Iter #50: [tensor([ 0.2297, -0.0218,  0.0477, -0.0395, -0.0094,  0.0306, -0.0191, -0.0409,
         0.0044, -0.2939,  0.0199,  0.0014, -0.0424, -0.0082,  0.0102, -0.0738,
        -0.0316,  0.0191, -0.0248,  0.0137, -0.0112, -0.0066,  0.0000,  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([-0.0779,  0.1605, -0.0171, -0.0195,  0.1480, -0.0225,  0.0168, -0.0203,
         0.1510,  0.0289, -0.0067,  0.0088,  0.0195, -0.0061,  0.0551,  0.1620,
        -0.0042, -0.0332,  0.0112,  0.0195,  0.0098, -0.0015,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.0834e-03, -1.3383e-01,  1.1742e-02,  1.9900e-02, -1.6339e-01,
         1.6947e-02, -1.1403e-02,  2.4364e-02, -1.0392e-01, -2.0010e-02,
        -4.3982e-03, -1.1465e-02,  4.9079e-03, -1.5799e-01,  6.0474e-03,
        -4.9064e-02, -2.8574e-04,  1.7473e-03, -1.4818e-01, -2.3522e-02,
        -1.6749e-02,  1.1264e-02,  5.4308e-03,  6.9950e-03, -1.7905e-04,
        -8.8145e-03,  7.9955e-05, -2.0448e-02, -5.6457e-03, -3.1919e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0188e-01, -1.4327e-01,  9.6325e-03,  1.8905e-02, -1.8484e-01,
         1.2041e-02, -1.9350e-02,  1.2659e-02, -1.5485e-01, -2.6550e-02,
        -1.0259e-03, -7.5907e-03, -8.1411e-03, -1.3458e-01, -1.1392e-03,
        -6.0459e-02, -2.2187e-03, -2.0291e-02,  7.8392e-03, -2.0038e-02,
        -3.2800e-02, -1.8595e-02, -1.2087e-04,  1.1877e-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([ 0.1357, -0.2502,  0.0138,  0.0459, -0.2663, -0.0086, -0.0159, -0.0064,
         0.0073, -0.0194, -0.0917, -0.0284, -0.0983, -0.0048, -0.0011, -0.0062,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  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([-0.1214,  0.2576, -0.0405, -0.0611,  0.2559,  0.0097,  0.0099,  0.0112,
        -0.0064, -0.0035,  0.0742, -0.0162,  0.1143,  0.0053, -0.0072,  0.0056,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  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([ 0.2180, -0.2350,  0.0036,  0.0350, -0.2266, -0.0031, -0.0156, -0.0111,
         0.0011, -0.0054, -0.0647, -0.0027, -0.0858, -0.0037,  0.0529, -0.0237,
         0.0108,  0.0010,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  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([-0.2453,  0.1950, -0.0202, -0.0379,  0.0120,  0.0059,  0.0817,  0.0023,
         0.0304, -0.0140,  0.0008,  0.0443, -0.0256,  0.0092, -0.0084, -0.0006,
         0.0210, -0.0070, -0.0100,  0.0642, -0.0027,  0.0394,  0.0092,  0.0631,
         0.0389,  0.0040,  0.0034,  0.0035,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5064e-01,  2.1109e-01, -6.1232e-03, -3.3856e-02,  3.9493e-03,
         5.0205e-03,  6.4604e-02,  3.1690e-03,  2.4407e-02, -6.2652e-03,
         2.1224e-03,  5.8973e-02, -3.1152e-02,  8.1768e-03,  8.3155e-05,
        -1.5846e-02,  2.2639e-02,  1.6771e-02, -1.1138e-02,  5.4062e-02,
         5.4039e-03, -4.8022e-03,  2.1124e-02,  1.9393e-03,  1.2811e-02,
         1.6598e-02,  4.7294e-03,  3.1311e-02, -1.6598e-03,  3.6135e-02,
        -4.0062e-02,  5.7374e-02,  1.6674e-02,  8.4532e-03, -1.0833e-02],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0774, -0.2248,  0.0374,  0.0394, -0.0227, -0.0133, -0.0580, -0.0060,
        -0.0188, -0.0132, -0.0098, -0.0530,  0.0107, -0.0157,  0.0130,  0.0065,
        -0.0370, -0.0074, -0.0052, -0.2256,  0.0017, -0.0004, -0.0340, -0.0100,
         0.0067, -0.0081, -0.0127, -0.0027, -0.0287,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.3074e-03, -2.1864e-01,  1.5807e-02,  2.1539e-02, -2.0750e-01,
         7.1825e-03, -1.1258e-02, -3.0832e-02,  5.0266e-05, -8.0324e-02,
        -2.2682e-02,  2.3615e-03, -1.9868e-02, -1.7083e-01,  1.7649e-03,
        -7.2142e-02, -3.8834e-03,  3.0490e-02, -3.6696e-02, -2.9145e-02,
        -4.4105e-03, -5.2930e-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([-0.0729, -0.2449,  0.0152,  0.0418, -0.2141,  0.0119, -0.0169, -0.0371,
        -0.0093, -0.0765, -0.0258,  0.0081, -0.0160, -0.0653,  0.0113, -0.0869,
        -0.0306,  0.0064,  0.0055, -0.0036,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #100: [tensor([ 1.1201e-04, -2.4600e-01,  5.5679e-04,  2.4730e-03, -4.1209e-03,
         4.0177e-02, -2.3397e-01,  2.1868e-03,  4.9642e-03,  7.5633e-04,
        -1.0511e-02, -3.7480e-04, -4.1623e-03,  1.3484e-02,  1.4415e-04,
         1.9897e-03,  1.8365e-03,  2.8364e-03, -6.4730e-02,  6.6055e-04,
        -6.0162e-02,  8.9876e-03,  3.7869e-02, -2.3217e-01, -1.0938e-02,
         1.0111e-02, -2.9001e-03,  8.1352e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.2806e-01, -2.3533e-01,  5.9540e-03,  3.1750e-03, -9.5396e-03,
         3.4904e-02, -2.0984e-01,  2.4315e-03, -2.1939e-04, -9.2850e-03,
        -1.5672e-02, -1.2669e-02, -1.7241e-02,  2.5264e-03,  4.6945e-04,
        -4.0822e-03, -9.7570e-04,  1.6022e-03, -2.4594e-01, -7.7535e-04,
        -5.0033e-02, -1.3603e-03,  6.8710e-03,  1.0428e-03,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6242e-02, -2.4367e-01, -5.5849e-03,  1.6292e-04,  2.9937e-03,
         4.2730e-02, -2.2466e-01,  7.3106e-04,  5.2394e-03, -2.5247e-03,
        -1.5753e-02, -4.7212e-03, -9.6191e-03,  2.7074e-03,  7.0822e-03,
        -4.8690e-03,  4.4834e-03,  1.2022e-03, -6.3576e-02, -3.5401e-03,
        -6.2568e-02, -6.8029e-03,  4.0428e-03, -2.2831e-01,  1.1709e-02,
        -1.5007e-02,  2.1760e-03, -7.2903e-03], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.2025e-02, -2.2440e-01,  1.0573e-02, -1.0053e-02, -3.5054e-03,
        -4.6761e-02, -1.8382e-04, -3.8715e-03, -1.9985e-01,  5.4745e-04,
         5.3317e-03, -1.7860e-01, -1.1717e-03, -6.1951e-02,  1.1282e-02,
        -1.4718e-02,  3.7678e-03,  3.1746e-03, -1.7807e-01, -1.7220e-04,
        -2.8755e-03,  7.1167e-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([ 0.0407, -0.1766,  0.0028, -0.0086, -0.0044, -0.0481,  0.0058,  0.0005,
        -0.1862, -0.0035, -0.0048, -0.1860, -0.0047, -0.0534,  0.0011, -0.0037,
        -0.0515,  0.0077, -0.0032, -0.1931,  0.0040,  0.0002, -0.0095,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-0.0490, -0.2058,  0.0086, -0.0105, -0.0014, -0.0472,  0.0019,  0.0006,
        -0.1845,  0.0080, -0.0039, -0.1983,  0.0026, -0.0496, -0.0014,  0.0180,
        -0.1906, -0.0110, -0.0030, -0.0034,  0.0006,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.1346,  0.2623, -0.0122, -0.0046, -0.0074,  0.0491, -0.0186, -0.0412,
        -0.0027, -0.0008,  0.2568, -0.0116,  0.0111, -0.0057, -0.0093,  0.0851,
        -0.0157,  0.0573, -0.0031,  0.0048, -0.0059,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.0420,  0.3029, -0.0039, -0.0052,  0.0035,  0.0628, -0.0193, -0.0301,
        -0.0037, -0.0023,  0.2526, -0.0053,  0.0214,  0.0018, -0.0072,  0.0733,
        -0.0060,  0.0200,  0.0783,  0.0219, -0.0075,  0.0226, -0.0054, -0.0007,
         0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 0.2008,  0.2603, -0.0159, -0.0015,  0.0020,  0.0518, -0.0053, -0.0375,
         0.0039,  0.0046,  0.2906, -0.0017,  0.0209,  0.0052,  0.0025,  0.0587,
        -0.0030, -0.0098, -0.0125, -0.0036, -0.0079,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2256e-02, -2.2986e-01, -3.5652e-03, -9.0520e-03, -2.0629e-03,
         1.0308e-02,  2.2965e-02,  8.3747e-04,  4.4042e-03, -2.3618e-01,
        -5.9515e-03, -1.4402e-02, -2.9239e-02,  3.1646e-02, -3.3202e-03,
         3.7324e-03, -5.6552e-02,  3.5278e-03, -5.7342e-02, -2.4176e-01,
         1.6021e-04, -8.7196e-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([ 0.0566, -0.3005,  0.0032, -0.0026, -0.0061,  0.0170,  0.0405,  0.0015,
        -0.0019, -0.2792, -0.0059, -0.0071, -0.0262,  0.0435, -0.0136,  0.0027,
        -0.0760, -0.0069, -0.0790, -0.0253,  0.0046,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.8231e-03,  3.0445e-01,  1.0145e-02,  1.0121e-02, -9.9616e-04,
        -1.9242e-02, -4.6289e-02,  5.6099e-03,  7.5881e-05,  3.0492e-01,
         1.0488e-02,  1.6291e-02,  2.9588e-02, -3.8853e-02,  1.4976e-02,
         6.9959e-04,  7.8657e-02, -8.9778e-03,  6.8906e-02,  3.4751e-03,
        -1.7412e-02,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #150: [tensor([ 1.7178e-02, -1.5907e-01, -6.9625e-04,  1.0928e-02, -1.5685e-01,
         2.0476e-03, -3.4216e-03, -5.3534e-04, -1.6814e-02, -2.1175e-04,
