Iter #50: [tensor([ 0.4783,  0.2036,  0.0061, -0.0015, -0.0069,  0.0097,  0.0093, -0.0065,
        -0.0089,  0.0083,  0.0109,  0.0060,  0.0059, -0.0104,  0.0059,  0.0062,
         0.0109,  0.0022,  0.0085,  0.0087,  0.0119, -0.0016, -0.0085,  0.0190,
        -0.0058, -0.0170, -0.0134, -0.0081,  0.0095,  0.0094, -0.0104,  0.0102,
         0.0055, -0.0014,  0.0169,  0.0031,  0.0127,  0.0091, -0.0047,  0.0073,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.4485e-01, -1.1148e-01,  1.1089e-03,  8.0308e-03,  7.8538e-03,
         1.3747e-02,  5.8387e-03, -7.1447e-03, -1.3542e-02, -6.3363e-03,
         4.8195e-03, -1.5518e-02,  5.9583e-03,  2.1371e-03, -1.0028e-02,
        -6.1550e-04, -8.5650e-03, -2.0237e-04,  3.7814e-03,  1.2190e-03,
        -1.3587e-02, -4.0997e-03, -1.1579e-02, -6.2430e-03,  4.6674e-03,
        -9.4652e-03,  8.4178e-03, -1.7028e-03, -1.2324e-02,  5.6041e-03,
        -5.2264e-03, -1.8074e-03, -7.0200e-03,  1.1299e-03, -3.5039e-03,
        -1.4332e-02,  3.0731e-03, -5.6398e-04, -1.7430e-03, -1.1968e-02,
        -1.5973e-02, -1.9193e-03, -1.0576e-02, -2.5323e-03,  9.9701e-03,
        -3.2544e-03, -1.8648e-03,  6.4910e-03, -6.4379e-03, -5.2702e-03,
         4.7513e-04, -1.2314e-02, -1.3651e-02, -3.2872e-03, -3.9243e-03,
         1.2278e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6324e-01, -2.6687e-01,  1.4500e-02,  3.1681e-03, -9.5154e-03,
        -2.7699e-02, -1.2877e-02, -9.2439e-03, -1.4136e-02, -1.0655e-02,
        -2.4980e-02,  2.3351e-02,  2.0414e-02, -1.6000e-03, -5.1004e-03,
        -2.3032e-03,  5.1943e-03, -3.0626e-04, -3.9563e-03, -2.3091e-02,
        -1.2352e-02,  9.2056e-03,  6.2575e-03, -2.3807e-02,  2.5344e-04,
         8.9672e-03, -1.9784e-02,  7.2621e-03, -8.3821e-03,  6.1206e-03,
         1.2806e-02, -2.2359e-02,  9.7150e-03,  1.0525e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.4354, -0.2858, -0.0072, -0.0524,  0.0094,  0.0105,  0.0437,  0.0196,
         0.0322,  0.0543, -0.0176,  0.0046,  0.0272,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.2084, -0.2872,  0.0336, -0.0033,  0.0013,  0.0136, -0.0368, -0.0109,
        -0.0145,  0.0173, -0.0062,  0.0024,  0.0051, -0.0410, -0.0092,  0.0359,
         0.0017, -0.0162, -0.0308, -0.0169, -0.0044, -0.0027,  0.0133, -0.0157,
         0.0374, -0.0486,  0.0146, -0.0325, -0.0179,  0.0206,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7828e-01, -2.8104e-01, -7.1052e-02,  2.1108e-02,  5.9736e-03,
         3.7071e-03,  3.3083e-03,  3.3149e-02, -9.9057e-03, -9.4453e-03,
         3.0130e-03,  1.1706e-02,  9.2460e-03, -2.4228e-02, -6.0117e-02,
        -2.7495e-04,  3.0572e-03,  2.5899e-02, -4.0903e-03,  9.5370e-03,
        -9.6580e-03, -1.9958e-02,  3.3359e-03, -5.5094e-02, -3.4398e-02,
         2.9041e-02, -2.2758e-02,  1.8737e-02, -1.4508e-02, -2.4376e-02,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5534e-01,  1.9609e-01, -1.0034e-02, -7.5747e-03,  1.3552e-02,
        -3.6606e-03,  2.6776e-02,  4.3972e-04, -1.5450e-02, -3.2152e-03,
        -5.9313e-03, -8.3682e-03, -1.9217e-02, -9.4896e-03,  1.3905e-03,
         4.8048e-03, -9.0077e-03,  6.3084e-04,  1.0443e-02, -4.4177e-03,
         7.1720e-03, -1.9893e-03, -2.2694e-02,  2.6985e-02,  1.7453e-03,
         1.1221e-02, -4.0379e-03, -3.3067e-03,  1.5014e-02,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0237, -0.5469,  0.0066,  0.0321, -0.0440, -0.0508,  0.0544,  0.0461,
         0.0571,  0.0465, -0.0400,  0.0349,  0.0168,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4287e-01, -3.6181e-01,  9.5295e-03, -3.0725e-04,  4.5170e-03,
        -2.1331e-02,  1.0102e-02,  3.4347e-03, -1.7597e-02, -4.8366e-03,
         1.1894e-02,  5.3285e-03,  2.1007e-03, -6.4333e-03, -1.1761e-02,
         2.3550e-03,  1.2911e-02, -6.5448e-03, -3.6123e-02,  7.2803e-03,
         1.6861e-03,  1.9239e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.4027, -0.2763, -0.0117,  0.0201, -0.0099, -0.0097,  0.0038,  0.0074,
        -0.0151, -0.0132, -0.0228, -0.0063,  0.0045, -0.0185, -0.0009, -0.0032,
         0.0229, -0.0025, -0.0084, -0.0231, -0.0173, -0.0161, -0.0076, -0.0170,
         0.0235,  0.0111, -0.0008,  0.0134, -0.0104,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.4188, -0.2646,  0.0024,  0.0081,  0.0169, -0.0016,  0.0054, -0.0166,
         0.0054,  0.0005,  0.0124, -0.0199,  0.0288,  0.0151,  0.0096, -0.0138,
         0.0012, -0.0171, -0.0036, -0.0242,  0.0202,  0.0084,  0.0122, -0.0193,
         0.0090, -0.0023,  0.0138,  0.0055, -0.0232,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1544, -0.3722, -0.0542, -0.0256, -0.0415,  0.0040, -0.0249,  0.0324,
        -0.0164, -0.0149,  0.0223,  0.0059, -0.0470, -0.0319, -0.0293, -0.0086,
        -0.0516,  0.0244, -0.0383,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  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([-1.3621e-02, -8.5100e-01,  1.6824e-03,  7.8856e-03,  1.8582e-03,
         4.7752e-03,  2.0115e-03,  4.0429e-03,  7.2906e-03,  2.8090e-03,
        -3.9835e-05,  8.7919e-03,  4.7265e-03,  7.3167e-03,  1.8745e-03,
         8.5004e-03, -3.7760e-03,  5.8281e-03,  7.6929e-04,  1.4963e-03,
         3.9789e-03, -4.1738e-03,  2.0657e-03, -3.6046e-03, -1.6696e-03,
        -1.7490e-03, -5.4413e-04,  5.2449e-03,  1.1611e-02, -1.3717e-03,
        -1.7316e-03, -6.1110e-03,  3.6895e-03,  1.3643e-03,  3.8295e-03,
        -5.3330e-03, -1.8344e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1908e-02, -8.1429e-01, -2.2279e-03, -2.6942e-03, -2.7635e-03,
        -1.9812e-03, -7.6441e-03, -4.8636e-03,  2.0032e-02,  2.7466e-03,
        -7.5092e-03, -1.0966e-03, -6.0563e-04,  9.2856e-03, -7.9046e-03,
         3.8249e-03,  7.2219e-03,  6.1622e-03,  6.7283e-04, -3.2130e-03,
        -2.0915e-03, -3.4390e-03,  7.3023e-03, -7.1373e-03, -7.5868e-03,
         1.3360e-02,  6.1879e-03, -2.7725e-03, -7.1495e-03,  5.8679e-03,
         1.3706e-03, -4.8950e-03, -2.1887e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7489e-02, -8.0795e-01, -1.2553e-02, -8.9213e-03, -7.6394e-03,
        -4.8175e-03,  4.4224e-03, -8.9194e-04,  1.8378e-02,  1.3995e-03,
         1.9670e-02,  2.3115e-03, -3.2697e-03,  9.4044e-03,  5.6340e-03,
         1.6261e-02,  1.6297e-04, -5.6649e-03, -1.1379e-02,  2.4191e-04,
         7.8302e-03,  3.7060e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7345e-01, -6.8997e-01, -2.0323e-03,  5.6810e-03, -4.8066e-03,
         3.2695e-03,  3.4600e-03,  1.4736e-02,  8.3683e-03, -3.8863e-03,
        -4.2830e-03,  1.7119e-04, -5.9050e-03, -2.4226e-04, -5.5886e-03,
         5.3772e-03, -4.9596e-03,  5.3649e-03,  8.1138e-05, -3.4262e-03,
        -2.4220e-03,  2.5038e-03,  2.5411e-03,  1.2164e-02, -5.3629e-03,
         1.7855e-03,  9.1969e-04,  2.1929e-03, -2.8968e-03, -5.0332e-03,
        -3.3436e-03, -4.0217e-03,  3.5318e-03, -6.2262e-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([ 3.6827e-02,  9.1437e-01, -1.7628e-03,  1.3470e-03,  3.6270e-05,
         2.7183e-03,  2.6352e-03,  1.5739e-03, -9.7019e-03, -4.4683e-03,
         2.6308e-04,  5.9614e-03, -1.1906e-02,  2.2504e-03,  4.1748e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2919e-02, -8.9828e-01,  2.2978e-03,  6.3038e-03,  4.6821e-04,
        -3.2195e-03, -2.6920e-03, -9.5335e-05, -5.6318e-04, -1.3726e-03,
        -8.1386e-03,  4.8627e-03, -1.1260e-03,  1.7576e-03, -2.2603e-03,
         3.7218e-03,  5.1332e-03, -4.9471e-04, -9.5355e-03,  1.6042e-03,
        -3.1569e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.2142e-02, -7.4988e-01, -7.4543e-03, -5.3936e-03, -3.5043e-04,
        -3.0960e-03,  8.4389e-03, -2.9634e-04,  6.2173e-03,  9.9642e-04,
        -2.0724e-03, -5.4350e-03,  1.2955e-02,  4.2555e-03,  4.9383e-03,
         2.7103e-02, -6.4628e-03,  5.3517e-03,  4.7712e-03, -6.9806e-03,
        -6.2402e-04, -8.0406e-03,  1.5584e-02,  7.4603e-03, -1.0145e-03,
        -9.6009e-04,  3.9190e-03, -1.0987e-03, -2.3032e-03, -2.8478e-03,
        -1.0187e-03, -4.9879e-03,  1.2448e-02,  2.4813e-03, -3.4326e-03,
         1.5133e-03, -2.0432e-04,  8.3934e-04, -4.3373e-03, -2.9022e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.0994e-02, -8.1560e-01,  1.4874e-02,  1.5099e-03,  3.0257e-03,
         6.3639e-03, -5.8869e-03, -1.5709e-03,  8.3542e-03, -2.9203e-03,
        -1.2033e-02,  1.0866e-02,  4.8573e-03, -1.1293e-02,  9.1235e-03,
         8.1652e-03, -3.8306e-03, -3.3668e-04,  3.2443e-03,  1.6741e-03,
         3.4800e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6668e-02,  9.0850e-01, -9.0135e-03, -6.5020e-03, -5.3508e-03,
         4.3297e-04,  4.1957e-03, -5.0155e-03,  3.2813e-03, -1.9896e-03,
        -1.3435e-03,  5.9418e-03, -2.0392e-02, -2.9803e-03, -5.0691e-05,
        -3.1023e-03, -5.2412e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3328e-02,  7.9374e-01,  2.6426e-03, -7.2808e-04,  1.7376e-02,
        -2.1633e-02, -9.8890e-03, -7.4897e-03,  3.3568e-03,  1.0589e-03,
         5.4698e-03, -1.3848e-02,  1.3289e-02, -4.6909e-03,  3.4973e-03,
        -1.0097e-02,  2.7530e-03, -4.4204e-03,  1.3923e-03, -3.4364e-03,
         1.4696e-03,  1.0070e-02,  2.2657e-03, -5.1228e-03,  4.2549e-03,
         2.6783e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1320,  0.7688, -0.0213,  0.0020, -0.0042, -0.0205, -0.0099,  0.0028,
        -0.0058, -0.0013,  0.0124,  0.0126,  0.0022, -0.0042,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6484e-03, -8.2330e-01,  1.2274e-02, -7.4148e-03,  1.6886e-03,
         3.6764e-03,  7.9204e-03, -1.9701e-03,  1.5358e-02,  2.9397e-04,
        -1.5470e-03, -1.2928e-03, -4.4316e-03,  1.4845e-02,  4.4909e-02,
         4.3874e-03,  2.1098e-03, -5.1369e-03, -4.4617e-03,  3.1935e-03,
        -1.2437e-03,  5.6257e-03,  3.0327e-03,  7.0352e-03,  4.7735e-03,
         1.5208e-03,  8.6655e-03,  4.2435e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+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([ 1.1272e-02,  9.7299e-01, -7.2262e-03, -1.9234e-03, -1.8348e-03,
        -2.3109e-03, -2.1063e-04,  3.4647e-04,  1.8897e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4502e-02, -9.5663e-01,  4.2320e-04, -3.4390e-03,  4.6099e-04,
        -1.7891e-03,  3.2905e-04,  8.2719e-04, -1.5751e-03, -1.5997e-03,
         4.1136e-03, -4.4371e-04,  1.0585e-03, -1.0603e-03, -1.0603e-03,
        -4.8391e-04,  3.7124e-03,  2.1147e-03, -3.6466e-04, -2.7994e-03,
        -1.2171e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1317e-02,  9.3765e-01,  2.9504e-03,  1.1152e-03,  5.9089e-05,
         3.7816e-03, -1.6479e-04,  2.5599e-03,  3.7118e-03,  4.1947e-03,
         2.3875e-03,  8.8754e-04, -1.5506e-03, -8.0058e-04,  1.2617e-03,
        -1.8340e-03,  3.1635e-04,  1.2043e-03, -1.2993e-03, -9.5288e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3773e-02,  9.5706e-01, -3.5216e-03, -9.4500e-04,  1.4499e-03,
        -5.6121e-04, -5.5895e-04,  7.2143e-04,  2.2232e-03, -4.9704e-04,
         1.0944e-03, -3.1014e-03, -6.4489e-04,  5.4461e-04,  6.9288e-04,
         2.6081e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6029e-02,  9.0934e-01, -5.7347e-03, -5.9825e-03, -3.5619e-04,
        -5.1905e-03,  1.2883e-03, -8.7115e-04, -8.6881e-04, -3.1976e-03,
         4.1197e-03, -2.7612e-03, -3.3025e-03, -3.2631e-03, -4.4130e-04,
        -1.2080e-03, -4.0531e-03, -1.9951e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3829e-02,  9.6297e-01,  1.0907e-03,  7.1582e-04, -1.9164e-04,
         1.1800e-03,  1.3602e-03,  1.1184e-03,  3.0578e-03,  1.6634e-04,
         3.1686e-03,  1.9260e-03, -8.9693e-04,  1.8917e-03,  4.5752e-04,
         1.2998e-03,  1.7175e-03,  1.3226e-03, -5.1421e-04,  1.1229e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1896e-02, -9.5003e-01,  1.2590e-03, -2.1102e-03, -1.3632e-03,
        -2.0106e-03, -2.4890e-04,  2.2595e-03, -1.4932e-03, -1.8462e-04,
         2.2642e-03,  1.7071e-04,  1.9201e-03, -2.7118e-03, -6.1691e-04,
        -1.4711e-03, -4.0320e-04,  2.5684e-03, -2.2851e-04, -7.1481e-04,
         1.7547e-03, -7.2224e-04,  1.6000e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1263e-02, -9.3248e-01,  8.1681e-04, -1.3127e-03,  7.9024e-04,
         1.8952e-03,  3.6054e-03,  1.6698e-03,  2.4381e-03,  1.4423e-03,
        -1.5561e-03,  2.3092e-03,  1.1012e-03,  9.2069e-04,  5.1002e-04,
        -8.8476e-04,  3.3459e-03,  1.2995e-03,  7.1614e-03,  3.8895e-03,
         1.8163e-03,  3.0505e-03,  2.0844e-03,  8.2868e-04,  3.1010e-03,
         3.1747e-03,  1.4316e-03,  1.8496e-03,  7.0086e-04,  4.2772e-04,
         8.4674e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.8300e-03, -9.6131e-01,  1.8322e-03,  2.4817e-03, -2.2050e-04,
         6.0015e-04, -2.0402e-04,  1.4300e-03,  1.2918e-03, -2.4633e-04,
         2.5249e-03,  4.1126e-03,  2.1902e-03,  2.5160e-03, -3.8149e-05,
        -7.6620e-04,  9.0246e-04,  1.5356e-03,  9.9975e-04, -1.0624e-03,
        -2.0580e-04,  3.0022e-03, -1.0114e-03, -1.3261e-03,  6.9752e-04,
         2.2937e-03, -4.1740e-05, -3.2698e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5485e-02,  9.6250e-01,  2.9682e-05,  5.8762e-03,  3.4254e-03,
        -3.0011e-03, -2.3375e-03,  1.0293e-04, -5.0258e-03,  1.5596e-03,
         6.5549e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0025e-02, -9.4954e-01,  6.6540e-04, -2.1546e-03, -3.1292e-03,
         3.0398e-03, -8.3222e-04,  8.6711e-04,  3.2784e-03,  9.8813e-04,
        -3.5003e-03,  6.5550e-04, -2.9112e-03, -1.6066e-03, -1.3615e-03,
        -9.3961e-04, -3.2131e-03,  1.1943e-04, -1.0867e-03, -8.1840e-04,
        -1.2991e-03, -3.9110e-04,  1.8990e-03, -2.4399e-04, -1.3311e-03,
        -3.7073e-04,  4.1676e-04,  5.3372e-04, -2.7859e-03,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.8892e-03, -9.5805e-01, -2.7194e-03, -5.6285e-03, -1.5993e-03,
        -3.1354e-03,  2.8229e-04,  1.2020e-03, -1.1318e-03,  3.6187e-04,
        -3.7205e-03,  2.6603e-03, -1.6353e-03, -3.7058e-04,  3.9201e-04,
         1.5447e-03, -8.7884e-05, -3.3727e-03,  2.1786e-03,  9.5858e-05,
        -9.4793e-04, -1.9933e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([-3.7970e-02,  9.4266e-01,  1.0106e-03,  7.8341e-05, -3.6346e-03,
        -3.4742e-03, -6.4442e-04, -2.3722e-03, -6.6202e-05, -2.4300e-03,
        -3.9886e-04, -4.4389e-03, -1.1681e-04, -7.0961e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.4729e-03, -9.6809e-01,  1.0620e-03,  1.2445e-03,  5.6438e-04,
        -9.2806e-04,  2.6791e-04,  7.9144e-04,  1.2243e-03,  1.8702e-03,
        -1.5372e-03,  1.2422e-04,  4.9833e-04,  8.9856e-04,  1.3627e-03,
         2.3776e-03,  2.2699e-03,  8.1450e-05,  2.1522e-03,  7.5066e-04,
         1.0108e-03, -2.1034e-03,  3.5378e-04,  9.9564e-05, -8.6639e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6272e-02,  9.5378e-01, -1.6927e-03,  6.0785e-05,  2.3827e-03,
        -2.5625e-03,  2.9226e-04,  9.7774e-04,  9.5685e-04, -2.2384e-03,
        -1.0151e-03,  7.7137e-04,  4.7507e-04, -1.5122e-04, -5.0759e-04,
        -7.1261e-04, -1.9684e-03,  9.7801e-04, -1.0500e-04, -3.1191e-03,
        -1.5135e-03, -1.4287e-03,  1.2409e-03, -5.8746e-04, -8.6252e-04,
        -1.6083e-03, -7.1443e-04, -4.8556e-04, -5.3590e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3491e-02,  9.7201e-01, -1.2003e-03,  2.0052e-04, -8.0880e-04,
         4.2414e-03, -2.8609e-04, -7.7412e-04, -8.3021e-04,  3.8324e-03,
         2.9440e-04,  2.0296e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3443e-02,  9.2788e-01,  1.7213e-03,  7.0284e-04,  1.6152e-03,
         2.1226e-03,  3.0903e-04,  4.4375e-03,  2.5548e-03,  4.8109e-03,
         2.3871e-03,  9.4381e-04,  8.8232e-04,  1.4649e-03, -2.4976e-03,
         2.2252e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2094e-02,  9.5065e-01, -1.6797e-03,  1.2005e-03,  2.0439e-03,
        -3.7756e-04, -3.7940e-04,  5.6904e-04,  1.6138e-03,  7.1044e-04,
         1.3903e-03,  5.6516e-04,  1.2098e-03, -3.8589e-04, -1.3434e-03,
         4.9151e-05, -2.0463e-04,  1.2528e-03, -1.7065e-03,  5.7843e-04,
        -8.7677e-04, -1.6029e-03,  2.1523e-03, -2.3230e-03,  7.2607e-04,
         9.2758e-04,  1.3882e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1048e-02, -9.5324e-01,  1.5117e-03,  1.7140e-03, -4.6904e-04,
        -4.1001e-04,  3.2068e-04,  1.0269e-04, -3.1367e-04,  5.3027e-05,
        -9.8680e-04, -1.5597e-03,  1.6201e-03, -3.1548e-05,  4.6089e-04,
         2.4119e-03,  8.6964e-04, -4.5023e-04,  2.6708e-03,  1.1047e-03,
         3.4558e-03,  5.8411e-04, -1.7683e-03, -2.8451e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8182e-02, -9.5473e-01,  1.2194e-03,  8.2988e-04, -1.1119e-03,
        -1.7876e-03,  1.0766e-03,  7.6057e-04, -1.2838e-03, -5.5753e-04,
         1.1058e-04, -6.5186e-04, -1.6508e-04,  3.5203e-04,  5.9989e-04,
         1.0233e-03, -3.4259e-04,  2.0627e-03, -3.6639e-04, -1.5550e-03,
         1.1559e-03, -1.5820e-03,  3.5037e-05,  1.7733e-06,  8.9142e-04,
         1.3235e-03, -8.4836e-04,  5.2818e-04,  1.3115e-03,  1.0456e-03,
         1.1922e-03,  5.3021e-04, -4.2329e-04, -3.6532e-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([-1.7553e-02, -9.5813e-01,  1.0375e-03, -5.4080e-05,  7.3238e-04,
         4.1048e-04,  1.3199e-03,  5.1673e-04, -4.6292e-04,  2.2831e-03,
         6.9141e-04, -1.6986e-04,  5.6695e-04,  7.3407e-04,  4.5501e-04,
         8.1763e-04,  1.7622e-03,  3.8210e-04,  3.9870e-04,  9.1021e-04,
         1.4820e-03, -3.7288e-04,  7.0420e-04,  5.6170e-04, -1.0603e-03,
         1.8610e-03,  2.5029e-03, -2.6247e-05,  5.1078e-04,  1.5313e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4210e-02,  9.4062e-01, -6.6472e-04, -2.6045e-03, -1.3471e-03,
        -1.6634e-03, -3.3867e-04, -3.2370e-03,  5.6369e-04, -1.5896e-03,
        -1.1409e-03, -1.2126e-03, -1.2884e-03, -2.6659e-03, -1.8520e-03,
        -2.4503e-03, -3.8581e-03, -4.1019e-03, -1.0048e-03, -6.2600e-04,
        -1.4706e-03,  2.3782e-04, -1.2487e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.8041e-03, -9.5752e-01, -1.8665e-03,  1.6603e-03,  6.4469e-05,
         1.1613e-03,  4.3585e-04, -1.9971e-04,  1.8718e-03,  2.2499e-03,
         8.4892e-04,  1.2120e-03,  1.7915e-04,  4.0948e-04,  3.3556e-04,
         3.6590e-04, -3.7239e-04, -5.3466e-04,  1.1246e-03,  5.6825e-04,
        -1.2923e-03,  3.7417e-04,  7.3063e-04,  4.2063e-03,  3.6790e-03,
         1.5590e-03,  8.2444e-04,  4.7442e-03,  1.1243e-03,  2.2025e-04,
         4.5980e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2366e-02, -9.6690e-01,  6.7142e-04,  3.9641e-04, -5.8611e-04,
        -1.2008e-03,  5.7038e-04,  3.1721e-04, -7.0690e-04, -4.7319e-04,
        -2.6186e-05,  1.5555e-03,  5.9475e-05, -6.9622e-04,  8.6606e-05,
        -4.5899e-04,  5.3761e-05, -5.3473e-04,  1.3999e-03,  5.9191e-05,
        -9.2889e-04, -6.4512e-05,  2.2903e-04,  3.8842e-04,  7.9111e-05,
        -4.4674e-04,  6.7419e-04,  2.9881e-04, -4.1449e-04,  3.6886e-04,
        -8.9579e-04,  1.2190e-03,  8.9114e-04, -2.6665e-05,  7.9620e-05,
        -7.8010e-04, -1.2078e-03, -3.3255e-04, -4.2236e-04, -1.1364e-03],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-4.2876e-03,  9.7772e-01, -4.5474e-04,  1.4328e-03,  1.4692e-03,
        -1.2111e-03,  4.4317e-04,  9.9120e-04, -2.0805e-03, -4.9895e-04,
         4.7043e-04, -1.3885e-03, -1.7178e-04,  1.2301e-03,  1.8780e-04,
         1.7437e-03, -9.0379e-04, -1.4683e-03, -6.7147e-04, -1.1754e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6374e-02,  9.6589e-01, -2.3654e-03,  7.6483e-04, -2.6018e-03,
        -4.7956e-03, -3.6913e-03, -3.5140e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.7356e-03,  9.8198e-01,  1.7920e-03,  2.7112e-05,  3.8408e-04,
         1.9895e-03, -1.3622e-03, -1.5003e-03,  2.3334e-03, -8.9388e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6320e-03, -9.7641e-01, -1.5390e-03, -2.8370e-04,  6.8947e-04,
        -3.7306e-04, -1.7720e-04, -1.2345e-03,  2.3806e-04, -1.0381e-03,
         4.4311e-04,  6.0098e-05, -6.6165e-04,  1.0956e-03, -8.0133e-04,
         5.5939e-04, -2.2007e-04, -1.4284e-03, -1.0229e-03, -9.4312e-04,
        -8.0708e-04,  2.9368e-04,  1.4003e-03,  6.6221e-04,  6.1302e-04,
         9.8255e-04,  1.8620e-04,  3.1587e-05,  7.0570e-04, -1.4714e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4718e-04,  9.8271e-01,  7.0222e-04,  3.6594e-04,  5.7366e-04,
         1.1104e-03,  2.9901e-04,  8.2524e-04, -1.4946e-03,  1.9813e-06,
         1.2985e-03, -4.0295e-04,  8.7220e-04,  1.1497e-03,  5.3180e-04,
         2.8196e-04, -8.4796e-04, -3.5839e-05,  1.2408e-03, -6.6929e-05,
        -1.5600e-04,  2.1139e-03,  2.9855e-04,  2.0676e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.8444e-03,  9.7191e-01, -1.5573e-04,  7.8271e-04, -5.6871e-04,
        -2.9374e-04, -1.0841e-03, -1.8382e-03,  9.1969e-04, -1.4519e-04,
        -8.0411e-04, -2.6355e-04, -3.0046e-03,  6.8617e-04, -7.7542e-04,
         4.6564e-04, -7.1393e-04, -3.6841e-04, -2.0201e-03,  5.5278e-04,
        -2.3896e-04, -8.7928e-04, -1.7185e-03, -6.9026e-04, -1.2767e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8809e-02,  9.5769e-01, -6.7304e-04, -5.7855e-04,  8.5621e-04,
        -4.9041e-04, -8.5761e-04,  3.2545e-04, -2.4043e-04, -1.6325e-04,
         1.4583e-04, -1.2438e-03, -4.5625e-05, -3.9313e-04,  9.1942e-04,
        -1.7108e-04, -1.2904e-03,  3.2549e-04, -1.0465e-03,  1.1716e-04,
        -3.3547e-04, -1.0943e-03,  3.0729e-04,  3.0343e-04, -3.1231e-05,
        -2.0536e-03, -7.7382e-04, -8.1140e-04, -3.9826e-04,  1.5848e-04,
        -1.7713e-03, -5.1350e-04, -5.0913e-04,  5.6582e-04, -1.1651e-03,
         5.2509e-04, -1.6980e-03, -2.3371e-04, -3.6660e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5736e-02,  9.7653e-01,  5.4328e-05, -5.9480e-04, -1.9647e-03,
        -2.7442e-04, -8.1518e-04,  2.2084e-03, -2.6473e-04, -3.9104e-04,
        -4.4802e-05,  1.1207e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.4874e-03,  9.8411e-01,  4.9669e-04,  9.2193e-06, -1.1864e-03,
        -4.8234e-05, -1.9684e-03, -4.7862e-03, -2.3563e-03, -3.7424e-04,
         1.8173e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.9688e-03,  9.6356e-01, -1.0537e-03,  1.9586e-04,  1.8705e-03,
        -1.4604e-03, -8.7963e-04, -1.0529e-03,  1.1105e-03,  3.4448e-04,
         1.2606e-03, -3.1174e-03, -1.6013e-03, -2.0049e-03, -1.7293e-03,
        -2.7812e-03, -2.0244e-04,  2.1524e-04, -2.8141e-04,  7.8729e-04,
        -2.0335e-03, -1.4061e-03, -4.5875e-04, -2.4884e-04,  3.7146e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7493e-03, -9.7683e-01,  1.0811e-03, -1.9266e-03, -2.4704e-04,
        -3.7985e-04, -2.9604e-04, -1.7779e-03,  2.1330e-03, -1.0822e-03,
         5.4501e-04, -5.9620e-04,  1.2467e-03, -1.1220e-03,  1.0335e-03,
        -4.3863e-04,  2.1807e-04,  1.4630e-04,  2.5697e-03,  1.2786e-03,
         1.1264e-03, -1.1771e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1504e-02,  9.6446e-01,  1.7969e-03,  1.7704e-03,  1.4182e-03,
         1.7790e-04,  1.1693e-03,  8.8959e-04, -3.0987e-05,  4.9741e-04,
         1.8787e-03, -1.6299e-04,  2.5780e-04,  1.1931e-03, -2.8182e-04,
        -3.6125e-04,  1.5948e-03, -8.4216e-05,  8.1848e-04,  9.1643e-04,
         2.6790e-03, -3.1075e-04, -1.3747e-05,  4.6465e-04,  4.8935e-04,
         1.7544e-04,  1.6655e-03,  7.3782e-04,  1.2362e-03, -9.9490e-05,
        -8.6399e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([ 3.1404e-03, -9.7878e-01, -2.1426e-04, -1.9773e-04, -3.5637e-04,
         2.2191e-04, -3.5911e-04,  2.1864e-04, -2.4631e-04, -3.0091e-04,
        -1.1823e-04, -4.4591e-04, -3.7298e-04,  4.5661e-05,  2.3496e-04,
        -4.6921e-04, -3.2571e-04,  8.0487e-04,  4.4743e-04,  7.7766e-04,
         7.5281e-06,  4.6114e-04, -3.7814e-05,  3.5039e-04,  4.7159e-04,
         8.0501e-04, -2.3141e-05, -1.7214e-04,  2.9044e-04, -4.2161e-04,
         3.9389e-04, -4.4616e-04,  2.7615e-04,  8.3520e-04, -1.5605e-05,
         1.6758e-04, -2.9468e-04, -3.3402e-04,  1.8737e-04, -1.0553e-04,
         1.1991e-04,  1.2924e-05, -4.9304e-04, -2.0324e-04, -1.7128e-04,
         4.0468e-04,  1.9663e-04,  1.7574e-04, -3.9221e-04, -6.3710e-04,
         6.0119e-04, -2.9201e-04, -6.8198e-04,  6.2201e-05,  1.9587e-04,
        -4.1993e-04, -7.6748e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8924e-02,  9.5065e-01,  1.4591e-03,  4.6938e-03,  3.2256e-03,
         1.4119e-03,  3.0534e-03,  5.3556e-03, -5.3385e-05, -1.1693e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.6647e-03,  9.6142e-01,  1.3100e-04,  1.8457e-03, -9.5438e-05,
        -2.1513e-03,  5.2769e-04,  6.9569e-04,  2.5019e-03,  3.6483e-03,
         1.1179e-03,  4.8204e-04,  1.8902e-03,  2.4287e-05,  6.4039e-04,
         9.6827e-04,  8.7640e-04,  3.9196e-04,  2.4797e-03,  3.1851e-03,
