Iter #50: [tensor([ 1.3340e-01, -5.2343e-01,  5.4916e-03,  3.4154e-03,  1.8527e-03,
        -1.1095e-02,  8.4534e-03, -1.1430e-02,  1.5404e-02, -5.6266e-03,
        -1.1034e-02,  8.7263e-03,  1.5402e-04,  1.1610e-02, -1.7026e-03,
        -6.1962e-04, -1.8996e-02, -1.6124e-02, -2.6410e-03,  3.4200e-03,
        -1.2259e-01, -1.8705e-02,  3.0778e-03, -3.8776e-03,  9.2963e-03,
        -2.5649e-03,  3.9345e-03,  3.0325e-03,  7.1437e-03, -1.1878e-02,
        -1.9277e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.2631,  0.1748,  0.0111,  0.0110, -0.0258,  0.1290,  0.0078, -0.0082,
         0.0093, -0.0112, -0.0155,  0.0017,  0.0006,  0.0890,  0.0180, -0.0180,
        -0.0135, -0.0235,  0.0077,  0.0093, -0.0109,  0.0892,  0.0131, -0.0083,
        -0.0271, -0.0029, -0.0003,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.1379, -0.2603, -0.0318, -0.0255,  0.0223, -0.1768, -0.0214,  0.0316,
        -0.0196,  0.0268,  0.0281, -0.0003, -0.0156, -0.1431,  0.0104,  0.0219,
        -0.0079,  0.0028, -0.0159,  0.0000,  0.0000,  0.0000,  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([ 4.1782e-01,  1.9547e-01,  1.1612e-02,  1.6291e-02, -2.5092e-02,
         1.2048e-01,  9.6641e-03, -1.1513e-02, -7.2628e-03, -1.1748e-02,
        -4.3035e-03,  1.1683e-02, -4.5357e-03,  8.7002e-02,  1.6206e-03,
        -4.5895e-03, -7.9818e-03,  1.0799e-02,  2.4727e-02, -1.5661e-02,
        -1.4392e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2117e-01, -1.4840e-01, -8.6625e-03, -2.8091e-04,  6.8516e-04,
         1.7382e-02, -9.8139e-02, -2.3104e-02, -2.3056e-02, -5.7624e-04,
         2.4432e-02,  7.7942e-03, -4.2995e-03,  1.1611e-02, -2.3631e-02,
        -1.0124e-01, -1.2659e-02,  5.6781e-03, -2.5381e-03,  7.3450e-03,
         9.7985e-03, -5.1855e-03,  3.8300e-03, -2.8347e-03, -4.7763e-03,
        -8.8737e-02, -1.5462e-02, -4.8369e-03,  1.7968e-02,  3.8816e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.1738, -0.1870, -0.0220, -0.0064, -0.0017,  0.0101, -0.1190, -0.0209,
        -0.0123, -0.0026,  0.0317,  0.0008, -0.0290,  0.0020, -0.0339, -0.1143,
         0.0110,  0.0060, -0.0073,  0.0120,  0.0053,  0.0124,  0.0124, -0.0197,
         0.0043, -0.0052, -0.1050, -0.0042, -0.0100,  0.0155, -0.0020,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1436e-01, -2.1991e-01, -2.4067e-02, -2.0303e-02,  3.5474e-05,
         9.7928e-03, -1.4864e-01, -3.3359e-02, -2.6874e-03, -2.7243e-02,
         2.7822e-02,  1.0751e-02, -1.1231e-02,  1.6158e-02, -9.3792e-04,
        -1.6961e-01, -6.8226e-03,  9.9828e-03, -1.7835e-02,  3.2566e-02,
        -2.1112e-02, -5.7402e-03,  3.8507e-03, -1.7070e-02,  2.8864e-02,
        -1.0522e-03, -1.8196e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0403, -0.4972,  0.0079,  0.0272,  0.0239,  0.0224, -0.0074,  0.0530,
        -0.0159,  0.0037,  0.0084,  0.0117, -0.1024,  0.0341,  0.0023,  0.0250,
        -0.0962,  0.0100,  0.0099,  0.0011,  0.0000,  0.0000,  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([-0.3314, -0.3201, -0.0029,  0.0094,  0.0014,  0.0150, -0.0137,  0.0260,
        -0.0061, -0.0126, -0.0032, -0.0009, -0.0726,  0.0117,  0.0143,  0.0095,
         0.0038,  0.0044, -0.0711,  0.0072,  0.0058,  0.0038,  0.0029,  0.0020,
         0.0082,  0.0041,  0.0127,  0.0030,  0.0159, -0.0042,  0.0000,  0.0000,
         0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2074e-01, -4.1200e-01, -9.3288e-03,  1.4617e-02,  1.2442e-04,
         3.2967e-02, -3.1852e-02,  3.6548e-02, -6.4644e-03, -2.9407e-02,
         9.4358e-03,  1.7723e-03, -1.0143e-01, -1.0954e-03, -7.2841e-05,
         3.0020e-02, -1.0404e-02, -1.8574e-02,  2.4383e-03, -1.5019e-02,
        -1.1738e-02,  3.9492e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7906e-01,  1.6237e-01,  1.5017e-02,  7.4703e-03, -2.0312e-02,
         1.2128e-01,  1.7962e-02, -2.7353e-02,  8.9952e-03,  3.1250e-03,
        -1.4118e-02,  9.1707e-02,  4.4439e-03, -7.4392e-03, -8.2629e-03,
        -2.1175e-02,  1.9165e-02,  7.9107e-03,  1.9071e-03,  1.2646e-02,
         1.7648e-03,  4.2760e-03,  1.0866e-02,  1.9460e-02,  9.9661e-02,
        -1.1491e-02, -1.6725e-04, -5.9501e-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([ 0.1817, -0.1291, -0.0029,  0.0024,  0.0176, -0.1094, -0.0066,  0.0339,
        -0.0068,  0.0006,  0.0183, -0.0859,  0.0105,  0.0004, -0.0081,  0.0107,
        -0.0031, -0.0017,  0.0154, -0.0752, -0.0055,  0.0239, -0.0067, -0.0113,
        -0.0015, -0.0923,  0.0022,  0.0068,  0.0028, -0.0869, -0.0021, -0.0210,
         0.0127,  0.0039], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 4.2065e-02, -8.9315e-01, -2.6133e-03,  3.0498e-03,  8.6420e-03,
        -1.5550e-04,  6.7636e-03,  4.1057e-04,  9.9637e-04,  1.0187e-03,
         3.5981e-03,  8.1695e-03, -2.0656e-04, -1.6901e-03,  1.1988e-04,
         6.0874e-03,  9.9525e-04,  4.4730e-03, -9.6624e-06, -3.0890e-04,
        -2.5632e-03, -9.8353e-04,  5.0229e-03, -1.0805e-04,  4.8458e-03,
        -1.2044e-03,  7.4613e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3299e-02, -9.0085e-01,  8.0807e-04, -6.9496e-04,  1.0616e-02,
        -2.2208e-03,  6.7026e-03,  1.6481e-03,  1.2443e-03,  5.6279e-05,
         7.7844e-03,  5.5231e-03,  1.0024e-03, -2.5227e-03,  4.1536e-03,
        -1.2067e-03,  2.1608e-03, -1.2833e-03,  5.0579e-03,  9.8240e-05,
         3.3291e-03,  1.3631e-02,  2.1461e-03,  9.3888e-04,  1.0251e-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([ 3.3707e-02, -9.2191e-01, -2.6789e-04,  1.2108e-03,  8.7375e-03,
        -1.1707e-03,  1.4814e-03, -2.2129e-03, -7.4579e-06, -6.5213e-04,
         8.4423e-03, -1.9767e-05,  9.3003e-04,  8.0425e-05,  9.1341e-04,
         1.6539e-03,  1.0522e-03,  2.3432e-04, -2.4437e-04,  3.0727e-03,
         1.0827e-03,  1.4655e-03,  8.9781e-04,  2.7178e-03,  2.1981e-03,
         4.8242e-04,  3.6827e-04, -2.3997e-03,  3.8394e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7293e-01, -5.4100e-01,  3.9934e-03,  8.7235e-03,  3.0790e-03,
        -8.6358e-03,  5.1120e-03, -2.1484e-03, -5.4686e-03, -1.5836e-03,
        -1.1054e-03,  3.8465e-03,  6.6742e-03,  7.1554e-03, -1.2857e-02,
         2.2653e-03, -5.1412e-03, -2.7029e-03,  5.3918e-03, -1.7608e-03,
        -9.0267e-03, -1.8809e-03,  2.8290e-03, -5.5845e-03,  3.3390e-03,
        -2.5866e-03, -6.9982e-03, -1.1598e-02,  2.1704e-04, -1.4409e-02,
        -1.7769e-03, -2.0988e-02,  5.5745e-04,  1.5738e-02, -4.1870e-03,
         2.2295e-02, -8.8691e-03, -7.3771e-03, -1.0434e-02, -1.3050e-03,
        -4.7711e-03,  8.1914e-03,  1.5963e-03, -7.3031e-03, -3.3935e-03,
         1.1362e-03,  5.1671e-03, -1.3058e-03, -1.1108e-02,  2.4627e-03,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5856e-02, -6.7171e-01, -9.4298e-03, -1.6640e-03, -6.8566e-03,
        -5.3966e-03,  1.0169e-02,  3.9286e-04,  1.0113e-03, -1.1727e-02,
        -4.3027e-03,  1.9073e-03, -1.8794e-03,  1.9643e-03, -6.6693e-03,
        -5.1200e-03,  5.1643e-03,  9.8098e-04,  3.5964e-03, -1.3537e-03,
        -4.3944e-03, -1.0189e-02,  2.5649e-03, -9.1670e-03, -4.9685e-03,
         3.6521e-03, -1.0872e-03, -1.3471e-02, -6.7030e-03, -2.1809e-02,
        -9.6790e-03, -4.0757e-03, -3.9817e-04,  9.6060e-03, -1.1325e-02,
         1.2653e-02, -1.0924e-02, -1.2349e-02, -8.3459e-03, -1.9043e-03,
        -2.5632e-03, -2.9618e-03, -8.2212e-03, -5.3053e-03, -1.5041e-03,
        -4.8585e-03,  7.2321e-04, -2.8680e-03, -1.1501e-03,  3.4211e-03,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4344e-03, -6.5881e-01,  7.1359e-03,  6.9923e-03, -7.8591e-03,
        -4.3170e-03,  4.8731e-03,  4.9664e-03, -5.9841e-03, -7.0561e-03,
         2.7490e-04,  3.4832e-03,  4.2548e-04,  3.3190e-03, -5.5683e-03,
         5.5608e-03, -3.5135e-03,  4.1250e-03,  3.4972e-03,  2.0178e-03,
        -1.7572e-02, -6.9980e-03,  1.4029e-02,  7.9804e-04,  7.2362e-03,
        -3.0701e-03, -4.6726e-03, -2.9316e-03,  1.0605e-02, -2.5132e-02,
        -3.7029e-03, -1.9973e-02,  2.6424e-03,  1.0296e-02, -1.8059e-02,
         1.2827e-02, -1.0147e-02, -1.9998e-02, -1.5885e-02, -1.0940e-03,
         2.8397e-03,  2.6459e-03,  6.2522e-03, -7.0386e-03,  9.7776e-03,
        -6.9663e-04,  3.4783e-03,  6.3522e-04, -9.9387e-03,  4.0823e-03,
        -7.3710e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.4304e-02, -9.0102e-01, -1.6281e-03, -2.1840e-03,  2.1548e-04,
        -6.9314e-03,  1.4841e-03,  2.9607e-03,  3.7418e-04,  1.1996e-03,
        -5.1814e-03, -3.6785e-03,  8.5077e-03, -9.0567e-04, -3.4935e-04,
         1.3436e-03, -5.4145e-03, -3.4766e-03, -2.8602e-03,  4.6362e-04,
        -1.9026e-03,  2.3578e-03,  4.9644e-03, -6.0737e-04,  1.9237e-03,
         1.9921e-03,  4.3370e-04, -2.2069e-04, -5.2499e-04, -5.9142e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7605e-02, -8.5863e-01,  5.7007e-04, -3.0066e-03, -5.9225e-04,
        -9.3683e-03, -3.3544e-03,  2.1429e-03,  3.2813e-03,  7.4086e-03,
        -9.8948e-03, -9.6588e-04,  6.0740e-03, -2.4384e-04, -4.7419e-03,
         4.0857e-03,  2.0622e-03, -3.1620e-03, -1.3324e-03, -5.0634e-03,
        -4.7862e-04,  4.0662e-03, -1.9294e-03, -4.7317e-03,  8.1683e-04,
         8.8253e-04,  2.9397e-03, -5.7574e-04, -9.8152e-04,  5.1712e-04,
         2.0161e-02, -4.9101e-03,  3.4271e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.8443e-03, -9.3500e-01, -2.2125e-03,  1.0519e-03,  3.2116e-03,
        -8.9085e-03, -2.9280e-03, -4.9343e-04, -8.9343e-04,  1.7565e-03,
         2.2309e-03, -1.3694e-03,  1.7818e-03, -2.1655e-03, -2.0538e-03,
        -2.1670e-03, -1.8223e-03, -1.0142e-03,  1.1143e-03, -4.7188e-03,
         2.8045e-03,  3.8565e-04, -2.4907e-04, -2.3675e-04,  3.1654e-03,
         1.2567e-03,  1.7684e-03, -2.5926e-03,  1.2757e-03,  1.2489e-03,
         2.7999e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.8993e-02, -8.6258e-01, -8.6027e-03, -2.3823e-03,  2.0982e-03,
         6.6199e-04,  6.9411e-03,  1.1877e-02, -1.3161e-03, -6.4183e-03,
        -2.3778e-03,  3.4806e-04,  1.6674e-02,  3.2158e-03,  9.5374e-03,
        -2.9195e-03, -7.2341e-03, -1.2707e-03,  3.4283e-03,  2.6923e-03,
         2.0701e-03,  1.3411e-03,  1.9149e-03, -3.0251e-03, -7.7435e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6726e-03, -9.1306e-01, -5.5799e-03, -2.4877e-03,  6.5370e-04,
        -3.8804e-03,  5.9137e-03,  8.2522e-03, -1.8932e-03, -3.6312e-03,
         2.1051e-03,  2.5275e-03,  7.7930e-03, -1.6991e-03,  9.6014e-03,
        -6.6700e-03, -1.6735e-03,  2.6953e-04, -4.4500e-03, -4.5410e-03,
        -1.5062e-03,  4.4336e-03, -2.3342e-03, -2.3674e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1142e-02, -8.6070e-01, -6.7880e-03,  1.6243e-03,  3.8263e-03,
        -2.6461e-03,  4.9162e-03,  9.1864e-03, -4.9509e-03, -4.1855e-03,
         2.0188e-04,  2.1510e-03,  1.0205e-02,  3.6313e-03,  4.9946e-03,
        -1.3411e-03,  3.9252e-04,  3.5691e-03,  4.2507e-04, -1.9010e-03,
         1.2476e-03,  7.1867e-04,  1.6259e-03, -1.5277e-03,  6.5862e-03,
         2.7919e-04,  3.5413e-03,  9.1389e-04, -2.5373e-03, -1.2605e-03,
        -9.8719e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #150: [tensor([-4.4724e-02,  9.0639e-01,  2.4146e-03,  2.3381e-03, -4.0945e-03,
        -5.3655e-03,  4.4102e-03, -1.0426e-03,  7.9564e-04,  2.0327e-03,
        -3.2725e-03,  1.1637e-03, -1.0180e-03, -2.2873e-03,  2.4117e-03,
        -1.8847e-03, -2.6296e-04, -2.5460e-03, -4.8258e-03,  1.3478e-03,
        -9.3439e-04,  1.0693e-03, -1.8483e-03, -5.4620e-04, -9.7465e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0541e-02,  9.5850e-01, -3.8549e-03, -7.3300e-04,  2.6393e-03,
        -8.0630e-03, -1.1352e-03,  1.3563e-04, -7.9833e-04, -1.7426e-03,
        -1.5291e-03, -2.4060e-03, -4.0390e-04, -7.8500e-04, -1.8675e-03,
        -9.4235e-04, -3.4977e-03, -4.2425e-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([-7.0604e-04,  9.6270e-01,  1.8814e-03, -1.7249e-03, -7.3983e-04,
        -8.1829e-03, -2.7377e-03, -2.6209e-03,  7.2255e-04, -6.8433e-04,
        -3.7498e-03,  7.5693e-04, -8.5590e-04,  1.4663e-03, -3.1441e-03,
        -2.0055e-03, -6.6708e-04, -4.1589e-03, -1.2185e-04, -3.7574e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9286e-03,  9.7474e-01, -3.3623e-03, -1.3630e-03,  9.8097e-04,
        -1.9717e-03, -1.2736e-03, -7.8353e-04, -1.1509e-04,  1.4486e-03,
        -3.7013e-05, -7.7864e-04, -7.6482e-04,  1.2954e-03, -2.6317e-04,
         2.6749e-03,  2.5236e-03,  1.5545e-05, -1.1623e-03,  3.1303e-04,
         1.2063e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3064e-02,  9.2979e-01, -1.9007e-03,  5.5744e-03,  1.1294e-03,
        -1.9873e-03,  3.0255e-03,  2.8321e-03, -1.7846e-04,  1.8371e-03,
         1.8753e-03,  3.5955e-03,  4.7637e-04,  4.6861e-03,  2.4992e-03,
         3.3680e-04,  2.9041e-03, -1.3968e-03,  9.1520e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.4335e-02,  8.7945e-01,  5.0477e-03, -7.2409e-04,  4.7414e-04,
        -2.1509e-03, -4.8907e-04, -2.4393e-03,  5.8153e-03,  4.4967e-03,
        -8.0419e-04,  4.0200e-03,  4.1870e-03,  1.0519e-03,  4.4004e-03,
        -2.6891e-03, -6.1813e-04,  2.8106e-03,  2.0952e-03,  1.8976e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6227e-02,  9.1112e-01,  6.3484e-03,  3.0945e-04, -2.5792e-03,
        -2.3128e-04,  4.1006e-03, -3.9119e-03,  1.7473e-03,  1.2096e-03,
         4.2318e-03,  1.1909e-03,  7.2003e-03,  3.4150e-03,  2.4609e-03,
        -3.4678e-03, -2.4911e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1137e-02,  9.3347e-01, -3.4593e-03,  6.0723e-04, -1.3717e-03,
        -4.9311e-03, -6.5891e-03,  1.6376e-03,  7.0872e-04,  1.9749e-03,
         7.8757e-05,  9.8419e-04, -2.1564e-04,  3.0101e-04,  1.1515e-05,
        -1.4454e-03, -8.3369e-05,  9.9455e-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([-1.2522e-02,  9.5638e-01,  1.1376e-04,  5.7106e-03, -1.8988e-03,
         1.9465e-03, -6.8322e-04,  2.5881e-03,  4.7952e-03, -3.7268e-03,
         3.4977e-03, -2.0457e-03, -3.0462e-03, -9.2561e-04, -1.2309e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5797e-02,  9.1408e-01,  1.8232e-03,  1.4227e-04, -2.3748e-03,
        -2.2226e-03,  1.4355e-03,  2.0973e-03, -1.9054e-03,  9.3632e-04,
         1.2586e-03, -1.3919e-03,  1.2487e-03, -3.0631e-03,  2.1088e-03,
        -2.1076e-03, -2.4280e-04, -5.0816e-04,  6.4840e-04, -1.5564e-03,
        -5.5411e-04, -2.4936e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7369e-02, -9.5577e-01,  1.6095e-03,  8.4617e-04, -3.4775e-03,
         1.7260e-03, -1.1739e-03, -1.1631e-03,  2.2764e-03,  2.7966e-03,
         9.0570e-04, -1.4314e-03, -1.5083e-03,  7.6137e-04, -8.8363e-04,
        -1.9056e-04, -2.4643e-03,  9.7663e-04, -1.0800e-04, -4.1248e-04,
         1.2194e-03,  2.5283e-04,  6.7658e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.5074e-02,  9.4000e-01, -4.4553e-03, -2.7026e-03, -1.5314e-03,
        -3.9754e-03, -3.4652e-03, -2.1555e-03, -2.5343e-03, -3.2705e-03,
        -1.5474e-03,  3.2686e-05, -2.7322e-03,  7.2933e-04, -1.1075e-03,
         1.5082e-03,  9.7157e-04, -1.6355e-03, -1.7284e-03, -4.3031e-03,
        -1.9205e-03,  2.1282e-03,  4.9131e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 4.1645e-03, -9.8277e-01,  1.5884e-03,  5.8615e-04, -2.8635e-04,
         7.0356e-05,  1.7018e-04, -6.2067e-04, -1.7753e-04, -8.1036e-04,
        -7.5052e-04, -1.0188e-03, -3.1357e-04, -2.6181e-04,  1.1730e-03,
         8.4228e-04, -7.8906e-04,  5.7559e-04,  3.5722e-04,  2.9464e-04,
        -8.3962e-04,  1.0476e-03,  4.9444e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8106e-03, -9.8463e-01,  1.3112e-03,  1.7594e-03, -5.7592e-04,
         9.2591e-04,  1.2413e-04,  1.8454e-04,  1.2837e-03,  8.7521e-04,
         9.7838e-04, -3.6891e-04,  1.6303e-03,  4.2611e-04,  8.2890e-04,
         1.9646e-05,  9.4798e-05, -1.5307e-03,  6.4292e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0164e-03, -9.8562e-01,  1.0233e-03, -2.2433e-04,  5.7159e-04,
         5.9277e-04, -3.0762e-04, -9.9478e-04, -1.2567e-04,  4.1998e-04,
        -2.9105e-04, -9.1326e-04,  1.6696e-05, -1.0322e-04,  1.2021e-03,
        -1.4633e-03, -4.3304e-04,  5.8758e-04,  1.3663e-04, -3.0637e-04,
         2.2897e-04,  1.3844e-05, -4.0409e-04,  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, -9.8855e-01,  5.1775e-04,  3.1433e-04,  4.9781e-05,
        -5.6746e-04, -4.0344e-04,  4.2546e-04,  3.1803e-04, -5.5644e-05,
         1.7554e-03,  2.4474e-04, -6.3305e-04,  7.8127e-04,  5.7262e-05,
         1.0364e-03, -2.6311e-04,  8.8275e-04,  3.3422e-04, -3.1977e-04,
        -2.6489e-04,  4.3082e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9159e-03, -9.7592e-01,  1.7675e-03, -1.1870e-04,  5.4516e-04,
         6.9363e-04, -4.4258e-04, -1.3026e-04,  1.0502e-03,  9.8799e-04,
         7.8559e-04,  3.6692e-04, -4.3897e-04,  1.0181e-03,  9.7732e-04,
         2.8137e-04,  1.8967e-03,  1.2668e-03,  1.4444e-03,  5.9583e-04,
         2.0017e-03,  1.1738e-03, -1.8373e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7465e-03, -9.8218e-01,  1.2491e-03,  1.2658e-03,  4.6495e-05,
         8.5971e-04, -5.5358e-04,  4.1485e-05,  1.6799e-03,  1.2431e-04,
         1.5665e-03, -1.7201e-04, -1.0342e-03,  5.8900e-04,  9.2084e-04,
         5.5533e-04,  3.6022e-04,  8.2174e-04,  1.2254e-03, -9.9224e-05,
         4.3930e-04,  8.5687e-04,  2.0543e-04, -4.0827e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.0275e-03, -9.7793e-01,  1.8951e-03, -2.5924e-04,  1.2123e-03,
         1.2366e-03,  2.3056e-03, -5.7014e-05,  1.6853e-03,  7.7589e-04,
         4.9600e-04,  2.3534e-04,  6.3864e-04,  8.1542e-04,  1.5766e-03,
         6.1699e-04,  3.1163e-04,  1.1427e-04,  1.1707e-03,  6.3827e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2952e-03, -9.8340e-01,  1.0024e-03,  9.5177e-05, -4.3923e-04,
         1.6480e-03,  1.2164e-03, -5.0155e-05,  1.0268e-03, -5.1919e-04,
        -1.9284e-04,  3.9686e-04,  5.0283e-04, -1.5667e-04,  1.0518e-04,
        -1.5011e-03, -1.0760e-03, -6.7517e-04,  5.3579e-04,  1.1647e-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([-4.7473e-04,  9.8624e-01,  3.5463e-04, -4.1032e-04,  1.0796e-04,
         4.9877e-04, -7.0763e-04, -4.1450e-05,  8.1353e-04,  2.0088e-03,
         5.0471e-04,  1.0060e-03,  1.1243e-04,  1.5082e-03, -3.0973e-04,
         1.1703e-03,  1.6952e-03,  7.2030e-04,  3.2221e-04, -9.9654e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.9519e-04, -9.9050e-01,  3.6394e-04,  7.7108e-05,  6.5803e-04,
         1.1616e-03, -3.2595e-04, -3.7769e-04, -2.9914e-04, -2.7099e-04,
         2.3105e-04, -1.4591e-04, -4.6085e-04,  5.8053e-04, -1.7386e-04,
        -5.6225e-04,  2.1493e-04, -1.5473e-04, -9.9268e-04, -2.7069e-04,
        -8.3508e-04,  3.4395e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.2521e-04,  9.8605e-01,  5.4557e-04, -1.1304e-03, -7.4883e-04,
        -1.4239e-04, -2.2884e-04, -3.1251e-04,  4.3740e-05, -1.1555e-03,
         4.2025e-05, -1.1304e-03, -1.3712e-03, -7.2277e-04, -4.2580e-04,
        -1.2248e-03,  1.6395e-04, -9.8083e-06, -1.4628e-04, -1.6984e-06,
        -9.5124e-04, -8.9942e-04,  1.5212e-04, -2.2213e-04, -4.8445e-04,
        -1.0637e-03, -2.7614e-06,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7253e-03, -9.8421e-01,  2.1972e-04,  6.5149e-04,  5.0919e-04,
        -3.2686e-06,  9.8737e-04,  4.5557e-04, -1.3212e-04,  9.9601e-04,
         7.8905e-04,  4.2489e-04,  1.8164e-03,  4.4971e-04,  8.3508e-06,
         4.0023e-04, -7.1857e-05,  9.7243e-04, -7.2389e-04,  2.9067e-05,
         7.1716e-04,  3.8880e-04,  1.3250e-04,  3.0486e-04,  3.4765e-04,
         1.2568e-03,  4.4045e-04, -8.3852e-04], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #250: [tensor([ 3.5381e-03, -9.8733e-01,  3.9880e-04,  3.4012e-04,  8.6313e-04,
         1.0389e-04,  2.5596e-05,  2.1377e-04, -4.5254e-05, -7.1054e-04,
         3.5958e-04, -3.9180e-04, -2.3170e-04,  8.1301e-04,  6.7347e-04,
        -1.8676e-04, -2.9800e-04, -3.4373e-06, -5.2555e-04,  6.7805e-05,
        -3.0487e-04,  4.9832e-05,  2.8733e-04,  2.5626e-04,  1.5583e-04,
        -5.3443e-05,  7.0372e-04,  2.8574e-04,  1.7525e-04,  6.0644e-04,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9970e-03, -9.8502e-01,  1.3310e-03, -2.3263e-04, -1.1191e-03,
        -3.9683e-04,  1.1491e-04,  1.4338e-04,  4.1530e-04,  8.5953e-04,
         6.9099e-04, -1.9626e-04,  5.2433e-04,  2.3375e-04, -1.3023e-03,
         6.3855e-04,  3.5044e-04,  3.3174e-04,  4.0804e-04,  5.2441e-04,
        -5.7505e-04, -5.5645e-04, -3.1777e-04, -7.2261e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7914e-03, -9.8421e-01,  1.1401e-03, -1.1670e-04,  1.1721e-03,
         3.8868e-04,  7.1927e-05,  1.3130e-03,  4.4095e-04,  3.0314e-04,
         1.1162e-03,  3.7864e-04,  2.8027e-04,  3.9426e-04, -9.2571e-04,
         2.4093e-04,  4.8798e-04, -2.9276e-04,  8.6714e-04,  3.6642e-04,
         1.3221e-04,  1.7661e-04, -9.5261e-05,  5.2832e-05,  6.5111e-04,
         5.9548e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3494e-03, -9.8815e-01, -2.9341e-04,  5.6537e-04, -2.1770e-04,
         5.7000e-04, -3.1904e-04, -8.2000e-04, -4.3900e-04,  4.6122e-04,
         2.2957e-04, -9.8564e-04,  2.4763e-04,  5.0785e-05, -3.7784e-04,
         2.2105e-04,  4.9777e-06, -5.7943e-05,  1.1087e-03,  6.5451e-04,
        -3.5140e-04,  5.2055e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.7393e-03, -9.8701e-01, -2.6089e-04,  1.7247e-05, -5.3985e-04,
         2.2067e-03, -5.5066e-04,  1.3417e-03,  1.6215e-04,  6.9708e-04,
         2.7680e-04, -2.1092e-04,  1.3018e-04, -9.3541e-05, -9.0132e-04,
        -9.0030e-04, -1.5122e-03, -1.6411e-05,  4.3538e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8393e-03, -9.8414e-01,  1.4154e-03,  4.8533e-04,  2.9631e-04,
         2.2560e-03,  6.5221e-04, -3.4961e-04,  6.4332e-04,  8.0168e-04,
         1.0607e-03, -2.6055e-04, -1.3140e-03,  6.3188e-04, -1.4177e-04,
         4.6627e-04,  4.5511e-04, -7.9116e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9120e-02,  9.3925e-01,  3.4991e-03,  8.4477e-04,  2.8763e-03,
         1.2534e-03,  5.5733e-06, -9.3513e-04,  2.4422e-03,  2.1525e-03,
         2.6485e-04,  2.9822e-03,  1.1903e-03, -1.5204e-03,  3.5933e-03,
         1.8442e-03,  5.7688e-03,  6.9011e-04,  2.6203e-03,  1.0303e-03,
         5.6187e-04,  2.9036e-04,  1.2293e-03,  3.1117e-05,  1.6382e-03,
         1.5869e-03,  5.2252e-04, -2.5564e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.5085e-03, -9.6860e-01,  1.7224e-03,  1.4936e-03, -4.5788e-04,
         1.1938e-03,  8.8224e-04,  1.6778e-03,  1.0993e-03,  1.4522e-04,
        -2.5526e-04,  2.9201e-04, -1.2478e-03,  1.0311e-03,  9.0400e-04,
        -7.0406e-04, -1.4794e-03,  8.6414e-04, -1.8542e-03,  1.2827e-03,
        -1.1811e-03,  1.2600e-03, -1.0333e-03, -7.0751e-05,  4.7575e-04,
        -8.3143e-04, -4.1222e-05,  6.2185e-04,  1.8879e-03, -1.9561e-04,
         7.0780e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.8466e-03,  9.6896e-01, -1.5180e-04, -1.8809e-03,  1.9941e-04,
        -1.8292e-03, -2.0573e-04, -1.8872e-03, -2.2696e-03, -4.1003e-04,
        -1.5173e-03, -7.9375e-04, -6.1787e-04, -1.6715e-04, -2.0402e-03,
         3.2614e-04,  7.4389e-04, -1.8312e-03, -9.5691e-06, -5.8935e-04,
         5.7375e-05, -1.1516e-03,  7.2948e-06, -7.0511e-05,  2.5333e-04,
         1.2856e-03,  4.7200e-04,  3.2820e-04,  9.3771e-05,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.1421e-03, -9.6589e-01, -1.8254e-03,  2.7701e-03,  1.5883e-03,
         1.3152e-03,  2.9239e-04, -1.5092e-04,  1.5291e-05,  2.0511e-03,
        -1.0814e-03,  3.8742e-04, -9.9243e-04,  1.3044e-03, -2.3740e-03,
         2.6412e-03,  2.4952e-05,  1.1205e-03,  9.9126e-04, -2.4177e-04,
         3.1306e-03,  8.2240e-04, -1.8465e-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([ 8.4137e-03, -9.6401e-01, -3.8228e-03,  3.4084e-03,  2.0690e-03,
         3.3103e-04,  1.4466e-04,  1.0493e-03,  2.6570e-04,  2.3494e-03,
         1.2148e-03, -7.6164e-05, -1.9026e-03,  8.6724e-05, -2.6315e-03,
         2.5458e-03,  1.0026e-03,  1.7361e-03,  3.5152e-05,  5.4738e-04,
         2.0022e-03,  1.3701e-04,  2.1982e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4211e-02,  9.6095e-01, -6.4967e-05, -3.1024e-03, -1.6712e-03,
        -7.8694e-04, -1.4654e-03, -1.7436e-03,  1.0649e-04, -1.7583e-03,
         8.1204e-05, -1.2553e-03, -4.5551e-04, -2.3032e-03, -2.3730e-04,
        -2.5538e-03,  1.0100e-03, -2.8141e-03, -5.3231e-04, -4.2643e-04,
        -1.5005e-03,  7.0596e-04,  2.6436e-04,  0.0000e+00,  0.0000e+00,
         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([-1.2058e-02,  9.6464e-01,  1.5397e-03, -1.5789e-03,  5.5802e-03,
         2.5778e-03,  1.3445e-03, -1.6117e-04,  6.6055e-04, -1.7053e-03,
        -2.7614e-03, -6.5334e-04,  2.7018e-04,  3.1519e-03, -1.0400e-03,
         2.7311e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3285e-04,  9.8862e-01,  4.3206e-05, -2.0571e-04,  1.0936e-04,
         2.8235e-05, -2.1055e-04, -7.0303e-04, -6.1110e-05,  1.0242e-03,
         5.6067e-04,  2.5318e-04,  4.8169e-05, -2.7741e-05,  3.8764e-04,
        -2.4465e-04,  1.1778e-04,  7.0193e-06,  2.5765e-04, -5.3540e-05,
         3.4984e-04,  1.4581e-04,  1.0039e-05,  1.4995e-04,  3.6814e-05,
         5.9175e-04,  2.2056e-04,  2.2346e-04,  1.7623e-04,  4.1978e-04,
        -5.0352e-04,  3.4543e-04,  5.5036e-04, -6.0878e-04, -2.8340e-04,
        -3.0512e-04,  7.8315e-05,  5.0736e-04, -8.4595e-04, -2.5019e-04,
         3.0415e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6702e-04, -9.8655e-01,  3.2984e-04, -1.8316e-04,  5.3570e-05,
         1.5745e-04,  5.5664e-04, -1.4986e-04,  3.8361e-04,  6.6160e-05,
         4.7114e-05,  2.0422e-04,  5.2790e-04,  2.8808e-04,  8.0853e-04,
         1.8816e-04,  3.6299e-04,  5.3784e-04,  5.0864e-04,  2.2468e-04,
        -2.9986e-04, -1.4327e-04,  7.7515e-04,  4.7069e-05,  8.1573e-04,
        -7.2524e-04,  3.6572e-04,  3.4879e-04,  3.0842e-04,  4.8283e-04,
         5.3914e-05, -8.3030e-05,  2.1181e-04, -2.1421e-05,  9.3734e-04,
         6.8578e-04, -1.1980e-04,  2.4426e-04, -6.6546e-05,  5.5247e-04,
         1.1770e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1794e-05, -9.9233e-01,  7.2664e-05, -1.0122e-07,  1.4278e-04,
         2.1645e-04, -2.6985e-05, -8.1665e-05,  3.3602e-04, -2.5574e-04,
         8.8795e-05, -3.2217e-05,  7.8128e-05,  1.1385e-04, -5.1434e-04,
         4.4182e-04, -3.0501e-04,  1.1146e-04, -1.3629e-05, -2.5780e-04,
        -1.5909e-04,  4.2992e-04,  1.0762e-04, -1.7442e-04,  3.5171e-04,
         2.8484e-04, -4.5363e-05,  2.2109e-04, -1.6917e-04,  1.6536e-04,
         4.8695e-05,  6.6127e-05,  1.4313e-04, -3.0043e-04, -2.9604e-05,
        -1.8725e-05,  8.9151e-05, -1.6881e-04, -1.1707e-05,  2.5685e-04,
        -5.3912e-05, -1.5232e-05,  2.2396e-04,  2.4841e-04,  2.1348e-04,
        -5.7146e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.2975e-03,  9.7458e-01,  2.1017e-04,  3.1490e-04, -1.2869e-03,
         8.8278e-04, -5.9929e-04, -3.4624e-04, -1.4295e-04, -7.1060e-04,
        -5.0006e-04, -2.1708e-04, -4.2786e-04,  3.3841e-05, -8.1359e-04,
        -1.3984e-03,  1.3003e-04, -3.0043e-04,  5.8604e-05, -2.8770e-04,
         1.5101e-04, -6.3160e-04, -5.9338e-04, -2.3919e-04, -5.5412e-04,
         4.2212e-04,  7.9083e-06,  4.3926e-05, -5.7970e-05, -3.3078e-04,
        -5.7713e-04,  3.0806e-04, -6.0639e-05,  8.9694e-06, -3.4999e-04,
         1.2552e-06, -8.0431e-04,  2.0631e-04,  1.3865e-04, -4.6993e-04,
        -7.8736e-04, -4.7354e-04, -6.7518e-05,  1.6687e-04,  3.7633e-06,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9606e-03, -9.8518e-01, -6.1151e-04,  1.5748e-04,  2.5291e-05,
        -1.2357e-04,  3.8916e-04,  4.3792e-04,  2.6713e-04,  1.8052e-04,
         4.3167e-04,  2.1101e-04,  2.0952e-04,  3.3044e-04,  3.3204e-04,
         2.5957e-04, -1.5342e-04,  1.3340e-04, -7.8362e-05, -2.6222e-05,
         2.2638e-04, -2.8607e-04,  8.0300e-04, -1.8767e-04,  1.8777e-04,
        -3.8898e-04, -4.4273e-05,  5.8822e-05,  1.2388e-04, -4.2680e-04,
        -1.3276e-04, -2.3423e-04, -3.9991e-04,  3.2564e-05,  1.4436e-04,
        -2.5627e-04,  1.1231e-04, -6.4615e-05, -5.4950e-05, -2.3908e-04,
         1.9891e-04,  2.9942e-04, -5.9732e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.3660e-03, -9.8276e-01,  1.5929e-04, -6.6712e-04, -1.5542e-04,