         3.6901e-04, -1.2038e-02, -1.3308e-01,  2.7099e-04, -7.2823e-04,
         5.8069e-05,  1.5043e-03,  1.6757e-03,  3.2994e-04, -1.4910e-01,
        -3.4403e-03,  1.3734e-03,  5.3414e-04, -1.5518e-01,  2.9342e-03,
         1.0979e-02, -1.4174e-01,  1.3999e-03, -3.3738e-03, -3.8369e-04,
        -4.3485e-03, -5.1702e-03,  1.8273e-03, -4.0461e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.3447e-02, -1.8244e-01, -3.4004e-03,  1.1789e-02, -1.4715e-01,
         4.5375e-03, -3.2874e-03, -2.2720e-03, -1.6869e-02, -1.0461e-03,
         8.6437e-05, -1.0849e-02, -1.3374e-01,  3.6564e-04, -3.9819e-03,
        -8.5268e-04,  1.0132e-04,  5.6861e-04,  6.7396e-04, -1.2354e-01,
        -2.5041e-03,  1.1213e-03, -1.1418e-03, -1.4273e-01,  1.3525e-03,
        -1.3472e-03, -9.5599e-03,  1.0576e-03,  7.5338e-04,  1.1159e-02,
        -1.6336e-01,  4.6333e-04,  1.2869e-03,  1.1780e-03,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 9.3927e-03, -1.8049e-01,  1.4590e-03,  1.0124e-02, -1.4375e-01,
         4.2877e-03, -3.3851e-03, -2.8836e-03, -1.5285e-02, -1.3421e-03,
         6.4813e-04, -7.1166e-03, -1.4241e-01,  2.2274e-04, -3.8481e-03,
         1.5767e-03,  2.0320e-03, -5.4293e-04,  7.9205e-05, -1.4228e-01,
        -4.2010e-03, -6.0623e-04,  1.1244e-03, -1.4717e-01,  2.2958e-03,
         1.3630e-02, -1.3176e-01,  8.3030e-04, -2.1411e-03, -1.9471e-03,
        -2.3758e-03, -1.3757e-02, -2.8135e-03, -9.4225e-05,  2.0909e-03,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-0.0445, -0.1842,  0.0016,  0.0111, -0.1526,  0.0028, -0.0031, -0.0028,
        -0.0205, -0.0009, -0.0010, -0.0108, -0.1809,  0.0012, -0.0016,  0.0003,
         0.0017,  0.0002, -0.0007, -0.1496, -0.0061, -0.0008,  0.0019, -0.0376,
         0.0003, -0.0219, -0.1522, -0.0046,  0.0014, -0.0010,  0.0002,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-7.4141e-03, -1.7369e-01,  9.7141e-04,  1.3800e-02, -1.8746e-01,
         3.6197e-03, -3.7852e-03, -4.1124e-03, -2.1867e-02, -1.7412e-03,
         1.4701e-03, -1.3538e-02, -1.4711e-01, -9.4031e-04, -2.9186e-03,
        -1.6916e-03,  1.4741e-03,  7.4818e-04, -8.5639e-04, -1.3462e-01,
        -4.1994e-03,  7.9851e-06,  5.7851e-04, -2.1723e-01, -1.9927e-03,
        -3.1904e-02,  2.8267e-04,  7.9864e-03, -8.5428e-03,  1.2055e-03,
         2.2371e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1920e-02, -1.2110e-01, -1.1184e-03,  1.0775e-02, -1.4387e-01,
         4.2833e-03, -3.6459e-03, -3.3377e-03, -1.4724e-02,  2.1689e-04,
        -1.0111e-03, -1.2241e-02, -1.0797e-01, -8.8376e-04, -7.7908e-04,
        -5.4356e-04,  1.1671e-03,  1.1264e-03, -3.7519e-04, -1.1949e-01,
        -4.0215e-03,  1.2658e-04,  9.7759e-04, -1.4525e-01,  2.6882e-03,
        -1.5402e-02, -1.2816e-01, -1.8547e-03, -1.1257e-03,  1.3763e-03,
        -1.1746e-03,  2.5159e-03, -1.2895e-01, -1.0520e-03, -4.0323e-03,
        -7.1443e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.0128e-02, -1.9185e-01,  7.0518e-04,  1.5700e-02, -1.8425e-01,
         4.6923e-03, -3.1504e-03, -5.2814e-03, -1.5603e-02, -1.6328e-03,
         1.5098e-04, -9.2860e-03, -1.5206e-01, -7.7912e-04, -1.8527e-03,
         1.5700e-03, -4.4791e-04,  1.2194e-03,  1.6485e-03, -1.5057e-01,
        -3.6819e-03,  1.1338e-03, -3.3810e-03, -4.1035e-02,  1.7884e-03,
        -2.3963e-02, -2.7550e-03, -1.6442e-01, -3.0840e-03, -1.5988e-03,
        -5.7887e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1207e-02, -1.6522e-01,  3.2834e-04,  1.3304e-02, -1.5738e-01,
         3.1225e-03, -4.6446e-03, -3.1703e-03, -2.0719e-02, -2.6816e-03,
        -1.3818e-03, -1.2384e-02, -1.7771e-01,  3.0106e-04,  1.4100e-03,
        -1.5978e-04,  9.2169e-04,  2.4850e-03, -1.1576e-04, -1.4917e-01,
        -7.7451e-03,  9.1176e-04,  2.4764e-04, -1.8121e-01, -5.3538e-06,
        -3.3525e-02, -3.3564e-03, -5.9386e-04,  7.8097e-03, -6.1797e-03,
         3.2483e-04, -2.9031e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5095e-02, -2.7738e-01,  1.9981e-03, -2.9017e-02, -1.0126e-03,
        -2.6267e-02,  2.1359e-04, -1.1577e-02, -2.3292e-01, -5.0209e-03,
         5.3343e-03, -1.1363e-03,  2.1631e-03, -1.0400e-03, -5.0240e-02,
         9.2436e-04, -3.7188e-02, -4.2852e-03, -2.8604e-01, -8.8776e-03,
         1.5845e-03, -6.9192e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.0007, -0.2658,  0.0035, -0.0309, -0.0004, -0.0433,  0.0028, -0.0058,
        -0.2877, -0.0014,  0.0073, -0.0017,  0.0010,  0.0026, -0.2903,  0.0011,
        -0.0445,  0.0012,  0.0035,  0.0022,  0.0021,  0.0000,  0.0000,  0.0000,
         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([ 0.0016, -0.4349,  0.0029, -0.0491,  0.0009, -0.0514, -0.0017, -0.0117,
        -0.3199, -0.0041,  0.0069, -0.0073, -0.0011, -0.0013, -0.0664, -0.0030,
        -0.0095, -0.0061, -0.0147,  0.0031,  0.0022,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  0.0000,  0.0000], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.6402e-02,  3.8238e-01,  2.8717e-04, -2.0889e-03,  1.8851e-02,
        -4.7221e-04,  1.2019e-02,  4.2475e-03,  3.6242e-01, -1.5023e-03,
        -7.7699e-03, -1.5898e-03,  4.6915e-03, -3.1679e-04,  6.1160e-02,
         1.5659e-03,  7.7048e-04,  6.8901e-02, -2.0201e-03, -2.0564e-02,
         7.5261e-03,  3.0198e-03, -8.0185e-03,  3.0562e-03,  2.2785e-03,
         6.0762e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #200: [tensor([ 0.0201,  0.1923,  0.0163,  0.0029, -0.0013, -0.0425,  0.0003, -0.0136,
         0.0037,  0.0027,  0.0003,  0.0197,  0.0064,  0.1268,  0.0035,  0.0005,
        -0.0011,  0.2364,  0.0220,  0.0153,  0.1552,  0.0058,  0.0039,  0.0256,
         0.0062,  0.0104,  0.0160,  0.0192,  0.0052,  0.0089, -0.0011,  0.0150,
         0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000,  0.0000],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.4381e-02,  2.5540e-02,  4.1318e-03, -1.3429e-03, -4.2850e-03,
        -1.6597e-02, -1.1973e-03, -3.3687e-03, -4.7400e-03, -2.2150e-04,
        -1.1036e-03,  4.9703e-03,  2.4462e-03,  4.1919e-02, -3.0961e-03,
        -1.6709e-05, -5.5702e-04,  8.1179e-01,  7.6402e-03,  3.4933e-02,
         1.8869e-04,  6.0937e-03,  2.8356e-03,  4.4505e-03, -2.1866e-04,
        -1.9371e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.4401e-02,  8.7172e-02,  1.3275e-02,  6.6895e-03,  6.2601e-04,
        -3.3660e-02,  6.2180e-03, -3.0827e-03, -2.6121e-03, -1.3521e-03,
         1.0025e-04,  1.3894e-02,  3.8884e-03,  1.0042e-01,  1.7906e-03,
         5.7471e-03,  1.8582e-04,  1.6343e-01,  2.4394e-02,  9.6296e-03,
        -1.1438e-05, -3.1156e-02,  5.5147e-03, -9.5182e-03, -1.8410e-04,
         1.2278e-01,  1.3892e-01,  5.4338e-03, -3.0855e-03,  1.5106e-01,
         6.6440e-03,  4.6140e-04, -2.6628e-03,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.6706e-02,  2.6304e-03, -3.4168e-02,  1.7505e-03, -6.2658e-03,
        -8.3947e-01, -4.1106e-03,  7.3156e-06, -3.1901e-07, -4.2882e-02,
         4.5343e-04, -4.3607e-02,  4.5150e-03,  1.9896e-03,  7.1559e-04,
         3.2176e-04,  4.1042e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.7542e-02,  2.4171e-03, -5.0798e-02,  7.9002e-04, -5.1765e-03,
        -7.9738e-01, -5.4908e-03,  1.5311e-03, -8.3854e-04, -6.3380e-02,
         7.2383e-05, -5.1023e-02,  5.0711e-04,  5.4251e-04,  2.5104e-03,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.0913e-03, -4.9392e-04, -5.6815e-02, -3.8868e-04, -5.8974e-03,
        -8.0988e-01, -4.0051e-03,  5.6891e-05,  6.7126e-04, -5.1031e-02,
        -1.8995e-03, -6.0821e-02, -8.2449e-04, -8.2135e-04, -3.0521e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.6046e-03, -2.5709e-01, -3.9278e-04, -2.6363e-04,  5.0255e-03,
        -1.9667e-01, -1.2572e-04, -9.4259e-04, -4.3155e-04, -1.9821e-01,
        -5.6868e-04,  4.4910e-03, -1.2972e-01,  1.0464e-04, -1.6203e-03,
        -4.7655e-03, -2.3722e-04, -1.7474e-04,  4.3074e-03,  1.0774e-03,
        -2.7622e-03, -4.2450e-04, -1.7088e-01, -2.8690e-04, -2.9601e-04,
        -1.8186e-04,  1.2716e-03,  7.0414e-04,  1.5670e-03, -1.7391e-03,
         1.5702e-03, -1.7353e-03, -8.0273e-04,  2.3193e-04,  1.0753e-04,
         1.8056e-03,  5.7055e-04, -2.2287e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.9925e-03, -2.3263e-01, -1.0084e-03, -1.0555e-03,  4.3389e-03,