         1.1350e-03,  1.2358e-03,  1.8274e-03,  3.0838e-03, -1.1157e-04,
         8.6657e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4661e-04,  9.7709e-01, -1.6861e-04, -2.3285e-03, -2.6033e-03,
        -4.0151e-03, -1.2299e-03, -2.5826e-03, -4.5761e-03, -4.6243e-04,
        -4.3345e-03,  5.8502e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.2986e-03,  9.7784e-01,  5.7806e-04, -9.7026e-04,  1.0326e-03,
        -1.5656e-03,  7.7689e-04, -8.4548e-04,  3.0801e-03,  1.8737e-04,
        -1.2595e-03,  1.4398e-03, -2.1995e-03,  4.3307e-05,  6.7459e-04,
         1.9162e-03,  2.5134e-04,  3.5004e-05, -1.0024e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3485e-03,  9.7535e-01,  2.8928e-04,  1.3761e-03, -7.5940e-05,
        -1.2951e-03,  1.2855e-03, -1.2811e-03, -3.8602e-04,  7.9678e-04,
         2.2216e-03, -1.8976e-04, -2.9878e-04,  1.1280e-03, -9.5398e-04,
         2.2583e-03, -6.3927e-04,  7.5755e-07, -8.1902e-04, -4.5752e-04,
        -5.2720e-04, -1.9471e-04, -1.8381e-03,  4.6566e-04, -5.2781e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.0778e-03,  9.7054e-01,  6.4056e-04,  5.5875e-04, -2.3852e-03,
        -1.7683e-04,  1.0683e-03, -7.1582e-04, -2.2512e-03,  2.9660e-04,
         9.9715e-04, -3.0149e-03, -3.6000e-03, -8.1087e-04,  6.7338e-04,
        -1.5840e-03, -1.2251e-03, -1.4813e-03, -9.0327e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5129e-02, -9.6269e-01, -1.1378e-03, -3.6087e-04, -1.2362e-04,
         9.1159e-04, -4.7844e-04,  7.4372e-05, -5.4538e-04, -9.6564e-04,
         1.0510e-03, -7.5798e-04,  6.1920e-04, -1.9887e-04,  9.8449e-04,
         8.8710e-04, -3.2988e-05,  6.9024e-04,  8.7188e-04, -4.6399e-04,
        -4.7906e-04, -5.1061e-04,  1.1837e-03,  1.2165e-03,  3.8838e-04,
         4.5191e-04, -5.2068e-04,  7.0714e-05, -1.5422e-04,  1.1462e-03,
         1.5905e-03, -1.5394e-05,  5.8040e-04,  9.4864e-04,  1.0234e-03,
         1.1343e-04,  3.1824e-04, -8.0676e-05,  2.2845e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6939e-03, -9.8707e-01, -6.3464e-04,  1.0298e-03,  2.4735e-04,
         2.6552e-03,  2.5194e-03, -1.1447e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0633e-03, -9.8488e-01,  5.2061e-04,  2.0188e-03,  1.3853e-03,
         1.2379e-03, -1.3467e-03,  1.6513e-03,  4.6532e-04,  1.5751e-03,
        -1.6387e-03,  1.3552e-03,  8.5759e-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,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0751e-02, -9.6707e-01,  3.7920e-05,  1.6648e-03,  1.6055e-04,
         5.7699e-04, -1.4820e-03, -6.2569e-04, -3.7035e-05,  1.8384e-04,
        -8.8916e-05,  9.1092e-04, -3.1634e-04, -1.3066e-04,  4.5552e-05,
         5.9149e-04,  1.1255e-03, -8.2938e-04,  1.4410e-03,  1.8023e-03,
        -1.4478e-03,  2.3248e-03, -1.2769e-03,  4.6887e-04,  1.3362e-03,
         5.1420e-05,  6.2500e-04,  2.0344e-03, -4.1615e-04, -1.5167e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2425e-03,  9.7948e-01, -3.6261e-04, -6.4894e-05, -3.1841e-04,
         2.2855e-04,  1.2825e-03,  2.0089e-03, -2.0125e-03, -2.4431e-04,
         6.3749e-04,  8.3147e-04, -3.2259e-04, -1.3887e-04,  9.1743e-04,
         8.4786e-04,  8.8036e-04, -1.6554e-03,  6.8937e-04, -2.4354e-04,
         1.0027e-03, -1.4303e-03,  1.9809e-04,  6.5043e-04,  1.3840e-04,
         1.5171e-04,  2.1475e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+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([-3.8349e-03, -9.8402e-01, -2.7289e-05,  6.6238e-04,  8.9612e-04,
         4.2388e-04, -3.1528e-04,  4.2662e-04, -1.0798e-03,  2.3381e-04,
         6.4076e-04, -4.0821e-04, -1.0088e-04, -9.4958e-04,  2.9138e-04,
        -1.0595e-04,  4.7779e-05,  1.3162e-03,  2.1496e-03, -3.5537e-04,
        -7.9519e-04,  1.9857e-04, -6.3009e-04, -9.2439e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6323e-04, -9.8575e-01,  2.1749e-04,  2.8386e-04, -1.7112e-04,
        -6.8721e-04, -8.0407e-04, -7.1870e-05, -5.1735e-04, -2.3007e-04,
        -4.3050e-04,  3.1542e-04, -1.8696e-04,  9.3646e-04,  6.0687e-04,
         7.8918e-04, -1.7192e-04, -1.5474e-03,  2.8421e-04,  7.2819e-05,
         1.0384e-03,  8.6505e-04,  6.2551e-04,  6.4481e-04,  7.0808e-04,
         7.5613e-04,  9.1572e-04,  5.3362e-06,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2337e-03, -9.8450e-01, -1.4864e-03, -1.2128e-03, -6.9202e-04,
        -1.9255e-03, -3.4001e-04, -8.4250e-04, -1.1544e-03, -6.6724e-04,
        -1.0228e-03, -1.2292e-03,  2.4390e-04,  8.1631e-05, -3.6373e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1672e-04,  9.7954e-01,  1.7131e-03, -2.4301e-04,  6.2312e-07,
        -2.5608e-04, -7.0054e-04,  3.5505e-04,  7.0651e-04, -5.2070e-04,
        -2.6628e-04, -1.0257e-03,  4.2208e-04,  8.6994e-04, -1.5248e-04,
         5.1476e-04,  7.4186e-04, -2.9101e-04, -4.1815e-04,  3.5651e-04,
        -2.6553e-04, -5.3733e-04, -1.5645e-04,  1.0020e-04,  4.8762e-04,
         1.0786e-03,  4.0820e-04, -1.4145e-03,  1.4299e-03,  3.2351e-04,
         5.2334e-04, -1.5373e-04,  1.1078e-03,  7.7868e-04,  3.0581e-04,
        -7.8252e-04,  8.3983e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7034e-04,  9.8230e-01, -4.0977e-04, -8.0218e-04, -5.7258e-04,
         1.9279e-03, -6.1264e-04, -1.0305e-03, -4.4097e-04, -1.6341e-03,
        -1.5397e-03, -8.3106e-04, -1.1308e-04, -1.0248e-03, -1.3802e-03,
         7.1574e-04, -3.0264e-04, -1.0621e-03, -1.3186e-03,  1.7025e-04,
         9.1771e-04,  6.2766e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.5912e-04,  9.8555e-01,  1.2301e-03,  2.5923e-05, -5.4655e-04,
         2.1241e-03, -6.2887e-04,  5.4892e-04, -4.2917e-03,  4.8834e-04,
        -2.3675e-03,  7.3852e-04, -2.7811e-04, -7.2448e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8051e-03,  9.8411e-01,  1.9018e-03, -8.2472e-04,  2.9790e-04,
        -2.9898e-03,  9.0178e-04,  1.2611e-03,  1.2454e-03,  1.7039e-03,
         2.9561e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3942e-02,  9.7548e-01,  8.8023e-04,  1.3503e-03, -1.2038e-04,
         2.4110e-03,  9.3924e-04,  1.3380e-03, -5.3576e-04,  1.5463e-03,
         1.4550e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3274e-03,  9.8770e-01, -7.9411e-04,  9.7479e-04,  4.8911e-04,
         1.4002e-03,  1.1602e-03,  1.5067e-04, -7.8016e-04,  1.4476e-04,
         3.4538e-04, -1.6006e-05, -2.3206e-03,  3.1982e-04, -5.8480e-04,
         4.1100e-04,  2.1252e-04, -3.6146e-04, -2.5057e-04,  2.5231e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.6132e-03, -9.8314e-01, -7.0621e-04,  1.0268e-03, -1.2864e-04,
         4.6053e-04,  6.0002e-04, -3.7702e-04, -1.8308e-04,  2.9218e-04,
         3.6744e-04,  9.4021e-04,  2.8135e-04,  1.7965e-03,  3.8619e-04,
        -4.3276e-05,  1.0442e-03,  4.5537e-04,  1.3085e-03, -5.0224e-05,
         3.4048e-04,  2.7371e-05,  4.2852e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8935e-03, -9.8804e-01, -6.9454e-05,  1.6281e-03,  1.2092e-04,
         2.5888e-04, -1.0230e-03,  3.1280e-04,  2.1434e-04,  9.3768e-06,
         4.2328e-05,  6.3016e-04,  1.4085e-05,  1.6467e-03,  1.4318e-04,
         1.6287e-05,  6.2050e-05, -6.6347e-04,  2.5889e-04,  2.8716e-04,
         1.6970e-04,  2.5010e-04, -3.9497e-04,  1.3695e-04, -4.6532e-04,
         8.3489e-05,  9.5348e-04,  1.1744e-04,  9.3859e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4793e-02, -9.6611e-01, -2.4745e-04, -7.3524e-04, -8.6817e-04,
         8.4385e-04, -3.9144e-04,  1.2116e-04,  7.3448e-05,  7.4339e-04,
         1.8994e-05, -1.2449e-03,  3.2163e-04,  1.9895e-05,  7.4172e-05,
         1.1840e-06,  4.6652e-04, -1.2511e-03,  3.2284e-04,  7.5673e-04,
         2.6257e-04,  5.7921e-04,  7.0910e-04,  1.0418e-03,  2.1638e-04,
        -2.9573e-04, -5.2671e-04,  4.6448e-04,  2.1527e-04,  1.0504e-04,
         1.2241e-03,  5.0052e-04, -6.3007e-04,  7.5713e-04,  8.9971e-04,
         8.6879e-04,  6.9109e-04,  6.0216e-04], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 7.9847e-04,  9.9046e-01, -6.8981e-04,  1.8195e-04,  3.9067e-04,
         1.8629e-04,  5.8851e-04,  1.4702e-03, -1.0315e-03, -4.9083e-04,
         2.8012e-04,  2.8174e-04, -2.5327e-04, -6.7328e-04,  5.2785e-04,
         2.0315e-04,  5.2307e-04,  1.5934e-04,  8.0879e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.5388e-03,  9.7424e-01,  3.7875e-03,  1.1715e-03,  2.1125e-03,
         9.3059e-05, -2.1926e-04,  3.5380e-04, -3.1295e-04, -2.1219e-04,
         1.2305e-03,  1.2418e-03,  3.3067e-04,  1.7617e-03,  2.0164e-04,
         5.4599e-05,  8.4779e-04,  3.4763e-03, -1.1651e-03, -6.4609e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1175e-02,  9.6111e-01,  1.3600e-03, -2.0951e-03,  5.9822e-05,
         5.8151e-04, -7.2640e-04, -7.6285e-04,  1.6065e-03,  6.2388e-04,
        -1.2097e-05,  1.7162e-03, -3.6617e-04, -2.4771e-04, -2.0905e-03,
        -3.2088e-04,  7.6549e-04, -1.4067e-03, -1.6560e-03, -9.5732e-04,
        -3.5760e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.3743e-03,  9.7871e-01, -8.8068e-04, -6.4309e-04, -5.6462e-04,
        -5.5206e-04,  1.2078e-03, -1.0209e-03, -9.3146e-05,  2.6062e-04,
         3.0405e-04, -1.9547e-05, -2.5706e-03,  7.6490e-04, -9.2414e-05,
        -1.9313e-03, -8.5920e-04,  2.0721e-03,  7.6146e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7512e-03,  9.7808e-01, -5.9255e-04, -8.4675e-05, -3.4225e-04,
         3.8025e-04, -4.8277e-04,  1.4807e-04, -1.7390e-05,  4.3750e-05,
        -1.0742e-03, -4.8481e-04, -3.2461e-04,  4.2647e-04, -2.3135e-04,
        -1.5977e-03,  6.1906e-04,  3.7177e-04, -1.5717e-03, -1.7251e-04,
        -9.0660e-05,  1.2331e-04,  5.3422e-04,  5.0610e-04, -4.5318e-04,
         2.7808e-04, -1.2578e-04, -1.0701e-03, -1.1822e-03, -1.1285e-04,
         2.2374e-04, -5.9212e-05, -6.7372e-04,  7.2084e-05,  5.4302e-04,
        -5.6586e-04,  1.6517e-04, -7.3528e-04, -2.5025e-04, -2.5350e-04,
        -5.4219e-04, -5.0778e-04, -1.2778e-04, -4.1131e-04,  3.1051e-04,
        -2.8177e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.0994e-03,  9.8132e-01, -3.5495e-04,  3.6650e-04, -5.5818e-04,
         5.6627e-04, -7.6180e-04, -4.4549e-04,  7.8637e-04, -1.3989e-04,
         6.3026e-04,  1.0226e-03, -3.6382e-04, -1.5418e-03,  4.6985e-04,
         6.3847e-04,  4.7496e-06, -9.1689e-04, -6.4957e-04,  3.6127e-04,
        -3.3719e-04,  6.6447e-05, -9.3802e-04,  1.5028e-03,  4.0377e-05,
        -8.8943e-04, -6.1374e-05,  1.6273e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5279e-04, -9.8085e-01, -8.4643e-04, -7.3906e-05,  1.9159e-04,
         1.8552e-04, -1.2585e-04,  9.9867e-04,  3.4548e-04,  9.2847e-04,
         7.0587e-04,  9.7478e-04, -3.9566e-04,  1.3400e-03,  1.0818e-03,
         3.5786e-04,  1.0713e-03,  3.9704e-04,  6.8997e-04,  1.1735e-03,
         1.6027e-04,  1.1859e-03,  3.3332e-04,  4.7332e-04,  1.3451e-03,
         6.5541e-04,  1.7637e-03, -3.0370e-04, -6.9306e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5403e-03,  9.8470e-01, -4.9516e-04,  1.1340e-04, -3.8107e-04,
        -5.5452e-04,  1.2322e-03,  7.7423e-04,  9.9628e-04,  3.4013e-04,
        -7.2517e-04, -2.5354e-04,  2.5718e-04, -2.2550e-04, -1.0699e-04,
        -3.6681e-04, -1.2281e-04,  1.0085e-04,  7.2308e-05, -7.6459e-04,
         9.9125e-04,  3.6201e-04,  5.1573e-04, -8.3863e-04, -4.3139e-04,
         1.3182e-03, -2.1378e-04,  2.0194e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.9707e-03, -9.7882e-01, -3.3401e-04,  5.2917e-04, -1.7733e-04,
        -1.6752e-04,  2.9995e-04, -1.9514e-05,  1.3754e-04,  1.6211e-04,
        -5.1854e-04, -2.4618e-04,  2.5941e-04,  3.1285e-04,  6.2512e-08,
         9.7283e-05, -9.3848e-05,  4.8571e-05,  3.9749e-04, -8.2604e-05,
        -5.6383e-04, -4.0079e-04,  2.2713e-04,  2.4246e-04,  1.6369e-04,
         3.6117e-04,  6.3577e-04, -4.0743e-05,  4.9994e-05, -4.8243e-04,
         2.4726e-04,  1.5033e-04, -1.3481e-04, -4.6942e-04, -2.6616e-04,
         2.6157e-04,  5.3188e-06,  8.1063e-05, -1.6466e-04,  1.3646e-04,
         3.6512e-04, -2.3103e-04,  5.8590e-06,  4.5522e-04, -6.3443e-04,
        -3.6995e-04, -1.7847e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5830e-03, -9.7793e-01, -7.0645e-04, -4.9636e-04, -1.6703e-04,
        -4.0020e-04, -3.9103e-04, -3.9694e-04, -3.9825e-04, -3.0873e-04,
        -5.2450e-04, -7.1027e-04, -2.4185e-04, -1.6634e-05, -2.2142e-04,
         7.6508e-05, -4.2671e-04, -2.8857e-04, -1.4156e-04, -8.2603e-04,
         1.6786e-05, -3.2159e-04, -1.0418e-04,  9.4069e-05, -9.5095e-04,
        -4.7315e-04, -6.1945e-04, -8.7544e-05, -9.5556e-04, -1.3219e-04,
         1.3479e-05, -4.5574e-04, -2.7356e-05, -5.1512e-04,  3.0297e-04,
        -1.2966e-04,  4.9156e-04,  2.3256e-04, -2.4329e-04, -5.0260e-05,
        -9.5406e-04, -1.0933e-03,  2.3454e-04, -8.8521e-04, -1.0269e-03,
        -3.3438e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5080e-03, -9.8533e-01,  4.2641e-04, -8.9043e-05, -6.4258e-05,
        -1.6500e-04,  5.4770e-04, -6.0963e-04, -2.6227e-04,  7.8196e-05,
         1.3080e-03,  3.1679e-05, -5.2788e-04,  8.7931e-04,  1.4072e-04,
        -3.1518e-04, -3.6420e-04,  2.5787e-04, -1.0342e-03, -1.9329e-03,
        -4.3599e-04, -6.9023e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.5195e-03,  9.7636e-01,  3.6104e-04,  4.8611e-05, -1.2563e-04,
        -8.1587e-04,  9.2646e-04, -8.5599e-04, -3.5214e-04, -1.5178e-04,
        -8.5886e-04, -1.5469e-04, -9.0079e-04, -3.0459e-04,  4.5011e-04,
        -7.6848e-04, -4.5057e-04,  1.4341e-05, -2.7287e-04,  3.6513e-04,
        -7.2701e-04,  3.0594e-04,  2.1615e-04,  7.7132e-04, -7.7374e-04,
        -2.9578e-04, -9.1308e-04, -7.6066e-04,  1.2862e-03,  8.9357e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([-7.6108e-03, -9.7858e-01,  1.2884e-03, -1.1266e-03, -1.2264e-04,
        -9.9487e-04,  5.5643e-04,  7.2480e-04,  5.2144e-04,  1.6890e-03,
        -1.5021e-03,  4.3315e-04,  1.9408e-03, -5.4683e-04, -1.2124e-03,
         8.8897e-04, -2.6396e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8817e-03, -9.7364e-01,  1.6760e-04,  1.5358e-03, -2.8541e-04,
        -7.1926e-05,  1.0942e-03,  2.6746e-03,  2.9257e-03,  5.2687e-04,
         1.1328e-03,  1.3267e-03,  1.6264e-03,  1.7181e-03,  7.0585e-04,
         1.1626e-03, -5.0045e-04,  1.1332e-03,  2.9228e-03,  9.6314e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5971e-03,  9.8431e-01, -9.0704e-04, -5.5513e-04,  8.6810e-04,
         1.2693e-03, -7.2721e-04,  8.2787e-04,  9.1024e-04,  1.8473e-03,
         1.9999e-03,  2.4837e-03, -6.9735e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1498e-03,  9.8310e-01, -1.6792e-03,  2.7157e-04,  1.2262e-03,
        -1.3634e-03,  6.7788e-06,  1.5210e-03,  3.0592e-04, -1.1905e-03,
        -2.9289e-04, -8.7487e-04, -1.2504e-03, -2.9308e-04,  3.1069e-04,
        -1.7155e-03,  1.8933e-04,  2.2542e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.0789e-04,  9.8956e-01, -1.1505e-05, -1.7416e-05, -1.0266e-03,
         3.1511e-04, -5.0361e-05, -3.1053e-04,  9.7759e-04,  4.9492e-04,
        -2.3892e-04,  2.1504e-04, -6.2673e-04,  7.5418e-04, -7.3488e-05,
         6.1323e-04, -1.1939e-03,  4.4520e-04,  2.0457e-04,  1.2894e-04,
        -5.3321e-04, -6.0608e-04,  1.4000e-04,  2.7238e-04, -1.4149e-04,
        -4.3765e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8226e-03,  9.8487e-01, -3.6021e-04,  2.4294e-04,  7.6114e-04,
         2.4419e-03,  8.3438e-04, -2.6350e-03,  1.1803e-03,  1.3337e-03,
         3.5196e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7142e-04,  9.9004e-01, -4.3619e-03,  6.3342e-04, -1.6378e-03,
        -5.9891e-04, -5.6954e-04, -8.8118e-06, -2.8254e-04, -1.3978e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1356e-02, -9.6836e-01,  2.9774e-04, -3.6681e-04, -6.4548e-04,
         3.5823e-04,  2.3805e-04,  1.7059e-04,  3.5473e-04, -1.7720e-04,
        -5.2170e-04,  8.5907e-05, -7.2719e-05, -5.5496e-06, -7.6821e-04,
         1.2593e-04, -5.3389e-04, -5.4197e-05, -3.3408e-05,  1.0755e-03,
         1.1187e-04, -2.5456e-05, -5.1922e-04,  3.8109e-05, -4.5936e-05,
         3.7482e-05,  1.3951e-03,  5.8418e-04,  2.1604e-04, -1.3138e-05,
        -4.5821e-04,  3.8842e-04,  5.5282e-04, -8.2918e-04,  6.6429e-04,
         1.0622e-03,  6.7574e-04, -2.2113e-04,  4.7468e-04, -2.0533e-04,
         4.0478e-04, -4.6013e-04, -3.0651e-04,  2.1077e-04,  1.4807e-04,
        -1.4854e-05, -8.4607e-05, -2.7277e-04, -4.9147e-04, -2.6062e-04,
        -1.7348e-04,  6.6959e-05, -3.6891e-04,  1.3385e-03, -5.3108e-04,
        -7.4089e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0583e-03,  9.8562e-01, -8.3852e-05, -8.5453e-04, -8.3005e-04,
        -1.8042e-03, -1.0020e-04, -7.7308e-04,  9.9343e-04,  2.7517e-04,
        -4.6910e-04, -1.2648e-03, -2.9091e-04, -1.4198e-03,  8.5957e-04,
        -6.6328e-04,  4.3906e-04, -7.9510e-04, -7.4572e-04, -6.5637e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9564e-03,  9.8445e-01,  1.6657e-03, -7.1069e-05,  3.2442e-04,
         2.8734e-04,  6.6818e-05, -3.5436e-04,  5.0564e-04, -2.9103e-05,
         1.7884e-04, -9.1240e-04, -4.2859e-04,  1.7827e-04, -4.4068e-04,
        -1.8498e-04,  8.0372e-04,  8.3674e-04, -3.1669e-04,  5.3523e-04,
         1.4750e-04,  1.1490e-03,  1.7919e-04, -4.0227e-04, -2.1535e-04,
        -1.7813e-04, -7.7883e-04,  8.6345e-04, -3.3884e-04,  8.7095e-05,
         1.3509e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3808e-03,  9.8249e-01, -4.3043e-03, -6.5391e-04, -3.2617e-03,
         1.8664e-03, -2.0471e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2950e-02,  9.7184e-01, -2.2753e-03,  3.5807e-03, -2.7292e-03,
        -7.5309e-04, -1.9136e-03, -1.7522e-03, -2.2060e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+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([-9.6698e-03, -9.7887e-01, -1.1450e-03, -1.0215e-04, -8.8898e-04,
         1.5794e-04,  2.0183e-04, -1.7880e-03,  3.8070e-05, -2.6306e-04,
         2.9935e-04, -2.8108e-04, -8.2604e-04, -1.7801e-04, -1.4723e-04,
        -2.0306e-04,  7.7675e-06,  2.0830e-04,  1.2183e-03, -3.8595e-04,
         3.5545e-04,  1.9882e-04, -2.8382e-04, -1.1142e-04, -7.9467e-04,
        -1.0033e-03,  2.1639e-04,  1.5511e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.9894e-03, -9.7441e-01, -1.1261e-04,  5.6581e-04, -4.4006e-04,
        -8.1191e-04,  7.8011e-04,  4.2623e-04, -2.2861e-04, -2.0218e-05,
         7.1616e-04, -7.9872e-04,  1.6960e-04, -3.1914e-04, -5.8130e-04,
        -2.6920e-04,  5.9403e-04, -1.8274e-03,  5.1430e-04,  8.8125e-04,
        -5.2705e-04, -5.9433e-04,  1.0575e-04, -4.0475e-04, -1.1807e-03,
         2.0620e-04,  1.0820e-03, -1.0584e-03, -6.7246e-04, -2.8908e-04,
         1.1322e-03,  1.2896e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6465e-04,  9.8957e-01,  2.2468e-04, -1.8783e-04, -1.2436e-04,
        -6.2161e-04, -5.8292e-04, -3.3749e-04,  7.1832e-04, -1.3473e-03,
        -1.3501e-04,  9.5129e-05,  9.8730e-04, -5.1928e-04, -3.6660e-04,
        -3.7225e-04, -8.7027e-04, -2.4961e-04,  2.6959e-04,  5.8703e-04,
        -3.5721e-04,  5.9475e-05, -4.6915e-04,  1.8700e-04,  3.9894e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5170e-03,  9.9038e-01,  1.4609e-03,  2.5600e-04, -9.6154e-08,
        -1.0406e-03,  1.9159e-04,  1.1574e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2792e-05,  9.9082e-01,  3.7084e-03, -1.1175e-03, -1.4710e-06,
         2.4647e-03, -1.8679e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6497e-03, -9.8630e-01,  8.4777e-04,  7.7186e-04, -3.7575e-04,
        -9.0015e-04, -4.9305e-04, -4.9116e-04, -6.8397e-04, -1.3361e-04,
        -7.7762e-05, -2.5349e-04, -1.1712e-03,  2.3797e-04, -2.0114e-03,
         2.7220e-04,  7.6515e-04, -6.3521e-04,  2.3747e-04,  6.9176e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4223e-03,  9.8724e-01,  7.8708e-04, -1.0674e-03, -1.2922e-03,
         7.9342e-05, -2.0880e-03,  1.2762e-03, -1.6564e-04,  2.0166e-04,
         4.2970e-04, -1.7371e-03, -1.0030e-03, -2.1029e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6289e-02,  9.4359e-01,  5.7168e-03,  7.0049e-03,  2.4029e-03,
         8.9301e-03,  4.7316e-03,  9.2413e-03,  6.2671e-04, -1.4655e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1902e-05,  9.8181e-01,  1.0640e-03,  2.6544e-04,  6.9556e-05,
        -6.2935e-04, -5.0116e-04,  8.6256e-04,  2.0960e-05,  3.8567e-04,
         1.0679e-03,  3.3951e-04,  1.7337e-03, -7.7501e-04,  1.1734e-03,
         2.2119e-03, -4.6859e-04,  6.6615e-04,  8.0443e-04, -5.2287e-04,
         1.1174e-06, -3.5271e-05,  7.4794e-05,  1.4216e-03,  2.2649e-04,
         3.0179e-04,  1.3310e-03,  1.1828e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5856e-03,  9.8817e-01, -1.9299e-04,  7.0937e-05,  9.2323e-04,
        -2.7891e-04, -1.2444e-03, -1.7702e-04, -8.9541e-04,  1.7111e-04,
        -7.7170e-04, -7.3837e-04,  1.7970e-04, -1.2501e-03,  2.7043e-05,
         9.3620e-04, -3.8732e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7518e-03,  9.8354e-01, -6.2604e-04, -3.4070e-04,  3.6156e-04,
         7.8745e-04, -6.1511e-06, -9.3998e-04, -4.9270e-04, -2.2676e-03,
        -3.9047e-04,  1.2612e-03, -4.6469e-04, -5.2547e-05, -1.2244e-03,
        -1.0620e-03,  3.7457e-05, -4.8431e-04, -9.0675e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6672e-03,  9.8784e-01,  1.6809e-03,  4.0022e-04, -3.4240e-04,
         9.5947e-04,  5.1002e-04,  7.0855e-04, -2.3424e-04, -5.5696e-04,
        -1.2448e-03, -5.6440e-04, -2.3126e-04, -2.9081e-04,  4.4163e-04,
         1.0572e-04, -6.2987e-04, -5.9378e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #550: [tensor([-2.8532e-04, -9.8406e-01, -1.4709e-03, -2.7418e-04, -6.4856e-04,
        -1.8872e-04, -7.9166e-04, -2.3306e-04, -8.8177e-04, -6.2184e-04,
        -1.3252e-03, -7.3910e-04, -6.0510e-04, -1.2102e-03,  9.4381e-04,
        -2.1505e-04,  5.6961e-05, -7.4477e-04, -4.5285e-04,  6.7812e-04,
        -1.3993e-03,  2.6654e-05, -1.5426e-04, -8.3667e-04,  4.1397e-04,
         2.4319e-04,  4.9937e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1112e-04,  9.8336e-01, -1.1969e-04, -9.3516e-04, -6.2223e-04,
        -4.3486e-04,  4.7833e-04,  2.0478e-04, -3.3207e-04, -1.0744e-04,
        -4.0144e-04, -3.9950e-05, -7.6714e-04, -3.0156e-04, -4.8738e-04,
        -2.9013e-04,  6.5529e-04, -9.7686e-05, -7.0298e-04,  1.0188e-04,
        -9.4941e-05,  5.8092e-05, -2.4866e-04, -5.9787e-04,  3.9016e-05,
        -2.9985e-04,  4.5453e-04, -6.1132e-04,  1.2639e-04, -2.1342e-04,
        -1.6402e-04, -3.4534e-04, -6.1570e-04, -6.3638e-04, -1.5136e-04,
        -9.2341e-05,  5.5755e-04, -5.8395e-04, -9.7144e-06,  4.1578e-04,
        -9.7446e-04, -3.5768e-04, -4.1982e-04, -7.1985e-04, -1.6009e-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([-2.1443e-03,  9.8428e-01, -6.5900e-04,  3.6507e-04, -2.8187e-04,
         7.0556e-06, -8.3975e-04,  4.1723e-04, -3.4440e-05, -5.3330e-04,
        -4.9226e-04,  3.8548e-04,  7.2260e-04, -1.4596e-03, -3.0333e-04,
         2.4465e-05, -3.3816e-04, -9.4651e-04, -5.7241e-05, -7.3596e-04,
        -4.2032e-04, -3.6434e-04, -9.1986e-04, -9.3445e-04, -8.5724e-04,
        -1.5313e-04, -3.2930e-04, -7.1836e-04, -2.7811e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.9399e-05, -9.9100e-01,  1.7946e-05,  1.2065e-03, -1.1316e-03,
         1.1730e-03,  1.0609e-03, -9.8539e-04, -5.6716e-04,  1.3336e-03,
         6.8899e-05,  3.8441e-05, -1.3140e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.7863e-03, -9.8255e-01, -4.8959e-05,  2.9159e-04, -3.1000e-04,
         1.4489e-04, -6.5344e-04,  4.1103e-04, -1.7830e-05,  1.5027e-04,
         1.5181e-05,  3.2167e-04,  1.0363e-05,  2.1178e-04,  3.1877e-04,
        -6.6415e-04,  1.4283e-03, -1.9658e-04,  3.8015e-04,  2.9162e-04,
         1.1284e-04,  3.0090e-04, -5.6968e-04, -3.9407e-04, -9.0097e-05,
         3.6733e-04,  5.9891e-04,  4.4941e-04, -2.7411e-04,  2.5498e-04,
         8.0204e-04, -4.1141e-04,  7.1519e-04, -5.9333e-04, -3.7561e-04,
         3.5111e-04, -1.3906e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7342e-03, -9.8365e-01, -1.5270e-04,  1.8276e-04,  4.1932e-04,
        -1.0131e-03,  1.7619e-04, -2.9133e-04, -2.3388e-04,  5.2826e-04,
        -1.5196e-04,  2.2090e-04,  4.0625e-05,  7.5256e-04,  6.1648e-04,
         1.1095e-03,  3.5692e-04,  5.7665e-04, -5.6556e-04,  4.7337e-04,
        -3.2412e-04, -1.6454e-04,  6.1203e-04,  1.0234e-03,  7.9765e-05,
         3.4438e-04,  1.1499e-03,  2.3340e-04, -5.9436e-05, -7.6906e-04,
         5.5317e-04,  5.4555e-04, -5.6034e-04,  2.9812e-04, -4.0132e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4511e-03,  9.7799e-01,  1.2351e-03,  3.2504e-04,  1.1196e-03,
         1.9877e-03,  9.0037e-05,  1.5001e-03,  1.0267e-03,  9.0360e-04,
