.4591e-04,  9.2595e-04,
         4.9756e-04,  1.6758e-03, -2.2252e-05, -1.1136e-04,  1.5097e-03,
         1.7154e-03,  1.9627e-04,  6.8014e-04,  3.9769e-04,  9.4275e-04,
         1.6836e-03,  6.9455e-04,  7.1601e-04, -5.6408e-04, -2.5319e-04,
         5.7400e-04,  6.7721e-04, -1.7144e-04,  1.5020e-03,  1.0528e-03,
        -1.7924e-04, -7.5452e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1142e-02,  9.7121e-01, -1.3472e-03, -2.9312e-04, -6.2298e-04,
        -4.8440e-04, -2.1483e-04, -7.2000e-04, -1.0632e-03, -3.6006e-04,
        -2.9621e-05, -3.5111e-04,  4.0724e-04, -6.4278e-04,  6.8354e-04,
        -8.7332e-04,  2.3100e-04, -5.7665e-04, -1.3998e-03,  2.3615e-04,
        -9.2791e-05,  1.2895e-04, -2.4840e-05, -4.6249e-04, -5.7639e-04,
        -1.8869e-04, -7.5116e-04,  1.3106e-04, -8.5754e-04,  5.1400e-04,
         4.3654e-04, -1.0964e-03, -4.5901e-04, -4.7257e-04, -9.1961e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.7682e-03, -9.8304e-01,  7.8627e-04,  1.9632e-04, -1.9694e-04,
        -2.0713e-04, -5.0510e-04, -2.8956e-04, -7.5534e-04, -3.0912e-04,
        -4.4962e-04,  4.5002e-04, -6.8513e-04, -2.3246e-04, -2.3835e-04,
        -8.0255e-04, -5.4028e-04,  1.4976e-03,  6.8207e-04, -5.7035e-04,
         3.7259e-05,  7.3561e-04,  7.3377e-06, -1.0024e-03, -4.9008e-04,
         2.4325e-04, -1.5342e-04, -1.2715e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1347e-03, -9.8735e-01,  4.6992e-04,  1.7616e-04,  8.4160e-04,
         7.7319e-04, -3.7536e-04, -1.7597e-03, -5.5414e-04,  8.6934e-04,
         9.7678e-04, -9.0567e-04, -5.6638e-05, -3.5419e-04, -4.0247e-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,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0565e-02,  9.7045e-01, -2.3965e-04, -3.4439e-04,  6.4980e-04,
         4.6601e-05, -6.3104e-04, -6.3365e-04, -4.7177e-04, -9.5953e-04,
         2.1820e-04, -1.2535e-03, -4.4421e-04,  6.2929e-05,  5.2261e-04,
         3.2276e-05, -1.4712e-03, -4.4219e-04, -2.7055e-04, -7.5237e-04,
        -2.5437e-04, -1.0296e-03, -8.3321e-04, -3.1115e-04,  2.7415e-04,
        -4.3025e-04,  3.5576e-05, -2.5083e-04, -1.9060e-04, -2.5418e-03,
        -1.3201e-03, -8.5045e-04, -7.6922e-04,  3.0763e-04,  1.4239e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4473e-03,  9.7788e-01, -5.4070e-04,  1.0722e-03, -2.9860e-04,
        -1.3488e-03, -1.6552e-03, -5.2322e-04,  1.0883e-03, -2.2738e-03,
        -5.8080e-04,  4.1865e-04, -9.6572e-04, -1.9213e-04, -4.1790e-04,
        -1.3623e-03,  1.8002e-04,  4.8191e-04, -6.5887e-04, -1.6428e-03,
        -1.1049e-03, -8.3436e-04, -1.0275e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2376e-03, -9.8139e-01,  6.9265e-04,  5.0037e-04,  9.6321e-05,
        -7.3044e-05,  1.9003e-03, -3.8525e-04,  9.0481e-04,  6.4511e-04,
         3.6514e-04,  1.9397e-04,  6.3682e-04, -2.5229e-04,  3.0699e-04,
         1.0621e-03,  6.0309e-04, -1.0366e-03,  1.1423e-03,  5.5348e-04,
         7.9882e-04,  5.6857e-05,  5.5243e-04,  9.6944e-04, -1.9210e-04,
         6.3288e-04, -5.3333e-05, -7.6189e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.9128e-04,  9.8239e-01, -1.5541e-04,  1.4867e-04, -4.0094e-04,
        -1.0912e-03, -1.5101e-04,  8.3357e-04,  1.8119e-04, -7.1224e-05,
        -1.4422e-03,  3.1767e-04, -8.3073e-04, -1.7883e-04, -5.0908e-04,
         1.9530e-04,  5.6320e-04, -1.0446e-03,  5.5903e-04, -1.1927e-03,
         1.1101e-07,  3.4077e-05, -5.4982e-04, -6.1060e-04, -3.9363e-05,
        -5.9609e-05,  1.4972e-04,  3.1259e-05, -6.2923e-04, -4.4057e-04,
        -1.1668e-03,  1.6344e-04,  3.1575e-04,  1.3850e-04,  1.7422e-04,
        -9.1663e-04, -3.6963e-04,  5.5802e-04, -6.0020e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([-8.1783e-03,  9.7195e-01, -3.6682e-04, -1.5763e-03, -7.9099e-04,
        -1.2724e-03, -1.4489e-03, -9.5515e-04, -4.9285e-04, -1.0064e-03,
        -6.8734e-04, -9.4195e-04, -1.3278e-03,  3.6400e-05,  1.5829e-03,
        -1.4990e-04,  4.5112e-04,  2.1812e-04, -1.2330e-03, -3.4344e-04,
         1.7845e-04,  1.8026e-04, -6.0143e-05, -1.3906e-03,  6.9749e-04,
        -3.6674e-04, -1.1616e-03, -1.5391e-05, -9.4268e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3797e-03,  9.7974e-01, -7.8566e-04,  1.3314e-03,  4.7258e-04,
         4.8038e-04,  2.5637e-04, -9.4877e-04,  9.7844e-04,  1.8943e-03,
        -4.7164e-04,  7.3686e-04, -3.2953e-04,  2.3634e-03,  7.6372e-04,
         7.3302e-04,  3.8572e-04,  1.6997e-03, -6.6488e-04, -6.2855e-04,
         1.8493e-03, -9.8375e-04,  1.2699e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4584e-03, -9.7973e-01,  5.1859e-04,  2.7111e-04, -1.3256e-04,
         3.4997e-04,  3.6934e-04, -1.3976e-04, -2.3720e-04, -4.2633e-04,
         1.1411e-05,  1.0300e-03, -5.0576e-04, -3.7309e-04, -2.4654e-04,
         4.4886e-04, -9.2506e-04, -5.8112e-04, -4.6400e-04, -1.2141e-03,
        -2.7196e-04,  3.4828e-04, -1.0957e-04, -1.1900e-03,  3.1369e-04,
         1.7924e-04, -2.9626e-04, -2.2521e-04,  6.5585e-05,  5.8751e-04,
         1.7066e-04,  1.0017e-03, -7.3860e-04,  4.8312e-05, -1.0232e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.7052e-03,  9.7624e-01,  2.8647e-04, -7.9711e-04,  2.4469e-04,
        -1.3881e-03, -5.3558e-04,  5.7224e-05, -4.5733e-04,  7.5780e-05,
         5.9555e-05, -2.7384e-04,  1.5472e-04, -1.2596e-03, -3.8883e-04,
        -1.7888e-04, -2.3339e-04, -2.4956e-04,  2.7591e-04, -1.7813e-04,
         3.5940e-04, -8.6585e-04, -1.4543e-04, -5.6111e-04, -2.9600e-04,
         2.5042e-04, -4.2296e-04, -9.9283e-05,  6.2133e-04,  1.5284e-05,
        -9.8445e-04, -7.2446e-04,  1.1068e-04,  7.5715e-05,  1.5866e-04,
         6.6334e-04, -1.9933e-04, -1.3034e-04, -1.4023e-04,  5.9043e-05,
         1.0743e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1941e-03, -9.8471e-01, -5.8170e-04, -5.2767e-04,  1.9947e-04,
         1.3567e-04,  1.0233e-05,  1.2196e-04,  4.2543e-04,  8.8763e-05,
        -1.3186e-04, -1.4341e-04, -1.2687e-04, -9.7415e-04,  2.5743e-05,
         2.8037e-04, -2.2921e-04,  1.2717e-03,  4.6348e-04, -3.3115e-04,
        -7.0531e-04, -7.7351e-04, -2.8717e-04,  4.5024e-04, -2.1369e-04,
        -3.9329e-04,  2.6759e-04, -1.6719e-04,  2.2591e-04, -1.0267e-03,
         3.7923e-04,  1.9444e-04, -1.6729e-04,  6.6469e-04, -4.9881e-04,
         3.8813e-04,  1.1267e-04, -1.1237e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2659e-03, -9.8479e-01, -5.5255e-04, -5.9681e-04, -3.9523e-04,
        -6.6796e-04, -8.1676e-04, -1.2929e-04, -6.7186e-04, -2.4738e-05,
        -8.8042e-04, -4.6264e-04,  1.1266e-04, -1.1565e-04,  4.0732e-04,
         1.1627e-04, -2.0933e-04, -1.1229e-04, -8.3801e-04, -1.9763e-04,
        -8.7470e-05,  5.1375e-04,  2.9336e-04, -3.0256e-04,  4.7526e-04,
        -1.1544e-03, -9.7313e-04, -5.5267e-04, -5.4730e-05, -1.5548e-04,
         5.3396e-04,  1.2878e-04, -4.1482e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.8031e-03,  9.8216e-01,  2.5452e-04,  1.7894e-04, -1.1153e-04,
        -4.9113e-04, -8.6544e-04, -1.7863e-03, -9.4980e-04, -1.9069e-04,
         4.9014e-05,  2.9047e-04,  3.3909e-05,  6.4997e-05,  8.4011e-05,
         6.1251e-04, -1.3325e-03,  5.0920e-05, -1.4601e-03, -1.4451e-03,
        -1.9476e-03, -3.9228e-04,  4.4055e-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,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1237e-03, -9.7902e-01, -2.8708e-03, -1.7074e-04,  8.6879e-04,
        -1.2195e-03, -1.8236e-03, -2.6018e-04,  8.2673e-04, -3.8306e-05,
        -1.1593e-03,  1.3813e-04, -1.3222e-03, -2.5060e-03,  2.2702e-03,
        -6.9696e-04, -3.9011e-04, -2.9349e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.9540e-03,  9.8116e-01, -3.4326e-04,  1.7920e-04, -1.8405e-04,
         3.0480e-04,  1.8394e-04,  5.0644e-04,  7.9895e-05,  1.9134e-04,
         1.4895e-03,  4.8120e-04, -1.1246e-06, -7.7758e-04,  4.6491e-04,
         1.9630e-04,  9.0698e-04,  7.3155e-04,  5.2245e-04,  3.6378e-04,
         1.2071e-03,  1.5147e-03, -3.0123e-04, -6.9001e-04,  3.6910e-04,
        -4.8577e-06, -9.5528e-05, -7.9547e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.3435e-03,  9.8214e-01,  1.7871e-04, -3.3385e-05,  2.3571e-04,
        -1.0185e-03, -8.4090e-04, -4.3764e-05,  7.5539e-04, -5.6569e-04,
         6.3264e-04,  1.1925e-04,  5.2751e-04, -1.1411e-03, -3.6235e-04,
        -8.5065e-04, -5.4703e-04, -1.2128e-04, -1.7001e-04, -4.5737e-04,
        -7.8616e-05,  3.9428e-05,  7.9880e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0326e-03, -9.8567e-01,  8.6944e-05, -2.8326e-06,  9.0060e-05,
        -4.5078e-04,  4.4871e-04, -2.7221e-04, -2.6235e-05,  1.2914e-04,
         9.2709e-05,  5.3700e-04,  2.9485e-04,  2.4022e-04,  1.9993e-04,
         5.6815e-04,  2.7838e-04, -3.5315e-04,  1.6005e-04,  4.9106e-05,
        -1.7638e-04,  2.1982e-04, -8.7236e-05,  1.8899e-04,  6.1231e-04,
        -1.2924e-04, -7.5627e-05,  1.4510e-04, -3.8789e-04, -2.2986e-04,
         2.9446e-04, -1.3599e-04,  1.5217e-04,  5.9680e-05, -5.2950e-05,
         7.6516e-05, -1.1618e-04, -7.4573e-04,  1.2611e-04,  3.0192e-04,
         3.4147e-04,  5.1333e-04,  5.3594e-05, -4.4431e-04, -3.1674e-04,
        -2.4433e-04, -7.9227e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1302e-03, -9.8388e-01,  1.6409e-03,  1.4529e-03, -1.8938e-03,
         6.8937e-04,  6.8152e-04,  6.0964e-04, -3.0311e-03, -1.0110e-03,
        -2.0936e-05, -1.0956e-03,  1.1417e-03,  3.3406e-04,  1.7255e-04,
        -1.2167e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 9.9446e-05,  9.8716e-01, -2.6405e-05,  2.1487e-04,  8.3438e-04,
        -5.8049e-04, -7.3359e-05,  4.8958e-04, -5.3678e-05, -5.6106e-04,
        -4.6045e-04, -3.1716e-04, -4.2519e-04,  1.0892e-03,  4.8480e-04,
        -1.5257e-03, -8.1833e-04, -1.3794e-03,  3.6373e-04, -7.1084e-04,
         7.7386e-04,  2.7970e-04, -1.2685e-03, -6.0669e-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([-5.2010e-03, -9.8177e-01,  1.5384e-04,  2.9435e-04,  1.3544e-03,
         3.2008e-04,  2.7343e-04, -2.8025e-04,  2.1145e-04,  2.2819e-04,
        -3.7918e-04, -3.9246e-07, -6.6953e-04,  1.5109e-03, -9.9752e-04,
        -2.6988e-04,  6.8001e-04, -2.9548e-04,  7.4822e-04,  8.4626e-04,
        -8.2684e-05, -1.7413e-04,  9.1346e-04,  3.8531e-04,  2.3430e-04,
         3.7846e-05,  7.4820e-05,  5.3785e-04,  2.1517e-04, -4.0491e-04,
         4.5977e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.5176e-03, -9.8284e-01, -5.1535e-04, -5.1341e-04, -2.2353e-04,
        -3.9094e-04,  6.7630e-04,  8.4337e-04,  9.9121e-04, -5.8546e-05,
        -1.5785e-03, -4.1260e-04, -6.5637e-04, -9.5621e-04, -8.2249e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.0868e-04,  9.7920e-01,  1.6186e-03,  6.0846e-04, -3.7421e-03,
        -3.2175e-03,  1.2271e-03,  2.1180e-03, -2.2812e-03,  1.7321e-03,
         3.3484e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0253e-03,  9.8336e-01, -1.7481e-04, -6.6698e-04, -9.7731e-05,
         1.1532e-04, -1.6472e-03,  1.9268e-04,  1.1413e-03,  9.3809e-04,
         2.2059e-03, -6.5030e-04,  1.0339e-03,  2.2282e-03, -9.9532e-05,
         1.5035e-03,  7.4556e-04, -6.3959e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4552e-03,  9.7656e-01,  6.8419e-04, -2.7546e-03, -8.0473e-04,
        -1.2979e-03, -2.3087e-03, -9.1426e-04,  3.7451e-04, -2.2893e-03,
        -2.5041e-03, -2.5912e-03, -1.7219e-03, -1.0580e-03, -6.8462e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1911e-03,  9.8343e-01, -1.5797e-04, -1.2228e-03,  5.5797e-04,
         3.2317e-04, -1.2295e-03, -1.5711e-03,  4.7223e-04, -1.0115e-03,
        -3.1639e-04,  5.0970e-05,  1.1908e-03, -1.5511e-03, -1.9458e-04,
         3.7083e-04, -1.3997e-03, -5.1590e-04,  2.3870e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.9833e-03,  9.8168e-01,  4.9581e-04,  9.6500e-04,  3.4275e-04,
         1.1069e-03,  8.1120e-04, -8.6338e-04,  8.6719e-04, -3.8932e-04,
        -2.3788e-04,  5.0025e-04, -1.2411e-04, -6.4003e-04,  8.2526e-04,
         3.4879e-05,  6.7134e-05, -3.3061e-04,  1.0673e-03,  2.9373e-04,
         9.7552e-04, -2.7973e-04, -1.1624e-04, -4.1306e-04, -5.8967e-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([ 6.1029e-03, -9.7707e-01,  2.4717e-04,  8.1019e-04, -2.0602e-03,
        -1.0271e-03, -1.7682e-03, -7.2479e-04, -2.2207e-04,  1.1274e-04,
         1.4343e-03,  1.4218e-03, -1.3932e-03, -1.1224e-03,  1.3209e-03,
        -1.4422e-03, -1.7180e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0971e-03,  9.8969e-01, -8.7480e-04, -4.9377e-05, -5.2281e-06,
         3.5467e-04, -9.8409e-05, -7.2857e-05,  5.9101e-05, -1.0685e-03,
         1.1014e-04,  2.2671e-05, -1.9212e-04,  9.6494e-06,  7.7770e-05,
         4.6561e-04,  8.3591e-05, -8.4538e-05, -3.4281e-04,  1.0856e-04,
        -3.3747e-04,  3.4526e-04, -1.7768e-04,  3.4270e-05, -9.7955e-04,
        -5.3412e-04, -1.7987e-04, -2.6167e-04, -6.4244e-05,  8.4267e-05,
         1.1065e-03, -2.5847e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0383e-03,  9.9030e-01,  2.8745e-04, -3.2326e-04,  4.6212e-04,
        -7.6421e-04, -6.1483e-04,  8.5258e-04,  5.4338e-04,  7.9256e-04,
        -1.9848e-04,  9.9413e-05, -7.6329e-04,  2.8887e-04, -2.1118e-04,
        -7.1694e-04, -5.8128e-04,  4.9480e-04, -6.6705e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1189e-03,  9.8271e-01, -3.2513e-03, -4.0922e-04, -9.5499e-04,
        -1.5930e-03, -1.8426e-03, -2.2983e-03, -5.8589e-05,  5.5917e-04,
        -2.0300e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 4.4781e-03, -9.8489e-01,  5.0852e-05,  2.8143e-04, -1.2333e-04,
         2.1112e-04, -2.8407e-04, -5.4091e-04, -6.3757e-04, -3.1505e-04,
        -3.1665e-04, -1.0037e-04,  7.1983e-04,  6.2691e-05, -4.8114e-04,
        -1.7970e-05, -2.1237e-04, -9.5317e-04,  2.3536e-04,  5.6920e-04,
        -8.3678e-04, -6.7081e-04, -1.2997e-04, -8.0039e-04,  5.8092e-04,
         2.7849e-04, -5.5076e-04, -5.6082e-04, -1.0708e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5525e-03, -9.8293e-01, -3.9722e-04,  2.4844e-04,  3.5674e-04,
        -5.9441e-05, -3.6649e-04, -2.4907e-04, -7.2757e-05, -2.4700e-04,
         3.7295e-05, -1.2858e-04, -1.1855e-04, -6.3096e-04,  4.6607e-04,
         1.1534e-04, -1.0885e-04, -1.8307e-04, -5.1474e-04,  3.5219e-05,
        -5.6381e-04,  5.4490e-05,  9.5052e-05, -2.4750e-04,  1.1592e-04,
         4.5306e-04, -4.2096e-05,  3.3185e-04,  4.1868e-04, -4.0757e-04,
        -5.0033e-04,  5.5739e-04, -5.2201e-04,  2.6670e-04, -9.1068e-04,
        -1.8869e-04, -1.8575e-04,  2.9109e-04, -8.5982e-05, -4.1912e-04,
        -7.4131e-05,  2.7518e-05,  7.4331e-05, -2.3686e-04,  3.7307e-04,
        -2.3821e-05,  4.1068e-04, -6.4929e-04,  2.3281e-04, -5.5767e-04,
         4.5114e-04,  4.1741e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0967e-03, -9.8588e-01,  8.1983e-04, -4.2072e-04,  8.9580e-04,
        -2.1487e-04,  2.5536e-04, -7.0370e-05,  1.9631e-03,  6.5444e-04,
        -1.0914e-04,  6.1152e-05,  7.2878e-04,  6.3228e-04,  2.1338e-04,
         9.7092e-04, -3.7485e-06, -5.6622e-04,  4.4241e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4595e-03,  9.8307e-01,  2.4122e-04, -9.8325e-05,  5.3975e-04,
        -1.9376e-03, -2.6153e-04, -8.0229e-04,  2.8605e-04,  1.4994e-03,
         1.0736e-05,  2.8546e-04, -5.7690e-04, -4.5770e-04,  2.7887e-04,
         3.6324e-05, -9.8674e-04, -4.8643e-04,  8.4582e-04, -1.0846e-03,
        -7.2565e-04,  3.5768e-05, -2.4032e-04, -2.3221e-04, -5.2051e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2106e-03, -9.9214e-01,  3.4539e-04,  4.6386e-04, -1.3397e-04,
        -4.9304e-04,  1.9433e-04,  1.9608e-04, -3.0684e-04,  1.2327e-04,
         8.4932e-04, -1.0764e-03,  1.1157e-04, -4.8205e-04, -1.8775e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2081e-03,  9.8344e-01, -8.0620e-04, -1.1948e-03, -7.6888e-04,
        -9.3614e-04,  7.6916e-04,  1.3284e-04,  8.6997e-04, -1.6868e-03,
        -5.1409e-05,  7.4034e-04,  2.9289e-04, -3.6106e-04, -1.7459e-03,
        -9.1315e-04,  9.5087e-05, -3.1790e-04,  1.7939e-03, -2.3074e-04,
        -8.5183e-05,  6.6108e-04, -2.4113e-04, -6.6005e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1897e-03,  9.8090e-01, -1.7294e-03, -1.3790e-03, -2.7906e-03,
        -3.1529e-03,  1.2654e-03,  1.1370e-03, -7.7884e-04,  1.0644e-04,
        -5.7356e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8582e-03, -9.8694e-01, -6.3884e-04, -2.3584e-04,  1.6462e-04,
         2.1723e-04, -4.5064e-04,  2.7419e-05, -3.9752e-04,  4.8339e-04,
        -6.4395e-04, -6.0595e-04, -3.1441e-05, -4.4081e-05,  1.6119e-04,
         3.2871e-04, -5.8955e-04,  1.6251e-04, -7.3415e-05,  2.9408e-04,
         7.8336e-04,  1.3123e-04, -3.4733e-04, -4.0499e-04,  1.2452e-04,
         2.2005e-04,  7.4232e-05, -5.0610e-04,  1.6454e-04, -3.5696e-04,
        -8.1124e-04, -7.2507e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7227e-04, -9.8410e-01,  2.3671e-03,  1.8422e-03, -4.3660e-04,
         1.8665e-03,  8.2076e-04, -7.4164e-04, -1.5566e-03,  1.3711e-03,
         5.6430e-04,  5.6308e-04,  1.0270e-03,  8.7000e-04, -9.2387e-04,
        -5.7237e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4566e-03,  9.7648e-01,  2.6314e-04, -3.5909e-05,  9.4587e-04,
        -1.0854e-04,  4.9615e-04,  1.1960e-03,  1.0381e-04,  6.0642e-04,
         7.6285e-04, -5.8489e-05,  7.4251e-04,  2.5099e-04, -6.5602e-05,
         1.1913e-03, -3.9264e-04, -1.6837e-04,  4.9580e-05, -1.8453e-04,
         1.0207e-03,  5.6218e-04,  6.8738e-05, -8.2089e-04,  1.1339e-03,
         2.2949e-03, -1.1671e-04,  1.8668e-03, -8.9976e-04,  2.6542e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8522e-03, -9.8032e-01,  7.6583e-04, -2.5402e-03,  1.7595e-03,
         2.4467e-03,  1.2560e-04,  2.4762e-04,  7.3323e-04, -3.5321e-04,
        -1.1675e-03,  4.1609e-04, -4.7752e-04, -1.6235e-03, -3.9717e-04,
        -1.5407e-03, -2.1798e-04, -4.1318e-04, -5.9911e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0626e-04, -9.8897e-01,  7.0467e-04, -3.2014e-04, -9.3844e-05,
        -8.3753e-04, -8.0246e-04, -1.1154e-04, -4.2736e-04, -3.9108e-04,
        -3.8182e-04, -7.3239e-04,  5.5350e-04, -7.5830e-04,  5.9875e-04,
         4.6143e-04,  1.1279e-03,  5.9508e-04,  4.5963e-04, -1.0583e-04,
        -1.7141e-04, -7.6941e-05, -1.6775e-04, -9.4331e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 6.1663e-03,  9.7647e-01,  1.6479e-04, -2.2231e-04,  3.0070e-05,
         3.3410e-04, -3.6763e-04, -4.1101e-04, -2.1098e-04,  3.4198e-04,
         1.0465e-04, -4.9823e-04,  8.6278e-04,  4.5732e-04,  2.9520e-04,
         7.1309e-04,  8.5625e-05,  3.2660e-04,  9.3654e-04, -3.1884e-04,
         2.8757e-04,  1.8589e-05,  5.7014e-04,  2.8306e-04,  3.3598e-07,
         6.4555e-04,  1.2396e-04,  2.8592e-04,  4.5154e-04,  6.9880e-04,
         9.0506e-04,  1.3606e-04, -3.8116e-04,  6.3462e-04,  1.0766e-03,
         2.3573e-04,  1.2831e-04,  3.4515e-04,  8.0324e-04,  7.2164e-04,
        -2.4595e-04, -2.1119e-04,  8.2566e-04,  2.8297e-04, -3.0131e-04,
         7.8299e-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([ 1.9103e-03,  9.8576e-01,  6.3969e-04,  2.9595e-04,  1.2460e-04,
        -1.6429e-03,  6.8430e-04,  1.4639e-03, -3.1808e-04, -1.9164e-03,
         7.8241e-04, -2.9684e-04,  1.3068e-03,  3.9898e-04, -5.2595e-04,
        -1.3295e-03, -6.0426e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.9342e-03,  9.7881e-01, -8.4828e-04,  1.9319e-04,  3.5879e-04,
        -8.8721e-04, -6.6963e-05,  1.1707e-03, -7.9474e-04, -3.3738e-04,
        -1.5353e-04,  3.3789e-04, -5.2628e-04, -3.2760e-04,  4.6886e-04,
         4.7838e-04, -1.0143e-04, -5.2465e-04, -1.2281e-04, -7.3635e-04,
        -2.8441e-04,  5.2949e-04,  2.4449e-04, -7.3736e-04, -4.1984e-04,
         2.6819e-04, -1.7976e-04,  2.7040e-04, -1.7056e-04, -5.0346e-04,
         1.2467e-03, -2.2892e-04,  9.2286e-05,  4.0544e-04, -2.3865e-04,
        -2.7475e-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([ 2.9080e-03,  9.8601e-01, -2.9382e-04, -8.6962e-04, -4.6414e-04,
        -2.1048e-03, -7.8049e-04, -5.3624e-04, -9.8022e-04, -2.8139e-04,
         4.8607e-04, -7.0325e-04, -2.1149e-04, -7.8506e-04, -3.4707e-04,
        -8.0362e-04, -7.7298e-04,  2.0940e-04, -4.4820e-04,  7.3442e-06,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7073e-03,  9.8592e-01,  6.1777e-04, -4.1731e-04,  9.5130e-04,
        -1.8011e-05, -1.4871e-04,  5.2822e-04,  4.1059e-04,  5.5912e-05,
        -3.2263e-04,  2.3267e-04,  8.0711e-04,  5.6873e-04,  1.7688e-04,