        -2.2361e-01,  4.7117e-04, -1.8119e-03, -1.1383e-03, -2.0165e-01,
         7.9785e-04,  5.1440e-03, -9.0708e-02, -5.8896e-04, -1.5889e-03,
        -3.2500e-03,  3.2616e-04, -3.4695e-04,  4.1502e-03,  8.1912e-04,
        -2.4162e-03, -3.6490e-04, -2.0096e-01,  9.4670e-04,  3.9167e-05,
        -1.2294e-03,  8.0549e-04,  6.3583e-04,  3.8420e-04, -3.0681e-03,
         1.3401e-03, -1.6292e-03,  4.4215e-04, -2.0192e-04,  1.2827e-03,
         1.4686e-04, -3.3208e-04,  1.3180e-03, -3.7735e-03,  2.4622e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3302e-03, -1.8611e-01,  2.7645e-04,  2.0554e-04,  4.5926e-03,
        -2.2618e-01,  2.7066e-04, -2.3972e-03, -1.1757e-03, -1.9553e-01,
         4.4971e-04,  3.8901e-03, -1.1917e-01, -4.6076e-04, -4.0941e-04,
        -3.2971e-03,  6.6455e-04,  9.6406e-05,  4.1925e-03,  3.7920e-04,
        -2.1983e-03, -2.8456e-04, -2.3002e-01, -7.0236e-04,  5.0324e-04,
         1.3829e-04,  1.0339e-03,  1.0695e-03,  4.2436e-04, -2.4100e-03,
         1.3677e-03, -1.3421e-03,  3.2900e-04, -4.8809e-04, -7.7578e-05,
        -9.9751e-04,  1.1884e-04,  1.2241e-03, -1.8762e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.4120e-02, -2.0061e-01,  8.0859e-04,  5.9063e-04,  7.6980e-03,
        -2.1425e-01, -4.5176e-05, -5.5552e-04, -1.8295e-03, -2.0120e-01,
         2.7200e-04,  6.3351e-04, -5.5444e-04,  1.5701e-03, -1.0653e-02,
        -1.7972e-01,  4.5571e-04,  9.2032e-03, -1.5461e-01,  2.4725e-04,
        -2.1406e-04, -1.5188e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.2910e-03, -2.1931e-01,  6.5589e-04,  2.5149e-04,  7.7493e-03,
        -1.8866e-01,  4.0619e-04, -7.9976e-04, -4.3830e-04, -2.0331e-01,
         2.7732e-04,  2.4259e-04,  2.0456e-04, -2.0877e-01,  3.2882e-04,
         7.5502e-03,  1.8516e-03, -4.9374e-04, -7.5853e-04, -1.2556e-04,
        -6.4404e-03, -2.7430e-03, -1.3914e-01,  1.2583e-03, -5.9679e-04,
         6.6247e-04,  6.7862e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.7872e-03, -2.1520e-01,  1.9572e-03,  1.1117e-03,  8.3164e-03,
        -2.0278e-01,  3.8204e-04, -4.1389e-05, -2.3153e-03, -2.2906e-01,
         7.1979e-04,  3.7524e-04,  3.4672e-04,  5.3998e-05,  1.6845e-03,
         1.1466e-03,  8.9183e-03, -1.4109e-01,  6.9707e-04,  3.9617e-04,
        -9.0252e-03, -1.6948e-03, -1.6404e-01,  1.0701e-03,  2.4794e-03,
         3.1059e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #250: [tensor([-1.1546e-02, -8.0217e-04, -2.4944e-04,  2.9675e-03, -1.9745e-03,
        -2.3695e-03, -1.0709e-03,  9.3049e-01, -1.4692e-04,  1.8957e-03,
        -4.7391e-04,  3.0655e-04, -9.1054e-04,  4.2348e-02,  1.1035e-03,
         7.5738e-04, -4.8513e-04, -1.0709e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.5674e-03,  8.8878e-04,  8.7821e-05,  2.6569e-03, -2.0513e-03,
        -3.5127e-04, -2.4506e-04,  9.7035e-01, -1.9470e-03,  3.5296e-04,
        -4.5751e-04, -1.5096e-03,  7.5031e-04, -1.2650e-03, -1.2343e-03,
         3.5451e-03,  4.4264e-03, -7.4551e-04, -5.7229e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.6322e-03, -9.7548e-02,  2.5857e-04, -4.8128e-04, -1.8020e-01,
         1.6847e-03, -2.7468e-04, -1.1713e-04, -2.0633e-01,  1.4758e-03,
         3.9547e-04,  1.1560e-04, -2.6253e-01,  4.4637e-04, -2.3395e-04,
         3.0269e-04,  1.0619e-03,  2.6556e-04, -2.4042e-01,  6.3025e-04,
         5.8540e-04,  7.9318e-06,  9.9251e-04, -1.3618e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.6860e-03, -1.6039e-01,  4.1409e-04, -2.3906e-03, -2.3855e-01,
         1.4197e-03, -7.5749e-05,  6.7168e-04, -3.2477e-01,  4.6681e-04,
        -3.7815e-04, -5.5637e-05, -2.5567e-01,  4.5408e-04, -8.1021e-04,
         2.5589e-05, -1.0549e-03,  1.3138e-03,  2.6975e-06, -3.9378e-04,
        -1.0106e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.4980e-03, -2.6580e-01,  8.1950e-04, -2.7084e-03, -1.8781e-01,
         1.4499e-03,  4.1371e-04,  1.5693e-03, -1.2757e-01,  1.6396e-04,
        -1.9225e-04, -9.2996e-04, -2.0633e-01,  7.9941e-04, -3.9618e-03,
        -4.9491e-05,  1.6789e-03, -1.9048e-01,  5.9509e-04,  5.9205e-04,
         5.8715e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.1349e-02,  6.9312e-03, -5.8248e-04,  1.7028e-03,  5.4895e-03,
         2.5237e-01,  7.8141e-03,  6.3382e-03, -6.4264e-03,  1.8138e-01,
         1.9463e-03,  2.2114e-05, -5.1098e-03,  1.5088e-04,  8.5529e-03,
         3.2198e-03, -3.7130e-03,  1.1095e-02,  1.8335e-03,  3.5840e-03,
         5.5120e-03, -2.4327e-02,  2.8017e-04,  1.3053e-02, -2.7579e-03,
        -4.7427e-03,  3.1717e-03,  5.5514e-03, -1.3353e-04,  2.9850e-03,
         8.5103e-03,  3.9107e-01, -1.0241e-03,  7.7206e-04, -5.4012e-03,
         1.0849e-03,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.2654e-03,  2.7649e-03,  3.3455e-03, -4.5750e-03,  5.7999e-03,
         2.4142e-01,  7.7050e-03,  2.8128e-03, -9.3013e-03,  2.9516e-01,
         5.6596e-03,  9.1673e-04, -2.1965e-03,  7.3942e-03,  1.2750e-02,
         7.9921e-03, -1.3770e-03,  5.5295e-03, -1.9160e-03, -1.2129e-03,
         1.2574e-02, -2.2097e-02,  1.9011e-03,  1.0445e-02,  2.6481e-04,
        -2.7121e-03,  7.1810e-03,  4.6000e-03,  2.4511e-01, -6.1370e-03,
         1.2855e-02,  1.3532e-02,  3.2139e-03, -1.2626e-02,  1.5709e-02,
        -1.1210e-03, -8.2581e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3344e-02,  7.0269e-04, -1.4286e-03, -1.2256e-03,  5.1064e-03,
         3.1144e-01,  4.5093e-03,  7.5686e-03, -6.4275e-03,  2.4187e-01,
        -2.0210e-03,  3.3971e-03, -2.1530e-03, -5.4805e-03,  6.3435e-03,
        -1.1474e-03,  7.5035e-04,  7.3649e-03, -1.0643e-03,  4.5277e-03,
         4.0556e-03, -2.4228e-02, -2.2132e-03,  9.1297e-03, -5.0004e-03,
        -3.7005e-04, -3.0080e-05,  8.2168e-04,  2.4584e-01,  1.8192e-02,
        -2.4658e-02,  2.4791e-02, -3.4477e-03,  6.9541e-03, -2.3989e-03,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.5423e-03, -3.0805e-05, -1.0043e-04,  4.2352e-04, -3.0499e-04,
        -1.7722e-03, -3.0301e-01,  7.4811e-04, -9.0253e-04, -6.0682e-04,
        -4.7399e-04, -3.5522e-03, -2.5193e-04,  1.8062e-04, -1.7782e-01,
        -3.8888e-04, -3.2650e-04,  3.7738e-04, -3.0309e-01,  3.7396e-04,
         1.5605e-04, -6.5401e-04, -1.8927e-02, -1.6711e-03, -2.5679e-04,
        -1.8913e-03,  6.2789e-04, -8.8482e-04, -1.7595e-01,  3.0261e-04,
        -1.9351e-03, -4.3833e-04,  2.9362e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-4.6839e-03, -1.4611e-03, -4.9971e-04,  8.3730e-04,  6.8731e-04,
        -1.7425e-03, -4.0085e-01,  3.7182e-04, -8.1586e-04, -1.4418e-04,
        -3.8285e-04, -4.9462e-03,  5.0262e-05,  2.8739e-04, -3.0081e-01,
         7.7208e-04, -3.1086e-04, -5.4596e-04, -2.7237e-01,  7.3408e-05,
        -5.5276e-03, -6.7240e-04, -3.2367e-04, -1.2919e-05,  4.6222e-04,
        -3.6150e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.4316e-03, -6.2732e-04,  5.0945e-06, -4.9378e-04, -1.0870e-04,
        -3.0974e-03, -3.8001e-01,  4.8786e-04, -1.2694e-04,  2.7391e-04,
         7.7207e-05, -2.8270e-03,  4.3288e-04,  8.8358e-04, -3.2481e-01,
        -7.6473e-04, -5.3173e-04, -8.5508e-04, -2.5933e-01,  3.0613e-04,
        -6.1362e-04,  8.2848e-04, -1.0625e-04,  4.3634e-04,  4.0279e-03,
        -1.0220e-02,  8.9343e-04, -3.9456e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6460e-02, -3.1880e-01, -1.0320e-03, -3.6000e-06,  6.5288e-05,
        -2.4463e-04,  1.3602e-03, -1.9734e-01,  1.8006e-04, -1.0428e-03,
         4.9102e-04, -5.1372e-04,  3.7493e-03, -1.1852e-01,  4.0787e-04,
        -6.6796e-05,  7.9448e-04, -1.1136e-02,  8.5493e-04, -2.7672e-03,
         3.6486e-05, -3.1873e-01,  6.8495e-04, -2.6449e-03, -1.3656e-03,
        -4.1698e-04,  7.6635e-05, -2.1695e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  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([-1.1705e-02, -1.2074e-03,  3.8967e-04, -6.3058e-04,  7.8956e-04,
        -5.4442e-01, -8.3162e-04,  1.3264e-04,  1.7024e-03,  1.5557e-04,
         2.3123e-05,  5.5872e-04, -1.1069e-03,  8.7477e-04, -2.5675e-04,
         1.1223e-04, -4.3870e-04,  7.7132e-04, -1.9805e-02,  5.3301e-04,
         5.4557e-03, -4.0619e-01,  8.5169e-04, -7.9743e-04,  1.2085e-04,
         1.3836e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.9772e-03, -1.1107e-01, -5.6750e-05, -8.5023e-04, -1.7226e-01,