        -7.7372e-04, -3.1848e-04,  2.4249e-04,  1.3409e-03,  2.4501e-04,
        -7.0689e-04, -3.2457e-04,  4.1431e-03, -3.5105e-04, -5.9232e-04,
         3.3725e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.3382e-03, -9.8305e-01, -8.1060e-04, -2.7152e-04, -1.8981e-04,
         2.2196e-04,  1.4304e-04, -1.2350e-04, -2.4088e-04,  1.2607e-04,
        -8.7724e-04, -8.2308e-04,  1.4426e-04, -9.3397e-04, -2.5858e-07,
        -1.0552e-03,  7.7713e-05, -7.8770e-04, -6.7969e-04, -1.5568e-04,
         1.1085e-03,  2.9146e-05,  2.9825e-04,  4.0172e-04, -2.5476e-05,
        -8.5305e-04,  3.8705e-05,  2.1273e-04, -3.4280e-04,  6.0897e-04,
        -2.7651e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3127e-04, -9.8651e-01, -3.1991e-04,  3.2405e-05,  3.8736e-04,
        -1.9428e-04, -2.9889e-04, -1.6397e-04, -6.4625e-04, -8.0222e-04,
        -1.2185e-04, -5.3141e-04, -5.3870e-04, -1.3649e-04, -5.8357e-04,
         8.6659e-04,  3.8216e-04,  2.0809e-04, -2.0783e-04, -1.5133e-04,
        -1.7478e-04, -1.9530e-04, -4.9594e-04,  2.2907e-04, -6.1908e-04,
        -5.0421e-05,  1.7958e-04, -3.3028e-04, -1.6978e-04,  3.0769e-04,
        -1.9119e-04, -3.4866e-04, -1.5205e-04, -1.3706e-04, -2.4616e-04,
        -3.4516e-04,  2.9476e-04, -5.6541e-04, -2.7230e-04, -4.5146e-04,
        -6.0945e-04, -1.1852e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6506e-04,  9.8964e-01,  7.0297e-04,  3.8587e-04,  5.8016e-04,
        -1.4500e-03,  1.4528e-04,  4.8606e-04, -2.4243e-04, -1.9427e-03,
        -1.0412e-03, -1.2454e-03, -1.8737e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.7300e-03, -9.7322e-01, -4.8286e-04,  1.9382e-04,  2.5021e-04,
         6.0550e-04,  1.3417e-05, -2.0002e-04, -1.6450e-04, -2.4599e-04,
        -3.0754e-04, -5.7899e-05, -1.1569e-04, -1.2830e-04,  3.1531e-04,
        -7.7243e-05,  4.6425e-05, -9.6498e-04,  3.4337e-04,  5.9987e-04,
        -6.2146e-04,  2.5428e-04, -8.2601e-04,  2.8776e-04, -2.1326e-04,
         6.8810e-04, -3.8187e-04, -1.6878e-04, -4.3493e-04,  3.5977e-05,
        -2.3793e-04, -2.8534e-04, -8.0664e-04,  1.6038e-04, -1.3440e-04,
         9.1285e-04, -2.2299e-04,  3.3138e-04,  6.6591e-05, -9.1680e-04,
         1.0277e-04,  5.6857e-04,  4.7423e-04,  7.5281e-05,  6.0476e-04,
        -4.8873e-04, -5.2992e-04,  7.7014e-06,  2.7435e-04,  1.0624e-03,
         4.4706e-04,  3.0993e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4940e-03,  9.8688e-01, -7.1736e-04, -3.1320e-04, -1.2736e-03,
        -1.0161e-03,  4.8623e-04, -8.2702e-04, -8.3960e-04, -1.7099e-04,
        -1.5371e-03, -7.9749e-04, -7.9622e-04, -1.8532e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+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([-3.2929e-03, -9.8524e-01,  4.1989e-05, -4.8664e-04,  1.4450e-04,
        -5.2495e-04,  9.1212e-05, -3.9233e-04,  4.4539e-04, -3.9609e-05,
        -8.1815e-04, -5.4636e-05, -3.0967e-04, -1.4993e-04,  3.4547e-04,
        -4.4879e-04,  2.9179e-04, -4.2125e-04,  1.3272e-04,  3.2497e-04,
        -6.0727e-05, -2.1693e-04, -2.2558e-04,  7.4573e-05, -3.2478e-04,
        -3.1134e-04,  3.9823e-05,  3.2386e-04, -1.8889e-04, -5.3420e-04,
        -1.1552e-03,  2.8470e-04, -5.7482e-04, -3.8254e-04, -2.7205e-04,
        -7.5270e-05,  4.5021e-05, -3.0041e-04, -4.4976e-04, -1.6451e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2167e-03, -9.8591e-01,  3.0274e-05,  1.5310e-04, -1.0878e-04,
         9.6072e-04,  1.8178e-04, -5.2079e-04,  4.0510e-05,  3.4510e-04,
        -1.9522e-04, -1.2616e-04,  1.2106e-04,  7.0310e-05,  1.5446e-04,
         5.6229e-04, -6.5407e-04,  2.7366e-04,  1.3018e-04,  3.1130e-04,
        -4.5675e-04, -2.5508e-04,  2.6754e-05, -3.9714e-04, -4.6080e-05,
         1.7698e-04, -1.5237e-04, -1.1627e-06, -4.8206e-04, -1.5749e-04,
         5.8627e-04, -9.6057e-05,  4.5297e-05,  2.9404e-04, -8.8985e-05,
        -2.8731e-04, -1.4083e-04, -9.0009e-05, -6.2288e-05,  2.2376e-04,
        -9.8081e-05, -3.6047e-04, -9.5540e-06, -3.7538e-04,  1.4900e-04,
        -6.4414e-04,  7.8673e-05,  1.3828e-04, -1.7489e-04,  1.5041e-05,
         8.8718e-06, -1.8459e-04, -2.0061e-04,  3.8296e-04, -7.4666e-04,
         2.9907e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2163e-04, -9.9142e-01,  6.3560e-04,  5.8235e-06,  3.2547e-04,
         2.0859e-04,  2.1835e-05,  5.9183e-04, -4.3577e-04,  2.7039e-04,
        -5.1046e-05, -1.8128e-04,  3.3605e-06,  9.8011e-05, -2.5657e-04,
        -3.2848e-04,  2.0535e-04,  2.3667e-05,  3.0795e-04, -8.5562e-04,
        -3.9024e-04,  6.6613e-04, -2.4654e-05, -3.1918e-04,  5.0150e-04,
        -2.8323e-04, -4.4709e-04,  4.6877e-05,  3.1532e-04,  7.5456e-05,
         1.3502e-04,  4.5910e-05, -1.6264e-04,  2.4039e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2525e-03,  9.8867e-01,  3.9449e-04, -1.7792e-03,  4.6990e-04,
        -8.5550e-04, -1.7876e-04,  6.6413e-04,  2.7739e-04, -3.0896e-03,
         3.7234e-04,  5.3033e-05,  9.4468e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5142e-03,  9.8231e-01, -3.3433e-04,  4.3657e-07, -9.9136e-04,
         1.7173e-04,  4.6361e-04, -1.6723e-04, -1.2281e-03, -2.5536e-04,
         5.4393e-05,  1.3189e-03,  5.8878e-04, -3.3516e-04, -1.3295e-04,
        -1.4097e-04, -8.7437e-05,  9.4170e-04, -2.4772e-04,  6.2565e-04,
         8.3551e-04,  7.0228e-04, -2.7222e-04, -6.6673e-04,  1.1457e-03,
        -1.4457e-03,  1.1794e-04, -1.0894e-03, -3.9412e-04, -4.1987e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4854e-03, -9.7972e-01, -4.8219e-04, -3.6927e-04,  7.6002e-04,
         2.9107e-04,  5.5021e-04, -7.1298e-05, -1.5332e-03, -7.7560e-05,
        -5.2368e-04,  5.8556e-05, -8.5800e-04, -1.0332e-03, -5.6261e-04,
        -1.6558e-04,  9.1135e-04, -6.0220e-04,  2.1279e-04,  1.5399e-03,
         2.1927e-04,  1.3422e-04, -9.1520e-04, -9.8382e-04, -9.1457e-04,
        -4.5877e-04, -1.2429e-03, -1.1558e-03, -1.5255e-04, -1.4655e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5850e-03,  9.8605e-01, -4.0569e-04,  1.0143e-04,  6.1437e-04,
        -3.6079e-04, -4.8688e-05,  4.5835e-04, -2.5833e-05,  7.0476e-04,
        -7.4590e-04,  9.5018e-04,  8.6301e-04,  5.1449e-04,  5.3450e-04,
        -1.0581e-04,  9.3967e-04, -5.8464e-04, -1.1196e-05, -7.0341e-04,
         8.2758e-04, -7.6932e-04, -3.6881e-05, -9.3611e-05,  9.5445e-05,
        -1.1764e-04,  4.1416e-04, -5.2003e-04, -8.2082e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0232e-03, -9.8751e-01,  6.3558e-04, -1.6803e-04,  1.8625e-03,
         4.6349e-04,  2.0023e-04, -1.4730e-04, -1.1767e-03, -3.2460e-03,
         1.0012e-03,  5.9090e-04,  9.7973e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8613e-03, -9.8671e-01, -7.1857e-04,  3.3540e-05, -5.9702e-04,
         5.6287e-04, -2.2483e-05,  9.5407e-04,  3.7609e-04, -7.8920e-04,
        -6.0416e-04,  9.2551e-04, -4.5308e-05,  5.2627e-04,  3.6472e-04,
         8.9418e-04, -8.9765e-04,  7.1653e-04, -9.1279e-05, -5.7023e-05,
         1.2373e-03,  1.0163e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.3728e-04, -9.9092e-01,  4.4813e-04,  1.1141e-04,  6.0179e-05,
        -1.2277e-04, -2.3468e-04, -3.1660e-04,  4.1877e-04,  1.4074e-05,
        -4.9504e-04,  2.9040e-04,  5.1582e-04,  5.4862e-04, -2.5850e-04,
         4.4218e-04,  3.2481e-04,  1.3251e-04, -1.7924e-04,  5.2607e-04,
        -2.5205e-04,  7.5822e-04, -5.1582e-04,  1.3750e-04,  6.5195e-04,
         2.7833e-05,  1.7985e-04, -1.2427e-05,  1.6792e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8130e-03, -9.8624e-01,  7.6536e-04,  4.1978e-04, -2.2268e-04,
         3.2444e-04, -1.5324e-04,  3.2171e-04,  3.0086e-04,  5.5055e-04,
         4.9420e-04,  6.6608e-04,  3.5724e-04,  3.2946e-04,  4.6464e-04,
         1.4473e-03, -2.1738e-04, -4.6904e-04,  4.3971e-04,  7.7120e-04,
         7.5772e-04,  2.7276e-04, -2.4843e-04, -9.3760e-05, -1.0045e-04,
         4.6630e-04, -2.1966e-04,  6.3514e-04,  4.3344e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9807e-03,  9.8162e-01, -2.2171e-03, -1.6391e-04,  9.7152e-04,
         1.0426e-03,  7.3677e-04, -3.5948e-04, -2.2483e-03, -4.3952e-04,
         1.0833e-04,  2.1841e-04,  8.9735e-04, -5.4668e-05,  1.4083e-05,
         4.7980e-04,  3.0769e-03,  7.2631e-05,  1.2981e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 7.2173e-04, -9.9023e-01,  3.8603e-04,  4.0991e-04,  2.9241e-04,
        -2.1038e-04,  4.4535e-04,  3.3388e-05,  7.2623e-05,  4.1251e-04,
         2.0215e-04,  1.0035e-04, -1.6890e-04,  3.7384e-04, -4.0656e-04,
         2.8991e-04,  5.2882e-04,  2.1563e-04,  8.9649e-05,  2.8599e-04,
        -2.6682e-04,  8.1282e-06,  4.2320e-04, -1.2571e-04, -3.3738e-04,
         5.9640e-04, -4.6438e-04,  7.2583e-05,  3.4678e-04,  3.2459e-05,
         2.7376e-04,  1.7106e-04, -4.3784e-05, -7.1819e-05,  1.2460e-04,
         6.8842e-04,  7.9969e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.3294e-03, -9.7378e-01, -4.2420e-04,  1.9452e-04,  6.8166e-05,
         3.0119e-04, -1.5812e-04, -3.1023e-04, -1.2820e-03, -3.3018e-04,
        -1.2846e-03,  8.3831e-04,  6.3402e-05, -5.3702e-04,  2.1769e-04,
        -8.6057e-04, -6.6235e-04, -1.0723e-03,  1.7547e-03, -8.8319e-04,
        -6.5600e-05,  7.1991e-04, -6.5899e-04,  3.6735e-04,  2.4855e-04,
         6.4876e-04, -8.4863e-04,  3.8116e-04, -1.3485e-03, -6.7329e-04,
        -1.0897e-03, -5.6644e-04, -2.6243e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2481e-04, -9.8684e-01,  2.1040e-04,  2.8233e-04,  8.5627e-04,
         1.1890e-03, -5.3800e-04, -5.9374e-04,  7.2594e-04,  2.0105e-03,
        -1.8575e-04,  1.2231e-03,  2.2144e-04, -1.7469e-04,  5.1953e-04,
         7.5175e-04,  2.2954e-04,  3.6726e-05,  7.6425e-04, -2.8977e-04,
         1.1381e-03,  6.9264e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6747e-03, -9.8511e-01, -2.4156e-04, -2.2416e-04,  2.6556e-05,
        -1.6048e-04,  8.9050e-04,  2.6356e-05, -2.6949e-04, -1.5829e-04,
         8.0983e-05,  7.8736e-04,  1.5621e-05,  7.2938e-04,  3.3834e-04,
         9.8438e-04, -2.5207e-04, -2.3952e-04, -6.8510e-04, -2.7009e-04,
         2.1371e-04, -2.3430e-04,  2.1626e-04,  8.0248e-05, -5.8403e-04,
         3.5502e-04, -1.6625e-04,  1.0786e-03,  7.6616e-04,  5.0491e-04,
         1.0027e-03,  3.6463e-04,  1.3053e-04,  1.3633e-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([-1.5816e-04,  9.9105e-01,  2.0190e-04, -1.1335e-04, -2.6472e-04,
         1.1441e-03,  5.8681e-04, -1.4986e-03, -1.3950e-03, -5.7981e-04,
         1.0672e-03, -5.0225e-04, -3.7673e-04,  9.2429e-04, -1.3856e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9116e-03,  9.8477e-01,  6.7877e-04,  3.1744e-05,  1.3829e-03,
         5.4226e-04,  5.8805e-04, -2.2646e-04, -1.7305e-03, -2.3209e-04,
         1.1129e-04,  9.4208e-04, -1.1368e-03,  9.5272e-04,  6.6482e-04,
         9.9336e-04, -1.2471e-03, -8.9657e-04, -3.6533e-04,  5.3964e-04,
         5.4236e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1003e-03, -9.8132e-01, -3.4788e-04,  4.7787e-05,  1.5290e-04,
         2.0758e-04, -2.3049e-04, -1.1652e-04, -1.9851e-04,  1.8732e-04,
        -2.7858e-04, -3.9866e-04,  1.0426e-03, -5.7584e-05, -1.4253e-04,
         1.3546e-03, -4.0437e-04,  2.9589e-04, -4.5436e-04,  3.0765e-04,
        -6.0373e-04, -3.0165e-04,  6.0534e-04, -4.8075e-04, -5.2839e-04,
         2.3979e-04,  5.1013e-04,  3.2217e-05,  5.7702e-04, -3.9837e-04,
        -3.2535e-04, -6.1077e-04,  7.0144e-04,  8.1147e-05,  4.6821e-05,
        -4.6779e-04, -1.0339e-04,  3.4710e-04, -3.7751e-04, -1.4881e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5640e-03, -9.8636e-01, -1.3164e-03, -9.4686e-05,  1.6573e-04,
        -6.6983e-04, -4.9353e-04,  3.3648e-04,  7.1296e-05,  6.4068e-04,
         3.7027e-04, -1.4832e-03,  7.4036e-05, -3.0160e-04, -3.8870e-04,
         8.7662e-04,  1.1059e-03,  1.3320e-03, -6.1371e-05, -1.9711e-04,
         9.4945e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3010e-03, -9.8633e-01,  1.9150e-04,  1.0128e-03, -6.4736e-04,
        -7.1564e-04,  3.1627e-04, -8.3127e-04,  8.1267e-05, -6.3011e-06,
         5.2004e-04,  1.0902e-03,  6.9181e-04, -4.2034e-04,  4.8273e-04,
        -6.3581e-04, -1.7237e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4756e-03,  9.8678e-01,  9.8633e-04,  8.6717e-05, -1.6213e-04,
        -6.5510e-04,  3.8377e-04,  1.9109e-04,  8.0572e-04, -2.4959e-04,
        -4.8487e-04, -3.2243e-04,  1.6433e-04, -1.3167e-03, -2.6354e-04,
        -3.5645e-04, -7.6656e-04,  8.1559e-04, -7.1169e-04, -1.4924e-04,
        -2.7732e-04,  1.5153e-04, -8.8251e-04,  4.8356e-04, -4.8710e-05,
         3.2990e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2846e-03,  9.8158e-01, -1.0021e-03,  1.5567e-04, -8.0032e-04,
        -4.2368e-03,  1.7853e-03,  1.6316e-04, -1.0418e-03, -2.0060e-03,
        -3.7585e-04, -1.4552e-03, -1.0393e-03, -1.0783e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9410e-03, -9.8541e-01,  4.0122e-04,  5.8018e-04, -6.8782e-04,
        -6.9088e-04, -1.0613e-03, -6.9649e-04, -4.4147e-04,  1.7409e-04,
         7.9121e-05, -3.6738e-04, -1.0059e-04,  3.9639e-04, -1.0744e-04,
         2.3315e-04,  2.4649e-04, -3.1969e-04,  3.7159e-04, -2.2077e-04,
         2.5924e-05, -8.6976e-04, -1.5355e-04,  5.9016e-05,  1.7670e-04,
         2.0766e-04, -4.0425e-04, -5.7514e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+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([ 1.9323e-03,  9.9312e-01,  1.7759e-03, -4.1725e-04, -3.4668e-04,
         4.3728e-04, -6.1599e-04,  6.2327e-04,  7.2874e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1375e-03,  9.9218e-01,  4.7895e-04, -2.5981e-04,  1.5401e-04,
         1.6414e-04,  3.8065e-04, -7.1767e-04,  8.5961e-04,  9.4955e-05,
        -1.2622e-03,  5.6709e-04,  2.8784e-05,  7.0947e-05,  1.1235e-04,
         1.0547e-04, -3.2602e-04,  3.1642e-06,  3.7348e-04,  6.6189e-04,
        -5.6521e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2979e-03,  9.8571e-01,  5.5083e-04,  8.2635e-04, -7.2710e-04,
        -7.2969e-05,  9.1240e-04,  3.7830e-04, -1.8756e-06, -5.6130e-04,
        -7.9236e-04,  1.9913e-04, -1.3561e-03, -2.2289e-03,  1.5332e-04,
        -1.7386e-03, -4.6684e-04, -1.5495e-04, -4.5533e-04, -4.1927e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5478e-03, -9.8356e-01,  6.7089e-04, -1.3513e-05, -8.1550e-04,
         1.3277e-03, -9.4584e-04, -3.5487e-04, -2.1922e-03,  2.0651e-04,
         1.2943e-03, -6.4843e-04,  8.5776e-04, -1.4196e-03,  6.0471e-04,
        -1.5412e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8343e-05,  9.8631e-01,  7.9625e-04, -1.1854e-03, -3.8520e-04,
         3.8923e-04, -8.1343e-04, -3.9939e-04, -2.2814e-03, -1.0433e-03,
         1.0830e-03, -1.2122e-03, -1.1319e-03, -6.8579e-04, -7.6866e-04,
         8.6359e-04, -6.3313e-05,  5.3768e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7960e-03,  9.8527e-01,  6.6666e-04, -1.3668e-03, -5.5039e-04,
         4.0675e-05,  3.9984e-04,  9.3161e-04,  7.4946e-04, -3.8389e-04,
        -3.2177e-04, -3.5964e-04, -8.5960e-04,  1.4642e-03,  7.6771e-04,
        -1.1084e-03, -4.4284e-04, -7.6963e-04, -1.5082e-03, -2.4463e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5201e-03,  9.8516e-01,  1.2059e-04,  9.4997e-05, -4.3798e-05,
        -6.7895e-05, -7.0275e-05, -5.9604e-05,  6.7716e-04, -9.6942e-04,
        -3.6583e-04,  8.4680e-04,  3.0381e-04,  1.1703e-03,  6.2015e-04,
         1.1612e-04, -5.3533e-04, -6.0170e-04, -3.7910e-04, -1.3876e-03,
        -1.1852e-03,  2.9716e-03, -7.3388e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0524e-03,  9.8234e-01, -4.2742e-04,  5.6187e-04,  4.7681e-04,
        -3.4699e-04, -6.7629e-04, -2.7501e-04,  1.9651e-04, -3.0031e-04,
        -9.4489e-04, -9.7189e-04,  6.6764e-04, -5.7043e-04, -1.9165e-04,
        -9.5495e-04, -6.3979e-04, -7.9803e-04, -4.3561e-04, -7.9421e-04,
        -3.8983e-04, -5.2174e-04, -7.3227e-04, -4.7607e-04, -4.1160e-04,
        -5.5360e-04,  7.3417e-04, -5.9424e-04, -1.1474e-03, -5.0239e-04,
        -3.1156e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.2658e-04, -9.9151e-01,  5.0005e-04, -6.9159e-04,  1.8786e-04,
         2.0390e-04,  4.5413e-04, -2.0004e-04,  2.3659e-04, -1.3925e-04,
        -3.9025e-05, -5.8062e-04,  6.2762e-05, -7.0690e-04, -3.3326e-04,
        -6.3568e-04,  1.3854e-04,  3.8335e-04,  3.5791e-04, -2.0429e-04,
         1.2533e-04, -2.1469e-04, -4.7032e-04,  2.8244e-04,  1.2472e-04,
        -4.4268e-05, -1.0366e-04, -4.4392e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9984e-03,  9.8951e-01,  1.2316e-03, -1.3520e-03, -1.2158e-04,
         6.7838e-04, -5.6394e-04, -6.1402e-04, -6.6651e-04, -2.3340e-04,
         3.0932e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1984e-03, -9.7969e-01, -5.5716e-04,  5.4640e-05, -1.4813e-03,
         1.8395e-04, -8.0319e-04,  2.0639e-04, -1.0749e-03,  5.3887e-04,
        -5.3568e-04,  5.6669e-04, -1.2074e-03, -8.3250e-04, -2.6158e-04,
        -1.1951e-03, -1.7881e-03,  9.2836e-04,  3.3247e-04,  2.6565e-04,
        -9.6354e-04,  3.9569e-05, -7.4085e-04, -3.1721e-04, -1.1732e-03,
         4.2500e-04, -1.3053e-04,  1.1667e-04,  3.8815e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0853e-03,  9.8894e-01, -4.2849e-04,  5.2906e-04, -2.9634e-04,
         9.3234e-05, -9.7327e-04,  3.3932e-04, -4.6455e-05,  6.7565e-04,
         5.4759e-04,  8.0936e-04,  3.6703e-04, -9.5708e-04, -2.9426e-04,
        -6.9618e-04,  8.3357e-04,  5.1008e-04,  1.7064e-04, -3.6382e-04,
         1.3713e-05, -2.5591e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+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([ 3.2902e-04,  9.8856e-01, -1.4327e-03,  2.8787e-04, -2.5305e-03,
        -1.8761e-04,  1.1434e-04,  1.6842e-04, -2.8781e-04, -4.2301e-04,
        -1.0082e-03, -1.9072e-03, -1.9031e-03, -8.6437e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8510e-04, -9.9114e-01, -5.6993e-04, -2.1507e-04, -1.0164e-04,
        -1.0730e-03,  4.8667e-04,  5.4549e-04,  7.1193e-04, -5.8182e-04,
         4.4110e-04,  4.0181e-04,  2.5224e-04, -2.9816e-05,  1.6171e-04,
         1.2689e-04,  2.7808e-04, -8.3062e-04, -3.9501e-04,  3.7951e-04,
         2.5799e-04,  4.8520e-04,  2.5052e-05, -1.8380e-04, -4.1330e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9433e-04,  9.8784e-01,  1.1548e-03, -1.0519e-03,  1.6177e-05,
        -7.8769e-05,  2.4158e-04,  8.8018e-04,  7.1506e-04,  2.8789e-04,
        -4.9282e-04, -1.7354e-04,  1.4882e-04, -2.0649e-05, -5.3844e-04,
         1.0940e-04, -9.5306e-04, -2.5232e-04, -1.1065e-05, -3.2270e-04,
        -3.4033e-04, -2.4154e-04,  2.4548e-04, -4.3174e-04, -9.9925e-04,
        -5.2623e-04, -2.4877e-04,  2.1484e-06,  7.7793e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1523e-04,  9.8450e-01,  1.6063e-03,  4.6932e-04, -7.4607e-04,
         2.7986e-03,  2.3152e-03, -1.4304e-03,  2.9416e-03,  1.1727e-03,
         1.0928e-03,  4.0677e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1239e-03,  9.8568e-01, -1.3742e-03, -2.4342e-03,  1.3081e-03,
        -8.5220e-04,  3.3695e-04,  6.3444e-05, -7.7466e-04,  1.0553e-03,
         1.2733e-03, -1.1701e-04, -8.0870e-04, -4.4147e-04, -5.9141e-04,
         1.7660e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.8040e-03,  9.7882e-01, -4.4269e-04,  6.9311e-04,  8.9837e-04,
         5.9940e-04,  2.7866e-04,  4.6930e-04,  1.2104e-04,  4.6674e-04,
         1.2829e-03,  3.8352e-04,  6.3405e-04, -3.2349e-04,  4.8843e-05,
         1.9878e-04,  2.2406e-03, -1.1276e-03, -5.2733e-04, -5.5116e-04,
        -6.4708e-04, -6.6709e-04,  4.9363e-04,  9.0392e-04,  6.2389e-04,
         5.8075e-04,  1.1708e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9751e-03, -9.8739e-01,  7.1161e-04,  2.8442e-04, -9.5625e-04,
         9.9078e-05,  1.8250e-04, -1.0432e-03, -1.3581e-04, -9.3261e-04,
        -2.2030e-04, -9.7674e-04, -3.8662e-04,  8.9098e-04, -8.3430e-04,
         4.5454e-04, -6.7465e-05,  6.3167e-05,  8.8412e-05,  1.8170e-04,
         4.2939e-04,  4.4372e-04, -9.2339e-04,  3.3283e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4330e-03, -9.8937e-01, -2.4236e-04, -5.1046e-04, -1.9966e-04,
        -2.9621e-04,  1.6097e-05,  1.1763e-04, -2.4059e-04,  9.1544e-05,
        -9.6551e-06, -2.6731e-04,  4.7086e-04,  6.0259e-04, -1.5914e-04,
        -3.6805e-04, -4.0414e-04,  4.3813e-04,  1.0989e-04, -5.5121e-04,
         1.5394e-04,  5.9392e-05, -1.6846e-04, -1.9285e-04, -5.7124e-04,
        -2.2523e-04,  2.0323e-04,  1.4688e-06, -1.1736e-04, -3.8379e-04,
        -5.5975e-04,  4.1634e-04, -2.4667e-05,  2.5712e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1329e-05,  9.8755e-01, -6.6943e-04,  5.7486e-04,  6.9538e-06,
         4.7391e-04,  3.9102e-04, -1.1421e-04,  5.0402e-04,  1.9364e-04,
         8.7743e-04,  6.9121e-04,  3.9193e-04,  4.1772e-04,  1.0055e-03,
        -1.6329e-04,  1.1728e-03,  5.1531e-04, -2.3585e-05,  1.0075e-04,
        -7.5225e-04,  7.8926e-04,  4.3019e-04,  4.1436e-04,  5.6558e-04,
        -3.9167e-04, -1.7212e-04,  3.0095e-04, -2.0465e-04,  1.2683e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3399e-03,  9.8695e-01,  4.0479e-04, -6.9613e-04, -2.6313e-04,
         1.0736e-03, -1.8377e-04, -9.3246e-04,  2.7938e-04, -2.8512e-05,
        -9.5937e-04,  1.1103e-04, -5.9503e-04, -9.5541e-04, -1.0074e-03,
        -4.5087e-04, -8.8708e-04, -8.2880e-04, -2.3619e-04, -2.4660e-04,
         3.9260e-04,  3.8899e-04, -7.8775e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2253e-03, -9.8868e-01, -7.3278e-04, -7.0670e-05, -6.0714e-04,
        -4.6879e-04, -1.0230e-05, -5.0046e-05,  1.3839e-05, -3.3418e-05,
        -2.9428e-04,  1.5008e-04, -5.8136e-04, -2.5287e-06, -3.3109e-04,
         2.1147e-05, -5.4076e-04, -3.6060e-04,  9.9204e-04, -3.2142e-04,
        -4.3124e-04, -1.4772e-04, -1.1293e-04,  1.2557e-05, -8.8035e-05,
        -6.3892e-04,  2.4486e-04, -2.2006e-04, -4.7779e-04, -8.1797e-04,
        -3.2477e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.1988e-04, -9.8872e-01, -2.9129e-04,  4.2426e-04, -4.1328e-04,
        -4.7879e-04, -3.4885e-04, -6.9186e-05,  3.2164e-04,  1.0402e-04,
         4.9628e-05,  5.8503e-05, -4.0911e-04, -3.9970e-04,  1.8376e-04,
         4.8970e-04, -9.2475e-05, -9.6871e-05, -1.6270e-04,  2.4728e-04,
        -5.3084e-05, -2.8470e-04, -6.0670e-04,  4.8428e-05, -2.6093e-04,
        -7.1112e-05, -1.7454e-04, -1.5063e-04,  4.1717e-04, -7.5805e-05,
        -5.4686e-05, -7.5515e-05, -2.5929e-04, -6.0399e-04, -5.8690e-04,
        -6.6059e-04, -2.6167e-04,  4.5986e-04, -4.6679e-04,  1.5115e-04],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 7.5435e-04,  9.8849e-01, -7.5337e-04,  1.2255e-03,  4.7440e-04,
        -1.2980e-03, -2.5823e-04, -8.1315e-04, -7.5470e-04, -1.0655e-03,
         7.1671e-04, -3.0653e-04, -7.3265e-04, -3.0259e-04, -5.5753e-05,
        -9.8338e-04,  1.2542e-04,  4.5187e-04, -1.7149e-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,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7587e-03, -9.8966e-01,  1.9980e-04, -6.8165e-04,  3.1917e-03,
         1.4159e-03, -8.4039e-04,  2.2496e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1034e-03,  9.9033e-01,  1.2896e-03,  8.8816e-04,  7.8425e-04,
        -6.5115e-04,  6.3073e-04, -2.3284e-03, -9.7849e-04,  1.2380e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3477e-03, -9.8674e-01, -8.4569e-04, -5.3568e-04, -4.0014e-04,
         8.7292e-05, -1.7962e-04, -5.9709e-05, -9.4636e-05, -7.0869e-04,
        -5.0553e-04, -1.2870e-05, -2.2945e-04, -8.1692e-04,  9.8591e-05,
        -3.4513e-04, -1.0122e-03,  2.9978e-04,  5.1796e-04,  5.0163e-04,
        -7.0575e-04,  3.1831e-04, -1.2258e-04, -2.0571e-04,  8.5156e-04,
        -5.4728e-04,  2.1004e-04,  5.7994e-04, -4.0589e-04, -7.1172e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3421e-03,  9.8796e-01, -5.1756e-05, -4.0069e-04, -7.0217e-04,
        -4.6809e-04, -2.5770e-04,  2.8699e-05, -7.2627e-04,  5.1628e-04,
         1.4631e-04, -2.0579e-05, -4.9639e-04, -8.5592e-04, -6.2961e-04,
        -8.4129e-04, -8.4018e-04,  2.3081e-04, -2.5534e-04, -1.0113e-03,
        -6.5567e-04,  3.6603e-04,  1.0050e-04,  9.3606e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7331e-04, -9.8906e-01, -9.9611e-05, -2.9841e-04, -3.4314e-04,
         2.6948e-04,  4.8521e-04, -4.9234e-04, -6.2071e-04, -6.9607e-04,
        -2.3803e-04, -4.7347e-04,  3.6870e-04, -1.9016e-04,  3.4319e-04,
         1.0690e-04, -8.0163e-04, -4.9210e-05, -7.8187e-05, -6.4947e-04,
         9.7256e-04,  1.1785e-03,  3.4163e-04, -1.0053e-04,  7.6804e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3612e-03,  9.8424e-01, -2.5238e-04, -3.2231e-04,  1.7171e-04,
         2.3431e-04, -1.8943e-04, -5.8030e-04,  1.4995e-04, -3.5566e-04,
         2.6599e-04, -1.1833e-04, -7.8171e-04,  3.9444e-04,  4.7191e-04,