        -5.2414e-05,  2.7134e-04, -8.4069e-05, -1.3854e-04, -1.3723e-03,
        -1.8216e-04, -3.8643e-04,  8.9755e-04, -1.3786e-04,  1.3522e-03,
         6.3162e-04, -1.6909e-04,  4.3666e-04, -1.7796e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9975e-03, -9.8413e-01,  3.2346e-04,  3.5855e-04,  1.4949e-04,
         1.3256e-04, -8.3960e-05,  2.1677e-04, -1.8849e-04,  2.7111e-04,
         6.5265e-05, -4.2044e-04,  4.7236e-04, -1.5951e-04, -6.8541e-04,
        -5.0108e-04, -6.1889e-04,  3.3330e-04, -8.3497e-04, -1.9375e-04,
         4.5308e-05,  9.0574e-04, -2.1392e-04,  4.8158e-04, -9.1402e-05,
        -6.6985e-04, -7.0167e-04, -9.1026e-04,  7.7061e-04,  3.4915e-04,
         5.4298e-04, -1.8710e-04, -7.0780e-04,  7.5026e-04,  3.1868e-04,
        -2.1288e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.4448e-04,  9.8799e-01,  5.2958e-04,  7.7096e-04,  1.1374e-03,
         1.1941e-03, -5.7137e-04, -5.5500e-04, -7.4613e-05,  4.1860e-04,
         7.9401e-04,  1.1697e-04,  1.5813e-03, -2.7304e-04,  6.4013e-04,
         8.7465e-05,  9.7256e-04,  1.8845e-04,  1.1289e-03,  5.2724e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4073e-04, -9.9161e-01, -8.5479e-05,  7.3713e-05, -3.2237e-04,
        -5.8145e-05,  4.0800e-04, -7.2823e-05,  1.5537e-04, -1.2842e-04,
        -2.1678e-04,  8.6879e-06,  1.0023e-04, -3.0481e-04,  4.0126e-05,
        -1.5554e-04, -9.0206e-05,  2.7480e-04,  3.9190e-04,  2.6686e-04,
        -3.2320e-05, -1.0421e-05, -5.3106e-04,  1.2416e-04, -5.5752e-05,
         1.6093e-04, -1.1206e-04,  1.6269e-04,  3.9801e-04, -1.3228e-04,
        -1.3972e-04, -2.1444e-04, -9.1406e-05,  2.4333e-04, -1.6870e-04,
         2.7829e-04,  2.3937e-04, -5.5598e-05,  5.9814e-04,  2.4078e-04,
         2.1811e-04,  2.6369e-05, -3.7438e-05,  3.2753e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.6680e-03, -9.7508e-01, -6.9857e-04, -3.0279e-04, -6.8972e-04,
        -3.1063e-04, -3.8423e-05, -8.9445e-04, -1.7817e-04, -4.0963e-06,
        -5.8505e-05, -2.8934e-04, -5.5109e-04, -6.4872e-04, -9.5036e-04,
         1.1127e-04, -9.0121e-05, -6.8091e-04, -6.3676e-04, -4.0852e-04,
        -3.5053e-04, -2.5666e-04, -7.5226e-04, -3.9798e-04, -1.5233e-04,
        -1.6479e-04, -1.8109e-04, -3.5924e-04,  4.7862e-06, -6.9817e-04,
        -5.6656e-04, -6.2999e-04, -2.1390e-04, -4.3143e-04, -4.3135e-04,
        -3.4168e-04, -6.9281e-04, -5.2980e-04, -1.1040e-04, -6.5133e-04,
        -9.1612e-05,  1.1302e-04, -6.1634e-04, -3.6750e-04,  6.9326e-05,
        -2.2631e-04, -6.1696e-05, -4.3007e-04, -6.3566e-04, -3.6095e-04,
        -2.8096e-04, -1.0269e-04,  4.3261e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1602e-03, -9.8994e-01,  8.0767e-06, -3.0854e-04,  4.2980e-04,
        -1.6863e-04,  5.0047e-04, -7.6355e-04, -5.7247e-04, -1.1554e-04,
         2.5619e-04, -2.5245e-05, -4.9776e-05, -6.5010e-04,  1.2188e-04,
        -6.0895e-05, -2.8862e-04,  2.8700e-04, -8.3420e-05, -5.7267e-04,
         1.6358e-04, -4.8493e-04,  4.3541e-04,  8.6110e-05,  3.7435e-04,
        -1.6314e-04, -1.3467e-04,  7.5994e-05,  2.9607e-04,  4.3556e-04,
        -1.8604e-04, -4.4200e-04, -2.4094e-04, -1.0606e-04,  1.2451e-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([ 1.0339e-03,  9.8651e-01, -5.7874e-05,  4.3963e-04,  2.9515e-04,
         5.5174e-04,  8.1034e-06, -1.9902e-04, -5.6104e-04, -2.8351e-04,
        -7.4360e-04,  2.4383e-04, -5.6777e-04,  8.2252e-04,  1.0360e-03,
         2.3107e-05, -7.2208e-04, -6.4211e-05,  1.9734e-04, -1.3851e-04,
         3.0547e-04, -5.0826e-04,  2.4206e-04,  1.7153e-04,  8.9686e-05,
         9.9781e-04, -1.0210e-03, -3.3832e-04,  1.1148e-03, -3.3474e-04,
         3.0635e-04, -6.6827e-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([-2.2324e-03, -9.8771e-01, -4.2226e-05,  7.8284e-04,  3.1377e-05,
        -1.7961e-04, -1.7162e-04,  3.5816e-04,  9.0693e-05,  6.3934e-05,
         1.5147e-04, -3.1613e-05,  2.6422e-04, -3.2233e-04,  4.9079e-04,
        -9.2141e-05, -7.6095e-04,  2.5315e-04, -2.9161e-04,  2.5066e-04,
         2.5583e-04, -9.0448e-05,  3.1591e-04, -7.2217e-04,  1.1008e-04,
        -7.6841e-05, -3.4545e-04, -3.0939e-04, -7.0192e-04, -4.6499e-04,
         2.4801e-04, -1.1238e-04,  1.5622e-04,  3.3970e-04,  1.9692e-05,
        -6.4460e-04,  1.3338e-04, -8.9632e-06, -3.8751e-05,  5.5289e-05,
        -1.9141e-04,  8.4753e-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>)]
Iter #100: [tensor([ 1.6453e-03,  9.9267e-01, -1.8178e-04, -7.7757e-05,  1.0458e-04,
        -4.3229e-04, -3.7210e-04, -6.3131e-04, -3.5722e-04,  3.9721e-04,
        -7.2381e-04,  9.4545e-04, -2.8805e-05, -4.5237e-05, -1.0755e-03,
        -2.8224e-04, -3.2957e-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([-7.1994e-04, -9.8992e-01,  2.1938e-04,  4.8287e-04,  7.9024e-04,
        -1.1060e-03, -6.6259e-05,  1.4693e-05,  1.1257e-04,  9.1570e-04,
        -1.4051e-04,  8.1089e-04,  4.8218e-04,  8.9474e-04,  7.9545e-04,
         4.8700e-04,  4.2672e-04,  2.1648e-04,  6.8599e-04,  3.9010e-05,
         1.2095e-04,  1.4252e-04,  1.0335e-04, -2.3907e-04, -2.5198e-05,
         4.3197e-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([-3.7638e-03,  9.8404e-01,  1.2091e-03, -1.3412e-03, -1.5813e-04,
        -1.9278e-03, -1.8632e-03, -1.5462e-03,  1.3363e-03,  1.8886e-03,
         9.2983e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6149e-03,  9.8351e-01,  1.0001e-03, -1.0701e-04, -1.0779e-03,
         3.9554e-04, -3.0104e-04,  3.5976e-04, -1.1538e-03,  1.1043e-03,
        -1.2348e-03,  1.6136e-03,  1.7882e-03,  8.7945e-04, -5.4187e-04,
        -1.7330e-03, -1.5705e-03,  1.2561e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8884e-03,  9.9056e-01,  8.9717e-04,  7.0904e-04, -5.0285e-04,
         1.0974e-04, -7.3418e-04, -7.4063e-05,  1.0731e-04,  1.3088e-03,
        -3.6760e-04, -9.1974e-04,  2.6090e-04, -5.0800e-04, -2.4962e-04,
         6.2693e-04,  1.7120e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3157e-03, -9.8016e-01, -2.8960e-04, -3.2560e-04,  4.2264e-04,
         5.0108e-05, -4.0764e-04, -1.0083e-03,  6.2991e-05,  2.8567e-04,
         2.5237e-04,  6.2125e-04,  4.7982e-04,  7.3582e-04, -5.7183e-04,
        -7.0264e-05, -1.8648e-05,  7.8984e-04, -1.7319e-04,  1.3929e-04,
        -5.6151e-04,  2.2175e-04,  7.8397e-04, -2.1058e-04,  9.9311e-04,
        -2.6598e-04,  5.6290e-04,  8.9556e-04,  2.7019e-04, -1.5061e-04,
         9.5235e-04,  1.8405e-04, -7.6611e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.7031e-03, -9.8839e-01, -4.3410e-04, -9.8505e-05, -2.1514e-04,
        -3.4255e-05,  2.7013e-04, -2.4347e-05, -1.3625e-03,  6.8649e-05,
        -3.6002e-04, -4.3890e-04,  6.4870e-05,  4.7012e-04, -6.0337e-04,
         1.1712e-03, -1.0069e-04, -1.6536e-06, -1.9130e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5786e-03,  9.9217e-01, -3.0899e-04, -8.2193e-04, -1.2618e-03,
        -1.5612e-03, -1.5048e-04,  4.6672e-04, -6.7615e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8167e-03,  9.8599e-01, -7.3710e-04, -2.6906e-04, -1.5919e-03,
        -6.8171e-04, -9.1685e-04, -4.6597e-04, -5.2790e-04, -1.7291e-03,
         2.1007e-04, -5.7696e-04,  3.1438e-05,  6.9629e-06, -7.5166e-04,
        -1.2422e-03, -1.6985e-04,  4.9205e-05, -2.1906e-04, -2.2140e-04,
         2.1171e-04,  1.5289e-04, -4.3355e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7995e-05,  9.8803e-01, -2.6180e-04, -4.9634e-04, -2.5647e-05,
         2.6067e-04, -2.2363e-04, -1.0968e-05, -2.3354e-04, -3.0804e-04,
        -2.7400e-04, -1.3039e-04, -5.7586e-04, -2.2264e-04, -2.2281e-04,
         2.1252e-05, -2.3211e-05, -1.9529e-04, -3.8221e-04, -1.0042e-04,
        -1.3019e-04, -2.7214e-04, -7.8810e-04,  2.6349e-04, -2.3022e-05,
        -6.1929e-04, -2.5867e-04, -1.5727e-04, -8.5546e-04,  5.8534e-04,
        -5.4614e-04,  4.2893e-05, -5.2640e-04, -2.5559e-04, -3.0520e-04,
        -8.9309e-05,  1.4501e-04, -4.0665e-04, -1.2722e-05, -3.7018e-04,
        -8.3871e-04, -2.5995e-04, -1.7747e-04, -4.3121e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6735e-03, -9.8421e-01, -7.6702e-04, -2.5959e-04, -3.9878e-04,
        -3.1988e-04,  3.6048e-04, -1.2204e-03, -2.8021e-04,  2.1258e-04,
        -3.6695e-04,  1.5062e-04, -2.5113e-04,  1.9804e-05,  7.0790e-04,
        -7.4904e-05, -1.9672e-04,  5.7895e-04, -1.0245e-03, -1.9168e-05,
         3.0330e-04, -1.0820e-04,  2.6807e-04,  1.0787e-04,  7.9568e-04,
         9.9776e-04,  2.3834e-04, -9.0651e-04, -5.9891e-04, -1.6298e-04,
        -2.6876e-04,  1.5121e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3294e-03, -9.8597e-01,  1.1840e-04,  2.2482e-04, -4.9519e-04,
         3.9993e-04, -3.7080e-04, -6.2775e-04,  1.5036e-04,  4.5893e-05,
        -6.9095e-04,  6.8947e-04, -4.6353e-05, -3.6909e-04,  7.6171e-05,
        -1.0627e-05, -5.5365e-04, -2.1129e-04, -3.9692e-04,  1.9798e-05,
         1.1633e-03, -1.1697e-04, -3.4704e-05, -4.4534e-04,  4.4828e-04,
        -2.0695e-04,  4.8291e-04,  2.7310e-04,  2.6754e-04, -1.4909e-04,
        -6.1060e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 1.8592e-03,  9.8628e-01, -1.0084e-03,  1.8770e-04, -4.5042e-04,
         9.0074e-04, -1.1401e-03, -4.6449e-04, -3.0785e-04, -4.5405e-04,
        -1.2203e-03, -8.6933e-05, -5.4556e-05,  9.1396e-04,  7.1645e-04,
        -3.7516e-04, -9.9416e-04, -6.9653e-04,  5.3020e-04,  3.5411e-04,
         5.5520e-04,  4.5341e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.9135e-03, -9.8002e-01,  1.9381e-04, -1.5496e-04, -3.6235e-04,
         3.8976e-05, -3.1313e-04,  3.3408e-05,  4.9426e-04,  3.8576e-04,
        -1.5064e-04, -3.8361e-04, -1.5359e-05, -2.6508e-04, -1.3296e-04,
         4.2585e-04,  4.3685e-05,  1.2022e-04, -1.5336e-04,  1.2026e-04,
         6.1764e-04,  2.3580e-04, -6.1139e-04,  6.5657e-04,  1.2030e-04,
        -1.5648e-04, -2.6983e-04,  1.1736e-04,  9.0328e-04,  3.4778e-04,
         4.4117e-04, -2.2281e-04, -5.2600e-05,  3.7640e-04, -7.0631e-04,
         4.7064e-04, -4.0224e-05, -8.2988e-04, -2.3240e-04, -1.3505e-04,
        -6.3334e-04,  3.6578e-04, -2.7530e-04,  2.5071e-05,  1.0164e-04,
        -1.7860e-04, -1.1678e-04, -3.5567e-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([ 6.6994e-03,  9.7983e-01,  1.3731e-03,  3.2776e-04, -6.2906e-04,
        -2.4384e-03, -7.7339e-04, -5.5741e-05, -1.5729e-03,  2.7335e-05,
        -5.8460e-04,  7.2383e-04,  1.3455e-03,  3.5602e-04, -2.9775e-03,
         2.8503e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0719e-03, -9.8456e-01,  5.4412e-04, -3.0957e-04,  3.7893e-05,
        -8.0963e-04, -2.9738e-04, -5.2920e-05, -3.8801e-05, -4.8516e-04,
        -3.2998e-04, -1.9708e-04, -2.8513e-04, -3.3428e-04, -3.7555e-04,
        -1.3138e-04, -5.7730e-04, -5.8644e-05, -1.7707e-04, -2.3158e-04,
        -2.2787e-04, -3.7725e-04, -9.5089e-05,  6.2751e-05, -1.2404e-04,
        -2.1197e-04,  3.8472e-05, -1.7680e-04, -1.8984e-04, -1.6086e-04,
        -1.3006e-04, -1.9712e-04, -2.6367e-04,  4.4015e-05, -4.2550e-04,
         2.8360e-04, -4.2501e-04, -1.1420e-04,  1.0096e-04,  1.7763e-04,
         3.0922e-04, -4.7560e-04, -1.5896e-04,  3.0367e-05,  1.5086e-04,
        -2.1252e-04, -4.9749e-04, -1.7947e-04, -1.2787e-04,  2.6462e-04,
        -1.2995e-04,  1.7145e-04, -2.0682e-05, -6.3572e-04, -3.8094e-04,
        -2.0329e-04,  2.2671e-04,  9.6992e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6055e-03, -9.8655e-01, -2.0795e-04,  2.7094e-04, -9.1647e-05,
        -1.0192e-04, -3.7947e-04, -4.1088e-05, -1.8843e-04, -1.8244e-04,
         2.6028e-04,  2.3606e-04, -4.4465e-04,  4.8972e-04, -2.1407e-04,
        -1.5031e-04, -6.4962e-04,  3.9530e-05,  4.6250e-04,  5.1100e-04,
        -5.8470e-06, -2.4236e-04, -1.1460e-04,  1.8497e-04, -1.8762e-04,
        -2.1373e-04, -4.7615e-04,  2.2250e-05, -5.9796e-04,  6.2401e-05,
        -7.4983e-05, -7.4949e-05,  1.8351e-04, -3.0326e-04, -1.3577e-04,
        -4.0307e-05, -4.4906e-04,  6.8519e-04, -2.1147e-05, -4.5254e-04,
         7.7631e-05, -2.4485e-04,  1.3405e-04, -1.7230e-04, -9.3961e-05,
        -5.5292e-06, -2.2050e-04, -4.2333e-04,  2.5157e-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([-1.2645e-03,  9.8653e-01, -4.4947e-04,  1.0931e-03, -8.6704e-04,
         1.7038e-03, -1.2441e-03, -9.9648e-04,  1.0538e-03, -7.6793e-04,
         2.4048e-04,  1.0418e-03, -7.6543e-04,  1.9851e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1105e-03, -9.8754e-01, -3.1547e-04, -2.7737e-04, -6.3674e-04,
         1.0809e-03, -6.9227e-04, -1.9897e-04,  1.2598e-03, -5.0574e-04,
         7.8318e-04, -5.7016e-04, -8.9659e-04,  7.6193e-04, -2.5888e-04,
         7.7105e-04,  5.2974e-04,  7.7801e-04,  3.6689e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.3427e-04, -9.8986e-01, -2.0917e-04, -8.0228e-04, -5.5425e-05,
         4.8108e-05, -6.3120e-06, -4.7795e-04,  5.7558e-05, -6.6265e-05,
        -1.9815e-04,  7.4150e-06, -6.6259e-05,  9.7105e-04,  2.0252e-04,
        -5.8962e-04, -4.7178e-04, -1.6258e-04,  6.8350e-05, -7.5234e-04,
         3.7665e-04, -6.7635e-04, -5.1022e-04,  3.3140e-04, -4.6350e-05,
        -2.8154e-04,  1.0760e-03,  2.5754e-05,  6.6423e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.2078e-03,  9.8193e-01,  3.1814e-03, -5.7602e-04,  3.0020e-04,
        -6.8055e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.2742e-04, -9.8814e-01,  4.1369e-04, -2.5338e-05,  3.9493e-04,
         2.3177e-04, -3.0006e-04, -3.4214e-04, -5.0184e-04,  1.7759e-05,
        -5.8396e-04,  7.9217e-05,  3.9080e-04,  4.6323e-05,  2.7025e-04,
         6.0636e-04,  5.9610e-04,  5.1767e-05,  7.5497e-04, -3.4454e-04,
        -1.0287e-04, -2.1176e-04, -4.2011e-04,  1.2596e-04,  4.1460e-04,
         3.3514e-04,  3.5212e-04,  2.4799e-04, -2.2954e-04, -2.0294e-04,
         2.1576e-04,  7.3152e-04, -3.0349e-04,  3.2536e-05, -4.4558e-04,
         1.7482e-04,  2.9516e-04, -2.3746e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0083e-03,  9.8979e-01,  3.4533e-04, -5.0142e-04,  1.1278e-04,
         1.6732e-04,  4.7450e-04, -4.8825e-05,  2.7620e-04,  8.0726e-04,
         7.4325e-04,  1.5092e-04,  1.8006e-04,  3.3981e-04, -5.3219e-04,
         7.0420e-04, -1.0771e-04, -3.7359e-04,  1.8160e-04,  3.2238e-04,
        -3.5163e-04, -1.3816e-04,  7.3364e-04, -4.4258e-04, -7.1228e-04,
        -8.8204e-05,  1.9551e-04, -1.2544e-04,  4.3676e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7219e-03, -9.8982e-01, -2.8132e-04,  8.9979e-05, -4.6434e-04,
         6.2354e-05, -2.7139e-04, -1.0394e-03, -3.2059e-04,  2.7881e-04,
        -7.0547e-04,  9.9775e-05, -5.0286e-04,  1.7903e-04, -3.5830e-04,
        -1.9240e-04,  9.7917e-04,  4.5270e-04,  1.1945e-06, -6.6761e-04,
        -5.3327e-04, -5.1745e-04, -3.8888e-04, -6.7719e-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 #200: [tensor([-3.3103e-03, -9.9003e-01,  1.4291e-04,  2.4004e-04, -6.0562e-05,
         3.1774e-04,  1.5302e-04, -2.4764e-04, -4.4558e-04,  3.1994e-05,
        -1.2603e-04,  1.9129e-04,  9.3129e-05,  1.0436e-04, -9.9672e-06,
        -1.1017e-04,  2.0308e-04, -4.0755e-04, -6.1276e-06, -1.7182e-04,
         7.2611e-04, -4.8725e-04, -2.4012e-04, -4.8575e-05,  2.4518e-04,
         4.7221e-04, -1.8269e-05, -2.6843e-06,  1.9981e-04,  1.3166e-04,
        -1.1770e-04, -3.5134e-04,  1.9473e-04, -6.1100e-05, -4.8124e-05,
         2.5343e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8322e-03, -9.8858e-01,  6.0206e-04,  6.8640e-04, -2.5085e-04,
         3.5657e-04, -1.8299e-04,  2.7317e-04,  3.7527e-04, -5.2603e-05,
        -2.9864e-04, -1.4041e-04,  3.6426e-05,  3.6263e-05,  7.9530e-04,
        -1.6441e-04, -1.3967e-04, -5.9915e-04,  4.2181e-05, -3.5497e-05,
         4.1640e-04, -5.9290e-05, -2.3894e-05,  4.3068e-04, -6.4144e-06,
        -1.5670e-04,  7.7956e-05, -5.2359e-04, -4.2146e-04, -5.9763e-04,
         4.2017e-04, -4.9371e-05, -2.6314e-04,  3.6932e-05,  3.3900e-04,
        -7.0005e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5227e-03,  9.8579e-01, -2.3354e-03, -4.3086e-04,  2.5769e-04,
        -1.2389e-03,  6.5456e-04, -9.3637e-04, -1.0036e-03, -3.0428e-04,
         1.9600e-04, -1.2937e-03,  2.5172e-04,  3.6997e-04,  1.2029e-03,
         1.0427e-04, -1.1487e-03, -8.2998e-04, -5.2102e-05, -7.4133e-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([-4.0756e-03,  9.8264e-01, -3.3748e-05,  3.0442e-04, -6.6958e-04,
        -2.5260e-04,  6.7802e-04,  3.7894e-04, -4.2564e-04,  8.3564e-04,
        -4.1289e-04, -1.1900e-04, -1.9212e-04, -5.6435e-04, -3.4378e-04,
        -3.4833e-04, -2.5325e-04,  3.0276e-04, -6.9717e-05,  2.3773e-05,
         2.6224e-04, -1.5227e-03, -1.0893e-04, -8.2570e-04, -6.5819e-05,
        -8.6544e-05,  2.1222e-04,  3.1715e-04, -4.6601e-04,  4.9051e-04,
        -5.0856e-04,  4.2141e-04,  3.0447e-04, -2.9659e-04, -6.8326e-04,
         2.8287e-04,  1.5148e-04, -6.9423e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6141e-04,  9.9031e-01,  1.5927e-04,  1.6177e-04, -1.5214e-04,
         1.9399e-04,  6.0085e-05,  7.2270e-04,  2.9003e-04, -3.1590e-04,
         1.2683e-04, -2.3124e-04,  9.2424e-05,  5.5920e-04, -1.1128e-04,
         6.0104e-04,  1.6118e-05, -5.6835e-05, -1.7365e-04,  6.1488e-04,
        -2.2777e-04, -5.2211e-05, -6.7762e-04, -1.1299e-04, -4.2504e-05,
        -2.6363e-05,  3.1613e-04,  2.0148e-04,  3.7054e-04, -2.1920e-04,
        -3.7361e-04, -4.0827e-05,  5.4788e-04, -4.4571e-04, -5.9574e-05,
        -1.1872e-04, -1.7179e-04,  5.5184e-05, -3.2627e-04,  1.4644e-04,
         2.5276e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2733e-04, -9.9371e-01,  9.0595e-05,  3.0458e-04, -2.6045e-04,
        -2.4984e-04, -3.0711e-04,  3.8632e-04,  5.1184e-04,  1.0129e-03,
         2.9901e-04, -6.7476e-04,  1.7334e-04,  2.9861e-04,  2.7565e-04,
         1.9405e-04,  1.1334e-04,  1.9944e-04, -1.4845e-04, -2.5430e-04,
        -2.6416e-04,  4.0435e-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,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3373e-04, -9.8401e-01,  7.0153e-04,  1.4130e-04, -5.2454e-04,
         8.2433e-04,  2.9985e-04,  1.2078e-03,  8.5086e-04, -2.9288e-05,
         2.3373e-04,  7.8431e-04, -1.3046e-04,  3.7699e-04, -6.4023e-05,
        -3.4451e-05,  4.0731e-04,  1.3996e-05,  9.9464e-04,  8.0177e-04,
         5.7416e-04, -4.9326e-04,  6.4030e-05,  3.8777e-04,  2.0222e-04,
         5.8661e-04,  3.4334e-04,  2.5792e-04,  7.3098e-04,  1.0990e-03,
         3.9195e-05,  5.0526e-04,  5.8683e-04,  4.5096e-04,  7.1554e-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([-1.7403e-03, -9.8414e-01, -3.5191e-04,  7.0990e-04,  1.0029e-03,
         9.9884e-04,  2.4402e-04,  1.0092e-03,  9.7463e-04,  3.7508e-04,
         8.7073e-05,  6.9224e-04,  5.5153e-04,  5.5666e-04, -9.3258e-05,
         1.0041e-03,  6.6324e-04, -4.4165e-04,  4.8857e-04,  2.6424e-04,
         6.2191e-04,  2.5095e-05,  5.5015e-04,  4.6270e-04,  6.7489e-04,
         9.5632e-05,  2.7209e-04,  3.3103e-04, -5.7368e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.5475e-04,  9.9021e-01,  4.0481e-04, -5.8533e-04, -6.1934e-04,
         1.3694e-04, -3.8461e-05,  1.5893e-04,  7.3743e-04,  1.0812e-05,
         1.1407e-03,  1.7561e-05,  2.5247e-05,  1.1724e-05, -3.8755e-04,
        -3.6496e-04,  1.4028e-04, -3.3871e-04,  1.2140e-03, -3.9508e-05,
         4.1458e-04, -7.4747e-04, -5.6800e-04, -3.9392e-04, -6.3399e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6022e-03,  9.8036e-01, -1.7651e-03,  3.2507e-04, -1.2703e-03,
        -7.8179e-04,  7.0991e-04, -5.8989e-04, -6.2907e-04, -1.0651e-03,
        -8.6079e-04,  4.0074e-04, -5.6235e-04,  3.4269e-04,  1.4108e-03,
        -1.0445e-03,  6.3131e-05, -1.5241e-03,  1.2700e-03, -6.9909e-04,
         7.6734e-04, -1.3519e-03, -6.0063e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.0695e-03, -9.8702e-01, -3.0464e-04, -1.5180e-04, -2.9437e-05,
         2.4619e-04, -4.5667e-04, -2.5430e-04, -1.1226e-04, -2.7256e-06,
        -5.2727e-04, -2.9141e-04, -3.4499e-04, -1.6674e-04, -1.1996e-05,
         1.1217e-04, -5.6996e-05, -7.7105e-04, -8.6088e-05,  1.4375e-04,
        -5.1407e-04,  2.9090e-04, -1.0963e-04, -1.7736e-04,  1.2094e-04,
         4.3955e-04, -1.6372e-04,  3.2426e-04,  4.2170e-04,  6.5940e-05,
        -7.4441e-04,  2.0411e-04,  1.6557e-04, -5.4688e-04,  3.8556e-04,
         2.8544e-04, -2.9658e-04, -3.1607e-04, -2.6867e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3013e-03,  9.8876e-01,  1.2166e-03,  1.3192e-03,  5.1200e-04,