        -2.9875e-04, -5.3608e-04, -2.2406e-04, -1.4705e-04, -3.8921e-04,
        -1.7402e-04, -2.4945e-04, -2.6734e-04, -4.3158e-04, -3.0565e-04,
         6.9603e-04, -6.5651e-04, -6.7365e-04, -6.1516e-06, -3.6393e-04,
         4.9161e-04, -3.2438e-04,  2.1413e-04, -4.9241e-04, -1.4287e-04,
         6.9769e-05,  2.2623e-04, -3.5616e-04, -3.1436e-04,  3.6866e-04,
        -1.1969e-04, -2.6604e-04,  3.7635e-04, -1.4677e-04, -3.3697e-04,
        -2.3987e-04, -6.4031e-04, -3.8350e-04, -7.3712e-04, -1.8941e-04,
         3.9018e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.4482e-03, -9.8235e-01,  2.8475e-04, -2.7270e-05,  3.5629e-05,
         5.5971e-04,  3.0501e-04,  4.1109e-05,  2.9736e-05,  9.2836e-05,
         7.3252e-05, -4.4688e-04, -1.3557e-05,  1.7225e-04, -1.1333e-04,
         6.5972e-04,  3.8561e-04,  7.1329e-04,  4.8534e-04, -2.0378e-04,
        -6.4490e-05,  1.3739e-04,  5.1661e-04,  1.0137e-04,  2.8872e-04,
        -4.1796e-04,  1.3486e-04,  4.7211e-04,  2.1733e-04,  1.7466e-04,
         1.0819e-04, -4.0881e-04,  6.3520e-04,  3.3302e-04,  5.9087e-05,
        -7.2303e-05,  3.7200e-05, -1.3945e-04, -8.8855e-05,  1.4703e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.1577e-03, -9.7787e-01, -2.7224e-04, -4.7710e-04,  8.6566e-05,
        -6.9556e-04,  9.9988e-04, -6.0992e-04, -7.5630e-05,  2.7106e-04,
        -1.7057e-04,  1.5245e-04,  4.7142e-04, -1.9406e-04,  5.4783e-07,
         2.6299e-04, -1.9066e-04,  1.1392e-03, -8.9904e-06, -3.4344e-05,
        -2.2631e-04, -1.8198e-04,  1.1119e-03, -3.7212e-05,  6.8669e-05,
        -6.0313e-04,  2.0633e-04, -3.2470e-04, -2.4233e-05, -3.9437e-04,
        -1.9897e-05, -3.5985e-04,  6.0957e-04, -3.0514e-04,  5.2112e-05,
        -2.5595e-04,  1.8293e-04, -6.6371e-04, -5.1605e-04, -3.1386e-04,
        -3.6650e-04,  3.8085e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4233e-03, -9.8959e-01,  2.1386e-04, -3.4454e-04,  3.7543e-04,
        -1.7948e-04,  3.0828e-04, -2.6409e-05, -1.8954e-04, -4.0589e-04,
        -1.7640e-04,  2.7221e-04,  4.5230e-04,  1.1330e-04,  1.4499e-04,
         4.1789e-04, -1.1357e-04,  5.8248e-05,  2.7101e-04,  1.5166e-04,
         1.8986e-04, -1.5904e-04,  3.4099e-04,  2.8802e-04,  3.3865e-04,
        -1.0646e-04,  1.4560e-04,  2.1272e-04,  5.5660e-05,  3.6120e-04,
        -5.4189e-05, -4.8021e-04,  2.1799e-04, -3.1853e-04, -2.3582e-04,
        -4.3869e-05, -2.3305e-04, -4.2551e-04, -3.2616e-04, -2.0110e-04,
        -3.4876e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.4857e-04, -9.8746e-01,  5.9221e-04,  1.6493e-05,  4.4913e-04,
        -5.1037e-04,  4.6760e-04,  3.6988e-04, -6.5193e-05,  3.6378e-04,
        -3.5021e-04, -1.9320e-04,  3.7543e-04, -8.6352e-05, -2.9120e-05,
         5.1874e-04, -8.4941e-05, -1.3821e-04,  4.8716e-04,  5.7863e-04,
        -1.7899e-04,  2.9222e-04,  3.9287e-04, -1.1916e-04,  5.1087e-04,
        -3.2764e-05,  1.1057e-04,  3.5984e-04,  2.6305e-04, -3.5047e-04,
         9.3856e-04,  1.0455e-05,  1.3956e-04,  1.3560e-04,  6.2675e-04,
         1.6589e-04,  4.6771e-05, -2.3810e-04, -9.3915e-05, -2.1143e-05,
        -1.1645e-04,  6.4138e-05,  4.4644e-04,  4.8343e-05,  2.1584e-04,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5118e-03, -9.9020e-01,  7.8092e-05, -3.1347e-06,  1.9824e-04,
         4.9471e-04,  5.1816e-04,  5.5048e-05, -3.6771e-05,  2.7664e-04,
         3.0726e-04,  4.6297e-04, -1.9665e-04,  9.3050e-05, -4.0869e-04,
         2.0580e-04,  4.6858e-04,  5.4167e-05, -6.2246e-06,  2.6707e-04,
         6.1812e-05,  7.5439e-05,  1.8375e-04,  1.6079e-04, -1.4178e-04,
        -3.5618e-04,  4.5924e-04, -2.7339e-05, -4.3370e-06, -2.3920e-04,
         2.6595e-04, -5.5025e-04, -2.4348e-04,  2.9418e-04,  1.0431e-04,
         2.8647e-04, -1.3023e-04,  1.0923e-04,  2.2493e-04, -2.0393e-04,
         3.8947e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-7.1268e-03, -9.8247e-01, -4.0478e-04, -5.3521e-05,  1.8833e-04,
        -2.9066e-04,  6.7496e-05, -7.1686e-04, -1.0451e-03,  6.1509e-05,
        -2.4418e-04, -4.7687e-05,  2.9467e-04, -6.7535e-04, -3.0021e-04,
        -9.3403e-04, -3.2148e-04, -7.1556e-04, -3.0003e-04, -2.5336e-05,
        -4.1307e-04, -3.2800e-04, -5.3635e-04, -2.5930e-04, -3.5507e-04,
         2.9522e-04, -4.3109e-04, -4.9014e-04,  3.6210e-04,  2.4126e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4866e-04, -9.9165e-01,  2.1094e-04,  4.7161e-05, -4.7989e-04,
        -9.2755e-05,  5.4837e-05, -4.7003e-04, -3.1750e-04, -4.5015e-04,
         2.1374e-05,  1.6850e-03, -3.3516e-04, -4.8720e-04, -5.7053e-04,
         3.3794e-04, -5.6323e-05, -1.5172e-04,  5.6686e-05, -1.8775e-05,
         1.2236e-04, -2.3275e-04, -3.7082e-04,  2.4545e-04,  2.1548e-04,
        -1.4263e-04, -5.0256e-04, -2.2896e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.8194e-04,  9.9054e-01, -5.0835e-04, -4.2309e-04,  3.5170e-04,
        -2.4244e-04, -5.6011e-04, -6.3468e-04,  2.2013e-04, -6.3970e-04,
         8.4586e-05,  3.3492e-04, -4.4374e-05, -3.9239e-04,  3.2070e-04,
         4.7976e-04, -5.1175e-05,  1.8474e-04, -5.3878e-04, -2.3967e-04,
        -1.0495e-04, -1.1240e-03, -3.4040e-04, -4.7882e-04, -1.8169e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.4114e-04,  9.9127e-01,  4.9690e-04, -6.3120e-04, -3.9462e-04,
         4.1222e-04, -4.2394e-04,  1.8577e-04, -1.0337e-03,  1.1028e-04,
         9.0423e-04, -6.2425e-04, -3.9722e-04, -1.3328e-04,  5.2598e-04,
         2.5617e-04,  4.1212e-04, -1.4047e-05,  3.9122e-04, -4.4469e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2285e-04,  9.8943e-01, -4.1410e-04, -8.2879e-04,  6.9301e-04,
        -5.0694e-04, -4.2970e-04, -1.0003e-03, -2.4328e-04,  4.3516e-05,
        -7.4818e-04, -3.9852e-04,  8.3894e-05, -4.6358e-04,  9.8761e-04,
         5.1008e-04,  8.1029e-04, -4.2944e-04,  6.2726e-05,  7.1538e-05,
        -1.1276e-03, -2.9240e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0457e-03,  9.7183e-01, -6.5598e-05,  7.8542e-04, -2.6217e-03,
        -1.4220e-03, -1.4343e-03, -2.3445e-03, -1.8883e-04,  8.7432e-04,
        -1.0621e-04,  1.0011e-04, -9.3750e-04, -1.3299e-03,  2.3275e-04,
        -3.4340e-04, -6.5871e-04,  8.7268e-04, -1.7521e-03,  8.1684e-04,
        -2.0933e-04, -5.1867e-04,  3.3583e-04,  1.2296e-03, -2.4213e-03,
        -1.4800e-03, -8.5506e-04,  1.1909e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.7163e-03,  9.7768e-01,  7.1621e-04,  4.0907e-04,  6.7575e-04,
         3.4036e-04,  1.3616e-04, -6.1078e-04,  1.6264e-03,  8.5474e-04,
        -3.2632e-04, -5.0604e-04, -8.3062e-05, -7.9596e-05,  2.5277e-03,
         2.9087e-04, -1.7668e-04,  1.3456e-03,  5.3737e-04,  9.5193e-04,
         4.2107e-04,  1.1568e-03,  2.0789e-03,  9.9631e-04,  8.8530e-05,
         3.4222e-04,  7.5307e-04,  4.6004e-04,  1.1330e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.7509e-03,  9.7967e-01,  1.0232e-03, -5.3775e-04,  3.4111e-04,
        -1.3095e-03, -6.3387e-04,  1.1390e-05,  1.0128e-04, -2.5166e-04,
         2.0951e-04, -5.5927e-04, -8.1182e-04, -4.6850e-04, -4.7227e-04,
        -2.4973e-04,  5.1361e-04, -9.3314e-04,  2.9240e-04, -4.5885e-04,
        -1.6034e-03, -5.2482e-04, -3.8166e-04,  4.2358e-04, -3.0646e-05,
         1.6980e-03, -6.5301e-04, -8.0324e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2764e-03, -9.8886e-01,  7.6477e-04, -4.4804e-04,  1.9635e-04,
         3.8350e-04,  4.5976e-04, -2.0748e-04, -1.4852e-04,  6.2047e-04,
        -4.0754e-04, -4.7918e-04,  3.3008e-04,  3.9239e-05, -1.5517e-04,
         7.8102e-04,  6.2038e-04,  6.6614e-05, -3.7875e-04, -2.0852e-04,
        -4.5253e-04, -5.6176e-04, -3.9672e-04, -3.6727e-04, -2.9440e-04,
        -9.3279e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2958e-03,  9.7881e-01, -1.1597e-03,  7.8704e-04,  9.2560e-05,
         1.2644e-03,  1.2468e-03,  4.8147e-04,  6.5621e-04, -5.9698e-04,
        -5.5078e-05, -4.1882e-05,  2.1565e-04,  1.0682e-03, -2.8929e-03,
         7.0457e-04, -6.0774e-04, -5.5812e-04,  9.4618e-05,  4.0675e-04,
        -6.1543e-04, -1.1177e-03, -1.4098e-04,  1.6423e-04, -9.8791e-04,
        -2.7381e-04,  7.1490e-05, -4.4128e-04,  1.4836e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4470e-03, -9.8909e-01,  9.3264e-05, -3.3660e-04, -2.1932e-04,
         4.2896e-04, -1.7031e-04, -2.8692e-04,  6.4541e-06,  1.9028e-04,
        -1.9217e-04,  6.9037e-05,  3.4839e-04, -4.9883e-04, -9.9983e-05,
        -9.7739e-05,  1.3709e-05, -1.2158e-04,  3.1047e-04, -5.6758e-04,
         2.5552e-04,  6.0844e-05,  1.6843e-04,  2.3453e-04, -5.0571e-04,
         5.6073e-05,  3.0590e-04, -1.6276e-05,  4.3909e-04,  4.8292e-04,
         2.8872e-04, -4.9613e-04, -1.0040e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1205e-03,  9.8557e-01, -4.6705e-04, -5.9701e-04,  1.4478e-03,
        -1.9085e-03, -9.6401e-04, -8.4201e-04, -1.6372e-03, -3.5475e-04,
        -4.7302e-04,  1.6866e-04,  4.6862e-04, -5.8228e-04, -1.2863e-04,
        -2.6221e-04, -4.6968e-04, -5.1330e-04, -2.7884e-04, -8.1602e-04,
         9.3067e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #400: [tensor([ 3.9359e-03, -9.8898e-01,  8.1130e-05,  2.7751e-04,  5.2356e-04,
        -3.4825e-04,  5.2773e-04, -1.7332e-04,  2.2608e-04,  6.8641e-05,
        -7.6286e-05, -2.9241e-04,  6.7241e-05,  9.5238e-05, -2.4164e-04,
        -4.9217e-04,  1.3624e-04,  5.7526e-04, -1.1472e-04,  1.0122e-04,
         4.4478e-04,  2.8804e-04, -5.6976e-04,  2.9928e-04, -5.9958e-04,
         1.1421e-04,  3.0671e-04,  4.5650e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.9426e-04, -9.9220e-01,  3.6807e-04,  3.3287e-04,  3.9000e-04,
        -2.1466e-04, -3.2219e-04,  2.3391e-04,  3.3257e-04, -3.7486e-04,
         2.8178e-04, -3.7073e-05,  5.2930e-04,  6.6144e-04,  3.6839e-04,
        -2.8938e-04,  1.5338e-05,  2.2817e-04,  2.1629e-05, -1.7498e-04,
         5.2493e-04,  7.3566e-05,  2.8005e-04,  1.5926e-05,  5.1146e-04,
        -1.1658e-04,  4.0394e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.1612e-03, -9.8631e-01,  2.2583e-04, -4.0408e-04, -2.1641e-04,
        -6.4979e-05, -3.7007e-04,  5.3241e-04, -3.6078e-04,  2.3477e-04,
         3.2611e-04, -1.6630e-05,  8.8168e-04, -2.5912e-05, -4.4560e-04,
        -3.2690e-04,  4.8087e-04, -1.9875e-04,  5.1031e-05,  8.0900e-04,
        -3.5742e-04,  8.3523e-04,  3.8214e-05,  1.0335e-03,  2.3475e-04,
        -5.5896e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3959e-03, -9.8745e-01, -5.4033e-05,  4.6214e-04,  3.8183e-04,
        -2.2063e-04,  2.5263e-04,  7.9932e-06,  3.0942e-04,  4.6077e-04,
         2.7177e-04,  5.4885e-04,  4.7011e-04,  4.6144e-04,  6.1975e-05,
        -6.6621e-05,  6.4229e-04,  3.0346e-04,  3.6371e-04, -3.6964e-05,
         2.0861e-05,  4.0014e-04,  3.8522e-04,  1.3236e-04,  3.1117e-04,
         2.2237e-04, -2.4098e-04, -3.1303e-04,  9.0167e-05,  1.0157e-04,
        -3.6117e-04, -2.5790e-04, -2.4977e-04, -1.6462e-04, -2.9841e-04,
        -3.8923e-04, -3.3728e-04, -2.0489e-04, -2.9355e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6679e-03, -9.8722e-01, -6.7434e-04, -3.3218e-04,  1.5541e-04,
         1.5322e-04,  2.1524e-04,  3.8371e-04,  1.7455e-05,  2.3822e-05,
         7.2583e-05,  1.8588e-04,  4.4880e-04,  3.1900e-04, -2.0525e-04,
         3.1482e-04,  4.3237e-04,  1.4879e-04,  1.8652e-04,  4.3953e-04,
         4.8133e-05,  1.2064e-04,  3.0302e-04,  4.4244e-04,  4.3657e-04,
        -1.8425e-04,  9.4473e-04,  2.4676e-04, -4.1501e-05, -1.3682e-04,
         3.6105e-04, -2.7941e-04,  5.9166e-04,  1.5839e-04,  3.1407e-04,
         3.6501e-04, -2.9207e-04, -1.3885e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7293e-03, -9.9118e-01,  4.9754e-05,  8.1616e-05,  1.6055e-04,
         2.9517e-04,  2.8499e-04,  9.3154e-05, -3.3928e-04, -7.3245e-05,
         4.6657e-04, -3.6772e-06,  1.5188e-04,  2.3886e-04, -2.2749e-04,
        -1.1868e-04,  1.0075e-04,  1.7893e-04, -1.5403e-04,  2.3113e-04,
        -2.4308e-04,  4.7243e-04,  1.3165e-04, -2.3847e-06, -1.6790e-04,
        -8.8006e-05, -6.8990e-05, -1.3708e-04, -5.9917e-04,  3.0313e-04,
        -1.3587e-04, -1.1121e-04,  8.4093e-05, -2.3044e-04,  3.1731e-04,
         2.0664e-04, -1.2197e-04, -1.4566e-04,  2.7706e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.3313e-03, -9.8820e-01,  2.0179e-04,  6.4885e-04, -2.1344e-04,
        -8.4089e-06,  2.4326e-04, -1.4869e-04,  5.6992e-04,  3.0317e-04,
         5.8084e-05,  7.9771e-05,  3.2132e-04, -3.2618e-04,  4.8839e-04,
         4.2127e-04,  5.3955e-04,  6.0408e-04,  1.0569e-05,  6.5383e-04,
        -1.7351e-04,  4.4757e-04,  4.6218e-05,  1.6149e-04, -3.5956e-04,
        -1.7321e-04, -3.9936e-05, -2.2837e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8820e-04, -9.9309e-01,  4.1528e-04, -2.6025e-04,  6.8697e-04,
        -5.9872e-06,  4.2057e-04,  3.6758e-04,  7.2919e-05, -2.6271e-04,
         7.7237e-05,  1.3151e-04,  2.6646e-04,  2.2842e-04,  1.0336e-04,
         8.7838e-04, -1.2910e-04,  7.1335e-04, -7.1509e-05, -1.6220e-05,
        -8.2813e-05,  2.3353e-04, -1.1751e-04, -2.5448e-04,  1.2392e-05,
         1.6858e-04, -1.3083e-05, -2.5662e-04, -2.2957e-04, -8.6291e-05,
        -1.9157e-05, -4.0834e-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([ 1.1450e-03, -9.9227e-01,  1.1605e-04, -1.0673e-04, -2.6885e-04,
        -3.5686e-04,  1.5943e-04,  1.1861e-04,  1.2898e-04, -2.0172e-04,
        -9.4475e-05,  1.9441e-05, -1.0675e-04, -1.7400e-04, -2.6712e-05,
         2.6816e-04,  3.5077e-04,  5.3451e-04,  2.6302e-04,  1.8053e-04,
        -1.5407e-04,  2.4207e-04,  1.5478e-04, -2.9665e-05,  5.9200e-04,
         2.4914e-04, -3.0424e-04,  1.9148e-04,  1.2907e-04, -5.6853e-04,
        -5.9011e-05, -1.4402e-04,  2.8872e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.5184e-04, -9.9112e-01, -2.5958e-04,  4.1915e-04,  7.3703e-04,
        -3.5096e-04,  5.6877e-04, -1.5582e-04, -5.4310e-05,  3.6804e-04,
         5.7942e-04,  3.3301e-05,  2.3847e-04,  5.3047e-04,  3.8808e-04,
         5.2825e-04,  4.8786e-04, -1.3537e-04,  5.8101e-04, -3.7878e-04,
        -1.6739e-04,  2.5923e-04,  3.8861e-04, -1.0556e-04,  1.5820e-04,
         5.8828e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6553e-03, -9.8892e-01,  3.8936e-04, -3.1924e-04,  2.0589e-06,
        -9.4717e-05,  4.6105e-04, -1.1063e-04, -7.3226e-04, -2.5944e-04,
        -1.8269e-04, -2.1210e-04,  4.9577e-04,  3.6349e-04,  1.5576e-04,
         2.2635e-04, -3.7614e-04,  2.2281e-04,  6.6369e-04, -2.8790e-04,
         4.4752e-04,  6.8650e-04,  3.0944e-06, -1.5602e-04,  1.3907e-05,
        -2.5904e-04, -5.8583e-04, -3.1900e-04,  3.9471e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8028e-03, -9.8992e-01, -1.2438e-04,  2.3630e-04,  5.5729e-05,
        -4.4348e-05,  1.0332e-04, -3.7229e-04,  5.3461e-06, -2.6938e-04,
         2.6399e-04,  2.1895e-04,  1.0457e-04,  3.0605e-04, -4.2876e-05,
        -3.1392e-04, -4.8909e-04,  6.8428e-04, -1.4567e-04,  5.7905e-05,
         4.3756e-04,  5.2713e-04, -2.0159e-04, -3.7950e-05, -1.2945e-04,
        -7.4816e-05, -1.6179e-04,  7.6931e-04,  4.4357e-04,  6.5928e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #450: [tensor([-1.0564e-02,  9.7856e-01,  8.0474e-05,  1.4976e-03,  1.1752e-03,
         1.6536e-04, -5.0430e-04,  8.5695e-04,  1.2493e-03,  1.3339e-03,
        -9.4191e-04,  8.9078e-04,  7.1829e-05,  1.3621e-03, -1.4911e-04,
         1.7269e-04,  1.6012e-04, -2.6246e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6554e-03, -9.8176e-01,  5.8993e-04,  7.0541e-04,  1.0038e-03,
         3.0183e-05, -1.5191e-03,  1.5003e-03,  7.2482e-04,  2.0825e-04,
        -8.0579e-04,  9.3193e-04,  3.0899e-04,  6.4704e-05, -3.3322e-04,
         3.0731e-03,  4.5469e-04,  3.2585e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1691e-06, -9.9399e-01,  2.8138e-05, -6.0024e-05, -5.0525e-04,
         3.6714e-04,  2.0918e-04, -3.9746e-04,  4.7597e-04,  8.1948e-04,
        -6.1894e-04,  6.4898e-05,  9.1574e-04,  1.1201e-04,  4.4329e-04,
        -1.3740e-04, -3.1546e-04, -1.5196e-04, -1.7433e-04, -1.4834e-04,
         5.6490e-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([-1.0648e-04,  9.9303e-01, -1.3747e-04,  4.2057e-04,  3.3722e-04,
         9.1366e-05, -2.4784e-04,  6.3213e-04,  3.8940e-04,  1.4600e-04,
         1.5243e-04,  2.6675e-04,  4.4157e-04,  5.8673e-05,  5.5850e-05,
         3.0006e-04,  9.6506e-04,  3.5613e-04, -1.7042e-05,  2.8576e-04,
        -1.1169e-06,  5.7575e-04,  8.1602e-05,  5.0945e-04,  4.0963e-05,
         3.5260e-04,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.6010e-03, -9.8668e-01, -5.0242e-04, -5.5883e-04, -6.4646e-04,
        -8.4664e-06,  2.4233e-04, -3.4892e-04,  1.0867e-04, -5.4516e-04,
        -4.4780e-04, -7.8175e-04, -5.6742e-04, -4.3053e-04, -2.1819e-04,
         2.8044e-04, -4.7820e-04,  1.8375e-04,  3.1739e-04,  5.3803e-04,
        -1.2751e-04,  3.8251e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7070e-03, -9.8853e-01,  2.8133e-04,  3.7599e-04,  4.5828e-04,
        -4.2126e-04,  2.7979e-04,  7.8881e-04, -3.3082e-04,  4.0826e-05,
         1.0843e-04, -2.4607e-05, -1.1508e-04, -2.5490e-04,  2.8312e-05,
        -1.6660e-04, -6.2015e-04, -7.1941e-05, -7.3886e-04, -5.2597e-04,
        -4.7543e-04, -4.8671e-04, -1.6579e-05, -1.5351e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8846e-03, -9.8899e-01,  4.1475e-04,  2.8319e-04, -7.0681e-04,
        -4.3753e-04,  1.7996e-04,  9.2301e-05,  2.4681e-04,  8.4848e-04,
         2.1568e-04, -4.5039e-04, -3.4604e-04, -2.7781e-04,  4.8999e-04,
         2.7712e-04,  3.9826e-05, -1.3359e-04, -6.2278e-04, -4.0520e-05,
        -3.1181e-04, -9.9312e-06, -6.9583e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6672e-03, -9.9103e-01, -1.9380e-05,  2.9070e-04, -1.4300e-05,
        -5.4207e-04, -5.7541e-04, -7.1979e-05, -2.3315e-04, -1.9274e-04,
        -2.3854e-04, -5.3163e-05, -4.6921e-04,  3.4344e-05, -4.5518e-04,
         4.9237e-05, -1.4359e-04, -8.5617e-04,  2.3411e-05, -4.7991e-04,
        -3.2358e-04, -1.6433e-04, -5.9999e-04, -6.5912e-05,  4.0747e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1255e-02,  9.6311e-01, -4.7659e-04,  1.2680e-03,  1.0302e-03,
        -9.7174e-04, -1.8495e-03,  1.7943e-04,  1.1607e-03,  1.3055e-04,
        -1.4396e-04, -7.7750e-04,  1.5996e-04,  1.2128e-03, -8.1484e-04,
         1.0937e-03, -8.2024e-04, -1.4562e-04,  1.7625e-03, -1.5168e-04,
        -9.4935e-04, -5.3487e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4818e-03,  9.7824e-01, -1.0359e-03, -1.1611e-03,  2.4666e-04,
        -1.7913e-04, -1.7505e-03,  1.2495e-04, -1.0013e-04, -8.3607e-04,
        -1.1816e-03, -6.3698e-04,  4.4581e-04,  1.1664e-03,  1.3182e-04,
         1.0063e-03, -7.7850e-04,  1.6894e-04, -6.7174e-04, -1.4716e-04,
         2.0223e-03, -7.2429e-04, -5.3925e-04, -1.7878e-03, -4.4116e-04,
         9.5358e-04,  1.0444e-03], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1632e-03,  9.8478e-01,  4.0344e-04, -2.1380e-04,  7.4655e-04,
         2.2852e-04,  7.0730e-05, -5.7177e-04,  1.9950e-04, -1.7153e-03,
        -2.8219e-04, -7.8905e-04, -8.2782e-04,  6.8414e-04,  7.8514e-04,
         1.0691e-03, -3.1775e-04, -8.6536e-04,  1.2546e-04, -3.7166e-04,
         3.4159e-04,  2.2044e-04, -2.2931e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.8729e-03, -9.7861e-01, -6.1533e-04, -1.6448e-03,  8.5322e-05,
        -2.3407e-04, -1.0829e-03, -9.3161e-05,  6.1531e-05,  2.8301e-05,
         4.7774e-05,  1.2533e-03,  1.6858e-03,  8.9173e-04,  1.8497e-04,
        -2.3511e-03, -1.6034e-03, -2.9066e-04, -1.1031e-03, -6.5843e-04,
         6.0346e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 6.4464e-03,  9.7207e-01, -2.9573e-03, -1.0912e-03,  1.9926e-03,
        -8.2137e-04, -1.0647e-03,  1.0101e-03, -1.3373e-03, -2.9859e-03,
        -1.3891e-03,  5.6795e-04, -1.8430e-03,  1.2499e-03, -1.1577e-03,
        -1.0248e-03,  9.9181e-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([-4.6993e-03,  9.8004e-01, -2.6506e-03,  7.4370e-05, -8.1036e-04,
        -1.1813e-03, -8.3151e-04, -1.6837e-03, -1.3254e-03, -2.7008e-03,
         9.0290e-04, -3.4127e-04, -1.0564e-03,  1.4751e-03, -2.3082e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3501e-03,  9.6876e-01, -2.2861e-03, -1.6098e-03,  1.0498e-03,
        -2.5346e-04, -1.5274e-03,  2.5054e-04,  1.9664e-03,  1.4947e-03,
        -1.1299e-03, -2.9397e-04,  2.8479e-03,  5.8947e-04,  8.9463e-04,
        -8.2322e-04,  2.2409e-03,  2.5383e-03,  2.7114e-04, -8.2167e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.3580e-03,  9.7560e-01,  1.1986e-04, -3.2104e-03, -1.0948e-03,
        -9.3909e-04, -1.1182e-03, -9.6896e-04,  1.7854e-04,  1.4700e-03,
        -4.5753e-04, -1.0617e-03, -3.7650e-04,  2.8101e-03, -1.0680e-03,
         2.3816e-04,  1.0313e-03,  6.4995e-04,  2.5198e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7156e-02,  9.6366e-01,  2.0475e-03, -1.5900e-03,  2.3689e-04,
        -7.7241e-04, -2.2569e-04, -8.0431e-04, -4.1378e-04, -8.8932e-04,
         3.2841e-03,  7.9568e-04, -4.7598e-04,  1.0908e-03,  7.4241e-05,
        -3.7972e-04,  1.6321e-03,  1.0461e-03,  2.3099e-03,  1.1112e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.0526e-03,  9.7710e-01, -8.0563e-04, -1.9848e-04, -2.0379e-03,
        -4.5683e-04, -9.1477e-04, -1.0771e-03,  1.1381e-03,  9.9566e-04,
        -1.3614e-04, -2.5391e-04, -1.4682e-03, -1.2325e-04,  3.8157e-04,
        -1.8264e-03, -4.8664e-04, -5.0519e-04, -1.3091e-03, -1.8557e-04,
         5.4381e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6297e-03,  9.7813e-01, -3.8705e-05, -1.5845e-03, -1.7809e-03,
         3.4882e-05, -9.0950e-04,  1.1900e-05, -8.7439e-04,  2.5438e-04,
         3.7263e-04, -5.5457e-04, -9.5644e-04, -1.4849e-03, -8.4932e-04,
        -3.0491e-05, -6.4425e-04, -3.3483e-04, -1.0329e-03,  3.8159e-04,
        -2.1946e-03,  3.4153e-04,  6.4944e-04,  1.9267e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1264e-03,  9.8148e-01, -2.6376e-03, -1.7771e-03, -1.6864e-03,
         2.5011e-04,  4.6140e-04,  4.6875e-04,  9.2685e-04,  4.9037e-04,
        -5.3784e-04, -1.6418e-03, -4.6636e-04, -1.2319e-03,  2.9558e-04,
        -6.1936e-04,  2.9321e-04,  4.8814e-04, -1.1161e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6166e-02,  9.6992e-01,  7.8041e-05, -1.7428e-03,  4.9910e-04,
         4.6909e-04, -2.4216e-03,  1.3919e-03,  9.6719e-05, -4.6101e-04,
         7.6583e-04, -1.0016e-03,  5.6439e-04, -8.2698e-04,  1.7228e-04,
        -9.8582e-04, -3.3957e-04, -4.4719e-04,  1.3980e-04, -1.3786e-03,
        -1.3507e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.3048e-03,  9.7324e-01, -2.0699e-03, -6.5013e-04,  5.0074e-04,
        -2.2226e-03, -7.0193e-04, -1.7038e-05, -2.6523e-03,  1.3715e-03,
        -9.2093e-05, -4.0750e-04,  6.7096e-04, -1.1881e-03,  2.3720e-03,
         3.4556e-05,  1.4669e-03,  8.7661e-04,  1.6300e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.9512e-03,  9.7882e-01, -2.1446e-04,  5.5576e-04, -6.3548e-04,
        -2.2303e-04,  8.4662e-04,  8.9719e-05, -1.6192e-03,  8.0334e-04,
         9.6472e-04, -1.3367e-03,  1.3333e-03, -1.1166e-03,  3.4580e-03,
         7.1650e-04, -5.0861e-04,  9.3890e-04,  8.6320e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2761e-03, -9.9095e-01,  9.8895e-05, -3.4827e-04,  6.8530e-04,
         2.7202e-04,  1.0184e-03, -1.7212e-04, -5.5767e-05, -4.1029e-04,
         2.4796e-04,  9.4623e-04,  3.1615e-04,  1.9801e-04, -3.7604e-04,
         1.0212e-04, -3.5991e-05, -4.2846e-04, -3.7346e-06,  6.5435e-04,
         3.5296e-04,  7.7678e-04,  2.7110e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #550: [tensor([-3.7247e-03,  9.9104e-01, -1.5594e-04, -2.8333e-04,  9.4827e-05,
         2.3313e-04, -4.2200e-04,  3.2781e-05,  3.5624e-04, -6.7945e-04,
        -5.9869e-04,  2.9587e-04, -3.5250e-04, -6.6201e-04,  3.4700e-05,
         7.9625e-05,  1.6223e-04, -1.9305e-04, -1.9461e-04, -4.0452e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1189e-05,  9.9447e-01, -3.6600e-04,  1.3191e-04,  6.7839e-04,
        -1.7396e-04, -1.6609e-04,  4.8517e-04, -7.1408e-04,  8.9517e-05,
         1.9792e-04, -1.6652e-04, -2.0124e-04, -4.0377e-04, -2.1548e-04,
         4.6485e-05,  5.2842e-04,  2.5165e-04,  1.0158e-04, -5.5777e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8198e-03, -9.8934e-01, -8.2895e-04,  3.5442e-04, -6.9723e-04,
        -5.2926e-04,  2.0514e-04, -2.1510e-04,  4.1413e-04, -2.9948e-04,
        -1.6493e-04, -1.9540e-04, -3.6651e-04, -8.4022e-04, -5.3250e-04,
         4.2797e-04,  2.4783e-04,  1.7198e-04, -7.8256e-04, -8.1124e-05,
         3.4030e-04, -1.4992e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1764e-03,  9.8793e-01, -1.9841e-04,  6.5868e-04,  4.5608e-04,
        -3.3149e-04, -1.1423e-03,  7.2689e-04, -4.7821e-04, -6.3182e-05,
         2.9913e-04, -3.8043e-04,  2.4718e-04, -1.6598e-04,  5.3573e-05,
        -5.9914e-04, -1.0895e-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([-4.0117e-03, -9.8998e-01,  2.9480e-04, -5.9672e-04, -6.3198e-04,
        -6.5090e-05,  6.3580e-04, -4.0675e-04, -9.2770e-04,  2.0553e-04,
         4.7054e-04,  8.0483e-05,  9.7488e-05,  4.2425e-04,  5.4605e-04,
         3.6838e-04, -2.5704e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.9563e-04,  9.9229e-01, -9.6741e-04, -3.3309e-04,  5.6317e-04,
         1.5183e-04, -5.0630e-04,  1.0969e-03,  7.5797e-04,  6.0766e-05,
         9.8002e-04, -6.4962e-06,  3.6853e-04, -1.0741e-03,  6.2408e-05,
        -8.8092e-05, -4.0014e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3632e-03, -9.8346e-01,  1.2171e-04,  3.2295e-04, -1.0084e-03,
        -6.5477e-05,  5.2751e-04, -4.6804e-04, -1.2821e-04, -1.0085e-03,
         7.3030e-04, -9.0649e-05, -1.1109e-04, -4.5560e-04,  1.6777e-03,
         1.0534e-03,  1.7618e-03, -1.3678e-04, -1.4890e-04, -7.7527e-04,
         2.5641e-04, -3.2586e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6430e-03,  9.7822e-01,  8.5590e-05, -1.1310e-04, -2.1358e-04,
         1.2177e-03, -1.9884e-03, -1.2363e-03,  1.5834e-03, -1.0104e-03,
         2.2965e-03,  9.0204e-04, -2.1484e-04,  2.1088e-03,  5.4294e-04,
         1.5109e-04,  1.7879e-03, -2.7952e-04,  1.1508e-03, -6.5836e-04,
         5.9182e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8031e-03,  9.8198e-01, -2.2101e-03,  4.3260e-04,  1.4516e-03,
        -6.4881e-04, -1.2744e-03, -1.0226e-04,  1.1090e-03, -1.1290e-03,
        -1.2224e-03, -2.0581e-05, -4.3064e-04,  3.3002e-04,  7.5749e-04,
        -2.2962e-04, -7.3437e-04, -1.0096e-04, -4.9277e-04, -1.6067e-03,
        -4.2419e-04, -3.7437e-04,  1.3417e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8598e-05,  9.9447e-01, -3.4860e-04,  5.2281e-04,  1.9407e-04,
        -1.8359e-04,  1.0456e-04, -2.6738e-04, -4.5480e-04, -4.3449e-04,
        -3.0256e-04,  1.2968e-04,  1.0500e-03,  2.3924e-05, -3.4777e-05,
        -6.6188e-05,  1.0151e-03,  3.5924e-04, -1.9428e-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([-7.5660e-04,  9.9255e-01, -1.3631e-05, -4.5646e-04,  2.6635e-04,
        -1.7645e-04, -5.1069e-04, -4.2038e-04,  9.0136e-04,  1.6210e-04,
         1.7176e-04,  2.9502e-04,  6.6004e-04,  5.9868e-05, -1.8082e-04,
         4.5126e-05, -5.2165e-04, -9.4136e-05, -3.1705e-04,  9.5432e-05,
         2.9490e-05, -7.0553e-04, -4.3405e-04, -1.0215e-04,  7.4992e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.1457e-03, -9.8978e-01, -1.8796e-04,  8.2529e-05,  5.2467e-04,
         1.3286e-04,  1.3604e-04,  3.7504e-04,  4.3340e-04,  9.0847e-04,
         4.5313e-04,  2.6321e-04, -3.9899e-04,  1.3994e-04, -2.4811e-04,
         2.6379e-04,  1.6315e-04,  2.3770e-04,  6.1973e-04,  8.1973e-05,
        -1.9943e-04,  1.1242e-05, -5.4260e-05, -1.5817e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #600: [tensor([ 4.2772e-03,  9.8176e-01,  9.0947e-04,  6.6668e-04, -6.4947e-04,
        -5.0831e-04, -1.0255e-03,  6.9652e-04, -5.8336e-04,  2.6565e-04,
         4.7954e-04, -7.0945e-04,  5.1544e-04,  8.5239e-05, -7.0368e-04,
         4.8105e-04, -3.6865e-04, -3.1786e-04,  1.4513e-04, -5.2175e-05,
        -8.3956e-05, -7.8936e-04,  6.9621e-04, -5.2687e-04, -1.6469e-03,