         1.2210e-04, -3.5920e-05,  1.6206e-04, -2.6021e-04,  4.5407e-04,
         4.0223e-04,  4.9654e-04,  1.3770e-04, -5.9204e-04, -3.1800e-04,
        -4.8368e-04, -3.0597e-04,  4.9905e-04, -1.2683e-04, -7.6651e-04,
        -1.3145e-03, -1.9295e-04, -4.4260e-04,  7.9000e-05, -5.8916e-04,
         1.8703e-04, -3.4476e-05, -7.3453e-06, -1.0308e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6122e-03,  9.8588e-01,  1.7200e-03,  1.2861e-03,  1.7973e-03,
        -1.1348e-03, -5.6734e-04,  1.7142e-03,  4.1621e-04,  7.9244e-04,
         1.6449e-03,  1.4328e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6157e-03,  9.8870e-01, -4.7140e-04,  1.6815e-03,  1.2127e-03,
        -7.9403e-04,  2.3666e-03,  5.4304e-05,  1.2769e-04, -6.7557e-04,
        -2.9749e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2809e-03,  9.8244e-01, -7.2390e-04, -3.6161e-04,  9.2658e-04,
        -1.8813e-04, -4.4732e-04, -5.9595e-04, -8.7732e-04,  4.3788e-04,
         3.7773e-04, -7.2765e-04,  1.1264e-04, -4.3764e-04, -7.5492e-05,
        -1.0961e-03, -6.6755e-04, -5.3956e-04, -9.5636e-04, -1.0006e-04,
        -8.6183e-04, -1.1643e-03,  6.0383e-04,  5.5243e-04, -4.4644e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1083e-04, -9.9241e-01,  3.3433e-04, -6.5101e-05,  3.1320e-05,
         5.2064e-04, -2.3915e-05,  8.6727e-04,  6.4211e-04, -5.8454e-05,
         4.2689e-04,  2.0801e-04, -8.6881e-05, -1.9162e-04, -2.7292e-04,
        -1.0133e-04,  1.0791e-03,  2.9578e-04, -5.1379e-05,  1.3335e-03,
         8.1744e-04,  6.8160e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7005e-03,  9.8186e-01,  5.1938e-05,  2.6375e-04, -8.2822e-04,
        -7.4862e-04,  1.6143e-03, -4.8414e-04,  1.2636e-03, -3.4796e-04,
        -4.0769e-04, -4.7652e-04, -2.9738e-04,  4.2323e-04, -8.2934e-04,
        -7.7398e-04, -5.5797e-04, -4.4375e-04,  9.7459e-05,  4.5517e-04,
        -5.2293e-04, -7.7041e-04,  2.2847e-04, -2.8670e-04, -7.6476e-04,
         4.7079e-05,  2.1323e-04,  1.6500e-04,  2.4336e-04, -1.4235e-03,
        -4.0954e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-3.5732e-04, -9.8873e-01, -1.8433e-04, -2.6211e-04, -3.2444e-05,
        -2.8436e-04, -2.6167e-04, -3.9268e-04,  1.9069e-04, -1.2505e-04,
        -1.8349e-04, -3.0204e-05, -3.5699e-04, -8.7555e-05, -4.0829e-04,
        -2.2456e-04, -2.6132e-04, -1.5960e-04, -4.0493e-04, -2.0423e-04,
        -4.1771e-04,  3.2430e-05, -1.9753e-04, -1.5840e-04,  1.7084e-04,
        -2.5960e-04, -2.1613e-04, -1.7017e-05, -2.5668e-04, -2.3325e-04,
         1.2577e-04, -3.6823e-04,  4.8930e-05,  4.9570e-05, -8.5411e-05,
         1.7870e-04, -2.5606e-05,  2.7350e-04,  2.7360e-05, -2.8619e-04,
        -1.4146e-04, -2.7309e-04, -2.1707e-04, -4.9131e-05, -1.4597e-04,
         2.5932e-04,  2.5407e-04, -1.1225e-05, -2.5257e-04, -3.0163e-04,
         1.8906e-04, -9.1475e-05, -2.7794e-04, -3.1423e-04, -3.2646e-04,
         5.0466e-05, -2.7213e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.2916e-03,  9.7415e-01, -4.3351e-03, -1.8368e-03, -5.6801e-03,
        -7.5145e-04, -9.0329e-04, -9.1273e-04, -4.9026e-03,  2.3284e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.0977e-03,  9.7931e-01,  4.1181e-04, -7.1618e-04, -3.0265e-04,
        -5.2242e-04,  7.6312e-04,  6.4368e-04,  5.0207e-04,  1.1860e-03,
         2.6725e-04,  8.5840e-04,  9.7243e-04, -2.4721e-04, -5.5104e-04,
         1.5357e-04, -1.1209e-03, -8.1456e-04,  7.8429e-04,  3.6778e-04,
         6.0856e-04,  1.3110e-03,  1.1683e-03, -6.7860e-04, -5.2871e-04,
        -1.1630e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4149e-03,  9.8980e-01,  1.3919e-03, -4.8846e-04,  7.5650e-04,
        -1.2660e-03,  1.1309e-03, -4.7408e-04, -5.8327e-05,  4.5552e-04,
        -1.8818e-03,  8.8104e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3509e-03, -9.8552e-01,  2.7505e-04, -6.0407e-05,  9.3733e-05,
        -3.2781e-05, -6.0023e-04, -5.3594e-04, -1.3372e-03, -2.0278e-04,
        -5.6508e-04, -4.1957e-04,  9.9795e-04,  2.0456e-04, -8.6868e-04,
        -3.6864e-04, -9.5524e-04, -3.6972e-04,  1.2378e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.4768e-04, -9.9196e-01, -3.5568e-04, -1.8779e-05,  3.5887e-06,
        -3.1835e-04, -1.3149e-04,  1.0707e-04,  2.0681e-04, -8.7009e-04,
        -1.6248e-05, -6.6220e-04,  4.6854e-04, -6.6887e-05,  1.3702e-04,
        -1.3342e-04,  6.4600e-04, -4.3834e-06,  4.9248e-04, -7.0738e-04,
         1.9579e-04,  5.4126e-04,  4.6491e-04, -4.6853e-04,  2.7435e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9038e-05, -9.9023e-01,  6.8275e-04, -7.4495e-04, -8.4095e-05,
         2.5553e-04, -1.0150e-03, -4.9273e-06, -4.1096e-04, -1.5540e-03,
        -1.2107e-03,  1.7531e-04,  9.5982e-04, -2.2798e-04, -1.1990e-03,
        -2.0409e-04,  3.1948e-04, -2.3758e-04, -4.5240e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0147e-03, -9.9043e-01, -2.8535e-04,  1.9057e-04,  2.4044e-04,
        -1.2591e-04,  1.8642e-05,  6.0568e-04,  9.1138e-05,  4.8913e-05,
        -1.6440e-04,  3.4284e-05,  3.5969e-04, -2.8830e-04,  6.5016e-05,
         1.0281e-04,  5.3475e-05,  4.4301e-05, -1.1378e-04, -9.2982e-05,
        -6.7244e-05,  2.1508e-04, -5.7249e-05,  5.9773e-04,  1.9295e-04,
         6.9363e-04, -2.4657e-04, -8.6619e-05, -7.9873e-05,  2.6949e-04,
         2.6086e-04, -3.4088e-04, -1.1371e-04,  1.8795e-04,  4.0467e-04,
        -4.2217e-04, -2.3710e-05,  2.8588e-04,  8.4918e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3616e-03,  9.9127e-01, -2.9216e-04,  8.8664e-04, -5.4190e-04,
        -1.0953e-03,  2.3081e-03, -2.4460e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3023e-03, -9.9078e-01,  2.9494e-04, -1.6132e-03,  3.6031e-04,
         8.2578e-04, -2.2791e-04,  1.3118e-03, -1.0051e-03, -1.5212e-04,
         4.5116e-04,  1.2975e-03, -3.7786e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7604e-03,  9.8526e-01,  5.7966e-04, -4.2096e-04,  2.8281e-04,
        -1.1207e-03, -2.3751e-04, -2.4737e-04, -3.7389e-04, -2.3025e-04,
         3.7798e-04, -9.0645e-04,  3.8224e-04, -1.2196e-03, -3.9999e-04,
        -1.6787e-04,  7.5405e-04,  3.7097e-04, -1.5359e-04, -3.1601e-04,
        -4.2285e-04, -8.1557e-04,  1.3470e-04, -1.2736e-04, -1.3955e-04,
        -1.0159e-03,  8.1200e-04, -6.2546e-04, -6.9036e-05,  2.7147e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3554e-03,  9.8247e-01,  9.4908e-04, -6.6822e-04,  4.7282e-04,
         3.9564e-04,  7.9258e-04,  1.1035e-03, -3.4271e-05, -2.2590e-04,
         7.2953e-04,  9.5059e-05,  1.2450e-03,  6.7051e-04,  9.0539e-04,
         8.1506e-04, -1.1864e-03,  8.0275e-04,  1.1047e-03, -3.5280e-04,
         4.5073e-04, -1.3515e-04, -3.8347e-04,  3.2801e-04,  1.7634e-03,
        -2.1973e-04,  3.4579e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+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([-1.4991e-03, -9.8708e-01,  1.3322e-03, -1.9760e-06,  1.9741e-04,
        -2.0371e-04, -2.6480e-04,  6.4718e-05, -9.4042e-04,  1.0961e-03,
        -2.2408e-04,  1.3855e-03, -7.7588e-04,  7.2132e-04, -2.7487e-04,
        -4.0550e-04, -3.4614e-04,  6.0282e-04,  1.9233e-04,  1.0722e-03,
        -1.9870e-04,  1.6327e-04, -7.0522e-04, -2.5634e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.0672e-04, -9.9198e-01,  3.6288e-04,  2.3895e-04,  1.6309e-04,
         3.1568e-04, -2.9444e-04,  3.9234e-04,  7.1692e-05,  3.9785e-04,
         1.1831e-04,  4.3458e-04, -7.1370e-04, -2.1558e-04,  2.8461e-04,
         7.9107e-04, -4.4102e-04, -1.4607e-04, -2.0182e-05,  2.7479e-04,
         2.4630e-05, -1.6736e-04, -1.7268e-04,  2.2863e-04, -4.3706e-04,
        -1.4024e-04, -2.3938e-04,  2.2982e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2059e-04,  9.9008e-01, -7.0151e-05, -4.3911e-04, -1.2982e-03,
         1.2809e-03,  9.1037e-05,  2.0108e-04,  1.6164e-04, -3.5450e-04,
        -2.4006e-05, -1.1130e-03, -1.6322e-03, -1.4239e-03, -1.5089e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6539e-03,  9.8486e-01,  5.8729e-04, -4.5255e-04,  3.8672e-04,
        -1.4621e-04,  2.5021e-04,  5.6101e-04, -2.1774e-04, -5.1388e-04,
         9.0823e-05, -5.4156e-04,  5.1635e-04, -2.7187e-04,  5.9761e-04,
         7.4544e-04, -5.3452e-04, -1.7162e-04, -3.0738e-04, -4.0738e-04,
        -2.8410e-04, -3.9289e-04,  1.1584e-05,  9.3516e-04,  6.2763e-04,
         9.1017e-05,  5.0552e-04, -2.3112e-04,  1.2260e-03,  1.2973e-04,
         4.1588e-04, -2.6503e-04,  1.1819e-04,  6.2765e-04,  2.3207e-04,
         7.4334e-05, -2.0834e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8864e-03, -9.9086e-01, -1.3096e-04, -4.6814e-04,  6.2075e-05,
        -5.0528e-04,  6.4010e-04, -2.2670e-04, -1.3891e-04, -5.9414e-06,
        -4.6272e-05,  8.6434e-04,  3.9784e-04, -2.3874e-04, -5.9549e-04,
        -4.5435e-04, -7.1194e-04, -2.5232e-04, -8.0171e-04, -6.0408e-04,
        -4.4934e-05, -6.0794e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.2361e-04,  9.9005e-01, -6.9675e-04,  2.1636e-04,  4.6804e-04,
         1.5334e-03, -9.5488e-04,  4.4990e-04,  2.9686e-04,  2.9501e-04,
        -2.4794e-04, -2.0589e-04, -1.6587e-03, -2.3070e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.0166e-03,  9.8227e-01,  2.6679e-03,  1.3271e-03,  2.9423e-04,
        -1.8564e-03, -1.6296e-03, -2.9180e-04,  4.1688e-04, -5.1109e-04,
         2.7143e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3658e-04,  9.9184e-01, -9.0757e-05, -4.2009e-04,  9.0904e-04,
         2.7464e-04,  5.2359e-04,  1.2322e-03,  2.2439e-04,  1.0274e-03,
         2.6193e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1868e-04,  9.9135e-01,  7.7841e-04, -4.2157e-04,  2.0908e-04,
        -2.7494e-04,  4.7398e-04,  5.4162e-04,  1.2676e-04,  7.4516e-04,
         3.0331e-04,  1.3405e-04, -4.3592e-04,  2.2407e-05,  2.4602e-04,
        -1.6011e-03, -2.0265e-04, -6.6695e-04, -2.1180e-04,  9.3088e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1228e-03, -9.8932e-01,  8.1976e-05, -6.6655e-04, -4.0767e-04,
         4.2687e-04,  4.8549e-04, -7.6607e-04, -2.8410e-04, -5.5668e-04,
        -2.8275e-04, -3.2398e-04, -1.0421e-03, -1.7604e-04, -1.5791e-04,
         1.3889e-04, -7.5404e-04, -8.2766e-04, -5.9699e-04, -6.5668e-04,
        -5.3610e-04, -3.0931e-04,  7.7303e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1083e-03, -9.8788e-01,  1.6381e-04, -9.0648e-05,  2.2309e-04,
         4.6295e-04,  5.8511e-04,  3.2632e-04, -2.6483e-04,  1.1136e-04,
         6.7942e-04,  6.6856e-04, -1.7570e-04,  3.2727e-05,  4.9661e-04,
        -1.3702e-04,  7.9586e-04, -4.4563e-04,  6.8638e-04, -8.9281e-06,
         4.2510e-04, -2.9600e-04, -3.3112e-04,  3.4927e-04,  2.6832e-04,
        -8.4753e-04, -3.3985e-04, -7.3501e-04,  6.0402e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1225e-03, -9.8667e-01,  6.4321e-05, -5.6721e-04, -3.6934e-04,
        -6.9524e-04,  1.7862e-04, -9.2614e-05, -6.3325e-05,  7.2898e-04,
         2.1634e-04, -4.2909e-04,  5.1488e-04, -6.4430e-05,  3.8610e-04,
        -6.7991e-04, -2.7609e-04, -1.5127e-04, -1.5333e-04,  8.2517e-04,
        -3.0109e-04, -4.1466e-04,  1.2337e-04, -1.8687e-05,  5.1845e-05,
        -5.5454e-05,  2.3332e-04, -7.9416e-06, -6.4857e-04, -1.7522e-04,
         4.9349e-04, -1.1176e-04,  3.5261e-04,  4.9751e-04,  5.3949e-04,
         6.2918e-04,  2.8329e-05,  7.0737e-05], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 1.2439e-03, -9.9141e-01,  1.3519e-04,  2.6142e-04, -8.9244e-04,
        -5.7422e-04, -9.5923e-05, -5.6003e-04, -5.9613e-04,  2.3262e-04,
         1.2901e-03, -3.3625e-04, -2.8288e-04,  9.3641e-05,  2.9824e-04,
        -7.6614e-04,  5.2236e-04, -2.6976e-04,  1.3925e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4547e-03,  9.8525e-01,  5.4150e-04,  3.8575e-06, -7.5543e-04,
        -4.8651e-04,  4.5421e-04,  5.3108e-04, -6.1222e-04,  4.5331e-04,
        -2.3871e-04, -8.9525e-04,  1.7833e-04,  7.4044e-04,  7.2481e-04,
         4.8945e-04,  1.4081e-04, -8.5888e-04,  4.4617e-04,  7.4324e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8948e-03, -9.9053e-01, -2.6446e-05,  3.1638e-04,  6.2634e-04,
         1.1211e-03,  4.6780e-04, -3.0640e-05,  1.9842e-04, -5.3135e-04,
         4.4179e-04,  3.6411e-04, -2.2179e-04,  5.8353e-04,  9.1476e-04,
         6.2336e-04,  1.1568e-04,  2.8545e-04,  7.2959e-05, -3.0639e-04,
        -3.2271e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3591e-03,  9.8671e-01,  8.1442e-05, -4.3539e-04, -1.6089e-03,
        -8.2784e-05, -4.4823e-05,  1.3794e-03,  9.4039e-04, -7.5764e-04,
         1.2755e-04, -5.7285e-04, -2.0548e-04,  3.9589e-04,  9.8345e-04,
         1.5811e-03, -7.4452e-04,  8.9137e-04,  1.1003e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5612e-04, -9.8551e-01, -6.8027e-05,  5.9004e-05, -3.1279e-04,
         3.1781e-04,  6.6147e-04, -4.6572e-05,  5.0259e-04, -3.4848e-05,
         7.9553e-04,  4.5053e-04, -1.0894e-04,  4.1786e-04,  3.3051e-04,
         3.4991e-04, -1.5497e-04, -6.4804e-04,  3.4424e-04, -1.0634e-04,
         6.4910e-04,  1.6507e-04, -5.8163e-04,  7.1234e-04, -5.9872e-04,
        -4.8886e-04,  7.9840e-05,  1.0981e-04, -7.0567e-05, -1.8676e-05,
         7.8199e-04,  6.8687e-05,  8.2657e-05, -1.1421e-04, -2.0363e-04,
         1.9527e-04,  2.2898e-04,  2.7510e-04,  1.2443e-04,  2.8573e-05,
        -6.1079e-04,  2.2416e-04, -4.3438e-04, -5.9731e-04,  4.4227e-04,
        -5.4108e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.8851e-04, -9.8989e-01,  1.0973e-03,  1.5195e-04,  5.8239e-04,
         1.9576e-04,  2.1749e-04, -3.7746e-04,  3.3605e-04,  1.2450e-04,
         7.6518e-06, -3.0886e-04,  2.2250e-04,  7.5246e-04,  3.2825e-04,
        -5.4040e-05, -1.7753e-04, -2.0320e-04,  1.7469e-04,  1.2650e-04,
         2.5327e-04,  1.0557e-03, -7.4648e-04,  4.8185e-04,  2.8409e-04,
        -3.7241e-04, -2.7964e-04,  4.0945e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0049e-04, -9.9189e-01,  2.6583e-04, -6.9898e-05,  1.9725e-04,
         4.4388e-05,  1.0028e-04,  1.0127e-04, -4.2982e-04, -1.0543e-04,
        -1.9885e-04, -6.7076e-05, -2.3610e-04, -3.8057e-04, -7.5933e-04,
        -2.6106e-04,  3.3933e-04,  1.6446e-04,  1.8664e-04,  6.5921e-04,
        -3.4873e-04, -5.7053e-05,  3.1333e-04, -1.1086e-03, -1.4727e-04,
         5.2459e-05,  7.7968e-05, -5.0265e-04, -4.3627e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.0972e-04,  9.8802e-01, -1.1262e-04,  9.5662e-04,  7.7538e-04,
        -8.4035e-04,  7.4753e-04,  6.4995e-04, -4.5890e-05,  4.4824e-04,
        -8.8658e-05, -2.2355e-04,  3.1637e-04,  4.7907e-04, -3.4223e-04,
         1.7500e-04, -7.7305e-05,  3.8323e-04,  1.0135e-04, -3.7309e-04,
         4.5440e-04,  1.9726e-04, -2.7488e-04,  7.7893e-04,  4.2999e-04,
         1.0914e-03,  6.5259e-04, -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([-3.9764e-03, -9.8617e-01, -3.7329e-04,  7.1741e-05, -1.9111e-04,
        -3.6059e-04,  1.7434e-04,  1.6192e-04, -2.0062e-04,  4.9332e-04,
        -1.3878e-04,  6.3155e-05,  6.6604e-05, -1.5072e-05, -5.6387e-05,
         2.6819e-04, -9.1205e-05,  2.3034e-04,  5.6475e-04, -1.4645e-04,
        -1.2223e-04, -4.0301e-04,  7.3332e-05, -3.7823e-04, -5.8453e-05,
        -2.3217e-04,  4.3008e-04, -4.3506e-04, -3.6848e-04, -9.3011e-05,
        -3.7154e-04, -5.5860e-05,  1.3241e-04, -2.9289e-04, -1.6097e-04,
         4.8898e-04,  2.7131e-06,  3.3285e-04, -3.0629e-04, -1.2939e-04,
        -1.0046e-04, -4.9166e-04, -2.2748e-05,  4.4646e-04,  1.4551e-06,
         1.8622e-04,  6.5341e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6755e-03,  9.9221e-01, -1.3506e-05,  1.9666e-05, -1.0704e-04,
         3.9163e-05, -5.3624e-05, -5.7308e-05,  2.7368e-04, -2.3806e-04,
        -1.8898e-04, -1.9725e-04,  1.6439e-04,  1.5169e-05,  3.9614e-04,
         5.4060e-05,  5.5546e-05, -1.1099e-04,  1.9360e-04,  3.4039e-04,
        -9.3255e-05, -1.0436e-04, -2.3249e-04, -9.9267e-05, -3.6461e-05,
        -1.1279e-04,  9.4689e-05, -1.4861e-04,  2.2731e-04,  1.1524e-04,
        -3.7755e-04,  3.2128e-04, -1.3863e-04, -1.9236e-04,  3.4796e-05,
        -3.3742e-04, -5.2790e-05, -1.2016e-05, -1.7442e-04, -2.8276e-05,
        -4.8240e-05,  2.2971e-04,  2.3243e-04,  8.3267e-05, -2.2882e-06,
         6.3788e-05,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.2608e-04,  9.8982e-01,  6.1508e-04, -5.0739e-04,  1.2617e-03,
         5.8352e-04,  9.9217e-05,  4.0021e-04,  6.2297e-04,  5.2792e-04,
         1.3931e-04, -2.3208e-04,  5.5783e-05,  4.1882e-04, -3.4870e-04,
        -7.5575e-04, -1.8827e-04, -3.3987e-04,  1.0047e-03, -1.0849e-03,
         7.4238e-07,  4.6663e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9061e-03,  9.8191e-01, -6.9333e-05, -6.2188e-04, -8.2970e-04,
        -5.1107e-04,  2.2634e-04, -7.5393e-04, -2.4105e-04, -8.9779e-04,
         2.2073e-05, -1.4618e-03, -9.9413e-05, -2.7560e-04,  1.6303e-04,
        -1.3655e-03, -7.5431e-04, -4.6269e-04, -3.5656e-04,  2.2971e-04,
        -5.9267e-04,  2.7594e-04, -2.1891e-04, -2.0291e-04,  1.8916e-04,
        -6.3312e-04, -7.7493e-04, -3.5079e-04,  4.8591e-04, -1.1438e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([-2.9140e-03, -9.8782e-01, -2.8211e-04,  1.9535e-04,  6.5667e-04,
         8.9860e-04, -4.3475e-04,  6.1830e-04,  1.1328e-03,  3.1194e-04,
        -5.6193e-04,  1.0788e-03,  5.1728e-04,  7.3800e-04, -4.5795e-04,
         6.2458e-04,  7.5320e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1903e-04, -9.9104e-01, -9.9552e-04,  5.0964e-04,  4.5651e-04,
        -2.0586e-04, -6.4596e-04,  1.7457e-04,  2.8150e-04, -2.6727e-04,
         1.8055e-03, -4.0514e-04,  3.2595e-04, -4.4734e-04,  1.9838e-04,
         6.0354e-04,  3.5494e-04,  4.2991e-04,  4.5221e-04, -2.8075e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2624e-04,  9.9334e-01, -1.0930e-03,  3.0613e-04,  1.0004e-03,
        -1.1071e-03, -6.6651e-04, -3.6846e-04,  4.7460e-04,  1.8112e-04,
        -1.0763e-04,  1.8260e-04, -6.4583e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.9330e-04,  9.9074e-01, -2.2784e-04,  3.2249e-04,  3.0902e-05,
        -2.4393e-04, -1.3691e-03,  4.5148e-04, -8.7178e-04,  2.0023e-04,
        -1.5723e-04, -5.2174e-04, -8.9143e-04, -1.0179e-03, -3.9872e-04,
        -1.1093e-03, -5.4969e-04, -2.0233e-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,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7754e-03, -9.8640e-01, -7.9420e-04, -3.0097e-04,  1.2781e-04,
        -8.8691e-04, -4.9897e-04, -5.2302e-04, -5.6849e-04, -1.7818e-04,
         3.1290e-04, -4.8256e-04, -4.0503e-04,  3.6249e-04,  1.1186e-05,
        -8.1193e-04, -8.8762e-04, -8.3920e-04, -1.2864e-04, -1.3051e-04,
         2.5493e-04, -2.0787e-04, -6.1234e-04,  6.6244e-04, -1.9763e-04,
        -6.4163e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4691e-04,  9.8700e-01,  2.1041e-04,  1.4461e-03, -1.6760e-03,
         4.7191e-03,  9.0064e-05,  2.6806e-03,  1.6510e-03, -3.1643e-04,
        -6.2643e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3588e-04,  9.9269e-01, -2.8486e-03, -8.2800e-04, -6.5242e-04,
        -1.2572e-03, -5.9608e-04,  5.4312e-04, -2.8683e-04, -1.5818e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1965e-03, -9.8979e-01, -1.2753e-04, -4.9216e-04,  2.0634e-04,
        -5.2090e-05,  1.4657e-04, -1.6342e-04,  9.9251e-05, -1.4908e-06,
        -1.8406e-04, -8.1168e-05,  3.2689e-04,  3.8289e-04,  1.9996e-05,
         1.3491e-04, -2.8148e-04, -3.4498e-05,  1.4800e-04,  7.9184e-05,
         2.3607e-05,  6.8576e-05,  1.2078e-04, -5.5904e-05, -2.5261e-05,
        -2.2623e-05,  6.6371e-05, -2.2262e-04, -2.0093e-05,  6.4971e-05,
        -1.7600e-04,  1.2335e-04,  9.1088e-05,  2.3628e-04, -4.1037e-05,
         1.5790e-04, -2.2861e-04, -2.8011e-04, -1.2334e-04, -1.6126e-06,
         1.8110e-04,  2.3622e-04,  8.4139e-04,  6.5092e-05,  1.2298e-04,
        -8.8462e-05,  3.6810e-04, -1.7650e-04,  3.3059e-04,  2.0846e-06,
        -3.7313e-04,  5.0424e-04,  1.9710e-05,  4.3022e-04,  2.9549e-05,
         1.3280e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3261e-03, -9.9274e-01,  1.2273e-04, -3.3934e-04, -1.5563e-04,
         9.9605e-04, -5.2735e-04, -1.7128e-04, -7.6046e-05,  4.2580e-04,
         5.1108e-04,  6.6793e-04,  1.3657e-04,  3.6270e-04,  1.1250e-04,
        -2.5291e-04,  9.8362e-05,  4.3782e-04, -4.7386e-04,  6.4409e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1833e-03, -9.8656e-01, -4.1648e-04,  1.4213e-04,  6.5434e-04,
         5.3852e-04,  2.7807e-04, -1.7687e-04,  4.8048e-04, -2.0075e-05,
         7.1968e-04,  3.2610e-04,  1.2978e-03,  8.5720e-06,  1.1916e-03,
         4.6731e-04, -5.1333e-04,  4.3746e-04,  9.0559e-04,  3.4393e-04,
        -3.6141e-05, -6.9316e-04, -4.1588e-04,  2.0157e-04,  1.4324e-04,
         2.5297e-04,  1.0527e-04, -7.6693e-04, -4.0835e-04, -1.1847e-04,
         2.0019e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1059e-03,  9.8818e-01, -9.4381e-04, -1.2746e-03, -3.8376e-03,
         1.2088e-04, -2.5324e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0396e-03,  9.9088e-01,  5.9147e-04,  1.1094e-03, -1.3116e-03,
         2.1479e-04,  3.9378e-04,  2.3407e-03,  1.1801e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 1.3921e-03, -9.9130e-01, -1.7926e-04,  1.0321e-04,  2.4223e-04,
        -1.3491e-05,  1.9781e-04,  1.6032e-04,  9.4773e-04, -5.6415e-04,
        -3.3505e-05, -7.1028e-04,  1.8113e-04,  1.2127e-04,  1.8772e-04,
         3.1639e-05,  2.6025e-04, -3.7553e-04,  3.2626e-04, -4.1645e-05,
         5.6057e-04, -2.6899e-05,  3.6684e-05,  2.0065e-04, -4.9790e-04,
        -1.5755e-04, -7.5059e-04, -4.0339e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.0389e-06, -9.8907e-01, -8.1822e-04,  5.4108e-04, -3.9102e-04,
        -2.7370e-05,  7.3799e-04,  2.4958e-04,  2.7412e-04,  1.0645e-03,
         4.6323e-04,  1.0682e-05,  4.8593e-04, -1.7406e-04, -3.0378e-05,
         4.9498e-04,  6.4478e-04,  1.7693e-04,  4.2508e-04, -2.7467e-04,
        -1.2603e-04,  5.4363e-05,  3.4311e-04, -1.1714e-04, -2.4422e-04,
         8.4901e-04,  3.3014e-04,  3.7093e-04,  2.3005e-04, -2.7477e-04,
         2.6772e-04,  4.3208e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3689e-04, -9.8784e-01, -3.6624e-04,  4.6405e-04,  5.6434e-04,
         7.9595e-04, -2.3172e-04, -3.3397e-04, -6.4864e-04,  2.2661e-04,
         8.3994e-04, -3.4687e-05,  9.9669e-05, -2.9395e-04,  8.3929e-04,
        -2.4390e-04,  1.1578e-03,  5.6373e-04, -2.2870e-04, -7.6284e-04,
        -2.9020e-04,  9.0013e-04,  7.9997e-04,  1.7885e-04,  3.6013e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4290e-03,  9.8563e-01,  1.7924e-03, -7.2659e-04, -1.0979e-03,
        -2.1151e-03, -1.1170e-03,  3.0952e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8117e-03,  9.9194e-01, -6.4905e-04, -2.2861e-03, -5.9444e-04,
         7.7646e-04, -9.4415e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9836e-04, -9.9234e-01, -1.5856e-04, -1.3707e-03,  9.4256e-05,
        -7.4781e-04,  4.0724e-05,  1.0081e-03,  9.4864e-04,  4.7954e-05,
         1.6840e-04,  7.3113e-04, -2.8989e-04,  3.0036e-04,  2.2000e-04,
         1.5428e-04,  3.4164e-04, -1.1282e-04, -2.0571e-04,  2.1859e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0791e-03,  9.9086e-01, -5.7858e-04, -1.6365e-04,  8.7213e-05,
         1.6437e-04, -1.2507e-04,  5.7676e-05, -1.4668e-03, -4.6996e-04,
        -1.5238e-03, -1.0557e-03,  7.5434e-04,  6.1498e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0395e-03,  9.8936e-01,  1.0674e-03,  5.4205e-04, -8.5781e-04,
         4.1795e-03, -1.3682e-04,  1.7271e-05,  8.1622e-05, -1.7184e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2765e-04,  9.9236e-01,  7.0902e-04,  4.9699e-04, -3.9553e-04,
        -2.1665e-07, -1.4776e-04,  6.6220e-05,  4.1027e-04,  2.1993e-05,
         5.4763e-05, -1.6823e-04, -2.5221e-05, -8.2808e-04,  2.5697e-04,
         6.2057e-04,  5.0527e-04,  1.8834e-04,  5.6089e-04, -3.1638e-04,
         1.2002e-04, -1.8363e-04,  3.7677e-04, -1.6048e-04,  3.7689e-04,
        -3.8795e-04, -1.9321e-06, -3.0146e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8685e-04,  9.9529e-01,  8.5924e-04, -5.9704e-04, -1.1005e-05,
         3.1055e-04, -2.3968e-04, -1.7299e-04,  2.0849e-04,  4.0882e-05,
         8.0631e-06, -2.9420e-04,  1.9124e-04, -5.1439e-04,  5.4325e-04,
        -1.4869e-04, -1.8412e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4074e-03,  9.8812e-01, -1.1371e-03,  2.4856e-04, -1.0397e-03,
        -1.2450e-03, -2.0431e-04, -4.3714e-04, -1.0104e-03, -2.7192e-04,
        -4.8718e-04, -4.3736e-04, -1.1703e-03,  4.8243e-04, -2.8186e-05,
        -6.4941e-04, -1.3228e-03, -6.5744e-05, -2.3058e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4160e-03, -9.8804e-01, -6.0883e-04,  3.2174e-04, -2.4375e-04,