         2.3143e-04,  1.0919e-03, -4.8560e-04,  2.4877e-03,  7.9926e-04,
        -8.1341e-04,  9.8595e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 5.4208e-04, -9.9094e-01, -2.7419e-04,  5.1779e-04,  5.6933e-04,
        -9.2006e-04,  1.4516e-04,  3.4816e-04,  2.5851e-04, -2.7559e-04,
         1.6875e-04, -1.4773e-04, -1.5135e-04, -4.1487e-05,  4.0657e-04,
         9.1928e-04,  5.9493e-06,  7.3632e-05,  8.1142e-04, -5.9019e-04,
         5.5155e-05,  4.1286e-04,  2.8121e-04, -1.5535e-04, -9.3950e-05,
        -5.8852e-05, -5.4158e-06,  2.6510e-04,  2.7423e-04,  2.8684e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1777e-03,  9.8840e-01, -9.4202e-05, -5.2276e-04,  2.1300e-05,
        -4.9325e-04, -4.5193e-05, -3.9082e-04, -3.3336e-04, -2.0937e-04,
         2.4884e-04, -6.5077e-04, -2.7560e-04,  5.4667e-06, -7.9842e-05,
        -4.2515e-04,  1.0255e-04,  3.1570e-05, -1.0509e-04, -1.0863e-04,
        -5.2415e-04, -3.0811e-04, -2.7460e-04, -5.7725e-04,  3.2435e-04,
        -9.0872e-05,  3.0182e-04,  2.7551e-04, -6.8238e-05,  3.4127e-04,
         8.2333e-05, -8.4109e-04,  5.2663e-04, -9.0050e-06, -7.1711e-04,
        -2.2035e-04, -2.7988e-04, -5.1893e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5384e-03, -9.9038e-01, -6.5545e-04,  2.6883e-04,  8.5904e-04,
         1.9894e-04,  3.0340e-04,  5.4311e-04, -1.3059e-05, -5.2351e-04,
         1.1739e-04,  1.4731e-04,  3.8385e-04, -7.8014e-05, -5.6840e-04,
         5.8853e-04, -3.8332e-05,  8.2650e-06, -5.2672e-04, -1.9260e-04,
         4.1248e-04,  3.1245e-04,  2.1455e-05, -3.3926e-04, -4.0745e-05,
        -3.5790e-05, -1.1272e-04,  3.3606e-05,  4.5379e-04,  3.0944e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1810e-04, -9.8975e-01, -1.6987e-04, -5.7830e-04, -2.8095e-04,
         6.4982e-04, -5.4349e-04, -1.8813e-03,  5.9611e-04, -5.7735e-04,
        -6.4114e-04,  6.7668e-05, -2.0378e-04,  3.3499e-04, -7.2389e-04,
         2.0267e-04, -1.0832e-03,  2.7406e-04,  1.1186e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.3648e-04, -9.8775e-01, -2.8553e-04,  2.9859e-04, -3.6087e-04,
         1.5702e-04, -2.5693e-04, -7.4982e-04, -9.7476e-04, -1.9629e-05,
        -5.1741e-04, -1.0537e-05, -1.3643e-03,  1.1387e-04, -1.9357e-04,
        -1.7965e-04, -6.4001e-04, -4.8914e-04, -9.0564e-04, -8.2448e-04,
        -3.9010e-04, -7.3983e-04,  3.4951e-05,  4.1882e-04,  4.2049e-04,
        -1.2672e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.3363e-04, -9.8468e-01,  6.5181e-04,  1.3776e-03,  4.9486e-04,
         1.7337e-03, -4.4789e-04,  2.2717e-04,  5.7315e-04, -2.4956e-04,
         5.5984e-04,  8.3242e-05,  4.2539e-05,  5.9643e-04,  5.7181e-04,
         4.4618e-04,  1.2944e-03, -3.7017e-05,  1.3803e-03,  7.7984e-04,
         2.0910e-03,  6.4872e-04,  4.0219e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1522e-03, -9.9133e-01,  9.9651e-04, -3.4410e-04, -6.9416e-04,
         1.6536e-03, -1.1144e-03, -3.6833e-04,  8.8931e-05, -9.0311e-05,
         8.2628e-05, -7.9983e-04,  9.9683e-05, -1.5582e-04,  1.0325e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3619e-04,  9.9239e-01, -1.2450e-04,  2.1180e-04, -6.0566e-04,
        -2.2092e-04,  2.6061e-04, -6.0155e-05, -7.1884e-04,  9.7909e-04,
        -1.4032e-03,  7.3668e-04, -8.7598e-04,  2.5582e-04,  1.5496e-04,
         6.2168e-05,  6.0746e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1616e-03, -9.9208e-01,  7.1763e-04, -1.0168e-04,  1.4301e-04,
         3.8666e-04, -3.2478e-04, -3.1869e-05,  6.6400e-04,  8.5680e-04,
         7.4719e-04,  9.2130e-04,  3.3754e-04,  5.9357e-04, -1.1125e-04,
        -5.6643e-04, -2.0909e-04,  4.5486e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4974e-03, -9.9034e-01,  3.0738e-05, -3.3078e-04,  5.1280e-04,
         1.9888e-04, -6.3679e-04,  6.5680e-04,  3.7224e-04,  1.0105e-03,
        -3.9910e-04,  2.8899e-04, -4.8014e-05,  6.8886e-04,  3.2967e-04,
         1.3248e-04,  3.5502e-05,  5.9117e-04,  1.6002e-04,  4.2874e-04,
        -8.4944e-04,  3.1269e-06,  3.4642e-05,  1.3381e-04,  1.4070e-04,
        -1.5267e-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,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7021e-03, -9.8012e-01,  1.7122e-04,  7.4567e-04,  2.9643e-04,
        -2.9294e-04,  4.2981e-04,  2.6526e-05,  9.3536e-04,  1.3288e-04,
         1.1202e-03, -1.6052e-05,  3.6325e-04, -1.0394e-03,  1.0684e-03,
         4.4125e-04,  4.8947e-04, -4.6327e-04, -5.5148e-04,  3.0646e-04,
         5.1808e-04,  3.6410e-05, -4.5185e-04,  3.6272e-04,  5.9793e-04,
         4.9056e-04,  1.2210e-04,  8.0271e-04, -8.7120e-04,  1.4087e-05,
         3.1747e-04,  7.5395e-04, -3.9051e-04,  5.0978e-04,  2.6905e-04,
        -1.5888e-04, -1.4604e-04,  2.2186e-04, -2.5752e-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([-2.4994e-03, -9.8606e-01, -2.5335e-04, -2.3528e-04, -3.8088e-04,
        -3.6332e-04, -3.3683e-04,  3.9211e-04,  1.1407e-05, -1.9814e-04,
        -2.6926e-04,  4.0532e-06, -3.0540e-04, -4.6577e-04,  9.6923e-04,
         2.2432e-04, -5.0323e-05, -6.7439e-04, -5.7463e-04, -2.0160e-04,
         8.7778e-05,  3.0982e-04, -8.0318e-05, -4.7691e-04,  5.3830e-05,
        -1.6415e-04,  2.0451e-04,  5.8855e-04, -4.8951e-05, -1.7247e-04,
         1.4252e-04, -3.0004e-05,  4.4867e-04,  2.5754e-04, -2.2519e-04,
         2.9685e-04, -5.1314e-05, -1.9062e-05,  1.1064e-04, -5.2070e-04,
         2.7560e-04,  2.7058e-04,  9.0878e-07, -2.6827e-04,  2.8972e-04,
        -1.4041e-04], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 3.0242e-03,  9.8563e-01, -4.8736e-05,  4.7461e-04,  9.2484e-04,
        -8.3512e-05, -6.4767e-04, -2.1583e-04,  3.3027e-04,  1.9735e-04,
         2.9165e-04,  9.9923e-05,  2.5424e-04, -6.0336e-04,  5.7832e-04,
        -5.5510e-04, -5.3738e-06, -7.1343e-05,  4.0764e-05, -5.3511e-04,
         3.0050e-04, -5.8306e-04, -1.3285e-05,  3.0197e-04,  2.3199e-05,
        -2.6127e-05,  2.9664e-04,  3.8695e-06, -2.9407e-04,  1.4887e-04,
         5.1389e-04,  1.0954e-04,  2.1875e-04, -5.5450e-04,  5.4633e-04,
        -5.8109e-04,  1.3456e-04, -3.2292e-04, -4.1278e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1666e-04, -9.8829e-01,  4.3029e-04,  2.3494e-05,  2.1428e-04,
         9.0692e-04,  2.3855e-04, -4.8247e-05, -2.7132e-04,  8.3765e-04,
         1.3709e-04,  5.4287e-04,  5.3066e-04,  1.1665e-03,  5.0795e-04,
        -1.6657e-04,  8.5902e-04,  4.5882e-04,  7.4068e-04, -2.0150e-04,
         3.4795e-04,  2.6742e-04,  4.2112e-04,  3.5350e-04,  2.1772e-04,
         5.3268e-04, -2.3252e-04, -5.6916e-05,  7.8299e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1557e-04,  9.9088e-01,  5.0674e-04, -7.7673e-04,  3.6718e-04,
        -5.5567e-05, -3.9706e-04, -1.3317e-04,  1.8796e-03, -1.4715e-04,
         1.3732e-03, -6.4475e-04, -8.6424e-04, -6.8987e-04, -9.4282e-04,
        -2.2628e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4114e-03, -9.9037e-01, -6.7410e-04, -3.1825e-05,  4.8820e-04,
         3.4104e-04,  8.5039e-04, -2.5368e-04,  3.5522e-04,  1.6258e-04,
        -3.0201e-04, -7.3463e-04, -3.9265e-04, -1.3364e-04,  2.8081e-04,
        -1.2760e-04,  2.0421e-04,  4.9206e-04,  4.8637e-05,  1.6416e-04,
        -3.9388e-04, -3.2442e-04, -2.3225e-05,  4.3560e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7309e-03,  9.9041e-01,  1.0904e-03, -1.7717e-04,  7.8906e-04,
        -7.6996e-04, -2.5318e-04, -2.0706e-03,  5.1883e-04,  4.2017e-05,
         8.9797e-04,  6.2065e-04,  6.2697e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4446e-03, -9.8439e-01, -6.2731e-06, -4.3357e-04,  3.7636e-04,
        -2.1068e-04, -4.6926e-05, -1.1036e-04,  3.6524e-04,  1.6144e-04,
         3.6648e-05, -1.6278e-04,  2.6862e-04, -4.6270e-04,  2.7473e-04,
        -1.5442e-05, -3.6701e-04,  8.4991e-04,  4.7099e-04,  1.8115e-03,
         1.1518e-03,  4.9004e-04, -6.8214e-04,  1.9933e-04, -1.1107e-04,
        -2.4775e-04, -5.6816e-04, -8.7639e-05,  2.7844e-04,  4.2535e-04,
        -2.9524e-04,  1.6508e-04, -9.1737e-04,  9.5691e-06, -2.3725e-04,
        -4.6804e-05,  2.0052e-05,  6.4194e-05,  2.1920e-04,  2.7314e-04,
        -2.4679e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1292e-04, -9.9211e-01,  1.8087e-04,  1.9652e-04,  1.2625e-04,
        -8.1421e-05, -5.2877e-04, -3.3200e-05,  8.0351e-05, -1.0973e-04,
        -5.9795e-04, -2.9821e-04, -1.7660e-04, -3.1891e-04, -5.3975e-05,
        -1.8584e-04,  1.2557e-04,  5.8218e-05, -4.0863e-04,  3.6099e-04,
         5.1457e-04, -5.4456e-05, -1.4013e-04, -1.1738e-04, -1.8467e-05,
         4.5290e-04,  8.7825e-05, -2.3205e-04, -2.2181e-04, -2.2779e-04,
        -2.0401e-06, -1.5422e-04, -2.9007e-05, -3.8803e-05,  1.9628e-05,
         4.2432e-05,  2.0254e-04, -3.8738e-04,  1.3940e-04, -4.6329e-04,
         1.5682e-04,  1.1339e-04,  4.2145e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.8274e-04, -9.9104e-01,  3.4765e-04,  5.5705e-04,  2.0628e-03,
        -3.7780e-04, -1.2257e-05, -3.3231e-05, -6.7002e-04, -2.5903e-05,
        -3.2513e-04,  4.8188e-04, -4.6497e-05, -3.2256e-04, -4.4823e-04,
         2.6479e-04, -2.3902e-05,  3.2635e-04,  2.5112e-04, -3.8695e-04,
        -6.6935e-04,  1.5947e-04,  1.8465e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0067,  0.9672,  0.0028, -0.0024,  0.0030,  0.0089, -0.0013,  0.0050,
         0.0027,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4732e-03,  9.8551e-01, -2.9716e-04, -6.1406e-05,  2.4698e-05,
        -1.2123e-04,  9.9868e-05,  5.7377e-05,  2.0368e-04, -2.5027e-05,
         2.9496e-04, -1.4658e-04, -3.1928e-04,  3.5286e-04, -1.7090e-04,
         2.5501e-04,  3.5686e-04, -1.6371e-06, -8.8068e-04,  4.2394e-04,
        -1.0947e-03, -5.2441e-04, -6.1470e-04, -1.7515e-04, -1.1876e-05,
        -3.5905e-04, -6.3241e-04,  4.1251e-05, -6.8282e-04, -6.1434e-04,
         8.1612e-04, -6.1296e-04,  3.7869e-04, -6.5573e-04, -6.5119e-04,
        -6.2963e-05, -3.5217e-04, -5.6606e-05, -9.7305e-05, -4.9396e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4614e-05,  9.9030e-01, -3.2051e-04, -9.8417e-04, -1.2362e-03,
         1.3260e-03, -6.0056e-04,  2.4232e-04,  2.0483e-03,  1.0884e-03,
         7.0615e-04,  5.7046e-04,  5.5121e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0458e-03, -9.9052e-01,  1.4200e-03, -5.7169e-04,  2.3312e-03,
         6.4847e-04, -1.7200e-03, -1.7459e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+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([-1.2136e-03,  9.9092e-01, -6.3931e-04,  2.9841e-03,  4.0848e-04,
         1.0749e-03,  1.9228e-03,  8.4137e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2596e-03, -9.9072e-01,  4.0242e-04,  6.1618e-05,  3.8800e-04,
        -9.5252e-05, -1.5618e-04, -2.6296e-05, -2.8906e-04,  8.5635e-05,
         9.8919e-05,  1.3067e-04, -5.0544e-04,  2.4267e-04,  4.8964e-05,
         3.4305e-04, -1.5917e-04, -2.6703e-04,  3.8704e-05, -6.7518e-05,
         7.6286e-05, -1.0226e-06,  1.7313e-04,  1.3199e-04, -8.6815e-05,
        -2.1083e-04,  1.2407e-04,  8.1784e-05, -2.4649e-04, -4.1078e-04,
         3.6939e-04, -5.5840e-04, -4.2580e-04,  2.8008e-04, -1.2201e-04,
        -1.7229e-04,  4.4780e-06,  4.2756e-04, -4.4476e-04, -1.9511e-04,
        -6.7934e-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([ 1.5720e-04,  9.9079e-01,  3.7804e-04, -2.0695e-04, -4.1648e-04,
        -3.0296e-04, -2.5893e-04, -2.1958e-04, -4.8933e-04, -2.5220e-04,
         8.5125e-05, -1.3519e-04, -6.7622e-04,  6.0177e-06, -1.4574e-04,
         2.1678e-04, -2.7871e-04, -5.8969e-04,  1.1890e-04, -2.4320e-04,
        -5.8458e-04, -4.4107e-04, -5.2954e-04, -4.5223e-04, -6.2479e-04,
        -3.3069e-04,  3.1368e-04, -1.8074e-04,  4.6395e-04,  1.1230e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5189e-03,  9.8394e-01, -1.2644e-03, -2.0998e-03, -1.3182e-03,
        -9.6515e-04, -4.2222e-04, -8.9839e-04, -6.8629e-04,  3.9924e-05,
        -8.5987e-04,  4.9085e-04, -2.0672e-03, -1.4444e-03,  4.5621e-04,
        -8.6853e-04, -3.5489e-04,  3.0019e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.7818e-04, -9.8864e-01,  5.9278e-05,  3.3911e-04,  1.4699e-04,
        -5.7425e-05, -2.4408e-04,  2.2684e-04, -3.4162e-04,  8.7931e-05,
        -5.5371e-05, -3.1928e-04, -2.6686e-05, -2.0613e-04, -5.6647e-05,
         1.4304e-05, -7.3532e-05, -6.1095e-04, -2.0151e-04,  3.1406e-05,
        -2.4459e-04, -1.6370e-04, -3.6628e-06,  5.1263e-05,  1.7598e-04,
        -6.9080e-04,  4.1397e-05, -5.9428e-06,  4.4745e-04,  1.0996e-04,
        -1.1927e-04, -1.9913e-04,  7.7449e-04, -3.6498e-04,  1.5765e-04,
        -1.5806e-04,  3.4487e-04,  2.3512e-04,  8.0127e-04, -1.0716e-04,
         4.4706e-05,  4.6837e-04,  2.1158e-04,  3.5661e-04, -6.2009e-05,
        -4.1901e-04, -5.6231e-06,  1.2995e-04, -2.8163e-04, -3.2937e-05,
         1.8423e-05,  4.2847e-05, -4.4433e-05,  1.6431e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3469e-03,  9.9360e-01,  5.4622e-05, -6.8394e-04, -1.0032e-03,
         9.2146e-04,  2.7344e-05,  1.2961e-03,  7.1330e-06, -7.1479e-04,
         3.4472e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2599e-04, -9.8997e-01,  7.9755e-04,  2.0666e-04,  6.6637e-04,
         7.5029e-04,  4.6769e-04,  8.0350e-05,  9.6664e-04,  2.6555e-05,
         6.5270e-04, -6.1667e-04, -2.7087e-04,  8.4863e-05,  5.8206e-04,
        -1.7292e-04, -6.4648e-04,  1.1976e-05, -3.1938e-04,  4.3462e-04,
        -4.9369e-04,  1.1148e-05,  1.2643e-04,  2.4057e-04,  5.5999e-04,
         3.5367e-04,  8.3093e-05, -1.7227e-04, -1.0537e-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([ 3.4048e-03,  9.8799e-01,  7.2784e-04,  6.4205e-04,  3.2458e-04,
         1.5390e-03, -1.4658e-04,  7.1284e-05,  1.4588e-03, -4.2885e-04,
        -5.8725e-04,  4.3515e-04, -1.0079e-03, -3.1014e-04,  9.2372e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2580e-03,  9.8665e-01,  1.2047e-04,  2.9470e-05, -6.8343e-04,
        -3.7773e-04, -1.0504e-03, -2.6617e-05, -1.4653e-04,  4.8514e-04,
        -6.7452e-04,  1.2495e-04, -2.4038e-04,  5.6476e-04, -1.5138e-03,
         7.1605e-04, -5.9301e-04,  2.0646e-04,  3.0786e-04, -1.2676e-04,
        -6.3008e-04,  5.0377e-05, -3.3309e-04, -5.0279e-04,  1.0385e-04,
        -3.3047e-05,  2.6915e-04, -1.0914e-04, -3.8904e-04, -5.0842e-05,
        -1.5178e-04,  2.4707e-06,  4.4244e-04, -3.3138e-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([-4.4195e-03, -9.8341e-01, -3.5716e-04, -1.7134e-04, -2.2109e-04,
        -6.5333e-04,  3.7065e-04, -7.6884e-04, -8.5604e-05, -3.7203e-04,
         4.9899e-04,  7.6327e-05, -1.2450e-04, -2.5794e-04,  6.7614e-04,
         4.7867e-04,  3.4687e-04, -1.2396e-04, -8.8639e-05,  3.3297e-04,
        -5.2602e-04, -7.2320e-05,  1.0703e-04,  1.6778e-04,  3.1010e-04,
         1.0286e-04, -1.7980e-04, -2.2614e-04, -4.6119e-04, -2.3286e-04,
        -4.4590e-05, -8.7270e-04, -7.6862e-05,  5.5020e-04, -3.2667e-04,
        -8.8243e-04,  4.5964e-04, -3.1218e-04,  2.5135e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7974e-03,  9.8815e-01, -5.9613e-04, -1.4320e-05, -6.3970e-04,
        -2.2973e-04,  1.2369e-04,  1.0991e-03, -8.7311e-05, -5.2272e-04,
        -1.0005e-04, -1.2171e-03,  1.9272e-04,  3.3520e-04, -2.7732e-04,
         1.5397e-04,  1.1948e-04,  7.8483e-04,  1.3713e-03, -3.4688e-04,
         1.8358e-04, -3.3354e-04,  3.4523e-04, -4.7331e-04,  3.6418e-04,
         1.4016e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1870e-03, -9.8908e-01, -4.0080e-04, -7.4571e-05,  7.5080e-04,
        -2.4067e-04, -5.1100e-04,  2.3368e-04, -4.0303e-04,  2.0235e-04,
         9.3130e-04, -5.6981e-05, -5.5668e-05, -2.9252e-04, -3.7417e-04,
        -4.9529e-04, -1.9202e-04, -2.5814e-04,  7.9508e-05, -3.0201e-04,
        -9.1889e-04, -6.2303e-04, -7.4660e-04, -2.2089e-04, -3.7227e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 6.9587e-04, -9.9148e-01, -3.3800e-04, -1.4826e-03,  1.4217e-04,
        -2.2916e-03,  9.7563e-04, -1.9179e-04, -7.5833e-04, -1.1081e-03,
        -5.3711e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5409e-03, -9.8932e-01, -1.7551e-04, -2.3021e-05,  3.0031e-04,
        -2.4362e-04,  4.7910e-05,  1.2332e-04,  2.7502e-04, -7.3364e-05,
         6.3819e-04,  9.2183e-05,  2.4510e-04, -2.3196e-05, -4.9316e-04,
        -1.0018e-04, -5.2338e-04,  3.4064e-04, -2.3712e-04,  1.3980e-04,
        -1.7630e-04,  6.8082e-05,  1.8120e-04,  5.2451e-05, -1.6128e-04,
         2.3486e-06, -2.9995e-04,  6.1788e-05, -8.9918e-05, -1.2152e-04,
         3.0896e-04,  8.4008e-05, -1.5671e-04,  4.5155e-04,  1.2208e-04,
        -2.7150e-04, -1.4677e-04,  9.0127e-05,  1.6531e-04,  1.1385e-04,
         3.6754e-05,  1.5258e-04, -2.4589e-04,  3.1010e-04,  1.2930e-05,
         1.2828e-04, -1.8212e-04, -9.3317e-05, -1.4350e-04,  6.0662e-05,
         6.7509e-05,  4.8079e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7414e-04, -9.9444e-01, -1.0469e-03,  9.1128e-05,  3.4811e-04,
        -1.1056e-04,  1.7079e-04, -8.7009e-05, -1.6119e-05,  1.9190e-04,
        -6.5464e-05, -2.1726e-04,  1.8101e-04,  6.8569e-04, -1.5948e-04,
        -2.3254e-07, -2.0919e-04,  4.4010e-04, -1.6069e-04,  1.2850e-04,
        -2.3764e-05, -3.1363e-04,  1.1620e-04,  1.2523e-04,  1.9197e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0494e-04,  9.8939e-01, -1.6122e-03,  1.0333e-04, -1.0405e-03,
         6.0940e-04, -3.7205e-04, -5.4058e-04, -1.2065e-05, -5.7840e-04,
         4.4418e-04, -5.3761e-05, -4.5417e-04, -5.9365e-05,  2.2977e-04,
        -6.2544e-05, -4.4978e-04, -1.2495e-03, -3.8612e-04,  2.1915e-04,
         1.1587e-03, -3.6623e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9658e-03,  9.8243e-01, -7.0581e-04,  8.4204e-04, -3.4840e-05,
         7.4311e-04, -6.2233e-04,  5.1340e-05,  3.9170e-05,  2.1797e-05,
        -8.5461e-04,  4.4594e-04, -6.6771e-04, -5.6116e-05,  4.2613e-04,
        -3.7345e-04, -9.1335e-04, -3.7878e-04, -2.3097e-04,  2.7556e-04,
        -6.4023e-04,  9.7355e-05, -4.0133e-04, -7.7421e-04, -5.8320e-04,
        -2.6660e-04, -9.9040e-04, -1.4380e-03, -4.9205e-04,  4.5324e-05,
        -6.4544e-04, -5.5172e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5528e-03, -9.8550e-01, -5.5961e-05,  1.2588e-04,  1.1040e-03,
         1.1876e-03,  1.6204e-04, -1.4884e-04,  2.0452e-04,  5.8258e-04,
        -3.5892e-04,  1.0192e-04, -5.5149e-04, -5.5124e-04,  3.5133e-04,
        -4.6974e-04,  4.5224e-05, -2.9052e-04,  1.3885e-03, -1.7101e-04,
         7.0071e-04,  2.3454e-04,  2.5227e-04,  4.0744e-04,  4.9649e-04,
        -5.8171e-05,  4.3893e-04, -4.9688e-04, -7.0509e-04, -4.0708e-04,
         1.3913e-05,  6.4374e-04, -4.7685e-05,  1.1149e-04, -7.9912e-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([-3.3030e-03,  9.8538e-01, -2.2123e-04,  3.1061e-04, -1.9877e-04,
        -1.3250e-04,  1.3672e-03,  7.4227e-05, -2.2223e-04,  1.1841e-03,
         3.0765e-04, -4.1672e-05,  9.0985e-04,  3.3977e-04, -8.4051e-04,
         4.6166e-04,  6.5225e-04, -7.4298e-04, -1.9027e-04,  6.0132e-04,
        -6.0613e-05,  3.1091e-04,  5.1671e-04,  3.0038e-05,  2.8363e-04,
         9.3365e-05,  4.4811e-04, -7.7103e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.8021e-04,  9.8979e-01, -7.9043e-04,  8.5660e-04,  2.3135e-04,
         9.6132e-05,  7.5073e-04, -3.2150e-04,  5.5192e-04, -1.4491e-03,
         4.6523e-04,  8.9183e-04,  1.7269e-03, -4.8761e-04, -6.0800e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.9723e-04,  9.8673e-01, -6.8989e-05,  2.2006e-05, -1.5177e-04,
         3.1110e-04, -6.7659e-04, -5.9426e-04, -3.6986e-04, -4.0281e-04,
        -3.6220e-04,  6.5133e-05, -3.2417e-04, -3.5029e-04,  1.6805e-04,
         6.9500e-05, -5.9739e-04,  2.0487e-04,  2.3965e-04, -5.5008e-04,
        -1.3488e-04, -3.3938e-04, -2.6861e-04, -3.1614e-04, -8.2749e-04,
        -9.7904e-04, -7.2044e-04, -2.3505e-04, -3.2097e-04, -6.1501e-04,
        -3.8777e-04, -1.9981e-04, -5.5163e-04,  6.5266e-04, -1.9146e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9581e-03,  9.8734e-01,  9.1453e-05,  7.6010e-07, -2.3181e-04,
        -2.2687e-04, -2.5233e-04, -4.2195e-04, -6.5747e-04, -1.3004e-04,
         3.2049e-04,  6.7643e-05, -1.1681e-03, -9.2001e-04,  1.1743e-03,
        -1.4922e-03,  6.7145e-04, -7.0371e-04,  1.4165e-05, -1.5095e-03,
        -6.0191e-04,  5.8741e-06,  4.1440e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4980e-03, -9.9076e-01, -4.2216e-04, -9.4568e-05, -2.7363e-04,
         5.6919e-04, -3.8989e-05,  1.9738e-04,  1.0521e-04, -1.1568e-04,