         6.6684e-04, -2.7693e-04, -1.0459e-04, -2.1990e-04, -1.4788e-04,
         6.9479e-04,  4.0775e-04, -1.3989e-04, -2.2033e-05, -1.5189e-05,
        -5.0077e-04,  1.3688e-04, -4.8620e-03,  2.9696e-04, -3.0052e-04,
         3.9515e-05, -4.2934e-02, -2.7753e-04,  4.2754e-04, -1.2475e-03,
        -1.2530e-01,  2.7881e-04,  1.6779e-04,  2.4284e-04,  7.6047e-04,
        -1.7281e-04,  1.0152e-05,  2.0744e-05, -9.3612e-05, -1.2059e-01,
         1.2805e-04, -3.8721e-04, -8.7876e-02,  4.2163e-04, -2.6725e-04,
         1.9273e-04, -4.6289e-04,  2.3990e-04,  8.4435e-06,  4.6914e-04,
         2.8218e-04,  1.0571e-04,  1.2464e-04, -9.8328e-04,  5.3539e-05,
        -1.1978e-02,  1.8964e-04, -3.3903e-04, -1.5178e-01, -1.9965e-04,
         6.0177e-04, -5.0574e-04, -1.4477e-01,  3.4479e-04,  1.2629e-04,
         1.0572e-04,  5.4789e-04, -2.1287e-04,  3.7372e-04,  3.7316e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0997e-03, -3.4276e-01, -5.6294e-04, -1.4188e-03, -1.8818e-01,
         2.7863e-04, -9.2567e-04, -2.1765e-04, -1.1311e-04,  1.3704e-04,
         3.2502e-04,  3.9521e-04, -4.7807e-04, -4.0844e-04, -3.0416e-04,
        -9.1923e-04, -4.7674e-04, -2.0594e-02, -5.9592e-05, -3.6846e-04,
        -7.5891e-04, -2.6633e-01, -8.1758e-04, -1.5125e-04, -2.3889e-03,
        -1.4624e-01, -3.3430e-05,  2.1214e-04, -3.6690e-04,  6.7107e-04,
        -2.1439e-04, -3.5188e-04,  3.9470e-05,  7.8625e-05, -2.0163e-02,
        -6.9555e-05, -9.4108e-05, -5.3180e-04, -2.0030e-05, -4.4402e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.8497e-03, -2.8109e-01, -2.4143e-04, -1.2212e-03, -1.1306e-01,
         2.5001e-04, -6.1767e-04, -1.8895e-04,  7.2372e-05, -1.0054e-04,
         5.3765e-04,  6.5719e-04, -8.2092e-04, -3.4341e-04, -1.7405e-04,
        -6.2303e-04, -7.0040e-04, -6.9535e-03,  1.4081e-04, -1.2828e-05,
        -6.4475e-04, -1.4166e-01, -8.0233e-04, -3.0597e-04, -2.1132e-03,
        -3.7130e-01, -1.1541e-04, -1.4634e-04, -7.0536e-05,  5.4648e-04,
        -2.4894e-04, -7.5945e-05, -8.1452e-05, -1.7364e-04, -3.0392e-02,
        -4.8419e-04, -7.8584e-04,  1.2241e-04, -1.1719e-03, -7.4944e-04,
        -3.6917e-02, -1.3757e-04, -2.9300e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6427e-02, -6.0169e-04, -2.1380e-04, -8.5576e-04, -1.5896e-02,
         2.2757e-03,  4.3232e-04, -7.8898e-04,  5.1261e-04, -5.9062e-01,
         1.6529e-03,  2.9693e-04, -6.9282e-04, -2.2338e-04,  1.2712e-03,
         7.5111e-04,  4.3518e-04, -3.6424e-01,  9.1150e-04,  8.8362e-04,
         1.8782e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.6116e-03, -1.4020e-03,  1.4140e-04, -2.3566e-04, -2.1308e-02,
         2.1676e-03,  2.2748e-03, -3.5244e-04,  1.5351e-04, -3.8691e-01,
         5.8931e-04,  6.6674e-04, -2.5235e-03,  5.9444e-04, -5.7205e-01,
         1.7262e-04, -1.6752e-04,  1.9301e-03,  1.1528e-03, -1.4508e-04,
         4.4804e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.7922e-03, -1.7066e-04, -2.1807e-04,  2.6690e-04, -2.7656e-02,
        -9.0841e-05,  1.1354e-03, -9.0943e-04, -1.4263e-03, -3.5513e-01,
        -2.4350e-04, -5.2160e-05, -6.1816e-04, -1.0475e-04, -5.5918e-01,
        -1.1412e-04, -7.1162e-05,  8.5937e-04, -3.1814e-03,  5.2618e-03,
        -3.3620e-02,  2.3633e-04,  3.4231e-04, -3.2544e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.1125e-03, -1.2651e-03,  3.7340e-04,  4.0459e-04,  4.7866e-04,
        -3.8992e-03, -3.3327e-01,  1.2997e-03,  5.2748e-03, -1.2686e-03,
         7.5827e-05,  3.3077e-04, -3.3268e-02,  1.7352e-04,  2.6725e-03,
        -7.6822e-03, -1.7226e-04, -5.9648e-01,  1.4176e-03, -6.0252e-04,
         4.8044e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.9475e-03, -1.9096e-03,  8.1613e-04, -1.3086e-04,  1.0344e-03,
        -3.0758e-03, -9.2156e-01,  1.3240e-03,  3.9937e-03, -4.0698e-04,
        -6.9453e-04, -1.0535e-03, -5.8843e-02,  9.0028e-04, -5.1348e-04,
        -6.8033e-04,  1.1506e-05,  3.4729e-04, -7.5390e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.6826e-02,  1.2666e-04, -1.0426e-04, -3.1588e-05,  8.0688e-04,
        -1.3469e-03, -4.9625e-01,  5.3811e-04,  6.8767e-03, -1.3720e-04,
        -2.6471e-04, -7.4795e-05, -2.6969e-02,  9.6840e-06,  9.7120e-04,
        -4.2786e-03,  1.2531e-03, -4.4140e-01, -6.1168e-04,  8.6715e-04,
        -2.5278e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7019e-03, -1.2991e-01, -3.7833e-05, -3.9152e-04, -3.5466e-04,
         9.9379e-04, -1.8559e-01, -3.9871e-04,  1.1485e-03, -1.5398e-01,
         3.2952e-04, -6.2496e-04, -9.9140e-05, -1.1747e-01,  2.5939e-04,
         3.6884e-04, -7.1652e-04, -6.3651e-04,  5.1319e-04, -4.4636e-02,
        -3.7809e-04,  3.1177e-05,  5.8580e-04, -2.8355e-04, -1.1709e-01,
         2.6892e-04, -2.3226e-04, -2.1617e-04,  8.9968e-04, -1.1386e-01,
        -2.1155e-04, -1.3902e-04, -1.2336e-01, -2.5942e-04, -4.0692e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.2259e-02, -2.2296e-01, -4.9849e-05, -4.1739e-04, -6.1573e-04,
         7.2298e-04, -1.2770e-01, -1.6513e-04,  3.3076e-04, -1.7929e-01,
         1.4189e-04, -4.3524e-04,  3.3502e-04, -3.5591e-02,  4.7784e-04,
        -3.2229e-05, -3.1421e-05, -2.4266e-04,  5.2182e-04, -6.4400e-02,
        -3.7904e-04, -1.2628e-04,  6.2835e-05,  2.9963e-05, -9.0778e-02,
         4.8363e-05, -4.1727e-04,  1.4469e-04,  5.1944e-04, -1.2568e-01,
         3.2778e-04, -3.3535e-04,  6.5112e-04, -1.3336e-01, -1.8555e-04,
        -2.2004e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #350: [tensor([-3.6107e-04, -9.8407e-04, -9.4774e-05,  3.2504e-02,  3.2837e-04,
        -8.5370e-04,  5.6862e-04, -2.8804e-05,  9.5975e-01,  2.7928e-04,
        -3.6572e-04,  2.6519e-04,  7.2432e-04,  2.5794e-04, -7.8381e-05,
         8.9687e-04, -4.2520e-04, -3.6769e-04,  4.9870e-04,  3.7075e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.2083e-03, -2.2950e-03, -1.3090e-04,  7.0715e-02,  1.1750e-03,
        -5.6770e-04,  4.5325e-04, -1.0206e-03,  7.1518e-01, -3.5694e-04,
        -8.3380e-04, -2.9225e-04,  1.6368e-04,  9.4363e-04, -4.7313e-04,
         1.2269e-03,  2.0001e-01, -6.6463e-04, -9.6378e-05,  1.8656e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 7.8961e-03, -1.3405e-01, -7.6807e-04,  3.5760e-05,  3.4076e-03,
        -3.5164e-01,  2.4061e-04, -2.9060e-04, -2.6346e-03,  4.0332e-05,
         4.1175e-04, -7.0789e-04,  7.8901e-04, -1.4476e-02,  1.6183e-03,
        -4.9080e-04,  4.9085e-04, -3.0342e-01, -1.5176e-04, -2.2459e-04,
        -2.7096e-04,  2.9858e-04, -1.7253e-01,  1.2960e-05,  5.1127e-04,
        -8.6743e-04, -2.8407e-04,  8.1277e-04,  3.4091e-04,  2.8881e-04],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.4154e-04, -2.3996e-01, -5.1745e-04,  3.4805e-04,  2.5625e-03,
        -3.3161e-01, -1.1583e-04, -3.1921e-04, -2.0703e-03, -1.6693e-04,
         2.9915e-04, -1.0532e-03,  2.4008e-04, -7.6169e-03,  1.0845e-03,
        -2.1635e-04,  3.7732e-04, -1.9495e-01,  4.2511e-04, -2.9778e-04,
        -4.1236e-05, -1.4078e-04, -2.1334e-01, -3.8403e-04, -2.8920e-04,
         1.0281e-04, -4.5643e-04, -6.4913e-05,  1.0894e-04,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.0365e-03, -1.4587e-01, -1.9660e-05,  1.3020e-04,  3.6374e-03,
        -2.5569e-01,  9.8920e-06, -3.5061e-04, -1.8995e-03,  2.8021e-04,
         4.0523e-04, -1.4563e-03,  1.1502e-03, -6.2180e-03,  1.4999e-03,
        -2.6037e-04,  7.8891e-04, -2.7532e-01, -1.1728e-04,  6.4363e-05,
        -3.8034e-04, -1.0761e-04, -3.0075e-01, -5.1633e-04,  3.3897e-04,
         4.5413e-04, -8.0112e-05, -1.5357e-04,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 0.1270,  0.1023, -0.0381,  0.0613,  0.0008,  0.1189, -0.0173, -0.0369,
        -0.0362,  0.1850,  0.0402,  0.2284,  0.0076,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  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([ 0.2820, -0.0088, -0.0931,  0.0250,  0.0135,  0.0112, -0.0312,  0.0191,
         0.3058,  0.1230,  0.0083,  0.0441,  0.0349,  0.0000,  0.0000,  0.0000,
         0.0000,  0.0000,  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([ 4.8810e-01,  1.2699e-02, -8.0897e-02,  4.5377e-02, -4.9627e-03,
         6.8873e-02, -5.1019e-03,  1.7409e-02,  1.2707e-01, -8.4434e-02,
         6.4816e-02, -2.5773e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1369e-03,  8.4736e-04, -7.7174e-04,  1.2587e-03, -7.2979e-04,