        -6.2045e-04, -1.2185e-03, -8.6271e-04,  6.9011e-04,  9.9241e-04,
        -3.9542e-05,  2.2639e-04,  4.1248e-05, -1.6784e-03, -3.6784e-04,
         1.2346e-03, -9.4988e-04, -4.5072e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+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([-9.5851e-04, -9.8988e-01,  2.5626e-04, -9.9232e-05,  6.4088e-05,
         3.0688e-04, -9.9870e-04,  1.5753e-04, -6.3143e-04, -5.9970e-05,
        -9.0093e-04, -4.8789e-05,  1.9948e-04, -7.6043e-05,  2.1428e-04,
        -1.2264e-03, -6.3938e-04,  5.5079e-04, -1.7905e-04, -5.4920e-04,
        -6.2664e-04, -1.7253e-04,  9.4956e-04,  1.7097e-05, -5.8859e-05,
         1.0793e-04,  6.8611e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.9977e-04,  9.8955e-01,  4.8742e-04, -2.7835e-05, -8.3136e-04,
         3.6840e-04,  2.6625e-04,  4.4288e-04, -2.3118e-04, -2.4919e-06,
        -1.2769e-05, -5.7867e-05, -1.0025e-04,  1.3691e-04, -2.7706e-04,
        -4.1855e-04,  3.8701e-05,  2.7419e-04, -7.3018e-05,  1.6922e-04,
         1.0794e-04,  1.9093e-04, -1.1632e-05, -4.5562e-04,  5.5459e-05,
        -6.6712e-06,  1.6543e-04, -1.0246e-04,  3.5009e-04,  2.0650e-04,
         3.2963e-05, -1.6124e-04, -4.1486e-04,  1.1285e-04,  4.9200e-04,
        -3.6585e-04,  4.5711e-04,  7.4822e-05,  3.4811e-04,  1.5811e-04,
         1.0669e-04,  3.4921e-04,  1.2254e-04, -2.4966e-04, -1.3417e-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([-1.7026e-05, -9.9040e-01,  3.9241e-04, -4.5552e-04, -1.2398e-04,
        -4.4252e-06,  7.1615e-04,  3.2866e-05, -6.0961e-04,  1.7335e-04,
        -3.6331e-04,  3.3735e-04, -9.6615e-04, -7.7603e-04, -2.2629e-04,
         1.1761e-05, -9.2453e-04,  4.4416e-04, -3.1415e-04,  4.0852e-04,
         7.7545e-05, -1.2749e-04,  3.4271e-05,  1.4704e-04, -3.8483e-04,
        -5.0699e-04,  6.2802e-04,  1.8346e-04,  2.1163e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0393e-04, -9.9308e-01,  1.4664e-03,  1.4304e-04, -5.1346e-04,
        -7.1570e-04, -1.3929e-03, -8.5826e-05, -1.0670e-03, -3.4809e-04,
        -5.0130e-05,  3.5780e-04, -5.7135e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.6534e-04, -9.9136e-01,  2.6980e-04,  1.9723e-04, -2.7482e-05,
        -1.1663e-04,  2.6344e-04,  5.9882e-04,  2.8227e-04, -5.1944e-05,
        -5.5834e-04,  2.5399e-04,  2.6443e-05,  2.7367e-04, -1.4506e-05,
         1.6157e-04,  6.6187e-04, -2.0956e-04, -4.5387e-05,  2.7455e-06,
        -8.8736e-05,  1.2032e-04, -3.8681e-04,  2.6179e-04, -2.6372e-04,
        -6.2680e-05,  2.2799e-04, -7.6736e-05, -1.0701e-04,  3.4990e-05,
        -4.2339e-04, -5.5175e-04, -3.8942e-04,  2.4348e-04, -2.1106e-04,
         4.4226e-04, -7.1608e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8117e-03, -9.8838e-01, -1.2263e-04, -3.4569e-04, -5.3264e-04,
        -3.2941e-04, -4.4324e-04, -3.4427e-04, -4.2704e-04,  4.2185e-04,
        -3.4528e-04,  4.4264e-04, -2.9672e-04,  5.7073e-04,  2.5927e-04,
         4.1818e-04, -2.3585e-04, -2.1719e-05,  2.8884e-04, -9.4581e-05,
        -3.5637e-04,  6.3460e-04,  3.9245e-05,  3.2915e-04, -4.4766e-06,
        -1.2072e-04, -2.4151e-04, -3.7607e-05,  2.4605e-05, -2.7343e-04,
        -1.2755e-04, -5.7511e-04, -2.3944e-04,  1.5879e-04, -7.0223e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7574e-03,  9.8332e-01,  1.0782e-03, -7.5621e-04,  7.2430e-04,
        -5.1017e-04,  3.9701e-05, -2.0121e-04, -2.7847e-04, -1.2967e-03,
         3.5312e-04, -1.0293e-04, -7.3066e-04,  1.3352e-03, -1.4869e-03,
        -5.3243e-04, -1.5851e-03,  1.7883e-03, -1.0625e-03,  3.8942e-05,
         2.1730e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4531e-04,  9.9067e-01,  5.3245e-04,  2.1097e-05,  6.2037e-04,
         7.1498e-05,  2.1410e-04,  1.6647e-05,  7.1111e-04, -3.3798e-05,
         4.9772e-05, -3.4494e-04,  2.0588e-07,  2.3300e-04, -3.5395e-04,
        -2.5378e-04, -3.0879e-04,  9.7629e-04, -2.4356e-04,  6.5557e-06,
        -6.9854e-04, -7.5562e-05,  5.0695e-04, -2.9257e-05, -1.0546e-04,
         1.8232e-04, -5.3336e-04,  3.4998e-04, -3.5857e-04,  7.7050e-04,
        -1.7937e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2866e-03, -9.8964e-01, -7.9379e-05, -7.4563e-05, -3.1850e-04,
        -2.4158e-04, -1.4511e-04, -3.8857e-04, -4.1023e-04, -5.2994e-04,
        -2.3636e-04, -4.8955e-04, -3.8982e-04, -1.3072e-04,  3.6996e-04,
        -3.1647e-04, -2.1230e-04, -4.8387e-04, -7.7846e-05, -1.0180e-04,
        -2.7917e-04, -5.1924e-05, -1.1378e-04,  2.6153e-04,  5.1619e-05,
        -2.6350e-04,  2.0457e-06,  1.3903e-04, -8.2368e-05, -1.2360e-04,
         1.3957e-05, -1.4175e-04,  2.1242e-05, -4.5496e-04, -1.5222e-04,
         2.4092e-04, -4.9983e-05,  2.7837e-04,  3.6839e-04, -3.1520e-04,
         4.3272e-04, -2.3553e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5921e-03,  9.8174e-01,  1.7715e-03,  2.4957e-03, -4.2195e-04,
        -2.3575e-04,  3.2602e-04, -1.3090e-03, -2.2753e-03, -3.5509e-04,
         2.1418e-03,  7.9654e-04, -5.3752e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1338e-04, -9.8704e-01, -4.9326e-04,  9.4219e-05,  2.0980e-04,
         5.1145e-04, -5.4273e-04, -3.1506e-04, -3.8567e-04,  1.6436e-04,
        -5.2496e-04, -2.2066e-04,  2.2248e-05, -3.6508e-04,  1.6327e-04,
        -1.4370e-04,  6.6600e-04, -1.2504e-04,  1.8555e-04, -1.5060e-04,
        -7.2971e-06, -3.4246e-04, -7.6784e-04,  7.8932e-05,  1.7782e-04,
         3.2586e-04, -2.1158e-04, -3.3772e-04,  3.4114e-05, -4.7611e-04,
        -2.5349e-04, -1.7967e-04,  2.2065e-04,  1.5463e-04,  1.7877e-05,
        -4.5591e-05,  1.1239e-05, -1.0661e-04,  3.9336e-04, -2.6296e-04,
        -2.1178e-04,  3.1909e-05,  5.1748e-05, -4.5107e-04, -5.6175e-04,
         3.1147e-04,  1.0652e-04, -7.1261e-05,  3.7026e-04,  8.6540e-05,
         1.3319e-04,  3.6818e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4866e-03,  9.8858e-01,  8.1740e-04,  3.0416e-04,  3.0005e-04,
        -4.8633e-04,  6.8302e-04,  5.1911e-04, -7.0752e-04,  1.2585e-03,
        -7.2743e-04,  1.3103e-03,  2.4576e-03,  3.5734e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+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([ 7.7094e-05, -9.9146e-01, -1.4155e-04, -2.3576e-04,  4.6726e-04,
        -2.2515e-05,  1.5514e-05, -3.6502e-04,  2.5304e-04, -3.9660e-04,
        -1.8556e-04,  4.1059e-04,  1.0498e-04,  4.1818e-04,  5.5695e-04,
         1.9942e-04,  2.0522e-05, -3.0549e-05, -6.3801e-05, -2.9445e-05,
         3.4964e-04,  1.2392e-04,  3.6205e-05, -2.3083e-04, -2.4310e-04,
        -2.7636e-04,  1.6652e-04,  2.4581e-04, -2.3937e-04,  3.7449e-04,
        -7.4669e-04,  1.1531e-05,  6.8595e-05, -2.2730e-05,  1.6532e-05,
        -3.9176e-04, -2.1038e-04, -7.9580e-05,  3.9435e-04, -3.1799e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1366e-03, -9.8566e-01,  6.4633e-05,  5.6098e-05,  2.9469e-05,
         4.1085e-04, -2.1489e-04, -1.9717e-04, -2.1462e-04,  1.1441e-04,
        -4.9543e-05, -3.2718e-04,  2.4567e-04, -2.6101e-04, -4.9563e-04,
         1.9391e-04, -5.2360e-04,  3.9146e-04, -1.0236e-04, -2.2802e-04,
        -4.4544e-04, -2.5356e-04, -4.7393e-04, -2.9166e-04,  4.2097e-04,
        -7.6504e-05,  1.0089e-04, -4.9610e-05, -2.7622e-05, -9.7056e-05,
        -1.1112e-04,  1.3380e-04, -1.2400e-04,  4.3310e-04,  3.7332e-05,
        -1.0146e-04,  3.0316e-04,  1.9496e-04,  3.3279e-04,  3.9482e-04,
         5.6200e-05, -1.6824e-04, -2.5421e-04, -9.6113e-05,  1.5942e-04,
         3.3052e-05, -3.0545e-04, -4.2187e-06, -4.2700e-04,  1.1011e-04,
        -1.5107e-05, -3.5822e-04, -1.3607e-04,  3.7244e-04, -1.4730e-04,
         3.8194e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9970e-04, -9.9490e-01,  1.1370e-04,  1.7493e-05, -1.4314e-04,
        -2.5360e-05,  1.2805e-04,  6.1451e-05, -1.4024e-04, -5.9401e-05,
        -1.5476e-04, -2.9152e-04,  1.1012e-04,  3.7502e-05, -7.2400e-05,
        -5.6176e-05,  4.6640e-05, -3.9387e-04,  2.5667e-04, -1.8611e-04,
        -6.5013e-06,  3.4885e-04,  3.9564e-04, -2.9534e-04,  6.9931e-05,
         9.2000e-05,  2.3088e-04,  2.1543e-04, -4.9882e-05,  1.7215e-05,
         1.2851e-04,  4.0996e-04,  1.1029e-04,  3.2995e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1101e-03,  9.9211e-01, -6.6378e-04, -1.1910e-04, -1.9537e-03,
        -9.8581e-04, -6.0440e-04, -8.1831e-06, -6.1741e-04, -4.3397e-04,
        -4.8025e-05,  2.1622e-05,  3.2247e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5225e-03, -9.8672e-01, -1.1410e-04,  2.6329e-04, -4.6038e-04,
         9.6580e-04,  7.5687e-04,  2.4523e-04,  9.4783e-04,  8.8680e-05,
        -1.9824e-04, -6.8854e-04,  5.5925e-04, -1.9327e-05,  1.8920e-05,
         1.0369e-04, -1.8299e-04, -5.4106e-04,  1.1139e-03, -1.9254e-04,
         6.0365e-05,  5.0880e-04, -4.6141e-05,  7.2950e-04, -4.4799e-04,
         4.8610e-04,  1.5350e-04,  1.9658e-04, -2.8851e-04,  3.7816e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8835e-03, -9.8808e-01, -4.4053e-04,  2.7560e-04, -1.3738e-04,
         6.8363e-04,  3.3629e-04,  1.0803e-04, -3.3458e-04,  2.0148e-04,
         9.0109e-04, -3.8022e-04, -1.2716e-04,  2.6731e-04, -7.1785e-05,
        -4.6294e-04,  1.2374e-04, -3.5986e-04, -6.3161e-04,  2.9958e-04,
         3.7209e-04, -5.9482e-04, -8.8901e-05,  3.2602e-04, -2.4206e-04,
         1.9103e-04, -7.1805e-04, -8.1040e-04, -2.2603e-04, -3.1928e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2122e-03,  9.8729e-01,  2.7626e-04, -6.2451e-04, -3.6615e-04,
        -3.3916e-04,  3.4351e-04,  5.5663e-04, -5.6745e-04, -6.2727e-04,
        -9.5009e-04, -5.3458e-04,  1.8157e-04, -5.0482e-04,  3.0418e-04,
        -3.7908e-04, -5.0612e-04,  6.1104e-05, -1.3604e-04,  8.1134e-06,
        -5.4547e-04, -5.7471e-04,  2.3346e-04, -3.6883e-05,  4.6347e-04,
         7.1738e-04, -1.0370e-03, -8.4227e-05,  5.4058e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.8759e-04, -9.9220e-01,  4.7097e-04,  2.4982e-04,  7.3963e-04,
         6.1607e-04, -5.5920e-04, -7.8267e-04,  1.5468e-03, -1.4423e-03,
         1.3830e-04,  2.5726e-04,  1.1271e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7312e-03,  9.8864e-01,  5.4866e-04, -7.5207e-04,  1.2592e-04,
        -5.1106e-04, -2.8713e-04,  2.9444e-05, -9.2270e-05,  9.8963e-05,
        -6.1128e-05,  3.0929e-05, -4.2856e-04, -8.1215e-04, -1.1260e-03,
        -5.5006e-04,  3.1709e-04,  9.0067e-04, -1.1361e-03, -3.4784e-04,
        -1.0708e-03, -3.9994e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3390e-03, -9.8998e-01,  5.9109e-04,  1.6708e-04,  1.0544e-04,
         8.2761e-05,  5.7192e-04,  3.4918e-04,  3.3228e-04, -2.5697e-04,
         5.9199e-04, -2.9971e-04,  4.2814e-04,  6.1573e-04, -1.9528e-04,
         5.2750e-04,  5.4054e-04, -1.3124e-04,  3.5308e-04, -3.3577e-04,
         5.2950e-05, -1.4517e-04, -2.0718e-04, -1.7833e-04,  5.1869e-04,
         9.0276e-05,  2.7951e-04,  3.0698e-04,  4.2729e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8568e-03, -9.8842e-01,  1.4707e-04,  1.0244e-04, -3.7423e-04,
        -1.5669e-05, -3.1038e-04,  2.4665e-04,  7.4549e-04, -1.1919e-04,
         5.8308e-04,  3.9545e-04,  4.3533e-04, -2.5235e-04,  2.7074e-04,
         4.0943e-04,  1.3833e-04,  1.5109e-04,  1.1591e-04,  3.2927e-04,
         3.1738e-04,  7.8920e-04,  2.7194e-04,  4.0020e-04,  3.8532e-04,
        -4.8038e-04, -4.7033e-04,  4.0683e-05,  4.2635e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8300e-03,  9.8419e-01, -8.9656e-04, -9.0147e-04,  1.8398e-04,
        -6.2889e-05,  1.6475e-05, -1.5545e-03, -1.6114e-03, -1.4839e-03,
        -2.7870e-04, -6.7518e-04, -5.8890e-04, -7.2198e-04, -8.0993e-04,
        -3.1786e-04,  8.1416e-04, -1.7282e-03, -3.3824e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 3.2428e-04, -9.9307e-01, -1.7861e-04, -4.1946e-05,  1.3255e-04,
        -2.5057e-05,  1.5075e-04, -1.1901e-04, -8.5530e-05, -2.3978e-04,
         8.6886e-05, -4.9365e-05, -1.0652e-05,  1.7935e-04, -1.7076e-04,
         6.4380e-05,  9.1008e-05, -1.3320e-04, -5.8904e-04, -7.3580e-04,
        -1.3743e-04, -1.4985e-04, -1.6881e-04, -3.7883e-04, -8.1957e-04,
         1.7237e-04,  8.5146e-05,  2.7559e-04,  1.4365e-04, -8.4267e-05,
        -1.4252e-04,  3.2453e-04, -1.2330e-06, -2.2242e-04, -1.8444e-04,
         1.8562e-04,  4.5809e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2688e-03,  9.8550e-01,  1.4150e-04,  8.5098e-04, -3.7076e-04,
        -2.7650e-04, -2.9219e-04,  2.5014e-06, -1.7439e-04,  1.3051e-03,
         1.0186e-03, -4.5222e-04, -1.9395e-04, -2.5491e-04,  4.3498e-05,
         3.7789e-04,  7.5148e-04,  9.0006e-04,  2.7682e-04,  7.6916e-04,
        -4.1008e-04, -2.4091e-04,  1.7438e-04, -2.5871e-04, -3.8981e-04,
        -1.9590e-04, -4.5191e-04,  1.5832e-04,  5.9649e-04,  4.6303e-04,
        -1.7656e-05,  3.3064e-04, -9.0861e-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([-1.2915e-03,  9.8747e-01,  6.0135e-04,  7.1970e-04,  8.4396e-04,
        -3.0250e-04, -1.7920e-04,  9.3920e-04,  3.6863e-04, -4.2036e-04,
         2.0442e-04,  4.1716e-04,  5.8167e-04,  2.2536e-04, -1.9669e-04,
        -5.4890e-04, -1.4323e-04,  1.3209e-03,  1.4161e-03,  1.3551e-03,
         2.1056e-04,  2.4127e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8175e-03, -9.9106e-01,  5.2417e-04, -1.2620e-04, -4.6779e-05,
         7.2663e-05, -8.0755e-05, -3.9079e-04, -1.3126e-04, -9.5308e-05,
         4.2268e-04,  5.1664e-04,  5.7721e-04, -2.2981e-05,  1.1175e-05,
         3.5524e-05, -6.4921e-05, -1.6524e-04, -6.4936e-04,  7.9899e-05,
         4.7096e-04,  1.2066e-04, -1.4135e-04, -8.6788e-05, -2.8758e-04,
         5.6251e-04, -1.3379e-04,  4.8950e-04,  3.2606e-05, -2.6095e-04,
         2.5632e-04,  3.2841e-07, -2.1917e-05, -2.4473e-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([ 1.0007e-03,  9.9376e-01, -8.0131e-05,  5.3985e-04, -2.8063e-04,
         5.2552e-04,  6.0325e-04,  1.8441e-04, -1.1695e-05,  2.9606e-04,
        -2.4500e-04, -9.0189e-04,  5.7782e-04, -7.7158e-04,  2.2309e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3145e-03,  9.8763e-01,  6.6429e-05, -7.0713e-04, -1.9633e-04,
        -6.4405e-04, -4.6568e-04,  2.6065e-04, -8.2137e-04,  7.0343e-05,
        -9.7558e-04,  3.4200e-04, -1.7676e-03, -4.7457e-04, -9.3049e-04,
        -5.2922e-04, -5.4035e-04, -8.3318e-04,  5.9007e-04, -4.9616e-04,
        -3.4075e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2589e-03, -9.9203e-01,  1.1560e-04,  1.3284e-05,  7.9852e-05,
        -2.5240e-04, -1.0917e-04,  6.2845e-05, -7.9281e-05, -1.6805e-05,
         3.3323e-04, -4.3112e-04, -3.7160e-04, -3.3839e-05,  3.2515e-04,
        -1.0475e-04,  4.4964e-06,  6.5064e-05,  6.6883e-05,  4.1764e-05,
         2.9340e-05, -2.3294e-04, -2.9155e-04, -5.1526e-04, -1.1512e-04,
        -2.8895e-04,  1.4534e-04,  4.5101e-05,  7.6428e-05, -2.5987e-04,
        -1.9367e-04,  4.2919e-04,  6.5574e-04, -2.2708e-05,  7.2063e-05,
        -1.9439e-04, -3.3344e-04,  2.9187e-05,  1.8064e-04,  9.5905e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7948e-03,  9.9320e-01,  2.0860e-04, -3.9278e-05,  3.3069e-04,
        -3.5902e-04,  5.3258e-04, -4.3786e-04,  1.0217e-04,  4.8057e-05,
         6.6807e-04,  9.1556e-05,  2.3121e-04, -6.5729e-05, -8.7746e-05,
        -5.7742e-04, -1.4437e-05, -5.5734e-04, -4.6346e-04,  6.3516e-05,
         1.2440e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7262e-03,  9.9138e-01,  8.5993e-04,  3.3533e-05,  5.4719e-04,
         7.4691e-05,  1.0673e-04,  8.5191e-04,  2.2990e-05,  1.6289e-04,
         1.1453e-03, -5.3228e-04, -5.4647e-05,  7.0296e-04, -6.9969e-04,
        -1.0757e-05,  9.2025e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3097e-03, -9.8732e-01, -7.1375e-04, -1.6904e-04, -3.6485e-04,
         6.5326e-04,  5.1734e-04,  7.8514e-05, -1.8457e-04, -1.9046e-05,
         5.8781e-04, -2.6534e-06,  1.5663e-04,  2.2075e-04,  6.9792e-04,
         7.7262e-04,  1.1405e-03, -6.1357e-04,  2.4519e-04, -9.1094e-04,
         1.3292e-04,  2.8038e-04, -1.4256e-04, -1.8446e-05,  1.6612e-04,
         5.8384e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9235e-03,  9.9223e-01,  1.0995e-04, -1.2394e-03, -7.8799e-04,
        -1.6154e-03,  4.8031e-04,  1.0280e-04, -4.9887e-04,  1.8613e-04,
        -2.0036e-04,  3.2757e-04,  1.4904e-04,  1.4803e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.5696e-04, -9.8971e-01,  6.9932e-04, -6.4723e-04, -3.6961e-04,
         1.2988e-04, -5.5939e-04, -4.2219e-04, -2.2758e-04, -1.9993e-04,
        -2.0720e-04, -6.3187e-05,  5.8640e-05, -1.6354e-04, -5.3234e-05,
        -5.0104e-04, -5.4435e-04,  1.6536e-04, -4.5470e-04, -5.2548e-04,
         3.9522e-04,  5.0681e-04,  3.1649e-04, -1.6951e-04,  7.9419e-04,
        -4.0248e-04,  3.0509e-04, -6.4676e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([-1.4681e-03,  9.9499e-01,  9.5821e-04, -2.1707e-04, -5.6146e-04,
         3.6828e-05,  1.5234e-03,  2.0729e-04, -3.2894e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.3014e-04, -9.9292e-01, -1.1438e-04, -1.4242e-04,  3.0333e-04,
        -9.6358e-05, -9.8026e-04, -9.8430e-05, -1.9861e-04, -5.0359e-04,
         5.0780e-04, -8.0916e-04,  9.9136e-05,  2.5972e-05, -4.8206e-04,
         4.2232e-04, -4.1144e-05, -5.4148e-04, -4.9760e-04, -2.6657e-04,
         5.2148e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6468e-03, -9.8805e-01,  1.4388e-04, -5.8850e-04,  1.1997e-04,
        -4.3991e-05, -4.7173e-04, -9.0113e-04,  7.3683e-05, -5.5399e-04,
         6.4128e-04,  2.1960e-05,  7.2438e-04,  2.8953e-04, -4.8153e-04,
         7.7931e-04,  7.6042e-04,  1.7734e-04,  4.0334e-04, -1.2747e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4053e-04,  9.9338e-01, -6.2140e-04, -2.0932e-04, -2.6561e-04,
        -2.1526e-04,  6.0928e-04,  2.0604e-04,  1.4616e-03, -3.5203e-04,
        -4.0762e-05,  3.9169e-04,  9.3990e-04, -1.9914e-04, -2.2054e-04,
         4.4749e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9822e-04,  9.9300e-01, -2.2554e-04, -6.3695e-04,  5.9943e-04,
        -9.4803e-04, -6.4115e-04, -3.7568e-05,  4.0726e-04,  6.0886e-05,
         2.9697e-04, -1.0771e-04, -5.7299e-04, -1.5120e-04, -1.5211e-04,
        -7.1021e-04, -6.8458e-04,  4.7077e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.1520e-04,  9.8947e-01, -1.7941e-04, -1.4264e-03, -4.6165e-04,
         9.3019e-04, -5.8940e-04, -1.3751e-03,  2.3361e-04, -8.8156e-04,
        -3.3482e-04,  5.5749e-04, -8.7582e-05,  1.0383e-03, -8.3433e-05,
         1.1744e-03, -1.8367e-04, -7.4601e-04,  2.5383e-05, -8.2072e-06,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6319e-03, -9.8766e-01,  1.9781e-05, -1.0921e-03, -6.0450e-04,
        -5.1042e-04,  7.9896e-04,  1.9844e-04, -8.4006e-04,  5.1814e-04,
        -5.4768e-04, -1.3882e-04,  1.6177e-04, -3.0641e-04, -5.7416e-04,
        -1.4817e-04, -3.9600e-04, -7.3552e-05, -3.0267e-04,  7.4092e-04,
         7.9957e-04, -7.1415e-04, -2.2388e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.2751e-04,  9.8648e-01,  5.2276e-04, -1.0288e-04,  1.9970e-04,
        -3.0288e-04,  3.0006e-04, -4.2852e-04,  1.9049e-04, -3.8621e-04,
        -1.1422e-03, -3.1197e-04, -8.2667e-04, -8.5622e-06,  9.4217e-05,
        -1.3520e-03,  8.1294e-04, -4.2285e-04,  4.3275e-05,  2.8162e-05,
         7.7486e-04, -5.8138e-04,  6.8621e-05, -3.3138e-04, -4.1719e-04,
        -4.4833e-04, -4.1558e-04, -2.9176e-04, -1.1802e-03,  5.8921e-04,
        -2.1295e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4556e-03, -9.8926e-01, -4.6671e-04, -5.8445e-04,  1.7476e-04,
        -1.3082e-04,  4.7363e-04, -5.2637e-04,  2.6110e-04,  4.6637e-04,
         7.2062e-06, -3.4498e-04,  7.3567e-05, -3.6263e-04, -9.5423e-04,
         1.8060e-04, -5.2822e-04,  2.6182e-04,  2.7850e-04, -2.8228e-04,
        -7.8219e-04,  5.2456e-04, -4.2783e-04, -8.4789e-05,  3.0665e-04,
        -6.8707e-05, -7.1223e-04,  1.4824e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1977e-06, -9.9369e-01,  1.1980e-03, -2.8840e-04, -4.4944e-04,
         4.0847e-04,  1.3007e-03, -1.1776e-03,  5.7532e-04, -1.2278e-04,
         7.8995e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3309e-03, -9.8689e-01,  7.3107e-04,  1.2115e-04, -1.6923e-04,
        -5.5457e-04, -2.1919e-04,  4.6694e-04,  4.7526e-04,  4.0498e-04,
        -9.3425e-04,  1.5993e-04,  3.6180e-04,  7.5431e-05,  3.5350e-05,
         1.6073e-04, -7.3758e-04,  6.9665e-05,  9.2583e-04,  1.0866e-04,
        -1.0215e-03,  4.5795e-04,  4.0005e-04,  5.0358e-05,  6.8361e-04,
        -2.8143e-04,  1.2457e-04,  2.6517e-05, -1.0252e-03,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0591e-04, -9.9111e-01, -4.6044e-04, -9.1744e-04, -8.1918e-05,
        -5.5904e-04,  1.7574e-04,  3.1461e-04, -3.1532e-04,  1.4789e-04,
        -5.8304e-05,  6.3029e-04, -6.0398e-04,  1.7411e-04,  1.9951e-04,
        -1.2906e-04,  5.1266e-04, -1.5136e-04,  6.3695e-04,  1.1482e-03,
         6.4443e-04,  2.2135e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+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([-7.9404e-04,  9.8889e-01, -1.9637e-03,  2.6099e-04, -2.7085e-04,
        -5.5064e-04,  4.2611e-04,  2.6782e-03, -1.3585e-03,  2.2634e-04,
        -1.6152e-04, -1.2692e-03, -1.0552e-03,  9.6740e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.9208e-04, -9.9195e-01,  9.7924e-05,  1.9403e-04, -4.1181e-04,
         5.7283e-05, -2.5141e-04, -6.8052e-04,  6.4311e-04, -1.5232e-05,
        -6.8257e-04, -2.0221e-04, -4.5470e-05, -5.4464e-05, -2.6210e-04,
         6.1576e-04,  2.7021e-04, -4.5024e-04, -2.7955e-04,  4.8235e-04,
         1.6716e-04, -7.9324e-04, -1.7547e-04, -5.1185e-04, -1.1024e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3833e-03,  9.8425e-01, -3.3503e-04, -4.4825e-04,  9.0596e-05,
        -6.3005e-04,  1.2800e-05,  3.1262e-04,  4.0550e-04, -5.9928e-04,
        -4.2489e-04, -1.2589e-04, -8.0783e-04,  5.1477e-05, -7.5741e-04,
        -9.9395e-04, -1.3049e-03,  5.9186e-04, -2.3948e-04, -3.9558e-04,
        -1.6697e-05, -7.3927e-04,  2.7409e-04, -3.8550e-04,  1.0483e-04,
        -7.4394e-04, -4.1721e-04, -3.1966e-04,  8.4047e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1991e-03,  9.8655e-01,  1.7721e-03,  1.8684e-03,  1.2770e-03,
        -1.9624e-03,  1.3608e-03,  1.7587e-03, -6.3375e-04,  2.2059e-04,
         9.3454e-04,  4.5968e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4754e-03,  9.8961e-01,  1.5076e-04, -1.0022e-03, -6.9466e-04,
        -4.0576e-04,  1.1237e-04, -6.5877e-04,  1.3062e-05,  4.5697e-04,
         6.6806e-04,  4.4916e-04, -1.2997e-04, -3.7473e-04, -1.8409e-03,
         9.5505e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0160e-05,  9.9080e-01, -8.4579e-05, -4.5251e-04,  1.7352e-04,
         6.3639e-04, -5.6710e-04,  1.4860e-04, -5.6563e-04,  3.8561e-04,
        -6.3032e-04, -2.0209e-04,  8.3483e-04, -6.1937e-05,  1.8673e-04,
        -7.9971e-05,  3.4639e-04, -6.0564e-04,  5.3417e-05, -4.3768e-04,
        -7.5225e-04, -1.1179e-04, -5.9566e-04,  2.3041e-04,  5.2311e-04,
         1.4829e-04,  3.7290e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.5202e-04, -9.8781e-01, -7.6474e-04,  2.5911e-04, -3.5824e-04,
        -4.7480e-04,  7.9773e-04, -9.9178e-04, -6.6802e-04, -3.6328e-04,
        -5.8854e-04, -1.0504e-03, -2.1791e-05,  5.5422e-05, -2.8990e-05,
        -5.6020e-04, -2.9054e-04, -8.6549e-04, -1.8696e-04, -1.0389e-03,
        -2.7204e-04, -1.2561e-04, -8.8563e-04, -6.9036e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7659e-04, -9.9206e-01,  2.2391e-04,  1.0316e-04,  2.7214e-05,
         1.6266e-04,  5.3509e-04,  2.7405e-04,  2.0388e-04,  1.4192e-04,
         7.7705e-04, -3.8233e-04,  5.2863e-04, -1.0093e-04,  2.7844e-04,
         1.6083e-04, -2.9592e-04,  6.1724e-05, -2.4385e-04, -1.8837e-04,
         2.9012e-04, -1.4505e-04, -3.1290e-04, -1.4784e-04,  2.6521e-04,
         1.0039e-04, -1.2996e-04, -6.1039e-05, -2.6047e-04,  2.9866e-04,
         3.2551e-04,  3.0841e-04,  2.0928e-05,  1.0565e-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([ 4.6824e-04, -9.9346e-01,  1.0391e-04, -2.7143e-04,  7.0349e-05,
        -6.5181e-06, -1.0933e-04,  3.8506e-04, -1.6065e-04,  4.8168e-04,
        -1.8312e-04, -8.1403e-05, -4.2372e-04,  9.9965e-05, -4.0496e-04,
         1.6257e-04, -4.6516e-04,  3.1776e-04, -1.6081e-04,  1.8069e-04,
         6.5795e-04, -9.2088e-05,  2.2770e-04,  9.3735e-05, -1.3408e-04,
         6.4296e-05,  1.6599e-04, -1.4521e-04,  2.9086e-04,  1.3371e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8800e-04,  9.9221e-01,  2.3017e-04,  2.7551e-05, -2.6485e-04,
         4.0823e-04, -1.6413e-04, -3.7104e-04,  5.8958e-04, -5.6288e-04,
        -5.4428e-05,  3.4786e-04, -2.4727e-04, -8.5627e-05, -2.5631e-05,