        -1.1366e-04,  9.8710e-04, -1.8404e-04, -8.1483e-04,  6.6046e-04,
         3.0693e-04,  3.9448e-04, -3.1390e-06,  3.0720e-05, -2.0299e-04,
         9.7011e-05, -1.8052e-04, -3.5082e-04,  2.5280e-04, -3.5646e-04,
         3.3126e-04,  6.7530e-05,  5.9143e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.3037e-05, -9.8617e-01,  7.3912e-04, -6.7676e-04,  5.3800e-04,
        -8.3885e-04,  3.0361e-04, -1.2782e-04, -1.9488e-04, -1.8382e-04,
         9.3899e-04, -5.1179e-04,  1.6082e-04, -3.7533e-05, -3.7737e-04,
         3.4795e-05,  1.4588e-04,  3.0745e-04,  4.7056e-04,  8.6097e-04,
        -3.0404e-04, -8.0793e-04,  2.4474e-04, -1.2031e-04, -5.1440e-04,
         1.5112e-04, -4.0746e-05, -3.2274e-04, -7.3737e-05,  5.4381e-04,
        -4.8283e-04, -9.7717e-05, -1.0171e-05,  4.8001e-04, -8.1133e-04,
         2.8287e-04,  5.1718e-04, -4.3605e-04, -7.9137e-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,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 2.1036e-03,  9.8474e-01, -2.6382e-04, -1.6406e-03, -9.0205e-04,
         4.2009e-04, -4.2131e-05, -2.5330e-04, -3.2513e-04, -8.0679e-04,
        -2.6788e-04, -4.6015e-04, -1.0304e-03,  1.2345e-04,  9.6592e-04,
        -7.0435e-04,  3.9752e-04,  1.3114e-06, -5.3989e-04, -7.4267e-04,
        -7.5614e-04,  5.3022e-04,  3.3977e-04, -6.4625e-04,  4.2169e-04,
        -1.0788e-04, -3.1204e-04,  1.5188e-04, -5.5077e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.7960e-04,  9.8922e-01, -7.4402e-04,  6.6282e-04,  2.7948e-04,
        -7.2341e-04,  3.6728e-04,  3.0702e-04,  1.6826e-04,  1.4175e-04,
        -6.0571e-04, -1.2351e-03,  7.6845e-04, -1.8033e-04,  4.4028e-04,
         4.2646e-04,  7.6658e-05,  2.6778e-04,  1.4770e-04,  2.5914e-04,
         7.5533e-04,  6.2186e-04, -8.2142e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.7463e-05,  9.8956e-01, -2.1384e-04,  9.4291e-05,  1.3467e-04,
        -1.2440e-04,  4.1709e-04, -9.1027e-04, -2.4265e-04, -3.1149e-04,
        -5.0005e-04, -5.0512e-04, -3.8147e-04, -4.4607e-04,  2.3336e-04,
        -3.1144e-05,  4.9205e-04,  3.3962e-04,  4.1331e-05, -1.5075e-04,
        -4.3552e-04, -5.4316e-04, -1.5607e-04, -2.8908e-04, -7.7758e-05,
         4.6179e-04,  5.2686e-04,  1.4391e-04, -1.6446e-04, -2.3924e-04,
         1.0514e-05, -8.1067e-04,  6.8011e-04, -5.1605e-05, -2.0338e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1502e-05,  9.9073e-01,  1.4554e-04,  3.7416e-04, -2.6497e-04,
         3.9798e-04,  7.0576e-05,  2.9959e-04, -9.5376e-05, -9.8226e-05,
         7.3382e-04, -1.4861e-04,  1.0465e-04,  1.4684e-04, -1.2201e-04,
         3.6323e-04,  1.9661e-04,  1.1596e-04, -3.8648e-06, -2.8428e-04,
         6.4029e-04, -1.0161e-04,  1.9800e-05,  4.2933e-04,  3.7937e-04,
         1.6698e-04,  4.7638e-06,  1.1849e-04, -1.1897e-04,  7.8515e-06,
         3.9861e-04,  4.6173e-04,  2.8474e-04, -1.6678e-04,  6.8811e-04,
         1.4437e-04,  8.8577e-05,  1.8519e-04, -4.7132e-04,  5.8545e-05,
         3.3038e-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([ 1.0644e-03, -9.9310e-01, -1.9169e-04,  3.2126e-04, -1.6045e-04,
         7.2115e-05, -3.4657e-04, -4.9360e-04,  1.4110e-04,  9.9576e-05,
         1.7882e-04, -6.7371e-05,  3.5537e-04,  2.1009e-04, -1.6462e-04,
        -4.6908e-05,  1.8099e-04,  9.5790e-05,  5.2845e-05,  1.6594e-04,
         1.6786e-04,  1.5648e-04, -3.5532e-05, -1.8927e-04,  2.7897e-04,
        -3.0233e-05, -1.2585e-04, -3.7666e-04,  2.5972e-05, -1.0950e-04,
         1.2127e-04,  3.7736e-04,  1.5535e-04,  1.3986e-05,  3.8811e-05,
         1.0642e-04,  7.1474e-05, -1.1269e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5203e-03, -9.8616e-01,  1.6741e-04,  9.0240e-05,  2.8259e-04,
         7.7765e-04, -3.3751e-04,  3.4328e-04,  1.9297e-04,  3.5157e-04,
         2.1013e-04, -1.7120e-05, -2.9043e-04,  2.1319e-04, -7.0744e-04,
        -4.2098e-04, -3.2431e-04,  6.3384e-05, -2.7648e-04, -2.4281e-04,
        -4.9230e-04,  3.9569e-04, -1.4584e-04,  3.3015e-04, -3.5825e-04,
         1.0703e-04, -1.7607e-04,  1.0171e-04, -7.7038e-04,  9.1174e-06,
        -2.8120e-04, -1.4424e-04, -6.9912e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.4891e-04,  9.9115e-01,  4.6754e-06, -2.5417e-04, -4.4481e-04,
        -6.2602e-04, -8.2212e-05, -1.0085e-03,  3.9839e-04, -4.9951e-04,
        -4.5954e-04,  2.8363e-04, -1.6751e-04, -2.2317e-04,  5.6827e-05,
         5.9922e-04, -1.1214e-03, -5.8574e-04, -5.1228e-04, -1.3748e-04,
         2.3670e-04, -2.6629e-04,  1.3522e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0578e-03,  9.9072e-01,  8.3237e-04, -1.5000e-04, -6.6013e-04,
        -2.2623e-04,  7.4591e-04, -4.9499e-04, -1.0454e-03,  3.7710e-04,
         8.3900e-04, -8.2790e-04, -3.1048e-04,  4.2796e-04,  1.5023e-04,
         7.0758e-05,  3.3717e-04,  7.2823e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1002e-04, -9.9159e-01,  5.3177e-04, -7.5172e-07,  8.6472e-05,
        -1.6802e-04, -2.1299e-04, -8.5732e-04, -1.2096e-04,  2.5467e-04,
        -1.1288e-05, -1.1192e-04, -1.9869e-04,  4.3057e-04, -6.0145e-04,
        -1.4082e-04, -4.4020e-04,  1.0534e-03,  2.8788e-04,  2.1821e-04,
         1.0783e-04,  1.4487e-04,  5.5609e-04,  1.3145e-05,  4.7748e-04,
         3.9345e-04, -3.8176e-04, -4.9940e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9880e-03, -9.8541e-01,  1.9609e-04,  5.6460e-04,  6.2314e-04,
        -1.1274e-04, -7.6575e-04, -3.5444e-04, -8.4124e-04,  7.0352e-05,
        -1.8859e-03,  6.3504e-05, -4.2766e-04,  7.8629e-04, -1.6220e-04,
        -9.0925e-04,  5.5669e-04,  6.0159e-04, -7.2698e-04, -9.8460e-04,
        -3.5869e-05, -5.6017e-04,  3.7345e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2573e-03, -9.9092e-01, -2.2401e-05, -1.2997e-04, -1.0525e-05,
         2.6613e-04, -2.2231e-04, -1.4107e-04, -1.2740e-04,  4.7373e-04,
         5.4076e-05, -3.3746e-05,  4.1282e-04, -2.7062e-04,  1.2382e-04,
        -2.4064e-04, -1.3832e-05, -1.6764e-04,  2.2047e-04, -1.1956e-04,
        -5.3793e-05, -1.6347e-04, -2.1106e-04,  1.9587e-04, -1.1779e-04,
        -2.8178e-05, -1.8173e-04, -2.9853e-05, -1.9154e-04,  3.3910e-04,
        -6.0367e-05, -3.0836e-04,  4.3219e-04,  1.3540e-04, -9.9849e-05,
         2.0662e-04, -3.2822e-04, -2.9292e-05, -2.0273e-04, -5.1869e-05,
        -2.2799e-04,  5.8154e-04, -1.3676e-04, -2.7295e-04,  6.6325e-05,
        -2.2697e-05,  9.3794e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.7095e-04, -9.8992e-01, -1.1400e-03,  6.4941e-04, -1.2672e-03,
        -1.5508e-04,  1.2578e-03,  6.1556e-04,  4.5059e-04,  6.5549e-04,
         7.6157e-04, -4.0494e-04,  6.5776e-04,  5.4950e-05, -1.1291e-03,
         2.0758e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 6.9487e-04, -9.9170e-01,  1.4653e-04,  9.0576e-04,  4.6260e-04,
         1.3201e-06,  2.4571e-04,  2.5410e-06, -1.0696e-03, -7.8161e-04,
         2.0535e-04, -5.5321e-05,  1.9005e-04, -2.3207e-04, -7.8946e-05,
         3.4202e-04,  7.0059e-04,  2.2587e-04,  1.6698e-05,  3.5175e-04,
        -1.4315e-04,  8.1737e-05,  1.2556e-03, -1.1131e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3480e-04,  9.9211e-01,  2.5092e-04, -1.8916e-05, -3.9883e-04,
        -4.4841e-04, -2.7403e-05, -3.0595e-04, -7.7599e-05, -1.7801e-05,
         3.5207e-04,  8.5065e-04, -1.3623e-04,  1.9642e-04,  5.3020e-04,
        -6.8716e-05, -2.5492e-05,  2.6935e-05,  2.2083e-04, -8.5726e-05,
         2.2245e-04,  2.8518e-04,  1.5297e-04, -2.4335e-04,  3.4955e-05,
        -1.1414e-04,  5.6057e-04,  8.4199e-04,  4.4278e-04, -5.4172e-04,
        -7.4070e-05,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9155e-04, -9.9249e-01, -3.8697e-04, -8.3931e-05,  4.5588e-04,
         8.4980e-04,  5.8014e-04,  1.0701e-04,  1.1417e-03,  8.2666e-04,
         2.6394e-04,  9.2599e-04,  2.5819e-04,  1.1321e-03,  2.0604e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3549e-03,  9.9105e-01, -4.3219e-04, -1.5433e-03, -1.1727e-04,
        -7.0039e-04, -1.0053e-03, -1.1518e-03,  9.7883e-04, -1.2894e-03,
         3.7689e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5626e-03, -9.8753e-01,  8.5257e-04, -3.1952e-04, -7.3347e-04,
        -1.1157e-03,  6.9227e-04, -8.4101e-05, -2.0853e-04, -1.7077e-04,
        -1.9385e-04,  3.5404e-04, -8.9997e-04,  1.1908e-04, -1.5795e-03,
        -1.0810e-03,  7.6294e-04,  8.4040e-04, -9.0007e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1773e-04,  9.9005e-01,  1.4935e-03,  1.9682e-04,  4.1778e-04,
         1.1712e-03,  8.2700e-04,  2.1414e-04, -2.7958e-04, -8.8265e-04,
        -1.2408e-03, -5.2641e-04, -2.1490e-04,  7.0600e-04,  1.4592e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5758e-04, -9.9234e-01,  1.1076e-04, -9.4134e-04,  1.7963e-04,
        -1.6902e-04, -2.1798e-04, -5.2019e-05, -7.1753e-04, -2.4114e-04,
        -2.4160e-05, -7.1951e-04, -9.6466e-04,  6.3490e-04,  8.4509e-05,
         3.7076e-04,  1.0250e-03, -6.8833e-04, -3.5728e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4852e-03,  9.9014e-01, -4.5133e-04, -4.5984e-04,  1.7228e-04,
         4.9693e-05,  1.1718e-03, -2.4835e-04, -3.2558e-04,  1.7705e-04,
        -2.1687e-04,  4.3276e-04,  6.2206e-04, -3.7843e-04,  2.3251e-04,
         1.0902e-04,  1.2961e-04,  2.6975e-04, -5.5353e-04,  4.2910e-04,
         5.0859e-04,  7.0893e-04, -2.6238e-04, -2.1327e-04, -2.4952e-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([ 1.8803e-03,  9.9122e-01, -7.5844e-04,  2.0274e-06,  4.8839e-04,
        -1.6442e-04,  3.0826e-05, -3.0718e-05, -4.9801e-04, -7.9713e-04,
         5.4654e-04, -1.2997e-03, -5.1319e-04,  8.2009e-04,  2.6788e-04,
         4.2122e-04,  2.5759e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.4243e-04, -9.9062e-01,  6.6309e-05,  8.6472e-04, -2.5595e-04,
         1.7046e-04, -1.4156e-06,  3.5538e-04, -7.3457e-05,  1.8328e-04,
         1.5260e-04,  6.2991e-05,  3.9738e-04,  1.1526e-04,  8.9868e-05,
         4.8613e-04,  1.5689e-04,  4.4325e-04,  1.8894e-04,  4.5320e-04,
        -5.6439e-04,  2.3292e-04,  9.5064e-05,  3.0498e-04,  6.4524e-04,
        -1.2490e-05,  6.8625e-04, -5.3002e-04,  3.2544e-04,  2.0941e-04,
        -1.3857e-04,  2.7581e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.4646e-04, -9.9277e-01, -3.6869e-04, -7.8035e-05, -1.9228e-04,
        -1.2413e-04, -1.5202e-04, -5.2115e-04,  6.5194e-04, -6.5879e-04,
         5.7140e-04, -1.2820e-03,  1.9514e-04, -3.2622e-05, -3.3298e-05,
         3.4849e-05,  5.4559e-05,  7.2043e-04,  6.1419e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.9380e-04,  9.9101e-01, -7.4908e-04, -7.7688e-04, -5.0615e-04,
        -3.8526e-04, -7.9796e-04,  1.1760e-03,  1.7267e-03, -1.9264e-03,
        -3.5023e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 6.3968e-04, -9.9371e-01, -4.9667e-04,  6.7454e-04, -4.7178e-05,
        -4.1311e-04, -1.2087e-04, -2.1997e-04, -1.8448e-04,  4.2610e-04,
        -2.5636e-04, -1.0164e-04, -2.9646e-04, -2.0969e-04, -1.2745e-04,
        -1.0070e-04, -5.7340e-04, -7.9331e-06, -1.6000e-04, -8.3285e-05,
        -1.5524e-04, -1.0335e-04,  6.4186e-05,  7.1423e-05, -2.3786e-04,
         2.4475e-04, -4.3930e-05, -7.1744e-06,  2.1762e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.7242e-04, -9.8944e-01,  2.1780e-04,  6.6918e-05, -5.7057e-05,
        -3.1975e-04,  5.0259e-05, -3.4150e-04, -2.8448e-04,  1.1030e-05,
         3.4512e-05, -2.9006e-04, -9.5963e-05, -1.3641e-04,  1.3801e-05,
         1.2393e-04,  8.0627e-06,  2.9348e-04, -1.5967e-04, -2.2427e-04,
         2.2733e-04,  7.1575e-05,  3.6110e-04,  4.2541e-05,  8.6530e-05,
        -1.6434e-05,  4.2098e-04,  3.4748e-04,  2.7143e-04,  2.7253e-05,
        -7.5620e-04, -1.8783e-04,  1.1200e-04,  4.6630e-05, -1.4811e-04,
         2.8840e-04, -5.3590e-04,  9.3045e-06, -3.5215e-04, -8.9845e-05,
         2.1806e-04,  2.0324e-04,  1.5233e-04, -2.7269e-04, -2.2751e-04,
        -1.6325e-04, -2.6977e-04, -5.2293e-05, -2.1323e-04, -3.0023e-04,
        -1.5771e-04,  2.3283e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6394e-04,  9.9161e-01,  9.8583e-04,  2.4752e-04,  1.7201e-04,
         2.6670e-04,  1.1865e-03, -1.7227e-04, -5.1459e-04,  1.4635e-04,
         2.4308e-04, -1.6787e-06, -3.0585e-04, -1.5427e-04, -5.6346e-05,
        -6.8260e-04, -1.8960e-03,  5.4921e-04, -5.4075e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3766e-04,  9.8907e-01, -5.0523e-04,  1.0714e-03,  6.9268e-04,
         1.8560e-04, -5.0392e-04, -3.5963e-04, -1.0001e-04,  6.3268e-04,
        -8.8036e-04,  5.5907e-05, -1.3250e-04,  5.3520e-04, -3.0760e-05,
         8.2834e-04,  6.6918e-04,  3.5547e-04,  5.2677e-04, -5.2602e-04,
        -2.6822e-04,  1.8868e-04, -7.3197e-04, -6.2713e-04, -2.8381e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3042e-04,  9.9189e-01, -6.0503e-04, -3.4096e-04, -7.1710e-04,
        -7.8028e-04, -8.9154e-04,  3.9505e-04,  1.1694e-04, -1.3034e-04,
        -1.0044e-03, -6.4574e-04, -8.4330e-04, -2.2297e-04, -4.8953e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.6445e-04,  9.8798e-01, -6.8761e-04, -1.1841e-03, -3.0498e-04,
         2.0912e-04,  3.5100e-04, -1.2204e-03,  1.5123e-04, -1.1860e-03,
         1.4858e-05, -1.1323e-04,  2.7175e-05,  6.6661e-04, -1.4164e-03,
        -5.0993e-04, -3.3068e-04,  5.3179e-04,  1.3067e-04, -8.8595e-05,
        -1.3179e-03,  6.7610e-05, -4.6547e-04,  2.7930e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9975e-03,  9.9163e-01,  9.4847e-05, -7.4254e-04, -4.8231e-04,
         1.1030e-03,  1.1855e-03, -6.1262e-04, -1.1318e-04, -5.2091e-04,
         5.1860e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6654e-03, -9.8634e-01, -2.4886e-04,  4.6116e-06,  2.1625e-04,
         2.6307e-04, -7.8290e-04,  6.7769e-04, -6.8169e-04, -4.0975e-04,
        -1.3586e-04, -6.5649e-04, -3.8892e-04,  4.3460e-04,  1.0864e-04,
         1.7946e-04, -1.0430e-03, -1.4010e-04, -1.6315e-04,  1.1618e-04,
        -6.6971e-04, -2.0725e-04, -3.4187e-04, -9.2695e-04,  1.5459e-04,
         4.0215e-04, -4.3099e-05, -2.7344e-04, -7.8014e-04, -9.3137e-04,
        -3.3036e-04, -2.7981e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5040e-03, -9.9030e-01,  1.5545e-03,  1.9685e-04,  1.9181e-04,
         4.5478e-04, -4.5668e-04, -3.6380e-04,  4.0099e-04, -4.9546e-04,
         1.1551e-03,  2.5967e-05,  1.7244e-03, -4.4302e-04, -3.5620e-04,
         3.7588e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.0932e-04,  9.8451e-01,  4.1267e-04, -9.7039e-04, -2.3474e-04,
        -1.4312e-04,  5.9092e-04,  9.2128e-04, -1.9780e-04,  9.3600e-04,
         9.7747e-04, -1.4260e-04,  3.9053e-04, -2.9289e-04,  1.4635e-04,
         8.6429e-04,  6.0357e-07, -2.2204e-04,  6.5530e-06,  4.7750e-04,
         7.2192e-04, -6.2325e-05,  8.1616e-04, -3.4686e-04,  9.8134e-04,
         1.4372e-03,  4.2915e-04,  2.9876e-05, -7.2870e-04,  1.2950e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1057e-04, -9.8822e-01, -3.2071e-04, -4.3011e-04, -7.3299e-04,
        -5.5141e-07, -4.2870e-04,  1.0115e-03,  7.2630e-04, -5.4492e-04,
        -8.2838e-04, -4.9049e-04, -8.9871e-04, -1.5841e-03, -1.0195e-04,
        -1.3087e-04, -9.8677e-04, -1.1751e-03,  1.0779e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+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-03, -9.9017e-01,  5.4472e-04, -4.8523e-04,  1.8771e-04,
        -8.8995e-04,  2.4846e-04, -1.3733e-04,  3.1041e-04,  6.5126e-05,
         7.2403e-04,  4.2316e-04, -1.7915e-04, -1.5218e-04,  4.2506e-04,
         4.6064e-04,  7.5021e-04, -2.3301e-04, -2.7615e-04,  5.2038e-04,
         6.5662e-04, -5.0054e-04, -1.6655e-05, -4.8583e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 2.8075e-03, -9.8547e-01,  6.1086e-04,  6.2011e-05, -1.3408e-04,
        -5.4837e-04,  6.0647e-04,  4.4620e-04, -1.4593e-04,  6.9556e-05,
         3.4791e-05,  2.1266e-04,  4.4020e-05, -2.9350e-05,  1.4063e-04,
        -8.7281e-05, -1.6645e-04, -3.5059e-04,  8.8159e-05, -1.1416e-04,
        -3.5265e-04, -3.5261e-04,  4.1593e-04, -3.2469e-04, -1.7366e-04,
        -4.4737e-04,  3.7645e-05, -1.2742e-04, -9.2746e-04, -5.0161e-04,
        -3.9603e-04, -2.5949e-04,  6.2057e-05, -1.8987e-04, -8.2147e-05,
         2.4318e-04,  6.1140e-04,  2.8572e-04, -6.2948e-04,  1.1001e-04,
         1.9616e-04,  1.7338e-04,  5.8188e-05,  3.3865e-06,  1.4642e-04,
        -7.2238e-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([ 7.2730e-04,  9.9150e-01, -5.0653e-04, -3.5665e-04, -4.1562e-04,
         9.8063e-05,  2.7134e-04, -1.6371e-04, -1.3588e-04, -1.0657e-03,
         6.4037e-04,  4.9432e-04, -1.6933e-03, -7.3767e-04, -6.0682e-04,
        -9.5305e-05,  4.9165e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.5447e-03, -9.8582e-01,  2.4276e-04,  1.1305e-04, -7.2251e-04,
         4.2562e-04,  6.2064e-05, -4.0301e-04,  2.0877e-04,  1.4711e-03,
        -3.7416e-04, -3.7119e-04, -2.8752e-04, -8.4563e-06,  2.7645e-04,
         3.5027e-04, -2.2426e-04,  5.5925e-04,  6.4091e-05,  2.2613e-04,
        -4.4806e-04,  4.0199e-04,  2.2240e-04, -2.2982e-04, -5.3017e-05,
         3.2787e-04,  2.6283e-04, -4.9955e-05,  9.7011e-05, -1.8421e-05,
        -2.1244e-05,  2.2528e-04,  3.0232e-04,  5.7240e-05, -3.1782e-05,
        -4.9300e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.5729e-04, -9.9228e-01,  5.5730e-04, -1.0265e-03, -5.2639e-06,
        -8.6496e-04, -4.4013e-04, -2.0523e-04,  1.7460e-04, -2.0018e-04,
        -3.9531e-04, -5.1395e-04, -8.9959e-04, -4.1168e-04,  9.0047e-05,
         4.4064e-04, -1.6499e-04, -1.3177e-04,  2.4212e-04, -3.0242e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1653e-03,  9.8794e-01, -4.8025e-04,  3.0751e-04,  2.8885e-04,
         2.9623e-04, -2.7740e-04, -4.7692e-04, -3.6993e-04, -2.7736e-06,
        -5.1004e-04,  2.8658e-04,  6.5028e-04,  9.7762e-05, -5.0195e-04,
         4.3718e-04, -6.8908e-05, -1.9995e-04,  2.0150e-04, -2.2431e-04,
         6.8686e-05,  7.8750e-05,  3.6927e-04,  4.1602e-04,  7.2462e-04,
         2.5418e-04,  1.1471e-04,  3.6003e-05, -1.5576e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5901e-03, -9.9051e-01,  2.7641e-05,  1.8187e-04,  2.2731e-04,
        -1.1292e-04, -5.1430e-04,  3.2636e-05,  2.8669e-05,  1.2619e-06,
         1.7493e-04, -2.4178e-04,  2.9263e-04, -9.2863e-05,  3.1465e-05,
        -4.1143e-04, -2.3416e-05,  5.0382e-04, -5.2392e-04, -1.3710e-04,
         2.0538e-04, -1.6946e-04, -2.8317e-04, -2.4098e-04,  1.9103e-04,
        -2.3009e-04, -2.9225e-04, -1.7090e-04,  3.4674e-04,  1.1758e-04,
        -1.5069e-04, -1.3751e-04, -2.9865e-04,  3.0156e-04,  1.3527e-04,
        -6.6618e-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([ 2.0906e-05,  9.9427e-01,  2.2150e-04, -6.7906e-04,  3.8947e-04,
         1.7450e-04, -3.8290e-04, -2.8573e-06,  2.4168e-04, -7.1588e-04,
         5.6119e-05,  2.4942e-05,  7.9507e-05,  2.4784e-06,  6.1042e-04,
        -3.3266e-04,  5.9716e-04, -6.2554e-05, -7.3241e-04,  4.0375e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.1066e-04,  9.9288e-01,  2.2902e-04,  3.8257e-04, -3.0437e-05,
        -2.5665e-04, -2.9055e-04, -3.4120e-05, -6.9664e-05,  1.2442e-04,
         1.3734e-04,  7.3352e-05,  3.9659e-05, -1.1420e-04, -2.5981e-04,
         9.3012e-05, -7.8037e-05,  3.5583e-04,  3.8007e-05,  1.5320e-05,
        -1.7891e-04, -3.3471e-05,  9.3292e-05,  2.4234e-04,  2.8108e-04,
        -2.8092e-05, -2.3074e-04,  3.6235e-04, -2.6264e-06, -7.3182e-05,
         5.0666e-05,  2.7668e-05, -2.6195e-04,  1.5183e-05,  2.8860e-04,
        -3.5753e-04, -1.1822e-04,  4.0599e-05,  1.4347e-04,  4.3408e-04,
         5.9642e-05,  2.1608e-04, -5.2956e-05,  9.7630e-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([ 1.9014e-03, -9.8508e-01,  4.8487e-05, -1.9373e-04,  4.7554e-05,
        -4.6840e-04,  2.4626e-04, -4.2214e-04,  1.1920e-04, -2.9786e-04,
        -1.5431e-04, -1.5534e-04, -8.3196e-05, -1.6176e-05, -3.5840e-04,
        -6.6590e-04, -5.7813e-05, -1.4231e-04, -2.2499e-04, -4.8689e-04,
         2.9137e-04, -1.2102e-04, -8.0186e-04,  2.6067e-05, -3.3945e-04,
        -3.6366e-04, -4.3150e-04, -3.2944e-04, -7.5961e-04, -5.4080e-05,
        -5.9691e-04, -2.6226e-04, -3.6833e-04, -1.4045e-04, -5.9252e-05,
        -4.2945e-04,  1.3423e-04, -2.7021e-05,  3.0755e-04, -2.5914e-04,
         1.2835e-05,  1.3380e-05, -6.4275e-05, -1.3405e-05,  5.6429e-04,
         5.7154e-06,  3.6110e-04, -2.3302e-04,  1.1198e-04, -2.5021e-04,
         3.1049e-04, -6.7707e-04,  1.1147e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.8878e-05, -9.9138e-01,  3.6782e-04,  2.7090e-05,  6.5693e-04,