        -3.8592e-05,  6.3609e-04, -3.8360e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.4448e-03,  9.8067e-01,  3.6899e-04,  2.0646e-04, -1.0237e-03,
        -1.1601e-03, -3.6484e-04, -6.1637e-04,  3.3512e-04,  2.6472e-04,
         1.5738e-05,  2.6836e-04,  1.5326e-03, -3.1838e-04,  9.2062e-04,
        -9.0752e-04, -7.3842e-04,  1.3176e-03, -5.2334e-04,  3.8557e-05,
         6.7890e-04,  1.9580e-03,  1.1803e-03, -1.4197e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.1603e-03,  9.8202e-01,  3.1840e-04, -1.1300e-03, -1.6579e-04,
        -4.5723e-04, -8.6831e-04, -2.8168e-04, -3.0521e-04,  1.1278e-04,
         8.5050e-04, -7.1267e-05, -7.4643e-04, -3.0919e-04, -4.4037e-04,
         2.2958e-04, -2.2687e-04,  1.0496e-05, -4.2999e-04, -5.8684e-04,
        -5.7954e-04, -1.5244e-04, -2.0367e-04, -3.4549e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9039e-03,  9.8252e-01, -8.5003e-05, -1.3744e-04,  1.1482e-04,
         6.3175e-05, -8.2563e-04, -3.7555e-04, -8.1671e-05, -1.0022e-03,
        -9.5910e-04, -2.0791e-04, -3.6698e-04, -1.9490e-03,  7.1195e-05,
        -1.1214e-03,  4.3147e-04, -7.8962e-05,  2.7825e-04, -6.2439e-04,
         5.5827e-04, -6.3172e-04,  3.8741e-04,  2.3648e-04,  7.1399e-04,
         1.6339e-04,  2.6291e-04,  1.1303e-03,  7.2046e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9599e-03,  9.7927e-01,  1.0304e-03,  1.0078e-03, -5.6626e-04,
        -1.7647e-03, -2.4436e-04,  4.9353e-04,  2.0668e-03,  3.7092e-05,
        -1.6649e-04,  1.3124e-03,  1.2026e-04, -6.8889e-04,  8.5794e-04,
         3.2526e-04,  1.3798e-04,  1.9012e-06,  1.1932e-03,  5.7238e-04,
         1.1043e-03,  6.4540e-04,  2.7410e-04,  1.7400e-03,  4.2222e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9423e-03, -9.9151e-01, -3.7491e-04, -1.9225e-04,  2.5018e-04,
         5.0858e-04,  4.1039e-04, -2.6509e-04, -4.2410e-04, -2.5262e-04,
        -1.7923e-04,  1.9094e-04, -1.5661e-04, -3.2455e-04, -3.9416e-05,
        -1.6359e-04, -3.0562e-04,  9.8595e-05, -2.4064e-04,  4.7519e-05,
        -1.3286e-04,  6.3828e-04, -2.4671e-04,  3.5604e-05,  5.4709e-05,
        -1.2529e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4894e-03,  9.9266e-01,  5.4484e-05, -3.5302e-04,  4.4140e-04,
        -2.2962e-04, -6.6928e-04, -6.0068e-04,  1.7528e-04, -2.7095e-04,
        -1.0181e-04, -1.6461e-04, -8.3470e-04, -7.6567e-05, -1.2939e-04,
         3.9052e-04,  2.8478e-05,  5.8520e-05,  2.0184e-04, -2.6676e-04,
        -2.2264e-04, -2.0096e-04, -2.5181e-04,  1.3115e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.6001e-04, -9.9507e-01, -3.5469e-04,  1.5850e-04, -4.5561e-04,
         1.6374e-04,  1.8236e-04,  2.3384e-04, -2.8466e-04, -1.7427e-04,
        -9.3896e-05, -6.0487e-04, -6.0491e-04, -3.3193e-05,  8.7401e-05,
        -1.7477e-05, -3.0100e-04, -7.8611e-06,  1.3798e-04,  8.9518e-06,
         9.2843e-05, -7.1741e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.2785e-04, -9.9475e-01, -1.9171e-04, -1.1492e-03, -1.0150e-04,
         1.5233e-04, -1.9592e-04, -9.5600e-04,  1.9285e-05, -7.9019e-05,
         4.0396e-04, -3.5642e-04,  2.8535e-04, -5.3503e-04,  3.7285e-05,
        -5.5501e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9511e-03, -9.9244e-01,  1.8323e-04,  8.9016e-04,  3.1876e-04,
        -5.0266e-04,  2.3457e-04,  2.7884e-04, -1.0483e-04,  8.2327e-04,
        -9.2388e-05, -1.8201e-04, -3.1442e-04, -6.4567e-04,  1.0385e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2192e-03, -9.9144e-01,  3.3392e-04,  5.6684e-04, -2.2378e-04,
         6.6425e-05, -1.7567e-04,  3.6132e-04, -6.0906e-04,  1.3169e-03,
        -2.1976e-04, -1.1593e-04, -2.1610e-04, -1.2949e-04,  9.3654e-05,
         4.1891e-04, -9.2039e-04,  1.0660e-07,  1.4156e-05, -1.4734e-04,
         4.1117e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7233e-05,  9.9362e-01,  2.0398e-04,  3.0013e-04, -1.1766e-04,
        -5.5528e-05, -3.3863e-04, -8.1754e-05,  5.5657e-04,  4.5909e-05,
         3.5731e-04, -2.6668e-04,  8.6950e-05,  1.3614e-04,  5.3626e-04,
         3.7642e-04, -3.6269e-05,  1.2967e-04, -3.2787e-04,  3.0741e-04,
         5.9007e-05,  1.6349e-04,  3.8101e-04, -8.2649e-05,  3.7415e-04,
        -4.3620e-05, -5.6619e-04,  4.2273e-05,  1.6063e-04,  1.9686e-04],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #650: [tensor([-1.6511e-03, -9.9317e-01,  4.2538e-04,  4.1418e-04, -4.7614e-04,
        -3.6400e-04,  4.5999e-04, -5.9456e-05, -8.7859e-04,  1.0851e-04,
         2.8838e-04,  1.0214e-04,  9.2682e-05,  6.2572e-04, -1.5175e-05,
        -6.1166e-04,  7.7953e-05, -1.2456e-04, -5.4868e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3208e-03, -9.9273e-01,  1.6596e-04,  9.7578e-05,  1.2520e-04,
         4.5181e-06, -3.6359e-04, -6.7840e-04, -9.7219e-05, -1.7325e-04,
        -1.7199e-05,  4.7392e-04,  8.2136e-05, -6.0467e-04,  2.6319e-04,
        -1.4079e-04,  4.9301e-05, -5.4298e-04,  4.0141e-04, -5.6550e-04,
        -1.0265e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4256e-03,  9.9130e-01,  1.1524e-04, -1.2861e-04, -1.0388e-04,
         1.3487e-04,  6.2127e-04,  8.1282e-04,  2.5849e-04,  6.1084e-04,
        -4.4960e-04, -2.8749e-04, -4.2821e-04, -3.9547e-04, -4.7349e-04,
         1.3969e-04,  1.4814e-05,  4.6559e-04, -4.1755e-04,  1.5647e-04,
        -2.5599e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.2153e-04, -9.9217e-01,  5.3015e-04, -5.4379e-04,  5.1540e-04,
         7.1641e-05,  7.2496e-04,  3.1308e-04,  4.1354e-05, -4.2355e-04,
        -2.9983e-04,  3.7803e-04,  1.4834e-04,  9.3105e-05, -7.4299e-04,
         1.9510e-05, -6.8229e-04, -2.4862e-04, -1.0149e-04,  8.9011e-05,
        -3.2705e-04, -1.0158e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.9226e-03,  9.7502e-01,  2.2975e-03,  9.3865e-04,  1.1211e-03,
        -8.9781e-04,  5.0136e-04,  4.1781e-04, -5.7459e-04,  1.4453e-03,
        -1.2060e-03,  7.8618e-04, -6.5610e-04,  4.8416e-04, -5.7218e-04,
        -5.6288e-04,  1.0398e-03,  4.2330e-03, -1.0583e-03,  2.6128e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4324e-04,  9.8650e-01, -3.3623e-04, -1.3453e-03, -4.1558e-04,
        -5.0365e-04,  5.7404e-04, -4.1841e-04, -3.4435e-04, -1.4981e-03,
        -1.0490e-03,  9.6715e-04,  7.8491e-04,  7.0543e-04, -3.4982e-04,
        -1.1065e-03, -3.9706e-04, -2.4197e-04,  1.2522e-04, -7.8767e-04,
         8.5106e-04, -4.5970e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8593e-03, -9.7777e-01, -8.4869e-04,  1.8853e-03,  1.3344e-03,
         1.2507e-03,  2.5340e-04,  1.4989e-03,  2.7145e-03,  8.9758e-04,
         5.9593e-04, -2.1160e-03,  1.4920e-03,  5.4807e-04,  1.0891e-04,
         3.0347e-03, -3.8918e-04,  3.0053e-04, -1.1033e-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([-1.1220e-03,  9.8678e-01, -1.2332e-03,  3.1935e-04,  5.4007e-05,
         1.9377e-03,  2.2285e-03,  6.7579e-04, -5.5830e-04, -7.8441e-04,
         1.1955e-03, -5.3199e-04, -2.1052e-04,  6.9893e-04,  1.0929e-04,
         1.0321e-04,  4.8001e-05, -2.9274e-04, -4.0848e-04, -7.1176e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1424e-02,  9.7949e-01,  2.0008e-04, -3.1879e-04,  2.2857e-04,
        -1.8117e-03,  1.2215e-03, -9.2612e-04,  3.0456e-05,  3.7955e-04,
         1.7190e-04, -4.0149e-04,  2.8846e-04, -7.6163e-04, -9.3034e-04,
         1.0777e-03, -3.4236e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.9003e-03,  9.8731e-01, -1.4369e-03,  1.5509e-04, -8.1112e-04,
        -6.2460e-04,  5.6123e-04, -2.5941e-04, -5.7252e-04, -1.5789e-03,
         4.9750e-04,  3.0788e-04,  4.1694e-04,  3.4129e-04, -2.2916e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5903e-03, -9.8902e-01, -3.8117e-04,  6.2892e-04, -5.7047e-04,
         2.6590e-04, -1.6884e-04, -3.9864e-04,  4.4435e-04,  1.5239e-04,
         1.2328e-04, -6.0281e-04,  7.7256e-04, -1.8490e-05, -3.3279e-04,
         5.3535e-04, -1.7042e-04,  2.7887e-04,  2.0060e-04, -1.7497e-04,
        -4.1942e-04, -1.2388e-04, -4.8355e-04,  1.4297e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5738e-03, -9.8634e-01, -8.9979e-04,  2.8547e-04, -3.3383e-04,
        -3.1690e-04, -3.1022e-04, -1.3667e-04, -9.0327e-05, -5.5880e-04,
         2.2612e-04, -2.0714e-04,  6.5800e-04, -1.0670e-03,  8.2697e-05,
        -4.9069e-04,  5.6219e-05,  7.5661e-04,  7.1920e-04, -7.7536e-04,
        -2.4456e-04, -9.9658e-04,  1.2115e-04, -1.2824e-03,  4.7372e-04],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #700: [tensor([-3.6064e-04, -9.9204e-01, -2.2246e-04,  8.3387e-05,  2.6455e-04,
        -1.6263e-04, -1.3280e-04, -2.7037e-04,  3.4527e-04, -1.5484e-04,
        -2.0158e-06,  1.8295e-04, -5.9027e-04,  2.0700e-05, -2.1385e-04,
        -5.3800e-04,  2.3634e-04,  7.5610e-05, -1.0197e-04,  1.0616e-04,
         6.8899e-05,  1.3573e-04, -3.8013e-05, -3.7924e-04, -3.2154e-04,
        -1.6928e-04,  2.4831e-04, -2.5712e-04, -4.5224e-06,  3.1657e-06,
        -4.4388e-04, -1.4709e-04, -6.2811e-05, -8.0025e-05, -7.0087e-05,
        -1.7815e-04, -3.8438e-04, -1.5677e-05,  1.9904e-04,  3.0992e-04,
        -3.6447e-04,  1.3687e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.6247e-03, -9.8476e-01, -2.4495e-04, -2.3106e-04, -1.4303e-04,
        -2.9836e-04, -1.6033e-04, -1.1196e-04, -3.1765e-04, -1.7623e-04,
        -3.1379e-04,  7.4150e-05, -2.0913e-04, -2.4411e-04, -3.1257e-05,
        -2.9694e-04, -2.5177e-04,  4.1293e-05,  1.5976e-05,  3.9246e-06,
         1.3491e-05, -2.9236e-04, -3.1016e-04,  5.7064e-04, -1.3388e-04,
        -2.6514e-04, -2.6792e-04, -9.2307e-05,  1.1058e-04, -9.1020e-05,
        -7.4596e-04, -3.1040e-04, -4.8131e-05,  5.5089e-05,  1.1219e-04,
         4.6330e-05, -1.3901e-04, -8.9591e-05,  5.5312e-05,  1.2026e-04,
        -1.2333e-04,  1.2037e-05, -2.8280e-04,  1.6190e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5853e-03, -9.9211e-01, -6.7620e-05, -1.5483e-04, -1.7442e-04,
        -4.1474e-05,  1.5755e-04,  4.5974e-05, -1.0371e-04, -7.3070e-05,
         1.1974e-04, -1.3017e-04, -1.9336e-04, -3.6108e-04,  1.2773e-04,
         7.7288e-05, -3.8280e-04, -1.8405e-04, -3.4840e-04, -3.0754e-04,
         9.8541e-06, -2.7435e-04, -1.2671e-04,  1.2858e-05, -1.8855e-05,
         7.2153e-05,  1.6984e-06, -2.8005e-04, -9.2611e-05, -2.0082e-04,
        -1.0345e-04,  1.4813e-04,  7.0005e-05, -4.9740e-04, -3.3745e-04,
         1.3513e-04,  2.0351e-04,  1.5574e-04, -4.3493e-04,  7.8077e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9409e-04,  9.9141e-01,  3.4179e-04, -2.0808e-04, -1.1230e-04,
         3.3005e-04,  8.5387e-05,  3.2066e-04, -5.1227e-04, -6.1724e-04,
         5.5952e-05,  1.8274e-04, -3.7911e-04,  1.0079e-04,  1.2385e-04,
        -4.9282e-05,  3.1527e-04,  2.0495e-04, -2.5930e-04,  1.5292e-04,
         3.0823e-04, -1.1870e-04, -3.0477e-04, -8.2086e-05,  2.1632e-04,
        -2.8128e-04,  6.7152e-05,  4.2000e-04, -1.0957e-04, -2.4174e-04,
         6.1781e-05,  1.7233e-04,  2.2358e-04, -1.1559e-04, -3.2074e-04,
         3.3095e-04,  1.8289e-04,  9.1087e-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([-1.6420e-03, -9.9066e-01,  1.4912e-04,  4.3904e-04, -6.5843e-06,
        -3.4544e-04, -2.1721e-04,  1.8130e-04,  1.7153e-04,  2.4588e-04,
        -2.7765e-04, -2.3606e-04, -4.5609e-04, -2.6267e-04,  9.2479e-06,
         2.2363e-04, -4.2333e-04,  1.0613e-04, -2.0915e-04, -2.4692e-04,
         2.8558e-04,  1.7456e-04, -3.3658e-04,  3.7570e-04, -1.7057e-04,
         1.9123e-04,  2.6870e-05,  1.8025e-05,  5.8463e-05, -9.9360e-05,
        -1.0248e-04, -4.5741e-04, -1.3308e-04,  5.5833e-05,  1.0843e-04,
         1.6426e-05, -3.9256e-04,  1.3248e-04,  3.5289e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1877e-04, -9.8499e-01, -1.1862e-03,  3.1872e-04, -3.8365e-04,
         6.6635e-04,  5.4223e-05,  3.0511e-05,  6.5415e-04,  5.4190e-04,
         2.3254e-04,  1.0040e-03, -3.1945e-04,  1.9146e-04, -6.1549e-04,
        -2.9611e-04, -1.2413e-03,  4.0988e-04, -1.8926e-04,  7.0829e-05,
         1.6550e-04,  2.3314e-05, -3.4878e-04,  4.3548e-04, -5.2799e-05,
         1.5511e-03, -1.3752e-03,  5.1475e-04, -7.1877e-04, -4.8220e-04,
        -7.2204e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7500e-03,  9.8541e-01,  3.2142e-04, -6.3337e-05, -3.8432e-04,
        -6.3961e-04, -3.5894e-04, -2.6365e-04, -4.0134e-04, -2.9209e-04,
        -4.6241e-04, -8.4420e-04, -3.3106e-04, -3.7748e-04, -2.7142e-04,
        -5.1611e-04,  8.7975e-04, -1.0345e-04, -4.5461e-05, -5.3815e-04,
         5.5361e-04, -3.5516e-04,  1.2826e-04,  9.3707e-04, -5.2083e-04,
        -1.6395e-03, -1.2066e-05,  3.9878e-04, -1.0982e-04, -9.3927e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5444e-04,  9.8841e-01, -1.8223e-04, -1.5860e-04,  5.6723e-04,
        -4.2484e-04,  5.7509e-05, -2.6167e-04,  4.6615e-04, -3.0750e-05,
         1.1249e-04, -5.7125e-04,  3.7472e-04, -1.3251e-04,  5.8477e-04,
        -3.4379e-04,  8.7655e-04, -1.1683e-03, -8.2988e-04,  2.8650e-04,
         1.1014e-04, -1.7590e-04,  2.2832e-04, -3.4033e-04, -5.0845e-04,
        -2.9334e-04,  2.1516e-04, -8.9124e-04, -2.0095e-04,  3.3252e-04,
         6.0748e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4161e-03, -9.9007e-01,  4.3564e-04, -2.8136e-04,  2.2613e-04,
        -3.0510e-04,  7.5579e-05, -1.6223e-04, -1.2400e-04,  2.4172e-04,
         6.2899e-04, -1.0003e-05,  2.5073e-04,  4.1406e-04,  5.5488e-04,
        -1.7717e-04, -2.5448e-05,  1.7899e-04, -2.9543e-04, -1.9243e-04,
         4.3304e-04, -4.5613e-04,  6.2956e-04, -1.7547e-04,  6.8555e-06,
         4.2818e-04,  5.1520e-04, -1.5803e-04, -4.1821e-05, -8.1317e-05,
        -4.5460e-04, -5.5546e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0348e-02, -9.7895e-01,  7.2288e-05, -5.3362e-04,  1.9283e-04,
        -7.5259e-04, -1.2269e-03, -9.8678e-04,  1.2962e-04, -2.4150e-05,
         7.8662e-04, -3.8257e-04, -1.4302e-03, -1.7801e-04, -1.8759e-04,
        -6.5453e-04,  1.0567e-04,  9.0964e-06, -3.8474e-04, -4.9595e-04,
        -8.1639e-04, -7.2183e-04,  5.1625e-05, -4.8524e-04, -5.6808e-06,
         8.5601e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.0000e-03, -9.8228e-01, -1.3178e-03, -4.3853e-04, -3.5961e-04,
        -5.8463e-04, -1.0483e-03, -7.9448e-04, -6.8308e-04, -1.3393e-03,
         3.0776e-04,  7.7419e-04,  4.3405e-04, -8.9185e-04, -6.4455e-04,
         1.9684e-04, -1.0618e-04, -3.5011e-05,  4.0913e-04, -2.0383e-04,
         1.2607e-04, -1.0838e-04,  2.4518e-04,  1.7405e-04, -1.0499e-03,
         1.7869e-04,  2.6902e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1240e-04,  9.8956e-01, -1.0524e-03,  5.6079e-04,  1.3398e-06,
         9.5409e-05,  1.5145e-05,  5.3272e-04, -2.4152e-04,  2.0735e-04,
        -4.0858e-04, -2.2297e-03,  4.5099e-04, -7.3178e-04, -8.5730e-05,
        -4.1894e-04, -3.4584e-04,  1.3442e-03,  1.3824e-05,  6.7879e-04,
         6.8600e-04, -1.3117e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #750: [tensor([ 3.9230e-04, -9.9487e-01, -1.0547e-04,  3.9973e-04,  3.4803e-04,
         1.4493e-04, -1.0649e-04,  6.0830e-04,  2.5946e-04,  1.7383e-04,
        -2.5775e-04,  1.1293e-04, -1.0952e-04,  1.2410e-04,  4.0609e-04,
        -3.4619e-04, -2.8791e-04,  2.1810e-05, -1.5200e-05, -1.9959e-04,
        -4.0453e-04, -1.3884e-04, -1.6498e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3342e-03, -9.9169e-01,  2.1781e-04, -6.2841e-05,  2.9668e-04,
         3.5767e-04, -3.4380e-04, -3.3359e-05,  7.8792e-04,  4.6972e-05,
         5.0779e-04,  7.0145e-05, -1.8073e-04,  1.7313e-04,  2.4768e-04,
         9.1929e-05, -5.3790e-04,  1.0394e-04, -6.8155e-04, -8.6097e-04,
        -2.7846e-04, -5.6677e-04, -2.7204e-04, -2.5584e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0428e-03, -9.9025e-01,  4.5378e-04,  2.0574e-04, -1.6257e-04,
         4.6867e-04,  2.3479e-04,  3.7669e-04,  5.0207e-05, -2.6359e-04,
         2.7508e-04, -1.7283e-04,  3.2572e-04, -2.4806e-04,  6.7949e-05,
        -1.9301e-04, -6.2803e-04, -2.4344e-04,  1.5405e-04, -1.1079e-04,
        -5.8228e-05, -6.7039e-05, -3.7259e-04, -3.1078e-04, -2.5377e-04,
        -1.5640e-04, -3.1165e-05,  1.8922e-04, -2.3677e-04,  1.1288e-04,
         2.3214e-04, -5.1528e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0225e-03, -9.8944e-01,  4.2349e-04, -1.3368e-04, -8.6890e-05,
         2.3480e-04, -2.1040e-05,  5.7644e-04, -8.8205e-05, -2.7456e-05,
        -5.0175e-04,  5.1592e-04,  2.1880e-04,  5.1505e-04,  7.1980e-04,
         1.4142e-04,  5.2354e-04,  2.9498e-05,  1.0572e-03, -4.1037e-04,
         2.0821e-04, -5.9217e-04, -2.4383e-04, -3.1391e-04, -6.1750e-04,
        -1.8276e-04,  4.6930e-04,  2.8418e-04,  1.7433e-04, -2.2230e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8749e-04, -9.9161e-01,  3.9879e-04, -6.6838e-05, -1.4590e-04,
         2.9479e-04,  7.0529e-05,  1.2278e-03,  4.3688e-04,  5.5251e-04,
         1.6924e-04,  8.6138e-05,  2.9721e-04, -3.3639e-05,  2.0137e-04,
         1.3867e-04,  4.4342e-04,  9.6486e-05,  5.4177e-04, -2.4546e-04,
         4.7196e-05,  1.2162e-04, -1.7013e-05,  2.5787e-04, -2.1731e-04,
         6.9015e-04,  1.5775e-04, -4.8673e-05,  3.5890e-04,  2.3679e-04,
        -3.7856e-04,  6.7669e-05,  6.0193e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9487e-03,  9.7025e-01, -1.7105e-03,  4.8713e-04, -1.9515e-03,
        -3.0947e-03,  1.6815e-03, -6.5452e-04,  1.0114e-04, -2.4475e-03,
        -1.2573e-03, -1.4717e-03,  4.5674e-04,  9.1942e-04,  2.3980e-03,
         1.5101e-03, -8.7568e-04,  2.0759e-03, -4.6595e-05, -6.6192e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6361e-03,  9.8413e-01, -6.5170e-04,  4.5481e-04,  1.7618e-03,
        -8.3793e-04,  2.5897e-03,  4.0361e-04,  5.0866e-04,  1.0019e-03,
         8.9113e-05, -2.6083e-04,  1.0450e-03,  7.3522e-04,  9.7646e-04,
         5.5528e-04,  1.4620e-04, -4.9958e-04, -6.5838e-04, -4.7164e-04,
         8.4060e-06,  5.7781e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.8176e-03,  9.7694e-01,  1.2184e-03,  9.7022e-04, -1.3156e-03,
         1.6575e-03, -2.2380e-04, -1.3511e-03,  6.9860e-05,  9.7618e-04,
        -2.1265e-03, -6.4944e-04,  7.1600e-04,  1.0028e-03,  2.3228e-04,
        -8.9996e-04, -2.4435e-04, -9.6066e-04,  7.7145e-05,  1.5503e-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([ 2.8830e-03, -9.8759e-01,  6.2576e-04,  4.6222e-04,  7.3794e-04,
         2.1425e-03,  2.0534e-04, -7.0503e-04, -8.5058e-04, -2.2594e-04,
        -2.6413e-04,  5.1695e-04, -7.7066e-04, -7.3363e-04, -4.8835e-04,
         7.8516e-04,  1.4914e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2787e-04,  9.8653e-01, -9.0236e-04,  3.5101e-04, -3.7172e-04,
        -2.3123e-03,  1.1438e-03,  1.0152e-03,  7.6184e-04,  8.2614e-04,
         8.5371e-04,  7.0983e-04,  1.8496e-03, -2.0563e-04,  1.1231e-03,
         6.1506e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.7453e-03,  9.8166e-01, -2.6293e-03,  8.4438e-04, -1.1447e-03,
        -8.8203e-04,  4.6669e-04, -1.6778e-03,  1.7014e-03,  3.6501e-04,
        -1.8174e-04,  1.7467e-04, -7.0537e-04, -1.3319e-04,  2.0175e-03,
         6.6972e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6383e-03, -9.8834e-01,  5.0118e-04,  2.7316e-05,  4.3628e-04,
        -3.0450e-04,  3.3317e-04,  3.4579e-04,  7.6458e-04,  6.1393e-04,
         6.1961e-04,  2.2442e-04, -3.1957e-04,  7.5533e-04,  2.2948e-04,
         4.8207e-04,  5.1585e-04,  1.2564e-04, -1.5379e-04, -1.9671e-04,
         5.8461e-05,  2.2157e-05,  1.0734e-05, -1.0493e-04,  2.4323e-04,
        -1.7139e-04, -4.6648e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #800: [tensor([-1.4592e-04, -9.9432e-01, -2.7736e-04, -7.3514e-05,  4.8418e-04,
         1.3248e-04,  2.5834e-04, -7.1712e-04,  5.6271e-04,  3.4329e-04,
        -7.7645e-05, -1.1074e-04, -3.1788e-04, -4.4231e-05,  6.6151e-05,
        -2.9287e-04, -3.1383e-04, -3.6380e-04, -2.8949e-04, -6.3328e-04,
         9.5099e-05, -7.8460e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3925e-04, -9.9100e-01, -8.5895e-04, -1.2954e-04, -7.5505e-05,
        -6.5584e-04, -1.0280e-04, -2.8990e-04, -6.5462e-05, -4.9715e-04,
        -2.4643e-04, -2.3733e-05, -5.0339e-04, -9.2116e-04, -5.5935e-04,
        -2.4746e-04, -4.4371e-04,  2.0288e-04,  1.5181e-04,  6.2676e-05,
         7.2903e-05, -1.0439e-04, -4.5773e-04,  9.6217e-05, -3.4196e-04,
        -8.3105e-05,  2.6092e-04, -6.0026e-04, -5.0367e-04,  2.2265e-04,
         7.8699e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.5363e-03, -9.9240e-01,  6.2973e-04,  6.3594e-04,  3.6870e-04,
        -7.9859e-05,  1.5918e-04,  1.4115e-04, -2.8355e-05,  2.9640e-04,
        -2.8238e-05, -6.3501e-04,  3.6042e-04,  2.2190e-04,  1.1039e-04,
         1.3154e-04,  1.1703e-04, -3.6858e-04,  4.0092e-05, -6.1430e-04,
        -3.4141e-05,  6.1451e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.8985e-03, -9.9333e-01, -1.3784e-04, -6.2442e-04,  1.7859e-05,
         3.9857e-04,  1.3967e-04,  3.2972e-05, -4.1599e-05,  2.1691e-04,
         3.5430e-04,  8.7894e-05,  4.7833e-04,  5.0799e-05,  4.2230e-04,
         4.9348e-07, -3.0124e-05, -2.4181e-04,  2.3371e-04,  2.2302e-04,
        -2.0489e-05,  5.4770e-04, -8.1663e-05, -1.4928e-05,  3.7710e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1345e-03, -9.8529e-01, -9.4445e-04,  3.2145e-04,  1.4591e-03,
        -4.3887e-04,  4.4327e-04, -3.0682e-04, -3.4945e-05,  6.1552e-07,
        -2.9907e-04,  1.2127e-03,  2.0311e-04,  6.9439e-04,  1.0041e-03,
        -3.4724e-04,  1.1488e-03, -1.8526e-04, -7.7295e-05, -5.2347e-04,
        -5.0553e-05,  1.0298e-03,  3.4415e-04, -4.7539e-04,  4.7529e-04,
        -2.5490e-04, -2.9826e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0100e-04, -9.8958e-01,  4.6834e-05, -8.7880e-04, -2.6410e-04,
         4.7483e-04,  2.1399e-04, -1.0376e-04,  9.7605e-04,  2.6106e-05,
        -3.6435e-04,  2.0089e-04, -3.7671e-04, -1.7781e-04, -7.9634e-04,
         6.5212e-04,  4.7538e-04, -5.9738e-04, -4.5159e-04,  2.4273e-04,
         2.3697e-04, -4.5735e-04, -4.9615e-04,  3.2535e-05, -5.1707e-04,
        -2.4917e-04, -5.3344e-04, -1.1930e-04, -2.6073e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.5798e-04, -9.8745e-01, -6.0313e-04,  1.2903e-04,  7.7727e-04,
         6.8420e-04, -1.4023e-05,  4.5758e-04,  2.2940e-05, -4.9822e-04,
        -3.5278e-04,  1.5242e-03,  3.1098e-04,  9.8098e-05,  9.5735e-04,
         3.8124e-04,  2.6839e-04, -8.7993e-05,  2.2440e-04, -6.2750e-05,
        -3.8244e-04,  5.9097e-04,  3.7693e-05, -4.6906e-04, -3.9820e-04,
         4.1533e-04,  5.8727e-04, -1.0432e-03, -2.1024e-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,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3785e-03,  9.8626e-01,  2.3024e-04, -1.3603e-04, -6.7752e-04,
        -4.1347e-04, -1.4691e-04,  9.0841e-04, -5.0644e-04,  2.4892e-04,
        -2.1976e-05, -6.2466e-05, -6.1242e-04,  2.6144e-04,  8.9659e-05,
        -6.4210e-04, -2.4531e-04, -3.4527e-05, -9.5073e-05, -3.7735e-04,
         1.7890e-04,  2.1916e-04,  4.2568e-04,  3.1188e-05,  6.3842e-04,
         1.1859e-04, -1.3963e-04, -7.1788e-04,  9.3550e-04, -4.0160e-04,
        -6.5984e-04,  2.4505e-04, -7.6795e-04,  5.4301e-04,  1.2729e-04,
         1.0553e-04,  3.5584e-04,  4.0079e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4944e-04, -9.8469e-01, -2.4470e-04,  5.8144e-04,  3.2099e-04,
        -2.4535e-04, -7.2739e-04, -3.4635e-04,  6.8152e-04,  3.9325e-05,
        -2.0939e-04,  7.4073e-05, -4.6086e-04, -9.8416e-05,  3.4929e-04,
         5.8284e-04, -3.9748e-05,  1.6387e-04,  1.9492e-04, -2.3098e-04,
        -3.7459e-04,  6.3141e-04, -3.4278e-04, -7.0544e-05,  1.1789e-03,
         8.5834e-05, -7.7916e-04,  4.2549e-04, -2.7190e-04, -3.6776e-04,
         1.2819e-04, -1.8171e-04,  9.8354e-05, -3.0045e-04, -3.1702e-04,
         4.6965e-04,  5.4939e-04,  3.4917e-04, -7.1053e-04,  7.6123e-04,
         1.3645e-04,  2.9780e-04, -2.0692e-04,  4.3653e-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([-4.0937e-03, -9.7916e-01,  3.2225e-05,  1.1474e-03,  9.7248e-04,
         1.1747e-04,  1.1791e-04, -2.3442e-04,  4.4242e-04,  8.7342e-04,
         1.9045e-06, -8.6586e-05, -3.8032e-04,  3.3562e-04, -2.1263e-04,
         3.0784e-04, -3.0313e-04,  3.0386e-04, -2.4296e-04,  2.6951e-04,
         5.7364e-04,  1.7076e-05,  2.1490e-04,  4.0010e-04,  8.9295e-04,
         3.7835e-04, -2.7690e-04,  6.4112e-04, -3.6871e-05,  5.1163e-04,
         1.3381e-03, -1.6352e-04,  6.6827e-04,  4.8264e-04, -3.1513e-04,
         8.6899e-04, -7.2402e-04, -5.6848e-05, -6.4392e-04, -7.8989e-04,
        -6.0833e-06,  3.6147e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3664e-03, -9.9116e-01, -1.0462e-04,  6.9158e-05, -7.0233e-05,
        -2.4583e-04, -2.4004e-04,  1.4653e-04,  8.6224e-05, -8.1668e-05,
         3.8242e-04,  1.3913e-04, -4.3539e-04,  2.6185e-04, -1.0753e-04,
         3.4655e-05,  2.0511e-04,  8.7406e-05,  1.2440e-05, -5.1474e-06,
        -1.3522e-04,  9.4171e-05,  1.5524e-04,  1.5302e-04,  4.0076e-04,
        -1.0299e-04,  1.7380e-04,  1.0105e-05,  1.2425e-04,  9.3186e-05,
         1.8346e-04, -6.4680e-05, -8.2661e-05, -3.4893e-05, -1.6229e-04,
         1.7671e-04, -1.7403e-05, -1.3764e-05,  2.9173e-04, -4.7411e-05,
        -1.1523e-04, -1.8390e-04, -1.4648e-04, -4.7369e-05,  1.6533e-04,
         1.1822e-04,  1.2448e-04,  2.7100e-05,  1.0790e-05,  1.5372e-04,
         1.4907e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0265e-03, -9.9290e-01,  1.9261e-04, -1.7730e-04,  4.4936e-04,
        -1.2624e-04, -2.3263e-04, -1.1288e-04, -3.7702e-04, -1.1106e-04,
         2.3718e-04, -4.4296e-05, -1.6698e-04,  7.5623e-05, -2.9281e-05,
         3.5257e-05,  4.3951e-05,  2.1641e-04, -9.5334e-05, -3.7194e-04,
        -4.0213e-04,  9.9953e-05, -5.7844e-05, -3.9802e-04, -2.7583e-05,
         1.3344e-04, -3.4926e-05, -5.1567e-05, -2.4163e-04,  9.8852e-06,
        -1.6000e-04, -1.0378e-05, -3.7981e-04,  4.0859e-05,  1.0365e-04,
        -1.7360e-04, -3.3368e-04, -1.2201e-05, -1.8122e-04, -3.5869e-05,
        -9.0277e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #850: [tensor([-2.3743e-03,  9.9375e-01,  2.5041e-04,  3.8790e-04, -1.9196e-04,
         2.2660e-05, -3.8508e-04,  2.8041e-05,  7.6300e-05,  2.0816e-04,
        -8.3537e-05, -3.3218e-05, -2.7657e-05,  3.1900e-04, -7.0441e-05,
         4.7475e-04,  1.9025e-04,  3.6694e-04,  3.0597e-04,  3.5982e-04,
         9.4498e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.7246e-04, -9.9535e-01, -9.4973e-05, -4.2416e-05, -2.7133e-04,
        -2.0461e-04, -3.9535e-04,  3.5073e-04,  5.6683e-06,  1.5467e-04,
         3.4602e-05, -2.5520e-05, -2.9738e-04,  8.6442e-05, -1.1924e-04,
        -9.6601e-06,  5.8932e-05,  2.5538e-04,  2.4128e-04,  2.6979e-04,
         2.3290e-04,  4.8777e-04,  4.2760e-05, -1.9201e-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([ 7.7623e-04, -9.9443e-01, -8.1142e-05,  1.7131e-04,  4.4198e-04,
        -1.0350e-04,  1.2962e-04, -6.2169e-04, -2.0002e-04, -2.2497e-04,
         3.7681e-04,  9.8924e-05, -1.2918e-04,  7.0219e-06, -2.4128e-04,
        -3.5275e-04, -1.4935e-05,  3.3065e-04, -8.6962e-05,  3.4795e-04,
        -3.7292e-04, -1.5835e-05, -2.9769e-04,  1.4424e-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([ 4.9032e-04, -9.9315e-01,  3.5477e-04, -1.0380e-04,  1.6950e-04,
         6.9618e-04,  4.1326e-04,  3.3577e-04,  4.8585e-04,  6.5018e-04,
        -1.1085e-04, -3.0127e-04, -3.8892e-04,  4.3332e-04, -5.6066e-04,
        -4.1204e-04, -5.5225e-04, -1.5960e-04, -1.4480e-04, -8.7213e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.6685e-04,  9.9306e-01, -4.3914e-04,  1.1478e-04,  1.9593e-04,
        -7.2738e-04, -5.7423e-04, -4.4727e-04,  2.5683e-04,  2.3008e-04,
        -4.9883e-04, -3.8324e-04, -2.8616e-04,  4.0347e-04,  3.1537e-04,
         3.6095e-04, -1.2915e-04, -4.5256e-04,  1.5133e-04,  1.0538e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5398e-03,  9.9318e-01,  1.3184e-04, -3.4168e-04,  7.2442e-05,
         2.3279e-04, -5.1891e-04,  4.3741e-04,  1.5496e-04,  4.6727e-04,
         2.4494e-04,  1.3329e-05, -1.3977e-04, -1.3084e-04,  6.1411e-04,
         1.6262e-04, -1.6277e-04, -2.4502e-04,  5.8617e-04,  1.4935e-04,
        -1.8994e-04, -2.8145e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.6650e-04, -9.9393e-01,  2.1683e-04, -2.5620e-04, -1.5738e-04,
        -2.6340e-04, -7.6932e-05, -1.0788e-04, -4.1318e-04,  2.1757e-04,