        -1.0388e-03, -1.1122e-03, -2.1057e-04,  2.3675e-04,  5.5011e-04,
        -2.3448e-04,  4.0736e-04, -3.3546e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.3736e-04, -9.8833e-01, -5.1776e-04, -9.0231e-04, -2.2463e-04,
        -9.1582e-05,  4.1967e-06, -2.5892e-04,  2.4701e-04, -8.8438e-04,
         1.7987e-04,  1.2847e-04,  1.8275e-04,  1.1735e-04,  4.3850e-04,
         1.5009e-04, -2.6048e-04, -6.8117e-04,  9.6188e-05,  5.2260e-04,
         9.5619e-04,  5.3333e-04, -1.1643e-03, -1.2167e-04, -2.7823e-04,
        -4.7257e-04,  1.4251e-04,  4.3459e-04, -2.3335e-04, -1.9492e-04,
        -5.1432e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2280e-03, -9.9203e-01,  1.1302e-04, -2.7152e-04,  1.7714e-05,
         1.0620e-04,  3.1067e-04, -2.8911e-04, -1.0575e-04,  3.2535e-04,
         1.9088e-04,  1.9052e-04, -3.1911e-05, -2.5338e-04, -1.3792e-04,
         3.0767e-04,  1.7141e-05, -1.7582e-05,  1.2439e-04,  1.5702e-04,
        -6.8082e-05, -3.3622e-04, -1.3912e-04, -4.7455e-04, -7.7604e-04,
        -5.5189e-06, -1.6129e-05,  4.7354e-05,  4.4099e-04, -3.8197e-05,
        -7.3438e-05,  2.1842e-04, -1.4915e-04,  1.2915e-04, -7.4326e-05,
        -2.5050e-04,  2.6629e-04, -1.5474e-04, -1.1800e-04, -1.5152e-06],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 2.2332e-04,  9.9205e-01, -7.2997e-04,  7.4166e-04,  9.6987e-05,
        -7.8556e-05, -1.6656e-04,  3.6384e-04, -2.6266e-04, -4.7753e-04,
        -3.5430e-04,  6.9260e-04, -1.0557e-03,  4.4548e-04,  6.6385e-04,
        -1.0054e-04,  8.4421e-04, -3.7919e-05, -1.5256e-04,  4.6394e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6034e-03,  9.9283e-01, -3.9458e-04,  1.1338e-03, -9.1576e-04,
         1.9466e-03, -6.4713e-04, -5.2393e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6796e-04,  9.9258e-01,  9.6386e-04,  7.7082e-04, -7.4425e-04,
        -7.4946e-04, -1.4196e-03,  7.9966e-05,  1.7196e-03, -6.0053e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2726e-03, -9.9133e-01, -3.8177e-04, -3.3155e-04,  8.0120e-05,
        -4.4790e-04,  2.7724e-04,  1.9686e-04, -2.2280e-04, -2.0497e-04,
        -2.4144e-04,  4.5383e-04, -2.3629e-04, -3.3118e-04, -1.6200e-04,
        -6.6571e-06, -1.5132e-04, -1.7865e-04, -1.3901e-04,  3.5693e-04,
        -1.4078e-04, -7.6545e-04, -1.1204e-04, -2.0722e-04, -1.6552e-04,
        -1.1492e-04,  5.4506e-04,  3.2558e-04,  3.1526e-04, -3.0790e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.1725e-04, -9.9253e-01,  1.4680e-04, -2.5416e-04,  1.1435e-05,
        -9.9062e-04, -3.5339e-05,  4.8051e-04, -1.3637e-04, -1.0899e-04,
        -3.6732e-04, -7.7350e-05, -1.5353e-04, -4.7422e-04, -9.6070e-05,
        -9.8667e-04,  2.4891e-04,  5.5285e-04,  5.7053e-04, -4.0832e-04,
         5.1414e-04,  6.8106e-05, -4.5263e-04,  2.2056e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7687e-09, -9.9067e-01, -5.7321e-04, -2.7739e-04, -2.5674e-04,
         7.0495e-04,  5.0527e-04,  4.0271e-05,  6.7796e-04, -7.3898e-04,
         3.9690e-04,  3.7340e-04, -1.7583e-04, -1.4336e-04,  2.8455e-04,
        -5.3352e-05, -2.7691e-04, -3.6851e-04,  1.0588e-03,  1.7206e-04,
         5.7775e-04,  1.8353e-04,  8.8960e-04, -2.0247e-04,  3.9513e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0423e-04, -9.8818e-01,  1.3455e-04,  8.6356e-04,  4.2448e-04,
        -3.3574e-04,  4.1401e-04, -1.2052e-05,  9.7985e-04, -2.3163e-04,
        -1.4342e-04,  4.9470e-04,  5.2101e-04,  3.2076e-04,  1.8057e-04,
         2.1750e-04,  5.7455e-04, -2.2691e-04, -9.4179e-05, -2.9919e-04,
        -1.1466e-04, -5.3225e-04,  4.7477e-05,  4.9477e-04, -5.9079e-06,
         2.2394e-04,  2.5042e-04, -1.8689e-04,  3.3224e-04,  6.3738e-04,
         7.8910e-04,  2.1514e-04,  3.8462e-04,  3.0678e-04,  6.6525e-05,
        -2.4352e-04,  1.4300e-04,  6.5262e-05, -1.1209e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6490e-03,  9.8880e-01, -5.9927e-04, -2.8009e-04,  1.3785e-03,
         1.5099e-04, -1.4040e-03,  1.7556e-03,  8.9647e-04,  1.2858e-04,
        -6.9429e-04,  2.6030e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.1845e-04, -9.9641e-01, -5.5801e-04, -1.9489e-04,  1.8879e-05,
        -6.0953e-06, -1.5974e-04,  3.4369e-04,  8.8099e-05, -9.0680e-04,
         8.9795e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9984e-03,  9.8762e-01, -1.2462e-03,  1.5111e-04,  5.9810e-04,
        -3.1552e-04, -5.4069e-04, -1.3696e-03,  1.4131e-04,  3.0134e-04,
        -6.5398e-04, -4.9170e-04, -2.9425e-04, -5.0363e-04,  1.8119e-04,
        -3.5044e-04,  8.9574e-05, -1.4058e-04, -1.0737e-03, -1.5199e-04,
         3.3068e-04, -1.6998e-04, -4.7318e-04,  2.9526e-04,  5.1641e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.0125e-04, -9.9199e-01,  6.3551e-04, -6.2612e-04, -4.8215e-04,
         3.0260e-04,  5.5655e-04,  3.3766e-04, -4.8345e-04,  2.1416e-04,
         5.3723e-04,  4.9661e-04,  3.9656e-05,  3.5092e-04, -4.8971e-04,
         7.9630e-05, -3.2470e-04,  2.6704e-04, -6.5488e-04, -1.9441e-04,
        -2.6941e-04,  2.6586e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9346e-03,  9.8701e-01, -5.6861e-05,  2.9603e-04,  7.0369e-04,
        -3.6125e-04,  4.1590e-05, -9.4000e-05, -8.2924e-04,  3.2167e-04,
        -2.5767e-04,  2.5768e-04, -4.1860e-04,  7.9901e-05, -3.3380e-04,
        -5.8435e-04, -6.9927e-04,  9.8394e-05, -5.1151e-04,  9.2245e-05,
        -4.2092e-04, -2.4103e-04, -2.6906e-04, -3.3874e-05, -8.0805e-05,
         2.3712e-04,  3.5503e-04,  4.8079e-04,  1.1662e-03, -3.5867e-04,
        -3.7594e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-3.1807e-04, -9.9386e-01,  1.9488e-04,  1.2731e-05,  1.0117e-04,
         2.2670e-05, -6.8718e-05, -8.0400e-06,  3.8637e-05,  5.6358e-05,
         9.3628e-05,  4.4035e-06, -7.9928e-06,  1.1676e-04,  5.9350e-05,
        -1.4194e-04,  6.9697e-06,  1.8022e-04,  7.5064e-05,  9.9392e-05,
        -8.1971e-05, -4.4391e-05,  1.8005e-05, -3.2241e-04,  2.3343e-04,
         4.7739e-05,  1.6526e-04,  2.2572e-06, -7.3373e-05,  2.9071e-04,
        -2.0112e-05,  1.2558e-04,  1.0732e-04, -9.9959e-05,  1.0138e-05,
         7.0099e-05,  3.2555e-05,  8.6706e-05, -1.3891e-04, -1.8447e-04,
        -5.5854e-05, -1.7342e-04, -3.1077e-04,  3.2647e-04, -1.6567e-04,
        -6.7328e-05, -9.3156e-05, -1.5440e-04,  3.2436e-05, -1.4608e-04,
        -1.0160e-04,  1.7247e-04,  5.0414e-05,  6.6828e-05,  6.1043e-05,
         2.4286e-04, -1.6096e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3829e-04,  9.9380e-01,  8.0106e-05,  4.4002e-04, -1.9640e-03,
        -4.5883e-04, -7.9294e-04, -1.9808e-04,  1.2377e-03,  6.9166e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4993e-03,  9.8668e-01, -7.4270e-04,  6.0254e-05, -4.4756e-04,
         4.0587e-04, -3.1254e-05,  5.4464e-04,  5.4708e-04,  9.4736e-04,
         2.7127e-04, -7.7356e-04,  1.0403e-03, -2.9577e-04, -7.8786e-04,
        -4.2282e-04, -3.4991e-04, -3.1568e-04,  6.8107e-04, -5.4329e-04,
         6.7397e-04,  5.0147e-04,  5.6224e-04, -3.4744e-04, -2.7226e-04,
        -2.5990e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4405e-04,  9.9462e-01,  1.0704e-05, -2.1134e-04, -6.0569e-04,
         1.2060e-03, -1.9892e-04,  6.1104e-05, -3.9417e-04, -8.4379e-04,
        -8.6070e-04, -4.4711e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.6761e-04, -9.9312e-01,  2.1673e-04, -1.3981e-04, -5.7592e-04,
        -7.1085e-04,  1.6868e-04, -4.8149e-04, -1.3477e-04, -1.7715e-05,
         7.9027e-04,  5.1658e-05,  5.9250e-04,  1.0054e-04,  3.4273e-04,
         9.0809e-04, -1.3827e-04, -1.9697e-04, -9.4891e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4289e-03,  9.8841e-01, -9.9529e-05,  7.2024e-04, -4.3047e-05,
         2.5157e-04,  6.6478e-05, -5.5638e-04, -3.9982e-04,  1.2367e-04,
        -3.0489e-04,  7.7475e-04, -3.3299e-04,  7.2963e-04, -3.5085e-04,
        -7.5451e-05,  5.7460e-04,  3.4458e-05, -2.1595e-05,  2.7411e-04,
        -4.1052e-04,  4.6822e-04,  6.2569e-04, -5.5275e-04,  3.6763e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1808e-04, -9.9167e-01, -4.3860e-04, -2.5324e-04,  1.0028e-03,
        -1.7337e-04, -6.8057e-04, -2.8298e-04,  2.8354e-04, -6.6653e-05,
        -2.6743e-04, -8.9458e-04,  3.3739e-04, -9.1120e-05,  3.4607e-04,
        -1.0371e-03, -8.0985e-04,  8.4565e-04,  1.0178e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0224e-05, -9.9295e-01, -3.5420e-04,  1.0781e-04, -2.5318e-04,
         1.3691e-04,  2.4978e-04,  1.0351e-04, -1.9108e-04,  1.7377e-04,
        -7.2057e-05,  4.3560e-05,  4.6739e-04, -7.0286e-05,  3.5177e-04,
        -3.4568e-04,  6.0008e-05, -6.3387e-05,  1.0163e-04, -5.1801e-05,
         3.7697e-04, -6.9824e-05,  7.5855e-05,  3.1839e-04,  1.4734e-04,
        -5.1207e-06,  1.1816e-04, -2.9689e-04, -2.4103e-04, -4.9551e-04,
         3.6781e-04, -4.2761e-05, -1.0981e-04, -1.1283e-04, -3.1052e-04,
        -1.4660e-04, -8.4157e-05, -1.8722e-04, -2.9021e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.7540e-04,  9.9556e-01, -1.8406e-03, -3.8451e-04, -9.3886e-04,
        -3.0601e-04,  2.0423e-05, -1.7502e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4304e-03,  9.9492e-01,  9.3797e-04, -1.4738e-04, -2.1682e-04,
        -1.8006e-04, -9.9782e-05, -9.3775e-04,  4.5053e-04, -7.5562e-05,
         1.6592e-04,  4.2146e-04, -1.8351e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2435e-03, -9.8916e-01, -5.0297e-05, -2.1521e-04, -1.6690e-04,
        -3.8382e-04,  1.6164e-06,  7.2980e-04,  6.3838e-04, -5.9855e-05,
         1.8395e-04,  4.4763e-04, -4.9201e-04,  4.8846e-05, -1.9044e-04,
        -1.0840e-04, -5.2437e-04, -8.5814e-05, -7.5074e-04,  1.1146e-05,
         3.6234e-05,  1.2123e-03, -5.3776e-04,  9.1267e-04, -1.3354e-04,
        -9.2118e-05, -1.1454e-03,  1.5411e-04,  5.4576e-05,  2.2531e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0694e-04,  9.9146e-01,  7.6918e-04, -2.2650e-04, -1.0712e-04,
        -6.8439e-05,  5.2644e-04,  1.6016e-04, -1.0170e-03, -2.6282e-04,
         1.0329e-03,  2.5144e-04,  3.7437e-04,  5.2302e-04,  1.6770e-05,
         3.3769e-04,  1.8012e-04,  4.4264e-04, -1.0748e-04, -3.1411e-04,
         1.8512e-04,  1.1300e-04,  1.2541e-04,  2.6110e-05,  4.7994e-04,
        -5.0197e-05,  2.3400e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+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([ 4.8870e-04,  9.9349e-01,  1.2565e-04, -2.4073e-04,  2.0370e-05,
         4.2276e-04,  4.0228e-04,  2.0922e-04, -3.3866e-04, -3.6587e-04,
         2.4832e-05, -7.8764e-04,  2.9650e-04,  2.1179e-04,  9.8345e-05,
         2.6258e-04, -1.4042e-04, -8.4887e-05, -5.7238e-04, -4.7446e-04,
        -3.0320e-04,  1.6046e-04, -1.0473e-04,  3.7161e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.6844e-04, -9.9297e-01,  1.2230e-04,  7.2287e-05, -3.4819e-04,
         5.5652e-05,  3.8027e-04,  8.6543e-05, -1.5440e-04, -5.7883e-04,
        -5.1398e-04,  2.5942e-04,  1.4501e-04, -4.6902e-05,  1.5161e-04,
         3.2613e-04, -2.5134e-04, -1.5207e-04, -4.2303e-04,  2.9670e-04,
        -2.2505e-04,  4.3488e-04,  2.3369e-04,  7.2546e-05, -3.0900e-04,
        -4.6953e-04,  9.5440e-05,  1.6053e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.1433e-04,  9.9044e-01, -6.2805e-04, -4.4142e-04,  1.9376e-04,
         1.2011e-03,  8.1011e-04, -3.4689e-04,  8.5217e-04,  6.5873e-04,
         8.2433e-04,  9.4583e-04, -7.3858e-04,  7.4733e-04,  2.5256e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0489e-03,  9.8878e-01, -5.6281e-04, -3.9579e-04, -2.8631e-04,
        -1.7655e-04,  1.2534e-04, -1.5636e-04, -7.1361e-04, -3.2839e-04,
        -4.7751e-04, -4.1056e-04, -1.4780e-04, -4.1670e-04,  3.3153e-04,
         2.2352e-04, -1.1150e-04, -4.4649e-04, -5.1608e-04, -8.7135e-04,
        -2.5789e-04,  5.8550e-06, -4.1338e-04, -4.1623e-04, -5.1247e-05,
         6.0948e-05,  2.8444e-05, -7.4719e-05, -1.7831e-06,  4.0690e-05,
         7.1691e-05, -5.4478e-04,  3.6574e-05, -3.0778e-04, -2.6106e-04,
        -5.0583e-04,  3.9454e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8592e-05, -9.9430e-01, -4.2900e-04, -1.1847e-04, -1.9346e-04,
        -1.0417e-03, -5.9063e-04,  2.3913e-05,  1.5685e-05, -6.4183e-04,
        -4.4862e-04,  4.8031e-04,  2.4900e-06,  6.4145e-05,  3.8644e-04,
         4.8443e-05, -2.7114e-04,  3.5024e-04, -1.3275e-04, -9.0202e-06,
         3.7636e-05,  3.8057e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.1658e-04,  9.9491e-01, -2.5448e-04,  6.8391e-04,  2.7511e-04,
         7.6238e-04,  2.0000e-04,  2.0888e-06, -5.0461e-04,  1.1387e-03,
        -4.1614e-06, -1.0343e-04,  1.4203e-05, -2.2705e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3661e-03,  9.9024e-01,  1.0733e-03,  1.0185e-03, -1.0233e-03,
         5.6524e-04, -4.0144e-04,  6.9710e-04,  1.2770e-03,  1.2919e-03,
        -4.6567e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.1035e-04,  9.9444e-01, -1.0697e-03,  2.6307e-04, -1.6332e-04,
         1.2627e-04,  1.3689e-03, -1.5655e-04,  1.6370e-04, -1.2855e-03,
        -3.5071e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3498e-04,  9.9124e-01,  3.8054e-04,  1.7854e-03,  5.8972e-04,
         2.9491e-04,  1.3564e-04,  4.6063e-04,  8.7379e-04,  8.7793e-04,
         3.0382e-04,  4.2785e-04,  5.1841e-04,  6.5080e-04, -2.7979e-04,
         6.5245e-05,  3.7249e-05, -6.2266e-05,  4.3790e-04,  2.4627e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7928e-04, -9.9368e-01, -5.2442e-04,  5.2552e-05,  6.7422e-05,
        -5.3493e-04,  1.2030e-04, -8.3090e-05, -2.2137e-04,  1.0871e-04,
         4.0363e-04,  2.3202e-04,  3.0946e-04,  2.9654e-04, -3.7028e-04,
         1.2199e-04,  4.4759e-04, -1.4374e-04,  5.4050e-04,  2.5809e-04,
         5.0950e-04, -6.8972e-04, -6.9939e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0202e-04, -9.9194e-01,  3.1483e-04, -6.5275e-04,  1.2526e-05,
         9.5665e-05, -9.7992e-05,  2.1094e-04,  1.5037e-05, -3.1404e-04,
        -3.8986e-04, -2.7923e-04, -4.4572e-04, -3.6812e-04, -3.6325e-04,
        -4.0587e-05,  2.3266e-04, -3.3379e-04, -5.1247e-04,  2.6119e-04,
         3.6850e-04,  1.3864e-04, -3.6432e-04,  3.3602e-04, -3.2383e-04,
        -6.9354e-04, -6.6730e-05, -2.7558e-04, -1.4609e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.0720e-06, -9.9156e-01,  1.1813e-04, -3.1635e-04,  8.8371e-05,
        -2.6984e-04,  1.1321e-04,  1.3675e-04, -3.1583e-04,  3.6165e-04,
        -4.5544e-04,  1.0800e-05,  9.2679e-04, -1.6053e-04, -2.3270e-05,
        -1.9279e-04, -2.5056e-04, -7.2667e-05,  8.4386e-05,  1.2429e-04,
        -1.4536e-04, -3.0215e-04, -4.2863e-04,  3.9029e-04,  1.2934e-04,
         1.3266e-04,  1.7221e-04, -1.6875e-04, -1.5856e-04,  1.4710e-04,
         2.7915e-04, -5.0613e-04, -6.1314e-05, -9.4788e-05,  7.8766e-05,
         8.7607e-04, -3.4652e-04,  2.8138e-07], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 7.8740e-04,  9.9487e-01, -1.2591e-04,  5.5610e-06,  8.7575e-05,
         2.7772e-04,  6.4728e-04,  2.6206e-04,  9.5557e-05, -2.0999e-05,
        -8.2319e-04, -3.9559e-05,  1.9863e-04, -1.1611e-04, -2.6383e-04,
        -4.3319e-04,  5.2394e-04, -2.2232e-04,  1.9616e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9562e-05,  9.9306e-01, -7.2537e-05,  6.3924e-04,  3.0313e-04,
         5.6814e-04, -5.5856e-04, -7.4443e-04, -1.4190e-04, -3.3747e-04,
        -3.3720e-04,  2.7157e-04, -6.8526e-04, -1.5173e-04,  1.8809e-05,
        -7.4594e-04, -1.9056e-04,  3.2802e-04, -6.2938e-04,  1.6270e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2577e-04,  9.9045e-01,  8.6766e-04, -3.6096e-04,  3.4015e-04,
        -2.0702e-05, -3.9046e-04, -5.0480e-04, -7.2886e-04, -2.4968e-04,
        -6.9416e-04,  1.9917e-04, -1.2720e-03, -7.7890e-04, -1.7061e-04,
        -7.2881e-04,  8.1134e-04, -3.8210e-04,  1.7789e-04, -6.2688e-04,
        -1.2098e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2373e-03, -9.9098e-01, -7.5887e-04, -6.9810e-04, -9.7473e-05,
        -2.3612e-04, -4.0881e-04,  1.2951e-04, -8.2643e-04, -4.9559e-04,
         1.4716e-04, -4.8233e-04,  5.2580e-04, -9.7130e-04, -3.0011e-04,
         3.1778e-04,  5.9433e-04, -1.1827e-04, -6.7549e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2738e-03, -9.8637e-01,  4.6021e-04,  3.9124e-05,  1.4042e-04,
        -3.0570e-04,  2.5442e-04,  7.3130e-04,  3.1563e-04,  6.1865e-06,
         4.4780e-04,  2.1985e-04,  1.5654e-04,  1.7854e-04,  3.0980e-04,
         7.3393e-04,  4.2259e-04,  1.1042e-04, -2.3918e-04,  3.1349e-04,
         2.2772e-04,  3.0094e-04, -4.3608e-04, -1.2580e-04,  2.6265e-04,
        -7.6770e-05, -2.0475e-04, -6.2580e-05,  1.9278e-04,  9.3480e-05,
         1.9271e-04, -2.0914e-04, -4.1752e-04, -9.5180e-05, -2.3715e-04,
         4.4036e-04,  2.4417e-04,  3.7266e-04,  1.2663e-04,  2.0143e-04,
        -1.4045e-04,  3.1270e-04, -3.3455e-04, -7.5029e-04,  5.9342e-04,
         3.2232e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4674e-04,  9.9370e-01,  2.6692e-04, -6.3718e-05,  1.8228e-04,
        -2.1989e-04,  2.8371e-04,  1.4107e-07, -8.1682e-06,  1.1825e-04,
         1.7459e-04, -1.6272e-05,  2.6925e-04, -1.9514e-04, -9.5039e-05,
         2.9289e-04,  2.7791e-04,  5.4502e-04, -3.7409e-04,  2.7342e-04,
         1.7006e-04,  3.8651e-04,  2.1198e-04,  4.8017e-04,  2.0933e-04,
         4.8937e-04,  2.2327e-04, -2.1848e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.9315e-04, -9.9436e-01,  1.2879e-04,  1.2633e-05, -2.3126e-04,
        -1.2899e-04, -3.8836e-05,  6.3746e-05, -2.3979e-04, -7.8733e-04,
        -5.9305e-05, -5.2550e-06, -3.5884e-04,  2.8357e-04,  6.9933e-05,
        -1.4886e-05, -1.6168e-04, -1.3555e-04, -2.2037e-04,  2.3303e-04,
        -1.4567e-04,  3.3864e-05,  2.7280e-04,  2.0525e-04, -4.2923e-05,
         1.3609e-04,  1.9266e-04,  4.4656e-04, -2.0189e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5030e-03,  9.8695e-01,  1.9536e-04,  1.7433e-04, -6.1380e-04,
        -3.5127e-05, -3.3179e-04,  6.0163e-04,  3.4970e-04, -1.4342e-04,
         4.7620e-05,  8.3687e-05, -6.6286e-04,  4.7127e-04,  6.0286e-04,
         3.8225e-04, -1.3371e-03, -8.0290e-05, -3.4547e-04, -3.9278e-04,
        -8.9006e-05, -2.2649e-04,  1.2167e-04,  2.7127e-04,  1.0092e-03,
        -1.1590e-04, -6.2056e-04, -2.4169e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4901e-03, -9.8983e-01, -1.7115e-04, -9.8270e-05, -2.4911e-04,
        -5.4802e-04,  5.2907e-05, -5.9287e-05, -1.2112e-04,  3.1759e-04,
        -1.2103e-05, -4.7358e-04,  2.3156e-04, -1.7143e-04, -7.8809e-05,
        -1.4880e-04, -1.8293e-04,  4.0193e-05,  1.5809e-04,  1.0279e-04,
         1.2589e-04, -4.6111e-05, -5.7842e-05, -5.6811e-05, -4.1380e-05,
         3.2434e-04, -3.6858e-04, -1.2943e-05,  2.8256e-04,  3.2450e-04,
         2.3518e-04, -3.8587e-05, -4.5209e-05,  2.3631e-04, -2.9638e-04,
        -3.4538e-04,  3.1131e-04,  3.1006e-04,  8.0023e-05,  4.5502e-04,
        -9.3818e-05, -2.2060e-04, -1.4252e-04,  5.7590e-04, -1.1942e-04,
        -3.0489e-04, -7.9880e-06], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0472e-03, -9.9266e-01,  1.8981e-04, -5.0308e-05,  4.4638e-05,
         1.1815e-04, -3.0888e-04, -3.1778e-06,  3.6174e-04, -3.0503e-05,
        -8.0418e-05, -1.2337e-05, -1.0306e-04, -1.3848e-04, -4.9298e-04,
        -1.3977e-04,  2.6102e-04, -7.6121e-05, -1.1873e-04, -7.1008e-05,
         9.7042e-05, -1.4655e-04,  4.7883e-06,  8.5536e-05,  5.1555e-05,
        -4.2650e-05, -1.6142e-04, -1.9406e-04,  1.6621e-04,  1.0683e-04,
         2.1392e-05,  3.2151e-04,  1.2314e-04, -1.1784e-04,  1.6230e-04,
         4.3676e-04,  2.5404e-04,  4.1464e-04,  3.8627e-05,  5.0699e-05,
        -8.8351e-05, -8.6212e-05, -1.9583e-05,  3.1556e-04, -1.3717e-04,
         4.6424e-05,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.0848e-04, -9.9258e-01,  3.4925e-04,  4.9593e-04, -5.9249e-04,
        -1.5670e-05, -3.0899e-04, -5.6203e-04,  1.2551e-05, -4.3385e-04,
        -5.5200e-04,  8.1974e-05,  1.0282e-04,  1.7569e-04,  4.0988e-04,
         2.7143e-04, -2.8433e-04, -3.3886e-04, -8.9683e-04, -2.7609e-05,
        -2.9815e-04,  6.0344e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4298e-03, -9.8996e-01, -5.9762e-05,  5.0952e-04,  4.4317e-04,
        -3.7525e-04, -9.0242e-04, -5.4543e-05, -9.4263e-05, -1.1367e-04,
        -7.5815e-05, -4.8057e-05,  4.7296e-04,  1.5050e-04, -1.3377e-04,
         8.5820e-04,  6.2098e-04, -2.2177e-05,  2.1493e-04, -4.3226e-04,
         3.3260e-05,  1.3389e-04,  6.8160e-04, -3.1594e-04, -4.0417e-04,
        -6.6765e-04,  9.3521e-05, -3.4909e-04, -3.0476e-04,  4.8392e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+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([ 2.2455e-03, -9.8892e-01, -6.9018e-04, -7.3705e-04,  2.0527e-04,
        -2.1172e-05, -5.2905e-04,  8.9816e-04,  6.2594e-04,  5.6213e-04,
         7.5044e-05, -1.8786e-03, -6.6543e-04,  4.8634e-04, -5.5272e-04,
         6.9654e-04, -2.0897e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0976e-03, -9.9099e-01,  4.9801e-04, -3.6876e-04, -4.5862e-04,
         1.9677e-04, -4.4456e-04, -5.2456e-05,  4.8754e-04, -4.6075e-05,
         8.1181e-04,  8.1003e-05,  6.5179e-04,  7.7214e-04, -1.1049e-04,
         1.0125e-03, -6.4552e-04,  3.9744e-04,  4.3774e-04,  4.3782e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0433e-03, -9.9249e-01,  2.1474e-04,  5.2218e-04, -6.7536e-04,
         1.6962e-04,  1.0823e-03, -2.5653e-04,  1.2428e-03,  1.2602e-03,
        -6.5917e-04,  3.6574e-04,  1.5260e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1676e-03, -9.9377e-01,  1.0263e-04, -1.8821e-04,  6.2271e-04,
         1.1353e-04, -1.0055e-04, -4.0966e-04,  6.9539e-05,  7.4850e-04,
        -5.1275e-04,  5.5500e-04,  3.6122e-05,  1.9850e-04,  8.0561e-04,
         2.1720e-04,  1.8772e-04,  1.9592e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1160e-03, -9.9199e-01, -7.6343e-05,  4.7473e-04,  3.4042e-04,
         5.0117e-04,  3.8033e-04, -1.4367e-04, -7.5192e-05, -4.7152e-04,
         1.2252e-04,  4.8365e-05, -1.8642e-04,  7.4864e-04,  6.6440e-05,
        -4.3906e-04,  5.6348e-04,  4.6219e-04,  1.6105e-04, -7.8104e-05,
         2.5939e-04, -1.7412e-04, -2.1839e-04,  1.9025e-04,  6.7817e-04,
        -2.9130e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5252e-03,  9.8946e-01, -9.3667e-04,  8.1807e-04,  1.2274e-03,
         2.3341e-03,  2.0421e-03,  9.4874e-04,  4.6983e-04, -1.6380e-04,
        -7.2605e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2632e-03,  9.9193e-01, -1.4728e-03, -3.7075e-05, -1.7011e-03,
         1.0175e-03, -3.4413e-04,  1.5388e-04,  2.6804e-04,  8.1495e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3566e-04, -9.9129e-01, -2.2163e-04, -9.1298e-05,  3.7954e-04,
         1.3798e-05, -4.8200e-06,  1.4691e-04,  7.8611e-05,  1.5240e-04,
         9.5631e-06,  7.6842e-05,  4.4927e-04, -3.6846e-04, -3.4474e-05,
         4.9262e-04, -1.1391e-04,  1.7434e-04,  3.9526e-04, -5.8408e-05,
         9.7740e-05,  1.4015e-04,  4.0757e-07, -1.5992e-04, -1.2978e-04,
        -1.1157e-04,  2.9030e-04,  2.1683e-04,  2.2634e-04, -1.5313e-04,
         8.6142e-05,  2.7743e-04,  7.0385e-05,  1.7938e-04, -3.7797e-05,
         1.4058e-04,  1.3026e-04,  6.3556e-05,  1.6399e-04,  9.6999e-05,
         7.1333e-05,  2.0261e-04, -9.6652e-05, -4.4589e-04,  1.0467e-04,
         2.5232e-04,  3.2140e-04,  8.6871e-06,  5.5199e-05, -2.6389e-05,
         3.9855e-05,  1.9826e-05, -2.3912e-04,  4.0417e-04,  2.2802e-04,
        -1.9741e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5469e-03,  9.9194e-01, -7.0621e-04, -5.8947e-05, -2.6119e-05,
         1.1999e-04,  6.1000e-04, -3.7365e-04, -3.5161e-04, -1.3808e-04,
         2.8877e-04, -2.2921e-04,  3.5275e-04,  6.3732e-04,  1.7628e-04,
         2.5155e-04,  3.0419e-04, -1.2361e-05,  6.9260e-04, -1.8451e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5382e-04, -9.9070e-01,  6.2422e-05,  4.4433e-06,  2.4002e-06,
        -2.9166e-04, -1.4521e-04, -5.0354e-04,  4.2780e-04,  3.7727e-04,
         5.0094e-04,  4.9912e-04,  8.1109e-04,  4.2723e-04,  4.3009e-04,
        -1.5177e-04, -4.7669e-04,  1.0848e-04,  1.4928e-04,  9.6828e-05,
         1.9630e-04,  3.1315e-05,  5.4334e-04, -2.5136e-04, -3.9688e-04,
         5.3399e-04,  3.6716e-04,  3.5499e-05,  1.6990e-04,  3.0369e-04,
         4.4947e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([3.4702e-04, 9.9796e-01, 1.9862e-04, 1.6393e-04, 5.4464e-04, 6.9599e-04,
        9.2550e-05, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00,
        0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00,
        0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00,
        0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00,
        0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00,