        -1.1397e-04,  6.4271e-05,  3.4525e-04,  3.3393e-04, -7.9386e-05,
        -2.9122e-04,  1.8608e-04,  2.0412e-04, -1.4805e-04,  2.3565e-04,
        -4.7800e-04,  3.7277e-04, -9.7756e-05,  3.1857e-04,  7.4093e-05,
         5.5790e-04,  1.3941e-04, -4.2526e-04,  5.8572e-04,  3.0230e-04,
         2.3143e-05,  4.2212e-04,  1.3608e-04, -2.0412e-04,  5.0605e-04,
        -2.3135e-05, -7.1406e-05,  3.1615e-04,  1.7816e-04, -2.6160e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5554e-03,  9.8432e-01,  1.3922e-04,  2.1334e-04, -7.0987e-05,
         8.7520e-04,  3.4247e-04,  6.2436e-04,  1.8705e-04,  1.3998e-03,
        -2.1237e-04, -6.9689e-04, -5.2934e-04,  1.7890e-04, -1.5509e-04,
         5.4176e-04,  7.4243e-04,  5.6127e-04,  4.7748e-04, -2.5694e-04,
        -1.8377e-04, -6.7998e-04,  5.9657e-05, -1.1509e-04,  2.8077e-04,
         7.8158e-04, -5.0219e-04, -2.3819e-04, -4.1969e-05,  7.3628e-04,
         8.9932e-04, -3.9979e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7673e-04, -9.9090e-01, -9.6663e-05,  9.0879e-05,  1.8954e-04,
        -1.5616e-04,  7.8373e-05,  1.3530e-04,  3.6141e-04,  1.6978e-04,
        -1.9287e-04, -4.8251e-04, -2.7170e-04,  9.1006e-05, -5.8422e-04,
         3.8677e-04, -2.9780e-04,  5.0024e-04,  2.0602e-04, -1.4557e-05,
        -2.7508e-04, -4.0165e-04, -1.6494e-04,  4.9780e-05,  3.9035e-04,
        -7.8949e-05, -2.8752e-04, -1.9926e-04,  4.1327e-04, -1.2526e-04,
        -3.5010e-04,  1.5034e-04, -1.6449e-04,  3.3905e-04,  1.6616e-04,
        -2.2522e-04,  7.2236e-05,  4.0344e-04,  2.5926e-04,  9.0170e-06,
        -4.2307e-05, -5.2827e-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>)]
Iter #100: [tensor([ 1.1415e-03, -9.9445e-01, -4.6708e-07, -3.2449e-04, -2.8695e-04,
        -5.4745e-04,  5.3548e-05, -2.6773e-04,  3.3325e-04, -8.4435e-05,
         2.8826e-06,  4.0270e-04,  2.4828e-04, -8.9144e-04, -4.1865e-04,
        -2.2195e-04, -3.2138e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2734e-04, -9.9405e-01, -4.0548e-04, -2.4501e-04,  3.3991e-04,
        -1.5604e-04, -1.5999e-04, -9.1768e-05,  5.5410e-04, -2.5241e-04,
         8.4212e-05, -2.9211e-04, -8.3358e-04,  6.3495e-05,  6.0239e-06,
        -4.5590e-04, -6.9317e-04, -1.7892e-04,  9.9969e-05,  1.0403e-04,
        -2.9368e-04, -1.9169e-05,  5.4728e-05,  7.0955e-06, -2.0738e-04,
         2.0161e-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([ 1.5209e-03,  9.8368e-01,  8.0674e-04, -8.9208e-04,  1.0312e-03,
        -3.0711e-03,  8.1173e-04, -2.5313e-03,  2.2788e-03,  2.2613e-03,
        -1.1128e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.0563e-03,  9.8414e-01,  7.2327e-04,  7.5026e-04, -1.2992e-03,
         7.0889e-04, -8.9291e-05, -3.2226e-04, -5.2676e-04,  1.3454e-03,
        -1.4134e-03,  4.6906e-04,  4.1268e-04, -4.3567e-04, -1.1155e-04,
        -6.8280e-04,  4.7061e-04,  4.5383e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.6655e-04,  9.9100e-01, -6.6963e-04,  7.2411e-04,  7.6635e-04,
         7.5848e-04, -4.9948e-04, -7.0053e-04,  4.1327e-04,  6.6583e-04,
         2.0942e-04,  3.7384e-04, -3.7905e-04, -2.7589e-04,  5.2636e-04,
        -4.8409e-04,  1.0912e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.4965e-04,  9.8615e-01, -3.8687e-04,  1.0375e-03, -7.6376e-06,
         6.9980e-04,  5.3995e-04,  2.9538e-04,  1.4867e-05,  8.3929e-05,
        -6.0664e-05, -5.8480e-04, -6.7130e-04,  1.3695e-05,  3.1556e-04,
         1.9302e-04, -4.3374e-04, -1.3935e-04,  9.9221e-04,  5.2428e-04,
         4.8153e-04, -4.7299e-04, -4.9155e-05,  1.1075e-03, -1.9000e-04,
         3.8699e-04, -4.0671e-04,  5.8482e-04,  5.8283e-04,  7.7178e-04,
        -3.6180e-04,  1.4484e-04,  7.6111e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7492e-03, -9.9222e-01,  9.2713e-05,  1.4400e-04, -3.3403e-04,
         3.6320e-04, -8.4313e-05, -1.1941e-04, -4.0885e-04, -3.4656e-05,
        -4.6306e-04, -3.5832e-04,  2.8699e-04, -6.2708e-04,  6.3423e-04,
         6.1305e-04,  3.9434e-04, -6.3753e-04, -4.3692e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6298e-03,  9.8683e-01, -9.4990e-04, -1.5847e-04,  1.9563e-03,
         4.1281e-03, -3.0896e-04,  1.7521e-03, -2.2838e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7387e-04, -9.9190e-01,  4.0087e-05, -1.5157e-04,  1.3033e-03,
        -1.3850e-04,  1.4219e-04,  7.6306e-04,  8.3378e-04,  1.9871e-04,
         1.3066e-04,  3.3774e-04, -5.3018e-04, -5.9997e-05,  2.6817e-04,
         1.4326e-04, -1.5307e-04,  3.9504e-04, -4.8286e-04, -1.7859e-04,
        -7.0543e-05, -3.5671e-04, -4.4823e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4165e-03, -9.8697e-01,  9.3145e-05,  1.5402e-04,  8.9607e-05,
         3.6098e-04,  3.0820e-04,  1.0253e-03,  9.3532e-05,  2.3829e-05,
        -1.9096e-05,  1.7024e-05,  5.8173e-04, -3.5022e-05,  1.9964e-04,
         5.2162e-04, -2.7858e-04,  2.8513e-04,  2.8906e-04, -1.0094e-04,
         4.8018e-04,  8.2660e-04,  9.4306e-04,  1.4127e-04,  9.7063e-04,
         1.6719e-05,  1.6381e-04,  1.5460e-04,  1.3734e-04, -1.3087e-04,
         3.1382e-04,  9.1401e-06,  1.3779e-04,  7.6878e-04,  7.7921e-05,
         1.6601e-04,  2.4308e-04,  1.4713e-04,  2.4788e-04,  3.0416e-04,
        -7.7150e-05,  3.8089e-04,  1.9508e-04, -1.0631e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.8234e-04,  9.8931e-01,  1.0153e-04, -2.7690e-04, -4.9054e-04,
        -1.0449e-04, -6.6854e-04,  4.4718e-04, -1.7770e-04, -8.4952e-05,
         5.6532e-04,  1.4496e-04,  3.8789e-04,  1.7209e-04, -4.6061e-04,
        -1.3119e-04,  2.3809e-04, -4.2351e-04,  4.9255e-04,  2.0360e-04,
        -8.4497e-04, -1.2289e-04, -3.9349e-04,  4.3687e-04, -7.7810e-04,
        -1.1203e-04,  3.0355e-04,  1.7396e-04,  1.2786e-04, -3.2199e-04,
         3.5901e-04, -6.5845e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3265e-04,  9.8690e-01, -3.6506e-04,  3.1032e-04,  3.0535e-04,
        -4.8861e-05, -1.5704e-04,  2.6849e-04, -3.7869e-04,  7.2168e-04,
         4.9236e-04, -7.5072e-04,  8.4846e-05,  8.2653e-04, -2.5943e-04,
        -8.2293e-04,  1.5193e-03,  4.7613e-04, -4.5500e-04, -3.4853e-04,
        -9.2743e-04,  7.8553e-05,  2.8729e-04,  2.1184e-04,  4.6145e-04,
        -2.0633e-04,  1.3909e-04, -6.2173e-04, -3.1648e-04, -3.6768e-04,
         3.5769e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 1.0598e-03,  9.9037e-01, -2.8283e-04,  7.6466e-04,  4.0162e-04,
         5.1643e-04,  1.4657e-04, -6.5138e-04, -1.1607e-04,  5.8097e-04,
        -2.7066e-05, -5.4676e-04, -3.6545e-04,  2.7534e-04,  1.0157e-03,
        -2.1061e-04, -6.2356e-04, -3.8176e-04,  3.1429e-04,  5.9081e-04,
         3.4528e-04,  4.1179e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3847e-03, -9.8718e-01,  2.5875e-04, -2.3931e-04, -1.4669e-04,
         1.2799e-04,  1.5297e-04,  3.0672e-05,  8.5289e-05,  4.0074e-04,
         1.0316e-04,  4.9125e-04,  1.4081e-04, -1.1500e-04, -3.5784e-04,
         2.6128e-04, -2.4246e-04,  1.4373e-04,  1.0002e-04,  4.8556e-04,
         1.1017e-04,  1.5002e-04, -8.4269e-04,  2.8199e-04, -4.3902e-04,
        -7.0657e-05,  2.5855e-04,  2.4488e-04,  5.9654e-04,  4.1004e-04,
         1.8191e-04, -2.9124e-05, -2.3512e-04, -6.2966e-05,  1.6206e-04,
        -1.7038e-04,  9.3432e-05, -2.4475e-04, -1.7894e-04,  7.3041e-04,
         3.7198e-06, -3.2860e-04, -3.3001e-04, -2.8983e-04,  5.1375e-05,
        -4.3715e-04,  1.6175e-04, -4.5299e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5852e-03,  9.8798e-01,  7.5604e-04,  4.0413e-04, -1.2316e-03,
         2.8353e-04, -6.9546e-04, -1.3185e-03, -6.4298e-04, -2.4359e-04,
         6.6344e-04,  2.4698e-03,  2.5931e-04,  3.2956e-04, -7.8263e-04,
        -3.5466e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.5264e-05, -9.9189e-01,  9.4041e-05, -2.5496e-05, -1.0448e-05,
        -3.1171e-05, -3.7237e-04,  2.6509e-04,  7.5332e-05,  1.7777e-04,
         2.9643e-04,  3.7667e-05, -1.5632e-04,  8.4554e-05,  3.0232e-04,
        -1.6049e-04, -2.4888e-04, -1.6192e-04, -3.0722e-05,  3.6422e-05,
         1.8642e-04, -5.9085e-05, -3.8123e-05, -4.5797e-05, -1.0430e-04,
         1.8319e-04, -3.2518e-04,  1.6783e-04, -2.5640e-04,  1.3412e-04,
        -7.7460e-05,  1.8136e-04, -5.1693e-04, -3.1394e-05, -1.5824e-04,
         7.5526e-05, -2.1583e-04,  3.3576e-04, -6.4664e-05, -1.8472e-04,
         4.2447e-06, -3.5536e-05,  1.3140e-04, -1.1263e-04,  4.9209e-05,
         2.1113e-04, -2.1018e-04,  1.9988e-05, -1.1599e-04,  1.9750e-04,
        -4.9177e-05,  8.7669e-05, -2.0338e-05, -4.0728e-04, -8.5539e-05,
         1.7903e-04,  1.9748e-04,  1.3814e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1881e-03, -9.8955e-01, -2.2828e-04, -1.3790e-04, -7.1144e-05,
        -2.9759e-04, -2.8136e-04,  3.3580e-04, -3.5772e-04,  3.1206e-04,
        -2.6195e-04, -9.8000e-05, -1.3317e-04,  1.1126e-04,  2.0781e-04,
        -3.9488e-05,  1.1190e-04,  1.9560e-04,  2.7058e-04,  3.2834e-04,
        -4.0173e-04, -2.3744e-04,  2.1444e-04,  8.5330e-05,  9.1552e-05,
        -6.5027e-07, -5.7242e-04, -1.2403e-04,  1.8816e-04,  4.7523e-04,
        -2.1797e-05,  1.5263e-05,  3.9061e-04,  1.5609e-05, -3.3145e-04,
        -7.1292e-06, -8.3305e-05,  3.6834e-04,  1.1526e-04,  1.9122e-05,
        -1.7960e-04, -1.5337e-04,  2.6848e-05,  1.4785e-04,  4.8020e-04,
         3.7787e-04,  3.1216e-05,  1.3096e-04,  1.9463e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2158e-03,  9.9407e-01, -3.8590e-04, -8.6676e-04,  3.0933e-04,
         8.4431e-05, -8.1291e-04, -1.8781e-04,  1.7343e-04, -5.6420e-04,
         2.0829e-04, -7.7923e-05, -5.4384e-04, -5.0121e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1705e-04, -9.9362e-01, -1.9960e-04, -2.2446e-04, -4.5525e-04,
         6.8808e-06,  7.0249e-04, -4.7876e-04,  5.5220e-04,  2.7502e-04,
        -4.4570e-04,  1.4499e-04, -3.3410e-04,  1.2432e-04,  7.4943e-04,
        -1.8424e-04,  5.4313e-05,  4.3914e-04, -5.9228e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6799e-03, -9.8962e-01, -4.5730e-04, -4.4656e-04,  1.1478e-04,
        -1.0343e-04, -5.4728e-04, -4.6600e-04,  7.3393e-04, -2.0631e-04,
        -6.2921e-04, -2.9534e-04,  5.2072e-05,  2.4681e-04, -3.8709e-04,
         6.2358e-05, -4.3495e-04, -7.8089e-04,  3.3126e-04, -2.0813e-04,
        -2.3273e-04, -4.3471e-04,  1.3770e-05,  5.7778e-04,  1.2244e-04,
        -5.7852e-04,  1.9026e-05, -1.1079e-04,  1.0246e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0522e-03,  9.9302e-01, -2.5411e-03,  7.0695e-04,  1.7944e-03,
        -8.8882e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.1588e-04, -9.9040e-01,  4.4736e-04,  1.3028e-05, -1.7382e-04,
        -3.2079e-04,  3.6337e-06,  4.8673e-04, -5.5672e-05, -1.9037e-05,
        -4.6744e-04,  6.8464e-06, -1.7404e-04, -2.3352e-04,  3.2978e-04,
         2.8002e-04, -7.3913e-04, -1.4538e-04, -1.2568e-04,  2.2879e-04,
        -6.6583e-05, -2.5582e-04, -3.3635e-04,  1.2810e-04, -4.7454e-05,
         4.6262e-04, -1.5331e-04, -9.6450e-05, -8.6263e-04, -4.7094e-05,
         2.0870e-04,  1.0234e-04, -7.9775e-05, -5.3986e-04, -7.7076e-04,
        -1.9772e-04, -2.0510e-05,  3.6189e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3276e-03, -9.9287e-01, -3.6595e-04, -2.6280e-06,  2.3735e-04,
        -4.2703e-04, -3.1859e-05, -1.9541e-04,  1.5742e-04, -1.3142e-04,
        -1.9134e-04, -9.7992e-05, -1.9121e-04,  3.3747e-05, -4.5599e-04,
        -3.7667e-04, -2.2113e-04, -5.0287e-04,  4.5441e-05,  6.8091e-05,
        -6.9667e-05,  4.0166e-04,  1.7665e-04,  3.9354e-04,  2.0816e-04,
        -8.7309e-05, -1.0005e-04, -2.9010e-04,  3.3808e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7035e-04, -9.8928e-01,  1.1005e-03, -2.2227e-04, -5.7258e-04,
        -7.6481e-04, -2.2466e-04,  3.8824e-04,  4.9232e-04,  2.4330e-04,
        -4.4374e-05, -9.9030e-04,  1.6986e-04, -1.7591e-04,  6.7532e-04,
         2.2763e-04,  1.2399e-03, -4.0866e-04,  1.9863e-04, -5.7608e-04,
        -6.2675e-04,  1.2587e-05, -3.2820e-04,  6.7756e-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 #200: [tensor([-8.3327e-04, -9.9270e-01, -1.1534e-04,  9.7079e-05,  4.1980e-04,
        -3.0890e-04,  7.6922e-05,  5.7646e-04,  6.6388e-05,  2.1794e-04,
         2.8588e-04,  4.3951e-04, -3.5716e-05,  1.1254e-04,  3.7840e-04,
        -2.6834e-05, -3.0257e-05,  2.5128e-04,  5.0851e-04,  1.2030e-04,
        -8.1550e-05,  1.1213e-04, -3.4750e-04,  9.9748e-05,  2.4886e-04,
         7.1482e-05, -4.8693e-05,  6.6539e-05, -1.4987e-04,  8.4132e-05,
         1.0034e-04,  1.7929e-04,  3.2266e-04, -2.8989e-04,  3.7962e-05,
         1.5316e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.1963e-04, -9.9240e-01, -2.9992e-04,  6.0674e-04,  2.0867e-04,
         1.1054e-04,  7.7632e-05,  1.9511e-04,  5.8044e-04, -3.6429e-04,
        -4.7442e-04,  1.0106e-04, -1.9952e-05,  3.5984e-05,  1.6234e-04,
        -6.9326e-05,  1.5629e-05, -1.7522e-04, -2.0105e-04, -2.1298e-04,
        -1.9445e-04,  1.0505e-04, -5.5978e-05,  2.0432e-04,  4.2061e-04,
        -6.2847e-05, -1.1183e-04, -2.0581e-04, -1.3075e-04,  3.1437e-04,
        -1.4400e-04,  2.1152e-04,  2.9111e-04, -2.6672e-05,  3.7967e-05,
         2.4712e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3773e-04,  9.9224e-01, -7.7394e-05,  2.5892e-04,  2.7173e-04,
         6.1128e-04, -3.7524e-04,  9.2184e-05, -2.6354e-04, -1.0904e-03,
        -2.3854e-04, -4.2888e-04,  4.3525e-04, -1.6909e-04, -9.3168e-04,
         3.3514e-04, -1.0496e-03,  9.4504e-05,  1.7964e-04, -3.2152e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0278e-05,  9.8887e-01, -4.7650e-04,  3.7997e-05, -3.9959e-04,
         3.5192e-04,  2.0972e-04,  1.7469e-04, -3.4189e-04,  3.1517e-04,
        -5.9176e-05, -4.3603e-04, -3.9825e-04, -8.1311e-04, -1.0707e-04,
         7.2551e-04, -7.9184e-05,  6.2152e-04, -1.4656e-04, -2.7512e-05,
         1.5203e-04, -2.9642e-04, -2.8097e-04, -5.4638e-04, -2.4420e-06,
        -1.7421e-04,  2.6183e-04,  7.2049e-04,  2.8245e-04, -1.3575e-05,
        -3.8406e-04,  9.0092e-04, -4.8554e-05,  1.2565e-04,  3.0098e-05,
         2.2142e-04,  4.0727e-04, -5.4323e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.7543e-04, -9.8965e-01,  2.3960e-04, -4.2808e-04, -1.1319e-04,
         1.4914e-04, -2.7732e-05, -5.3384e-04,  3.1821e-04,  1.4963e-04,
         1.8931e-04,  5.3144e-04, -9.5777e-06, -5.6675e-06, -5.9531e-05,
        -6.9220e-04,  8.6433e-05,  6.0603e-04,  1.5973e-04,  3.3051e-04,
         4.8714e-04,  1.5208e-04,  2.7176e-04, -3.0156e-04, -3.8285e-04,
         1.8212e-04, -2.0236e-04, -7.1075e-05, -4.5528e-04,  2.9712e-04,
         1.7625e-04,  5.6286e-04,  1.9498e-05,  4.5607e-05,  6.0080e-04,
         9.9700e-05,  4.1574e-05,  4.1122e-04, -2.4865e-05, -1.8610e-04,
        -1.7782e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7994e-04, -9.9205e-01, -4.1790e-04,  1.0327e-03, -2.3742e-04,
        -8.5851e-05, -2.8047e-04,  5.9211e-04,  7.0706e-04,  5.8399e-04,
         9.7952e-04,  8.1358e-05, -9.2654e-06,  2.9862e-05, -7.3617e-04,
         2.3788e-04,  2.7074e-04,  5.2287e-04, -3.0463e-04, -4.2867e-05,
        -6.0694e-05, -1.5679e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4267e-03,  9.8422e-01, -1.8030e-04,  2.5184e-05,  3.1233e-04,
        -3.6077e-04, -6.5205e-04,  3.0843e-04, -7.2550e-04, -7.9593e-05,
        -5.4986e-04, -4.9754e-04,  2.9945e-05, -2.5853e-04,  4.8355e-04,
         4.7656e-04, -4.4297e-04,  4.5502e-04, -9.5907e-04, -5.6629e-04,
         1.5842e-04,  1.6226e-04, -3.9875e-04,  1.7903e-05, -4.2776e-04,
         2.4353e-05,  9.3698e-05,  5.5964e-04,  7.8951e-04,  3.0937e-04,
         2.0736e-04, -2.1371e-04, -4.1436e-05,  1.6164e-04, -4.2734e-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([-2.2402e-03, -9.9023e-01, -9.2844e-06,  7.8081e-05, -3.8187e-04,
        -4.4845e-04,  2.0115e-04, -1.5224e-04, -1.0644e-04, -1.0501e-06,
        -1.9717e-04,  3.7931e-04, -2.8357e-04,  1.0821e-04, -2.1167e-04,
        -2.4085e-04, -4.8072e-04, -1.3816e-04,  8.4711e-04,  4.2565e-04,
         4.7735e-04,  4.8010e-05, -1.4328e-05, -2.9397e-05,  7.5236e-04,
        -9.3396e-05, -1.9470e-05,  1.0415e-03, -3.5952e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.3339e-04,  9.9056e-01, -3.7152e-04,  2.2162e-04, -1.0512e-04,
         5.5046e-04,  1.3681e-03, -2.3790e-04, -3.9289e-04, -5.2424e-05,
        -1.7844e-04, -4.3519e-04,  4.8264e-04, -5.8097e-04, -2.3119e-04,
         5.9054e-04,  3.5007e-04, -2.2809e-04,  5.0453e-04,  4.0820e-05,
         2.3733e-04, -2.8181e-04, -4.6869e-04,  4.2125e-04,  2.7252e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3134e-04,  9.8622e-01, -4.3482e-04, -1.0913e-03,  8.3516e-05,
         8.8571e-04,  5.2640e-04,  1.6041e-04,  3.5727e-04,  7.8013e-04,
        -7.4359e-04,  1.0197e-03, -6.8917e-04,  1.7387e-03,  1.4445e-03,
        -3.4196e-05,  5.1669e-04,  2.2507e-05,  1.0976e-03, -2.5904e-04,
         6.9771e-04, -5.3453e-05, -6.0792e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0047e-04, -9.9311e-01,  1.6423e-04,  1.1018e-06,  3.5261e-04,
         2.3978e-04, -1.8863e-04,  1.8423e-04,  4.0560e-04,  2.8329e-04,
        -5.8016e-06, -4.2975e-04, -3.9985e-05,  8.3459e-05, -1.7600e-05,
         1.6842e-04, -1.7940e-04,  8.7558e-05, -3.3867e-04,  3.4942e-04,
        -1.7845e-04,  7.5433e-05, -1.4197e-04, -1.3942e-04,  1.6116e-04,
         2.6329e-04,  3.7907e-04, -2.1531e-04,  1.4333e-05,  1.5718e-04,
         9.3092e-05,  4.3444e-05,  3.3368e-04, -3.8077e-05, -1.4768e-04,
        -1.5367e-04, -2.3509e-04,  1.9479e-04, -1.0189e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8405e-04,  9.8990e-01,  4.0149e-04, -1.4492e-04,  8.5563e-04,
         1.0839e-03,  6.1051e-04,  1.5185e-03,  1.0715e-03,  2.6849e-03,
        -9.2217e-04,  6.1977e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-1.6793e-03, -9.9149e-01,  8.3842e-05, -1.9169e-04, -3.7624e-04,
        -4.9907e-04, -2.2208e-04, -1.0671e-04,  5.7285e-05,  1.4438e-04,
         1.1584e-04,  4.3120e-04, -1.6484e-04, -3.1780e-04, -3.7258e-04,
         2.8973e-04,  5.6811e-05,  9.7332e-05, -3.2168e-04, -3.2166e-04,
        -4.7519e-04,  8.6408e-05, -2.2414e-04,  3.0638e-04, -3.0081e-05,
        -2.1217e-04, -6.3317e-04,  1.9804e-05,  4.7631e-04, -1.9461e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6990e-05, -9.9103e-01,  3.2021e-05, -1.1021e-04,  1.7374e-04,
         1.3282e-04, -3.5771e-04,  3.3400e-04, -4.7920e-05,  1.1818e-04,
        -1.0857e-04, -6.0886e-05,  5.1555e-05,  6.0411e-04,  6.0927e-04,
        -3.7468e-04,  1.4682e-04, -4.7811e-04,  4.0705e-04,  1.4259e-04,
         1.4977e-04, -2.8127e-04, -1.0281e-04,  7.8488e-05,  2.0151e-04,
        -2.1365e-04,  1.3095e-04,  1.8994e-04, -3.8904e-04,  5.7123e-04,
         6.2311e-04,  2.2464e-04,  3.2177e-04,  2.2164e-04,  2.5415e-04,
         1.9412e-04,  1.4450e-04,  3.5201e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0565e-04, -9.9196e-01, -1.2341e-05,  1.7617e-04,  1.7635e-04,
        -1.8022e-04,  6.1698e-06,  4.9231e-04, -2.3127e-04,  3.3491e-04,
         6.1301e-04, -2.4147e-04, -9.3468e-04, -1.3444e-05, -1.9895e-04,
        -2.8918e-04, -5.9639e-05, -4.1152e-04,  1.3746e-04,  2.0202e-04,
         1.1066e-04,  2.0555e-05,  6.0329e-05, -4.0945e-04,  2.7672e-04,
        -4.0799e-04, -1.7266e-04,  1.1613e-03,  3.9017e-04,  1.1440e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1191e-04,  9.8955e-01, -1.2450e-03, -1.9038e-04, -3.7518e-04,
        -7.8861e-04,  1.3344e-04,  2.9120e-04,  5.1056e-04, -6.1125e-04,
         8.9381e-04,  7.6950e-04,  8.4630e-04, -1.9304e-04,  9.1431e-04,
         5.3529e-04, -1.0224e-04,  3.4079e-04,  1.2961e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.2258e-04,  9.8695e-01, -1.0088e-03, -2.5494e-05,  8.0541e-04,
        -1.6286e-05,  9.6305e-04, -2.4053e-04,  2.0763e-04,  3.7472e-04,
        -6.2765e-04,  9.5223e-04, -5.8430e-04,  4.9538e-04, -7.6921e-04,
         8.3331e-04, -3.9817e-04, -2.0961e-04, -8.0567e-05, -2.6953e-04,
         5.4907e-04, -2.9001e-04, -9.7248e-04, -1.1815e-03,  1.4097e-04,
         3.2978e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1933e-04, -9.9074e-01,  2.1768e-04,  1.0095e-03,  4.8988e-05,
         1.0360e-03,  3.2550e-05,  3.1170e-04, -1.6555e-04,  8.4003e-04,
        -3.3325e-04,  1.0539e-03,  4.2431e-04, -4.0456e-04, -5.2749e-05,
         7.2511e-04,  2.6449e-05, -3.8133e-04,  5.0982e-04,  5.0160e-04,
         6.0959e-04, -1.8084e-04, -2.7043e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6143e-03,  9.9053e-01, -3.9940e-04,  1.7061e-03, -4.2210e-04,