         4.2863e-04,  3.0780e-04, -4.6702e-04, -3.4612e-04,  2.2838e-04,
        -8.8456e-05,  1.4727e-04, -1.6731e-04,  2.9438e-04, -1.1003e-04,
        -8.9929e-06,  1.6279e-04, -7.1934e-05,  4.2190e-04,  3.0682e-05,
         3.2156e-04,  1.8740e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6154e-03, -9.9158e-01, -4.1491e-04, -2.7705e-04, -3.9736e-04,
        -4.1120e-04, -2.4103e-04,  1.7066e-04,  2.5389e-04, -2.1584e-05,
         1.8343e-04, -1.3061e-04, -2.1151e-04,  5.3998e-05, -7.9044e-05,
         1.8018e-04,  1.8682e-04,  4.2942e-04,  4.0350e-04, -9.1370e-05,
         1.3389e-04, -1.5205e-05,  1.6393e-04,  2.8076e-04,  1.5639e-04,
         1.4703e-04, -3.4116e-05,  1.0974e-04, -1.8058e-04, -6.3279e-05,
         3.8291e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.4994e-05,  9.9424e-01,  1.4695e-04, -1.2082e-04,  6.7388e-04,
         3.4529e-04, -1.0951e-05, -3.5680e-05, -2.8213e-04, -5.9398e-04,
        -4.6435e-04,  3.7779e-04,  1.3759e-04, -1.5066e-04,  3.6320e-04,
         8.6991e-05, -2.7676e-05,  8.9821e-05,  9.5746e-05,  3.0708e-04,
         4.6995e-04, -2.6127e-04,  1.7918e-04, -3.4488e-04, -6.5534e-05,
        -5.2917e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7968e-04, -9.9221e-01, -6.4438e-04,  6.7767e-04,  9.4352e-05,
         2.0539e-04, -5.8953e-04, -2.0794e-04,  1.3926e-04, -1.4931e-04,
         2.5284e-05, -2.4709e-04,  1.2288e-04,  1.2725e-04,  7.2423e-05,
         6.1841e-04,  3.8784e-04,  6.9733e-04, -5.2419e-04,  3.5936e-04,
         2.7967e-04,  2.2264e-04,  2.7988e-04,  4.1036e-04, -2.0260e-05,
         3.0459e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.3552e-06, -9.9200e-01, -4.7645e-04, -4.5595e-04, -1.9284e-04,
        -9.5985e-05,  4.6654e-04, -6.4699e-04, -1.1313e-03, -1.1745e-04,
         1.0133e-03,  2.2331e-04,  8.0076e-04, -1.4463e-05,  7.4149e-04,
         2.5867e-04, -2.9750e-04, -1.0375e-04,  6.6331e-04,  2.9393e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4805e-03,  9.9106e-01, -1.4366e-04, -3.7471e-04, -3.4383e-04,
         3.6844e-05,  1.8757e-04,  2.2026e-04, -6.4513e-04,  2.1588e-04,
         3.6905e-04, -2.7986e-04, -2.5792e-04,  1.3878e-04, -5.1115e-04,
        -2.3292e-04, -5.4336e-04, -6.9633e-04, -3.6710e-04, -6.3946e-04,
         6.6017e-04, -6.1275e-05, -5.3469e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #900: [tensor([ 1.2510e-03,  9.9194e-01, -2.2017e-04,  3.1928e-04,  4.6960e-04,
         2.1840e-04,  4.1143e-04,  5.8775e-04,  2.4104e-04, -1.0384e-04,
         6.4554e-05, -3.5892e-04,  3.1870e-04, -1.5085e-04,  1.0431e-03,
        -5.1540e-05, -3.1485e-04, -8.2455e-05,  1.7568e-04,  2.4513e-04,
        -9.5250e-05,  2.8446e-04,  5.4200e-05,  5.1666e-04,  1.1088e-04,
         4.6903e-06,  1.8918e-04, -8.8240e-05,  8.3669e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1917e-04,  9.8926e-01, -6.2491e-04,  6.7492e-04,  5.6695e-04,
        -4.6958e-04,  5.4538e-04,  3.0442e-04, -3.3453e-04, -3.7114e-04,
        -2.6101e-04,  5.5852e-04, -5.1786e-05,  1.7478e-04,  9.0457e-04,
         1.0411e-03,  8.1070e-05,  4.9048e-04, -9.3075e-05, -6.4089e-04,
        -1.2447e-04, -4.2888e-04, -1.4744e-04, -1.0951e-03, -4.7474e-04,
         5.7795e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.1324e-04,  9.9092e-01, -7.4489e-04, -7.3777e-04, -4.0747e-04,
        -7.1512e-04, -8.5018e-04, -4.4608e-04, -3.8162e-04, -4.1241e-04,
         5.3226e-05, -2.3012e-04, -1.8074e-04,  2.3281e-04, -7.8502e-04,
         3.8706e-04,  4.0574e-05, -1.1994e-04, -2.8080e-04,  1.4069e-04,
        -5.8551e-04,  2.4591e-04,  1.1505e-04, -4.4368e-04,  3.3381e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.4657e-04,  9.9153e-01,  1.9841e-04, -9.5886e-05, -1.0416e-05,
         6.3041e-04, -8.1125e-04, -2.4122e-04, -1.7013e-04, -1.6858e-04,
        -3.1230e-04,  7.3127e-04, -3.4126e-04,  1.0461e-04, -7.3997e-05,
         3.5487e-04, -2.7798e-04, -9.2353e-05, -8.0107e-04, -7.6549e-04,
        -4.1418e-04, -1.9498e-04, -4.1760e-04,  3.5308e-05, -6.4952e-04,
         3.2806e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.9668e-03,  9.8943e-01, -6.2651e-05,  1.4223e-04, -2.2180e-04,
         5.3287e-05,  2.2351e-05, -1.3713e-04, -3.1695e-04, -1.2530e-04,
        -7.3700e-04, -3.6200e-04, -3.7383e-04, -9.6433e-05,  5.9281e-04,
         1.9428e-04, -5.9076e-05,  6.9918e-04,  2.2340e-04,  4.2519e-04,
         7.0838e-05,  1.5081e-04,  4.9716e-04,  3.5003e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2415e-03,  9.8777e-01, -3.4481e-04, -1.4711e-03, -1.2766e-03,
        -2.8407e-04,  2.9631e-04,  1.0963e-04, -1.9839e-04,  3.1449e-04,
        -5.2036e-04,  4.4369e-04, -1.7577e-04,  5.1514e-06,  3.6546e-04,
         6.4145e-04, -7.8729e-04,  1.7059e-04,  4.8142e-04,  5.5672e-04,
         7.4714e-04,  5.9562e-04,  2.0362e-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([ 1.8920e-03,  9.8946e-01, -4.3964e-04, -4.2191e-05,  4.8500e-04,
         4.9293e-04,  6.3209e-04,  1.3920e-04,  3.7284e-04, -8.3966e-04,
        -6.3995e-05, -3.6388e-04,  6.5370e-04, -3.9630e-04,  4.4503e-04,
         5.7512e-04, -2.3266e-04,  2.3329e-04, -2.5653e-04, -6.0920e-05,
         8.8883e-04, -1.4814e-04, -1.5748e-04, -3.4108e-04,  3.9155e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.2740e-04,  9.9075e-01, -4.0613e-04, -1.2521e-04,  4.0833e-04,
         6.7786e-05, -8.5497e-04, -7.9566e-04, -5.7176e-04, -4.1313e-04,
         1.1619e-04, -4.0260e-04,  6.4846e-05,  2.8760e-04,  2.1154e-04,
         5.7487e-04, -1.7900e-04,  1.3233e-04,  1.1560e-04, -5.6158e-04,
        -1.1831e-04,  6.2412e-05,  3.7167e-04, -4.5268e-04, -9.0366e-05,
         2.1798e-04, -6.9967e-04,  2.4858e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.0851e-03,  9.8937e-01, -5.0574e-04, -5.6262e-04,  1.1470e-04,
         6.8118e-04, -4.5663e-04,  1.0561e-04,  1.5789e-04, -7.1150e-04,
        -1.6111e-04,  2.6596e-05,  3.4952e-04, -3.9970e-04,  1.0354e-03,
         4.7542e-04,  4.6273e-04,  1.9133e-04, -2.6305e-04, -1.2988e-04,
        -5.9298e-04, -4.5532e-05,  1.5103e-04, -1.1333e-04, -3.4071e-04,
        -1.3291e-04,  1.2275e-04,  2.0883e-04,  4.8497e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.1442e-05, -9.9187e-01, -1.8983e-04, -2.5008e-04,  1.2855e-04,
        -9.8328e-04,  5.5177e-05,  4.7196e-04, -4.4439e-04, -1.2322e-04,
         7.1645e-04, -1.9447e-04,  8.8464e-05, -7.6787e-05, -8.7200e-04,
        -2.9360e-04,  1.4772e-04, -2.0148e-04, -6.2037e-04, -3.1100e-04,
         1.5541e-04, -4.3628e-04,  5.1020e-05,  3.9070e-04,  2.4688e-04,
        -3.8767e-04,  5.9659e-05, -5.8785e-05,  1.5760e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.3638e-04,  9.9169e-01, -3.0599e-04,  3.2272e-04, -4.9067e-04,
        -1.4693e-04, -2.3828e-04,  3.7180e-05, -2.2025e-04, -1.1696e-04,
        -5.4027e-04,  7.0536e-04,  8.9871e-05,  6.0502e-04,  9.7787e-06,
         2.3262e-04, -2.8682e-04,  6.2733e-05,  6.4199e-04,  6.1796e-04,
        -7.9317e-04,  2.8650e-04, -5.6552e-04,  5.8008e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3024e-03,  9.8987e-01, -3.2312e-04,  7.5716e-04,  2.5384e-04,
         1.5133e-04, -7.8175e-04,  4.7547e-04, -1.2436e-05,  1.1510e-04,
         4.4809e-04,  8.7074e-04,  2.4271e-04, -4.8568e-04,  1.5718e-04,
         6.9858e-04,  5.3737e-05, -4.0660e-04, -4.5587e-05,  4.9065e-04,
        -1.9198e-04, -3.3682e-04, -5.5892e-05, -4.7350e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #950: [tensor([-1.1990e-04,  9.8739e-01, -1.1366e-04, -1.2818e-03, -1.1076e-03,
        -8.9516e-04, -1.0326e-03, -7.4228e-04, -3.2241e-04, -4.4355e-05,
         9.3757e-04, -2.9376e-04,  3.8246e-04,  5.7610e-04, -1.5824e-03,
        -6.4937e-04,  1.8307e-03, -5.2408e-04, -1.7024e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.5648e-03,  9.7545e-01, -2.5952e-03, -6.1919e-05, -1.7211e-03,
        -1.5437e-03, -2.5018e-03,  1.8190e-03, -2.3708e-03, -1.8460e-03,
         1.2371e-03, -7.1882e-04, -5.6910e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.8663e-03,  9.8328e-01,  1.3782e-03,  1.8756e-04, -1.0673e-03,
        -8.0339e-04, -6.5995e-04,  8.6461e-04, -1.4349e-03, -7.7875e-04,
         3.4694e-04,  1.2684e-03,  9.3568e-04,  6.2495e-04,  2.1757e-04,
         1.9973e-04, -1.3339e-04, -7.1795e-04, -1.9127e-03,  3.2047e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.0571e-03,  9.8146e-01,  2.1332e-04,  2.5208e-04,  4.3533e-04,
        -9.3711e-04, -6.7288e-04,  8.9464e-04, -1.6769e-04, -1.2422e-03,
         1.6337e-04, -5.2308e-04, -2.7019e-03, -1.9449e-04, -1.6338e-03,
         7.4396e-04, -5.4480e-04, -2.0664e-03, -9.5247e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.7668e-03,  9.8259e-01,  2.6703e-04, -8.1022e-04, -3.7702e-04,
         3.7058e-04, -3.4685e-04,  3.6091e-04, -1.1032e-04,  4.4391e-04,
        -2.2822e-04, -9.3441e-04, -2.0051e-04, -1.0572e-04, -1.5094e-04,
         1.2550e-04,  2.5108e-04,  9.0143e-05,  6.8647e-05,  3.2852e-04,
         1.7926e-05,  1.0255e-03, -1.5054e-03, -9.8114e-05,  5.2983e-04,
         8.9352e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6887e-03,  9.8152e-01, -2.5711e-04,  6.8674e-04, -7.5487e-04,
         1.0546e-03, -3.6186e-04,  3.5930e-04, -1.8467e-03,  9.1030e-04,
         3.2991e-05, -6.0351e-05, -1.1731e-04,  2.5680e-04, -4.1687e-04,
         6.2449e-04, -2.5496e-04,  4.4947e-04,  8.7010e-05,  2.2124e-04,
         1.0453e-03,  9.5460e-04,  9.5847e-04,  5.4212e-04,  2.6643e-05,
         1.7379e-03,  1.0082e-03, -7.6194e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0484e-03,  9.8559e-01, -8.1216e-04,  1.6689e-04, -2.9898e-04,
         4.3657e-04, -1.6021e-05, -3.3580e-04,  3.4746e-04,  1.3763e-03,
         3.5001e-04, -7.0401e-04, -4.7278e-04,  5.5023e-04, -6.1142e-04,
         2.7414e-04,  9.3313e-04, -7.6038e-04,  2.0097e-04, -4.4608e-04,
         1.4717e-05,  8.7939e-04,  9.2749e-04,  5.1601e-05,  3.9338e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3945e-02,  9.6343e-01,  6.6382e-04,  1.9288e-04,  2.0823e-04,
        -1.7348e-03, -2.9900e-04,  2.1943e-04,  2.6108e-03,  6.1735e-04,
        -4.0827e-04, -5.4283e-04, -1.2331e-03,  1.0187e-03, -4.6475e-04,
        -2.1500e-04, -1.0788e-03, -1.8087e-04, -3.5441e-05, -9.4050e-04,
         1.5906e-04, -5.8155e-04, -2.3666e-04, -5.0110e-05,  1.9985e-04,
         6.0056e-04,  2.2269e-04,  5.3619e-04,  4.6462e-04,  5.9100e-04,
         1.2914e-03,  6.9421e-04, -4.3726e-04, -6.1974e-04,  9.0029e-04,
        -1.8280e-04, -6.6726e-06,  3.6533e-04,  6.6922e-04, -1.9241e-04,
         6.5411e-04,  3.0897e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3841e-02,  9.6743e-01,  4.6459e-04, -7.5658e-04, -2.4296e-05,
        -8.8554e-04,  3.9564e-04,  2.9029e-04,  3.1746e-04,  8.2771e-04,
         5.3784e-05,  2.9407e-04,  1.8651e-05,  8.8509e-04, -4.5692e-04,
         2.3656e-04, -5.5008e-04,  5.6144e-04,  4.1973e-04, -4.1848e-04,
         2.4004e-04, -5.3370e-04,  3.4953e-04, -2.9597e-04,  3.6142e-04,
        -9.7892e-05, -2.2122e-04, -1.0319e-03,  3.5331e-04,  1.2441e-03,
         1.6709e-03, -4.5421e-05, -5.4267e-04,  2.0179e-04,  7.2926e-04,
        -3.4699e-04,  6.3948e-04,  5.9220e-04,  2.2315e-04,  2.6199e-04,
         6.7941e-04,  2.1343e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8207e-03,  9.7943e-01,  2.9950e-04, -6.4339e-04, -4.2775e-04,
        -7.1734e-04,  2.7715e-04, -3.1788e-05,  1.0973e-03,  3.8416e-04,
         1.1982e-04, -8.9303e-06,  2.0959e-04,  7.9629e-04, -3.0365e-05,
         6.1963e-04,  5.4480e-04,  4.8496e-04, -4.9128e-05, -3.9498e-04,
         2.2696e-03, -3.5569e-04, -2.0377e-04, -4.7055e-04,  6.6947e-04,
        -2.2082e-04, -2.5673e-04, -1.0106e-04, -8.8073e-05,  1.5057e-05,
         5.3969e-04,  1.2191e-04,  3.8649e-04,  1.0212e-04,  1.1369e-03,
         1.1814e-04,  1.0384e-03,  1.6050e-03, -2.5989e-05, -1.6489e-04,
        -1.3817e-04, -4.9144e-05,  2.2564e-04,  3.0470e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2273e-02,  9.7412e-01, -1.3061e-04, -7.8400e-04, -8.1320e-04,
         2.5110e-04, -1.2243e-03,  3.8452e-05, -5.4730e-04, -1.0950e-03,
        -1.3616e-03, -5.0490e-04,  2.0066e-03, -1.7106e-04, -1.4917e-03,
        -5.9301e-04,  4.1979e-05,  3.8357e-04, -2.3405e-04,  4.7666e-05,
         4.3863e-05,  8.3404e-04, -5.1508e-04, -4.9520e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.1214e-03,  9.7429e-01, -1.4429e-03, -6.4932e-04,  1.3109e-05,
         8.5424e-04, -1.2999e-03, -8.1558e-04, -1.1006e-03, -1.7820e-03,
        -1.7720e-03, -5.9562e-04,  1.1819e-03, -1.1898e-03, -1.2689e-03,
        -9.0551e-05,  8.7299e-04, -5.4064e-04,  3.6851e-06, -2.8248e-04,
         3.4142e-04, -4.7263e-04, -1.1413e-03,  8.7295e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1000: [tensor([ 3.3164e-03,  9.7938e-01, -1.1495e-03,  6.1875e-04,  1.3050e-03,
         6.1621e-04, -7.0431e-04,  7.7101e-04,  2.4713e-04, -6.1260e-04,
         1.5524e-04,  1.0077e-03, -6.4406e-04,  2.6103e-04, -2.6967e-04,
         1.0217e-04,  8.8505e-04,  2.2137e-04, -9.0932e-04,  1.0764e-03,
         1.4520e-03, -7.2556e-04, -3.4526e-04,  1.0184e-03, -1.1866e-03,
         1.0391e-04,  5.1996e-04,  4.0009e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3733e-03,  9.8772e-01, -3.3147e-04,  5.6427e-04, -4.9967e-04,
        -1.1814e-04, -6.7104e-04, -5.9699e-04, -7.1259e-04, -1.6614e-04,
        -3.9998e-04, -3.0062e-04,  1.9601e-04, -8.3593e-04,  8.6248e-04,
        -1.4262e-03,  1.8820e-04, -1.1934e-03, -1.1812e-04,  7.2231e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5453e-03,  9.8615e-01,  3.0825e-04,  1.1517e-03, -1.0129e-04,
         4.2204e-04, -1.0532e-03, -9.0958e-04, -6.5727e-04,  5.4462e-04,
        -3.6763e-04, -2.5328e-04,  3.2357e-04,  3.4490e-05, -1.2826e-04,
        -5.9439e-04, -4.5430e-04, -3.7295e-04, -1.0171e-03,  3.9822e-04,
        -2.0852e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4892e-04,  9.9119e-01,  5.2177e-04, -3.9050e-04, -5.0096e-04,
        -5.9808e-04, -6.9103e-04,  5.5178e-04, -5.0676e-04,  1.2614e-04,
        -7.1491e-04,  4.4454e-04, -6.2543e-04, -8.2683e-05, -2.3973e-04,
        -7.8532e-04, -1.4723e-04, -3.0311e-04,  2.8177e-04, -2.6667e-04,
         4.4319e-04, -1.2807e-04, -3.1531e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.3551e-04,  9.9602e-01,  2.2954e-04,  2.8030e-04,  3.8841e-04,
         9.9269e-05,  7.8191e-04,  8.6531e-05, -2.2633e-04, -3.2004e-04,
        -4.9263e-05, -1.0141e-05, -2.7196e-04, -2.2165e-04, -3.3825e-04,
         1.1402e-04, -4.8951e-05, -7.7544e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+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,  9.9540e-01,  4.1460e-05, -2.3368e-04,  1.4641e-04,
         1.9635e-04,  3.6902e-04, -1.8966e-06,  7.4440e-04, -2.6207e-04,
        -3.5080e-04,  4.0590e-04, -5.4210e-04,  2.7364e-04, -3.2878e-04,
        -1.7770e-04, -9.7680e-05,  7.4822e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0102e-03,  9.9280e-01,  2.4849e-04,  1.0912e-03,  4.7711e-05,
         5.4933e-04,  4.2964e-04, -6.8707e-04,  3.6882e-04,  1.7753e-04,
         1.6918e-04,  4.2303e-05,  6.4182e-05,  2.9947e-04,  1.7571e-04,
         4.8606e-05,  7.8721e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4952e-03,  9.8269e-01, -2.6622e-04,  4.4293e-04,  1.0377e-03,
         5.5248e-04, -9.3936e-05, -2.2982e-04,  1.2233e-04,  6.5435e-04,
         2.7183e-05,  8.9636e-04, -8.8951e-05,  6.6167e-04, -1.5288e-04,
         2.3774e-04,  2.4332e-04, -3.1964e-04,  5.2672e-04,  2.9761e-04,
         4.5617e-04, -2.4756e-04,  6.4909e-04,  4.7416e-04,  5.3894e-04,
         2.1060e-04,  7.3264e-04, -9.1146e-04, -1.9431e-04, -1.3589e-04,
         4.5795e-04,  6.1353e-04,  1.1969e-03,  7.3412e-05,  4.6075e-04,
         2.9606e-04, -3.1310e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.5544e-04,  9.9014e-01,  2.8064e-04, -8.4171e-04,  1.7290e-03,
        -1.4145e-04,  4.2330e-04, -2.4222e-04,  2.4150e-04, -4.7180e-04,
        -5.2956e-05,  2.4903e-04, -1.6874e-04, -5.8564e-05, -9.1874e-05,
        -3.4369e-04,  6.8215e-06,  9.8357e-04,  8.5372e-05,  4.8884e-04,
         6.7051e-05, -6.8825e-04,  6.9016e-05,  7.1335e-04,  3.8141e-04,
        -2.8011e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9713e-03,  9.8875e-01,  1.1836e-04, -1.2147e-03,  1.1343e-03,
        -1.8964e-04,  6.2548e-05,  2.3805e-04,  5.9748e-04,  1.0201e-04,
        -7.4259e-04,  9.9842e-04,  2.6345e-04,  2.4752e-04, -6.0984e-04,
         8.1216e-04,  3.8605e-04,  5.6182e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.2397e-03,  9.8361e-01, -6.8377e-04,  7.4855e-04,  5.4319e-05,
        -9.2433e-04,  1.0810e-03,  3.7224e-05,  3.0483e-04,  5.0591e-04,
        -4.1809e-04,  8.0356e-06, -5.8798e-04,  2.5039e-04, -1.9845e-04,
        -8.0274e-04, -6.4445e-06, -1.1396e-04,  2.9222e-04, -9.2892e-05,
        -5.6349e-04,  1.3643e-03, -5.2814e-04,  7.1239e-04, -2.2580e-04,
         4.7119e-05,  1.0184e-03,  5.8064e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.1479e-03,  9.8067e-01, -5.6221e-04,  1.3623e-03,  9.8744e-05,
        -7.5946e-05, -3.2856e-04, -3.1605e-04, -1.1749e-04,  7.1459e-04,
         8.1543e-04,  7.3665e-04,  3.8111e-04,  1.2587e-03,  4.4969e-04,
        -2.9095e-04, -3.8579e-04, -1.1411e-04,  5.6280e-04,  1.1808e-04,
         1.2985e-04,  7.4125e-04, -5.7312e-04,  1.5784e-04, -1.9237e-04,
        -1.3522e-04,  5.3657e-04,  1.9848e-04, -1.2996e-04,  6.9691e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1050: [tensor([-6.2611e-04,  9.9512e-01,  2.6236e-04,  2.1280e-04, -1.4347e-04,
        -1.6466e-04,  3.6922e-04,  2.7922e-06,  1.4683e-04, -1.4486e-04,
        -3.5948e-05,  2.8878e-04, -8.4358e-05,  2.1746e-04,  3.4493e-04,
         1.7701e-04, -1.6769e-04,  1.6889e-04, -4.6886e-05,  3.9447e-04,
         3.9048e-05, -6.7266e-05, -8.7524e-05,  1.1189e-04,  1.7583e-04,
        -2.0319e-04,  6.5775e-05,  7.0130e-05,  5.7999e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3427e-04,  9.9442e-01, -2.3444e-04, -1.7687e-04,  5.8706e-04,
        -9.2971e-06,  3.7815e-05,  1.9034e-04,  1.5889e-04,  1.0264e-04,
        -1.6922e-04,  3.9547e-04,  3.1964e-04,  1.7166e-04, -4.3956e-05,
         9.6019e-05,  4.3711e-04, -1.3600e-04,  4.0284e-04,  1.1184e-04,
        -2.8215e-04,  1.5079e-04, -1.3534e-04, -1.8881e-04, -1.7687e-04,
         5.3267e-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([ 4.8467e-04,  9.9513e-01, -3.2423e-04,  1.8787e-05, -2.3535e-04,
         2.7154e-04,  8.1201e-05, -1.9437e-04, -5.6746e-07, -2.0583e-04,
         2.7351e-04,  2.6992e-04, -1.5383e-05,  2.8278e-04, -3.5691e-05,
        -2.1658e-04, -2.8451e-04, -2.4106e-05, -1.5289e-04,  4.7542e-05,
         4.6827e-04, -2.0350e-04,  1.1301e-04,  2.4367e-04,  2.0045e-04,
        -2.1712e-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([ 1.5463e-03,  9.9375e-01, -1.3927e-04, -2.3340e-05,  5.0712e-05,
        -2.8057e-04, -3.7349e-04,  1.2430e-04,  1.4761e-04, -3.7537e-05,
        -6.9509e-04,  2.0221e-04, -3.5479e-04,  7.1251e-05,  9.9628e-05,
         4.2694e-05,  3.5219e-05,  3.5180e-04, -7.6197e-05, -3.9467e-04,
        -6.2013e-06, -3.3577e-04,  7.2574e-05, -2.1842e-04, -1.0453e-04,
         4.6691e-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([ 4.1216e-04,  9.9190e-01,  1.0308e-04, -7.2831e-04,  8.1446e-04,
         9.7086e-05,  1.8857e-04,  4.9302e-04, -2.1326e-04, -1.4198e-04,
         7.0494e-04,  3.4215e-04, -3.6946e-04,  1.0259e-05, -2.1348e-04,
        -1.0232e-04,  1.1744e-04, -4.1966e-04, -1.8284e-04,  3.2595e-04,
         1.5965e-04, -3.7210e-04, -1.7706e-04,  1.4907e-04, -7.8479e-04,
        -3.2452e-04, -1.2901e-04, -2.3933e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1773e-03,  9.9119e-01, -3.8084e-04, -4.2233e-05,  3.0808e-04,
        -5.0141e-04, -1.1213e-04, -4.2116e-04, -4.2038e-05, -2.9012e-04,
        -4.9421e-04,  8.2389e-04, -7.6445e-04,  5.3586e-04,  1.3672e-04,
        -1.6014e-04,  2.7913e-04, -3.6652e-04,  2.2658e-04, -3.5465e-04,
        -5.7133e-04,  3.3385e-05,  3.3726e-04,  1.0503e-04, -3.4164e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9529e-03,  9.9030e-01,  2.4444e-04,  1.3220e-04, -3.0109e-04,
        -9.9278e-05, -4.5161e-06, -5.0730e-04, -3.8052e-04, -3.9678e-04,
        -3.6310e-05, -3.9337e-04,  8.6300e-05,  8.4989e-05, -2.2689e-04,
         7.7997e-05,  1.0390e-04, -7.3588e-04, -4.3049e-04,  3.7236e-04,
         1.5101e-03, -5.4728e-04, -3.5272e-04, -4.3426e-04, -3.4549e-05,
        -2.5293e-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([-2.1923e-03,  9.8829e-01,  1.4008e-04,  3.2348e-04, -6.6686e-04,
        -4.7204e-04, -2.1431e-04,  7.2088e-05, -1.8645e-04,  2.7306e-04,
        -1.2189e-04, -1.7537e-04,  4.3252e-04, -7.9444e-04, -2.5499e-04,
        -1.5246e-04, -7.4628e-04,  2.1251e-04, -3.2095e-04,  1.2635e-04,
        -2.0604e-04,  2.2099e-04,  3.8909e-04, -1.2931e-05, -3.5890e-04,
        -4.4226e-04, -4.7741e-04,  3.1263e-04, -2.3199e-04, -8.3441e-05,
         4.2509e-04, -4.9217e-04,  1.7446e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.4176e-04,  9.9094e-01,  2.7223e-04,  4.0377e-04,  4.5292e-04,
         5.4858e-04,  3.5716e-04, -8.1279e-05,  4.5789e-04, -2.7284e-04,
         8.3937e-04,  7.5993e-04, -3.5689e-04, -4.7180e-04, -2.1478e-04,
        -1.2803e-05,  1.3292e-04, -1.6128e-04,  9.0248e-04,  4.1352e-04,
        -2.4546e-04,  1.7102e-04, -3.4528e-04, -5.3740e-06,  2.0475e-04,
         3.3247e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.9008e-04,  9.9278e-01,  2.7038e-04, -1.0662e-04, -1.1825e-05,
        -3.6969e-04,  6.6850e-04,  3.9229e-04,  2.5693e-05, -5.0767e-05,
         9.1134e-05,  3.4122e-04,  4.6308e-04,  1.3905e-04, -2.6877e-04,
         1.8223e-04, -5.4123e-04, -8.6413e-05,  2.1088e-04,  2.6313e-04,
        -5.6407e-05,  6.6491e-04, -5.2140e-06,  1.1774e-04,  1.9869e-04,
        -1.1559e-04, -1.0912e-04,  5.7107e-05,  1.1878e-04,  1.4556e-04,
        -1.0527e-04, -3.1155e-05,  2.0252e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1259e-04,  9.9401e-01, -9.2059e-06, -7.9201e-04,  7.2793e-04,
         1.0134e-03, -8.1631e-04, -7.6182e-04, -2.4874e-04, -6.7955e-05,
         1.6595e-04, -1.3727e-04,  2.6760e-04,  1.2694e-04, -2.4605e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2613e-03,  9.9456e-01, -5.0854e-05, -3.5455e-04, -1.2993e-04,
        -1.1812e-04, -1.1467e-03,  1.6228e-04,  3.0731e-04,  1.7905e-05,
        -5.9837e-04,  3.2523e-04,  2.8785e-04,  3.3951e-04, -3.4380e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1100: [tensor([ 3.7824e-04,  9.9350e-01, -4.0998e-04,  3.4164e-05,  2.4960e-04,
         2.7363e-04,  2.8020e-04, -2.0238e-04, -2.6506e-05,  3.7445e-04,
         5.3839e-04, -2.2490e-04,  2.2803e-04, -2.4828e-05,  3.8298e-04,
         1.0795e-04,  1.0475e-04,  4.9422e-05,  4.0136e-04,  2.6230e-04,
         2.6226e-06, -3.4720e-05, -8.7702e-05,  4.7108e-05,  2.7926e-06,
        -2.6326e-04, -9.7356e-05,  4.9573e-05,  2.0615e-04, -2.0576e-04,
        -4.0370e-05,  1.8723e-04, -5.1010e-05,  5.1364e-05, -2.1217e-04,
         3.3433e-04,  3.8490e-05, -3.6171e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5650e-03,  9.9436e-01,  4.7917e-05, -1.0697e-04,  1.8829e-05,
        -1.4141e-04, -1.0409e-04, -2.1504e-04, -4.7269e-04,  5.3463e-05,
        -3.5125e-05, -1.0474e-05, -1.1120e-04, -8.8470e-05, -4.1074e-05,
        -6.8370e-05, -2.7099e-04, -3.2974e-04,  4.6025e-05, -2.8443e-05,
         1.3622e-05, -1.6752e-04,  1.2659e-04, -2.7274e-06,  4.4975e-04,
         2.2438e-05,  1.5268e-04, -3.4579e-05, -2.9093e-04,  1.1587e-04,
         5.9210e-05, -1.8803e-04,  2.4859e-04, -5.6914e-07,  9.1514e-06,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3302e-03, -9.9045e-01,  1.9799e-04, -2.1946e-04,  4.5785e-04,
        -1.0803e-04, -2.5128e-04,  1.7989e-04, -5.2401e-05, -4.3646e-05,
         3.0753e-05,  1.2318e-04, -2.9226e-05,  7.3750e-05, -1.4737e-05,
         1.5419e-04, -1.5521e-06, -7.9358e-05, -3.8450e-04, -3.7643e-04,
        -3.5365e-04,  4.4446e-04,  1.7112e-04, -7.3368e-05, -1.7218e-04,
         8.2021e-05,  2.1024e-04, -7.1597e-04, -8.4086e-05,  1.3399e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6762e-03, -9.8836e-01,  3.8207e-04, -4.2236e-04,  1.9677e-04,
        -4.3876e-04, -1.9654e-04, -3.2431e-04, -8.6164e-04, -5.6150e-04,
         1.6415e-04, -3.3407e-04, -3.5073e-04,  2.3139e-05,  2.0412e-05,
        -1.1448e-05,  1.2662e-04, -3.3224e-04, -2.2707e-04,  5.9940e-05,
        -1.8035e-04, -2.8542e-04,  4.9034e-05,  1.3395e-04,  2.9644e-04,
         3.6777e-04, -5.1483e-04,  2.2102e-04, -4.5803e-04,  2.3519e-04,
         1.8588e-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([-4.9367e-04,  9.9515e-01, -9.1585e-05,  5.0978e-05, -2.2549e-04,
         2.8197e-04,  3.0976e-04, -4.9748e-06,  3.6359e-04,  2.5284e-04,
         1.8987e-04, -2.4794e-05,  1.7252e-04, -2.9947e-04, -1.8509e-04,
        -1.3017e-04,  1.1389e-04, -1.1305e-04,  4.4067e-04,  2.4434e-04,
         1.3638e-04, -1.8334e-05, -7.2993e-05,  1.3755e-04, -2.8198e-04,
         9.7788e-06,  1.0444e-04, -3.4555e-06,  9.3848e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1461e-03, -9.9197e-01,  1.1527e-05,  1.5157e-04,  4.0839e-04,
         3.2590e-04,  2.6317e-04,  6.5443e-05,  1.6338e-04,  2.1326e-04,
         7.8391e-05,  1.2809e-04,  1.7935e-04,  6.8193e-04,  2.6598e-05,
         4.2989e-04,  4.8576e-04,  3.4555e-04, -1.2219e-04, -2.2518e-04,
         3.9873e-04,  1.6522e-04, -1.7110e-04, -3.6693e-04,  1.6991e-04,
         9.4667e-05, -2.2231e-05,  8.3197e-05, -4.6352e-05,  5.9762e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1309e-03,  9.9473e-01,  3.7699e-04,  2.9311e-04,  5.8534e-05,
        -9.2366e-05,  2.4547e-04, -6.8302e-05, -3.6608e-05, -9.2290e-05,
        -1.3361e-04,  6.6554e-05,  3.1378e-04,  1.3020e-04,  9.0713e-05,
         1.2859e-04, -1.8266e-04, -1.3240e-04,  8.6509e-06,  1.2978e-04,
         1.4197e-05, -4.3886e-04,  9.8196e-05,  7.5668e-05,  1.7650e-06,
         1.2492e-04, -1.7927e-04,  2.1627e-04,  9.8027e-05, -3.1469e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.4617e-04,  9.9590e-01,  2.7558e-06,  6.1203e-06, -1.5622e-04,
        -1.5589e-04,  3.8001e-04,  5.2145e-05,  3.2211e-04, -1.1387e-05,
         5.5221e-05, -5.3430e-06, -1.6035e-04,  1.3335e-04,  2.1842e-04,
        -1.3633e-04,  6.9722e-05, -4.0503e-05,  2.1428e-04,  1.1323e-04,
         2.4063e-04,  1.0423e-04, -1.0263e-04, -1.4470e-04,  6.2102e-05,
         1.0654e-04, -9.3487e-06,  4.4003e-04,  6.4781e-05, -1.5049e-05,
         1.8062e-04, -1.4724e-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([ 1.2884e-03,  9.9396e-01, -1.1549e-04,  4.6152e-04, -2.4844e-04,
         1.4639e-04, -3.3698e-05,  1.7801e-04, -1.9655e-04, -2.6837e-04,
        -2.5212e-04,  9.6335e-07, -1.0672e-04, -3.4701e-04, -1.4195e-04,
        -1.8127e-04,  6.4362e-05, -2.4995e-04,  1.3851e-04, -2.0586e-04,
        -1.2455e-04, -6.9634e-06,  2.1076e-04,  7.6683e-05, -8.9097e-05,
         4.2471e-05,  1.1253e-04,  1.7543e-04,  2.0806e-04, -1.2068e-04,
         2.5183e-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([-7.2980e-04,  9.9348e-01,  5.5478e-05, -4.6421e-04, -2.4547e-04,
        -4.2857e-04,  1.1236e-04, -4.7604e-05,  1.4299e-04,  2.6863e-04,
        -5.6442e-04,  1.6537e-04,  4.7681e-04, -5.7057e-04,  1.3375e-04,
        -2.7877e-04,  6.7998e-04,  1.3414e-04,  1.3438e-04,  2.7762e-04,
        -3.0233e-04, -3.0830e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0981e-04,  9.9422e-01, -2.3155e-04, -9.9056e-05,  8.0103e-04,
        -3.7461e-05, -5.0580e-05,  1.5551e-04, -2.0783e-04, -3.7961e-04,
         5.5034e-04, -2.4848e-04, -1.3175e-04,  6.0744e-04,  5.4316e-04,
        -3.4316e-04,  4.6906e-06,  1.4521e-04,  1.2872e-04,  5.3589e-04,
         3.3965e-05,  4.3467e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5869e-03, -9.9405e-01,  4.3046e-04,  4.4274e-04,  6.6928e-05,
        -1.2842e-04,  4.3498e-04, -1.1409e-05, -3.8750e-04,  2.8243e-04,
         1.5269e-06,  1.1960e-04,  3.9259e-04, -2.8517e-04, -1.4475e-06,
         3.6197e-04, -4.2130e-05, -1.0799e-04, -2.3709e-04,  2.4111e-04,