        0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00,
        0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00,
        0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00, 0.0000e+00,
        0.0000e+00, 0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1420e-03,  9.8828e-01, -4.3749e-04, -7.5317e-04, -3.1576e-03,
        -8.5301e-04, -8.2334e-04, -2.0845e-03, -4.6822e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 8.6420e-04,  9.9241e-01,  3.6530e-04, -4.2041e-04, -2.5465e-04,
         1.7785e-04,  5.0088e-05, -1.8392e-04, -2.8976e-04,  1.7392e-04,
        -4.0509e-06,  2.1781e-04, -2.3880e-04,  1.4092e-04, -4.7710e-05,
         5.2927e-04,  5.3148e-04,  1.1019e-04,  1.7572e-04, -5.3720e-04,
        -5.5428e-04, -4.3385e-04,  1.0188e-05,  3.0329e-04, -2.3420e-04,
        -8.3329e-05,  4.5935e-04,  1.9898e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.8885e-04,  9.9174e-01,  8.8790e-04,  4.8723e-04,  5.0247e-04,
        -2.3064e-04, -2.7869e-04,  4.6437e-06, -8.9045e-05, -4.1338e-04,
         2.9032e-04,  7.0144e-05, -2.6884e-05,  6.9689e-06,  1.4496e-04,
        -1.0320e-04, -1.5271e-04,  1.5529e-04, -4.0809e-04, -1.0005e-04,
        -2.2378e-04, -9.8544e-05, -2.0079e-04, -5.0925e-04,  1.4211e-04,
        -5.1679e-04,  1.5005e-04,  3.3305e-04,  2.2349e-04, -1.3579e-04,
        -1.4931e-04, -4.3465e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5932e-04, -9.9237e-01, -4.9202e-04,  3.4882e-04,  3.7361e-04,
         2.0968e-04,  3.9308e-05,  4.1770e-04, -8.8291e-05,  2.9092e-04,
         1.4096e-04,  1.6080e-04,  1.5566e-04,  7.3991e-04,  7.7230e-04,
         5.9488e-04,  5.6528e-04, -2.6081e-04,  2.2939e-04, -1.0497e-04,
         3.6200e-04, -3.2901e-04,  2.6241e-04,  4.8878e-05, -1.8600e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.2029e-04,  9.9376e-01, -2.2353e-03, -7.8470e-04, -6.4327e-04,
         7.2899e-04, -1.2995e-03,  2.6908e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2081e-03,  9.9277e-01,  1.7217e-03, -7.7884e-04,  2.0037e-03,
         2.6361e-04,  2.5377e-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,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1727e-03, -9.9346e-01, -3.4964e-04,  1.8712e-04, -3.0720e-04,
         2.4795e-04,  4.2138e-04,  1.3312e-04, -6.6051e-04,  5.0492e-04,
        -3.3295e-04, -3.5512e-04,  1.0743e-04,  7.5247e-05,  4.0786e-04,
         2.4653e-04, -1.8199e-04,  4.3439e-05,  1.7765e-04, -6.3024e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2096e-03,  9.9307e-01, -6.7057e-04,  9.2636e-04,  3.5889e-04,
        -7.3431e-04,  3.2269e-05,  6.3400e-04, -3.0578e-05,  5.5588e-04,
        -8.1066e-05, -9.0056e-04,  5.8991e-04,  2.0458e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4573e-03,  9.8936e-01,  2.5360e-04,  2.2482e-05, -9.8089e-04,
         1.5815e-03,  9.5500e-05, -2.0166e-03, -4.8087e-04,  7.5154e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1133e-03, -9.9176e-01, -9.2161e-05,  1.0856e-04, -9.5656e-05,
         1.1389e-04,  2.7216e-04,  2.1022e-05, -5.6488e-04, -4.0175e-04,
        -2.7582e-04,  1.6653e-04,  6.3807e-05,  6.3608e-05, -3.8344e-04,
        -2.9671e-04,  2.6966e-04, -3.3621e-04, -4.3201e-04,  7.2430e-04,
        -3.5821e-04,  2.0984e-04,  4.8790e-04,  9.7923e-04,  3.2085e-05,
        -4.3425e-05,  2.3792e-04, -9.7238e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1942e-04,  9.9378e-01,  3.4759e-05, -1.4053e-04, -1.7081e-04,
        -6.2968e-04,  1.1159e-03, -6.9802e-05,  8.7640e-04, -3.3090e-06,
         2.7792e-04,  3.5147e-04,  2.5794e-04, -3.8331e-04,  7.1855e-04,
        -6.3352e-04, -3.2169e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4363e-04, -9.9224e-01, -2.9519e-04, -7.4574e-04,  5.6157e-04,
        -7.3586e-04,  3.9706e-04,  2.3313e-05,  6.9026e-04, -7.5242e-04,
         4.9641e-04, -7.7647e-05,  2.4953e-05, -3.0531e-04, -4.3122e-04,
         1.1692e-03,  3.4233e-04,  8.1007e-05,  1.8345e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.4778e-04, -9.9278e-01,  3.4178e-04,  2.0993e-04,  7.5284e-04,
        -6.3498e-04, -5.7496e-04, -6.0937e-04,  2.0064e-04,  4.5361e-04,
         1.9343e-04, -4.9990e-05,  8.1600e-04, -3.2210e-04,  2.6145e-04,
         1.7017e-04, -2.0963e-04,  7.7300e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+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([ 4.1119e-05, -9.9515e-01, -1.0731e-04,  2.9570e-04, -2.2760e-04,
        -8.2318e-05, -6.9663e-04, -3.9461e-05, -2.4666e-04,  1.7686e-04,
        -4.1268e-05,  3.5617e-04,  1.2789e-04,  2.7649e-04, -1.5386e-04,
         8.8144e-07, -2.0942e-04,  2.2896e-04, -1.9213e-04,  2.3786e-04,
         2.2514e-04,  1.2091e-05, -1.8915e-05, -1.7588e-04,  5.2643e-05,
         6.7014e-05,  5.6180e-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], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3682e-03, -9.8902e-01, -1.9657e-04, -1.1349e-04,  6.6963e-04,
        -4.1763e-04, -2.6303e-04, -5.2973e-04,  4.0343e-05,  2.2124e-04,
         4.8295e-05, -8.3703e-05, -2.0952e-04,  4.3701e-04,  4.7112e-04,
         8.1105e-05,  5.2963e-05, -3.0718e-05, -6.1296e-05, -2.7173e-05,
        -2.1206e-04, -1.6114e-04,  2.4902e-04,  3.6815e-04, -1.5976e-04,
        -7.1375e-06,  4.7257e-05,  3.2087e-04,  2.3046e-04, -1.2255e-05,
         2.1777e-04, -6.3651e-05, -4.7702e-04,  4.4678e-04, -1.9976e-04,
        -3.1545e-05,  1.8261e-04,  1.5847e-04,  1.4853e-04, -4.7297e-04,
         1.1502e-04, -5.7529e-04,  1.7319e-04, -3.9461e-04, -2.2898e-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([ 7.9553e-04,  9.9110e-01, -6.5317e-04,  2.6089e-04,  1.8077e-04,
         4.9409e-04, -6.5173e-04, -7.6084e-05,  6.2345e-04, -4.9508e-04,
         8.6661e-05, -4.0184e-04,  1.5889e-04, -5.7577e-06, -4.6145e-04,
         1.1730e-04, -3.4790e-04, -3.1185e-05,  5.5409e-04,  7.8304e-05,
         1.8649e-04,  5.2502e-05, -1.1263e-04, -6.7797e-04, -4.2350e-04,
         5.1414e-05, -2.8890e-04, -3.0837e-04, -3.2616e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5178e-03,  9.9289e-01,  5.0516e-04, -4.1877e-04,  7.3953e-04,
        -1.5770e-04, -1.2439e-03,  9.4760e-04,  6.3058e-04, -5.0961e-04,
        -5.5415e-05,  3.2337e-04,  5.7992e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5742e-03, -9.9219e-01,  6.1373e-06,  2.6994e-04,  8.5922e-05,
         1.9733e-04,  1.6435e-04,  3.0168e-04,  2.3810e-04, -8.5030e-06,
         3.7840e-04, -7.8854e-06, -1.9481e-04,  1.2706e-04, -4.1205e-05,
        -3.7839e-05,  4.1853e-04, -3.3719e-04,  9.4015e-05,  1.7304e-05,
         7.5733e-05, -2.3948e-04, -1.9557e-04, -2.5257e-04, -3.3937e-04,
        -1.0173e-04, -1.3548e-04, -4.8796e-05, -9.9141e-05,  1.7122e-05,
        -7.1572e-05, -3.7249e-04,  1.7573e-04,  2.2328e-04, -6.4071e-05,
         5.4463e-04, -3.5177e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2470e-03, -9.9109e-01,  3.1110e-04,  2.3866e-05, -5.9038e-05,
        -2.3101e-05,  1.1471e-04, -1.5621e-04, -2.9185e-04,  6.7529e-04,
         1.7665e-04,  3.4086e-04,  1.0804e-04, -2.0830e-04,  2.5550e-04,
         6.8037e-04, -4.3725e-05,  1.8848e-04,  1.4902e-04,  2.8390e-04,
        -1.3786e-04,  3.6166e-04,  4.5283e-04,  1.7717e-04,  7.5894e-05,
         1.4922e-05, -2.2123e-04, -2.1476e-04, -1.5369e-05, -2.8609e-05,
         2.1325e-04,  3.0339e-04, -6.2279e-04, -5.0491e-04, -2.2680e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.1829e-04,  9.8618e-01, -2.3132e-04, -7.9889e-04,  2.8928e-04,
         1.3879e-03, -5.5196e-05,  7.0849e-04,  8.1341e-04, -1.4076e-03,
         3.6076e-04, -1.6144e-03,  5.1775e-04, -3.8988e-04, -1.1709e-03,
         2.3216e-04, -8.4145e-05,  7.4256e-04, -3.0188e-04, -1.2541e-03,
        -6.4250e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3959e-04, -9.9330e-01, -1.4383e-06,  3.9118e-04, -2.8736e-04,
        -1.8061e-04,  1.1318e-04, -1.0543e-04,  1.8927e-04,  1.8939e-04,
        -1.9306e-04,  1.8637e-04, -1.3515e-04, -1.9534e-04,  6.9993e-05,
        -1.5407e-04, -6.0524e-04,  1.0555e-04, -1.9685e-04,  4.4224e-04,
         2.9243e-04,  3.4893e-04, -3.0081e-04,  9.8163e-05,  4.3046e-04,
        -1.1834e-04,  1.7283e-04, -2.5494e-04,  1.4735e-04, -2.9894e-04,
         1.6043e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.8483e-04, -9.9352e-01, -9.4220e-05, -3.6432e-05, -1.1847e-04,
         9.7955e-05, -1.2384e-04, -1.7123e-05, -2.9734e-04,  4.1991e-04,
        -1.5051e-04, -1.3791e-04, -1.3835e-04,  1.8060e-04, -6.5159e-07,
        -2.0088e-04, -3.0432e-04, -1.4192e-04, -4.4818e-05,  3.0003e-04,
        -2.4322e-04, -2.6066e-04, -3.6948e-05,  1.2053e-04, -3.4665e-04,
        -1.4756e-05, -2.1811e-05, -5.4506e-05,  8.2902e-05,  1.7903e-05,
        -1.1332e-05,  5.7057e-05, -1.1733e-04, -3.1191e-04,  7.2200e-06,
        -1.2134e-04,  4.0839e-04,  5.1984e-05, -2.4988e-05, -2.6906e-04,
        -1.8735e-04, -2.5818e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.4088e-04,  9.9032e-01,  1.5228e-03, -6.7061e-05, -8.9930e-04,
        -1.4873e-03, -9.6182e-04, -9.2712e-04, -1.4511e-03,  3.1467e-04,
         7.4756e-05,  7.1389e-04, -3.1876e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2134e-03, -9.8482e-01,  1.4513e-04,  4.0792e-04, -1.7448e-04,
         8.8506e-05,  5.5494e-05, -5.4826e-04, -2.3398e-04, -4.5939e-05,
        -1.1380e-04, -1.2966e-04,  1.3387e-04, -1.2677e-04, -4.4377e-04,
         2.2584e-04,  3.2740e-04, -1.3622e-06,  1.9814e-04, -2.2602e-04,
        -5.7006e-04,  4.0036e-04, -5.2238e-04,  4.9850e-04,  3.6359e-04,
        -1.8256e-04,  1.2022e-04, -5.3382e-04,  8.2106e-05, -4.1558e-04,
         1.5533e-04,  8.2670e-05, -5.1272e-05, -7.7700e-05, -2.6853e-04,
        -2.5679e-05, -2.8096e-04,  6.4122e-05,  1.1787e-04,  1.6358e-04,
        -6.7105e-05, -2.0068e-05,  8.2400e-04, -4.1598e-04,  1.7610e-04,
        -3.5241e-05, -3.4030e-05, -3.1322e-04, -2.7738e-04,  2.9263e-05,
        -5.2158e-05, -1.2138e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.5259e-04,  9.8982e-01,  8.7060e-04,  1.0294e-03, -2.3590e-05,
         1.8330e-03,  9.4279e-04, -1.0562e-03,  6.1635e-05,  1.2758e-03,
         1.3451e-03,  2.4785e-04, -7.0187e-04, -4.4209e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+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([-2.5532e-04, -9.9366e-01, -4.8851e-05, -1.6251e-04, -2.6138e-04,
        -1.3257e-04, -1.0498e-04, -3.2449e-04,  3.0732e-04,  2.1030e-04,
        -2.2325e-04, -1.5040e-04,  7.2708e-05, -3.6174e-04,  2.2313e-04,
         8.5033e-05,  5.1895e-05,  8.4617e-05, -7.9492e-05, -2.1407e-04,
        -2.9685e-05, -2.1238e-04, -2.8229e-04,  1.2331e-04, -8.2327e-05,
         1.3331e-04,  2.9681e-04, -1.4689e-04, -1.2247e-04, -2.2962e-05,
        -9.4767e-06, -1.6608e-04,  2.1737e-04, -3.4828e-04, -2.4207e-04,
        -3.3591e-05,  7.7380e-05, -2.1231e-04, -1.8963e-04, -3.4122e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2849e-03, -9.9064e-01, -2.5982e-04, -3.8988e-04, -4.6613e-04,
        -8.1397e-05, -1.4855e-04, -4.6512e-05, -1.9689e-04,  1.7407e-04,
         7.4038e-05, -5.8067e-05,  5.8633e-05, -1.6291e-04, -1.8191e-04,
         2.6616e-04, -5.3563e-05, -1.4158e-04, -3.2114e-05,  2.0093e-04,
         2.7394e-05, -1.1883e-04,  1.0550e-04, -2.7276e-04, -4.2316e-04,
        -1.1104e-05,  8.5007e-05, -6.0747e-05, -1.4180e-04, -2.5844e-04,
        -2.5913e-05, -6.4259e-05, -4.3615e-05, -1.6540e-04,  4.6116e-05,
        -4.0728e-04,  1.4838e-06, -2.0337e-04,  2.0473e-04,  2.3527e-04,
        -4.8061e-05, -1.0836e-04,  2.7837e-05, -1.2888e-04,  2.4475e-04,
        -2.1613e-04, -4.1149e-05, -3.8825e-04, -1.2862e-04,  1.2479e-04,
        -2.7500e-05, -2.7211e-04, -1.1408e-04, -1.2993e-04, -7.3475e-05,
        -1.0557e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9084e-04, -9.9305e-01, -6.6633e-05,  3.1801e-04,  7.8345e-05,
         1.9985e-04, -2.7267e-04, -1.1402e-04, -2.8889e-05,  1.0316e-04,
         7.2072e-05,  3.3239e-04,  2.8758e-04,  3.3926e-04, -8.7203e-05,
        -1.6498e-04,  2.9391e-04,  8.8486e-05, -4.6184e-05, -6.5631e-05,
         3.7990e-04,  1.0463e-04, -1.1707e-04, -1.4886e-04,  1.2063e-04,
         1.9990e-04, -3.3145e-05,  1.3167e-04,  6.0504e-05,  2.4762e-04,
         3.8935e-04,  4.9469e-04,  3.7276e-04,  2.9625e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3535e-04,  9.9393e-01, -4.1904e-04,  5.4019e-04,  1.0508e-03,
         1.3712e-05,  1.1123e-03,  7.9461e-04, -2.5410e-04, -6.3191e-04,
         6.0323e-04,  4.1623e-04, -9.8682e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6900e-03,  9.9010e-01,  2.0566e-04,  3.7195e-04,  1.6002e-04,
        -1.9013e-04,  1.5680e-04, -1.2007e-05, -2.8586e-05, -8.1926e-05,
        -1.4900e-04,  9.2046e-06,  8.0487e-05,  2.2930e-04,  7.5559e-05,
         7.5654e-04, -3.0934e-04,  7.8510e-04, -3.0723e-04,  3.3097e-04,
         5.8952e-04, -3.1015e-04,  8.1850e-04, -3.3615e-04,  1.6031e-04,
        -5.3400e-04,  6.2736e-04, -4.6097e-05,  2.5597e-04, -2.9144e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9913e-06,  9.9125e-01,  7.0117e-04, -2.4965e-04,  2.5065e-04,
        -7.0641e-04, -4.7315e-04, -5.3867e-05,  2.4164e-04,  5.0138e-05,
        -3.2803e-04,  6.9169e-05, -1.0596e-04,  1.1245e-04,  2.2294e-04,
        -3.8527e-04, -5.4704e-05,  3.4191e-05,  5.2431e-04,  1.2299e-04,
        -2.0542e-04,  3.4358e-04, -1.0958e-04, -1.3899e-04, -5.9895e-04,
        -4.8215e-04,  7.2491e-04,  5.9548e-04, -4.6416e-04,  3.9836e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.8006e-04,  9.8890e-01,  4.3969e-04, -6.8408e-04,  5.4118e-04,
        -1.6071e-04,  4.4152e-04,  3.2048e-04, -3.2354e-06, -7.0297e-05,
        -5.0414e-04, -6.6543e-04,  1.8699e-04, -3.1188e-04,  5.1727e-04,
        -8.9443e-04,  3.6732e-04,  3.7196e-04, -6.6482e-05,  2.3649e-04,
        -3.9769e-04, -8.4946e-04,  1.2920e-04,  3.1423e-04,  4.1046e-04,
         5.9035e-04, -3.0736e-04,  1.8587e-04,  1.4890e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.8947e-04, -9.9203e-01, -2.7763e-04, -8.7552e-04,  1.3530e-04,
        -1.0465e-03, -5.0831e-04, -4.3914e-04, -2.4289e-04, -2.4713e-03,
         2.9695e-04, -7.5211e-04, -1.3100e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0028e-03,  9.9288e-01,  7.8268e-04, -2.9874e-04, -2.8481e-05,
        -3.6941e-05,  1.1728e-04, -3.9506e-04,  2.7855e-04,  5.6261e-04,
         4.6752e-04,  2.2279e-04,  2.2525e-04,  1.1009e-04,  4.0164e-04,
         5.3181e-05, -8.2518e-04,  1.6327e-04, -4.6262e-04, -3.7854e-04,
        -1.7749e-04, -1.2574e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2495e-04,  9.9448e-01,  4.5446e-04, -1.4881e-04, -1.6589e-04,
        -4.1070e-04, -2.0289e-04,  2.7775e-04,  1.1777e-04, -9.9104e-06,
        -2.6038e-04,  3.1810e-04,  1.6529e-04, -1.8518e-04, -1.8115e-04,
         5.9394e-05, -2.0866e-04,  2.8390e-04, -6.3310e-05,  2.2522e-05,
        -6.4101e-05,  1.9685e-04,  2.7905e-05,  6.0075e-05, -3.4773e-04,
         3.4910e-04,  1.6199e-04,  1.9548e-04,  1.5524e-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,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2676e-03, -9.9214e-01,  2.0184e-04,  4.5161e-05, -2.6099e-04,
        -1.2100e-04, -3.6784e-04,  2.7242e-04,  4.6058e-04, -2.2372e-04,
         5.2899e-05, -2.8087e-04, -3.9227e-04, -5.2061e-04,  1.7394e-04,
         6.2806e-05, -1.6710e-04, -5.4268e-05,  4.9733e-05,  1.4719e-04,
        -3.6366e-04, -1.0365e-04,  1.0666e-05,  2.8461e-04, -1.8530e-04,
        -3.0332e-04, -3.0064e-04, -1.0637e-04, -7.5605e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4007e-03,  9.8939e-01, -2.2027e-03,  8.6281e-04, -2.1730e-04,
        -7.4894e-05, -9.2670e-04, -9.3188e-05, -4.8387e-04, -4.5302e-04,
        -2.6315e-04, -4.1221e-04,  3.0669e-04,  3.5462e-04,  1.8703e-04,
         4.9055e-04,  1.0383e-03, -7.5909e-05,  7.6718e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([-8.6718e-04, -9.9374e-01, -1.7722e-05,  4.4940e-04, -1.3973e-04,
         1.7492e-04,  9.8315e-05, -2.6233e-04,  6.7907e-05,  3.3508e-04,
         8.7530e-05,  2.0222e-04, -5.5923e-05, -2.2579e-04, -1.8868e-04,
         1.7364e-04, -1.0351e-04, -1.9888e-04,  9.3955e-05,  5.0907e-05,
         1.8186e-04,  1.2302e-05, -1.2743e-04, -1.4762e-04, -6.4531e-05,
        -1.8769e-04, -4.4336e-05, -2.9563e-04, -4.6483e-04,  1.2813e-04,
         6.7258e-05, -6.8292e-05,  1.7819e-04,  1.1591e-04,  3.0182e-05,
         1.2594e-04,  2.2425e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1458e-04,  9.9063e-01,  1.5342e-04,  2.5004e-04,  8.5853e-04,
        -5.3025e-04,  1.2959e-04,  2.5308e-04, -3.1778e-04,  3.3782e-04,
         4.4347e-04, -2.1809e-05, -4.5706e-04, -4.5735e-04,  3.1723e-05,
        -5.0198e-05,  2.4776e-04,  5.4838e-04,  1.7222e-04,  6.1124e-04,
         5.0317e-05, -6.1696e-04,  3.4600e-04, -5.2804e-04, -2.9112e-05,
         2.9889e-04, -2.6786e-05, -3.0835e-04,  7.2751e-05,  3.5445e-04,
         6.1750e-04,  4.9966e-05,  7.9911e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5909e-04,  9.9267e-01,  4.7284e-04, -8.6195e-06, -5.3260e-04,
         3.7768e-04,  4.9759e-04,  9.3388e-05, -4.1979e-04, -8.0452e-04,
         1.0433e-04, -6.1546e-05, -2.3363e-04, -7.7513e-04,  4.6878e-04,
        -7.7666e-05,  6.6135e-04,  1.8723e-04, -6.9112e-04,  4.0689e-04,
         1.4429e-04, -4.7106e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4219e-03,  9.8792e-01,  1.3279e-04,  2.3905e-04,  8.2693e-05,
        -2.0709e-04, -1.5823e-04,  5.2404e-04, -2.7804e-04, -4.1340e-04,
        -1.2977e-04,  2.5112e-05, -6.5326e-05, -2.5317e-04, -3.7074e-04,
        -5.7369e-04,  9.6416e-05,  2.2864e-04,  3.5023e-04, -3.0110e-04,
        -4.4337e-04,  1.1347e-04,  9.6739e-05, -3.8289e-04, -1.1617e-04,
         8.5025e-05,  4.7128e-04, -7.3254e-04, -5.5605e-04,  3.4872e-05,
        -6.0980e-04,  2.7797e-04,  2.6221e-04, -4.6271e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.3208e-04, -9.9200e-01, -4.7475e-04, -5.5122e-04, -2.5828e-04,
         3.3703e-04,  1.9781e-04,  4.3240e-04,  1.4592e-03,  4.6699e-04,
         3.9538e-04,  1.4211e-03,  7.5338e-04, -2.6630e-04,  1.5525e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0294e-03, -9.8861e-01, -2.5709e-04, -5.5021e-05,  2.4564e-04,
        -3.4619e-04,  3.6976e-04, -2.6365e-04,  8.8992e-04,  2.1083e-04,
         1.0461e-03, -4.5301e-04,  9.3483e-04,  1.2008e-03, -5.9825e-04,
         3.5147e-04,  1.0385e-03,  7.0060e-04,  6.4231e-04,  5.5245e-04,
         2.0071e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.9587e-04, -9.9382e-01,  2.2295e-04,  3.3551e-06,  1.0663e-04,
        -1.9060e-04,  1.3120e-04,  7.8300e-05, -9.2865e-05, -7.6905e-05,
        -1.1408e-05, -1.4679e-04, -2.0524e-04,  2.5554e-04, -1.2823e-04,
        -2.8291e-04,  7.1251e-05,  8.2728e-05, -7.9366e-05,  5.1588e-04,
        -1.1228e-04, -2.7039e-04,  8.8766e-05,  1.0858e-04, -1.4740e-04,
        -6.0688e-05, -4.0045e-05, -2.3843e-04,  2.8541e-04, -6.6982e-05,
        -2.0834e-04, -7.4072e-05,  3.2498e-04,  1.3471e-04, -3.7676e-04,
        -2.0902e-05,  1.1080e-04, -8.1229e-05,  1.1371e-04,  3.3464e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1128e-03, -9.9363e-01,  8.8913e-04,  6.6660e-04, -4.2816e-04,
         1.7036e-05, -1.4346e-04,  4.4942e-04, -2.3977e-04, -1.7450e-04,
        -1.5505e-04, -7.1897e-04, -4.9582e-05, -6.9387e-05,  3.1928e-04,
        -1.5043e-04,  1.8498e-04, -2.0095e-04, -1.0601e-05, -3.7769e-04,
         9.4885e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.4628e-04,  9.9307e-01,  1.0177e-03,  4.8833e-04,  4.4118e-04,
        -2.7988e-04,  1.9066e-04,  2.5540e-04,  2.1539e-04, -6.3571e-06,
        -1.0415e-03, -1.2376e-04, -5.9049e-04,  8.7181e-04,  5.9699e-04,
         1.2514e-04,  4.2905e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6697e-04,  9.8825e-01, -6.9490e-04,  2.9510e-04,  2.1244e-04,
        -9.8181e-04, -2.6964e-04, -6.9393e-04, -2.3077e-04, -1.0047e-04,
        -6.9749e-04, -6.4894e-04, -1.1128e-04, -6.0697e-04, -1.1401e-04,
        -8.7795e-04, -9.2628e-04, -1.3328e-04,  1.3163e-04,  8.8655e-04,
        -5.9557e-04, -9.1321e-04, -4.3565e-04,  6.6567e-04, -1.8037e-04,
        -1.8223e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1021e-03,  9.9129e-01,  9.2437e-04,  8.4955e-05,  4.2581e-05,
        -7.9798e-04,  2.1797e-04, -2.3469e-04, -8.4934e-04, -7.9942e-04,
        -1.4151e-04, -3.9528e-04, -9.4399e-04, -1.7103e-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,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4604e-04,  9.9390e-01,  1.0387e-04, -3.3411e-05,  4.4790e-04,
         9.1291e-05,  1.2624e-04,  4.6827e-04,  2.8325e-05,  2.1672e-04,
        -5.2638e-04,  1.3414e-04,  4.0235e-04,  1.1764e-04, -1.7802e-04,
         3.1724e-05, -7.2648e-06, -3.6977e-04, -3.3102e-04, -4.2884e-05,
        -2.8674e-04,  1.7397e-04, -3.7619e-04, -4.5525e-04, -4.2134e-04,
         1.5794e-04,  2.1053e-04, -2.1570e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+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([-1.2241e-03,  9.9400e-01,  1.1095e-03, -4.8722e-04,  6.8710e-04,
         4.4978e-04, -2.8798e-04, -1.6653e-03, -9.3648e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.4682e-04,  9.9335e-01, -1.3332e-04, -7.1979e-05, -4.4030e-04,
         3.4580e-04,  2.5225e-04, -3.9751e-04,  2.3763e-04, -1.9667e-04,
        -5.5150e-04, -2.4382e-04,  4.9532e-04, -2.8821e-04, -5.1598e-04,
        -1.8922e-04,  1.7812e-04, -2.9902e-04,  1.1626e-04,  1.7682e-04,
         7.7490e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7064e-03,  9.9143e-01,  9.9011e-04, -3.5088e-04,  2.6198e-04,
        -1.5708e-04,  1.6657e-04,  6.0248e-04,  7.4738e-04,  3.2035e-04,
         4.1388e-05, -4.6085e-04, -5.7428e-04, -3.0090e-04, -6.1137e-05,
        -7.6673e-04,  1.2400e-04,  6.1822e-05,  7.9551e-04,  7.5571e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5864e-04,  9.9514e-01, -9.1166e-05,  4.1820e-04, -1.6396e-04,
         3.7030e-04, -4.3217e-04,  9.2947e-05, -7.7178e-05, -1.4985e-04,
        -4.1526e-05,  8.9641e-04, -7.7907e-04, -2.8702e-04, -1.3521e-04,
         4.6954e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2526e-05, -9.9197e-01,  2.9681e-04,  1.0261e-03, -2.0203e-04,
        -3.5818e-04, -3.0736e-04,  5.5382e-04, -1.8190e-04,  4.2237e-04,
        -6.7306e-04, -6.9661e-04, -8.8407e-04,  8.5587e-04, -5.6378e-04,
        -2.7423e-04,  3.0414e-04, -4.0387e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5348e-03,  9.8912e-01,  6.1475e-04, -5.5225e-04, -6.2684e-04,
         3.9800e-04, -3.5475e-04, -5.2290e-04,  4.9183e-04, -6.1088e-04,
        -1.2636e-04,  6.0679e-04,  1.0879e-04,  1.2039e-04,  8.8587e-04,
         1.2281e-03,  6.3859e-04, -1.7570e-04, -1.0018e-04,  1.8425e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4051e-03,  9.9077e-01, -4.4826e-04,  1.6383e-04,  5.9910e-04,
        -1.1215e-04, -8.6025e-04,  5.0517e-05,  2.0059e-04, -4.6067e-04,
         1.0825e-03,  9.9653e-05,  3.2583e-04, -3.4830e-04,  8.2161e-04,
        -5.3467e-04, -8.4910e-05, -3.8542e-04,  2.0368e-04, -2.0669e-04,
        -2.3633e-04,  3.3092e-04, -2.7093e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.1607e-04,  9.9214e-01, -3.0169e-05, -1.9982e-04, -4.4064e-04,
        -4.1482e-04,  6.2145e-05, -1.2990e-04,  1.1377e-04, -3.6493e-04,
        -6.6369e-05, -8.7628e-05, -1.2763e-04, -2.0942e-05, -3.6141e-04,
        -6.2293e-04, -7.3249e-05, -6.7989e-05,  1.5354e-04, -5.5464e-07,
        -8.4600e-04, -6.1237e-04, -6.0837e-04, -6.6217e-05, -6.0414e-04,
        -3.0199e-04,  1.7486e-04, -3.1873e-04, -5.1148e-04,  4.6291e-06,
         1.5326e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4471e-03, -9.9270e-01, -9.8201e-05, -4.8838e-04, -4.6246e-04,
         6.2331e-05,  4.5568e-04, -6.0839e-04, -2.7451e-05,  1.9247e-04,
        -4.3154e-04, -3.5027e-04, -4.6698e-05, -2.1106e-05,  4.7886e-05,
         4.6655e-04, -2.1170e-04, -3.3855e-04, -8.5128e-05, -5.6384e-05,
         1.0203e-04, -1.2858e-04,  1.4807e-04, -2.1837e-04, -2.4793e-04,
         3.1420e-04,  5.2869e-05, -1.8922e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8547e-03, -9.9166e-01, -1.3729e-03, -2.7401e-04, -9.3155e-04,
         1.9996e-04, -9.0392e-04,  6.8461e-04, -5.2253e-04, -5.6445e-04,
         1.0323e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2244e-04,  9.9121e-01, -1.4419e-04,  3.1103e-04, -9.4880e-05,
         2.2916e-04, -5.9582e-04,  3.5375e-04, -4.6948e-04,  1.6506e-04,
        -4.1806e-04,  1.1669e-04, -1.0912e-05, -2.2664e-04, -8.4315e-05,
        -2.9886e-06,  7.6520e-04, -3.2385e-04,  2.5162e-04,  4.3179e-04,
         1.0739e-04, -3.4387e-04, -3.9089e-04,  1.8325e-04, -6.6381e-04,