        -2.7210e-04, -2.7971e-04, -7.2138e-04,  1.1278e-03, -1.1623e-04,
        -5.5390e-04,  6.0953e-04, -4.3923e-04,  1.0264e-03, -1.7704e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.6345e-04,  9.9354e-01, -2.8182e-04, -6.3454e-04,  2.8380e-04,
        -1.4591e-04, -6.7574e-04, -3.1898e-05,  4.0096e-04, -1.9744e-04,
        -4.2408e-04,  2.9974e-04, -7.2292e-05, -7.2828e-04, -4.6479e-04,
         2.9515e-04, -5.5967e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.9879e-04, -9.9487e-01, -1.1499e-04,  2.9943e-04, -5.9097e-04,
         2.0212e-04,  1.1251e-04, -2.6577e-04, -3.8964e-04, -2.3069e-04,
        -3.6335e-04,  2.9685e-04, -2.0415e-04,  2.1328e-04, -7.1595e-05,
         2.9079e-04, -3.2222e-04, -2.6607e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6652e-04,  9.9211e-01,  4.8480e-04,  2.8098e-05,  4.1405e-04,
         1.9356e-05,  4.0665e-04,  3.2780e-04,  7.8317e-04,  1.5542e-04,
         6.1797e-04,  2.5274e-05, -1.0678e-05, -2.2430e-04, -3.1544e-04,
         9.3352e-04,  2.1500e-04, -2.2686e-04,  1.9177e-04,  9.5799e-05,
         2.9040e-04, -4.9506e-05, -9.4945e-05,  5.7149e-04,  2.2465e-04,
         6.1862e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8255e-04,  9.9021e-01, -1.4047e-05, -5.8709e-04, -3.1793e-04,
        -1.8125e-04, -2.6496e-04,  6.5338e-04, -2.8702e-04, -4.5619e-04,
        -1.2750e-04, -2.6055e-04,  1.8536e-04,  2.6850e-04, -4.5667e-04,
        -5.0352e-04,  1.0798e-04,  2.7615e-04, -1.9834e-04,  3.1552e-04,
        -2.4099e-04, -2.6819e-04,  5.9893e-05, -1.4892e-04, -8.0065e-05,
         2.1450e-04, -2.8450e-04, -6.1633e-06,  1.2350e-04,  2.4335e-04,
        -1.7907e-05,  6.0015e-05,  2.5226e-04, -2.7097e-04, -2.2162e-04,
         5.8460e-04,  5.4291e-04, -4.2367e-05,  2.8029e-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([-1.2495e-04, -9.9171e-01, -3.6544e-04, -1.0512e-04, -6.0061e-05,
         1.3113e-04,  1.0533e-04,  2.6266e-04,  5.0345e-05,  1.3466e-04,
        -1.7544e-04,  1.0904e-04, -3.0704e-05,  1.5716e-04,  7.3520e-04,
        -1.6784e-04, -2.1862e-04,  1.4930e-04, -2.6529e-04, -1.6352e-05,
        -2.1855e-04,  8.2604e-05,  1.7108e-04, -1.0755e-04,  2.2612e-04,
        -4.9682e-04, -1.8732e-04, -7.3786e-05, -1.2062e-04, -1.8796e-04,
        -2.8541e-05,  2.4524e-04,  1.7152e-04,  4.6703e-06, -1.7078e-04,
        -1.1374e-04, -4.9366e-04,  1.5708e-04,  3.8287e-04, -1.8092e-04,
        -1.3634e-04,  2.1723e-04, -2.0798e-04,  5.9277e-05, -1.6978e-04,
         3.1419e-04], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 5.5696e-04, -9.9127e-01,  5.5538e-05, -2.0348e-04, -2.5404e-04,
         1.4457e-04, -1.8895e-04,  1.6540e-05, -2.0335e-04, -4.7049e-04,
        -8.6429e-05, -2.3497e-04,  1.6380e-04, -3.7471e-04, -4.9740e-04,
        -2.4003e-05,  2.9030e-04, -1.5414e-04, -4.6578e-04,  9.2421e-05,
         2.4280e-04,  1.6159e-04, -2.4980e-04, -8.1517e-05, -1.7343e-04,
        -1.7885e-04, -2.4504e-04, -2.3271e-04, -1.0347e-04, -2.8416e-05,
         4.6907e-05, -2.5482e-04, -5.0163e-04,  1.1758e-04, -5.8515e-04,
         2.9977e-04, -1.6926e-04, -4.0253e-04, -1.7489e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.7684e-05,  9.9135e-01, -9.7873e-05,  2.3672e-04, -4.4248e-04,
        -2.8141e-04, -3.4310e-04, -5.4363e-04,  1.8669e-04,  1.2389e-04,
        -5.6677e-04,  3.8339e-05, -3.5894e-04, -3.0711e-04, -6.8756e-05,
        -2.6055e-04, -3.1695e-04,  2.5158e-04, -7.4483e-04, -1.9086e-04,
        -1.1271e-04,  2.7463e-04, -8.0097e-05, -4.0118e-04, -9.9367e-04,
         2.7673e-05,  5.2766e-04,  2.6781e-04, -5.3718e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.5525e-04,  9.8930e-01, -3.6726e-04, -1.3132e-03, -1.0107e-03,
         4.8039e-04, -3.7256e-04, -8.4206e-04, -3.0215e-04, -9.4561e-04,
        -5.7334e-04, -1.8743e-03, -1.7826e-05, -1.2641e-03, -6.8428e-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,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8930e-04,  9.9313e-01, -4.0920e-04, -1.6113e-04, -5.7627e-04,
        -2.9617e-04, -2.3673e-04, -6.3452e-05, -3.2536e-04,  1.6041e-04,
        -6.4426e-05,  1.3890e-05, -3.2185e-04, -1.5378e-04, -1.7343e-06,
         3.4617e-04, -7.5001e-04, -2.1699e-04, -2.4196e-04,  5.3982e-04,
         5.7495e-04, -7.9986e-04, -3.2938e-04, -1.0121e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3600e-03, -9.9094e-01, -6.1761e-04,  8.0329e-04,  5.8838e-04,
        -7.1678e-04, -2.1270e-05, -8.1309e-04, -7.9979e-04, -1.0406e-03,
        -7.3173e-04,  3.5830e-04, -1.2115e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.6722e-04, -9.9090e-01,  3.9430e-04,  1.6894e-04, -9.5398e-05,
        -3.4586e-05,  3.6510e-04,  1.4685e-04, -1.4658e-04,  3.4083e-04,
        -4.3590e-04,  2.4858e-04,  2.2065e-04,  1.3511e-04,  1.7871e-05,
         4.6117e-04,  1.6646e-04, -2.2398e-04,  2.3637e-04, -1.3664e-05,
         4.9319e-04,  4.7611e-04, -8.2675e-05, -1.3751e-04,  1.3670e-05,
        -6.0366e-05,  1.4171e-05, -3.9350e-04,  1.7732e-04,  1.3702e-04,
         7.4580e-04,  2.8694e-04, -3.4706e-04, -1.7843e-04,  2.7484e-04,
        -1.7895e-05,  2.2883e-04,  9.2202e-05, -2.2069e-04, -1.0160e-04,
         1.9840e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.9870e-05, -9.9468e-01, -3.6192e-05, -9.3439e-05,  1.1992e-04,
         1.7333e-05, -1.5942e-04,  6.6857e-07, -2.0563e-04, -1.7778e-04,
        -2.7697e-04, -3.2200e-06,  3.7353e-05,  6.8942e-05,  5.1458e-04,
        -6.4376e-06,  7.5759e-05,  1.5436e-05,  8.4051e-05,  4.2949e-05,
         1.7824e-05,  3.2038e-05, -3.0548e-04,  1.0559e-04,  8.9131e-05,
        -9.4986e-06,  6.4095e-05, -2.8568e-05,  1.9284e-04,  3.0865e-05,
         1.5946e-05, -2.0088e-04,  2.8433e-04, -3.0516e-04, -4.1985e-04,
        -1.6683e-04,  1.1040e-05, -1.8244e-04,  3.8191e-04, -1.1685e-04,
         9.5273e-05,  1.5718e-04,  7.4874e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.4247e-04, -9.9281e-01,  1.1442e-04, -2.7816e-04,  3.9298e-04,
        -2.9306e-04,  3.6501e-05, -1.6553e-04, -3.3540e-04,  2.0552e-05,
        -4.6092e-04, -2.4533e-04, -4.7395e-04, -3.7235e-04, -1.8132e-04,
         8.8652e-04,  2.8599e-04,  2.7467e-04,  3.1490e-04,  3.1180e-04,
        -2.7773e-04, -2.0411e-04, -3.1828e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.9900e-03,  9.7701e-01,  2.2256e-03, -4.4339e-03, -8.4286e-04,
         2.0416e-03,  3.3680e-03,  1.0916e-03,  9.9547e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1674e-04,  9.9004e-01, -7.7689e-04,  4.3893e-05, -2.4733e-04,
        -1.9492e-04,  2.0036e-04,  5.9716e-05,  8.9715e-05, -3.7707e-04,
         6.9765e-05,  2.0861e-05,  1.6435e-04, -1.5155e-04, -9.0891e-05,
         5.2640e-04, -5.2441e-04,  3.0800e-05,  6.9527e-04, -4.3543e-05,
        -3.1674e-04, -1.4439e-04, -3.8718e-04,  3.3403e-04,  5.6884e-04,
        -8.0210e-05,  2.0296e-04,  8.8841e-05, -9.4801e-05,  8.9743e-05,
         4.6227e-05, -8.0484e-05,  2.4205e-04, -7.0263e-04,  6.7698e-04,
        -2.9452e-04, -3.3197e-04, -2.0174e-04, -4.6215e-04, -9.0682e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.5952e-04,  9.9032e-01, -1.2801e-03, -5.0893e-04,  2.3059e-04,
        -2.6950e-04,  1.2605e-03, -9.2391e-05,  1.1829e-03,  2.0976e-03,
         1.2691e-03,  4.5374e-04, -5.7122e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.9261e-04, -9.9458e-01, -5.2149e-05,  1.5882e-04, -7.3516e-04,
         1.7949e-04, -2.9549e-03, -5.4931e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+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([-1.8224e-03,  9.9294e-01, -9.6422e-04,  7.1386e-04,  1.4566e-03,
        -3.3913e-04,  7.3501e-04,  1.0252e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5854e-03, -9.9126e-01,  2.6380e-05,  3.4063e-04,  1.1805e-04,
        -4.2449e-05,  1.2001e-05,  4.4444e-05,  2.0674e-04, -1.7499e-04,
        -5.1011e-04, -3.3423e-04, -2.6569e-04, -1.0972e-04, -8.4553e-05,
         4.0957e-04, -1.5477e-04,  9.3856e-05, -1.4551e-04, -5.5697e-05,
         1.4158e-04, -1.2420e-04,  7.2863e-05,  2.4659e-04, -2.0526e-04,
         2.4830e-04, -4.8340e-05, -8.7361e-05, -1.1319e-04, -1.3811e-04,
        -8.4578e-05, -3.3384e-04, -9.2179e-05, -4.9454e-04, -5.9528e-04,
        -1.9603e-04,  9.6374e-05, -3.0818e-04, -3.7068e-04,  1.8976e-05,
         1.1181e-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([-5.4584e-05, -9.9505e-01, -3.6975e-04,  1.1435e-04, -8.8351e-06,
         7.5707e-05,  1.8063e-04, -1.5075e-04, -6.2594e-05,  1.7087e-04,
         5.7578e-05,  5.5232e-04,  1.8417e-04, -2.1890e-04,  2.1986e-04,
        -2.4675e-04, -1.5971e-04,  2.7338e-05,  3.8781e-04,  1.1298e-05,
        -1.7279e-04, -2.5820e-04, -2.3850e-04,  9.7046e-05, -1.5000e-04,
        -2.1453e-04,  1.0305e-04, -1.3235e-05,  3.2247e-04, -1.2365e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.3029e-04,  9.9146e-01,  4.8969e-04,  1.9113e-04,  2.0456e-05,
         5.1255e-04, -1.0547e-03, -3.5304e-04, -6.6230e-04,  4.3874e-04,
        -2.7685e-04,  4.1657e-04,  1.7361e-04, -4.6355e-04,  1.0819e-03,
        -3.3433e-04, -2.0406e-04,  1.1405e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8094e-03, -9.8924e-01,  2.1729e-04,  2.2668e-04, -1.1922e-04,
         3.0709e-04,  2.5907e-05,  3.1276e-04,  1.8655e-04, -1.0081e-04,
         3.5782e-04, -7.8262e-05,  2.6879e-05,  4.9094e-05,  1.6310e-04,
        -3.3676e-05,  5.5074e-05, -2.0648e-04,  2.2305e-04, -5.6292e-05,
        -1.8055e-04, -2.1595e-05,  3.1171e-04, -1.6506e-04,  6.9620e-05,
        -4.2822e-06,  2.3629e-04,  1.1417e-04,  6.8589e-05,  4.0635e-04,
         1.5168e-04, -9.4711e-05,  4.4491e-04, -1.9831e-04,  3.1265e-04,
         8.6497e-05,  2.4775e-04,  2.1063e-04,  1.2466e-04, -9.5544e-05,
        -3.5020e-05,  2.3650e-04, -8.7965e-05,  2.4955e-04,  3.8494e-05,
        -1.4064e-04,  2.5430e-04,  7.3479e-05,  3.0027e-04,  3.1751e-04,
         2.7663e-04,  5.1232e-04,  1.1791e-04, -2.0315e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3809e-03,  9.9331e-01, -4.5408e-04, -6.4651e-04,  1.6238e-04,
         1.9369e-04,  9.7766e-04,  7.7643e-04, -4.4272e-04,  3.3755e-04,
         3.1769e-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,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6337e-04, -9.9300e-01,  5.0554e-04,  3.9783e-04, -1.4671e-05,
         6.4073e-04,  5.4198e-05,  1.0711e-04,  1.5750e-04, -5.0259e-04,
         2.0291e-04, -1.6319e-04,  8.4825e-05, -5.8561e-05,  4.1412e-04,
         1.3660e-04, -2.7025e-04, -1.4639e-04, -1.8227e-04,  4.6537e-04,
         2.6178e-04, -6.5013e-04, -1.1630e-05,  1.8209e-04,  2.3802e-05,
         2.5284e-04,  4.6855e-04, -3.1748e-05, -1.4575e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.8950e-04,  9.9466e-01, -5.5935e-04, -3.6587e-04,  1.9328e-04,
         1.7722e-04,  3.1510e-04, -5.9126e-05,  2.1699e-05, -4.7405e-04,
        -2.8195e-04, -1.6888e-04, -5.6018e-04, -8.0833e-04, -3.6180e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4903e-03,  9.8868e-01, -4.4325e-04, -1.9292e-04,  1.6946e-04,
        -3.4113e-04, -5.6269e-06, -3.3667e-05,  8.9006e-05, -1.9313e-04,
        -9.3978e-05,  4.4465e-05, -5.7596e-05, -1.3917e-04, -7.7387e-04,
         6.0235e-04, -2.9720e-04,  4.3418e-04, -5.5055e-04, -1.9679e-04,
        -5.6284e-04,  1.4163e-04, -2.3811e-04, -8.4342e-04, -5.5322e-04,
         4.6807e-04,  2.7727e-04, -5.4311e-04, -4.1845e-04,  3.2785e-04,
         7.1828e-05,  2.3761e-04,  1.5511e-04, -3.2995e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3295e-03,  9.8851e-01, -2.8868e-04, -1.6959e-05,  1.8557e-05,
         2.3414e-04,  3.4412e-05,  6.0760e-04,  2.1858e-04, -5.3541e-05,
        -2.2936e-04,  4.2386e-05,  3.7258e-04,  2.5296e-04,  3.9559e-04,
        -6.0632e-04, -4.5486e-04,  2.0293e-04, -5.0431e-04, -8.2376e-04,
        -1.2543e-04,  4.6294e-04, -6.0037e-05, -3.0014e-05, -5.6127e-05,
        -3.9230e-04, -5.1722e-04,  1.0002e-04,  5.3133e-04, -6.4662e-04,
        -8.3558e-05,  2.5709e-06,  5.7655e-05, -7.9162e-04,  5.8780e-05,
        -5.1204e-05, -4.9060e-04,  2.8076e-04, -6.4549e-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([ 1.8084e-03,  9.8667e-01, -1.5483e-04,  4.9531e-04,  4.2387e-04,
         7.4475e-04, -5.8187e-04,  9.5333e-04,  7.9658e-05, -7.2747e-04,
        -5.1678e-04,  1.1346e-04, -2.5679e-04,  6.9909e-04, -6.2162e-04,
        -2.6908e-04, -6.1664e-05, -6.4697e-04,  7.8875e-04,  3.8020e-04,
         4.6613e-04,  2.5993e-04,  8.2441e-04,  5.6445e-04, -9.4881e-05,
         7.9335e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7231e-03, -9.8652e-01,  4.2426e-04, -2.2746e-05,  1.2959e-03,
        -4.1705e-04, -8.0331e-05, -6.2639e-04, -4.5088e-04, -6.6337e-04,
         9.0103e-04, -1.0377e-04, -8.9094e-05, -5.7321e-04, -5.5844e-04,
        -3.7453e-04, -8.4599e-05, -1.5144e-04,  3.9735e-04, -2.8355e-04,
        -2.3224e-04, -1.6601e-03, -6.5557e-04, -1.9736e-04, -5.1538e-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,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 5.0131e-04, -9.9374e-01, -8.1285e-04,  1.0420e-03,  4.5239e-04,
         2.2910e-04,  8.0456e-04,  6.4290e-04, -1.6041e-03,  1.5861e-04,
         1.4173e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4785e-04, -9.9372e-01,  1.8834e-04,  2.6025e-05,  1.3889e-04,
         2.9931e-05, -1.5519e-04, -2.7936e-05,  2.4201e-04,  2.1107e-04,
         5.5571e-05,  1.3360e-06,  4.7631e-05, -2.6931e-05, -2.4641e-04,
         1.5140e-04, -6.3287e-05, -2.7177e-05,  1.4829e-04,  2.1246e-04,
        -1.6172e-04,  3.2467e-05,  2.5582e-04, -4.2822e-05,  1.9986e-04,
        -1.0475e-04,  1.7160e-04,  1.3891e-04,  1.2749e-04,  9.4420e-05,
         3.6989e-04,  2.2354e-05,  7.1764e-06, -1.4742e-04,  3.6707e-06,
        -4.1128e-04, -8.0490e-05,  3.5063e-05,  4.3701e-05,  4.0225e-05,
         6.7742e-05,  2.1452e-05, -9.2401e-05, -2.3063e-04,  1.8810e-04,
         5.1547e-05, -2.2506e-04,  1.0766e-05,  1.8098e-04,  6.0707e-05,
         1.5937e-04,  1.5390e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.2405e-06, -9.9433e-01, -2.2120e-04, -5.1375e-04, -2.2154e-04,
         2.8288e-05, -2.3794e-04,  9.0053e-05, -4.5955e-06,  6.5963e-04,
         1.3400e-04,  1.0767e-04, -3.0193e-04, -1.5760e-04, -6.4985e-04,
         3.9680e-04,  7.1622e-05, -4.7554e-04, -5.1288e-05, -3.9261e-04,
         1.3364e-04,  2.4943e-04, -3.2389e-04, -1.6036e-04, -8.4377e-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([-4.4265e-04,  9.9233e-01,  4.9102e-05,  3.7463e-04,  3.6992e-04,
         3.5985e-04, -3.3233e-04,  6.3456e-04,  4.9030e-05,  2.8064e-04,
         1.0721e-03, -2.3595e-04,  1.8097e-04, -4.2973e-04, -2.1763e-04,
         1.2936e-04,  7.0566e-05, -1.4178e-03,  1.0066e-04,  1.1149e-04,
         5.2793e-04, -2.8549e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.3509e-04,  9.9081e-01, -2.8331e-04,  3.2709e-04, -2.1834e-05,
         2.9417e-04, -5.4082e-04,  1.5288e-04, -1.2349e-04,  1.2599e-04,
         1.0824e-04,  3.7657e-04,  3.4694e-05, -2.7905e-04,  2.4668e-05,
         1.0693e-04,  2.2467e-04, -4.9030e-04, -3.4113e-04,  3.8818e-04,
        -9.6656e-04, -7.2416e-05,  1.1816e-05, -5.7564e-04, -4.6721e-04,
        -1.2083e-04, -1.7161e-04, -3.1137e-04,  1.3483e-04,  4.5368e-04,
         2.6605e-04, -5.6103e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.5512e-04,  9.9120e-01,  4.6007e-04,  4.3883e-06, -1.7379e-04,
         3.5088e-04,  5.7407e-05,  1.5398e-04,  2.1511e-05, -3.4722e-04,
        -5.2335e-04, -7.8702e-04,  1.2385e-04, -3.7449e-04, -3.8888e-04,
        -6.5886e-05, -2.1491e-04, -1.0393e-04, -3.6021e-04, -1.7805e-06,
        -7.8936e-05, -8.9364e-05, -9.5628e-05, -3.8044e-04,  8.3096e-05,
        -7.5462e-05, -2.0363e-04,  2.7738e-04, -5.1360e-04, -3.4628e-05,
         3.9986e-04, -1.9608e-04,  6.8488e-04,  2.6831e-04,  1.5377e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6218e-04,  9.8969e-01, -5.7795e-04, -1.2157e-04,  3.3191e-04,
        -3.1489e-04,  6.1965e-04,  9.9743e-04,  9.1931e-04, -3.7240e-04,
        -3.9526e-04,  1.9431e-04,  5.1240e-04,  3.2834e-04,  4.1430e-04,
         2.1524e-04, -2.0336e-05, -2.1077e-05, -1.5946e-04,  1.3029e-04,
         4.3730e-05, -1.1315e-04,  1.9788e-04,  3.5342e-04,  9.1618e-04,
        -1.6785e-04,  4.2151e-04,  9.8791e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5222e-04,  9.9421e-01,  3.7548e-04,  2.4791e-04,  1.5557e-04,
         6.4731e-04, -6.6454e-04,  5.5305e-05,  2.1571e-04, -9.1240e-04,
         8.7159e-04,  2.4424e-04, -8.2231e-06, -5.9253e-04, -3.4967e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8966e-04, -9.9168e-01,  5.6592e-04,  1.0614e-04, -3.4019e-04,
        -1.2551e-04, -4.1322e-04,  5.9559e-04, -9.6593e-05,  1.3772e-04,
        -5.3817e-05, -5.5734e-04, -1.7973e-04,  1.5135e-04, -1.9554e-04,
         1.3406e-04,  2.9348e-05, -1.1253e-04, -2.8829e-05,  3.8584e-04,
        -8.6904e-05,  1.4767e-04, -2.1243e-05, -3.2422e-04,  9.7585e-06,
         3.1451e-04, -2.8048e-04, -4.6046e-04, -3.4958e-04,  4.3918e-04,
        -3.8322e-04, -2.2996e-05, -2.3288e-04,  2.1083e-05, -5.2327e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.6586e-04,  9.9296e-01, -2.7217e-04,  4.5633e-04, -6.4551e-04,
        -7.3365e-04, -7.6367e-05,  1.1511e-04, -1.8046e-04, -1.0876e-04,
        -4.2890e-04,  1.5018e-04, -6.1918e-04,  9.3289e-04, -3.5242e-05,
        -4.6157e-04,  7.9270e-05, -2.3580e-04,  5.0994e-05,  1.7083e-04,
         7.8480e-05, -1.8880e-04, -3.5319e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6394e-04, -9.9363e-01, -1.8336e-04,  2.9080e-04, -2.2765e-04,
         3.4309e-04,  1.3768e-04,  3.5852e-04, -1.5358e-04, -6.8862e-05,
         6.5755e-05, -1.5190e-04, -8.1180e-04, -3.2582e-04,  1.5115e-04,
         4.4909e-06, -1.9188e-04,  1.2525e-04,  2.0835e-04,  3.6715e-04,
         3.7839e-04,  6.5341e-05,  6.2092e-05,  1.8826e-04, -3.2267e-04,
        -1.2696e-04,  6.0674e-04, -9.3384e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.3862e-04, -9.9262e-01, -9.5779e-05,  1.3148e-04, -4.4791e-04,
        -7.9665e-05,  4.4505e-05, -1.9876e-04, -8.3455e-05, -3.6728e-04,
         2.6891e-04, -2.0508e-04,  3.7587e-04,  7.9793e-05,  1.0056e-04,
         3.3202e-05, -8.5209e-05,  2.1442e-04, -5.2204e-04,  9.6127e-05,
        -2.5023e-04,  3.4524e-05, -1.7495e-04,  1.4080e-04, -1.1350e-04,
        -2.0241e-04, -1.5806e-04, -2.0724e-06,  2.2282e-04, -1.1376e-05,
         1.3906e-04, -5.8752e-04, -1.9889e-04,  2.0049e-04, -4.3551e-05,
        -1.1545e-04,  3.0666e-04,  1.5960e-04,  4.4491e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 1.0303e-03, -9.9087e-01,  3.2388e-04,  5.9792e-04,  2.5111e-04,
        -1.9914e-04,  2.5605e-04, -2.1142e-06,  3.7326e-04,  5.0660e-04,
         1.7052e-04,  1.0054e-04, -3.8102e-04, -1.6030e-04, -7.6744e-04,
         7.7981e-04,  2.1754e-04, -1.1071e-04, -4.5404e-04,  2.3418e-04,
        -1.0105e-04,  7.5446e-05, -3.6098e-04, -3.0425e-04, -4.8961e-04,
        -1.2609e-04,  4.3548e-04, -7.6342e-05,  2.3977e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5186e-03,  9.8639e-01, -6.6446e-04,  9.0167e-04, -2.4748e-04,
         2.7240e-04, -6.1133e-04, -2.6339e-04,  4.5065e-04,  1.2601e-04,
         3.4937e-04,  1.6237e-04,  2.4946e-04, -1.0562e-05, -2.6115e-04,
         8.1891e-05,  1.2994e-04,  1.1978e-03, -6.7312e-04, -1.2083e-03,
         9.8709e-04,  2.3170e-04, -1.4141e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0948e-04,  9.9097e-01,  2.7929e-05,  6.0269e-04,  3.3674e-04,
         5.6579e-05, -5.0872e-04, -2.1176e-04,  1.6356e-04, -7.5338e-05,
        -1.4874e-04, -4.8899e-04, -5.3267e-05, -1.2921e-04,  7.9962e-05,
        -1.3285e-04,  3.1044e-04,  4.6635e-04,  2.0081e-04,  1.4039e-04,
        -1.2063e-04, -7.6677e-04,  1.5236e-04,  3.3039e-04, -4.5844e-04,
        -3.8313e-04, -2.6258e-04,  3.5479e-04, -1.3722e-05, -2.2209e-04,
        -7.4864e-05,  2.1808e-04,  3.2896e-04,  1.0075e-04, -5.0096e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3010e-04, -9.9095e-01, -4.7245e-04,  3.2101e-04, -7.1684e-05,
        -7.0277e-05, -3.1890e-04, -2.6595e-04, -1.0643e-04,  1.0184e-04,
        -2.0366e-04, -1.6834e-04, -2.9158e-04,  4.5968e-05, -1.9235e-04,
        -4.8252e-04, -8.9052e-06, -1.4875e-04, -1.0730e-05,  4.1066e-04,
        -9.1045e-05,  5.1737e-04, -1.9754e-04, -5.4537e-04, -4.0592e-04,
        -1.9863e-04, -8.1974e-05,  3.7911e-04, -1.1157e-04, -3.5320e-04,