         1.9036e-04,  1.3425e-04, -6.6404e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1150: [tensor([ 2.2981e-03,  9.9415e-01,  1.2287e-04, -3.7692e-05,  1.0327e-04,
        -2.5730e-04, -1.0099e-04, -6.6869e-05,  2.5504e-05, -2.3895e-05,
        -1.9095e-05, -4.8473e-05, -1.7777e-04, -1.3132e-04,  2.3079e-04,
        -9.8373e-05,  1.0146e-04, -2.1580e-05,  8.2263e-05,  1.5002e-04,
         2.5215e-04, -5.5804e-05,  2.2350e-04, -8.4857e-05,  1.2387e-05,
         9.4070e-06, -2.1316e-04,  1.9672e-04,  1.0353e-04, -2.0053e-04,
        -7.8697e-05, -1.5616e-04, -1.6806e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5930e-03, -9.9424e-01, -1.5180e-04,  9.1934e-05,  1.4667e-04,
        -2.0815e-04, -3.1661e-04,  2.4000e-04, -7.0290e-06, -3.0069e-04,
        -1.5098e-04,  2.3950e-04,  6.7805e-05, -3.2519e-05, -1.4237e-04,
         9.7903e-05, -4.4994e-06,  3.0259e-05,  2.4496e-04, -8.3868e-05,
        -3.7791e-04,  2.1142e-05,  9.1278e-05, -4.4236e-05, -1.9581e-04,
        -3.5112e-05, -1.9436e-04,  3.7166e-06, -2.9740e-05,  3.9285e-05,
        -2.6743e-04, -2.1463e-04, -3.4515e-05, -5.5876e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3452e-04,  9.9201e-01, -1.4873e-04,  1.3073e-04, -4.2355e-04,
        -1.1759e-04, -2.7963e-04, -3.4476e-04,  2.7827e-04,  3.5164e-04,
        -4.6407e-04, -4.2401e-04,  5.6230e-05, -1.8985e-04, -3.9852e-04,
        -8.6979e-05, -1.6876e-04, -4.2033e-04, -4.2672e-05,  3.1406e-04,
         4.2519e-04, -2.8340e-04, -1.6872e-04, -5.5831e-05,  3.9508e-04,
        -3.0395e-05,  4.6681e-05, -4.4427e-04,  5.1570e-05, -2.1874e-04,
         4.0019e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1238e-04, -9.9275e-01,  1.5584e-04,  5.6743e-04,  2.5421e-04,
         6.9908e-05,  8.8053e-05,  3.0144e-04,  2.4588e-04,  1.9378e-04,
        -6.4852e-06,  8.4413e-05, -1.1132e-04,  3.9833e-04,  2.8438e-04,
         5.8483e-04,  2.2023e-05,  4.8539e-04,  9.2461e-05, -2.2524e-04,
        -1.8886e-05, -2.0489e-04,  2.2395e-04, -9.8526e-05,  8.2584e-06,
         5.7316e-04,  3.0542e-04,  5.2934e-04,  4.9954e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0482e-04,  9.9346e-01, -3.2073e-04,  2.1608e-05,  5.0609e-04,
        -2.2421e-04,  2.0819e-05,  1.8584e-04,  5.7739e-04,  1.2949e-04,
        -3.2194e-04, -1.5486e-04,  8.5153e-04, -1.1197e-04,  3.5896e-04,
         2.7378e-04, -2.0904e-04,  1.4008e-05,  3.8579e-04,  2.8933e-05,
         1.4233e-04, -2.5946e-04, -3.4788e-05,  1.6243e-04, -1.5636e-05,
        -1.0627e-04, -1.7074e-04,  5.6355e-04, -8.6707e-05,  1.3199e-05,
         7.9899e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0973e-03,  9.9044e-01,  3.4956e-04, -6.9435e-04, -5.6170e-05,
         6.0863e-04, -4.6184e-04, -2.1253e-05,  7.2077e-04,  1.2716e-04,
         2.6018e-04,  1.5625e-03,  5.9801e-04, -3.1818e-04, -5.2285e-05,
        -2.8310e-04,  1.4699e-03,  2.1233e-04, -3.1661e-04,  3.1389e-04,
         3.2699e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.5109e-04,  9.9245e-01,  3.4619e-04,  1.3912e-04,  2.8832e-04,
        -7.2573e-05, -1.6375e-03,  7.6959e-04,  5.5206e-04, -7.6195e-04,
         3.2331e-05, -2.5270e-04,  1.4677e-04,  9.6313e-06,  8.8554e-06,
         9.0052e-04, -4.4790e-04,  7.2840e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.8376e-04,  9.9281e-01,  6.1690e-04,  4.3786e-04, -1.0820e-04,
         5.0012e-04,  4.3510e-04,  4.9615e-04,  7.4918e-04,  1.7849e-04,
         3.8373e-04,  5.5322e-04, -8.8917e-05, -2.6710e-05, -1.5317e-05,
        -4.0285e-04, -8.9874e-04, -3.2766e-04,  2.9189e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.2888e-04,  9.9158e-01,  4.8877e-04, -7.6751e-05,  7.9645e-04,
        -9.2252e-04,  5.9348e-04, -2.8842e-04,  1.7574e-04,  3.3114e-04,
        -6.8976e-04,  6.4379e-04, -2.5419e-04,  3.3126e-05,  8.3434e-04,
         4.5397e-04,  8.7389e-05, -7.8812e-04,  4.2844e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0827e-03,  9.9018e-01,  4.1286e-05,  5.0533e-04,  1.0066e-03,
         3.3523e-05, -8.8101e-04, -1.0078e-03,  7.0566e-04,  9.7820e-04,
        -6.3691e-04, -1.8782e-04,  3.3072e-04, -7.3926e-04,  3.1216e-04,
         4.6671e-04, -3.5250e-04, -7.4115e-06, -4.5497e-04, -9.2728e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.5048e-04,  9.9565e-01, -5.3193e-04,  7.1757e-05,  1.5508e-04,
        -1.4701e-04, -1.7716e-04, -3.2372e-05, -2.3682e-04,  1.0155e-04,
        -3.3681e-04, -3.8599e-04,  1.8513e-04,  3.2362e-05,  7.8077e-04,
        -2.0481e-04,  2.9040e-06,  2.5953e-04,  2.9709e-05, -3.3159e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8610e-03,  9.9097e-01,  2.8742e-04, -1.0547e-04,  1.4295e-04,
         6.8731e-04,  4.3110e-04,  2.5096e-04,  6.8450e-04,  7.6054e-04,
         1.7396e-04, -1.2684e-04, -4.5059e-04,  4.3077e-04,  1.9869e-05,
        -5.0476e-04, -1.9270e-04, -6.7663e-04,  2.3752e-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], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1200: [tensor([-6.3604e-04,  9.9214e-01,  2.2404e-03, -8.5903e-04,  2.1942e-04,
        -2.1508e-04,  7.7160e-04,  8.3958e-05,  1.7749e-03, -5.8883e-04,
         4.6874e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.2650e-04,  9.9071e-01,  2.2630e-03, -6.8618e-04, -5.2814e-04,
         8.7422e-04,  5.1235e-04,  2.0300e-03,  4.1322e-04,  5.7504e-04,
         6.8469e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3168e-04,  9.9119e-01,  9.0367e-04, -9.6944e-04,  1.2040e-05,
        -2.4834e-04, -4.4568e-04, -6.5063e-04, -8.2619e-04, -2.7649e-04,
        -1.2094e-03, -6.2521e-05, -5.3929e-04,  2.0466e-04,  1.6895e-04,
         4.0003e-04,  8.6108e-05,  1.9120e-05,  1.6760e-04, -7.2390e-05,
        -5.3554e-04, -3.7669e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3619e-04,  9.8833e-01, -2.0502e-04, -1.5142e-03, -2.3122e-04,
         5.6788e-04, -3.1583e-04, -5.6833e-04,  1.1787e-03, -4.9681e-04,
        -1.3383e-04,  7.9338e-04,  5.0139e-05, -1.1496e-04, -1.4548e-03,
         2.9349e-04,  3.2928e-04, -8.3411e-05, -1.1186e-03, -8.2443e-04,
         1.7725e-04, -3.8346e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0585e-03,  9.8913e-01, -1.6234e-04, -1.0556e-03,  1.4982e-04,
         3.1467e-04,  5.4373e-06, -2.4742e-04,  6.8216e-04,  8.5435e-04,
         2.6305e-04,  2.9236e-04,  4.1809e-04,  7.8328e-05, -7.1255e-04,
        -1.6869e-04,  1.5136e-04, -1.3706e-04, -4.0879e-04, -1.7538e-05,
         5.4367e-04,  3.2681e-04, -7.0372e-04, -1.6329e-04, -4.1498e-04,
        -5.8188e-04, -6.0240e-04,  1.2470e-04,  2.3036e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.9235e-04, -9.9227e-01, -2.1524e-04, -2.7481e-04,  4.2024e-04,
        -1.3739e-04,  1.1510e-04,  3.0372e-04, -3.3960e-04, -1.3694e-04,
         2.2263e-04,  2.0620e-04, -5.3378e-05, -3.8006e-04, -5.2388e-04,
        -4.9220e-04,  5.7417e-04, -3.0173e-04, -1.8925e-04, -2.3843e-04,
        -3.2731e-04, -6.5834e-05, -3.1956e-04,  1.4244e-04, -7.3465e-05,
         3.6688e-04,  1.2078e-04, -2.4138e-04, -3.5037e-04, -7.9507e-05,
        -1.7866e-04,  1.4437e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8315e-04,  9.9476e-01, -2.3462e-04,  1.9454e-04, -3.2045e-05,
        -1.8974e-04, -2.5426e-04, -2.9907e-04,  2.6240e-04, -3.1258e-04,
        -1.8268e-04, -1.4031e-04, -1.3531e-04,  1.5951e-04, -2.4825e-04,
         6.9689e-05,  3.6047e-04, -1.1471e-05,  3.9415e-05, -3.7522e-04,
        -4.0605e-04,  1.7942e-04,  1.1443e-04, -7.9981e-05, -3.1302e-04,
        -8.1015e-05,  8.7481e-05, -1.3806e-04, -1.5748e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1510e-04,  9.9626e-01, -1.5099e-04, -1.5291e-04,  7.4934e-05,
         1.9054e-05, -5.9503e-05,  5.0849e-04,  4.2835e-05,  3.5992e-05,
         1.7467e-05, -1.9090e-05, -6.6166e-04,  9.7349e-05,  2.7231e-04,
        -1.6476e-04,  4.2480e-04,  1.1414e-04,  6.2066e-05, -1.2575e-04,
         2.1219e-04,  2.2963e-05,  2.2691e-04, -1.1922e-04,  3.6210e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0172e-03,  9.8746e-01,  7.3491e-05, -6.7370e-05, -4.9368e-04,
         8.4971e-05, -2.3477e-04,  8.4406e-05, -8.4108e-04,  4.6128e-04,
         3.9540e-04,  4.2202e-04,  2.1526e-04,  4.1665e-05,  4.0735e-04,
        -1.5158e-04, -3.0837e-04,  8.0861e-05,  2.6794e-04, -1.0228e-03,
        -3.3972e-04, -6.3479e-04,  1.5840e-04,  1.0637e-04,  3.1568e-04,
        -7.0680e-04, -8.4716e-04, -1.2216e-04,  9.2007e-05,  3.6766e-04,
        -1.7654e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1752e-03,  9.8898e-01, -5.3309e-04,  9.5104e-05,  4.3038e-04,
        -4.3542e-04,  4.8950e-05, -1.6334e-03, -2.6878e-04, -3.6742e-04,
         8.0804e-05, -6.1050e-05,  5.0918e-06, -5.5970e-04,  3.0553e-04,
         7.8900e-04, -1.3322e-04,  3.2514e-04,  7.1057e-04, -4.5353e-04,
         8.2101e-04, -7.0723e-04, -4.3116e-04,  9.7227e-06,  6.4412e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.8440e-04,  9.8970e-01, -3.2227e-04,  4.4468e-04,  9.7658e-05,
         9.2865e-06, -2.9895e-04, -1.9111e-04, -6.0476e-04,  4.1095e-04,
         1.7919e-05,  3.9374e-04, -3.6120e-04,  7.3113e-05, -4.1934e-04,
        -6.8933e-04, -4.9097e-04,  7.9854e-05, -8.2024e-04,  9.0121e-05,
        -2.8144e-05, -3.2800e-04,  2.3126e-05,  9.2765e-05, -2.4412e-05,
        -2.6773e-04, -1.1548e-03,  1.6296e-04, -8.3445e-04, -1.7804e-04,
        -3.9041e-04, -1.6613e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4792e-04, -9.9461e-01, -5.1835e-04,  6.1574e-05, -1.7561e-04,
        -1.3824e-04, -4.7554e-04, -1.5478e-04, -1.3078e-04, -7.7717e-05,
         3.2951e-04, -1.3780e-04, -2.3040e-04, -1.8668e-04, -8.3481e-05,
        -1.8755e-04, -1.4967e-04,  5.1474e-05, -3.2606e-05, -1.0425e-04,
         4.7394e-05, -4.4732e-05, -5.7561e-05,  1.3282e-04, -4.1851e-04,
        -8.0297e-05,  2.1596e-04,  2.2553e-05,  1.7430e-04, -2.3519e-05,
        -2.5874e-04, -5.4377e-05, -3.0400e-04, -1.8091e-04], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1250: [tensor([-6.3212e-04,  9.9438e-01, -2.8444e-05,  5.6870e-06,  4.5450e-04,
         2.9821e-04,  3.9234e-04,  3.5350e-04,  8.7823e-05,  8.7750e-05,
         2.7302e-04, -8.8151e-05,  1.8559e-04, -4.1948e-04, -1.7289e-04,
         4.7565e-04, -7.7820e-06, -4.0448e-04,  2.7498e-04,  1.5432e-04,
        -7.5449e-05, -3.3315e-05,  2.7764e-04, -4.2115e-04, -1.2820e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4679e-04,  9.9546e-01, -1.7162e-04, -1.9324e-04,  2.4314e-04,
         1.7245e-04,  3.1393e-04, -2.9366e-04, -1.9576e-04, -1.4362e-04,
        -1.1434e-04, -2.9359e-06,  1.1955e-04,  4.3946e-04,  6.0260e-05,
        -2.2235e-04, -1.9436e-04,  5.7003e-05,  2.0015e-04,  2.6484e-05,
         2.4015e-04, -1.3503e-04, -1.8715e-04,  9.2759e-05,  1.1028e-04,
        -6.4523e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6648e-04,  9.9691e-01, -4.0843e-04, -1.0512e-04, -3.3625e-05,
        -2.5231e-05,  1.4751e-04, -4.9074e-06, -1.3406e-04, -1.4745e-04,
        -5.2223e-04, -1.3981e-04,  9.7979e-05,  5.7250e-05,  1.0853e-04,
        -3.4827e-05,  8.6938e-05,  8.5283e-05, -9.4364e-05,  1.6200e-05,
        -7.1669e-05,  1.3012e-04,  1.8775e-04, -8.5443e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.0951e-04, -9.9463e-01,  2.4958e-04, -5.8365e-05, -4.0193e-04,
         2.3538e-04,  3.0220e-04, -2.3799e-04, -2.6674e-06, -3.1445e-04,
        -1.8153e-04, -3.5231e-04,  1.3997e-04, -1.8998e-04, -1.3577e-04,
         1.7190e-04,  1.0905e-04, -1.8881e-04,  2.7800e-05, -1.6957e-04,
        -1.5283e-04, -8.8277e-05, -3.5411e-04,  1.0914e-04, -3.3637e-04,
         3.5658e-05, -1.8500e-04,  7.3080e-06, -3.2377e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6447e-05,  9.9588e-01, -1.7475e-04,  2.6584e-04,  3.7182e-04,
         3.4583e-04,  1.6254e-04,  2.2037e-04, -6.3187e-05,  3.7582e-04,
         2.2184e-04,  2.1944e-04,  2.0488e-04,  1.9906e-04, -4.1670e-04,
        -4.8287e-05,  2.2164e-04,  7.1934e-05,  2.8057e-04,  8.7606e-05,
        -1.1404e-04, -3.1315e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.6269e-04,  9.9439e-01, -1.6936e-05, -6.2151e-04, -4.1524e-05,
        -3.3622e-04, -5.1013e-04,  1.2168e-04, -2.0972e-05,  3.1759e-04,
         1.6873e-04, -1.8316e-04, -4.5938e-05, -3.3658e-04, -3.8711e-04,
         9.2990e-05,  3.1526e-04, -2.8036e-04,  1.3453e-04, -1.5453e-04,
        -1.1475e-04, -1.7187e-04, -2.7133e-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([ 1.2321e-04,  9.9598e-01, -1.9733e-04,  4.3702e-04, -1.9566e-04,
        -1.7355e-04,  1.1357e-05, -4.8390e-04,  3.1260e-04, -3.7279e-05,
         1.8352e-04, -3.0546e-05,  8.0314e-05,  2.0612e-06, -2.9614e-04,
         2.6719e-04, -2.0455e-04,  2.9865e-05,  5.5211e-04, -9.4888e-05,
        -3.0532e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0912e-03,  9.8839e-01,  5.6120e-04,  1.4997e-04,  2.2548e-04,
         8.0972e-05,  4.7502e-05,  5.9366e-05, -5.1071e-04, -1.3538e-04,
        -7.5150e-04,  2.1680e-04,  9.5445e-04,  1.2510e-04, -6.4217e-04,
         6.0710e-04,  2.8025e-04, -4.4961e-04,  5.6210e-04,  4.8766e-04,
         1.5882e-04, -2.7781e-04,  3.1452e-04, -4.7364e-04, -5.9004e-04,
        -1.6834e-04,  5.6003e-04,  8.8357e-04,  1.3337e-04, -1.0926e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4147e-03,  9.8934e-01, -1.9273e-05,  9.8114e-04, -3.4702e-04,
         3.6897e-04, -6.0964e-04, -1.9044e-05, -7.4120e-04, -1.8199e-04,
        -6.4549e-04, -4.1107e-04, -1.7086e-04, -3.4389e-04, -6.4206e-04,
        -4.0901e-04,  2.4908e-05,  3.4903e-04,  4.8412e-05, -3.9255e-04,
        -8.9757e-05, -1.1815e-04,  6.9316e-04,  3.3639e-04, -9.9459e-05,
        -1.4675e-05,  1.8908e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.2084e-04,  9.8995e-01,  6.4917e-05, -3.0384e-04, -2.9126e-04,
         7.3766e-04, -5.0553e-04,  3.6244e-04, -6.2474e-04, -2.0083e-04,
        -4.8588e-04, -1.1497e-04, -3.4862e-04,  1.5489e-04, -2.6898e-04,
         1.0343e-04,  4.8181e-04, -1.3422e-03, -1.0599e-04, -2.3882e-04,
        -4.5217e-04,  5.4298e-04, -2.7664e-04,  9.4961e-05, -1.5927e-04,
         5.2071e-04,  3.4901e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6401e-03,  9.8064e-01, -8.0954e-04,  5.5295e-04, -3.8221e-05,
        -5.5504e-05,  9.9553e-04,  1.7874e-03, -6.0066e-04,  4.8429e-04,
         3.9465e-04,  1.7914e-03,  1.0441e-03, -1.4043e-03,  1.5026e-04,
        -1.0418e-03,  6.8175e-04,  1.8468e-04,  8.6374e-04, -3.8502e-04,
        -4.3505e-04, -3.5554e-04,  6.9845e-04,  5.5808e-04, -4.0232e-04,
         8.5639e-06,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0300e-03,  9.7565e-01, -1.6497e-03,  1.0902e-03, -5.1357e-04,
         3.4445e-04,  2.7184e-03,  1.8436e-03, -1.7150e-04,  2.0368e-03,
        -7.4586e-04,  3.8925e-04,  9.5402e-05, -2.2106e-04, -3.5425e-04,
        -9.1374e-05,  5.9819e-04,  1.8212e-03,  2.4799e-03,  2.8387e-04,
         9.1258e-04,  1.0073e-03, -1.4684e-03, -8.8569e-04,  1.9300e-04,
         7.0361e-04,  3.4377e-04, -3.5987e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1300: [tensor([-4.2144e-04, -9.9548e-01,  6.0525e-05,  4.3262e-04, -1.4984e-04,
        -1.6321e-04,  2.5367e-05, -2.9406e-04,  4.1717e-04, -9.6691e-05,
        -2.4657e-04,  4.4827e-04, -2.1617e-05,  9.2397e-06, -2.6815e-04,
        -1.4335e-04, -2.3395e-07,  2.3785e-05,  2.4542e-04,  3.4052e-05,
        -1.9062e-05, -1.7978e-04,  4.7598e-07, -1.8313e-04, -6.4755e-05,
        -1.5732e-05,  1.0423e-04, -3.3195e-04, -1.1815e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.3288e-05, -9.9598e-01, -2.7070e-05,  8.0299e-05, -1.4567e-04,
         3.7892e-04, -1.1557e-04, -9.7301e-05, -2.2402e-04,  4.4085e-04,
        -1.7891e-04, -2.8739e-06,  9.7475e-05,  1.7616e-05,  2.0340e-04,
         2.3221e-04,  8.2426e-05, -1.0231e-04,  4.0905e-05,  1.2264e-04,
         8.8327e-05,  1.9982e-04, -1.1036e-04, -1.0898e-04,  1.2273e-05,
        -6.3489e-05,  2.0828e-04,  9.4948e-05,  4.4799e-05,  1.2541e-06,
         2.1033e-04, -2.7117e-06,  2.0010e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2258e-03, -9.9469e-01,  6.8170e-04,  3.4600e-04, -1.2011e-04,
        -6.0940e-05, -2.5200e-04,  2.2334e-04, -3.1894e-04,  1.9733e-04,
        -2.9974e-05, -1.1771e-04,  7.2316e-05,  5.8901e-05, -9.8481e-07,
        -1.2337e-04, -2.8802e-05, -6.2475e-05,  6.7943e-05,  1.9936e-04,
         1.2315e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3243e-03,  9.9506e-01,  7.3664e-05, -8.9912e-05, -2.3143e-05,
        -3.1119e-04, -3.0484e-04,  1.2705e-04, -3.6281e-04,  1.8970e-04,
        -1.7295e-04, -1.2944e-04,  1.1056e-04, -6.2360e-05, -5.0183e-05,
         2.0310e-04,  3.8132e-04,  1.2374e-04, -9.2917e-05,  2.9982e-04,
         2.1857e-04,  1.7909e-05,  2.6791e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.4981e-04, -9.9535e-01, -6.3927e-05,  1.3724e-04, -2.6661e-05,
         8.0243e-04, -2.3542e-04, -1.4651e-05, -9.4229e-05,  4.3988e-05,
         7.4425e-05, -1.1449e-04,  3.7020e-05,  3.7242e-04,  3.5017e-04,
         2.4803e-04,  9.1576e-05,  3.2321e-04,  3.3813e-04,  4.2934e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2608e-05,  9.9470e-01, -8.3742e-04, -2.4086e-04,  6.1483e-05,
        -4.3898e-04, -3.7666e-04, -9.0125e-05,  1.8836e-04, -3.3140e-04,
        -3.6613e-04, -3.4506e-05, -3.4714e-04, -6.2462e-04, -3.3540e-04,
        -2.7762e-04,  8.3484e-05, -5.0877e-06,  1.0675e-04, -3.7047e-04,
        -1.0194e-04, -6.8772e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4869e-04,  9.9554e-01, -4.7318e-04, -9.6315e-05,  3.4411e-04,
         4.1574e-05,  1.9656e-04,  1.9695e-04,  1.8568e-04, -2.0005e-04,
        -1.6509e-05, -1.0009e-04, -3.4662e-04, -3.3584e-05,  2.0123e-04,
         2.3872e-04,  1.5665e-04,  2.7157e-04,  1.7723e-04,  7.2660e-05,
         8.0469e-05,  2.3532e-05,  1.1395e-04,  2.3285e-04, -1.1690e-04,
         9.5619e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.9365e-04,  9.9374e-01, -4.8122e-04,  7.9619e-04,  9.3905e-05,
        -8.3204e-05, -1.8699e-04, -2.9283e-04,  6.3067e-04,  1.1189e-04,
        -2.0813e-04,  6.6451e-05,  1.0558e-04, -2.3293e-04, -1.6065e-04,
        -1.3974e-04,  2.3141e-04,  4.6446e-04, -1.3148e-04,  3.4739e-04,
        -6.3184e-04, -2.7380e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7096e-04, -9.9589e-01, -2.0179e-04,  1.0301e-04,  1.3606e-04,
         6.0834e-04,  1.4870e-04, -2.0395e-04, -3.6627e-04,  5.6014e-05,
         1.9781e-04, -1.4374e-04,  2.3088e-05, -1.8850e-05,  4.8445e-04,
        -7.1428e-05,  3.0083e-04,  5.9874e-04, -2.7360e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2147e-04,  9.9641e-01,  3.8703e-05, -1.7356e-04, -1.9092e-04,
        -3.9948e-04, -5.4670e-04, -1.9047e-04, -9.5993e-05,  6.6256e-05,
         1.5499e-04,  6.6980e-05,  3.2018e-04, -9.7054e-05,  2.3038e-04,
         5.7797e-04, -1.1853e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6176e-04,  9.9713e-01,  2.0160e-04,  3.8648e-05,  1.3822e-04,
         2.8868e-04, -1.6646e-04, -5.3118e-05,  1.4344e-04,  3.6330e-07,
         3.4656e-04, -2.2626e-04,  2.0130e-04,  4.0155e-04, -4.1405e-04,
        -7.0891e-05, -2.0418e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.2538e-04,  9.8656e-01, -4.9707e-04,  1.3257e-04, -1.1866e-04,
        -6.8318e-04, -5.2411e-05, -5.9968e-04,  4.9967e-04,  1.8530e-04,
         1.3911e-04, -3.9024e-04,  3.5232e-04, -7.3703e-04,  3.2328e-04,
        -8.3351e-04, -3.6339e-04, -4.1709e-04, -1.9093e-05, -3.4522e-04,
        -3.6445e-04, -2.5311e-04,  6.6411e-05, -2.5425e-04,  5.3208e-04,
        -4.9917e-04, -3.3331e-05, -7.9855e-04,  3.9252e-04, -5.6589e-04,
        -5.8123e-05, -2.8321e-04, -4.1081e-04, -3.4241e-04, -3.4119e-06,
        -4.2511e-04,  7.8329e-05,  1.4747e-04, -3.1327e-04], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1350: [tensor([ 1.4468e-03,  9.8394e-01,  1.0461e-03,  1.2379e-03, -5.3182e-04,
        -3.1283e-04,  2.5325e-04,  1.9716e-04,  1.7631e-03,  1.0426e-03,
        -2.3739e-04, -1.7167e-04,  4.4849e-04,  3.0179e-04, -8.0035e-04,
         3.3101e-04,  3.7686e-04,  4.9837e-04,  8.8340e-04, -1.3621e-03,
        -4.8413e-05,  7.2332e-04,  9.7688e-04,  5.2221e-04,  5.4334e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0492e-03,  9.8512e-01, -6.5519e-04, -2.4216e-04, -1.5548e-04,
        -2.5879e-03, -7.0563e-04,  5.7604e-04, -7.9419e-04, -6.5336e-04,
        -8.5863e-05, -6.9342e-04,  4.4592e-05, -2.3373e-04,  7.1372e-04,
         5.2353e-04, -7.9460e-04,  6.2931e-04,  6.1522e-04,  4.2784e-04,
         7.2249e-05, -3.4902e-04,  2.7924e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9511e-04,  9.9602e-01, -2.7946e-04,  4.0677e-04,  8.1777e-05,
        -4.0784e-04,  1.8872e-04,  4.9913e-05, -1.2874e-04,  4.7356e-04,
         1.4840e-04,  1.6401e-04,  1.9154e-04, -3.3598e-04,  5.1545e-04,
         1.8142e-04, -7.8946e-05, -1.5580e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.2161e-04,  9.9626e-01,  1.1647e-06,  6.4416e-05, -3.8592e-05,
        -2.5208e-04, -6.4841e-04,  2.8141e-04,  1.4000e-05,  5.2628e-05,
         1.8738e-04, -3.0501e-04, -5.2043e-05, -1.2861e-04, -2.0482e-04,
         1.2881e-04,  2.8837e-04, -1.6878e-04, -1.9824e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0103e-03, -9.9441e-01,  2.8144e-04, -5.1710e-04,  2.4409e-04,
         2.9563e-04,  2.1919e-05,  6.2521e-04,  4.9120e-04,  5.3765e-05,
         5.3110e-04,  2.9488e-04, -3.4475e-04,  1.3299e-04, -1.7213e-04,
         1.2265e-04,  1.1695e-05,  3.8185e-04, -5.3713e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0845e-04, -9.9524e-01,  8.3826e-05,  1.9009e-04,  1.6134e-04,
         3.9195e-04,  3.1994e-04,  1.3624e-04,  2.9589e-04, -1.2550e-04,
        -2.2607e-04, -1.4631e-04,  7.6032e-05, -1.1349e-04,  6.6433e-05,
         1.5024e-05, -1.8830e-04, -1.0032e-04,  9.2413e-05, -2.2508e-05,
        -5.6117e-05, -9.4877e-05, -4.7098e-04, -1.6042e-04, -1.1886e-04,
         1.2464e-04, -2.0590e-04,  5.7053e-05,  1.0242e-04,  1.0035e-04,
        -2.1572e-04, -6.6623e-05, -1.2170e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.8480e-04, -9.9511e-01, -1.8877e-05,  1.1366e-04,  3.4506e-04,
         3.3507e-04,  1.4992e-04, -1.2473e-04, -2.1373e-04,  3.3995e-05,
        -2.2400e-04, -4.3919e-05, -7.0412e-05, -1.3483e-04,  3.0877e-04,
         1.3792e-04,  5.0473e-04,  1.8657e-04,  7.6332e-05, -3.8621e-05,
         9.9032e-05, -6.8628e-05, -1.4314e-04, -1.3050e-04,  4.5344e-05,
        -2.4764e-04,  3.0729e-04, -3.5437e-04, -1.4773e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6300e-03,  9.9454e-01,  1.1559e-04, -1.4715e-04,  3.0528e-04,
         2.5246e-04, -1.6307e-04,  3.6298e-04, -1.4025e-04,  4.2302e-06,
        -1.3822e-04,  2.0680e-04, -1.7633e-04, -5.6560e-05, -3.6725e-05,
         5.8136e-06,  1.3833e-04,  1.5496e-04, -2.2064e-04, -1.2269e-04,
        -5.9271e-05, -1.9801e-04, -6.5084e-05,  3.1363e-05,  6.9786e-05,
        -7.8075e-05,  2.3104e-05,  1.8068e-04, -1.4934e-04, -3.5035e-05,
        -9.1291e-05,  1.0289e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8297e-03, -9.9467e-01,  7.0844e-05,  1.9329e-05, -2.4703e-04,
        -1.0202e-04, -2.1540e-04, -1.3482e-04,  2.1065e-04,  4.8197e-05,
        -4.6702e-05, -4.0570e-05, -2.1115e-04, -1.0827e-04,  1.3523e-04,
        -1.2758e-05, -2.7766e-04, -2.1291e-04,  4.8832e-06, -1.3394e-04,
        -1.1381e-04,  2.5856e-05,  6.2155e-07, -8.4135e-05,  1.4274e-04,
        -1.8671e-04, -2.0231e-04, -4.0926e-05,  6.4304e-05,  2.0738e-04,
        -1.8557e-04,  1.4881e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3030e-04, -9.9495e-01,  3.4967e-04,  2.9303e-04,  9.0317e-05,
         2.4099e-04,  6.9001e-05, -5.7193e-06,  3.7546e-05,  7.3232e-05,
        -3.5142e-05,  5.3672e-04,  1.8551e-05, -4.4373e-04,  1.0318e-04,
        -7.7836e-05,  1.7663e-04,  5.4792e-04,  1.1831e-04, -5.6064e-05,
         1.2898e-04, -3.6400e-04, -1.2066e-04, -1.6536e-05,  3.1146e-05,
        -2.4186e-04, -1.4455e-04, -2.8556e-04, -3.7291e-05, -1.0012e-04,
         7.9787e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.3460e-04, -9.9536e-01,  8.5954e-05,  1.1714e-04,  5.0036e-05,
        -1.8645e-04,  1.3156e-04,  2.9715e-05, -6.2485e-05, -9.1956e-05,
        -7.4147e-05,  1.7266e-04, -1.9929e-04,  2.3197e-05,  3.8288e-05,
         1.3958e-04,  2.6326e-04,  1.0834e-04,  1.3652e-04, -2.7614e-04,
         1.7324e-04, -1.3139e-05,  9.8521e-05, -1.2186e-04, -2.0940e-04,
         1.3714e-04, -7.0185e-05, -9.9729e-05,  1.4270e-04,  1.4727e-05,
         9.9185e-05, -1.6747e-04,  2.1517e-05, -1.5237e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4747e-04,  9.9490e-01,  3.6377e-04, -4.8001e-04, -2.3372e-04,
        -1.3785e-04,  3.7947e-04, -3.2896e-05,  1.3743e-04,  1.2643e-04,
         2.9213e-04,  3.9862e-04,  5.7114e-04, -3.4664e-04, -2.0607e-04,
         4.4777e-04, -8.0115e-05,  4.8094e-04,  3.4368e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1400: [tensor([-1.1594e-03,  9.9328e-01,  3.2224e-06,  8.7197e-04,  2.9714e-05,
        -4.4946e-04,  2.4428e-04,  6.4599e-04,  4.8750e-04, -2.7782e-04,
         5.7453e-05,  1.8282e-04,  1.9615e-04, -4.5518e-04,  9.0986e-05,
        -2.7784e-05, -1.4626e-04, -2.9538e-04, -4.6029e-05, -6.5815e-04,
        -3.8353e-04, -8.0137e-06,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4103e-03, -9.9534e-01, -1.4011e-04, -4.7846e-04,  1.5488e-05,
        -9.4660e-05, -1.1147e-05,  8.5832e-05,  1.4424e-04, -1.6425e-04,
        -4.9245e-04,  2.1284e-04,  1.2692e-05,  1.8868e-04,  2.0508e-05,
        -5.4226e-05, -4.8038e-04, -4.1437e-04, -2.3936e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9535e-03,  9.8642e-01,  6.7714e-04, -4.0708e-04, -1.7586e-04,
        -1.8730e-03, -6.9607e-04, -3.7688e-04,  1.8817e-05, -6.0885e-04,
         7.6604e-04,  2.8475e-04, -7.7640e-04, -1.6817e-03, -1.6598e-03,
        -6.2187e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5380e-03,  9.8580e-01, -1.4249e-04, -1.3521e-03,  2.8579e-04,
        -2.4859e-03, -1.4536e-03,  1.8030e-04, -1.0850e-04,  2.8662e-04,
        -2.7813e-04,  5.8872e-04, -1.0212e-03,  1.3183e-04, -1.0948e-03,
        -8.8887e-04,  3.6818e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0833e-03,  9.8766e-01, -1.4292e-04, -7.6069e-04, -3.8839e-04,
        -6.0668e-04, -2.6098e-04,  1.0120e-03, -4.8617e-04,  6.1230e-04,
        -1.1097e-03,  7.1007e-04, -1.5951e-04, -5.8390e-04,  7.5114e-04,
        -6.6169e-04, -3.5006e-04,  3.1738e-04, -1.8658e-04, -1.9953e-04,
        -1.0470e-04, -6.9866e-04, -3.1001e-04,  6.7912e-04, -1.6461e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7463e-05,  9.8959e-01,  4.3764e-04, -3.4631e-05, -8.7891e-04,
        -1.6526e-03,  6.7473e-04, -1.1952e-04,  2.4742e-04, -6.3372e-04,
        -2.0694e-03, -7.5688e-04, -1.2821e-03, -6.9504e-04, -5.9230e-04,
         3.0339e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.8677e-03,  9.8885e-01, -1.1427e-04,  1.9948e-03,  1.5163e-04,
        -1.7667e-03, -5.0696e-04,  6.5986e-04, -5.7562e-04,  1.1720e-03,
        -6.9184e-04,  1.6074e-04,  9.1275e-04,  5.7128e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3358e-04,  9.9005e-01, -1.7136e-04, -4.7643e-04, -6.8136e-04,
        -3.7606e-04,  2.6338e-04,  8.8239e-04, -3.8826e-04,  6.1600e-04,
        -2.2687e-04, -7.6265e-04, -1.1856e-03, -6.0121e-04,  1.8491e-04,
        -6.9942e-04, -5.4964e-04,  1.5204e-04,  9.9210e-04, -6.0314e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.5177e-03,  9.8647e-01, -1.9489e-04, -5.9683e-04,  1.4483e-04,
        -7.5542e-04, -1.0785e-03,  9.1230e-04,  8.0010e-04,  2.4429e-04,
        -4.5208e-04, -5.5837e-04, -1.3308e-04,  3.8982e-04,  4.4481e-04,
        -9.3934e-04, -1.3579e-04,  7.6094e-05,  1.9586e-04,  5.0408e-05,
         1.1939e-04,  9.7688e-04, -7.9262e-05,  6.7091e-04,  6.5793e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1290e-04,  9.9228e-01, -7.0677e-05,  1.7748e-04,  2.8779e-04,
        -1.0529e-03, -8.7572e-04,  5.0681e-04, -1.1656e-03,  4.2159e-04,
         6.5041e-04,  1.7597e-04,  1.3099e-03,  1.6992e-04, -2.5439e-05,
         1.3015e-04,  3.8580e-05,  1.4451e-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([-2.8222e-03,  9.8725e-01, -9.1372e-04, -1.1534e-03, -3.0133e-04,
        -2.2843e-03,  1.5189e-04, -5.6679e-04,  9.0747e-04,  2.2522e-04,