        -1.0174e-04, -6.4434e-04, -5.1003e-04, -2.6570e-05,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.2556e-04,  9.9342e-01, -3.6229e-05,  1.5885e-04,  4.8459e-04,
         2.7642e-04, -4.5926e-04, -4.1862e-05, -2.5211e-04,  2.1858e-04,
        -1.3982e-04, -2.0581e-04,  2.8099e-04, -3.2707e-04, -7.8540e-05,
         8.3706e-04,  5.3526e-04,  6.7690e-04, -2.0271e-04,  1.8675e-04,
        -2.0746e-05,  2.3056e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 1.5530e-03,  9.9118e-01,  7.4291e-04, -3.9412e-04, -8.3493e-04,
        -7.1833e-04,  8.0472e-06, -2.7319e-04,  2.9323e-04, -1.3153e-03,
        -7.8368e-04, -6.0996e-04, -5.3975e-04, -7.5229e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0679e-03,  9.9333e-01, -4.0024e-04,  3.2445e-04,  1.1777e-04,
        -3.8602e-04, -2.7498e-04,  1.3612e-04,  5.1810e-04,  1.3344e-04,
         7.5660e-04, -2.1343e-04,  2.4502e-05,  3.9909e-04, -5.2406e-05,
        -4.0708e-04,  2.1922e-04, -9.0180e-05,  6.3505e-05, -2.4692e-04,
         3.4177e-04, -1.1858e-04,  1.5003e-04, -2.3284e-05,  1.9974e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6160e-03,  9.9129e-01, -5.4311e-04,  2.1969e-04,  9.3769e-05,
        -3.2405e-04,  1.1576e-04,  1.2972e-04,  1.9794e-04, -3.3908e-04,
        -8.7301e-05, -6.7032e-05,  4.3381e-04,  5.2202e-04,  1.7082e-04,
         1.0074e-04, -5.7567e-04,  3.8911e-04,  1.0938e-04, -9.5990e-04,
        -6.1822e-05, -2.6752e-04, -5.1557e-05, -2.9664e-05,  6.1056e-05,
        -6.3091e-04, -1.3952e-04, -2.8319e-04,  1.8517e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1095e-03,  9.9222e-01,  4.7725e-04, -5.0594e-04,  5.6271e-04,
         1.4475e-03,  7.5463e-04, -6.6923e-04, -5.9629e-06, -2.1516e-04,
         1.6253e-04,  1.8740e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.3188e-03,  9.8801e-01, -1.4303e-04, -5.6063e-04, -6.9232e-04,
         9.6818e-04,  1.8803e-04,  4.2273e-04,  6.0225e-04, -1.2102e-03,
        -3.7297e-05, -3.2055e-04, -1.3797e-03, -4.3447e-04,  3.3676e-04,
        -3.7800e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.4050e-04,  9.9126e-01,  6.2075e-05,  5.2314e-05,  3.3559e-04,
        -1.4406e-04, -2.9967e-04, -7.5804e-05, -1.3718e-04, -3.5899e-04,
         6.2048e-04,  4.5266e-04,  7.2571e-04, -4.4825e-04, -4.8128e-04,
        -2.7126e-04,  1.9318e-04, -3.3767e-04, -1.8868e-04, -2.6290e-04,
        -5.0336e-04, -1.4557e-03, -2.0374e-04, -7.1423e-05, -2.7624e-04,
         6.7470e-05,  7.5836e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7835e-04, -9.9043e-01, -3.4318e-05, -5.8438e-05, -8.2924e-04,
        -2.1104e-04, -8.1278e-05,  2.0826e-04, -3.6061e-04, -7.7570e-04,
        -8.7926e-04, -8.5460e-04, -2.2658e-04,  2.4503e-04, -7.4675e-04,
        -4.7716e-04,  1.2501e-04, -1.3414e-04,  8.4042e-05, -3.1038e-04,
        -5.7674e-04,  1.8673e-04, -1.2933e-03, -4.9457e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.7034e-04, -9.9385e-01, -4.3420e-05, -3.6989e-04,  3.8776e-05,
         1.4451e-04,  1.0967e-04, -6.0169e-06, -1.7556e-04,  1.7045e-04,
         9.5732e-05, -1.1281e-04, -3.3659e-04,  3.5993e-04, -1.2474e-04,
        -2.4905e-04,  4.9784e-06,  4.2226e-05, -5.4315e-05, -3.3004e-04,
        -1.8010e-04,  1.5407e-04, -7.5106e-05,  2.2169e-04, -2.9235e-04,
        -3.9296e-04,  1.8132e-04,  4.0566e-04, -1.4432e-04, -1.2324e-05,
         5.4897e-05, -2.3919e-05, -1.5624e-04, -4.2070e-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([-1.3208e-03, -9.9301e-01, -8.0892e-05, -3.4698e-04, -2.1876e-04,
         1.0287e-04,  2.9802e-04,  5.9773e-05,  1.4385e-05, -1.3593e-04,
        -1.2656e-04, -1.5026e-04,  1.2024e-04,  1.7079e-04, -4.7918e-04,
         2.2320e-04, -1.4723e-04, -1.9817e-04, -3.9662e-04, -3.6404e-04,
         1.4334e-04, -1.0510e-04, -3.5550e-04, -3.6448e-04, -7.1852e-05,
         1.2943e-04,  7.8169e-05, -2.3688e-04,  2.0803e-04,  3.4075e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0712e-03, -9.9238e-01, -4.5339e-04,  2.7751e-04, -7.7279e-04,
        -5.7216e-04, -6.9701e-05,  3.8272e-04, -1.0943e-04,  1.5279e-04,
         2.0713e-04, -2.7895e-04, -3.3877e-05,  1.8079e-04,  4.6292e-04,
         2.3939e-04, -2.2001e-04,  2.6541e-04,  5.6474e-04,  5.0115e-05,
         8.7154e-05, -5.6507e-04, -6.0008e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.7989e-04,  9.9368e-01,  4.1144e-04,  5.3598e-07, -1.8670e-05,
        -1.2390e-04, -6.2246e-05,  1.4658e-04, -5.4089e-04,  2.8512e-04,
        -4.8237e-05,  1.2828e-04,  5.2737e-04,  1.8425e-04, -2.1233e-04,
        -2.2037e-04, -9.3745e-07,  3.0461e-04, -1.6200e-04,  1.5260e-04,
         2.0088e-04,  1.3787e-04, -6.0913e-06,  9.7139e-05, -1.4992e-04,
        -4.0512e-04, -2.7881e-04, -1.7855e-04, -6.7583e-05, -1.7543e-04,
        -3.1510e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2248e-04, -9.9307e-01, -1.8051e-04, -7.6002e-05, -2.7046e-04,
         3.3960e-05,  7.8251e-05,  6.5340e-05,  1.4191e-04,  2.4701e-04,
         4.6684e-04,  1.3077e-04, -1.5014e-04, -3.0063e-04,  9.4375e-05,
        -8.7257e-05,  1.4922e-04, -1.7730e-04,  2.4148e-05,  6.0114e-05,
         5.2585e-05, -1.0615e-04, -1.4342e-04,  2.7523e-04, -3.0719e-04,
        -1.2046e-04,  2.6938e-04, -9.2160e-05, -1.2509e-04, -3.7262e-05,
        -3.9753e-04,  4.1577e-04, -1.4111e-04,  1.7423e-04, -2.0192e-04,
        -4.8489e-04, -6.9365e-05, -8.5199e-05, -2.2443e-04,  1.4802e-04],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 7.8863e-04, -9.9043e-01,  6.8557e-04, -1.0464e-04,  5.3314e-04,
        -8.4801e-04,  3.4634e-04, -9.0456e-04, -3.6094e-04,  1.1533e-03,
        -7.2938e-04, -5.3856e-04, -2.7531e-04, -4.2810e-05, -5.1686e-04,
         5.4034e-05,  7.2833e-04,  4.1498e-04, -1.3168e-04, -4.1729e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3261e-04,  9.9162e-01, -1.8712e-03, -1.1833e-03, -2.9243e-03,
         6.3068e-04,  5.8406e-04,  3.5770e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2675e-04,  9.9391e-01,  8.8616e-04,  5.1262e-04,  8.3959e-04,
        -4.4786e-04, -3.6332e-04, -1.7960e-03,  1.0125e-03, -5.2803e-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>), tensor([ 6.1042e-04, -9.9426e-01, -1.3848e-04,  1.3024e-04, -6.6174e-05,
        -7.6106e-05,  1.9556e-04, -9.0029e-05, -1.4747e-04, -8.4276e-05,
        -1.6615e-04,  2.7819e-04, -6.6979e-05,  1.6727e-04,  6.4109e-05,
         9.1026e-05, -4.9812e-04,  1.2764e-04, -4.6810e-04, -1.5829e-04,
         1.8140e-05, -1.3266e-07,  3.6382e-04, -2.5162e-04,  7.0774e-04,
        -5.8577e-05,  3.8591e-04,  2.1016e-04, -6.3935e-05,  5.6137e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1749e-04, -9.9369e-01, -5.3554e-04,  6.1399e-05, -1.2101e-04,
        -9.5273e-05, -4.5868e-05,  4.1333e-04,  6.6402e-04, -1.8105e-04,
         3.8207e-05, -7.9097e-04, -2.2403e-04, -1.8310e-04, -3.0957e-04,
        -4.5793e-04,  6.0197e-04,  1.6260e-04, -2.1162e-04, -2.4554e-04,
        -7.1630e-05, -1.4197e-04,  3.4034e-04, -9.8551e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3457e-04, -9.9398e-01, -7.6296e-05,  1.6982e-05, -3.6517e-04,
        -4.2112e-04,  3.6582e-04, -2.8671e-04, -8.1080e-05,  5.1085e-05,
         3.2140e-04,  5.2537e-04,  3.2627e-05, -2.7049e-04,  3.8211e-04,
        -2.9355e-04, -4.1076e-04, -4.7311e-04, -3.6241e-04,  3.4591e-04,
         2.4260e-04, -1.8043e-04, -9.6390e-05, -1.9815e-04,  8.1850e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4684e-03, -9.8995e-01, -2.0183e-04,  5.9707e-04,  5.2893e-04,
        -5.3643e-04, -1.7621e-04,  2.4999e-04, -5.7878e-05, -2.4413e-05,
        -5.5524e-04, -3.0675e-04, -2.0802e-04, -3.1612e-04, -1.6213e-04,
         2.1166e-04, -2.0633e-04, -1.7433e-04, -2.6022e-04,  1.1288e-05,
        -1.9104e-04, -2.5614e-05, -1.1501e-04, -3.9026e-04, -4.2915e-04,
        -4.0608e-06, -2.8149e-04,  7.3195e-06, -1.9609e-04,  1.2063e-04,
         5.6043e-04,  1.7128e-04, -1.4102e-04,  4.0014e-04, -2.0138e-04,
        -2.1739e-04, -1.8915e-04, -8.4531e-05, -7.4765e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.2755e-04,  9.8964e-01,  6.4914e-04, -1.0977e-03,  1.6769e-03,
        -1.3601e-03, -1.3594e-03,  1.8140e-03,  6.9647e-04, -2.7085e-04,
        -2.1963e-04, -6.9061e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8761e-03,  9.9353e-01,  7.2510e-04, -6.4573e-04, -4.5767e-05,
         1.4940e-03,  1.9689e-04, -1.5130e-04,  2.6493e-04,  4.9774e-04,
        -5.7443e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1982e-04, -9.9253e-01,  5.9070e-04,  1.2389e-04, -1.5204e-04,
         4.7547e-04,  2.4422e-04,  1.4110e-04,  4.5449e-04,  1.9107e-04,
        -4.1549e-04,  1.2563e-04,  3.2318e-05,  2.2479e-04,  3.2021e-04,
        -3.2458e-04, -8.4803e-05,  8.7886e-04,  6.1038e-04,  2.4687e-04,
         5.8791e-05,  1.0749e-04,  3.0775e-04,  8.1502e-05, -6.5706e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5967e-03, -9.9179e-01,  4.2762e-04, -4.8888e-04, -1.7178e-04,
         3.8740e-04,  2.4647e-04, -2.4918e-04,  2.6927e-04,  5.9760e-04,
         1.4287e-04,  3.8599e-04, -3.3123e-04, -2.9929e-04, -1.5956e-04,
        -2.4543e-04,  6.3000e-04, -1.3168e-04, -2.1862e-04, -9.3723e-04,
        -7.4634e-05,  2.1779e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.8789e-05,  9.9066e-01,  4.1826e-04,  2.2228e-04,  2.5487e-04,
         1.0322e-04,  2.6104e-04,  3.7133e-04,  6.6796e-04, -7.3318e-05,
         7.8611e-04,  2.9498e-04, -8.1265e-05,  3.8330e-04, -5.8182e-04,
         2.9836e-04, -2.6951e-04,  1.9776e-04,  1.9824e-04,  1.8523e-04,
        -1.7536e-04,  3.8551e-06,  8.4639e-04,  5.0361e-04, -9.1076e-05,
         2.7340e-04,  4.8167e-04,  7.1033e-04,  2.2454e-04,  7.8837e-05,
        -2.1737e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #300: [tensor([-1.9075e-04,  9.9524e-01,  4.2167e-05, -7.0710e-05,  9.5875e-05,
         9.0985e-05,  8.5027e-05,  1.2058e-04, -3.7438e-06, -2.2799e-05,
        -8.7599e-05, -2.2480e-05, -2.9890e-06, -1.3043e-04,  1.3183e-04,
         1.2771e-04,  7.4536e-05, -5.8849e-05, -3.4254e-05,  1.6176e-04,
        -2.2563e-04, -1.3042e-04,  1.3830e-05,  9.5123e-05,  1.9810e-05,
         6.5094e-05,  1.9446e-06,  1.1367e-04,  1.4246e-05,  1.1270e-04,
        -1.3845e-04, -7.0458e-05, -1.2779e-04, -1.5609e-05,  1.0046e-04,
        -2.1898e-04,  9.1739e-05,  6.7655e-06, -1.1829e-04,  3.6671e-05,
         6.3980e-05,  2.0681e-04,  8.6281e-05,  6.5910e-05,  2.1907e-04,
         2.1390e-05, -2.3081e-05,  2.2475e-04,  1.4305e-05,  1.0467e-05,
         3.6089e-05,  1.0997e-04, -8.5391e-05, -9.9165e-05, -8.9871e-05,
         2.6857e-05,  1.0025e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.2954e-04,  9.9395e-01,  3.4580e-04,  4.7249e-05,  2.8662e-04,
         2.9803e-04,  5.2223e-04,  1.0791e-03, -1.6697e-03,  1.0703e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4207e-03,  9.9116e-01, -3.2921e-04, -5.6535e-04, -3.3511e-04,
         1.4190e-04, -2.3299e-04, -1.6765e-04, -2.8168e-05,  1.1009e-04,
         1.6103e-04, -2.5358e-04, -8.8978e-04, -9.0061e-05, -3.9763e-04,
         4.9347e-05, -3.4461e-04, -4.1956e-04,  2.8459e-04, -1.4039e-04,
         3.7471e-04, -9.6984e-05,  7.6802e-04,  5.3217e-05, -1.8552e-04,
        -1.4792e-06,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0841e-03,  9.9329e-01,  4.3134e-04, -6.9372e-04, -3.6578e-04,
         5.2238e-04,  1.7446e-04,  1.2446e-04,  1.0349e-03, -4.6172e-04,
        -3.4789e-04,  1.4712e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6817e-04,  9.9260e-01, -1.0246e-03, -4.4576e-04, -1.4440e-04,
        -4.4824e-04, -4.5823e-05, -2.6863e-05, -6.9306e-05, -5.5966e-04,
        -1.5526e-03,  9.6797e-05, -2.2089e-04, -5.8639e-04, -1.6852e-04,
        -4.0163e-04,  5.4340e-04, -3.7650e-04,  3.2525e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.5570e-04, -9.9281e-01, -1.7450e-04,  1.4011e-04, -1.9397e-04,
        -7.2632e-05,  2.5477e-04,  1.6627e-04,  4.6283e-04, -1.3835e-04,
         1.6911e-04, -3.2672e-05, -1.6791e-04, -6.3149e-04,  2.6041e-04,
         4.8799e-04,  4.5693e-04,  5.0690e-04,  4.0274e-04, -1.5366e-04,
         2.0258e-04,  2.4677e-04,  4.9246e-04, -3.5200e-04,  3.6491e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.2986e-04, -9.9086e-01, -5.1123e-05, -3.8670e-04,  5.8479e-04,
        -2.3133e-04, -7.3340e-04, -1.0569e-04, -3.3709e-04, -5.8880e-04,
         1.2477e-03,  7.3287e-04,  1.5981e-04, -4.8313e-04, -6.5642e-04,
        -6.6507e-04,  2.5937e-04,  3.0067e-04,  6.8217e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6310e-04, -9.9360e-01, -1.6897e-04, -2.7065e-04, -2.8277e-04,
         1.6953e-04,  4.3392e-05,  1.8992e-04, -1.6728e-04, -2.7919e-04,
        -2.6859e-04, -1.6601e-04, -1.0984e-05, -2.6800e-05,  2.9068e-04,
         1.1544e-04,  8.7896e-05, -1.5961e-04,  1.0449e-04, -1.1016e-04,
         2.1770e-04,  1.3036e-04, -2.3351e-04, -1.5083e-05, -8.0860e-05,
        -4.2536e-04, -1.7302e-04, -3.6426e-04, -2.2303e-04, -1.8847e-04,
        -1.1426e-05,  3.7843e-05,  1.0867e-04,  1.5450e-04, -1.2115e-04,
         9.8310e-05,  1.1632e-04,  5.5445e-05, -1.7235e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9046e-04,  9.9447e-01, -5.5524e-04,  1.7525e-03, -1.0397e-04,
        -1.2569e-03, -4.6949e-04,  5.0236e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.7988e-05,  9.9684e-01,  9.7408e-05, -4.5313e-04,  1.7388e-05,
         4.1030e-04,  3.7323e-04, -3.4490e-04, -6.5733e-04, -1.0585e-04,
        -1.5253e-04, -2.9875e-04, -1.6067e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1013e-03, -9.9247e-01,  1.6573e-04,  8.9540e-04, -5.9359e-05,
        -2.1685e-06, -2.4535e-04, -1.3669e-04,  3.9815e-06, -1.4115e-04,
         6.8284e-06,  2.8606e-04,  5.1043e-05, -4.0862e-04,  2.8570e-04,
         3.2828e-05,  7.9968e-05,  3.9759e-04,  1.6375e-04,  1.1123e-04,
         3.5620e-05,  1.0782e-04, -1.2344e-04,  3.8127e-04, -7.1133e-05,
        -1.7358e-04, -1.0730e-03, -2.7628e-04,  3.7746e-04, -3.3504e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3496e-03,  9.8915e-01,  7.9923e-04, -5.1154e-04,  1.2151e-04,
        -1.2952e-04, -1.4169e-04,  4.8129e-04, -9.7898e-05, -8.5275e-04,
        -1.0956e-04, -9.9236e-05,  9.5362e-05,  3.4304e-04,  9.0704e-04,
         4.1983e-04,  2.0871e-04,  3.6983e-04,  3.4719e-04, -4.6509e-04,
         2.1225e-04,  1.8658e-04, -1.8729e-05, -8.8352e-06,  7.0760e-04,
         3.2271e-04,  5.4406e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-8.5943e-06,  9.9393e-01,  2.7034e-04, -3.3108e-04, -9.4319e-05,
         3.1080e-04,  3.9968e-04, -5.2923e-04, -2.9424e-04, -2.1924e-04,
        -1.6075e-04, -5.8244e-05,  9.6913e-05,  1.5407e-04, -4.0442e-04,
        -3.5785e-04,  2.9610e-04, -2.9396e-04, -4.6214e-04, -3.6478e-04,
        -2.7447e-04,  6.5833e-05,  3.8589e-04,  2.4079e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3364e-03, -9.8980e-01, -2.1156e-04,  4.7667e-05,  8.6858e-05,
         7.9834e-05, -1.1481e-04,  4.3543e-04,  3.4159e-04,  4.5890e-04,
         5.9838e-04,  3.3478e-04,  4.7857e-05,  7.8070e-04,  3.0480e-05,
        -7.8284e-04,  2.2235e-04,  4.4624e-04,  4.3947e-04, -1.2443e-04,
         1.2126e-04,  4.8304e-04,  3.4431e-04, -7.2316e-05,  3.1907e-04,
         2.8436e-04,  3.4358e-04,  3.1492e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4532e-03,  9.9379e-01, -6.1792e-04, -3.2337e-04,  2.2128e-05,
         8.7355e-04,  6.8282e-04, -7.5791e-04,  3.9459e-04, -3.7750e-06,
         5.2134e-04,  9.8562e-05, -4.0597e-04, -3.6122e-05,  1.4136e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6146e-04,  9.9191e-01, -2.3473e-04, -2.3295e-04,  8.1646e-06,
        -4.7242e-04, -2.3975e-04, -3.4513e-04, -1.6966e-04, -3.8474e-04,
         4.3537e-05, -3.2288e-04, -9.1019e-05,  1.0029e-04,  1.5379e-04,
        -7.4515e-05, -1.3498e-04,  1.5609e-05,  3.2484e-04, -1.8691e-04,
        -2.1440e-04, -2.0957e-04, -3.1470e-04,  4.5880e-04, -6.7301e-05,
         1.9961e-04,  8.2481e-05, -3.9255e-04, -4.5560e-04, -2.8891e-04,
        -2.0690e-04, -1.2251e-04,  1.5157e-04, -1.3486e-04, -2.3032e-04,
        -2.3305e-04,  3.2669e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.3333e-04, -9.9339e-01, -7.7983e-05, -6.4783e-04, -3.0882e-04,
        -3.4053e-05, -6.3202e-05, -4.0540e-04,  5.0104e-04, -5.9297e-04,
        -6.4339e-04, -4.1018e-05, -1.1816e-04, -8.5348e-05,  3.5910e-04,
         8.8837e-05,  3.8968e-04, -2.4868e-04, -5.7966e-04,  3.6878e-04,
         7.3758e-05, -2.5232e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.5978e-04,  9.9273e-01, -2.4929e-04,  9.4496e-04, -1.3448e-04,
         6.2149e-04,  3.3214e-04,  1.3632e-03, -3.3000e-04,  1.2228e-03,
        -8.1110e-04, -4.1635e-06,  1.7003e-05,  4.7607e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.8084e-04,  9.9558e-01,  8.2293e-04,  4.9930e-04, -5.6579e-04,
         5.7983e-04,  6.3051e-05, -4.1503e-04, -2.6788e-04, -2.1739e-04,
        -5.1004e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.1088e-05,  9.9546e-01, -2.0856e-04,  3.5668e-04,  5.2275e-04,
        -7.8798e-04,  5.0686e-04,  1.1852e-04, -1.0817e-03,  8.1295e-04,
         9.4625e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.7249e-04,  9.9344e-01, -4.4708e-06, -5.6567e-04, -1.4273e-04,
         7.7435e-04,  2.3892e-04, -4.1598e-04,  4.8451e-04, -2.1144e-04,
        -1.1831e-05, -2.4935e-04, -5.6314e-04,  5.0487e-04,  8.8918e-05,
        -1.6327e-04, -7.9450e-04,  2.5012e-04, -2.1332e-04,  3.0657e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0842e-03, -9.9622e-01, -2.8422e-05,  7.2402e-05, -2.6924e-06,
        -7.6577e-05,  2.5344e-04, -4.6679e-05,  3.0244e-05, -1.0903e-04,
         2.0380e-04,  1.6732e-04,  3.7554e-04, -1.8846e-04,  1.7121e-04,
        -5.8977e-05, -3.7008e-05, -1.6479e-04,  1.6861e-04,  1.6186e-04,
         8.7408e-05,  1.6643e-04,  1.2017e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8962e-04, -9.9416e-01, -1.3163e-04, -5.6667e-04, -1.8009e-04,
        -1.1470e-04, -1.8316e-04, -2.5968e-04,  7.6540e-05, -1.7691e-04,
        -2.2384e-04,  3.3891e-05,  4.4654e-04, -3.2732e-04,  3.8348e-04,
         5.7182e-04, -1.6073e-04, -5.2825e-05, -5.6609e-06,  4.1843e-04,
        -4.6585e-04,  5.5282e-06, -1.0900e-05,  4.8482e-04,  6.4980e-05,
        -2.0219e-04,  2.7150e-05, -1.3167e-05,  6.1752e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1419e-04, -9.9236e-01, -5.9738e-05, -2.9681e-04, -5.9943e-05,
        -4.9163e-05, -6.7209e-05, -1.6217e-04,  9.5366e-05,  7.4373e-05,
        -1.6492e-04,  2.6912e-05, -2.0563e-04, -6.5323e-05, -2.8885e-04,
         7.4694e-05, -4.8008e-04, -1.6748e-04,  4.4910e-05,  3.0366e-04,
        -2.8546e-04,  2.1806e-04, -2.6347e-05, -6.3201e-04, -7.2140e-05,
        -2.4164e-04,  3.1245e-05, -1.5521e-04,  3.6606e-04, -2.2445e-04,
        -4.8689e-04,  8.6063e-05, -1.8091e-04,  1.4213e-04, -5.9264e-04,
        -6.6772e-04, -2.6728e-04,  1.6400e-04], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([-5.8001e-06,  9.9446e-01,  5.0928e-04, -1.0200e-04,  4.4312e-04,
         3.9066e-04, -2.1392e-05,  3.0668e-04, -7.4123e-04, -3.7223e-05,
        -5.9025e-04,  7.2621e-05,  7.8227e-05,  1.7253e-04,  2.3979e-04,
         6.4187e-04,  5.6154e-04,  2.4288e-04,  3.7858e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8847e-04,  9.9240e-01,  5.1979e-04,  4.4641e-04, -2.3949e-04,
        -6.2137e-05,  9.1425e-04, -6.8650e-04,  8.5164e-04, -4.2953e-04,
         4.5457e-04,  1.7023e-04,  2.6863e-04,  3.8961e-04,  9.9982e-04,
         2.9643e-04, -1.5408e-04, -2.7387e-04, -3.0356e-05, -2.2677e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3173e-03, -9.9022e-01, -1.0305e-03, -1.6556e-04, -1.4605e-03,
         3.5186e-04,  9.2775e-05,  1.7193e-04, -7.3185e-04, -5.8088e-04,
         2.2680e-04, -4.6253e-04, -1.8902e-04, -7.7519e-05, -8.6360e-05,
         4.7205e-05, -6.0360e-04, -7.6392e-04, -3.0379e-04, -1.0166e-03,
        -1.0060e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9481e-04,  9.9256e-01, -4.1328e-04,  5.8574e-04, -3.9967e-04,
        -5.7174e-04,  1.2548e-04,  1.0626e-04,  2.4823e-04, -5.0592e-04,
        -5.7866e-04,  1.9471e-04, -5.2359e-04,  6.8653e-06,  5.4064e-04,
        -4.2749e-05, -1.3782e-03, -3.0474e-04, -4.1998e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.0001e-04, -9.9173e-01,  2.3581e-04,  2.9631e-04,  2.3054e-04,
         1.5338e-04, -4.0437e-05,  1.3339e-04, -2.5978e-04,  1.2540e-04,
         2.3615e-04,  1.4175e-04, -4.3772e-05,  6.0078e-05,  1.4376e-05,
        -4.1761e-05,  4.3109e-05, -1.7840e-04, -3.3618e-05, -6.1127e-05,
         2.7000e-04,  5.0005e-05,  9.2112e-05,  1.2782e-04,  2.9909e-04,
        -2.3826e-04,  9.0767e-05,  3.2784e-04,  3.8039e-05, -2.9431e-04,
         1.3232e-04,  1.3908e-04,  3.4841e-04,  2.3855e-04,  4.9545e-05,
         4.7914e-04,  1.4848e-04,  2.2015e-04,  1.7107e-04,  2.2277e-04,
         3.4442e-04,  3.3237e-04, -2.4013e-04, -9.1080e-05,  6.6877e-05,
        -9.1243e-05,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.0937e-04, -9.9437e-01,  2.7387e-04,  1.3335e-04,  6.9352e-05,
        -2.0448e-04,  2.8015e-05, -2.9271e-04, -3.1501e-04, -8.3658e-05,
        -1.2453e-04, -2.3310e-04, -1.9853e-04, -2.0341e-04, -3.0316e-04,
        -8.3733e-05, -1.4639e-04,  1.6639e-04,  8.2787e-05, -1.7145e-04,
        -1.8046e-06,  3.5153e-04,  2.8491e-05, -2.9823e-04, -9.9457e-05,
        -3.2986e-04, -4.6036e-04, -1.3321e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2574e-04,  9.9400e-01,  8.6617e-05, -7.6746e-05, -1.3571e-04,
        -6.7924e-05,  2.2662e-04,  3.3983e-04,  3.7577e-05, -1.3155e-04,
        -4.2272e-04, -1.2615e-05,  3.1979e-04, -1.9850e-04, -2.6250e-04,
        -5.7489e-05, -4.0203e-04,  9.4652e-06,  1.8280e-04, -4.7168e-04,
        -4.7677e-04,  2.2540e-04, -1.2259e-04,  1.4520e-04, -1.3440e-04,
        -1.8420e-04, -9.4912e-05,  1.3649e-04,  2.1313e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5176e-03, -9.9007e-01,  3.4042e-04, -1.4337e-04,  2.0653e-04,
         6.0332e-04,  5.0260e-05, -8.2220e-05, -1.8020e-04, -2.5729e-04,
        -7.2218e-05,  3.4648e-04,  3.8852e-04,  2.4090e-04, -6.5097e-04,
        -4.4607e-04, -2.4069e-04, -5.4382e-04, -2.3290e-04,  1.4872e-04,
         5.1531e-05, -3.4784e-04, -2.6585e-04,  1.1854e-04, -3.8555e-04,
        -8.3546e-04, -1.2467e-05, -2.2012e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4628e-04, -9.9321e-01, -8.8015e-05, -1.4859e-04, -2.1726e-04,
        -7.3810e-05,  5.4787e-05,  8.8611e-06, -4.2009e-04,  3.4810e-04,
         3.6120e-04, -8.0836e-05,  2.5518e-04, -8.5991e-06,  3.3197e-04,
         2.9929e-04,  9.8615e-05,  1.6829e-05,  2.8752e-04,  4.3188e-05,
         9.2760e-05, -4.9570e-04,  2.5653e-05,  5.5715e-05,  1.4644e-04,
         2.5419e-04, -1.5193e-04, -2.4992e-04,  1.6943e-04,  7.1988e-05,
         1.1509e-04, -2.9201e-05, -3.9870e-05,  6.7120e-05, -2.1223e-04,
         2.2537e-04,  1.0798e-04, -1.4501e-04,  3.2413e-05,  9.6612e-05,
        -1.2636e-04, -1.4418e-04, -4.5044e-05, -1.9218e-04,  9.2030e-05,
        -1.1614e-06,  1.1519e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.8245e-04, -9.9111e-01, -3.3061e-04, -1.6198e-04, -2.6816e-04,
         9.6046e-05, -1.5003e-04, -1.3586e-04, -1.9066e-04, -1.1625e-04,
        -1.4539e-04, -1.7374e-04, -1.8026e-04, -6.1381e-05, -1.1297e-04,
        -1.6077e-04, -5.0852e-06, -3.3873e-04, -2.4782e-04,  1.6781e-06,
        -2.8498e-04,  2.7099e-04,  9.9799e-05, -2.1256e-04, -1.0962e-05,
        -2.5616e-04,  9.8554e-05, -2.1423e-04, -1.5764e-04, -3.0677e-04,
         8.1832e-05, -2.3016e-04,  3.4306e-05,  2.4179e-04,  1.3581e-04,
        -7.5759e-05, -2.2209e-04, -8.9153e-05,  1.4220e-04, -1.9751e-04,
        -2.5750e-04, -3.8627e-04, -3.4509e-04, -3.0030e-04, -3.1808e-04,
        -1.6219e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.6718e-05, -9.9426e-01, -3.1934e-04,  1.0282e-04, -2.2390e-04,
        -5.4323e-04, -2.9805e-04, -3.4752e-04, -5.5786e-04, -1.7571e-05,
         3.0583e-06,  5.2490e-05,  7.5007e-05,  7.5357e-04,  2.3906e-04,
        -2.2452e-04, -2.8116e-04, -7.4207e-05, -6.6109e-04, -3.9771e-04,
        -2.1813e-04,  2.8322e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7291e-04, -9.9434e-01,  3.5487e-04,  3.7666e-04, -1.8565e-04,
        -9.9431e-05,  1.8117e-04, -1.7332e-04,  2.1169e-05,  1.5148e-05,
        -1.4693e-05,  5.1555e-04,  1.4104e-04,  4.4648e-05,  5.9751e-05,
         2.5872e-04,  7.0695e-04,  8.1893e-05,  2.6763e-04, -4.8453e-05,
         2.0411e-04,  9.9576e-05, -1.3990e-04, -4.2874e-04,  1.0247e-04,
        -1.7855e-04,  1.2529e-04,  9.8876e-05,  8.8208e-05,  7.8235e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