         3.3367e-03,  8.4880e-01, -4.5019e-04, -5.0757e-03, -2.7458e-04,
        -3.1525e-04, -4.8163e-04, -2.5919e-04,  1.3197e-01, -6.4365e-04,
        -5.3320e-04,  3.1650e-04, -6.9631e-04, -1.0400e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-7.9550e-03,  1.4380e-03, -2.2284e-03,  5.9535e-04, -8.8080e-04,
         4.7764e-03,  7.5150e-01,  3.3213e-04, -5.1166e-03,  8.7523e-04,
        -1.4051e-03, -9.0716e-04, -6.9497e-05,  1.2653e-01, -6.0824e-04,
         5.1135e-04,  8.7570e-02,  1.9235e-04, -5.0209e-04,  1.4503e-03,
         1.6507e-03,  1.3947e-03,  5.6599e-05, -7.8816e-04, -6.6877e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 6.3176e-03,  7.6319e-05,  1.3143e-03, -7.2063e-04,  6.2512e-04,
        -3.1071e-03, -3.1715e-01,  5.6814e-04,  3.9636e-03,  2.2227e-04,
         5.1468e-04, -6.0064e-04,  5.3350e-04, -4.6605e-02,  2.8576e-04,
        -2.4200e-05, -6.1023e-01,  4.5037e-04,  3.8774e-03, -1.0920e-03,
         3.0008e-04, -5.8588e-04,  8.3470e-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([-1.3603e-02,  2.3878e-03, -4.0505e-03,  1.3760e-03, -2.2525e-03,
         4.3798e-02, -6.0172e-04, -1.0060e-02, -1.9947e-03,  2.1529e-03,
         4.9793e-03, -1.2420e-03, -2.2956e-03,  1.6606e-04, -1.7259e-03,
         8.0386e-01, -1.6089e-03,  9.5072e-02, -2.3643e-03,  1.7962e-03,
         7.6885e-04,  1.8478e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #400: [tensor([-9.5881e-03, -1.6799e-01,  4.0565e-04, -2.1815e-03,  5.5862e-04,
        -3.6866e-04,  8.8695e-05, -1.3453e-04,  2.2584e-04, -3.7851e-01,
         2.8821e-04,  9.0376e-05,  1.9764e-04,  4.0366e-04, -2.8608e-01,
         4.0181e-04, -2.2048e-03, -2.8656e-04,  9.6688e-04, -1.4856e-01,
        -6.2457e-05, -3.7704e-04,  2.3393e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.6638e-03, -4.2984e-01, -4.7550e-04, -1.9129e-04, -6.8041e-06,
         2.2641e-04, -1.1242e-05,  8.4746e-05,  4.0375e-04, -2.8177e-01,
         5.1593e-04,  6.7837e-05,  4.7208e-04,  7.1516e-06, -2.7289e-01,
        -5.7145e-04, -1.5882e-04, -2.5923e-04,  6.5763e-04, -4.8088e-03,
         6.9568e-04,  1.8985e-04, -3.7222e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.7964e-03,  2.1606e-01,  7.2900e-04,  2.0239e-03,  3.8341e-04,
         5.1129e-04,  7.7393e-04, -1.7837e-05, -6.4725e-04,  3.2511e-01,
        -2.0908e-04,  3.6961e-04, -2.0145e-04, -2.4512e-04, -4.9094e-04,
         4.7571e-03,  4.4166e-01,  2.0913e-04, -3.4290e-04,  1.5289e-03,
         1.1903e-04, -7.0189e-05, -3.7194e-04, -3.7627e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.2629e-04, -4.3443e-01, -2.4299e-05, -8.6178e-04,  8.4747e-04,
         7.4287e-05,  1.9125e-04, -3.8727e-04,  4.7804e-04, -1.8898e-01,
         7.5767e-04, -3.2495e-04,  1.0446e-04,  5.7874e-04, -2.0806e-01,
         2.4844e-04, -1.2800e-03, -1.1240e-04,  5.4674e-04, -9.4279e-03,
         5.2344e-04, -3.3671e-04, -1.5019e-01,  1.3239e-04, -6.3048e-05,
         2.0346e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.5007e-03,  6.7697e-02,  8.6001e-04,  1.5498e-05,  1.2840e-03,
         8.2616e-04, -3.0851e-04, -6.5211e-04,  1.6163e-03,  5.1093e-02,
        -3.5994e-04, -2.6805e-04,  2.8906e-04,  8.6335e-01, -4.8535e-04,
         3.0196e-03,  1.6701e-03,  4.2123e-04,  1.7548e-03, -9.9878e-05,
         1.4247e-03,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.9742e-03,  4.6055e-02, -2.5200e-04,  2.4245e-04,  1.4870e-03,
         5.7786e-04, -4.2726e-04, -7.7218e-06,  1.2447e-03,  7.3538e-02,
        -5.1751e-04,  9.3302e-05, -6.7262e-04,  8.6384e-01, -4.3702e-04,
         6.1602e-03, -2.6102e-04,  5.8265e-04,  1.1893e-04,  8.0270e-04,
        -2.1705e-04, -1.2836e-04, -3.6115e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1808e-03,  7.5018e-02,  1.0617e-03,  5.3291e-04,  1.1294e-03,
         7.4043e-04,  4.6721e-04,  8.8454e-04,  1.5161e-03,  1.4489e-01,
        -1.6242e-03, -2.8671e-04,  2.8290e-05,  7.6217e-01,  1.6699e-03,
         4.2040e-03,  9.4065e-04, -1.2782e-03, -3.6942e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.4318e-03,  2.0589e-04, -1.9900e-05,  3.5379e-04, -5.1697e-04,
         6.0866e-01, -2.7286e-04,  6.5409e-05,  9.1838e-06, -5.2747e-04,
         1.9666e-02,  3.1808e-05,  2.3328e-04,  1.4863e-01, -2.6489e-05,
        -1.4986e-04,  2.4370e-04, -2.0007e-03,  1.1284e-01,  3.7716e-05,
         2.2343e-04, -6.1181e-04,  1.2628e-04, -3.1796e-04, -3.4302e-05,
         3.4428e-04,  4.3824e-04,  3.8226e-02, -1.2372e-04,  8.2406e-05,
         1.2030e-04,  5.5959e-02, -6.0237e-04, -1.5074e-04, -2.5510e-04,
         1.6854e-03, -3.8401e-04, -4.2417e-05,  3.5182e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.6008e-03, -2.0511e-05,  5.3008e-06, -5.6528e-05,  4.2880e-04,
        -3.2122e-01,  1.1689e-04, -3.1426e-04, -1.0811e-04,  1.3982e-04,
        -1.1209e-02,  1.8144e-04,  3.7820e-05, -1.1900e-01,  4.7816e-04,
         1.3810e-04, -2.4567e-05,  1.4394e-03, -1.3359e-01, -2.7668e-05,
        -3.5443e-04,  5.1249e-04,  6.1397e-05, -1.9527e-04, -1.5220e-04,
         1.5885e-04,  5.6303e-05, -3.7076e-02,  1.2185e-04,  2.4925e-04,
         1.6638e-04, -1.2534e-04, -7.3999e-05, -6.9716e-02,  3.9286e-05,
         3.6387e-04, -2.8823e-01,  2.2740e-04, -1.9534e-04, -8.7374e-04,
         1.2623e-04, -7.5586e-05, -2.3576e-04,  1.1937e-04, -1.2188e-04,
         1.2497e-05, -2.5721e-03, -2.1255e-04, -2.7100e-04, -1.6116e-04,
         3.8243e-06,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.3620e-03, -3.0635e-04, -1.3157e-04,  3.0697e-04,  2.4691e-04,
        -3.5959e-01,  3.8193e-04, -8.1000e-05, -2.1284e-04,  1.2436e-04,
        -1.9599e-02,  1.8475e-04, -7.6550e-05, -1.3722e-01,  4.4531e-04,
         1.9228e-04,  2.2228e-05,  1.4050e-03, -1.5901e-01,  9.5745e-05,
        -1.4770e-05,  4.9570e-04,  4.3855e-04, -1.7943e-04,  6.8741e-05,
        -1.5423e-04,  7.8045e-05, -4.4810e-02,  5.1250e-04,  1.4683e-04,
         5.5419e-04, -2.6637e-04, -3.0569e-04, -3.6687e-02, -3.1159e-04,
         3.7893e-04, -2.3118e-01,  2.8883e-04,  1.3353e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1231e-03,  7.8799e-05, -9.0674e-05, -4.7280e-05, -6.0186e-04,
         3.2673e-01, -1.7906e-04,  2.3020e-05,  7.7241e-05, -2.2713e-04,
         1.2660e-02, -1.1738e-04,  2.9286e-04,  1.3729e-01, -7.6088e-04,
        -2.7545e-04, -8.2548e-05, -2.0443e-03,  6.6656e-02, -6.4680e-06,
         2.4409e-04, -4.6550e-04,  9.4200e-05,  3.4739e-05, -3.6381e-06,
         3.2666e-04, -1.1934e-04,  4.5302e-02, -3.5294e-04,  2.0933e-04,
         5.8425e-05, -8.9562e-05,  4.2041e-03,  3.4066e-01, -4.9759e-05,
        -1.0860e-04,  3.3472e-05, -5.0937e-04,  5.7472e-02, -1.1289e-04,
         4.1229e-05, -1.4703e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7507e-03, -7.9098e-05,  7.3975e-05, -3.2396e-04,  4.4734e-05,
        -3.1809e-01,  8.5164e-06, -4.4896e-05, -2.6992e-04, -9.2442e-05,
        -4.2324e-03,  1.8310e-04, -1.1917e-04, -1.3539e-01,  1.2626e-04,
         1.7584e-05,  2.6638e-05,  1.5317e-03, -5.3210e-02,  1.0380e-07,
        -9.6172e-05,  2.5151e-04, -1.9227e-04, -2.2850e-04,  2.3588e-06,
        -3.1208e-04, -1.7171e-04, -1.7641e-02,  1.4465e-04, -2.2632e-05,
         7.1092e-05, -1.9331e-01,  9.1131e-06, -1.2267e-03, -1.4836e-04,
        -6.4119e-04, -3.7626e-04,  3.3170e-04, -8.2837e-04, -3.5932e-02,
         1.5126e-04,  2.3285e-05, -1.2004e-04, -1.2783e-04, -8.6884e-04,
        -1.4666e-04,  1.1369e-04, -2.2760e-01, -4.0065e-05, -9.2725e-04,
        -2.1510e-04,  1.0828e-04], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #450: [tensor([ 8.9205e-03, -1.9473e-01, -7.3066e-05, -4.2739e-05,  1.0356e-04,
         3.3023e-04, -2.7860e-04,  1.5475e-04, -1.4601e-04,  5.2214e-04,
        -6.1527e-02,  1.3559e-04, -2.1917e-04,  1.6662e-03,  1.0117e-04,
        -2.6445e-04, -2.7978e-01,  7.1079e-05,  1.2142e-04,  4.1373e-04,
        -1.2945e-01,  2.5927e-04,  9.4385e-05,  1.3739e-04, -1.3809e-01,
        -1.9585e-04,  1.6116e-03,  1.1625e-04,  2.7426e-04,  4.0621e-04,
        -5.7770e-02,  1.1260e-04, -2.1209e-05, -3.6022e-04,  2.1319e-04,
         1.0384e-04, -1.2056e-01,  2.9718e-04,  7.3403e-05,  3.4883e-05,