        -1.1665e-03, -6.0191e-04, -4.8561e-04, -7.2637e-04, -4.4359e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7602e-04,  9.8713e-01, -5.1939e-04,  1.0742e-03, -7.2658e-04,
        -1.2345e-03, -1.4968e-04,  1.1125e-03,  2.5815e-04, -5.1430e-04,
        -1.5897e-04, -7.1938e-04,  2.3327e-05, -4.0817e-04, -1.7199e-03,
        -1.7788e-03, -7.3666e-04, -3.0777e-05, -5.8034e-04, -7.4871e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1450: [tensor([ 6.6308e-04,  9.8851e-01,  6.7947e-04,  5.8456e-05,  2.0331e-04,
         2.9822e-05,  1.4927e-04,  5.8190e-04, -3.5028e-04,  9.8163e-05,
         2.7896e-04, -6.5432e-04, -1.8985e-04,  1.8433e-04, -6.2903e-04,
         6.5417e-04,  3.5762e-04, -3.7547e-04,  1.6295e-05, -5.7138e-04,
         3.0848e-04,  3.0680e-04,  1.1554e-04, -4.1793e-04,  2.1421e-04,
        -8.7786e-05, -8.5444e-04, -5.1456e-04, -7.8354e-04, -2.3630e-04,
        -2.1302e-04, -1.4752e-04, -1.6998e-04, -3.9413e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.4704e-04,  9.9549e-01,  6.9636e-04,  5.0889e-05, -1.5716e-04,
         1.0985e-05,  8.0588e-05,  4.1683e-04, -2.8721e-04, -3.9229e-04,
         4.3995e-04,  8.2768e-04, -5.3927e-05, -4.8527e-05,  1.8497e-04,
        -1.5366e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.6679e-04,  9.9612e-01,  5.9284e-05,  1.6270e-04,  6.6447e-04,
         2.8430e-04,  5.0752e-04,  1.1434e-04,  6.4229e-04,  5.8578e-04,
        -5.6780e-05,  2.9050e-05, -6.0451e-06, -2.5971e-04, -2.3748e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9939e-04,  9.9532e-01,  2.8878e-04,  4.7120e-04,  6.0983e-04,
         6.0922e-04,  2.9886e-04,  1.6033e-04, -3.7690e-04,  9.5572e-05,
        -8.7242e-05,  1.8885e-04,  7.5587e-04,  3.3313e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.5133e-04, -9.9468e-01, -4.8176e-05, -5.1156e-04,  1.0407e-04,
         3.1787e-04,  5.0884e-05,  2.8680e-04, -2.5818e-04,  8.2484e-05,
        -5.0112e-05, -4.5892e-04,  8.5203e-05, -1.5000e-04, -2.3421e-04,
        -6.3152e-07,  4.4949e-05, -3.9795e-04,  2.6934e-04,  2.3843e-04,
         3.2144e-04,  3.4158e-04,  2.4251e-04, -7.0733e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.0363e-04, -9.9677e-01,  2.0863e-04,  6.6921e-05,  9.1156e-05,
        -1.4112e-04,  2.8806e-04, -4.9057e-05,  4.4376e-04, -7.0864e-05,
         8.6575e-06,  3.1953e-04, -1.6635e-05, -2.0256e-05,  2.3083e-05,
         2.3790e-04,  1.4969e-04, -1.4540e-05, -3.2384e-04,  1.5012e-04,
         1.3487e-05,  1.8029e-04,  1.5845e-05,  3.0237e-05,  1.6507e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1033e-03, -9.9377e-01,  1.3441e-04, -6.6792e-05, -7.6890e-04,
         2.9785e-04, -1.5401e-04,  8.4387e-05,  3.1286e-04, -1.4323e-04,
        -2.7516e-05, -3.0498e-04, -2.1675e-04, -3.0500e-04, -1.3485e-04,
         8.7769e-05,  2.2622e-04,  2.4441e-04,  1.0926e-04,  1.0394e-04,
         2.2719e-04, -4.6776e-04,  2.1078e-04, -1.8166e-04,  1.6205e-04,
        -9.6737e-05,  5.4751e-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([ 2.0835e-04, -9.9579e-01,  6.3717e-05, -5.2367e-04, -4.3796e-04,
         6.5705e-05,  3.5216e-04, -6.5533e-06, -7.4178e-05, -1.3178e-04,
        -9.2499e-05, -4.6317e-05, -8.2042e-05, -6.2941e-04,  7.6992e-05,
         5.4401e-04,  2.0035e-04, -2.2617e-04,  8.8257e-05, -3.6160e-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], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.9616e-03, -9.9354e-01,  1.1752e-04,  5.1398e-04, -7.9793e-05,
        -3.5730e-04,  8.1159e-05,  2.3437e-05, -7.2966e-05, -1.1882e-05,
        -3.2614e-05,  7.4467e-04, -6.3783e-05, -1.6297e-04, -2.9324e-04,
        -2.3842e-04, -4.3757e-04, -4.4040e-05,  2.2510e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.5995e-04, -9.9387e-01, -8.9066e-04, -5.6189e-04,  1.2749e-04,
         4.0735e-04,  3.1979e-04,  6.5307e-04,  7.8400e-05, -2.4234e-05,
        -1.1574e-04, -2.8499e-04, -2.7148e-04,  2.3261e-04,  3.8964e-04,
         4.5740e-04,  3.8734e-04,  4.2337e-04,  1.4217e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.8247e-03, -9.9385e-01,  9.2443e-05, -9.8104e-05, -4.4251e-04,
        -2.1483e-05, -2.8067e-04, -1.0659e-04,  1.3025e-05,  9.9226e-05,
        -8.8838e-05,  4.0766e-05,  1.9185e-04, -8.8422e-05, -1.4278e-04,
        -1.9251e-04,  2.1497e-04,  2.1365e-04, -2.3841e-04, -9.5531e-06,
        -2.4438e-04, -2.6056e-04, -2.4200e-04,  1.3714e-04,  1.7918e-05,
        -6.5759e-05,  3.7679e-04, -9.8278e-05,  4.1448e-05,  1.1432e-04,
        -3.9508e-05,  1.1017e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.2373e-04,  9.9396e-01, -1.2681e-04,  3.7708e-04,  3.3982e-04,
        -3.1175e-04,  6.1320e-04,  3.1130e-04, -3.7976e-04, -3.3506e-04,
         1.6827e-04,  3.1182e-04, -3.0832e-04,  7.0081e-05,  1.7506e-05,
         9.8621e-05,  3.2075e-04,  3.3350e-05,  7.1616e-04,  2.3339e-04,
         1.4530e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1500: [tensor([-1.4056e-03,  9.9496e-01,  1.0888e-04,  4.2797e-05,  2.2826e-04,
        -6.0431e-05, -2.4474e-04,  4.8618e-04, -3.9232e-05, -5.9350e-04,
        -2.1880e-04,  2.4310e-06, -1.0239e-04, -2.1143e-04, -2.8293e-04,
         3.6224e-05, -1.1838e-04, -5.4648e-05, -6.5644e-05, -6.0715e-06,
         7.2241e-05,  1.2138e-04,  1.3007e-04,  1.5253e-04,  1.1246e-04,
         1.4670e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.9139e-04,  9.9641e-01,  6.6377e-05,  3.5083e-04, -2.2778e-04,
         3.4207e-04,  2.2177e-05,  2.5794e-05, -1.3786e-05,  3.1712e-04,
        -3.3524e-04,  2.5007e-04,  1.1965e-04,  1.4276e-04,  3.0430e-04,
        -2.5153e-06, -1.2822e-04,  8.5773e-05, -2.0633e-04,  2.2179e-04,
        -3.5624e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.8158e-05,  9.9601e-01, -3.6090e-04, -2.3576e-05,  8.7576e-06,
         8.6469e-05, -1.5176e-04, -1.2581e-04, -4.5163e-04, -1.7050e-04,
         1.0479e-04, -3.8097e-04,  1.6613e-04, -5.1275e-05,  1.3530e-04,
        -9.8126e-05, -1.4063e-04,  8.8637e-05, -2.6391e-04,  2.2777e-04,
        -2.5597e-05,  2.7667e-04,  2.3797e-04, -3.3883e-05, -2.4460e-04,
        -8.9266e-05,  1.1613e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.8401e-04, -9.9445e-01,  1.7163e-04,  1.0877e-04,  8.7796e-05,
         3.0193e-04,  1.7495e-04, -3.1475e-04,  1.7723e-04,  2.6063e-04,
         5.2184e-04,  5.5259e-04, -5.3273e-05,  1.4999e-04,  1.1705e-05,
        -3.5430e-04,  2.8439e-04,  2.1522e-04,  1.8188e-05,  1.6143e-04,
         7.3952e-05,  1.6086e-04,  2.2307e-04, -1.9489e-04,  2.4512e-04,
         1.9088e-05,  7.1401e-05, -1.7697e-05, -4.1559e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1648e-04,  9.9526e-01,  2.4928e-04, -4.3068e-06,  1.6993e-04,
        -1.7618e-04,  1.0633e-04, -3.4017e-04, -4.2794e-04, -1.4082e-04,
         3.2672e-05,  2.0749e-04, -7.5261e-06, -7.6327e-05, -4.8193e-04,
        -2.9741e-04, -9.5089e-05,  2.4193e-04,  1.1268e-04, -8.2178e-05,
        -3.9291e-04,  2.0307e-04, -4.7450e-05, -3.4844e-04,  1.1110e-04,
        -4.3652e-05, -2.3106e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.0894e-05, -9.9582e-01,  4.2260e-04,  4.2135e-05,  1.9113e-05,
         1.8041e-04,  7.3517e-05,  1.5484e-05,  3.7358e-04, -2.0120e-04,
        -1.3065e-04, -1.5482e-04, -3.8361e-04, -2.7820e-04,  4.0667e-04,
         6.9512e-05,  3.5227e-04, -5.1381e-05,  2.5089e-04,  1.0974e-04,
         3.1845e-05, -2.0721e-04,  3.7534e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.6614e-03, -9.9208e-01, -3.4112e-04,  1.0332e-05, -2.0823e-04,
         4.5277e-04,  3.9864e-04,  2.0769e-04, -8.9299e-05,  1.9543e-05,
        -7.8450e-06, -8.9638e-05,  2.3845e-05, -4.8764e-05, -1.3728e-04,
        -4.6658e-05, -3.3152e-04, -1.8041e-04,  2.3872e-04, -1.2205e-04,
        -1.1937e-04, -2.6140e-04, -5.4991e-04, -2.1607e-04,  6.1056e-05,
        -4.4080e-04, -4.5186e-04, -5.4416e-05, -1.4657e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.5287e-04,  9.9531e-01, -7.2509e-05,  1.8587e-04, -1.2602e-04,
         4.5745e-04,  1.3722e-04, -2.2999e-04,  4.3443e-04,  1.5600e-04,
        -2.2852e-04,  3.2974e-04, -4.9426e-05,  2.1483e-04, -2.4497e-04,
        -7.9634e-04,  3.7270e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.2754e-03, -9.9336e-01, -1.4966e-04, -1.5336e-04, -1.9041e-04,
         2.2863e-04,  6.4499e-05,  1.1087e-04, -7.1802e-04, -2.5934e-04,
        -3.8525e-04,  3.8040e-04, -6.8424e-04, -2.2217e-04,  8.0243e-05,
         3.0135e-04, -7.2340e-05, -2.5040e-04,  1.1593e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.7682e-04, -9.9485e-01, -6.0000e-05,  2.3348e-04,  4.1443e-04,
         2.9627e-04, -7.8196e-04, -2.9012e-04, -6.3224e-07,  9.9347e-04,
         2.1256e-04,  2.8616e-04,  1.8286e-04,  1.0647e-04,  4.9346e-04,
         7.5512e-05, -2.4161e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1337e-03, -9.9576e-01, -1.5589e-04,  1.5739e-04,  1.3214e-04,
        -3.7071e-04, -7.7306e-05,  3.0968e-04,  1.2800e-04,  5.6131e-05,
        -1.0506e-04,  2.0618e-04,  4.8796e-05,  1.6261e-04, -3.0820e-04,
         1.7391e-04, -2.0256e-04,  1.6308e-04,  2.2857e-05, -4.1594e-05,
        -1.2201e-04, -1.6152e-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([ 4.6318e-04,  9.9661e-01, -1.7101e-04, -2.2430e-04, -1.6197e-04,
        -9.6217e-06,  3.2881e-05, -1.5960e-04, -2.2511e-04,  3.9836e-06,
        -7.3200e-05,  3.1145e-04, -1.1480e-04, -2.0668e-04, -1.0344e-04,
         1.8814e-04, -4.2733e-05, -1.0293e-05,  2.5045e-05, -2.7376e-04,
        -7.3188e-05, -2.3643e-05,  2.1691e-05, -1.1037e-04, -2.4682e-05,
         1.5941e-04, -1.6331e-04, -8.0562e-06,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1550: [tensor([ 7.2101e-04,  9.9075e-01, -5.2550e-04, -6.1039e-04, -7.0621e-04,
        -2.3339e-04,  4.5346e-04,  4.1090e-04, -1.0711e-04, -5.5907e-04,
         2.3784e-04, -4.9495e-04,  1.7935e-05,  7.4491e-05,  1.8508e-05,
        -6.4094e-04, -4.6138e-04, -8.8297e-04, -1.2170e-04,  6.2990e-04,
        -3.6305e-04, -3.3487e-04, -1.4119e-05,  4.2194e-04,  1.3686e-04,
        -7.0317e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.8646e-04, -9.9583e-01, -4.8423e-04,  3.6050e-04, -2.0451e-04,
         3.7296e-04, -3.9517e-04,  6.8993e-05, -2.0394e-05,  6.4438e-04,
        -7.1289e-05, -8.9386e-05,  4.0155e-04,  7.3107e-04, -3.5661e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4370e-04,  9.9659e-01,  1.1351e-04,  1.3974e-04, -2.0733e-04,
        -4.6842e-05, -6.4584e-04,  2.1518e-04,  3.6322e-05,  3.3353e-04,
         4.5707e-04,  8.7858e-06,  5.2852e-04, -1.2100e-04,  1.4184e-05,
         7.9225e-05,  2.2226e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.4182e-06, -9.9504e-01, -6.7359e-04,  2.7483e-04, -7.3277e-05,
         4.1131e-04, -3.1512e-04, -5.7670e-04,  5.2062e-04, -1.5361e-04,
        -5.1277e-04,  4.0216e-05, -2.5510e-04, -4.7271e-04, -2.6064e-05,
        -6.3202e-04,  2.4634e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.2413e-03,  9.8978e-01,  6.4031e-04,  1.2290e-03, -6.0292e-04,
        -8.0532e-05,  5.4182e-04,  8.9863e-04, -3.6013e-04,  6.0345e-04,
         2.1489e-04, -4.8469e-05, -2.8497e-04, -2.9826e-04,  3.9595e-05,
        -7.8308e-04, -1.9889e-04, -8.6570e-04, -2.8332e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1332e-03,  9.8632e-01, -1.1909e-03,  4.6497e-04,  2.4138e-04,
         1.7958e-04,  6.7895e-04,  3.4863e-04, -1.1814e-03,  7.7670e-04,
         1.2927e-04,  8.6676e-04, -5.6730e-04,  1.0105e-03, -9.9478e-04,
         6.6098e-04, -1.8451e-04,  8.4780e-05, -8.3173e-04,  1.5486e-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([-1.7975e-04,  9.9050e-01,  5.9272e-04,  1.4710e-03,  1.1492e-03,
        -2.1570e-04,  1.8127e-04,  9.4747e-04, -4.5366e-04, -8.0489e-04,
         8.4496e-04, -5.0871e-05,  5.8098e-04, -2.3581e-04, -5.3871e-04,
        -1.0710e-04, -1.0665e-03,  8.3349e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.0163e-04,  9.9561e-01,  3.3635e-04,  1.9792e-04, -3.8259e-04,
         7.2763e-05,  3.3699e-04, -2.4054e-04, -1.0824e-04, -2.7633e-04,
         2.1248e-04,  1.1965e-04,  2.3544e-04,  3.4478e-04,  8.4266e-05,
        -8.2890e-05,  5.3001e-04, -1.7133e-04,  2.5206e-05,  1.7122e-04,
        -2.5938e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.8871e-04,  9.9640e-01, -2.6696e-05,  1.0436e-04, -3.9726e-04,
        -2.2157e-04,  2.0175e-04, -1.4491e-04,  4.0881e-05,  2.0389e-04,
        -6.3145e-05, -1.8850e-04, -1.3165e-04, -1.4408e-04, -4.0204e-04,
        -1.8885e-04,  4.8245e-05,  1.7206e-04, -2.3245e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.6456e-04,  9.9565e-01,  3.1087e-04, -1.6154e-04, -1.0364e-04,
         2.8589e-04, -9.0018e-05,  4.0047e-05, -1.7869e-04,  2.2660e-04,
         2.9430e-04,  5.2135e-04, -1.5336e-04,  1.8626e-04,  5.0443e-04,
         1.1716e-04, -8.9573e-05, -1.9655e-05, -9.8170e-05,  1.8811e-04,
        -1.1779e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0747e-04, -9.8953e-01,  3.0445e-04, -5.1455e-05, -6.7322e-05,
        -8.4919e-04,  9.4442e-04,  5.0027e-04, -1.7390e-05,  5.5560e-04,
         5.6241e-04, -9.3430e-04,  9.8864e-04, -1.5435e-04, -1.2829e-04,
         1.5606e-04, -7.2552e-04, -3.7501e-04, -5.8208e-04, -4.7863e-04,
         1.4617e-03, -2.0912e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1061e-04,  9.9377e-01, -9.2436e-05,  4.2042e-06,  9.7150e-05,
        -1.8482e-04, -1.3441e-03, -1.8889e-04,  1.7061e-05, -1.6359e-04,
        -1.2378e-03, -9.3644e-06,  1.2597e-05,  7.9380e-05, -2.2420e-04,
        -1.4488e-04,  4.2260e-05, -9.4956e-04,  3.3809e-05,  3.9027e-04,
         4.9267e-04, -2.1307e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1600: [tensor([ 5.5684e-04,  9.9594e-01, -6.2110e-04, -2.0054e-04, -2.2836e-05,
         4.9849e-05, -4.0969e-05,  2.1786e-05, -9.4031e-05, -2.7522e-04,
        -1.1893e-04, -1.8651e-04,  1.3598e-04, -3.0343e-04, -1.2771e-04,
         6.1467e-05,  1.0609e-04,  9.3603e-05, -1.2774e-04, -7.7784e-05,
        -1.3421e-04, -1.6699e-05, -2.0928e-04,  3.1313e-04,  1.6496e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.1107e-04, -9.9547e-01, -2.4693e-04,  9.9039e-05, -2.9039e-04,
        -7.0005e-05,  2.4875e-05,  1.2659e-05,  1.7392e-04,  6.4680e-05,
         1.0247e-04, -4.6434e-04,  1.1887e-04, -1.1015e-04, -1.6661e-04,
        -8.5448e-06,  1.7638e-04,  3.7705e-05, -9.3245e-05,  2.3617e-04,
        -9.5615e-05, -3.8215e-05,  1.5706e-04,  2.2245e-04,  5.1697e-05,
         9.2484e-05, -1.9607e-04, -1.1598e-05, -2.0346e-04, -4.9784e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.5234e-03,  9.8538e-01, -6.0763e-05, -9.0633e-05,  2.8125e-04,
         3.6762e-04,  4.0576e-04, -3.5660e-04,  7.5351e-04, -1.0052e-04,
        -9.0180e-04,  2.6273e-04,  5.4660e-04,  4.3381e-05, -2.4309e-04,
         2.4414e-04,  1.1898e-04, -6.4489e-04, -9.6856e-05, -2.2055e-05,
        -1.7307e-04, -3.5159e-04,  2.3985e-04,  4.1373e-04, -5.3712e-04,
         1.4539e-04,  6.9085e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.4758e-03,  9.8338e-01, -3.0110e-04, -1.0210e-03, -2.1247e-04,
         5.4047e-04,  5.5154e-04, -7.7566e-04,  8.5923e-04,  4.5066e-04,
        -6.6501e-04, -4.3038e-04,  3.2171e-04,  3.5830e-04,  9.5063e-04,
         8.4393e-04,  2.9608e-05, -6.3201e-04,  4.1068e-04, -1.1330e-03,
        -6.2393e-04, -1.0151e-04,  1.7458e-04, -1.4738e-07,  1.6717e-04,
        -2.5079e-04,  2.0504e-04,  6.5260e-04,  4.8028e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3693e-03,  9.8820e-01, -3.1879e-04,  4.5208e-04, -1.9185e-04,
         1.7685e-04,  1.9189e-03,  1.3291e-04,  8.0896e-04,  2.0937e-04,
        -2.2381e-04,  5.3524e-05,  4.8317e-04,  1.9208e-04,  1.0590e-03,
         3.8470e-04, -3.9805e-04, -4.2536e-04, -9.8883e-04, -2.0263e-04,
         6.4278e-05, -2.8668e-04,  9.7666e-05, -2.1456e-04, -8.6932e-04,
         2.6625e-04, -9.5091e-06,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.1907e-04, -9.9420e-01,  1.3253e-05,  1.4054e-04, -3.3077e-04,
        -3.6504e-05, -1.2977e-04, -6.1258e-05, -1.2073e-04, -4.2940e-04,
        -7.6289e-05, -3.3471e-04,  2.7132e-04, -1.2350e-04,  4.0201e-05,
         1.3524e-05,  6.9578e-06, -9.3121e-05, -6.9557e-05,  3.1120e-05,
        -7.3851e-05,  1.3661e-04,  1.0596e-04,  1.9660e-04,  3.2345e-04,
         6.3234e-05,  1.8700e-04,  2.7798e-05, -6.6619e-05,  1.8854e-04,
         9.9229e-06,  5.9095e-05, -5.1840e-06,  2.2838e-04,  1.5637e-04,
         1.0106e-04, -2.4155e-05,  1.3501e-04, -3.3278e-05,  1.9699e-04,
         1.0610e-04,  2.4995e-04,  8.6580e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1419e-03, -9.9443e-01,  9.3567e-05, -3.4359e-04, -1.2527e-04,
         6.4968e-05,  1.7007e-04, -2.5255e-04,  6.7756e-05,  1.9955e-05,
         1.4426e-05, -8.4995e-05, -4.0974e-05, -1.2283e-04, -1.6431e-04,
        -3.3116e-04, -1.1744e-04, -6.4128e-05, -1.5919e-04,  7.5904e-05,
        -5.9348e-05, -1.4750e-04,  9.8958e-05,  3.1788e-04,  2.7476e-04,
        -6.3373e-05,  3.6786e-05,  9.4621e-05,  1.3814e-04,  3.2324e-05,
        -3.4075e-05,  6.4143e-05,  7.0768e-05,  1.9691e-05, -3.8067e-05,
         6.0882e-05, -3.4740e-05, -1.0587e-04,  2.3808e-05, -1.0545e-04,
        -1.1751e-04, -1.9137e-05, -7.6967e-05,  8.0916e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4897e-03, -9.9469e-01,  3.9921e-05, -1.3003e-04, -1.0624e-05,
         1.4777e-05, -3.4459e-04,  3.0925e-05, -1.9261e-04, -9.6652e-05,
         1.1873e-04, -3.1944e-05,  1.1485e-04, -1.6891e-04,  7.0198e-05,
         8.8540e-05, -7.3394e-05, -1.5314e-04, -1.2610e-04, -7.6350e-05,
        -1.4273e-04, -5.0053e-05, -5.5639e-05,  1.1319e-04,  2.7724e-05,
         1.5143e-04, -1.0889e-04, -2.5124e-04, -2.1510e-06, -2.5332e-05,
         6.4788e-05,  1.5870e-04, -7.1583e-05,  9.2006e-05, -7.9844e-05,
        -2.9075e-05,  8.9153e-05,  6.7674e-05,  4.0310e-05,  6.0197e-05,
         2.0697e-04,  8.0987e-06,  4.4950e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.8272e-03,  9.8060e-01,  3.3842e-04,  2.8927e-04, -7.4320e-04,
        -7.1480e-04, -8.3831e-04, -6.4528e-04, -1.3984e-03,  2.5667e-04,
         7.6112e-04,  1.4050e-03, -7.1842e-04, -2.0960e-04,  2.7109e-04,
         3.9819e-04,  1.6174e-03,  4.5461e-05, -3.5008e-04, -1.0762e-04,
         8.6922e-04, -1.2133e-03, -1.0557e-04,  2.8078e-04, -2.9771e-04,
        -9.5938e-04,  3.9727e-04,  1.1358e-03,  4.9435e-05,  5.3508e-04,
        -4.4716e-04,  1.6935e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7909e-03,  9.8752e-01, -5.7547e-05,  1.4199e-04, -4.8123e-04,
        -2.5149e-04, -1.1001e-04,  3.1748e-06, -8.5708e-04, -4.2065e-04,
        -7.3491e-05, -3.3884e-05,  1.3022e-04, -5.3267e-04,  1.6377e-05,
        -2.3848e-04, -1.8880e-04, -4.2448e-04,  1.2314e-04,  2.8439e-05,
         1.1382e-03,  1.8023e-04, -1.6984e-04, -7.0110e-05, -4.2805e-05,
        -7.7148e-04,  6.1757e-05,  7.6902e-04,  4.8589e-04,  1.0904e-04,
         3.8866e-04,  3.5299e-04,  1.0364e-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([ 6.7484e-04,  9.8388e-01,  2.9087e-04, -3.6423e-04,  2.1498e-04,
        -4.9420e-04,  8.2694e-04,  3.0321e-04, -5.5906e-04,  7.4467e-04,
         3.6476e-04,  1.2742e-03,  5.0299e-04, -1.0167e-04, -8.1378e-05,
         1.7996e-04,  1.0133e-03,  2.1466e-04, -1.6253e-03, -7.0281e-04,
         1.4809e-03, -9.2477e-04,  8.6422e-04,  4.4811e-04, -2.7116e-04,
        -5.0679e-04, -2.8257e-04, -3.4175e-04,  4.6959e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.1197e-04,  9.8275e-01,  8.3254e-04,  1.6096e-04, -1.0441e-04,
         7.7093e-06,  4.6101e-04, -7.7642e-04, -5.1199e-04,  5.3217e-04,
         6.9722e-04,  9.8492e-04,  8.3442e-04, -5.7957e-04,  1.5188e-03,
        -3.9571e-04,  7.5871e-04,  1.2979e-03, -7.5487e-04, -2.0155e-04,
         1.4151e-03,  1.3326e-05,  3.1895e-04,  5.9374e-04, -4.2986e-04,
        -3.7360e-05,  1.0056e-03,  2.1138e-04, -9.3626e-05, -7.9561e-04,
        -6.3631e-04, -8.0988e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1650: [tensor([ 1.4806e-04, -9.9679e-01,  9.8598e-06,  2.7765e-04, -3.9741e-05,
        -3.8910e-06,  1.4698e-04,  7.8380e-05, -2.9948e-04, -1.5142e-06,
        -2.3266e-04, -9.7705e-05, -6.5691e-05, -1.1200e-04, -2.7069e-04,
         6.2252e-05, -6.6519e-05, -6.1764e-06, -2.3021e-04, -1.2683e-04,
        -9.2499e-05,  3.4760e-04, -8.9022e-05,  9.2408e-06,  2.2446e-04,
        -8.9815e-05,  7.7962e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.5268e-03, -9.9492e-01,  1.3555e-04, -1.0300e-04, -1.5609e-04,
        -1.6251e-04, -1.3674e-04, -3.1309e-04, -1.5758e-04, -6.7222e-05,
        -3.7727e-05, -5.0354e-05, -2.9597e-04, -2.0273e-05,  3.5710e-04,
         1.7778e-05,  1.0216e-04, -1.8524e-05, -2.8094e-05, -1.6246e-04,
        -6.4540e-05, -1.5658e-04, -1.2215e-04, -2.0683e-04,  2.4868e-04,
        -2.6361e-04, -8.2605e-05,  1.7923e-05,  6.3160e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5514e-04, -9.9616e-01,  7.2596e-05,  1.6745e-04, -9.2384e-05,
         1.1282e-04, -3.2302e-05, -1.3899e-04,  2.5479e-04, -7.1519e-05,
         1.5394e-04,  3.3875e-04,  1.6290e-04,  2.9638e-04, -1.7233e-05,
        -7.4896e-05,  1.6530e-05, -1.5275e-04,  3.5380e-05, -6.0755e-06,
        -1.8711e-04,  3.0082e-05,  1.0744e-04,  4.2422e-05, -1.2306e-04,
        -4.0670e-04, -7.1537e-07, -9.9256e-05,  1.1635e-04, -1.6474e-04,
         3.6497e-05,  7.0079e-05, -1.0581e-06,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.0216e-04, -9.9477e-01,  2.1919e-04,  9.9644e-05,  2.5795e-04,
         2.5521e-04, -3.6811e-04,  9.0182e-05,  2.8589e-04,  3.9605e-04,
         1.9167e-04,  8.4305e-05,  1.6277e-04, -2.9010e-05,  1.4262e-04,
         1.6456e-04,  5.5620e-05, -2.0219e-05,  8.2193e-05, -1.5446e-04,
         1.4573e-04,  8.0909e-05, -2.2884e-05,  2.9584e-05, -1.1234e-04,
         1.1999e-06,  3.5207e-05,  1.2362e-04, -3.8710e-05, -1.5765e-04,
         2.9965e-06,  9.3689e-05, -6.7847e-05,  2.2519e-05,  1.9208e-04,
        -1.4794e-04,  1.9211e-04, -2.0144e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.3124e-05,  9.9615e-01,  7.4975e-05,  1.5828e-04,  2.3355e-04,
         5.6084e-05,  1.4931e-04,  1.5528e-04,  2.1472e-05,  1.5406e-04,
         2.8375e-04,  2.0084e-04,  1.7158e-04,  1.0733e-04, -7.5301e-05,
        -1.0088e-05,  2.0928e-04,  1.7714e-04,  1.6184e-04, -4.8685e-05,
        -9.7672e-05, -2.6084e-04,  1.1269e-04, -8.9434e-05, -4.6680e-05,
        -1.8000e-04,  2.1554e-04,  2.2722e-06,  1.2569e-04,  1.5874e-04,
        -2.3287e-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([-2.6204e-04,  9.9618e-01,  1.7428e-04,  1.3267e-04,  4.1098e-04,
         2.7419e-04, -7.4843e-04,  3.0660e-04, -1.6340e-04, -4.0235e-05,
        -2.9225e-04, -1.7096e-05,  6.6367e-06, -3.5452e-04, -1.6303e-04,
        -1.5480e-04,  4.8136e-05,  2.0340e-04,  6.6153e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5474e-03, -9.9415e-01, -2.4542e-04,  3.3815e-04, -4.1142e-04,
         2.9510e-04, -3.5580e-04, -4.4294e-04,  1.4184e-04,  1.6590e-04,
         3.5926e-05,  2.7365e-04,  1.1266e-04, -3.9186e-05, -1.1531e-04,
         3.5900e-05, -8.9019e-05,  2.0362e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4334e-04,  9.9615e-01, -1.8769e-04, -6.5511e-04, -6.2473e-05,
         1.6827e-05,  1.3674e-06, -4.2293e-04,  1.1236e-05,  4.3415e-05,
         2.0364e-04, -6.6480e-04, -2.2726e-04, -7.7575e-05, -3.3920e-04,
        -2.4796e-04, -7.9258e-05, -2.9352e-04,  1.6841e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.2154e-04, -9.9556e-01, -1.8042e-04,  1.3892e-04,  2.1474e-04,
        -9.2763e-06, -2.5226e-06,  2.5566e-04,  1.1157e-04,  2.7264e-04,
         1.5645e-04,  1.6085e-04, -1.0878e-04,  5.5886e-05, -1.0265e-05,
         4.4142e-05,  1.7736e-04,  1.7407e-04,  4.0729e-05,  1.3027e-04,
         9.3203e-05,  4.7403e-05,  1.4433e-05,  1.0586e-04,  6.0224e-05,
         1.4999e-04, -1.2803e-04,  4.2199e-05, -7.1255e-06,  6.0929e-05,
        -8.1038e-05,  8.7155e-05,  2.5667e-04,  1.2108e-04, -2.3159e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 9.5858e-04, -9.9381e-01,  2.9435e-04,  1.1740e-04, -3.0889e-04,
         2.3194e-04,  1.6402e-04,  1.3093e-04,  2.0063e-04, -5.1261e-05,
        -2.0826e-04,  8.3807e-05, -9.9090e-05, -3.3271e-05,  3.1874e-04,
        -2.6712e-04, -2.6674e-04, -5.9018e-05,  2.5043e-04,  3.7388e-04,
        -1.8915e-05, -2.4112e-04,  3.5517e-04,  2.8024e-05,  2.0724e-04,
        -1.1019e-06, -1.2330e-05,  1.5778e-04, -9.0745e-05,  1.3745e-04,
         1.1814e-04, -4.6221e-05,  7.0507e-05,  1.4460e-04,  1.3839e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.5829e-04, -9.9585e-01, -1.4254e-04, -1.7216e-04,  1.7161e-04,
         1.3015e-04,  1.4668e-04,  5.5045e-05,  1.3383e-04,  1.2953e-05,
        -2.3357e-04,  3.7173e-05,  1.4345e-05, -5.5700e-05, -2.1299e-04,
         1.3832e-05, -2.0925e-04,  1.0202e-04,  2.1831e-04,  1.7299e-05,
        -2.4537e-04, -1.8836e-04,  2.6748e-05,  5.5814e-05,  4.1188e-05,
         1.0476e-04, -1.5827e-04,  7.0654e-05,  2.9047e-05, -1.0498e-04,
        -2.2836e-04, -8.2755e-05, -3.2696e-04, -1.5133e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.7354e-05,  9.9630e-01, -9.0737e-05,  6.8401e-05,  7.0903e-05,
         2.3650e-04, -2.0346e-04,  3.2715e-04,  1.2114e-04, -2.7933e-04,
         4.2481e-04,  2.4736e-05, -2.1166e-04, -1.2561e-05, -1.1389e-04,
        -2.1934e-04,  1.8228e-04, -8.6489e-05, -1.1874e-06, -1.9776e-04,
        -1.4361e-04, -1.2604e-04,  1.3170e-04, -2.2500e-04,  1.2820e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1700: [tensor([ 1.3431e-03, -9.9483e-01, -1.3785e-04, -1.5937e-04, -4.9619e-04,
        -2.0620e-04, -1.5764e-04,  1.9373e-04, -1.1740e-04, -4.1813e-05,
         2.1294e-04,  1.1064e-04, -3.4013e-04, -2.1646e-04, -2.4314e-04,
        -4.0891e-04, -2.9064e-05, -1.1536e-04,  3.7620e-04,  2.6350e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.5109e-05,  9.9549e-01, -5.0745e-04, -4.7160e-05,  2.2725e-04,
        -3.4722e-04, -2.4244e-04,  1.4100e-04, -2.5039e-04, -3.3866e-04,
         2.1949e-04, -3.8293e-04, -2.1288e-04,  5.1220e-04,  5.9826e-04,
         9.1646e-05,  3.4862e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.8753e-04, -9.9592e-01, -5.3259e-04, -2.0534e-04, -1.3333e-04,
        -5.5633e-04,  1.9956e-04, -1.5067e-04,  3.7531e-04, -3.6623e-05,
        -1.7273e-04, -1.3106e-04, -1.7164e-04,  1.1901e-04,  3.6081e-04,
        -8.9307e-05, -2.5683e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1407e-03,  9.9478e-01,  2.6000e-04, -5.0772e-04,  2.6151e-04,
         2.3246e-04,  2.5924e-04,  1.2152e-04,  3.5623e-04,  2.2200e-04,
         4.1668e-04,  5.9028e-05,  5.3078e-05,  1.5907e-04,  1.4615e-04,
         1.9490e-04,  9.5883e-05, -1.7564e-05, -2.3565e-04,  4.5884e-04,
        -2.6558e-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([ 5.9643e-04, -9.9518e-01, -6.1096e-05,  4.0182e-04,  2.0756e-04,
         2.1420e-04, -3.4857e-04,  2.0753e-05, -4.3288e-04,  5.0241e-04,
        -8.3222e-05, -1.0451e-04,  3.0081e-05,  1.4297e-04, -1.3956e-04,
         3.1915e-04,  2.3948e-04, -1.4402e-04,  3.9067e-04,  1.3611e-04,
        -3.0310e-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([ 2.2574e-03,  9.9420e-01,  1.4657e-04,  8.9958e-05, -1.7620e-04,
         1.8386e-04, -9.1973e-05,  1.5999e-04, -7.9196e-05,  9.6701e-05,
        -9.2867e-05, -1.9557e-04,  2.7091e-04, -8.1569e-05, -7.7071e-05,
         1.8078e-04, -1.3332e-04,  2.1937e-04,  2.7791e-04, -1.1251e-04,
         5.4374e-05, -3.0704e-04,  2.7735e-04,  1.9342e-04,  1.8976e-05,
        -2.1540e-05,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.5291e-03,  9.8307e-01, -5.2594e-04, -4.4844e-04,  4.8307e-04,
        -7.4591e-04, -4.1152e-04, -1.9570e-04, -9.7781e-04, -7.5091e-04,
        -1.6316e-03,  7.2484e-04, -7.9189e-04,  1.7516e-03, -1.0532e-03,
        -1.6667e-04, -1.2479e-05, -2.6252e-04,  7.2402e-04, -1.4876e-03,
         7.0112e-05,  6.9672e-04, -1.0839e-03, -4.0026e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.1910e-03,  9.8877e-01, -3.7978e-04,  1.0226e-04,  1.9029e-04,
        -1.1130e-03,  1.1267e-04,  1.9447e-04,  8.3379e-04, -1.7359e-05,