        -5.6423e-04,  8.5793e-05,  7.5897e-05,  4.5004e-04,  9.7101e-05,
         5.5458e-05, -8.7136e-05, -2.2513e-04,  2.6211e-04, -2.0447e-04,
        -2.3795e-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([-4.4992e-04, -9.9400e-01, -6.0932e-05, -1.4723e-05, -5.1747e-04,
         9.0746e-05,  9.0646e-05, -9.8228e-05, -3.6378e-05, -1.2424e-04,
         7.2566e-05, -8.3383e-05, -3.8911e-04, -5.0617e-05,  2.2728e-04,
        -5.4798e-05,  1.6950e-04,  4.4987e-04, -4.1877e-05, -4.4176e-04,
        -5.7521e-06,  1.8968e-04, -1.9365e-04,  6.2122e-06,  2.9139e-05,
        -2.2690e-04, -4.7221e-04, -6.8145e-05,  1.6193e-04, -3.7104e-04,
         6.2317e-05, -1.5277e-04, -8.6723e-05,  3.9198e-05,  7.9387e-05,
        -7.4739e-05, -8.8187e-05, -2.2663e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8125e-04, -9.9446e-01,  5.4461e-04,  1.9814e-04,  1.8771e-04,
         1.9277e-05,  4.2338e-04, -2.2218e-04,  1.9236e-04, -1.7611e-04,
        -1.9385e-04,  2.1469e-04, -2.0270e-04,  2.2978e-05,  1.1055e-04,
        -1.9843e-04,  2.8884e-04,  8.1521e-05, -1.2786e-04,  3.8459e-05,
         6.9307e-05,  2.7637e-04,  2.0471e-04, -9.6777e-06,  1.1883e-04,
         2.1865e-04, -2.3707e-04,  6.9056e-05, -7.5471e-05,  3.0729e-04,
        -1.0026e-04, -1.9815e-05, -4.2403e-06,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.7425e-04,  9.9068e-01,  5.5140e-04, -2.2445e-04,  2.5267e-04,
         4.1912e-05,  1.5608e-04, -8.0063e-04, -3.6262e-04,  2.4563e-04,
        -4.5681e-04,  4.6807e-04,  2.7860e-04, -6.6284e-05,  6.1679e-04,
         2.8563e-04,  4.3787e-06, -5.5716e-04, -7.9793e-04, -1.1043e-03,
        -5.7990e-04,  4.0932e-04, -1.8430e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2839e-03, -9.9093e-01, -7.9167e-04,  4.3387e-04,  5.9340e-04,
        -4.2623e-05, -7.9573e-04,  5.8275e-05,  5.0791e-04, -4.6518e-04,
        -1.8602e-04,  7.0538e-04, -9.0842e-04, -6.1121e-04, -3.5687e-04,
        -1.0673e-03,  3.5711e-05, -2.2197e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8858e-04,  9.9381e-01, -5.5890e-05,  5.6281e-04,  2.0065e-04,
        -1.3460e-05,  2.3251e-04,  8.1953e-05,  7.0564e-04, -3.1992e-05,
        -8.8316e-05,  2.8820e-04,  3.2020e-05, -2.7644e-04,  2.6140e-04,
        -1.4436e-04, -5.3369e-05, -5.4064e-04,  4.9632e-04, -1.7622e-04,
        -5.0789e-05,  2.6863e-04, -2.7135e-04,  6.5829e-05,  1.0209e-04,
        -2.5485e-04,  3.4366e-04,  1.0233e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2834e-04, -9.9365e-01,  4.0691e-04,  1.4712e-04,  1.5442e-04,
         1.0534e-04,  3.3141e-04, -4.0921e-04,  1.3454e-05, -5.2244e-04,
        -7.2286e-04,  8.2073e-05, -1.6980e-04, -1.2330e-04,  3.5350e-04,
        -4.9230e-04, -2.1682e-04, -2.3918e-05,  1.6482e-04,  2.3692e-04,
         5.1104e-04,  2.9691e-04,  3.2061e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2827e-04, -9.9078e-01, -1.7402e-04,  1.0688e-04,  3.1200e-04,
         7.8202e-05,  4.8722e-05,  2.1735e-04, -2.4728e-04,  2.4160e-04,
        -6.1477e-05,  4.1885e-04,  1.2630e-04,  7.4590e-05,  3.4819e-04,
         2.5532e-05, -9.8124e-05,  1.2110e-04, -1.0580e-04, -1.6333e-04,
         2.2067e-04, -2.6128e-04,  5.0142e-04,  4.1982e-04, -1.3524e-05,
         2.6878e-04,  8.7460e-05,  1.3298e-04,  3.9749e-05,  5.6481e-04,
        -4.5097e-05,  1.5340e-04,  3.8005e-04, -1.6561e-04, -4.7335e-04,
         3.9473e-05, -2.3545e-04, -1.2302e-05,  1.5440e-04,  3.0913e-04,
         5.4445e-05,  2.1716e-04,  4.4749e-04, -1.2872e-04,  2.6882e-04,
        -3.9792e-04, -3.3187e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5323e-04,  9.9387e-01,  1.4141e-03, -8.0669e-05,  5.3903e-04,
        -6.7270e-04,  7.7089e-05, -2.5834e-05,  2.6457e-04,  3.6515e-04,
        -5.6879e-04, -3.6745e-04, -1.0809e-04, -9.5184e-05, -1.0304e-03,
        -3.6433e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 6.9251e-04,  9.9262e-01, -2.1405e-04, -2.5695e-05,  1.3539e-03,
        -2.8559e-04,  2.2128e-04,  3.8893e-04,  3.6944e-04, -1.4382e-04,
        -4.7085e-04,  2.5379e-05, -8.0945e-04, -1.0691e-04, -1.8380e-04,
        -1.5483e-04, -2.8021e-04,  2.4705e-04, -3.1967e-04,  8.3444e-06,
        -3.7588e-04, -3.3643e-04,  2.2649e-04,  1.3840e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3747e-04, -9.9531e-01, -2.3108e-05, -1.6840e-04, -2.7887e-05,
        -1.5376e-05,  3.4649e-04, -1.6721e-04,  1.1582e-04, -1.9835e-04,
        -6.8131e-05,  2.1786e-04,  1.9501e-04,  1.3458e-04, -1.6900e-04,
        -4.1709e-04, -5.7377e-05, -2.5594e-04,  9.4572e-05,  9.8554e-05,
        -7.6122e-05, -2.2956e-04,  1.6144e-04, -1.6961e-04, -2.3870e-04,
        -1.3682e-04, -6.3262e-05,  3.5585e-05,  1.3897e-04, -1.5403e-04,
         1.7350e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0925e-04,  9.9172e-01, -6.3826e-04, -1.2346e-03, -8.9285e-04,
        -4.8122e-04, -4.9424e-04,  3.9611e-06, -2.0035e-04, -1.2037e-03,
        -6.7728e-04,  4.9753e-04,  2.8200e-04,  1.7539e-04, -6.8951e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3278e-04,  9.9395e-01,  2.4543e-04, -1.1872e-03, -1.8524e-03,
        -2.9652e-04, -6.6691e-05, -4.7310e-05,  3.5087e-04,  4.6108e-04,
         6.0865e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6702e-04,  9.9188e-01,  4.6093e-04,  1.8865e-05,  7.0151e-04,
         3.2687e-04, -2.6996e-04,  4.8494e-04, -2.1257e-05,  4.4075e-04,
         6.9096e-04, -1.5189e-03,  1.0313e-03,  3.1127e-04,  1.8741e-04,
         5.3711e-04,  2.5038e-04, -3.7885e-04, -2.1226e-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([-7.1490e-04,  9.9343e-01,  5.7885e-04, -8.9948e-04, -8.4220e-04,
         4.2499e-05,  4.7430e-04, -5.9496e-04,  5.4939e-04, -3.4520e-04,
        -4.2952e-04, -5.6942e-04, -1.8706e-04,  1.8667e-04, -1.5490e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2936e-03, -9.9309e-01,  6.0270e-04,  7.6545e-05,  3.6769e-05,
         1.9116e-04, -4.2203e-04,  7.9159e-04, -6.1230e-04, -1.7706e-04,
         7.3261e-05,  2.1691e-04, -1.0619e-04,  4.4114e-04,  4.5237e-04,
        -5.3048e-04,  4.8036e-04, -3.1537e-05,  3.7851e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4626e-04,  9.9142e-01,  1.9268e-04,  4.9786e-05,  2.6236e-04,
        -4.8692e-04,  8.9470e-04,  4.4174e-05,  4.9154e-04, -1.9369e-04,
         4.5252e-05,  9.7425e-04,  6.0624e-04, -3.7284e-04,  7.8407e-04,
         4.4029e-04, -3.5026e-04,  6.9550e-05, -1.0880e-04, -1.3748e-05,
         3.8318e-04,  2.2069e-04, -4.8000e-04,  7.4161e-04, -2.7253e-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([ 1.1325e-03,  9.9104e-01, -5.8153e-04,  6.2167e-04, -1.0142e-03,
         8.2689e-04, -3.0997e-04,  1.5635e-05,  4.0651e-04,  2.8733e-04,
         9.1029e-04, -2.4082e-04,  5.4668e-05, -3.3770e-05,  2.4246e-04,
         1.3401e-03,  9.4572e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7536e-04,  9.8931e-01, -3.7520e-04, -4.2718e-04, -3.3467e-04,
        -1.9403e-04,  9.9715e-05, -5.9050e-05,  2.1386e-04, -1.4783e-04,
        -2.1807e-04, -4.0949e-04, -2.5859e-04, -4.8835e-04,  1.2537e-04,
        -1.3665e-04, -9.7006e-04, -2.3236e-04,  2.4488e-04,  1.9453e-04,
        -4.6527e-04,  2.3514e-04,  4.3842e-04, -3.1696e-05, -7.3405e-04,
        -5.9284e-04, -3.4875e-04, -3.4013e-04, -9.3309e-04,  9.9540e-05,
         7.7096e-04, -9.2154e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9599e-04, -9.9360e-01, -4.7210e-04,  2.4012e-04, -1.6960e-04,
        -3.2097e-04, -5.3500e-05,  4.3366e-04, -6.6098e-04,  3.0426e-04,
         2.5794e-04, -2.9338e-04,  5.5024e-04, -4.9775e-04,  1.6608e-04,
        -2.0477e-04,  1.1788e-03, -2.0850e-04, -1.8760e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8174e-03, -9.9271e-01,  8.7891e-04, -1.5292e-03, -7.6542e-04,
        -1.0424e-03,  2.9608e-04, -9.3269e-06, -3.1351e-04, -1.2320e-04,
        -5.1235e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #550: [tensor([ 6.5847e-04, -9.9355e-01, -5.2409e-05, -2.6178e-04,  3.7486e-04,
         2.8170e-06,  2.8696e-04,  1.2880e-04,  8.7696e-06, -1.5719e-04,
        -5.7633e-04, -1.2426e-04,  2.3672e-04, -1.4670e-04, -3.2058e-04,
        -3.4663e-05,  6.0830e-04,  2.1565e-04,  1.8899e-04, -2.9377e-04,
         1.7078e-04, -1.0132e-04,  5.0430e-05, -2.5489e-04,  2.3161e-04,
         3.7840e-04, -2.2412e-04, -2.3648e-04,  1.2779e-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,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3563e-04, -9.9505e-01,  1.0279e-04,  4.4727e-05, -1.4621e-04,
        -1.7490e-04,  6.1742e-06, -4.7074e-05, -4.4637e-06, -1.2037e-04,
         6.4320e-05, -3.1900e-04, -5.6614e-05, -2.3908e-05,  9.6131e-05,
        -5.3137e-06,  1.8491e-04, -1.8768e-07,  1.0943e-05,  2.3179e-04,
         7.4073e-05,  4.8439e-05,  7.2828e-07, -8.7697e-05, -1.1893e-04,
        -6.4057e-05,  1.9107e-04,  1.6819e-04,  3.2389e-05,  5.3753e-05,
         5.1534e-05,  1.8519e-04,  1.2414e-04, -6.7207e-05, -1.1437e-04,
         2.5719e-05, -1.6275e-04,  2.8257e-04,  1.3345e-04, -1.3474e-04,
         7.0720e-05, -1.3700e-04,  2.6163e-05, -8.8309e-05,  9.8789e-05,
         2.8676e-05,  1.0411e-04,  6.3310e-05,  6.9075e-05, -1.8723e-04,
         1.4812e-04,  3.2697e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6775e-04,  9.9615e-01,  1.5051e-04,  2.9449e-04, -2.4908e-04,
        -4.9451e-04,  1.4881e-05,  4.9239e-05, -2.1608e-04, -3.8995e-06,
        -5.5915e-04, -5.5605e-05, -6.3529e-04, -1.4009e-04, -2.8348e-04,
        -1.5382e-04,  1.5221e-04, -9.9255e-05,  2.7043e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5956e-04,  9.9218e-01, -5.6519e-04, -8.9802e-04,  3.7507e-04,
         9.8280e-05, -3.0647e-04, -7.7054e-04, -4.1473e-04, -6.6136e-05,
        -3.4116e-05, -1.6504e-04,  1.4875e-05, -1.3849e-04,  2.3210e-04,
        -1.5148e-04, -3.2831e-04, -6.2624e-04,  5.9867e-05, -7.0602e-04,
         1.5257e-04,  5.4611e-04,  9.6415e-05,  1.5064e-05, -4.9720e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5383e-04,  9.9383e-01, -9.9529e-04,  3.3872e-04, -2.1256e-04,
         6.9233e-04, -1.8922e-04,  7.7466e-04,  4.5835e-04, -4.5058e-04,
         3.7484e-04, -2.3400e-04, -3.2231e-04,  6.1532e-05,  8.1003e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1853e-04, -9.9112e-01,  2.4406e-04,  1.9968e-04,  2.5136e-04,
         2.0485e-04, -9.5002e-04, -6.0975e-04,  7.5448e-06,  5.9353e-04,
        -3.4226e-05, -3.4806e-04, -5.6583e-04, -3.7398e-04,  2.4392e-04,
        -3.5417e-04,  1.8725e-04, -8.1777e-04, -8.1625e-04, -1.1154e-04,
         1.1758e-04, -2.5392e-04,  5.4173e-04,  4.3839e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6358e-03,  9.8956e-01, -1.3895e-03,  1.3619e-03,  2.5819e-04,
        -1.2689e-04,  2.0005e-03, -4.3506e-04,  3.1437e-04, -3.4167e-04,
         5.7520e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7630e-04, -9.8858e-01, -9.4058e-05,  1.3362e-04, -4.0559e-04,
         6.3817e-05,  7.5182e-04,  7.5782e-04, -3.2920e-04, -1.0077e-04,
         2.5989e-04,  1.3647e-04,  7.5201e-04,  2.2011e-04,  6.9725e-04,
         7.2299e-04, -2.7484e-04, -1.3001e-04,  1.0768e-03, -1.1570e-04,
         1.6252e-04,  3.4596e-04,  3.4521e-04,  3.3911e-04, -4.0364e-04,
         6.1238e-05,  3.6668e-04,  7.3875e-04,  3.8152e-05,  5.7503e-04,
        -2.8289e-04,  1.6619e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0967e-04, -9.9281e-01,  4.9064e-04,  1.0288e-03,  4.8651e-04,
        -4.9565e-04, -3.8009e-04,  9.4760e-04, -5.1620e-04,  1.0613e-04,
         6.9325e-04, -4.2336e-04,  1.2692e-04, -6.4633e-04, -6.5591e-04,
        -8.7950e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6747e-04,  9.8978e-01,  3.6632e-04, -5.2066e-04, -2.5191e-04,
         1.0849e-03, -1.6179e-04, -2.5507e-04, -3.0694e-04, -2.3266e-04,
         4.9545e-04, -3.5259e-05,  4.5760e-04,  5.3185e-04, -1.0796e-04,
         2.1086e-04, -3.7018e-04, -2.9404e-05, -5.5390e-04, -1.1064e-05,
         3.0391e-04,  5.3735e-05,  3.3708e-04, -9.8534e-04, -3.5538e-04,
         8.4930e-04,  8.0319e-04,  7.4289e-05, -2.1681e-04,  8.9901e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2224e-04, -9.9338e-01, -5.6649e-04,  2.7448e-06,  5.4933e-04,
         6.6295e-04,  2.3161e-05, -2.6336e-04, -5.1950e-05,  6.9379e-04,
        -7.3022e-05, -2.5925e-04,  2.7683e-04, -1.0152e-03,  3.3790e-04,
         1.4140e-05, -6.6318e-04, -2.8087e-04, -6.6741e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.2388e-04,  9.9397e-01,  6.0852e-05, -4.6793e-05, -3.7197e-04,
         9.5757e-05,  5.1401e-05, -1.6300e-04, -4.0311e-05,  2.2039e-04,
        -4.3204e-04, -2.7845e-04, -1.0822e-04,  7.5093e-04, -2.0120e-04,
        -2.5792e-04, -4.2355e-04, -9.2347e-05, -5.4700e-04,  5.7133e-06,
        -3.7070e-04, -3.9851e-04,  3.2888e-05,  1.5722e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #50: [tensor([ 4.4777e-04,  9.9056e-01, -5.6595e-04, -6.8615e-05, -1.3073e-05,
         3.6234e-04, -3.4752e-04, -1.4796e-04,  4.6168e-05, -2.1148e-04,
        -1.7148e-04,  1.4478e-05,  1.4903e-04,  2.3711e-05, -8.4949e-05,
         2.0618e-04, -4.0702e-04, -1.6788e-04,  7.8686e-05,  1.6180e-04,
         1.1020e-04,  1.6187e-04, -6.4535e-04,  3.5605e-05, -7.8369e-05,
         1.7616e-04,  2.0393e-04, -1.1732e-04,  3.5102e-04,  2.4554e-04,
         1.6808e-04,  9.9251e-05,  2.6963e-04, -1.3035e-04, -1.0108e-04,
        -6.5461e-05,  7.6903e-05, -5.3041e-04,  8.2667e-05,  4.0436e-04,
         1.9385e-04, -4.8771e-04,  5.7102e-05,  4.7326e-04, -1.5512e-04,
        -3.4492e-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([-1.5879e-03, -9.9127e-01,  3.0537e-04,  4.7394e-04, -8.9413e-04,
         5.9002e-04, -5.8096e-04, -2.7209e-04,  2.2136e-04,  1.8689e-04,
         2.7343e-04,  3.9197e-04, -9.4797e-04, -4.7472e-04, -3.5705e-04,
        -2.3934e-04, -9.3600e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2841e-03, -9.8955e-01, -1.1041e-03, -4.6543e-04,  1.7190e-04,
        -1.1549e-04, -6.6061e-05, -8.1081e-04,  3.6815e-04, -2.6390e-04,
        -4.1694e-04,  1.5863e-04, -4.3584e-04, -1.0832e-04, -2.6975e-04,
        -3.5966e-04,  1.1134e-04, -1.9282e-04, -3.3573e-04,  5.1418e-05,
        -6.0572e-05, -4.5850e-04,  9.4842e-05,  9.4807e-04,  4.1414e-06,
        -1.1365e-04,  3.6148e-05, -1.2141e-04, -1.2187e-04, -3.7551e-04,
        -3.4203e-04,  5.2245e-05, -2.9252e-04,  2.2269e-04,  5.7496e-05,
        -5.5486e-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([-1.2274e-04, -9.9236e-01,  2.4262e-04, -5.9892e-06, -4.5033e-04,
        -8.4483e-04, -1.3193e-04,  1.7920e-04,  5.2090e-04, -4.9709e-04,
        -5.9783e-04, -1.0596e-03, -3.4456e-05, -1.6378e-04,  4.4775e-05,
         6.5971e-04, -1.1643e-03,  1.3382e-04,  5.3209e-04, -2.5052e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0045e-03,  9.9274e-01, -3.2038e-04,  2.4063e-05,  5.2281e-04,
        -3.0543e-04,  9.5266e-05, -2.0944e-04,  1.3629e-04, -3.7304e-04,
        -1.2527e-04, -5.9334e-04,  9.1965e-06,  3.5639e-04,  6.5075e-05,
        -6.2677e-05,  9.7069e-05, -2.3438e-04,  2.8753e-05, -4.2576e-04,
         7.4554e-05,  1.7971e-04, -4.3853e-06, -4.8476e-04,  3.8445e-04,
         1.9473e-04, -2.0889e-04, -3.7068e-04,  3.6566e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7879e-04, -9.9288e-01, -1.8844e-04,  2.4839e-04,  1.8621e-04,
        -1.0035e-04, -2.4936e-05, -4.8627e-04,  1.8490e-04,  1.1973e-04,
         5.2610e-05, -5.4320e-04,  5.1419e-05, -5.1668e-05, -2.8926e-04,
        -1.5028e-04,  2.0102e-05,  8.9652e-05,  1.1399e-04, -3.6274e-04,
        -9.4465e-05,  2.8476e-04,  1.2375e-04, -3.5287e-04, -5.2051e-04,
         4.5393e-04, -2.1787e-04, -9.5856e-05,  2.5619e-04, -2.9978e-04,
         2.5908e-04,  7.8859e-05, -5.3097e-05, -1.5567e-04, -1.5822e-04,
        -2.7171e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4826e-04, -9.9406e-01,  5.7744e-04,  3.9325e-04, -1.2509e-04,
        -1.6982e-04, -7.0346e-04, -5.3205e-04, -3.1075e-04,  3.2789e-04,
         4.0876e-05, -8.1968e-05,  3.4071e-04, -5.7791e-04, -3.9284e-05,
         2.3645e-05,  4.5289e-04,  5.9191e-04, -2.1017e-04,  9.4401e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3138e-04, -9.9399e-01,  6.8404e-05,  2.6565e-04,  8.5944e-06,
         4.5976e-05,  3.8886e-04,  2.2515e-04, -3.0768e-04, -2.1443e-04,
         5.4544e-05,  5.3896e-05, -2.3233e-04,  4.9477e-05,  1.7318e-05,
        -1.3942e-05, -2.9313e-04,  6.4403e-05,  6.1016e-06,  1.3336e-04,
         1.1326e-04, -4.3336e-05, -9.4342e-05, -7.5821e-05, -1.0544e-04,
         1.1833e-04, -1.3990e-04, -5.4991e-05,  1.5131e-04,  9.6889e-05,
         2.3474e-04, -1.3524e-04,  2.5544e-04, -3.6674e-04, -2.2480e-05,
        -8.1313e-05,  2.4235e-04, -1.0580e-04,  4.0253e-06,  3.9092e-04,
         1.8554e-04, -3.5227e-05, -1.2547e-05, -7.1004e-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([ 1.1568e-03,  9.8866e-01,  1.0371e-04, -2.8865e-05, -1.0521e-04,
         1.8187e-04,  1.0870e-04,  3.0793e-04,  2.0655e-04, -2.7150e-04,
        -1.0933e-04, -1.5349e-04,  2.4470e-04,  2.7718e-05,  4.0561e-04,
         1.5526e-04,  3.8677e-05,  1.7506e-04, -2.9954e-05, -1.7966e-04,
        -7.0243e-04, -2.5561e-04,  7.5712e-04, -1.2902e-04, -2.5338e-05,
         4.7083e-05,  7.9601e-05,  2.1015e-04, -5.2860e-05,  1.5933e-04,
         2.4963e-04, -5.1754e-05,  7.8290e-05,  1.4811e-06,  1.7448e-04,
         3.5737e-04, -4.9711e-04, -7.0185e-05, -1.1139e-04,  8.5288e-05,
        -1.6127e-04,  1.4383e-04, -5.2088e-04, -1.2580e-04, -5.0524e-04,
        -2.1738e-04, -3.9269e-04, -3.8076e-04,  1.0542e-04, -4.0984e-04,
        -2.6214e-04,  1.8405e-05,  1.5104e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.2067e-04, -9.9335e-01, -4.5867e-05, -3.0931e-04, -2.4354e-04,
         3.4378e-04, -7.8923e-05, -1.9469e-04, -5.2115e-05, -6.6368e-06,
        -1.4258e-04, -3.3635e-04, -3.3081e-04, -2.4306e-04, -1.4084e-04,
        -4.5073e-05,  7.7860e-05,  1.0294e-04,  1.5609e-04,  5.9996e-05,
        -3.4777e-04,  1.8129e-04,  4.1790e-05, -1.0792e-04, -2.5680e-05,
        -2.6261e-04,  3.3338e-04,  7.2760e-05,  3.2188e-04,  2.4434e-04,
        -1.8964e-04, -6.1702e-05,  6.5798e-05,  3.1467e-04,  2.4368e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8766e-03, -9.9043e-01,  9.4576e-05,  1.3897e-04, -5.8048e-04,
        -9.1093e-05,  3.2089e-04,  4.0946e-04,  2.3357e-04,  4.2017e-05,
        -3.7997e-04,  3.0384e-04,  3.0507e-04,  4.9232e-04,  8.4459e-05,
         1.5186e-04, -7.2073e-06,  2.0418e-04,  2.3232e-05,  5.4978e-04,
        -1.3959e-04,  5.2746e-04,  1.1009e-04, -4.2796e-04,  1.5605e-04,
        -3.7050e-04, -3.9706e-05,  4.8407e-04,  2.5448e-04, -1.1859e-04,
        -2.1750e-04, -4.3878e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.5033e-04, -9.9330e-01, -2.6232e-04, -5.2966e-05, -1.3913e-04,
         7.5765e-06, -1.9265e-05, -1.7030e-04, -1.0996e-04,  6.7768e-05,
        -8.9269e-05,  1.6649e-05,  3.6244e-04, -6.6860e-05, -1.6215e-04,
        -1.7928e-04, -1.1391e-04,  4.3441e-04, -1.8677e-04, -3.0730e-07,
        -1.4058e-04, -3.4600e-04, -1.1670e-04, -4.4384e-04, -1.1294e-04,
        -6.8149e-05, -8.4759e-05, -5.0895e-05, -4.3555e-04, -4.3288e-05,
         7.0535e-05, -2.2852e-04, -1.0205e-04, -1.0108e-04,  1.1684e-05,
        -3.8464e-04,  6.1600e-06,  1.9359e-04, -1.4992e-04, -2.4600e-04,
        -9.2172e-05, -7.8559e-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>)]
Iter #100: [tensor([ 2.7384e-04,  9.9651e-01, -8.7424e-04, -8.4890e-05, -1.6504e-04,
        -1.4738e-05, -2.5571e-04,  1.5301e-04, -1.6140e-04, -3.8951e-04,
        -1.1377e-04,  2.4831e-05,  8.2557e-05, -1.8353e-04, -3.8042e-04,
         1.4263e-04,  1.8571e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.9798e-04, -9.9595e-01,  1.4075e-04, -8.1152e-05, -6.0243e-05,
         4.4058e-06, -4.7392e-04, -8.8755e-05,  2.0714e-04, -2.4994e-04,
        -1.6208e-04,  1.1695e-04, -1.2309e-04,  6.8168e-06,  1.6160e-04,
         3.1758e-04,  1.2450e-04, -1.5360e-04, -6.7602e-05,  1.2698e-04,
        -1.3588e-04, -2.8919e-04,  2.1386e-04,  2.6619e-04, -9.8643e-05,
         7.8267e-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([-1.0173e-03,  9.9379e-01,  2.3756e-04, -2.7833e-04,  7.1183e-04,
        -5.7970e-04,  3.2667e-04, -9.6245e-04,  1.3286e-03,  5.9126e-04,
         1.7450e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6278e-04, -9.9268e-01,  6.5481e-04,  2.4614e-04,  8.9079e-04,
         4.4727e-04,  5.0900e-04,  5.3540e-04,  3.5973e-04,  7.4993e-06,