         2.1784e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0484e-02, -3.4834e-01, -2.3091e-04, -3.3356e-04, -2.4183e-04,
         8.1173e-05, -1.9480e-04, -5.4973e-04, -1.6744e-05,  4.3257e-04,
        -1.3128e-01, -1.1015e-04, -6.2235e-05,  1.1260e-03, -3.7053e-05,
         9.3851e-05, -1.7114e-01,  5.5086e-05, -2.7147e-05,  4.3415e-04,
        -1.7171e-01, -1.3828e-04,  2.7856e-04, -1.4528e-04, -1.4946e-01,
        -1.6522e-04, -1.4191e-04,  1.3821e-05, -1.2125e-02, -3.8943e-04,
         7.1724e-05,  8.7387e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.0555e-03, -1.0975e-01,  3.9968e-04, -2.3315e-04,  4.9014e-05,
         3.1954e-04, -4.0381e-05, -2.9569e-05, -4.7871e-05,  4.1661e-04,
        -1.6116e-01,  2.7995e-05, -2.1414e-04,  9.5570e-04, -3.8221e-05,
        -1.0689e-04, -2.1269e-01,  3.5156e-04,  1.9605e-04,  7.6301e-05,
        -2.1631e-01, -1.2767e-04,  1.3315e-04,  1.2544e-04, -2.0067e-01,
         3.9373e-04, -1.1631e-04,  8.9184e-04, -2.2057e-04, -3.1411e-05,
        -8.7119e-02,  2.2399e-04, -1.3362e-05,  9.1510e-04, -2.6148e-05,
         5.9051e-05, -5.4020e-05,  2.5639e-04, -1.1516e-04,  4.4943e-05,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.1668e-04, -4.3145e-01,  7.3117e-05, -1.4457e-04,  7.7471e-05,
         2.0541e-04,  2.2341e-04,  3.3248e-04,  3.9553e-04,  3.4900e-04,
        -1.2820e-01,  2.0235e-04, -9.8873e-05,  1.6160e-03,  5.7072e-05,
         2.9753e-04, -2.0144e-01,  7.0927e-05, -4.5880e-05,  3.1028e-04,
        -1.3355e-01, -3.8719e-04, -7.8015e-05,  1.7590e-04, -9.6412e-02,
         2.3768e-04, -1.9766e-03,  2.1336e-04, -5.8561e-04, -2.5280e-04,
         3.0507e-04, -5.1662e-05, -7.0804e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0376e-03, -2.2444e-01,  7.3859e-05, -2.7299e-04,  2.9913e-04,
        -2.9151e-01,  1.3555e-04, -1.9551e-04, -3.0430e-04, -1.2050e-01,
         9.7940e-05,  2.3800e-04,  6.5463e-04, -8.4503e-03,  3.7954e-04,
         1.0430e-04, -2.7129e-04, -1.6017e-01, -3.6132e-04, -1.9008e-04,
         1.2552e-04,  4.3544e-05, -5.2812e-04,  2.4368e-05, -1.2200e-04,
        -4.7740e-04, -1.2674e-04,  3.0736e-06, -8.0358e-02,  1.6630e-04,
        -2.2241e-07,  8.2529e-06,  3.4630e-04,  1.1973e-03, -1.0614e-01,
         1.8685e-04, -4.1548e-05, -4.2153e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.4286e-03, -1.3471e-01,  2.5550e-05, -2.9092e-04,  3.4807e-04,
        -2.1439e-01,  8.8243e-05, -2.0280e-04,  5.8308e-04, -3.2578e-01,
        -9.6655e-07,  7.0968e-05,  2.2980e-03, -4.3092e-03,  3.9614e-04,
         4.7831e-05,  2.8330e-04, -2.1554e-01,  1.6980e-04, -1.2450e-04,
         2.4260e-04,  1.9771e-04, -4.6409e-02,  6.2014e-04,  2.3894e-04,
         1.8950e-04,  4.6313e-04,  7.9081e-04, -4.2434e-02, -1.6161e-04,
         2.3972e-06,  1.4889e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.3428e-04, -1.7806e-01,  2.9844e-04, -2.2728e-04,  5.7003e-04,
        -2.5807e-01,  4.5684e-05, -5.8806e-04,  6.1318e-05, -2.0632e-01,
        -1.5250e-04,  6.4601e-04,  1.1044e-03, -4.5126e-03,  4.0386e-04,
         2.8749e-04,  2.1912e-04, -1.4313e-01,  1.1654e-05,  3.5170e-04,
         2.6029e-05,  2.2225e-04, -1.8664e-01, -3.8478e-04,  3.1283e-04,
         2.3033e-03, -1.3818e-02, -1.4063e-04,  4.5541e-04, -1.1324e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.0219e-03, -3.9836e-01, -4.0083e-05, -3.9841e-04, -4.5639e-06,
        -4.0137e-04, -3.3029e-01,  3.9866e-05,  4.8678e-04, -3.2954e-04,
         7.3421e-04, -2.6817e-04, -2.9720e-04, -1.7279e-04, -1.4447e-01,
        -1.5989e-04,  1.7697e-05, -1.6428e-04, -1.1486e-04, -4.4255e-04,
         6.0944e-04, -4.2500e-06, -1.1667e-01, -3.4512e-05,  4.8543e-04,
         8.8015e-05, -3.4495e-04,  3.2886e-06,  5.2814e-04,  2.3380e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.0436e-02, -4.1166e-01,  6.2584e-04,  1.7010e-04, -1.4487e-04,
        -3.3700e-04, -2.5005e-01,  3.2127e-04,  4.1968e-04, -4.6579e-04,
         7.9548e-04, -2.3102e-04,  2.0490e-04, -2.0828e-04, -1.5127e-01,
        -6.6960e-05, -3.9607e-04,  2.1299e-04,  3.0040e-04,  2.9065e-04,
        -1.4614e-04,  4.1411e-04, -1.6847e-01,  4.2818e-04,  6.5278e-04,
        -4.1705e-04, -4.7968e-04, -1.2322e-04,  4.2801e-05,  2.2284e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 3.7880e-03, -2.1301e-01, -1.3353e-04,  2.6873e-04, -4.8491e-04,
        -5.2894e-04, -1.9589e-01,  7.9986e-04,  2.7845e-04,  3.7300e-06,
         3.9072e-04, -3.1013e-04, -7.1718e-05, -1.0557e-04, -3.5879e-01,
         1.0596e-04, -2.3243e-04,  1.3947e-04,  3.1768e-04,  8.2922e-05,
         1.9271e-04, -2.3038e-04, -2.2163e-01,  5.5390e-04,  6.0542e-04,
        -5.1718e-04, -5.8629e-05,  4.7356e-04,  1.9924e-05,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-9.0865e-04,  8.5397e-01,  6.7195e-04,  9.5032e-04, -3.3220e-04,
         5.8757e-05, -2.7639e-04,  1.8184e-03, -2.8741e-03,  5.1702e-02,
         9.3704e-06,  9.5668e-04,  2.2737e-04,  7.7767e-02, -3.9000e-04,
        -9.3818e-04,  3.8816e-03,  8.1967e-04, -5.8448e-05, -8.4743e-04,
        -5.3651e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.5002e-03, -3.9973e-01, -5.9142e-04, -3.9708e-04,  3.9621e-05,
        -1.6892e-04, -4.4239e-05, -2.2870e-03,  2.2928e-03, -5.3478e-02,
         5.3266e-05, -2.5500e-05, -8.7513e-06, -5.1675e-01,  2.2272e-04,
         1.0561e-04, -4.1052e-03,  3.6870e-04,  2.2229e-03, -1.1012e-02,
         2.1327e-04,  3.8902e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #500: [tensor([ 8.7581e-04,  2.5266e-01,  1.0422e-03, -2.6295e-04,  3.5990e-01,
         1.2444e-04,  9.7571e-04,  2.1070e-04,  1.5671e-04, -3.5880e-05,
        -1.7549e-04, -2.6543e-04,  2.2368e-04, -3.4077e-05,  1.5707e-05,
         3.8193e-01, -2.9980e-04,  7.3382e-05,  5.3174e-04, -9.2689e-05,
        -1.0924e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.3312e-03,  5.7191e-01, -1.4526e-04, -1.3977e-04,  2.6925e-01,
         5.8337e-04, -1.0182e-04, -4.6595e-04, -2.2021e-04,  2.4193e-03,
        -2.5944e-04, -7.2890e-04, -9.4091e-05, -2.0898e-04, -8.5657e-04,
         1.4617e-01, -9.9278e-05,  3.2210e-04,  1.1217e-04,  2.2525e-03,
         4.5298e-06,  1.5895e-04, -9.6695e-04,  1.3099e-04,  6.8084e-05,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.1584e-04, -2.0181e-04, -9.3880e-05,  1.4223e-04, -8.3447e-04,
         1.4149e-03, -3.3495e-05, -3.8908e-04,  9.7513e-01,  2.8329e-04,
         1.6323e-03,  4.5371e-04, -2.7779e-04,  2.8193e-04,  1.5588e-02,
        -1.6132e-03, -5.7618e-04,  1.8602e-04,  1.6189e-05, -5.3276e-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([-5.7820e-05, -2.4508e-04, -6.6870e-04, -9.3946e-04, -1.1213e-03,
        -4.9691e-04, -7.5197e-04,  1.1748e-03,  9.8059e-01, -8.7640e-04,
         3.9450e-04, -6.8568e-04, -9.8059e-04,  3.4987e-04,  8.7723e-03,
         3.5821e-04, -6.8080e-04, -3.1959e-04, -1.5872e-04, -3.7306e-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([-4.0482e-03, -2.8084e-04, -2.0216e-04, -3.3000e-04, -1.7161e-03,
         9.5911e-04, -1.4269e-04, -3.4498e-04,  1.6185e-01, -6.7879e-04,
         6.9366e-04,  1.9716e-04, -7.0559e-04, -7.5312e-04,  3.1484e-03,
        -1.8440e-04,  8.2146e-01, -1.1300e-03,  3.2010e-04,  1.8154e-04,
        -6.7237e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 5.1004e-04, -1.8908e-01,  9.6882e-05, -4.2310e-04, -8.1392e-02,
        -5.1376e-04, -3.0259e-04, -8.0397e-04, -1.4384e-05, -3.2909e-04,
         3.6250e-04, -3.3641e-01,  4.5833e-04, -6.5520e-04, -3.8669e-01,
        -3.3639e-04,  3.3058e-04, -4.1538e-04,  4.1979e-04,  4.5486e-05,
        -4.1470e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1554e-03, -2.3651e-01,  2.0524e-04, -6.6075e-04, -3.3392e-01,
        -6.1089e-04,  3.9250e-04, -7.3360e-04, -9.9471e-04, -7.5915e-04,
         6.1818e-06, -4.2101e-01, -7.7723e-05, -9.2932e-04,  8.8214e-04,
         6.7932e-06, -4.8266e-04, -2.8849e-04, -3.6900e-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([-3.0709e-04, -3.5913e-01,  4.6138e-04, -1.7401e-04, -2.2163e-01,
        -1.1585e-03, -4.8573e-04, -3.1809e-04, -2.0460e-04,  2.6224e-05,
        -4.0109e-05, -4.1425e-01, -1.0352e-04, -4.4751e-04, -1.8546e-04,
        -5.7490e-04, -1.0925e-04, -5.6508e-05, -3.4015e-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([-2.8439e-03,  8.3571e-01,  9.5937e-05,  6.1203e-04, -4.6032e-04,