        -5.7882e-04, -1.1559e-04, -6.4593e-05,  9.9666e-04, -2.5719e-05,
        -1.5068e-04, -6.4513e-04, -3.4019e-04,  4.3919e-04,  1.4781e-04,
         8.8181e-04, -2.9759e-04,  1.2631e-04, -7.1155e-04, -7.3084e-04,
        -3.8037e-04, -7.7901e-05, -3.8878e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4076e-03,  9.8792e-01, -7.9070e-04, -6.3501e-04, -1.3105e-03,
        -4.7545e-04, -6.3456e-04,  5.0794e-05, -1.4818e-04, -1.3790e-04,
         9.1580e-04,  3.4941e-04,  2.9537e-05, -1.1513e-04, -3.2185e-07,
        -4.5014e-04, -4.8231e-04, -8.6267e-04, -3.0658e-04, -1.2666e-03,
         4.8630e-04, -1.5225e-04, -1.3439e-04, -7.5142e-04, -5.1423e-05,
        -1.3526e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.4457e-04,  9.9587e-01, -5.2598e-05, -2.5957e-04, -2.5255e-04,
        -4.3956e-04,  1.3032e-04, -8.2808e-06, -2.3618e-04, -2.6682e-04,
         1.4171e-04, -7.9569e-05, -2.1967e-04,  2.8217e-05, -3.1437e-04,
        -3.3892e-04,  1.1259e-04, -3.9562e-04,  9.5667e-05, -2.6674e-05,
         1.8791e-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([-5.1952e-05,  9.9635e-01, -1.0278e-04,  5.7370e-05, -1.2025e-04,
        -9.3859e-05,  3.9380e-04,  1.9861e-04, -3.0371e-05,  1.7186e-04,
         1.0109e-04, -1.7737e-04,  4.1087e-04, -2.1010e-04, -2.9264e-04,
        -1.6115e-04, -4.7423e-04,  7.7190e-06, -5.6052e-05, -1.2646e-04,
         1.8112e-04, -1.4965e-04,  7.7943e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2489e-03, -9.9555e-01,  5.9614e-04,  4.3573e-05,  5.6678e-05,
         2.3406e-05, -2.6352e-04,  1.6448e-04,  4.8571e-05,  9.0167e-05,
         4.1237e-04, -3.1178e-04,  1.7068e-04,  1.1449e-04,  1.0232e-04,
         2.3715e-05,  3.7256e-05, -3.1570e-04, -3.2747e-04,  1.0307e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1750: [tensor([ 1.0762e-03, -9.9228e-01,  1.3127e-04, -4.2797e-04, -6.3234e-04,
        -6.9490e-04, -5.0643e-04, -8.2763e-04, -2.6542e-05,  1.1207e-05,
         2.2803e-04,  4.4702e-05, -2.1759e-04,  5.4903e-05, -5.8819e-04,
        -6.3356e-04, -1.1925e-03,  2.7218e-04, -1.4937e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-8.7466e-04, -9.9165e-01,  2.1850e-04,  6.5429e-04, -3.6366e-04,
         6.2165e-05, -7.8430e-06, -1.1665e-03,  3.5947e-04, -4.0938e-05,
        -4.4750e-04,  2.4652e-04,  6.8833e-05, -3.6597e-04, -5.3349e-04,
        -7.4442e-04, -1.5538e-03,  4.7310e-04,  1.6543e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.6355e-04,  9.9216e-01,  6.2062e-05,  2.0288e-04,  1.2144e-03,
         3.5947e-04, -3.2813e-04,  5.5139e-04, -1.2608e-04, -7.8942e-04,
         5.4478e-04, -6.8139e-04, -1.4678e-04,  5.0110e-04, -7.1623e-04,
        -4.4349e-04, -4.0663e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3101e-04, -9.9510e-01,  1.6081e-04,  3.6054e-04,  4.5648e-05,
         1.6664e-04,  1.7920e-04,  5.5713e-04,  3.9455e-05,  6.9179e-05,
        -1.1610e-04,  4.5594e-05,  3.1915e-04,  3.2685e-04,  3.7251e-04,
         3.5398e-04,  2.6229e-04, -5.0345e-05, -2.1344e-04, -1.9271e-04,
         2.8429e-04,  3.0300e-04,  1.0112e-04,  3.2267e-06,  1.1024e-04,
         7.6688e-05, -6.1569e-06, -5.2397e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.1675e-05, -9.9510e-01,  2.1928e-05, -2.1671e-04, -2.4779e-04,
         1.9006e-05, -2.3966e-04, -4.1859e-04, -8.9205e-05,  5.6206e-05,
        -2.7460e-04,  1.7723e-04, -1.3490e-04, -6.1240e-04,  1.2723e-05,
        -4.3852e-04, -2.4062e-04,  3.2236e-04, -3.9848e-04,  3.6639e-04,
        -1.0777e-04, -1.2438e-04, -2.2961e-04, -7.1781e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.3380e-04,  9.9607e-01, -5.9952e-05,  1.2793e-04,  1.1255e-04,
         1.8294e-04,  6.0421e-05, -5.4699e-05,  2.8705e-04,  4.4753e-04,
        -5.7805e-05, -1.0448e-04,  1.9831e-04,  3.3196e-04, -3.1741e-06,
         3.1077e-05, -1.2238e-04, -3.1712e-04,  2.1959e-04,  8.6183e-05,
         3.2195e-05, -1.3917e-04, -4.1862e-04,  3.9425e-06,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2298e-04,  9.9562e-01,  3.4804e-05,  2.3424e-04,  4.0194e-04,
        -1.4082e-04,  3.4721e-04,  9.6568e-05,  4.6542e-04, -2.8478e-04,
        -8.3785e-05,  5.2171e-05,  4.5401e-04, -1.6669e-04, -1.3482e-04,
         2.7159e-05,  2.2099e-04, -2.0947e-04,  4.3924e-05,  8.6258e-05,
         1.7638e-04, -1.4861e-04, -1.4244e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.0069e-04, -9.9586e-01, -4.2440e-04,  5.4231e-05, -3.8899e-04,
        -3.3723e-04,  4.6626e-05, -3.0155e-06, -5.3283e-04,  2.6562e-04,
        -2.3558e-04, -5.2656e-05, -4.2007e-04,  4.3101e-05, -3.7012e-04,
        -2.5378e-04,  2.4542e-04, -1.6322e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4261e-03, -9.9311e-01, -6.3646e-04, -3.2193e-04, -4.6265e-04,
         5.1621e-04,  7.7280e-04, -1.3137e-05, -4.9785e-04,  6.4437e-04,
        -8.5343e-05,  2.2016e-04,  1.4175e-04, -3.5895e-05,  4.3227e-04,
        -4.6791e-04, -2.1893e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6070e-04, -9.9362e-01,  4.6117e-04, -4.4569e-05, -5.4550e-04,
         3.7500e-04,  3.5094e-04,  2.3297e-04,  5.9197e-04,  3.3927e-04,
         2.2021e-04, -5.5735e-04,  6.1418e-04, -4.6794e-04, -4.9209e-04,
        -4.5870e-04, -1.8069e-04,  2.4395e-04, -4.4953e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3229e-03,  9.9481e-01, -2.2330e-04, -6.4461e-05, -1.0509e-04,
         5.5775e-04, -2.2497e-04, -3.7976e-05, -9.2090e-04,  1.1498e-04,
         2.4174e-04,  3.2320e-04, -3.2483e-04,  2.0311e-04,  8.6658e-05,
        -2.2891e-04,  2.1348e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0849e-04,  9.9558e-01, -1.1313e-04,  1.2623e-04,  7.3917e-06,
        -4.2978e-04, -1.8852e-06, -1.6091e-04, -1.9170e-04,  9.7213e-05,
         4.9721e-04, -6.0718e-04, -1.1479e-04, -1.1677e-04, -1.2841e-04,
        -3.3058e-04,  1.0069e-05, -1.5398e-04,  2.3299e-05, -3.6427e-04,
        -4.7524e-04,  6.5676e-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>)]
Iter #1800: [tensor([ 7.5574e-05, -9.9398e-01,  7.1592e-05,  3.8727e-05,  3.4361e-04,
        -6.9513e-04,  9.1458e-05, -6.7086e-05,  5.1298e-04, -4.9107e-04,
         3.3300e-05,  1.1056e-03, -2.3545e-04,  3.0518e-04,  1.9947e-04,
         5.7027e-05, -5.7288e-04,  4.9151e-04,  5.0482e-04,  1.2812e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.3619e-04,  9.9329e-01, -6.6241e-04,  1.6083e-04,  4.3608e-04,
         4.3139e-04,  4.0021e-04,  1.6269e-04, -3.8678e-04,  4.7422e-04,
        -2.8947e-04, -7.1534e-04, -1.1073e-04,  1.0899e-04,  1.9140e-05,
        -5.9908e-04, -6.3485e-04, -8.8698e-05, -3.8973e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.4463e-03, -9.9223e-01, -5.0185e-04,  3.4479e-04, -7.5992e-04,
         3.8772e-06,  1.8928e-04, -6.3391e-04, -8.0825e-04, -3.8125e-04,
        -1.2272e-04, -4.1230e-05, -1.0057e-03, -9.2666e-04, -6.0692e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.0174e-04, -9.9221e-01, -1.6651e-04,  1.9755e-04,  3.6659e-04,
        -4.3876e-04,  3.7610e-04, -8.2867e-04,  5.1960e-04,  1.2189e-03,
         3.5524e-04,  2.2062e-04, -5.8515e-04,  7.3336e-04,  3.6122e-04,
         6.3512e-05, -2.0200e-04,  3.5616e-04, -2.2846e-04, -1.7198e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 6.5133e-04, -9.9057e-01, -5.3528e-04, -1.1104e-05,  4.7780e-04,
         8.1363e-04,  1.6339e-04,  5.2980e-04,  6.7685e-04,  1.2904e-03,
        -4.9632e-04,  8.8058e-04,  2.2439e-04, -1.3994e-04, -9.6291e-04,
         4.1991e-04, -4.5097e-04, -2.6532e-05, -3.3713e-04, -3.4589e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0743e-03,  9.9074e-01,  6.2565e-04,  6.8819e-04,  2.7494e-04,
         5.2074e-04,  9.7070e-04,  8.7880e-04, -9.1286e-04,  7.4639e-04,
         4.7723e-06, -3.2154e-04,  3.1784e-05,  3.8336e-05,  1.9563e-04,
        -5.5072e-04, -6.5991e-04, -4.8111e-04, -2.8110e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.3180e-03, -9.9347e-01, -2.5389e-04, -5.5956e-05,  7.5010e-04,
         6.1083e-05,  7.0182e-05,  1.2979e-05, -2.6092e-04,  2.4124e-04,
         3.5789e-04, -1.9375e-04,  2.0022e-04,  2.0975e-04,  8.4253e-05,
         4.9948e-04,  3.2809e-04,  3.6542e-05,  3.5360e-04, -1.2711e-05,
         1.4641e-04,  7.9722e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3610e-04,  9.9732e-01,  3.0503e-04, -2.9273e-04, -1.2086e-04,
         3.3714e-05, -4.6267e-04,  1.4407e-04, -3.3261e-05, -2.5170e-04,
         9.2306e-05,  9.9571e-05,  2.3631e-04,  2.3686e-04,  1.2575e-05,
         4.7400e-05, -9.2908e-06, -1.3276e-04, -2.8969e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.7940e-04, -9.9317e-01, -4.3808e-04, -3.5178e-04,  1.7763e-04,
        -2.9903e-04,  2.3274e-04,  3.0077e-04,  4.1696e-04,  8.6594e-04,
        -2.2358e-04, -2.0502e-04,  1.9061e-04, -2.3525e-04,  2.0300e-04,
         7.7492e-05,  3.1820e-04,  6.2933e-04,  3.3240e-04,  1.2070e-04,
         2.6179e-04, -3.1134e-05,  3.9832e-04,  1.3745e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9832e-05, -9.9095e-01, -1.4480e-04,  1.1024e-04, -1.5088e-04,
         1.7068e-04,  2.0721e-04,  5.5653e-05,  1.8640e-04,  3.3320e-05,
         1.8117e-04,  3.4869e-04, -2.9488e-04, -7.0291e-05,  4.5297e-04,
        -5.7201e-04,  3.5760e-04,  3.1324e-04,  3.8531e-04, -1.5425e-04,
        -3.6926e-04, -2.5874e-04, -1.4380e-04,  5.7450e-04,  9.4555e-05,
         1.4853e-04, -6.9631e-05, -5.7707e-05, -1.0104e-04,  6.0378e-05,
        -1.2926e-04,  9.8251e-05,  7.8281e-05, -2.9786e-04,  1.9293e-04,
         5.1868e-04, -5.4947e-05,  1.2958e-04,  2.1155e-05, -1.5800e-04,
         4.1233e-04,  6.0877e-04, -2.2418e-04,  3.4062e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6215e-03, -9.8701e-01,  1.4145e-04, -2.3724e-04,  5.3454e-04,
         1.9956e-04,  1.8556e-04,  4.2341e-04,  3.8294e-04,  3.8081e-04,
        -5.2064e-05,  3.3838e-04,  6.3414e-04, -2.1462e-04,  2.4614e-04,
         9.2748e-04, -2.2825e-04,  3.1695e-04,  1.0434e-04, -8.6349e-05,
         1.9246e-04, -2.5980e-04,  3.6579e-04,  3.6255e-04, -6.4799e-05,
         5.0801e-04,  7.4759e-06, -2.0445e-04,  5.6549e-04,  5.1827e-05,
         1.4436e-04,  2.7101e-04, -1.5603e-04,  2.8530e-04,  2.9757e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.6693e-03, -9.8920e-01, -3.1646e-04, -2.2759e-04,  6.7377e-04,
        -3.7129e-04,  1.7263e-04, -8.0661e-05,  2.2276e-04,  1.0128e-04,
         2.1190e-04, -3.4437e-04, -4.0601e-04, -1.2365e-04,  8.9453e-04,
        -3.1685e-04, -5.3521e-04, -5.8061e-05, -9.9133e-05, -5.3968e-04,
         4.9029e-04,  7.8542e-05, -2.4698e-04,  3.0005e-04, -1.9775e-04,
        -1.5025e-04,  3.3726e-05,  2.5847e-04, -3.2491e-04,  8.8807e-05,
        -3.6806e-04,  6.4554e-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], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #1850: [tensor([ 1.0362e-04, -9.9694e-01, -1.2376e-04,  1.0059e-05, -1.2471e-04,
         1.8123e-06, -1.3706e-04, -5.8413e-05,  9.9512e-05, -3.8420e-05,
        -6.1246e-05,  4.1993e-05,  8.5880e-06, -1.3900e-04, -1.2027e-04,
        -3.2889e-05, -1.4189e-04, -4.6300e-04, -1.0087e-04,  2.7408e-05,
         2.5755e-04, -2.0135e-04,  1.5405e-05,  3.6243e-04,  1.1638e-04,
         2.3863e-05, -2.4999e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2757e-04, -9.9578e-01,  2.1912e-04, -1.1898e-05, -1.0912e-04,
        -8.3169e-05, -2.0370e-04, -5.3998e-04,  1.7306e-04, -1.5111e-04,
         3.5947e-04,  1.3020e-04,  2.0273e-04,  1.9998e-04, -8.2604e-05,
        -2.8785e-05, -5.8759e-06,  1.9388e-04,  2.8103e-04, -1.0715e-04,
        -6.5692e-05, -9.0835e-05, -2.8895e-04,  3.7042e-05,  6.3390e-05,
        -5.9374e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.8651e-04, -9.9497e-01, -5.1765e-05,  3.6976e-04, -4.8221e-04,
        -6.3013e-05, -3.2304e-04, -1.2584e-04,  1.9897e-04,  1.9990e-04,
         2.6109e-04,  3.4045e-04, -2.9530e-04,  4.4110e-04, -2.0752e-05,
        -3.1381e-04, -1.5888e-04, -8.3223e-05,  1.7823e-04, -2.7172e-04,
         2.6791e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9455e-04,  9.9614e-01,  1.6128e-04, -1.8214e-04,  1.1378e-04,
         3.0486e-04, -4.0786e-05,  1.5283e-04, -1.5567e-04,  8.5801e-05,
         5.9244e-05, -2.9907e-04,  9.9897e-06,  1.3525e-04,  7.6566e-05,
         7.8299e-05,  2.2968e-04,  1.7743e-04, -2.9220e-04,  8.3838e-06,
        -5.8090e-05, -2.9793e-04, -1.7046e-05,  4.2758e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6952e-04, -9.9474e-01, -9.2444e-05, -2.9932e-04, -1.5987e-04,
         4.5064e-04,  3.6869e-05,  2.2447e-04, -1.1320e-04,  2.8148e-05,
         2.2718e-04,  4.3748e-04, -1.8544e-04,  2.8503e-04,  2.2122e-04,
        -1.8827e-04,  1.4512e-04,  2.6739e-04,  2.1019e-04,  1.4772e-04,
        -5.2779e-04, -2.6354e-04,  3.8317e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.6476e-04,  9.9730e-01,  5.2208e-07,  1.3302e-04, -1.1101e-04,
        -3.1172e-05, -1.8653e-04, -8.2523e-05,  1.3629e-04, -5.6103e-05,
         1.2109e-04, -2.7691e-04, -5.4321e-05, -2.3504e-05, -2.4930e-05,
        -5.3877e-06, -2.4287e-05, -1.3226e-04,  1.0327e-04, -5.2283e-06,
         2.6201e-05, -2.6153e-04,  9.0226e-05,  9.4192e-05, -3.1037e-05,
         1.1369e-04, -1.9024e-04,  1.1717e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8555e-04, -9.9631e-01, -2.7589e-04, -1.2519e-04,  3.2314e-04,
         2.7552e-04,  6.6681e-05,  5.3580e-05, -2.0549e-04,  1.5260e-04,
         1.5610e-05,  4.6918e-05, -1.7273e-04, -1.4376e-05,  7.7582e-05,
        -7.6299e-05, -6.2303e-05,  8.0403e-05, -1.8767e-05, -1.1508e-05,
        -1.9078e-04, -3.6553e-04,  9.3405e-05,  7.8399e-06, -2.3414e-05,
        -8.4800e-05, -2.6812e-04,  1.4672e-04, -6.9655e-05, -6.1460e-07,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.2137e-04, -9.9612e-01, -8.9489e-05, -8.0895e-05,  4.6621e-05,
         6.7081e-05,  1.4084e-04, -1.5923e-04,  5.2768e-05,  1.1924e-04,
        -1.0205e-05, -5.5428e-05, -7.7892e-06, -3.7825e-05,  2.9001e-05,
        -2.1625e-05,  9.8248e-05, -7.8224e-05,  1.1922e-04,  7.7888e-05,
        -2.8212e-06, -1.1537e-04,  5.4097e-05, -2.3168e-04,  1.6368e-04,
         2.8663e-04, -1.0331e-04, -4.5875e-06,  4.4747e-05, -9.1765e-05,
        -6.3827e-05,  2.7645e-04, -1.6990e-04, -2.1788e-04, -5.0470e-05,
        -9.0097e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6481e-04,  9.9561e-01,  4.1394e-04,  6.4929e-04, -1.2954e-04,
        -1.1083e-04,  3.6291e-05,  9.0760e-05,  6.2464e-05, -3.8951e-04,
         5.5545e-05, -3.8836e-04,  3.1678e-05, -1.1477e-04, -3.5137e-05,
         2.6194e-04,  1.1981e-04,  2.3407e-04,  5.4127e-05,  1.9591e-04,
         6.8637e-05,  3.2711e-04,  2.3841e-05,  9.7989e-05,  3.2387e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.8695e-04, -9.9504e-01,  3.2206e-04,  3.3293e-04,  5.5205e-04,
        -1.1900e-04,  2.6761e-04, -4.3149e-04, -4.1752e-05,  3.8969e-05,
         3.0275e-04,  3.2295e-04, -1.4404e-04,  1.0533e-05,  6.0232e-04,
         6.2999e-05,  5.3996e-05,  1.0097e-04,  2.2585e-04, -4.0258e-06,
         2.3175e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6024e-05,  9.9653e-01, -1.1644e-04,  6.8520e-05, -4.0181e-05,
        -1.2475e-04,  4.5691e-04, -1.3127e-04,  2.4660e-04, -2.3319e-04,
        -6.8456e-06, -2.0256e-04,  4.6558e-05,  2.3660e-05,  1.0756e-04,
         2.0488e-04, -2.7288e-05,  1.6155e-04, -2.7524e-04, -5.0587e-06,
         1.5303e-04,  2.2660e-04, -1.8868e-04,  5.6938e-05, -1.2766e-05,
        -1.4893e-04, -4.9576e-05,  1.0080e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0188e-04, -9.9489e-01, -2.7053e-05,  4.0539e-04,  1.8657e-04,
        -3.6891e-04, -9.3456e-05,  2.7519e-04,  3.7998e-05,  7.2124e-04,
        -1.6940e-04, -2.5951e-04,  3.0996e-04,  1.7374e-04, -9.1690e-05,
        -1.3548e-04, -1.9327e-04, -3.5801e-04, -1.4517e-04, -3.4462e-04,
         7.6857e-05,  1.3234e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1900: [tensor([-1.4955e-04, -9.9543e-01, -1.4728e-04,  3.4323e-04,  2.3396e-04,
         1.0958e-04, -2.2609e-04,  3.9047e-04, -5.7476e-04, -3.3415e-05,
        -3.8050e-04,  5.2810e-05, -1.0822e-04, -6.8010e-04, -1.1898e-05,
         4.1807e-05,  1.3950e-04, -9.4556e-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([-2.8028e-04,  9.9415e-01, -1.5244e-04, -1.0818e-03, -6.2572e-04,
        -2.7695e-04,  9.6328e-04, -6.1767e-04,  2.4596e-04, -1.5767e-04,
        -2.9224e-04, -1.3772e-04,  6.6005e-04,  2.7434e-04, -8.7207e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.9619e-05,  9.9442e-01, -3.3293e-04, -2.7168e-04, -5.1704e-04,
        -6.1792e-04,  1.5519e-04,  9.6319e-04,  3.9075e-04, -2.8053e-04,
         4.1828e-04, -6.8751e-05,  1.8542e-05, -3.2786e-04,  7.9300e-04,
         2.7839e-04,  1.1019e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2060e-03, -9.9573e-01,  1.6938e-04,  1.6350e-05, -4.5435e-04,
         2.0872e-04, -9.1028e-05, -1.9881e-04, -1.2638e-04,  4.7151e-04,
        -2.2941e-04, -1.1566e-04, -5.4558e-06,  1.7651e-04, -4.2639e-04,
         8.2156e-05, -4.4010e-05,  1.4635e-04,  9.8833e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.2368e-05, -9.9746e-01, -1.9038e-04,  8.3894e-05, -2.1173e-05,
         1.9385e-04, -4.2692e-05,  2.4545e-04, -1.7543e-05,  5.4649e-04,
         3.4531e-04, -6.9178e-06,  1.5039e-05, -3.0314e-04,  1.3321e-04,
        -2.1927e-05, -1.4030e-04, -9.5512e-06, -1.6830e-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([ 1.6664e-03, -9.9560e-01, -1.1208e-05,  1.0881e-04,  1.3322e-04,
         2.4920e-04, -2.2072e-04,  5.9434e-05,  4.8705e-05, -1.4679e-04,
         8.6282e-05, -9.2043e-05,  1.6850e-04,  9.0424e-05, -2.5152e-04,
        -9.5490e-05,  1.1543e-04,  3.7173e-04, -1.8089e-04,  7.6450e-07,
        -3.5238e-06,  1.9770e-04, -1.0533e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.2735e-04, -9.9296e-01, -3.1694e-04,  2.3368e-04, -6.1249e-06,
         5.6071e-04,  6.2626e-04,  4.0730e-04, -6.4749e-04, -3.2417e-04,
         2.9053e-04,  4.6113e-06, -9.1135e-05, -1.0370e-03,  1.6044e-04,
         4.2492e-04, -1.5383e-04, -4.9576e-04,  1.3265e-04, -1.4456e-04,
        -4.5866e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2865e-03,  9.9214e-01, -5.3880e-04, -8.1467e-04,  7.4939e-04,
        -1.4096e-03, -5.3758e-04,  2.3141e-04,  2.6679e-04, -3.0488e-04,
        -1.0653e-04, -9.6700e-05, -2.0995e-04,  3.4316e-04,  6.9097e-05,
        -2.0520e-04, -6.8901e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.3797e-04,  9.9306e-01,  5.2012e-04, -3.0236e-04,  6.7195e-04,
        -8.7414e-04,  8.8715e-05, -3.6503e-04, -3.5901e-04, -4.5860e-04,
         8.5881e-05, -3.4984e-05, -6.2991e-04, -2.8455e-04, -4.4300e-04,
         7.6168e-05,  3.0125e-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],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4111e-04, -9.9648e-01, -1.4192e-04,  6.1248e-05,  1.2250e-04,
         2.4534e-05, -4.0714e-04, -9.1456e-07, -2.8358e-04, -1.6649e-06,
         2.3194e-04, -2.9793e-04, -9.9904e-05, -2.5814e-05, -1.2809e-04,
        -2.7155e-04,  1.0279e-04, -1.5884e-04,  8.7616e-05, -1.5627e-04,
         1.8692e-04, -1.6149e-04,  2.4067e-04,  1.7770e-04,  7.1184e-06],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.6514e-04, -9.9473e-01,  1.4276e-04, -1.6671e-05,  1.8520e-04,
         4.4583e-04, -4.0714e-05,  6.9714e-06,  2.0250e-04,  3.2474e-04,
         1.5367e-04,  3.5740e-04,  1.2857e-04,  4.3278e-04, -1.0680e-04,
         4.1109e-04, -9.9508e-05, -2.6013e-04,  3.8007e-04, -5.3992e-05,
         4.0813e-04,  1.5203e-04,  9.1514e-05,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.9398e-05,  9.9690e-01,  1.4658e-04, -1.6279e-04, -2.0182e-04,
        -3.9912e-05,  2.5775e-05, -3.0236e-04, -2.6250e-05, -2.4388e-04,
        -1.0009e-04, -6.0630e-05, -1.7183e-04, -2.1082e-04, -1.2675e-04,
         1.6887e-04, -4.5864e-04,  1.8555e-05,  1.2250e-04,  8.8151e-05,
         1.3647e-04,  1.6980e-04,  4.5278e-06,  3.8136e-05,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #1950: [tensor([ 5.6784e-04,  9.9373e-01,  2.8334e-04,  1.5817e-04,  5.6929e-04,
        -1.9747e-04, -3.3585e-04,  5.7359e-05,  4.8181e-04,  1.9293e-04,
         3.9906e-04,  2.9208e-04, -1.3626e-04, -4.1458e-04, -4.5818e-04,
        -1.5709e-04, -4.5085e-04,  5.9968e-05, -1.1235e-04, -4.2207e-05,
         2.6487e-04, -1.9944e-04, -2.3223e-05, -2.0689e-05, -2.2694e-04,
         9.7228e-05, -6.6068e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1380e-03, -9.9188e-01, -7.7323e-05,  6.7336e-04, -1.9344e-04,
        -5.5329e-04,  5.8547e-04,  4.6840e-05, -2.8414e-05,  2.2302e-04,
        -6.0264e-04,  5.7301e-04, -2.1243e-04,  2.1878e-04,  3.3241e-05,
         2.8486e-04,  1.4262e-05,  2.3464e-04,  4.7046e-04,  5.6742e-04,
        -2.7981e-04,  1.3225e-04, -6.8500e-04, -2.9591e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.0063e-04, -9.9460e-01, -5.2012e-05,  1.4004e-04, -1.5253e-04,
         6.8204e-05,  3.9029e-04,  5.6566e-04, -1.3480e-04,  1.9405e-04,
        -6.6533e-04,  1.5566e-04, -1.6690e-04, -7.6170e-05,  1.4631e-04,
        -4.6225e-04, -2.2906e-04, -1.6017e-05,  7.8451e-06,  2.5149e-04,
         2.3863e-04, -1.9676e-04,  1.8657e-04, -1.9858e-04,  7.8637e-06,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.4628e-04, -9.9319e-01,  3.6355e-05,  9.0053e-04, -4.6563e-04,
        -4.2743e-05,  5.0815e-05, -2.5883e-04,  7.8580e-04, -5.8165e-04,
        -4.8355e-05,  4.8265e-04,  1.3028e-04, -5.1622e-04,  6.5122e-04,
         1.7676e-04,  2.4511e-05,  5.6921e-04, -3.3717e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.7135e-03, -9.9564e-01,  1.0185e-04,  1.8279e-04, -8.8063e-05,
         1.3280e-05,  2.3188e-04, -9.4488e-05, -4.2498e-04,  9.2171e-05,
        -6.2082e-05,  6.6714e-06, -2.2473e-04, -2.4400e-04, -1.2029e-04,
        -6.0162e-05,  4.3116e-05, -1.1330e-04, -3.1914e-04, -7.3048e-05,
         4.9503e-05, -1.0134e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.7953e-04,  9.9612e-01, -5.6134e-05, -1.4375e-04, -4.2063e-04,
        -1.5178e-04,  1.9577e-05,  2.0920e-04,  3.4291e-04,  1.1125e-04,
        -8.8801e-05,  1.4696e-04, -1.4921e-04, -1.1127e-04,  2.7125e-04,
        -1.5516e-04, -2.9957e-05, -8.1952e-05, -2.0204e-04, -1.1385e-04,
         9.4099e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0678e-04, -9.9597e-01,  7.5359e-04, -1.7441e-04, -3.9455e-05,
         1.8253e-04,  8.2594e-04, -1.7435e-04,  3.4174e-04, -2.2028e-04,
        -1.7501e-04, -2.1218e-04,  2.1619e-05,  4.7052e-05, -1.2944e-04,
         1.3684e-06, -1.6092e-04, -1.2398e-04,  2.6486e-04,  7.6056e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.9676e-04,  9.9386e-01, -6.0764e-05, -1.1491e-04, -3.0592e-04,
        -1.6855e-04, -2.2092e-04,  4.3585e-04, -2.6575e-04, -4.5071e-05,
        -3.0208e-04, -1.5627e-04,  2.1626e-05,  1.0600e-04,  2.4288e-04,
         3.0514e-05, -3.1341e-04,  2.4287e-04, -3.5079e-04,  1.7364e-04,
         9.5733e-05,  2.3929e-04, -1.4254e-04,  3.0339e-05,  3.1968e-05,
         4.7916e-05, -3.5117e-04,  2.0940e-04,  1.9256e-04,  4.0408e-04,
         3.9356e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.2942e-03, -9.8993e-01, -2.5442e-04, -5.1369e-06,  7.9657e-05,
        -3.7590e-06, -3.5747e-04, -4.1994e-05,  1.8726e-04,  5.5104e-04,
        -5.9665e-05, -2.2852e-04,  2.3002e-05,  7.8114e-05,  2.1838e-04,
         3.6367e-04, -3.6013e-04, -5.0327e-04, -2.2187e-04,  3.1100e-04,
         2.2835e-04, -2.0684e-04, -3.2117e-04, -5.5372e-05, -7.0768e-06,
         2.5923e-05,  1.3109e-04, -1.6493e-05, -1.3910e-04, -6.9659e-05,
         1.7263e-04, -6.7458e-06,  2.0148e-04,  1.8522e-04,  5.4185e-05,
        -1.0387e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.1586e-04, -9.9339e-01, -1.8704e-05,  2.3087e-04,  3.8977e-04,
         3.1217e-04,  2.7876e-04,  1.3836e-04,  1.3700e-04, -1.3074e-04,
         3.2114e-04, -2.8540e-04, -2.4251e-04, -1.5407e-05,  1.2304e-04,
         1.4587e-04, -4.8961e-05,  6.3300e-05, -3.1700e-06, -1.0055e-04,
         3.8267e-04,  1.2669e-04, -2.2512e-04,  8.3352e-05,  3.1388e-04,
        -7.0238e-05,  2.9057e-04, -2.5663e-04, -1.5284e-05,  3.8719e-05,
         4.5337e-05, -1.2669e-04, -9.1955e-05, -4.1163e-05, -1.8168e-04,
        -1.5828e-04, -3.0553e-04,  9.4284e-05, -1.6459e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.8102e-04, -9.9588e-01, -1.5799e-04, -1.7712e-04,  1.0763e-04,
         1.3523e-05, -1.6422e-04, -1.1336e-04, -5.4160e-05, -3.4261e-05,
        -1.0913e-04,  4.7786e-05, -1.3606e-04, -4.5176e-05,  5.3586e-05,
         4.8423e-05,  1.6266e-04, -4.6608e-04,  2.1231e-06,  4.6395e-04,
         3.0825e-04, -5.0009e-05, -1.7534e-04, -4.9671e-05,  6.8636e-05,
        -8.1166e-05, -6.1970e-06, -1.4373e-04,  1.4472e-04,  1.9697e-04,
         5.3575e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.0067e-04, -9.9598e-01, -2.3193e-04, -2.1285e-04, -2.5965e-04,
        -2.5168e-04, -5.8262e-05,  2.1635e-04, -2.8877e-04, -1.6204e-04,
        -1.4534e-04, -7.5710e-05, -6.3604e-05, -4.5367e-05, -1.3159e-05,
        -3.3446e-05,  6.8506e-05, -2.0331e-04, -1.2801e-04,  4.4782e-05,
         8.0917e-05,  3.3947e-06,  9.8252e-05,  7.0848e-05,  1.5136e-04,
        -2.4760e-05, -6.9556e-05, -1.5477e-04, -1.8688e-04, -2.3561e-04,
        -3.9541e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #2000: [tensor([ 3.0303e-03, -9.9072e-01, -3.1918e-06,  7.2932e-05,  9.4051e-04,
         4.3352e-04, -2.9775e-05, -2.6731e-05,  5.3888e-04, -5.2574e-04,
         5.6046e-04, -1.4715e-04,  3.1312e-04, -5.1239e-04,  9.4741e-04,
        -1.5261e-04, -7.0013e-04,  3.4069e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6822e-04,  9.9330e-01, -2.4785e-04, -5.7213e-04, -2.9287e-04,
         1.5933e-04, -5.5141e-04,  1.6196e-04, -4.5042e-04,  2.8386e-04,
        -3.4098e-04,  1.8945e-04,  3.1515e-04, -1.1964e-04,  5.9021e-04,
        -3.9203e-04, -1.0414e-03,  1.1281e-04, -5.0731e-04,  1.0718e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.5949e-04,  9.9416e-01,  5.1692e-04,  4.4546e-04, -1.3024e-04,
        -4.5415e-04, -1.1883e-04,  4.5917e-04,  2.2514e-04,  1.1587e-04,
         8.0409e-05,  2.1998e-04,  2.1169e-04, -8.4521e-04,  1.9430e-04,
        -3.0694e-04, -8.6309e-04,  2.8831e-04,  1.2874e-06,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.2249e-04,  9.9678e-01,  7.5273e-04,  2.5363e-04,  2.2992e-04,
        -6.5039e-05,  3.8839e-04,  3.6587e-04, -4.1331e-04,  4.6093e-05,
        -2.5693e-04, -1.0513e-04, -1.4423e-04, -7.3622e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2282e-04, -9.9608e-01,  3.9472e-04,  4.3985e-04,  2.5490e-04,
        -4.4674e-04, -3.4484e-04, -2.3537e-04,  1.7864e-04,  4.7774e-04,
         2.7795e-04,  3.0303e-04,  7.3364e-05,  7.5481e-05,  5.1495e-05,
         4.7962e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2794e-04, -9.9638e-01, -3.8195e-05, -4.3135e-04, -4.3596e-04,
         1.7952e-04, -1.2609e-05, -1.7720e-04,  1.1276e-04, -7.2086e-05,
        -4.5267e-04, -1.5703e-05,  1.2311e-04, -2.0085e-04, -1.4228e-04,
         2.6340e-05, -8.8197e-05,  2.4572e-04, -2.9120e-04,  2.4841e-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([ 1.1747e-03, -9.9528e-01, -2.4583e-04, -6.5998e-05,  1.6008e-06,
        -8.8771e-05,  3.1132e-04, -1.7592e-04, -4.8327e-05,  1.0806e-04,
        -1.1181e-04, -5.5032e-06, -1.8137e-04, -3.8971e-04,  1.0619e-04,
        -2.4800e-05,  4.1395e-04, -1.5508e-04,  2.5668e-04,  6.1963e-04,
        -1.5862e-04,  2.2447e-05, -5.1556e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5024e-04, -9.9715e-01, -1.4936e-04,  1.5366e-04, -7.3485e-05,
         5.1215e-05, -2.9670e-04, -6.6851e-05, -1.8124e-04, -1.9495e-04,
        -1.1686e-04, -9.3736e-05,  1.2416e-05,  2.7733e-05,  2.1274e-04,
        -2.2888e-04,  4.4857e-05, -5.0611e-05, -2.0793e-04, -5.0557e-05,
        -5.6051e-06,  6.6168e-05,  2.4307e-04, -7.2353e-05,  1.0170e-04,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.3014e-04, -9.9530e-01, -1.5290e-04, -3.9447e-04, -3.6135e-04,
         7.3715e-05, -1.6271e-04,  2.9703e-04, -2.0309e-04, -2.0818e-04,
        -5.8875e-04,  2.6126e-04, -3.2473e-04,  6.3658e-05, -7.8718e-05,
         3.2689e-04, -7.2575e-05, -8.7959e-05,  3.7787e-05,  4.6111e-04,