         8.7796e-04,  7.4317e-05, -6.5956e-05,  6.1274e-04, -2.4943e-05,
         1.1715e-03,  4.9484e-04, -8.2576e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.3099e-04, -9.9383e-01, -3.2191e-04, -5.2507e-04, -9.7049e-04,
        -7.8677e-05, -3.0000e-04,  5.4703e-04,  7.2799e-04,  7.7950e-05,
        -7.9796e-04, -5.0383e-04,  9.1475e-06,  3.8101e-04, -2.4665e-04,
         4.7917e-06,  5.0080e-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([ 5.4950e-05, -9.9288e-01, -4.4786e-05, -2.1597e-04, -2.9884e-04,
        -1.6598e-04, -1.7933e-04,  4.7919e-04, -1.2106e-04,  5.0861e-05,
        -4.1819e-05,  3.5755e-04,  4.5552e-04, -2.9531e-05, -4.2402e-05,
        -9.7889e-05,  5.4158e-04,  8.8464e-05,  1.3675e-04,  8.5075e-05,
        -4.7072e-04,  7.8583e-05, -4.7329e-05,  2.4007e-04,  1.8721e-04,
         4.3552e-04,  2.6105e-04, -4.0122e-04, -6.3925e-05,  3.7061e-04,
         8.1935e-04, -1.7657e-04, -7.6013e-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([-2.9487e-04,  9.9414e-01,  8.2827e-04,  7.6142e-05, -1.2293e-04,
         3.5389e-04,  4.2932e-04, -6.8023e-05,  8.4820e-04, -1.6286e-04,
         8.7890e-05,  1.1676e-04,  3.8300e-04, -4.1721e-04, -3.8738e-04,
        -1.5016e-04,  8.4618e-04, -8.4219e-05, -1.9895e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.6230e-04,  9.9519e-01,  3.9425e-04,  5.3975e-04,  1.1478e-03,
         5.8912e-06,  2.9456e-06, -7.9009e-04, -1.3704e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.7240e-04, -9.9341e-01,  1.3056e-04, -6.4671e-04,  8.0746e-05,
        -1.0407e-04, -5.5749e-04,  2.4418e-04,  4.4701e-05,  1.7135e-04,
         9.8950e-05,  5.1864e-04, -4.3206e-04,  1.8399e-04, -1.5624e-04,
        -2.0068e-04,  2.7718e-04,  3.9687e-04, -2.4463e-04,  1.8769e-04,
        -6.1673e-04, -4.9836e-04, -2.2273e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9396e-04, -9.9254e-01,  2.8470e-05,  1.1197e-04, -5.7243e-06,
        -8.5663e-05, -8.2169e-06, -3.8489e-04, -4.6812e-05,  9.0850e-05,
        -2.3358e-04, -7.1638e-05,  3.5880e-04,  3.5922e-05,  2.3171e-04,
         6.9578e-05,  1.2594e-04,  2.0491e-04,  2.6268e-04,  4.7438e-04,
         2.0280e-04, -2.5435e-04,  2.5459e-04, -3.9254e-05,  1.0055e-04,
        -2.0619e-04, -1.2934e-04, -2.2465e-04,  1.5594e-04,  9.1330e-06,
        -1.0217e-05, -9.0466e-05,  2.5007e-04,  2.8033e-04,  1.4607e-04,
        -5.6338e-05,  1.2017e-04, -5.2094e-04, -3.8296e-05,  2.7612e-06,
        -4.2331e-05, -6.9475e-05,  4.7186e-04, -3.5607e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8342e-04,  9.9294e-01,  9.4381e-05, -2.2204e-04,  3.3326e-04,
        -2.6521e-04, -6.2867e-04, -5.9157e-05, -5.1145e-04,  5.3725e-05,
        -2.4360e-04, -1.9177e-04,  2.3941e-04, -2.1873e-04, -2.0082e-04,
         1.0387e-04,  5.4045e-04, -3.9647e-04,  1.9376e-05, -6.9149e-05,
        -2.9820e-04, -8.6691e-05, -2.9243e-04, -1.7816e-04, -1.4804e-04,
         3.5125e-04,  1.8823e-04, -2.7308e-04,  1.8557e-05,  4.7482e-05,
        -1.6368e-04, -4.4071e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0680e-05, -9.9392e-01, -2.1311e-04,  6.1252e-05, -2.8432e-04,
        -3.3263e-05,  3.6870e-04,  1.5022e-04, -2.2138e-04, -5.5738e-04,
         2.4392e-04, -7.2769e-05,  1.6127e-04,  2.9309e-05,  6.5596e-04,
         3.1991e-05, -1.4431e-04, -1.2295e-04,  2.0063e-04,  5.7992e-04,
         2.7678e-04, -6.3527e-05,  6.9312e-05, -1.4122e-04,  2.6106e-04,
        -2.8666e-05,  1.0624e-04,  1.2085e-04,  6.7085e-04,  1.4702e-04,
         5.3259e-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 #150: [tensor([-2.0360e-04,  9.9448e-01, -5.4393e-04, -6.9255e-05, -1.0536e-04,
         1.7562e-04,  1.1952e-05,  1.4132e-04,  1.2529e-04, -2.8085e-04,
         1.1605e-04, -6.2528e-04, -1.6604e-04,  2.3714e-04,  1.8422e-04,
         4.3590e-04, -2.2291e-04,  4.1141e-04, -1.8947e-04,  8.9298e-04,
         3.5041e-04, -2.9517e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.5541e-04, -9.9329e-01, -4.5052e-05,  1.4006e-04, -2.0283e-04,
         1.9772e-04, -4.2690e-04,  2.7644e-05,  8.7462e-05,  8.6293e-05,
        -2.3200e-05,  1.3126e-04, -1.2489e-05,  2.5700e-04, -2.1628e-04,
         5.1583e-05,  8.5812e-06, -9.5794e-05, -1.4167e-04,  1.1728e-04,
         1.8583e-05,  2.1018e-04, -1.2973e-04,  2.5458e-04, -2.2276e-04,
         1.9177e-04, -1.0470e-06,  3.4160e-05,  5.2819e-05, -1.2160e-05,
        -1.6645e-05, -1.6709e-04,  2.7959e-04,  2.4879e-04, -1.9974e-04,
        -1.8414e-04,  1.1161e-04, -1.3724e-05, -9.4758e-06, -3.2091e-04,
        -3.3849e-04, -7.4588e-05, -1.9160e-04, -1.4277e-04, -9.0497e-05,
        -5.2948e-05, -3.8035e-05, -7.7527e-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([-6.6080e-04,  9.9120e-01,  1.1122e-03, -5.3954e-05, -1.0120e-03,
         9.1797e-05, -6.6470e-04, -3.6695e-04,  2.4025e-04, -1.1161e-03,
        -7.2985e-04,  1.0196e-03, -4.1153e-04, -1.4631e-04, -4.6433e-04,
         7.0757e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5699e-04, -9.9393e-01, -2.9158e-05, -1.0847e-04,  5.1049e-05,
         1.0898e-04, -1.6366e-04, -1.0759e-06, -1.8338e-04, -7.8989e-05,
        -1.1122e-04,  8.8677e-05, -4.3454e-05,  4.0325e-05,  1.4067e-04,
        -2.0672e-04, -1.6627e-05,  2.7326e-05, -4.5027e-05, -1.2364e-04,
        -1.4214e-04, -1.2719e-04, -4.5715e-05,  2.4891e-05, -1.6467e-04,
        -1.3783e-04, -5.7949e-05, -5.9605e-05, -1.1097e-04, -4.1466e-05,
         2.1328e-04, -1.8788e-04,  8.7189e-05,  9.4950e-05,  5.8486e-06,
        -6.4972e-05, -1.0345e-04,  4.3859e-05, -1.2007e-04, -1.1244e-04,
         1.7728e-04,  3.8901e-05,  3.9029e-05, -1.6209e-04, -1.2085e-04,
        -1.2188e-04, -3.2356e-04, -1.1478e-04, -1.4477e-04,  6.5587e-06,
         4.1840e-05, -1.0384e-04, -8.4643e-07, -2.9013e-04, -6.1093e-05,
        -3.2652e-04,  2.2958e-04, -9.1421e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8911e-04, -9.9171e-01,  1.0739e-04, -7.2924e-05, -3.8828e-04,
        -5.4397e-04, -1.3524e-04,  2.5998e-04,  1.6821e-04, -7.7402e-05,
         3.1609e-04,  4.5318e-05, -2.3248e-04, -4.9451e-04, -6.5464e-05,
         9.6503e-05, -1.9250e-04,  8.5073e-05,  1.3714e-04,  2.3934e-04,
         1.2747e-05, -1.3222e-04,  1.1883e-04,  4.6684e-04, -1.1658e-04,
        -1.2738e-04, -1.1035e-04, -1.3414e-04,  9.7119e-05, -2.3607e-05,
         2.0895e-04,  3.7012e-05, -1.2817e-05, -3.0377e-05, -1.0901e-04,
         6.1133e-05, -3.9142e-05,  1.8149e-04, -9.7535e-05,  4.7233e-04,
        -3.1664e-04, -3.5854e-05,  1.3082e-04, -5.4495e-04,  2.0596e-04,
         2.0075e-04,  1.2455e-04, -9.5458e-05, -1.0203e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1603e-04,  9.9187e-01,  8.1189e-05, -1.0550e-03, -4.4699e-05,
        -2.4795e-04,  1.9072e-04,  3.3334e-04,  3.0090e-04,  1.7956e-03,
        -3.4082e-04, -1.3839e-03, -9.2932e-04, -1.0087e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.3537e-04, -9.9362e-01, -1.7666e-04,  2.5031e-04,  5.1276e-04,
        -3.2331e-04, -4.4158e-04, -3.0420e-04,  4.5605e-04, -2.0399e-04,
        -9.0538e-04, -1.4784e-04, -7.2635e-04, -4.8500e-06,  7.2153e-04,
         3.1012e-04, -1.0117e-06, -1.0362e-04,  5.9196e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5771e-04,  9.9459e-01, -2.5076e-05,  9.5683e-06,  1.1817e-05,
         8.5274e-05, -1.0319e-04,  1.2601e-04, -2.2424e-04,  1.5248e-04,
        -1.3910e-04,  1.3739e-04,  4.4645e-05, -5.2656e-04, -2.9572e-04,
        -2.6622e-04,  7.7954e-04,  9.2260e-05, -3.8097e-04, -5.5224e-04,
         8.9329e-05, -1.0746e-04, -2.8318e-04, -1.9939e-04,  7.3629e-05,
         1.4621e-04,  7.9821e-05, -2.1180e-04, -1.1265e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2054e-04,  9.9382e-01,  1.2242e-03, -1.9220e-03,  9.6881e-04,
         1.8478e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7809e-04, -9.9394e-01,  2.1420e-04,  2.3183e-04,  1.3685e-04,
        -4.9733e-04,  1.3661e-04,  9.5858e-05,  2.0393e-04, -1.7367e-04,
         3.4603e-05, -7.7790e-05,  1.2128e-05, -4.8669e-04, -1.9588e-04,
        -1.8042e-04, -3.7525e-05,  2.6118e-05,  4.8759e-05, -1.6510e-04,
        -3.9323e-05,  9.7028e-05, -2.3244e-04, -2.3391e-04, -2.6095e-04,
         9.4862e-05, -1.4102e-04, -8.0456e-05,  2.6198e-04,  1.8570e-04,
        -1.9000e-05,  5.5556e-05,  1.2930e-04, -2.7925e-04,  5.1687e-05,
        -2.0742e-04, -3.4890e-05, -2.2505e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9185e-04, -9.9551e-01, -2.0928e-04,  3.1875e-04,  2.2415e-04,
        -4.3217e-04, -1.6253e-04,  1.3090e-04,  1.8248e-04, -9.4025e-05,
        -4.1736e-04, -1.7948e-04,  1.7272e-04,  2.1868e-05,  2.8544e-04,
        -1.1942e-04,  1.4271e-04, -3.3836e-05,  2.5421e-04, -9.5371e-05,
         4.3548e-05,  1.3933e-04, -2.6892e-05,  4.6804e-06, -1.0727e-05,
        -1.7059e-04, -1.4902e-04,  1.8514e-05, -1.6108e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1834e-03,  9.9175e-01, -5.7320e-04,  6.5331e-05,  3.5438e-04,
         1.2653e-04, -1.1047e-04,  5.3964e-04, -3.1959e-04,  3.9261e-04,
         1.2221e-04,  5.1733e-05, -1.8525e-05,  3.5139e-04,  6.4697e-04,
         3.0136e-04, -5.8687e-04, -6.4415e-04, -1.8938e-04,  2.1528e-05,
        -4.7127e-04,  5.2326e-04,  3.1713e-04, -3.3990e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #200: [tensor([-3.3805e-04, -9.9483e-01,  6.7820e-05,  2.0701e-04, -9.0053e-05,
        -1.2846e-05,  7.9402e-05,  6.1230e-05, -1.5919e-04, -8.7725e-05,
        -9.6627e-05,  3.3177e-04,  2.1390e-04,  4.5936e-05, -6.5300e-05,
        -1.8960e-04,  7.7351e-05,  1.6712e-04,  5.5027e-05,  1.8248e-04,
         4.9585e-05, -1.2449e-04, -1.4840e-04,  1.7569e-04,  2.3489e-04,
        -1.8871e-04,  3.3760e-04,  3.0655e-04,  7.1241e-05,  2.2128e-04,
        -7.0266e-06,  2.0865e-04,  1.8729e-04,  1.2947e-04, -4.5137e-05,
        -2.0745e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.3853e-04, -9.9238e-01,  2.3647e-04,  1.3832e-04,  3.1580e-04,
        -1.9467e-04, -1.5246e-04,  2.9670e-04,  4.6338e-04,  2.4550e-04,
        -8.8583e-05,  5.0338e-05, -2.4032e-04,  1.4025e-04,  5.6635e-04,
         2.2323e-04, -2.4478e-04, -1.2138e-04, -1.9232e-04, -1.1700e-04,
         6.7731e-05,  6.3994e-04, -2.1330e-04,  3.0861e-04,  1.6348e-04,
         1.2818e-04,  2.9132e-04,  6.1086e-05,  6.4733e-05, -1.6028e-05,
         4.2845e-04,  1.0039e-04,  9.7479e-05, -2.0974e-05,  1.4455e-04,
        -4.1589e-06,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8477e-04,  9.9594e-01, -2.3307e-05,  8.2693e-05, -4.7937e-05,
        -3.7428e-04,  1.0822e-04,  1.8841e-04,  2.6422e-05, -4.2103e-04,
        -2.6257e-04, -1.7994e-04,  1.4541e-04,  9.8792e-05,  5.6514e-05,
        -5.3610e-05,  9.3455e-06,  6.8413e-04,  7.1358e-04, -3.0137e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.1508e-04,  9.8761e-01,  1.1720e-04,  4.0973e-04, -5.1232e-04,
         4.2181e-04,  5.9242e-05,  1.6294e-05, -4.6047e-04,  2.2596e-04,
        -1.1432e-04, -1.0736e-03,  7.7886e-04, -5.1557e-04, -4.0467e-04,
         1.8581e-04, -7.4338e-05,  1.8380e-04,  9.6482e-05, -4.8627e-05,
         2.7360e-04, -2.4350e-04, -3.8071e-04, -9.9262e-04, -1.1543e-04,
         6.5566e-04, -2.2120e-04, -1.3305e-05, -5.0931e-04,  3.2789e-04,
        -5.0486e-04, -2.6690e-04, -2.5127e-04, -6.2750e-04,  2.0409e-06,
        -3.8416e-04, -1.5640e-04,  5.1983e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6077e-04, -9.9175e-01,  1.9520e-04, -5.4453e-05, -5.3049e-04,
        -7.0610e-05,  5.2321e-04, -1.3013e-04,  1.0967e-04,  1.3673e-04,
         7.8634e-05, -1.9846e-04,  2.5449e-04, -1.9560e-04,  4.7007e-04,
        -1.0565e-04, -3.6350e-05,  6.8416e-05,  4.7754e-04, -2.2155e-04,
        -3.1349e-04, -3.0617e-05,  3.1302e-04, -1.7574e-04,  2.7875e-05,
         1.0362e-04,  7.0671e-05, -1.1759e-04, -1.8481e-04, -8.4389e-05,
        -1.7738e-04, -3.0866e-06, -2.4890e-04, -4.5011e-04, -6.4407e-05,
        -1.3310e-04, -1.1200e-05, -3.6504e-04,  5.2341e-04, -1.3794e-04,
        -2.9611e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.6724e-04, -9.9450e-01, -1.1738e-04, -4.3722e-04,  1.0165e-04,
        -3.6697e-04,  1.1677e-04, -3.8787e-04,  4.1887e-04,  1.2445e-04,
        -2.5973e-04, -6.6134e-05, -2.7346e-05,  2.4619e-04, -3.9864e-04,
        -3.7552e-04,  3.5057e-04,  3.3084e-06,  5.9829e-04,  1.4877e-05,
         3.2874e-04,  9.5913e-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([-5.5335e-03,  9.8210e-01, -3.1505e-04,  5.2652e-05, -8.7278e-04,
        -9.6684e-05, -1.0205e-04, -7.2518e-04,  4.2726e-05, -5.9264e-04,
        -1.0375e-03,  2.1214e-04,  5.5365e-04,  2.0139e-04, -5.4689e-04,
         3.0314e-04, -6.4013e-04,  9.4375e-04, -4.8763e-05, -4.7041e-05,
         4.8969e-04,  5.6928e-04,  1.5112e-04,  2.1076e-04,  7.2482e-05,
         8.0160e-04,  4.4137e-04, -5.4906e-05, -2.0220e-04,  4.1538e-04,
        -3.9983e-04, -2.9583e-04, -1.9939e-04,  3.1810e-04,  4.1412e-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([ 6.0548e-05, -9.9278e-01,  5.1504e-04, -5.1435e-04,  4.0418e-04,
        -2.8729e-06, -1.7356e-04,  1.9399e-05,  3.1041e-04,  6.0363e-04,
        -2.3803e-04, -1.2956e-04, -3.3970e-05, -2.0375e-04, -4.9977e-04,
         7.7087e-04,  4.3957e-04, -2.7466e-04,  3.0240e-05, -5.1329e-05,
         8.6418e-05,  2.9100e-04, -2.2234e-04,  2.5579e-04,  2.3049e-04,
        -1.1237e-04,  3.2289e-04, -2.7243e-04,  1.5362e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0236e-04, -9.9351e-01,  2.7543e-04,  8.6951e-05,  1.2291e-05,
         1.0901e-04, -6.3352e-04,  1.1045e-04,  2.2677e-04,  1.7868e-04,
         2.9384e-04,  2.1336e-04, -5.2501e-04,  3.3800e-04,  2.5249e-04,
         2.8971e-04, -2.5258e-04,  1.6985e-04, -2.8193e-04,  7.5018e-04,
         4.2337e-04,  5.6544e-04,  1.6334e-04, -6.3200e-05, -1.7479e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1596e-03, -9.8691e-01,  3.4061e-04,  3.8491e-04,  1.2835e-04,
         4.5889e-04, -5.5631e-04,  2.2072e-04, -7.8941e-05, -8.2770e-05,
        -9.5968e-05, -2.8968e-04, -4.0533e-04, -6.9815e-04, -5.8888e-04,
        -7.0814e-06,  5.5171e-04, -1.2331e-04, -7.8518e-04, -6.2253e-04,
        -6.7691e-04, -9.7268e-06, -8.2158e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8630e-04, -9.9338e-01, -9.0792e-05, -2.2148e-05,  9.3560e-05,
        -5.4545e-04,  4.5304e-05,  8.3195e-05, -1.9331e-04, -2.4999e-04,
        -2.1618e-04,  9.5821e-05, -1.8163e-04,  5.4145e-05,  1.4078e-04,
        -6.0961e-04, -3.3542e-05, -8.8973e-05, -1.5653e-05,  3.3511e-04,
        -7.3465e-05,  1.5658e-04, -9.1020e-05, -5.5307e-05,  2.4699e-04,
         7.0005e-05, -7.4750e-05, -8.2636e-05,  3.8298e-04, -2.7129e-04,
        -5.8021e-05,  2.5869e-07,  2.9030e-04, -2.8359e-04, -2.5003e-04,
        -3.1524e-04,  2.0740e-04, -1.3754e-04, -9.5715e-05,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6037e-03,  9.9318e-01, -1.5520e-03,  3.4830e-04,  5.7563e-04,
         1.3026e-03, -7.2491e-05,  5.1733e-04,  2.8872e-04, -4.5406e-04,
        -1.0291e-04, -4.3656e-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], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 6.7492e-04,  9.9296e-01,  4.5753e-04, -1.5800e-04,  8.2522e-05,
         3.2439e-05, -1.8176e-04, -5.2184e-04, -2.4535e-04,  3.9563e-06,
        -3.4834e-04,  1.2079e-04, -5.8625e-05,  1.1019e-05, -3.9346e-04,
        -3.7652e-04, -4.7789e-04,  3.3151e-04, -1.7388e-04, -9.7454e-05,
        -7.3883e-05, -2.3694e-04, -1.2736e-04,  5.5758e-04,  1.2662e-04,
        -5.0988e-04,  3.1687e-04, -7.9755e-05, -2.0714e-04,  5.3818e-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([ 4.9395e-04,  9.9341e-01,  8.2375e-05, -1.2689e-04,  2.0971e-04,
        -1.0802e-04, -3.1642e-04,  1.7709e-04,  3.1640e-04, -1.9870e-04,
         1.6104e-05, -4.6398e-05, -1.3301e-04, -3.2743e-04,  8.5367e-05,
        -1.5770e-04, -2.4451e-04,  1.1759e-04,  3.8820e-04, -2.9856e-04,
         2.3533e-06,  5.6361e-05, -1.7437e-04, -2.5030e-04, -1.0052e-05,
        -9.8335e-05,  8.3640e-05,  3.9655e-07,  1.0927e-04, -1.4565e-04,
        -2.7494e-04, -3.1433e-04,  3.1859e-04, -3.2025e-04, -2.8203e-04,
        -1.9049e-04, -3.1411e-06, -1.0775e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7644e-03, -9.9160e-01, -3.5660e-04,  1.1329e-04,  1.7678e-04,
        -3.9924e-04, -3.4306e-04,  1.0068e-04,  2.1024e-04, -4.5450e-04,
        -1.2979e-04, -2.0493e-04, -2.4149e-04, -3.7155e-04, -2.9635e-04,
        -2.3090e-04,  1.6092e-04, -2.5055e-04, -2.6355e-04, -1.0516e-04,
        -1.7105e-04, -3.2130e-04, -4.1238e-05, -8.4528e-05,  2.1923e-04,
        -1.6814e-04,  8.8090e-05, -2.7313e-04, -5.1709e-04, -3.4228e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1750e-05, -9.9407e-01,  5.7111e-04,  2.1018e-04,  1.2913e-03,
        -1.2564e-04, -6.4539e-05,  7.8980e-04,  3.8065e-04,  2.1670e-05,
        -3.9979e-04, -1.8572e-04, -3.0414e-04,  2.1061e-04, -5.1827e-04,
        -2.3466e-04,  1.7995e-04, -2.4811e-06,  3.9478e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0848e-04, -9.9218e-01,  5.0459e-04,  7.6750e-04,  1.7871e-04,
         3.9613e-04, -6.3805e-05,  8.2122e-07, -6.4200e-04, -4.4453e-05,
         7.4023e-04,  2.6661e-04,  5.4063e-04,  3.5739e-04, -4.6968e-04,
         5.6062e-06,  5.4573e-04,  1.3933e-04,  5.9799e-05,  2.4679e-04,
         1.6992e-04,  1.2519e-04,  2.8217e-04,  5.1659e-04,  3.5748e-04,
        -1.9159e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.0825e-04, -9.9234e-01, -4.2292e-04,  4.1334e-04,  1.1422e-03,
         1.0359e-04, -1.2211e-04,  2.0059e-04, -7.9520e-05, -3.6217e-06,
         6.2596e-04,  2.2537e-04,  8.9907e-05,  7.4122e-05,  2.1864e-04,
         2.2657e-04, -6.3001e-04, -4.2455e-04,  2.1362e-04,  2.8817e-04,
         9.9594e-04, -1.6634e-04, -8.1989e-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([ 6.5042e-04, -9.9182e-01, -1.3738e-04, -4.0085e-04,  6.6399e-04,
         1.2005e-03, -5.2107e-04, -8.5176e-04, -6.3121e-04, -1.3167e-04,
        -4.0282e-05,  2.4749e-04,  2.8700e-04, -1.5555e-03, -8.5653e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2609e-03, -9.9410e-01, -3.8610e-04,  5.9713e-05, -9.7143e-05,
        -6.3616e-04,  1.1412e-04,  8.5788e-04, -1.5157e-04, -1.8240e-06,
         2.0667e-04,  7.6582e-04,  3.8885e-05,  2.1525e-04, -4.7721e-04,
        -3.1356e-04,  3.1408e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.4892e-04,  9.9555e-01, -8.7335e-05,  1.8599e-04, -3.8567e-04,
         1.2833e-04,  2.3667e-04, -1.4008e-04, -7.7857e-05,  6.5589e-04,
         8.2023e-05,  3.1910e-04,  2.1438e-05,  1.3978e-04,  2.6972e-05,
        -6.1149e-04,  3.0462e-04,  1.9937e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0275e-05,  9.9516e-01, -2.6689e-04, -4.6150e-04,  2.9322e-04,
         5.2421e-04,  4.9828e-05, -9.9642e-05,  1.9532e-05,  3.0086e-04,
        -1.9582e-06, -2.5322e-04, -2.1811e-04, -3.9266e-04, -1.0494e-04,
        -2.4391e-04,  1.1505e-04, -1.7441e-04, -1.7641e-04,  1.1544e-04,
         2.2291e-04,  2.2322e-04,  2.5252e-05,  1.6278e-04,  1.3795e-04,
         2.1280e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4269e-04,  9.9036e-01,  6.0187e-04, -4.1639e-04,  1.9700e-04,
        -6.4597e-05, -6.3926e-04,  1.7345e-05, -2.1294e-04, -3.8008e-04,
        -2.3446e-04, -2.1866e-05, -2.0957e-04, -1.5653e-04,  4.6954e-05,
        -3.9669e-04, -2.3096e-04,  2.7070e-04, -3.1181e-05,  4.5941e-04,
        -1.9288e-05,  1.6450e-04, -5.4060e-05, -7.8834e-04, -5.3607e-05,
         2.8377e-04,  1.9915e-04, -8.2146e-04,  3.9306e-04,  1.2063e-06,
        -3.7979e-04, -5.4993e-04,  2.1359e-04, -1.0469e-04,  1.5884e-04,
         1.9772e-04,  2.1214e-04, -8.6034e-05, -3.3075e-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([-1.3344e-04, -9.9432e-01, -1.9357e-04, -7.3648e-05,  4.0177e-05,
         9.1596e-05,  9.0796e-05,  3.2747e-04, -9.4047e-05,  3.6744e-06,
         9.4389e-05, -1.4161e-04, -3.6448e-04,  1.0738e-04,  1.5290e-04,
        -5.3147e-05, -4.8850e-04,  1.8902e-04,  1.4962e-04, -3.1384e-05,
         2.5884e-04,  1.7920e-04,  1.6354e-04,  5.0510e-05, -1.2954e-05,
         2.6610e-04,  2.2582e-04, -8.3667e-06,  1.0836e-04,  7.0761e-05,
         7.3354e-05,  5.9519e-05,  2.6337e-06,  9.3229e-06, -7.8912e-05,
        -1.5404e-04,  3.0806e-05, -1.1577e-05,  4.3854e-05,  2.4618e-04,
        -2.0099e-04,  7.3895e-05,  2.4555e-04,  1.9248e-04, -4.0228e-05,
        -5.4678e-05], device='cuda:0', grad_fn=<DivBackward0>)]