        -7.6530e-04, -7.1879e-04,  5.5955e-04, -9.7417e-05, -1.9031e-04,
         7.9869e-05, -1.4751e-04, -8.7798e-05,  9.4398e-04, -2.7994e-04,
         2.0140e-04,  2.5085e-04, -2.8633e-04,  1.5144e-01, -4.8944e-04,
         2.2887e-03,  5.1613e-04,  2.5720e-05,  6.1121e-04, -9.5377e-05,
         1.9686e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 8.1509e-04,  7.1117e-01,  6.7610e-05,  1.0481e-03, -6.2805e-04,
        -2.4600e-04, -5.2549e-05,  1.2361e-03, -9.0215e-04, -2.6220e-04,
         7.9202e-04, -3.9489e-04, -1.1601e-04,  8.5425e-04, -4.7292e-04,
         4.2667e-04,  3.0913e-04, -8.5158e-05,  2.7236e-01, -1.5952e-04,
         6.2868e-03, -4.0017e-04,  5.2993e-04,  3.8644e-04,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.3451e-03,  5.0470e-01,  7.0755e-05,  1.0142e-03,  1.3423e-04,
        -2.9433e-05, -7.4393e-04, -1.6973e-04, -5.6477e-04, -4.3196e-04,
         4.8489e-04, -5.5645e-04,  3.8676e-04,  1.1413e-03, -6.6908e-05,
        -5.6927e-04, -1.5177e-07,  4.0642e-04,  4.8407e-01,  1.5258e-04,
         8.7706e-04, -1.6288e-04, -5.9190e-04, -8.5687e-04, -2.8554e-04,
         1.8594e-04], device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.5786e-03, -2.8394e-01, -2.0702e-04,  3.8567e-05,  6.2709e-04,
        -6.4110e-05,  3.9991e-04, -5.2358e-04, -6.1211e-01, -5.1957e-04,
         8.9995e-04, -4.2828e-05, -2.0338e-04, -4.8895e-02, -8.7876e-04,
         3.7141e-04,  5.4019e-04,  6.5379e-04,  2.4438e-03, -4.0998e-02,
        -4.7633e-04, -1.6622e-04, -3.2689e-04,  9.5749e-05,  0.0000e+00,
         0.0000e+00], device='cuda:0', grad_fn=<DivBackward0>)]
Iter #550: [tensor([-1.1272e-03,  5.1196e-05,  7.0574e-03, -3.8191e-04,  9.1934e-04,
        -1.9654e-03, -7.5434e-04, -1.1875e-04, -2.0779e-03,  1.6139e-04,
        -5.2309e-04,  7.3345e-04,  9.2571e-01, -1.7916e-04,  4.8566e-02,
         2.5773e-03, -2.1863e-03, -3.3539e-03, -1.2298e-03, -3.2253e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.9009e-03,  4.3270e-04,  9.4543e-04, -7.8045e-04, -7.9498e-04,
        -2.7359e-03, -2.1349e-04, -1.1845e-03,  1.4952e-04,  3.1408e-04,
        -2.3272e-04, -7.4555e-04, -1.7708e-03,  5.5863e-03,  9.8034e-01,
        -6.0172e-04,  2.7308e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.3347e-03, -6.2627e-01, -2.8940e-04, -8.2767e-05,  1.6974e-04,
         2.7463e-05, -3.3623e-04, -1.4699e-01,  8.3461e-04, -5.7250e-04,
        -1.9087e-05,  2.2686e-05,  1.0578e-04, -4.1221e-04,  6.2112e-04,
         1.0498e-05,  3.1365e-05, -1.3730e-01,  1.7151e-04,  4.3607e-04,
         2.0349e-04,  4.4026e-04,  1.5460e-04, -1.1353e-04,  1.3400e-04,
         2.1437e-04, -5.8098e-02,  1.5610e-04, -7.7900e-04,  8.6634e-04,
        -1.9584e-02,  9.1361e-05,  1.3283e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-6.0759e-04,  2.0117e-01, -3.8736e-04,  1.2246e-03, -1.6607e-04,
         2.3170e-04,  1.0711e-04,  4.1390e-01, -6.0862e-04,  1.9561e-04,
        -1.1873e-04,  3.4304e-03, -4.8124e-04,  2.5261e-04, -3.8865e-04,
         6.1183e-05, -1.4335e-04,  2.9892e-01, -4.4513e-05, -6.5347e-04,
        -1.8849e-04, -5.5812e-04, -5.5718e-04, -4.8755e-05, -9.7739e-05,
        -2.8180e-04,  7.4730e-02, -2.1404e-04, -7.4503e-05, -1.6475e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.9483e-03,  6.2215e-01,  2.0654e-04,  1.1041e-03, -5.5918e-04,
         6.1781e-05,  2.2465e-04,  9.7702e-02, -4.9318e-04,  6.6535e-04,
         1.3875e-04,  3.4501e-03, -5.1331e-06,  1.6264e-04, -2.6633e-04,
         2.0558e-04, -3.6058e-05,  1.4565e-01,  1.6999e-04, -4.0421e-04,
         9.3146e-05, -5.7989e-04, -2.3671e-04, -1.5891e-04, -4.0980e-05,
        -1.6287e-05,  9.0013e-02,  2.9785e-04,  4.6333e-04, -2.0363e-04,
         3.1512e-02, -7.6473e-04,  9.7711e-06], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.3796e-04, -3.0018e-04,  1.0221e-04, -3.0063e-03, -3.5894e-04,
        -5.9586e-01,  2.9350e-05, -1.1983e-04, -1.4824e-03, -2.2736e-01,
         4.6926e-04, -1.3380e-04, -3.8890e-05, -2.2547e-05,  8.3820e-05,
        -3.6238e-02, -2.5615e-04, -1.4464e-03,  4.7139e-04, -1.3046e-01,
        -3.2783e-04, -4.8023e-04,  6.2207e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 8.6300e-03,  2.7913e-04,  6.6285e-04, -6.8714e-04, -5.5706e-04,
        -4.1016e-01,  7.0993e-04,  4.6994e-04, -3.7300e-05, -4.5659e-01,
         4.3119e-04,  2.3010e-04, -3.1795e-04,  3.0195e-04,  4.0847e-04,
        -1.1681e-01,  6.0633e-04, -6.0573e-05,  4.2073e-04,  1.0848e-03,
         4.1359e-04, -1.3290e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 5.9428e-03, -7.4540e-04, -3.5617e-04, -1.5824e-03, -9.6600e-05,
        -3.1640e-01,  1.3827e-04, -2.6075e-05, -5.8594e-03, -5.9555e-01,
         1.6669e-04, -2.8153e-04, -2.7799e-04, -2.2274e-04, -2.8089e-04,
        -6.8169e-02, -3.1783e-04, -1.2647e-03,  1.3835e-04, -1.0760e-03,
        -3.4682e-04, -4.8540e-04, -2.0770e-04, -2.4514e-05,  3.9591e-05,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.7967e-03, -3.5388e-01,  2.8407e-04, -1.2314e-05, -6.0279e-02,
         1.0891e-03, -4.5294e-04,  8.5435e-05, -4.0205e-04, -5.4810e-01,
         1.5207e-04,  1.0087e-04,  1.4301e-04,  5.7080e-05,  4.2243e-04,
        -2.9118e-02,  1.7417e-04, -1.9658e-04,  5.7469e-04, -9.1566e-04,
         9.9762e-05,  2.9781e-04,  3.6748e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2121e-03, -1.5834e-01,  2.5829e-04, -3.7398e-04, -3.2859e-01,
         2.2735e-04,  1.6286e-04, -1.9892e-04, -1.5519e-04, -3.5494e-01,
        -6.1491e-04,  1.6885e-04,  5.9456e-04, -1.3799e-04,  2.8122e-04,
        -5.6978e-02, -1.3889e-04, -8.9165e-04, -9.2129e-02,  9.9008e-05,
         9.7175e-06, -1.3707e-04,  1.9038e-04,  1.6336e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.0712e-04, -2.7510e-01, -3.2652e-04, -4.3731e-04, -9.4668e-02,
         2.4460e-04,  2.3108e-05,  1.5394e-04, -3.6702e-04, -4.9551e-01,
        -7.3095e-05, -3.0254e-04,  6.3956e-04, -2.3194e-04,  4.9266e-04,
        -2.3839e-02, -3.4231e-04, -6.6186e-04,  7.4262e-04, -1.0431e-01,
         7.6725e-04, -2.8267e-04,  1.7259e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.7327e-03,  7.4768e-01, -7.7417e-04, -1.2763e-04, -3.4790e-04,
         1.2788e-01, -1.2051e-03, -5.6480e-04,  3.3544e-04,  1.1540e-01,
        -3.3951e-05, -2.4120e-05,  9.4072e-04,  7.2986e-04, -9.0844e-04,
        -3.1376e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #600: [tensor([ 6.9580e-03, -2.1019e-01, -4.6092e-05,  1.2781e-03, -1.9048e-01,
        -4.2393e-04,  3.9422e-05, -1.8351e-04, -7.4485e-04, -3.3240e-01,
        -4.4256e-04, -1.6567e-04, -2.7069e-04, -3.0979e-04, -4.0258e-04,
        -3.6574e-05, -4.9491e-05, -2.4334e-04, -6.3975e-02, -4.0392e-04,
        -3.4606e-04, -4.2387e-04,  3.4688e-04, -1.8948e-01, -6.0081e-05,
         9.4619e-05, -2.1560e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-5.5357e-04,  2.0240e-01,  3.2949e-04, -1.9111e-04, -1.6173e-04,
         4.6870e-05, -1.1753e-04,  6.8251e-04, -3.1043e-04,  3.8597e-04,
         1.6626e-05,  1.5039e-04, -3.7955e-05, -1.8031e-04,  4.4109e-02,
        -6.9370e-05,  2.1045e-04, -5.5304e-05,  5.7451e-04,  2.3668e-04,
         1.8641e-04,  7.3295e-02,  2.4892e-04,  8.0653e-05,  2.4696e-05,
        -8.5400e-05,  1.8458e-01, -2.0287e-04,  3.7043e-04, -4.8890e-04,
        -9.2139e-05,  3.5263e-01,  2.5818e-05, -4.7196e-04,  1.3615e-01,
        -1.1263e-04, -4.7432e-05, -9.6089e-05,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.4184e-03, -5.8307e-01, -3.0593e-04, -2.5622e-04,  1.3096e-04,
        -7.8391e-05,  3.8343e-04, -3.1680e-04, -1.6573e-04, -1.3308e-03,
        -3.0017e-04, -6.4302e-05, -2.9639e-05, -4.7135e-04, -1.7990e-02,
        -2.0292e-05, -5.4087e-04,  6.4673e-04, -5.0930e-04, -4.0178e-04,
        -2.3862e-04, -4.0397e-02, -2.1120e-04, -5.8491e-05, -1.9699e-04,
        -6.6134e-05, -3.0045e-01,  1.1117e-04, -3.7305e-04,  1.0446e-04,
        -2.8721e-04, -4.7572e-02,  1.7849e-04, -1.4586e-05,  2.4637e-04,
        -5.7477e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-1.7122e-03, -2.1849e-01, -9.1268e-05, -5.3952e-05,  6.3123e-05,
        -9.2741e-05,  2.3728e-04, -8.8580e-05, -4.8748e-05, -1.4580e-03,
         8.4261e-05, -2.3313e-04, -3.8628e-05,  3.0405e-04, -7.8630e-02,