         4.0566e-05, -3.1420e-05, -1.3769e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6013e-04, -9.9586e-01,  1.6599e-04, -2.2679e-04, -1.2873e-04,
        -3.5506e-04,  8.1573e-05,  1.4179e-04, -5.0384e-04,  4.0059e-05,
        -2.7452e-04, -8.5576e-05, -2.5621e-04, -1.0217e-04, -1.4610e-04,
        -6.6964e-05, -1.3876e-04, -1.8212e-04,  2.0035e-05, -3.2266e-04,
         5.7311e-05,  3.7759e-04,  2.9627e-05, -1.7451e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3259e-04,  9.9497e-01,  1.1480e-04,  6.1247e-05,  6.0093e-04,
         6.2046e-04, -6.2916e-05, -9.8713e-06, -1.1999e-04,  2.5827e-04,
         1.1900e-04,  3.4963e-04,  2.3471e-04, -1.1617e-04,  1.2522e-04,
        -8.0829e-05,  2.2310e-04, -6.2841e-05, -2.3738e-04, -6.0872e-05,
         4.3441e-04, -3.7920e-05, -1.8366e-04, -1.9952e-04, -1.1702e-04,
        -3.1456e-04,  5.4948e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3390e-03, -9.9606e-01, -1.9232e-04, -2.0335e-04,  2.0501e-05,
         7.7034e-05, -2.0118e-05, -1.2990e-04,  2.7396e-04, -6.7350e-05,
         8.6686e-07, -2.4505e-04, -1.0237e-04, -6.0761e-06,  1.5096e-04,
        -1.5305e-04, -7.0948e-05,  1.6779e-04, -2.5534e-05, -1.1739e-04,
         4.8199e-05,  1.2048e-04,  1.9600e-06, -1.2512e-04,  2.1872e-04,
        -6.0612e-05,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #2050: [tensor([ 5.7871e-04, -9.9238e-01, -2.5895e-04,  4.3154e-05,  2.7633e-04,
         3.2233e-04, -1.0515e-04, -1.8452e-04,  3.8031e-04,  1.4077e-04,
        -3.0254e-04,  3.4447e-04, -2.7230e-04,  3.7653e-04, -1.3111e-04,
        -3.4846e-05, -8.2885e-05,  8.5297e-05,  4.8875e-04,  3.1534e-04,
         2.5503e-04,  5.5916e-04, -8.5504e-05, -1.1572e-06,  2.4552e-05,
        -1.2332e-04,  5.9284e-05,  9.2412e-05,  1.5375e-04,  8.0214e-05,
         1.1893e-04, -4.5289e-04,  4.8655e-04,  1.6898e-04,  2.3874e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1894e-04, -9.9101e-01,  4.2090e-04,  3.5284e-04, -1.5554e-04,
         3.3672e-04, -9.6535e-05,  5.3283e-05,  4.1729e-04,  3.8062e-04,
        -2.1377e-04,  4.4100e-04, -6.2881e-04,  9.3839e-05,  1.3902e-04,
         9.5309e-04, -1.7864e-04,  1.1981e-04, -5.4600e-04, -3.9212e-04,
         3.8042e-04,  1.9914e-04,  4.5789e-05,  3.2412e-04,  5.3234e-05,
         5.8114e-05,  4.3771e-04,  1.5051e-04, -4.4811e-04,  7.0349e-05,
        -5.3286e-04,  2.4595e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.1033e-04,  9.9228e-01,  8.3663e-05,  2.5733e-04,  5.3804e-04,
         1.2078e-04,  4.6197e-04,  7.4201e-04,  4.0396e-04, -2.6127e-04,
         1.0419e-04,  1.2222e-04,  1.4780e-04,  4.5583e-05, -5.9830e-05,
        -2.9478e-04,  3.9097e-04,  2.5437e-04, -7.2253e-05, -6.1240e-05,
         1.3325e-04, -3.3155e-04,  1.6453e-04,  3.2270e-04,  5.1638e-05,
         3.1906e-04, -4.2503e-04, -6.2628e-04,  8.0504e-05,  1.3453e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.8738e-04,  9.9199e-01, -1.0530e-04,  8.6780e-05,  2.6026e-04,
        -2.3224e-04,  3.3303e-04,  2.6137e-04, -5.2811e-05, -9.6225e-05,
        -2.0740e-04, -2.4543e-04,  1.3754e-04, -4.0963e-04, -3.3225e-04,
         1.0094e-04, -1.7011e-04, -2.7949e-04, -4.2665e-04, -4.6957e-04,
        -2.3949e-04, -2.8833e-04, -2.3348e-04,  7.3833e-05,  1.1540e-04,
        -2.5509e-04, -2.7986e-04, -5.8782e-05,  2.0825e-04, -1.8410e-04,
         1.0564e-04, -5.1273e-04, -9.3326e-05,  3.7898e-06,  2.4946e-04,
        -1.9332e-04, -2.2463e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.1452e-04,  9.9095e-01, -5.1134e-04,  9.8791e-05, -1.4852e-04,
        -7.9985e-04,  3.8085e-04,  6.4013e-04,  2.6796e-04, -4.1862e-04,
        -2.7863e-04,  2.6264e-04,  1.8065e-04, -6.9554e-04, -3.0871e-04,
        -2.9962e-04, -1.3331e-04,  2.5712e-04, -4.7370e-04,  2.9359e-04,
        -3.0250e-04, -9.6478e-05, -3.3186e-04,  1.2048e-04,  2.6716e-05,
         2.1350e-04,  4.1990e-05,  1.1277e-04, -3.6387e-05,  3.6573e-04,
        -3.3676e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.4282e-04,  9.9136e-01, -5.0853e-04,  3.8377e-06,  3.1283e-04,
        -2.4861e-05,  1.6915e-04, -2.7752e-04,  3.0704e-04,  7.9017e-05,
         3.6340e-04,  6.3609e-05,  2.6200e-04,  2.1378e-05, -2.3301e-04,
        -4.1545e-04, -1.3229e-04,  2.4464e-04, -4.0203e-05,  1.9086e-05,
         1.9316e-04,  1.0287e-04, -2.8692e-04,  3.3558e-04, -1.0482e-04,
        -2.8433e-04,  1.2541e-04,  1.9443e-04,  1.4833e-04,  3.5466e-04,
         1.0535e-04, -9.3948e-05, -7.7881e-05,  9.8401e-05, -4.5019e-04,
         1.9881e-04, -2.8636e-04,  2.7421e-04, -3.9916e-04,  1.8357e-04,
        -1.5686e-04, -3.5926e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9456e-04,  9.9223e-01, -9.1492e-05, -2.8645e-04,  1.6876e-04,
         5.9615e-05, -1.6667e-04,  1.1912e-04, -2.8738e-04, -8.1507e-04,
        -5.8008e-04,  8.4718e-05,  6.0132e-04, -4.6683e-04,  4.4459e-04,
        -2.4682e-05,  3.1624e-04, -1.7963e-04, -2.0461e-04, -2.7015e-04,
         1.3656e-04, -6.8382e-04, -3.2828e-04,  7.0077e-05, -6.7080e-07,
         1.5837e-06,  3.9774e-05, -2.2727e-04,  2.4415e-04, -3.8432e-04,
         1.9486e-05,  1.6394e-04,  1.1187e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.1522e-04,  9.9043e-01, -1.7846e-04, -2.1073e-04,  2.2056e-04,
        -7.4798e-04, -2.4660e-04, -2.1127e-05,  1.8005e-04, -1.0358e-05,
        -6.1516e-04, -1.6493e-04,  5.6979e-04, -7.7618e-04, -5.4588e-04,
        -2.4048e-04, -3.0492e-04, -3.2351e-04, -3.0546e-04, -3.2547e-04,
        -3.5296e-04,  9.4422e-05,  5.7944e-05, -8.7869e-05, -2.8264e-04,
         3.7472e-04,  1.4677e-04, -4.5365e-06,  1.5487e-04, -4.4467e-04,
        -4.5700e-04,  2.0458e-04, -2.6929e-04, -3.0942e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.7217e-05,  9.9276e-01, -5.7168e-04,  2.0894e-04, -3.3925e-05,
         4.2375e-04, -3.1675e-04,  9.4830e-05,  2.4915e-05,  2.0413e-05,
        -2.6073e-04,  6.0813e-05,  9.2272e-05, -2.3492e-04, -3.6975e-04,
        -2.0991e-04,  3.0012e-05, -2.6924e-05, -3.3353e-04, -1.8703e-04,
        -8.3108e-04,  2.2318e-04, -9.6255e-06, -1.1391e-04, -3.4125e-04,
        -1.6996e-04, -5.4736e-04, -2.4768e-04, -1.5785e-04,  3.3500e-04,
        -1.7932e-05,  2.3877e-04,  3.0271e-04,  1.4109e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6469e-05,  9.9274e-01, -8.2313e-06, -3.6872e-05,  6.2727e-05,
         7.0515e-05,  3.9366e-04, -9.1637e-05, -1.2032e-04, -1.4836e-04,
        -3.6500e-05,  4.8247e-05,  3.6907e-04, -1.1085e-04, -8.5989e-05,
        -1.9368e-04,  4.1903e-04,  5.2289e-05,  1.4513e-04,  4.8987e-06,
         2.1418e-04, -1.2220e-04, -5.8863e-05, -2.6414e-04,  9.1690e-05,
        -2.2207e-04, -1.0346e-04, -1.4262e-04, -5.8506e-05, -2.7509e-05,
        -2.2698e-04, -4.2745e-04, -3.1645e-04,  3.8994e-04,  3.1110e-05,
        -1.2865e-04, -5.3409e-04, -3.2248e-04,  2.7308e-04,  4.5654e-05,
        -5.5373e-06, -5.1445e-04,  7.8453e-05, -2.5006e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0191e-03, -9.8971e-01,  2.4185e-04,  1.0229e-04, -2.8083e-04,
        -1.1878e-05, -1.4358e-04, -1.2397e-04, -2.2455e-04, -4.7829e-05,
         3.6107e-04,  5.3022e-05,  3.9599e-04,  1.0842e-04,  1.8079e-05,
        -3.6597e-04, -9.6983e-05, -2.8044e-04, -2.5729e-05, -5.1641e-04,
         1.4172e-04, -4.4527e-04, -1.8419e-06, -2.0006e-04,  1.3433e-04,
         1.1132e-04, -3.1657e-04,  1.6728e-04, -2.7032e-05, -1.5674e-04,
        -6.1346e-04,  2.3388e-04,  6.3004e-05, -3.3632e-04, -3.0233e-04,
        -3.5499e-04, -2.5016e-05,  6.6558e-05, -6.7608e-05, -3.3183e-04,
         3.3491e-04,  1.1753e-05, -2.6325e-04,  7.7176e-05, -9.2014e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7355e-03, -9.9152e-01,  1.6274e-04,  1.1042e-04,  5.7306e-04,
         1.6374e-05,  1.0984e-04, -5.1490e-04,  6.2126e-05,  5.9571e-04,
        -1.9915e-04,  2.3199e-04,  2.3717e-04,  6.1309e-05,  2.5447e-05,
         4.3621e-04, -1.5731e-04,  8.5729e-05,  1.6405e-04, -2.5591e-04,
         3.7392e-04, -2.8845e-04, -1.0973e-04, -6.8834e-04, -4.5499e-05,
         2.6430e-05, -8.9099e-05, -7.0848e-05, -3.4317e-04,  1.8023e-04,
        -2.5961e-04,  2.7297e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #2100: [tensor([-2.6699e-04, -9.9575e-01,  6.6360e-06, -5.6658e-05, -1.4213e-04,
         4.6782e-04, -5.9369e-05, -1.6656e-04, -4.8925e-05, -4.0167e-05,
         1.0081e-04, -1.9458e-06,  2.3472e-04, -4.0145e-05, -1.0601e-04,
         4.2986e-05,  6.0817e-04,  1.1855e-05, -3.6991e-05, -1.8950e-04,
        -4.4976e-05,  3.6003e-05, -4.3435e-05,  8.0824e-05, -5.1750e-05,
        -4.7632e-05,  1.2382e-05, -3.2887e-05, -1.2420e-04, -3.1154e-04,
         9.7064e-05, -1.2259e-04,  1.5670e-04,  1.5054e-04, -1.3743e-04,
         1.7142e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7784e-04, -9.9633e-01, -1.1175e-04,  8.6075e-05,  1.4206e-04,
         2.9527e-04,  4.5994e-05, -2.4753e-05,  9.9642e-05, -1.3093e-04,
        -9.1014e-05, -2.7104e-05,  2.9026e-04,  6.9549e-06,  1.1417e-04,
        -9.6233e-05,  2.3426e-04, -5.1976e-05, -1.0755e-04,  7.8668e-05,
         9.0058e-05, -1.3806e-04, -7.7014e-05, -1.9372e-05,  7.3434e-06,
         2.8795e-04,  8.1662e-05, -1.3980e-04, -8.9986e-05,  3.0237e-05,
         4.2378e-05, -2.9319e-05, -1.6820e-04, -5.3109e-05, -5.5955e-06,
         9.6816e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6586e-04, -9.9658e-01, -1.0279e-04, -4.5375e-05, -2.8506e-05,
         5.8553e-05,  8.7770e-05, -1.8616e-04,  1.1350e-04,  1.1116e-04,
        -1.8928e-04, -1.5788e-04, -9.9089e-06, -1.4563e-04, -5.8201e-05,
        -8.3704e-05,  4.7656e-05,  1.9974e-04,  2.7832e-05, -1.9955e-04,
         8.3234e-06, -3.1480e-04, -3.4342e-05, -2.5640e-04,  1.4173e-04,
         1.5465e-05, -1.7802e-04,  5.9525e-05, -4.4968e-05,  1.2091e-04,
         1.7506e-04, -5.2997e-06, -4.5573e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1949e-04,  9.9614e-01, -6.0255e-05, -7.7976e-06,  5.0795e-05,
         7.6013e-05, -2.1886e-06,  6.3529e-05,  1.3466e-04, -2.3035e-04,
         4.3031e-04, -1.6592e-04,  3.0811e-04, -3.8122e-05, -8.7124e-05,
         2.3729e-04,  1.4027e-05,  2.3364e-04,  6.1623e-05,  1.5301e-04,
        -1.6009e-04, -1.4548e-05,  3.3739e-05, -7.9907e-05,  3.1167e-05,
        -8.9754e-05, -1.9229e-04, -1.5077e-05,  5.8015e-06, -6.8923e-06,
        -3.1059e-04,  6.7311e-05, -1.8204e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.8209e-05,  9.9654e-01, -6.0041e-05,  1.4601e-04, -1.8141e-04,
         1.7822e-06,  2.8017e-05, -3.2811e-05, -1.5787e-04, -1.1718e-04,
        -3.4801e-05, -2.0438e-04, -3.2175e-04,  5.7571e-05,  9.2048e-05,
         1.4089e-04, -1.3015e-04,  5.4528e-06, -1.2180e-04, -1.1897e-04,
         2.7156e-04,  1.6348e-05, -4.3160e-05, -1.0044e-06,  1.4812e-05,
         5.5395e-05,  2.3505e-06,  6.1811e-05, -3.0021e-04, -4.4536e-05,
         1.0672e-04, -2.5532e-04, -1.1967e-04, -5.5515e-05, -2.5625e-05,
        -5.2688e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0095e-04, -9.9633e-01, -8.9730e-05, -1.7003e-06, -6.6383e-05,
         1.9551e-04, -1.9416e-04, -7.4851e-05,  6.7046e-05,  7.3964e-05,
         2.0718e-04,  2.1788e-04,  2.1860e-04, -2.5861e-04,  5.3485e-05,
        -1.4526e-04,  1.5476e-04, -2.7281e-04, -2.6179e-04,  7.6984e-06,
         9.9629e-05, -1.3631e-04,  1.1354e-04, -1.5710e-04,  7.9957e-05,
        -1.7290e-07,  5.9265e-05,  6.2616e-05,  4.2457e-05,  6.4417e-05,
        -8.1231e-05, -6.0859e-06,  1.2919e-05,  1.6416e-05, -7.3723e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.9596e-04,  9.9675e-01, -4.1385e-06,  2.5999e-05,  4.1298e-05,
        -7.5209e-05,  4.5935e-05,  1.2313e-04, -1.4654e-04, -2.0390e-04,
        -5.1575e-05, -1.2542e-04, -1.5614e-04, -3.3256e-05,  6.4401e-05,
        -2.6240e-05, -1.2878e-06,  1.0484e-04,  6.8396e-05, -1.0718e-05,
        -6.6596e-05, -6.3965e-05, -1.0313e-05, -2.3278e-04,  8.9914e-05,
         1.2054e-04, -1.7637e-04, -9.2326e-05,  9.8815e-05, -5.9589e-05,
        -1.4079e-04, -8.9899e-05, -2.0049e-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([ 1.0834e-04,  9.9667e-01, -2.3433e-04,  7.9511e-05,  1.8974e-04,
         1.5213e-04,  6.0565e-05,  6.7256e-05,  5.0085e-05, -1.1051e-04,
        -1.0612e-05, -1.1388e-04,  1.3294e-04, -1.2206e-04, -7.7246e-05,
         4.4215e-05,  2.3408e-05,  3.1320e-05,  8.8839e-05, -2.0201e-04,
         4.6140e-05, -4.9566e-05, -1.4881e-04, -3.3080e-05,  8.7344e-05,
        -1.3907e-04, -1.2949e-04, -3.9309e-05,  7.2811e-05,  1.0778e-04,
         4.7550e-05,  1.9173e-04,  4.6660e-05, -1.0886e-04, -1.0604e-05,
        -1.1450e-06,  1.4239e-04,  2.5744e-05,  2.7288e-06], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.3493e-05,  9.9628e-01, -9.9104e-05, -1.1316e-04,  1.5045e-04,
         4.8345e-05, -3.0986e-05,  2.9803e-04, -2.5681e-05,  1.1646e-04,
        -1.3939e-04,  6.5498e-05,  1.5219e-04,  8.0882e-06, -2.4793e-04,
         2.6555e-04, -1.2289e-04,  1.3545e-04,  1.2339e-04,  1.2846e-04,
        -8.4906e-05,  1.0137e-04,  1.5412e-04, -1.4127e-05, -7.9226e-05,
        -2.1948e-05, -1.4728e-04,  1.9788e-04,  1.2150e-04,  2.0933e-04,
        -1.6637e-04,  6.0960e-05,  8.4329e-06,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.9676e-04, -9.9582e-01, -2.8837e-05, -9.6295e-05,  1.0035e-04,
        -1.3578e-04, -9.2188e-05, -1.5300e-04, -1.5249e-05, -6.3833e-05,
        -2.8451e-04, -4.7445e-05,  1.7175e-04, -2.5194e-04,  2.2063e-05,
         1.4323e-04, -1.5009e-04,  7.4659e-05, -2.0657e-04,  1.2113e-04,
        -1.5138e-04,  8.1584e-06, -4.1940e-05,  1.7257e-04,  4.4370e-05,
         2.3002e-04, -5.7995e-05,  8.5083e-05, -8.8994e-05,  1.7484e-04,
         1.7049e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.5538e-04, -9.9543e-01,  2.2872e-05,  1.0044e-04,  1.0767e-04,
         3.9058e-04,  7.5677e-05,  6.0692e-05,  2.2803e-05,  8.9040e-05,
        -6.0396e-05, -2.0547e-05,  2.3171e-05, -2.0483e-04, -5.6962e-05,
         3.4132e-05,  1.9643e-04,  5.4107e-05,  4.2486e-04, -1.2231e-04,
         2.0028e-04,  2.3284e-04,  2.5113e-04,  3.8227e-04,  1.6363e-04,
        -6.4632e-05,  2.2860e-05,  1.3557e-05, -1.0614e-04,  4.4399e-06,
         1.1668e-05, -1.6585e-04,  1.8869e-04,  3.5828e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.6859e-05, -9.9604e-01,  8.2545e-05, -3.1345e-04, -1.1038e-04,
        -2.7265e-04, -1.7781e-05,  9.6065e-05, -7.2022e-05,  1.5517e-04,
        -5.0714e-05,  2.4706e-04, -2.2109e-04, -7.8554e-06, -1.6605e-05,
         4.0990e-05, -6.9761e-05, -1.2466e-04, -1.7643e-04, -1.6924e-04,
        -1.1541e-04,  5.2325e-06, -1.4758e-05, -3.0440e-04,  1.2605e-04,
         5.0640e-05, -1.4128e-04,  1.0589e-04, -3.5617e-05,  1.6764e-04,
         2.5665e-04, -1.0623e-04,  2.1822e-04, -2.0108e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #2150: [tensor([ 1.1970e-05,  9.9651e-01,  7.3932e-06,  1.8964e-04, -1.2760e-04,
         1.2917e-04, -2.4894e-05, -2.9081e-04,  5.5914e-05, -1.7925e-04,
        -1.5060e-05, -1.9036e-04, -1.1145e-04, -2.3377e-04, -4.0616e-05,
         1.9991e-04, -9.4581e-05, -4.2349e-04, -3.6000e-05,  1.1228e-04,
        -8.6930e-05,  1.0126e-04,  4.6886e-06,  1.3497e-04,  4.6538e-06,
         1.3909e-04,  2.0571e-04, -1.2780e-04,  1.3847e-04,  7.4051e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1373e-04,  9.9687e-01, -3.0015e-04, -3.3828e-04, -1.3429e-04,
        -1.3652e-04, -9.3163e-05, -2.7181e-04, -1.6149e-04,  6.9824e-05,
         3.0249e-05, -3.3282e-04, -7.4785e-05, -1.3282e-04, -1.4498e-04,
        -6.4365e-05,  5.3561e-05, -8.8223e-05,  1.9396e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.2582e-04,  9.9691e-01,  3.2386e-04,  2.5078e-06,  4.4944e-05,
         1.7360e-04, -9.7926e-05, -4.5649e-06, -3.4984e-05, -6.3823e-05,
        -2.1901e-04, -3.3332e-04,  1.8362e-04, -2.1711e-04,  3.8672e-04,
        -2.7889e-04,  4.0187e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.4631e-04,  9.9509e-01,  3.2630e-04, -4.0299e-04, -8.6084e-04,
        -5.0056e-04, -4.9720e-04, -1.6498e-04, -2.7788e-04, -1.9926e-04,
         2.5320e-04, -4.9240e-05,  1.0836e-04,  3.6651e-04, -6.1707e-05,
         1.9489e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.9609e-04, -9.8726e-01, -5.5154e-04, -4.1936e-04,  7.2998e-05,
        -3.4746e-04, -1.1238e-03, -4.9845e-04, -1.4957e-04, -5.4643e-04,
        -2.0783e-04,  3.4920e-04, -1.0223e-03, -3.8043e-04, -2.4515e-03,
        -3.8918e-04,  6.0455e-04, -3.8604e-04, -7.9995e-05,  1.9679e-06,
        -4.8231e-04, -6.1459e-04,  8.1918e-05, -5.6264e-04,  4.1909e-05,
        -1.0238e-04,  4.2021e-04,  3.5676e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7387e-04,  9.9734e-01, -3.6322e-05,  1.6366e-04,  3.2468e-05,
         1.5055e-05, -4.8528e-05, -1.0916e-04,  7.5406e-05, -9.4498e-05,
         6.8388e-05,  1.4262e-04, -1.2828e-04,  1.3341e-05, -2.5989e-04,
         9.6898e-05, -1.9314e-05,  6.0856e-05, -1.0010e-04,  2.7101e-04,
        -2.7135e-04,  3.7235e-05, -3.5754e-05, -1.2012e-04, -1.0645e-04,
        -4.5294e-05, -5.0325e-05,  7.9554e-05,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0294e-03, -9.9567e-01, -3.9438e-04,  4.9075e-05,  6.1275e-05,
         5.0392e-05,  6.3136e-05,  3.1337e-04, -2.9123e-04,  2.1383e-04,
         2.5171e-04,  4.3379e-05,  5.0354e-05,  2.3510e-04,  7.9015e-05,
        -2.3592e-06, -1.3855e-04, -2.8549e-05, -1.5177e-05, -6.6840e-05,
         3.6470e-05,  3.1978e-04, -7.7128e-05,  2.6468e-05,  1.1946e-04,
         2.0134e-04,  2.3234e-05, -1.4774e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.9877e-04,  9.9281e-01,  5.0769e-05,  3.5208e-04, -2.4177e-04,
        -6.2754e-04, -2.4709e-04,  6.3317e-05, -4.6114e-04,  4.0005e-04,
        -4.7796e-05, -1.2548e-04,  9.8193e-05,  6.6781e-04, -1.6839e-04,
        -2.6392e-04, -4.7877e-04,  1.6123e-04, -3.1333e-04, -3.0463e-04,
         3.1977e-05,  1.2671e-04,  9.3492e-05, -2.6357e-04,  3.2721e-04,
        -4.4847e-05, -3.8684e-05,  3.5757e-04, -4.3422e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3432e-03, -9.9173e-01, -2.9268e-04,  1.6029e-04,  4.9325e-05,
         1.6826e-04, -6.9224e-04,  2.4405e-04,  3.7555e-04,  2.4174e-04,
         2.3765e-04,  2.2036e-04, -1.2789e-04, -1.7990e-05,  2.8556e-04,
        -7.2799e-04,  4.0694e-04, -1.2651e-04, -3.2530e-04, -9.8970e-06,
         1.2728e-04, -7.0826e-04,  1.5190e-04, -4.6601e-04, -1.9630e-04,
        -5.6394e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3060e-03, -9.9297e-01,  3.1026e-05, -1.9386e-04, -6.3060e-05,
         1.4148e-04, -6.5195e-04,  1.2750e-04, -4.5737e-04, -1.7710e-04,
        -1.1184e-04, -4.0035e-04, -7.2991e-04, -1.4389e-05,  1.5162e-04,
        -4.3762e-04, -2.2592e-04, -4.9822e-04, -8.1027e-05,  8.9893e-05,
         5.2593e-05, -2.1971e-04, -9.7351e-05,  3.6238e-04, -2.2884e-04,
         1.7887e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.5329e-05,  9.9488e-01, -4.5116e-05,  2.6975e-04, -7.3003e-05,
        -3.0170e-04,  3.7464e-04, -6.7578e-05,  1.4379e-04, -3.3337e-04,
         1.2389e-04,  3.1900e-04,  3.5093e-04, -2.4058e-05,  1.0469e-04,
         5.6878e-04, -1.3870e-04, -1.9529e-04, -1.1117e-04, -3.6807e-04,
        -1.7349e-04, -1.5911e-04, -1.5411e-04,  2.5230e-04, -1.6541e-04,
         1.2882e-04, -7.9569e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.7231e-04,  9.9437e-01,  1.4456e-05,  3.2152e-04, -3.1599e-04,
         1.3543e-04, -1.1058e-04,  1.2178e-04,  3.4888e-04, -5.0278e-04,
         9.3082e-05, -1.7658e-04,  3.2134e-04, -1.7724e-04, -1.7791e-04,
         6.2081e-05,  1.8266e-04,  2.2351e-05,  1.0743e-04, -3.5861e-04,
        -6.6702e-04, -3.5438e-04, -1.4181e-04, -7.4431e-05, -2.4818e-05,
         3.4447e-05, -1.0627e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #2200: [tensor([-1.1511e-03, -9.9659e-01,  4.5220e-05, -3.6260e-05,  2.4840e-04,
        -1.1616e-04,  1.0014e-04, -1.0423e-04, -2.6352e-04, -2.4280e-04,
         1.1812e-04, -2.3843e-04,  6.7098e-05, -2.0023e-04,  1.8142e-04,
        -4.4345e-05, -6.0884e-05,  4.3482e-05,  2.9551e-05, -4.7526e-05,
        -3.6334e-06,  6.5192e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4904e-04,  9.9561e-01, -2.9200e-04, -1.7028e-04,  7.4824e-04,
        -5.0437e-04, -2.2488e-04, -2.4487e-04, -1.0708e-04, -2.1617e-05,
        -1.2928e-04,  1.9930e-04, -2.6452e-04,  3.8832e-04,  3.3240e-05,
        -5.1126e-04,  1.9938e-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([-4.4524e-04, -9.9639e-01, -1.6900e-04, -3.9244e-04,  1.1803e-04,
        -2.5102e-04, -7.4158e-05, -4.0630e-04, -3.3345e-04, -3.5963e-04,
         1.0623e-04, -1.9061e-04, -2.1557e-04, -8.5621e-05,  1.1808e-04,
        -1.4434e-04, -1.9961e-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([-1.1004e-04, -9.9788e-01, -7.8670e-05, -1.6830e-04,  2.3445e-04,
         3.0845e-04, -9.0468e-05, -1.8530e-04,  2.3456e-04,  7.6213e-05,
        -3.0321e-04, -2.5697e-04, -4.6270e-05,  6.4856e-06,  2.2446e-05,
        -2.7018e-07,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.4966e-04,  9.9670e-01, -3.9484e-04,  9.2225e-05,  2.3178e-05,
         4.0660e-05,  6.3801e-05,  8.1512e-05, -1.3232e-04,  4.8422e-05,
        -1.2673e-04,  1.8900e-04, -5.4532e-04, -8.8811e-05,  4.3851e-04,
         2.3134e-05, -1.0058e-04, -1.6724e-04,  1.9791e-04,  1.6750e-04,
         2.0994e-06,  2.9827e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.3082e-04, -9.9611e-01, -2.4104e-04, -1.4141e-04, -1.1850e-04,
         1.7286e-04,  2.2336e-04,  1.8441e-04, -9.0624e-05, -2.0018e-04,
        -6.0935e-06, -1.8100e-04, -1.9258e-04,  3.1585e-05, -6.6323e-05,
         3.4628e-05,  4.0202e-04, -3.5816e-04,  2.1337e-04,  2.9368e-04,
        -2.0676e-04, -2.8562e-04,  1.1373e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-5.5445e-04, -9.9741e-01,  1.7953e-04, -8.9378e-05,  6.5658e-06,
         1.1469e-04, -4.7455e-05, -4.0471e-05,  1.2918e-04, -1.6524e-05,
        -1.6091e-04, -1.8815e-04, -2.2676e-05,  8.9776e-05,  1.4271e-05,
        -1.8886e-04,  1.0706e-04, -2.7402e-04, -1.2715e-04, -1.6502e-04,
         3.3560e-06,  6.7895e-05,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0704e-03,  9.9306e-01, -2.8321e-04,  3.9198e-04,  3.9563e-04,
        -2.7785e-04, -2.8946e-04,  1.7340e-05, -5.2482e-04, -1.1300e-04,
        -1.6109e-04, -2.9150e-05, -1.9788e-04,  7.2874e-04,  7.7274e-04,
         8.3235e-05,  3.9462e-04,  3.3342e-04,  7.1431e-05, -3.4815e-06,
        -3.3099e-04,  3.9124e-04,  7.7087e-05], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.3995e-03, -9.8940e-01,  1.0644e-03, -7.4626e-04,  3.8863e-04,
        -2.4853e-05, -5.9734e-04,  5.6897e-04, -5.3070e-04,  2.3040e-04,
         5.2202e-04,  3.3743e-04,  1.0323e-04, -9.0752e-04, -3.9047e-04,
        -6.1658e-04, -3.4875e-04,  6.8968e-04, -1.8629e-04,  1.0484e-04,
        -1.3459e-04,  2.1939e-04,  4.8801e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.1796e-05,  9.9369e-01, -6.2377e-05,  1.1619e-03, -1.1084e-03,
        -2.3522e-04, -6.3934e-05, -2.3516e-04,  9.5890e-04,  1.2643e-04,
        -3.3943e-04,  6.5273e-05, -1.1582e-03,  1.4382e-04, -3.5322e-04,
        -2.0045e-04, -6.9921e-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([ 1.6603e-05, -9.9726e-01,  1.6418e-05, -1.5075e-04,  2.6402e-04,
        -2.4649e-04, -4.0856e-04, -2.6771e-04, -1.6875e-04,  8.0108e-05,
         4.1397e-05,  2.2366e-04,  3.3556e-05,  1.1890e-04,  1.8061e-05,
         6.8154e-05,  1.0103e-04,  3.3579e-04,  3.3149e-05, -4.0224e-05,
        -1.0245e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.6794e-04, -9.9661e-01, -2.5191e-06, -3.8315e-04, -5.6820e-05,
        -3.3408e-04, -3.6189e-04, -4.1634e-04,  1.3353e-05, -3.0886e-04,
        -6.4248e-05,  3.4024e-04,  7.9846e-06,  7.9101e-05, -1.3458e-04,
         3.0227e-04, -1.9306e-04, -2.3557e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #2250: [tensor([-7.5882e-04,  9.9282e-01, -5.0918e-05, -2.2261e-04, -7.7696e-04,
         2.8276e-04,  4.1071e-04,  3.7637e-05, -4.5416e-04, -3.9290e-04,
        -4.7943e-04, -1.7016e-04, -5.5581e-05, -4.4954e-04, -9.0336e-05,
         3.1312e-04,  5.7667e-04,  1.8325e-05, -3.8150e-04, -5.1656e-04,
         4.6160e-04, -2.7967e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.1809e-03,  9.8780e-01, -3.9790e-04,  9.4153e-04, -5.8961e-04,
        -4.4103e-05,  5.1017e-05, -4.4604e-04,  1.7942e-03, -9.2911e-04,
         9.6300e-04, -1.0725e-04,  3.7110e-04,  3.6436e-04, -1.7582e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6056e-03,  9.8562e-01, -1.8798e-03, -1.5635e-03,  5.4752e-04,
         6.7339e-04, -1.0465e-03, -1.9885e-04,  1.0647e-03,  2.6067e-04,
        -2.6155e-04,  1.6167e-03, -9.1400e-05, -9.0975e-04,  1.3946e-04,
         1.8725e-04, -3.1585e-05, -2.9919e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.8505e-03,  9.8170e-01, -4.8868e-04, -1.7817e-03, -5.7953e-05,
        -5.6412e-04, -1.0620e-03, -9.2323e-04,  1.9584e-03,  4.5358e-04,
        -1.1053e-03,  5.7764e-04, -1.9554e-04, -6.5349e-04, -1.5416e-04,
        -3.4446e-04,  5.4662e-04,  3.7918e-04, -2.0857e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8721e-04,  9.9405e-01, -4.9875e-04, -8.9656e-04,  1.9748e-04,
         4.7261e-04,  5.6570e-04,  1.8592e-05, -1.4944e-05,  1.0603e-04,
         2.8695e-04,  1.6167e-04,  1.7875e-04,  3.7470e-04, -3.1029e-04,
        -3.1273e-04,  1.1976e-04, -5.5972e-04,  4.0966e-04,  7.7751e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.2541e-04, -9.9402e-01,  3.7912e-05, -6.7959e-04, -1.0720e-04,
         1.3633e-04, -7.3213e-04, -4.1467e-04,  1.7488e-04, -1.3935e-04,
        -3.8197e-04,  5.1504e-04,  3.4873e-04, -9.3735e-05, -1.4934e-04,
         1.2052e-04,  1.7478e-05,  9.2922e-06, -1.9629e-04,  6.8485e-05,
        -5.0802e-04, -2.3855e-04,  8.2223e-05,  1.4248e-04,  4.0264e-05,
         1.9734e-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([-8.0617e-04,  9.9154e-01, -3.5320e-04, -4.9567e-04, -3.4880e-04,
        -6.0810e-04,  1.1300e-05, -1.7180e-04, -3.2641e-04,  1.6843e-04,
        -1.4603e-04, -1.4622e-04, -8.4759e-04,  1.5237e-04,  1.4756e-04,
        -3.2396e-04, -3.6462e-04, -1.4552e-04,  2.1665e-04, -1.2763e-04,
        -2.1306e-05, -1.7093e-04,  1.1328e-04,  2.2327e-04, -1.6029e-04,
        -8.6153e-05, -3.7665e-04, -2.6117e-04, -3.0734e-04, -4.1581e-04,
        -3.8629e-04, -2.6252e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.3759e-04,  9.9410e-01,  1.4148e-04, -1.0295e-04,  1.0136e-04,
        -1.0025e-06, -1.9459e-04, -6.6117e-04,  1.1690e-03,  4.7668e-04,
        -1.2202e-04, -4.0789e-04, -2.9281e-04, -8.0107e-04, -5.0293e-05,
         3.2922e-04, -1.1940e-04,  1.9465e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.8994e-04,  9.9340e-01, -6.1659e-04,  4.3091e-04,  5.6374e-04,
        -5.5227e-05,  1.9839e-04, -1.1464e-03, -1.2471e-04,  2.6883e-04,
        -9.1192e-05,  2.5936e-05,  8.1385e-04, -4.1846e-04,  2.6729e-04,
         4.0200e-04, -2.4217e-04, -9.0730e-05,  4.5737e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0693e-03,  9.9055e-01, -9.2001e-04, -1.5398e-04, -2.9499e-04,
        -4.4744e-04, -5.4182e-04, -1.6676e-03, -1.3078e-04,  6.3211e-04,
        -3.9862e-04, -1.8274e-04,  3.1076e-04, -5.2832e-04,  8.4383e-04,
         3.8308e-04, -5.0838e-04,  4.3362e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3102e-03, -9.9330e-01,  1.5148e-04, -1.0774e-04,  4.0791e-04,
         8.2935e-05,  1.2554e-04,  2.8463e-04, -1.4663e-04,  3.7221e-04,
         1.5696e-04, -1.3742e-04,  8.7039e-05,  1.7583e-05, -2.6292e-05,
        -1.9737e-04, -8.8558e-05, -3.2252e-05, -1.5646e-04, -1.3728e-04,
        -1.2939e-04, -3.8913e-05,  1.9728e-05,  1.7472e-04,  9.7415e-05,
        -1.9075e-04,  6.9203e-05,  1.3406e-04, -1.9399e-04,  3.1962e-04,
         2.5866e-04,  4.4886e-05], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.8721e-04, -9.9488e-01, -3.3134e-04, -9.4464e-05, -5.6543e-06,
        -2.0067e-04,  3.2219e-04, -7.9395e-05,  2.5723e-04,  9.1785e-05,
         1.1858e-05, -6.3415e-05, -6.4165e-05,  3.3285e-04,  1.4610e-04,
        -3.3625e-04,  2.3074e-05,  2.8429e-04,  1.6414e-04,  1.4035e-04,
        -1.7558e-04,  1.9325e-04, -1.8938e-04,  3.2596e-04,  2.1130e-04,
         2.4572e-04,  4.1820e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