         5.9579e-05, -1.8688e-04,  3.3917e-04, -1.6030e-04,  1.8849e-05,
         1.5484e-04, -2.3881e-02,  1.8996e-04,  5.9185e-05, -3.5650e-04,
        -1.8846e-04, -3.3322e-01,  2.2976e-04, -4.0743e-04, -1.1921e-04,
        -3.5644e-04,  1.1938e-04, -3.3707e-01, -1.7257e-04,  3.4158e-04,
         6.1866e-05,  8.8351e-05,  2.2498e-04, -3.5121e-04,  1.4398e-05],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-6.5782e-04, -3.6531e-01, -4.6233e-05,  5.7432e-04, -1.0519e-01,
         1.1288e-04, -5.7832e-05,  3.0617e-04, -5.0657e-04, -2.8376e-04,
        -3.3942e-04, -4.0497e-04, -2.2753e-01, -3.0541e-04,  1.1706e-04,
         2.4962e-04,  1.9139e-04, -1.6084e-01, -2.5611e-04,  5.8537e-04,
        -6.7166e-02, -1.2406e-04, -8.8539e-06, -3.4996e-04, -1.7615e-04,
        -3.2804e-04,  7.3286e-05, -7.7309e-04, -6.6799e-02,  5.2441e-05,
         1.3603e-04,  1.4900e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 4.1558e-03,  4.0150e-01,  1.0669e-04, -1.3219e-03,  1.3463e-01,
         4.3279e-04,  8.9378e-05, -8.0450e-04,  2.9154e-04, -1.8387e-04,
         8.0235e-04, -6.2401e-04,  3.1053e-02,  2.3192e-04,  1.2374e-04,
        -2.7605e-04,  5.4670e-05,  4.2226e-01,  1.9627e-04,  8.0678e-05,
         1.1072e-04, -4.0047e-04, -2.7303e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-3.7012e-03, -5.4418e-01, -3.1897e-04,  6.8204e-04, -1.6549e-01,
        -6.1513e-05, -1.3485e-05,  1.5488e-04, -2.3872e-04, -1.2956e-04,
        -5.9490e-04, -3.6359e-05, -4.7670e-02, -1.5979e-04,  1.7328e-05,
        -1.4112e-04, -1.1966e-04, -1.1505e-01, -3.7378e-04,  6.2037e-04,
        -1.1716e-01, -3.8205e-04, -1.9903e-05,  1.1926e-04, -7.8662e-05,
         6.2301e-05, -6.8796e-04, -1.6350e-04, -2.1247e-04, -8.0475e-04,
        -3.3747e-04, -1.0352e-04,  1.1912e-06, -1.0860e-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([ 1.6299e-03, -1.4075e-04, -1.0220e-03, -4.5255e-05, -7.9953e-04,
        -3.2886e-01, -4.5306e-04,  3.8110e-04,  3.3325e-04, -2.8422e-01,
        -4.1710e-04, -3.8147e-01,  1.0887e-04,  1.2652e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-2.3559e-03,  1.1399e-04,  1.3125e-03,  1.1093e-03, -6.3832e-04,
         6.7635e-01, -1.8759e-04, -6.2847e-04, -3.8945e-04, -7.8460e-04,
        -2.8992e-04,  3.1549e-01,  1.9054e-05,  3.2421e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([-8.9018e-04, -4.8040e-04, -1.2419e-03, -1.4512e-04, -3.7863e-04,
        -1.6735e-01, -4.5411e-04, -8.1329e-04,  1.1472e-04, -3.1528e-01,
        -1.5360e-04, -4.1251e-01,  1.9485e-04,  2.8879e-03, -9.6158e-02,
        -5.4631e-04,  3.9998e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 2.6575e-03, -6.9280e-04, -3.0085e-04, -5.7407e-04, -1.0778e-03,
        -8.9457e-04, -1.5095e-04, -9.3301e-04,  8.9027e-01,  6.1818e-05,
         3.0666e-04, -6.5185e-04, -7.7712e-04,  5.8393e-04,  1.3126e-04,
         1.6453e-05, -8.5127e-04,  9.7909e-02,  6.1701e-04, -3.7753e-05,
         3.5965e-04, -1.4588e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>), tensor([ 1.1141e-03, -4.8387e-04, -5.1842e-04,  2.9072e-04, -5.3501e-04,
        -1.2356e-03, -3.0739e-04, -2.7914e-04,  2.2496e-01, -3.6329e-04,
         3.9749e-04, -6.2601e-04,  7.4010e-05, -4.6997e-05,  4.5220e-05,
        -2.1232e-04, -3.2009e-04,  6.0016e-01, -4.8501e-04,  1.6656e-01,
         5.1583e-04,  1.8761e-04, -2.7883e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00],
       device='cuda:0', grad_fn=<DivBackward0>)]
Iter #650: [tensor([ 2.4895e-03, -7.0426e-04, -5.7028e-04, -8.6937e-04, -2.7263e-04,
        -5.2078e-04, -1.2693e-01,  1.1025e-03, -4.0792e-06,  4.3842e-04,
        -3.3791e-01, -1.1737e-03, -5.2412e-01,  1.6724e-04, -2.0297e-03,
        -2.0470e-04, -3.5324e-04,  7.0473e-06, -1.2882e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2605e-03, -3.1793e-01,  2.1804e-04,  1.8239e-04, -2.4201e-01,
         5.1592e-05,  3.2930e-04,  3.7493e-04, -7.4681e-05,  6.4953e-04,
         1.4172e-03, -2.1351e-02,  1.4954e-04,  5.4902e-04, -3.2652e-01,
         5.2724e-04,  3.4244e-04, -1.1798e-04,  4.1897e-04,  1.5105e-04,
         3.8848e-04, -8.1436e-02, -3.6927e-04, -6.6921e-04, -2.7481e-04,
         1.4279e-04,  9.1660e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.4627e-03, -1.8396e-01, -1.6036e-04, -9.3576e-05, -3.0909e-01,
        -1.5568e-04, -1.8483e-04,  1.3954e-04, -2.5943e-05,  2.9018e-04,
         3.2295e-04, -1.7090e-03,  1.8429e-04,  3.0009e-04, -1.6561e-01,
        -4.3561e-04,  3.3234e-04, -8.0541e-05,  6.7625e-05,  1.5404e-04,
         3.0836e-04, -1.2308e-01, -3.9787e-04, -3.6182e-04, -1.6597e-04,
        -2.0735e-01, -6.0415e-05,  1.4025e-04,  1.2077e-04, -1.0042e-04,
        -1.5345e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.2518e-04, -2.4145e-01,  2.5110e-04, -1.1751e-04, -8.2470e-02,
        -2.3838e-04, -1.2243e-05,  3.3822e-07, -3.3847e-04,  2.4500e-04,
         9.3131e-04, -1.2728e-02, -2.5668e-05,  1.0875e-04, -4.7581e-01,
         1.6447e-04,  1.2819e-04, -2.3923e-04,  1.8637e-04, -2.2891e-04,
        -1.0849e-04, -4.8750e-02, -1.1989e-04, -4.2777e-04, -1.3401e-01,
        -2.6893e-04, -2.6333e-04, -4.7720e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 7.9543e-03, -9.3434e-05, -1.4215e-04,  3.4208e-01,  5.7201e-04,
        -1.6483e-03,  3.5473e-01,  5.4629e-04,  2.2518e-03,  5.9066e-04,
         7.6161e-04,  2.8410e-01, -1.1399e-03,  1.9564e-04,  1.2922e-03,
         1.3352e-03, -3.2847e-04,  1.3674e-04,  1.0049e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.7024e-03,  2.2828e-04,  3.8650e-04,  1.2512e-01, -2.8025e-04,
        -5.4755e-05,  1.0490e-01, -1.9595e-04,  7.7868e-04,  1.9818e-05,
         5.9011e-05,  4.0208e-01,  3.3429e-04,  3.4437e-04,  3.5953e-01,
         1.6882e-04,  5.0423e-04, -2.9546e-04,  4.5799e-04, -3.4149e-04,
         1.1389e-04,  7.2689e-04,  6.0733e-04,  4.8854e-05, -1.6396e-04,
        -1.5695e-04,  3.9219e-04,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.4051e-04,  2.5510e-04,  5.6999e-04,  1.5732e-01, -2.3713e-04,
         7.9740e-04,  2.5005e-01, -2.0065e-04,  4.1553e-04, -3.0313e-05,
        -3.0933e-04,  4.0661e-01, -2.6385e-04,  1.0033e-04,  1.8216e-01,
         2.7429e-04,  1.4479e-04,  1.1706e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 3.6844e-03, -4.5976e-04, -1.1444e-06,  5.8504e-05, -3.4101e-04,
        -1.6568e-03,  1.3639e-01, -4.7561e-04,  2.1683e-04, -1.8927e-05,
        -5.4065e-05,  8.0114e-05, -7.0266e-05,  6.9920e-04,  8.2771e-02,
         2.0089e-05,  9.2490e-06,  7.8154e-02, -3.8829e-05,  1.5983e-03,
         9.8410e-02, -6.6092e-04,  1.9706e-04, -1.1029e-06,  7.1347e-05,
        -1.6538e-05,  2.5242e-01, -1.1504e-04,  1.5759e-03, -3.0393e-04,
         4.6039e-06, -1.7269e-03,  3.3695e-01, -5.4855e-04,  1.2982e-04,
         6.2071e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 4.2117e-04,  3.2772e-05, -4.7705e-04,  7.9641e-04, -5.4247e-04,
        -2.1966e-03,  2.5237e-01, -6.5578e-04, -3.8374e-04,  9.9223e-04,
        -2.2877e-04, -1.0453e-04, -2.0090e-04,  6.4558e-04,  2.2429e-01,
         4.4982e-05, -2.5743e-04,  1.4410e-01, -9.3159e-06,  3.8744e-03,
         3.6264e-01,  1.2700e-04,  4.7983e-04,  2.9985e-04,  1.2673e-04,
         1.7529e-03, -4.2696e-05, -8.1361e-07, -3.6402e-04,  3.3474e-04,
        -4.6979e-04,  2.4984e-04,  3.1054e-04, -1.2594e-04, -4.2437e-05,
        -6.8797e-06,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-4.1940e-03,  1.2444e-05,  8.0042e-05,  1.5710e-04,  4.3670e-05,
        -6.7655e-05, -1.4640e-04,  6.0061e-02, -1.1720e-04,  2.2075e-05,
         4.5118e-04,  1.4446e-05, -3.1211e-05, -1.3924e-03,  3.3537e-01,
        -8.0139e-05,  8.0390e-05, -4.2434e-04, -3.8580e-05, -4.9177e-05,
        -4.7723e-04,  4.1150e-04, -2.9442e-04, -4.9535e-05, -6.8074e-05,
         8.7847e-05, -2.4691e-04,  2.1224e-04,  1.7229e-01,  1.4891e-05,
        -1.5460e-04,  4.3797e-05,  2.0003e-01,  1.6296e-04,  2.1870e-04,
         7.5502e-05,  8.0726e-05, -1.0700e-04,  2.2183e-01, -1.8448e-04,
         8.1263e-05,  4.9913e-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([-6.2460e-03,  3.1603e-04,  5.2587e-04,  2.3700e-04, -1.1432e-05,
         2.6790e-04,  1.0441e-04,  7.4036e-02, -5.0334e-06,  2.9941e-04,
         1.9575e-03,  2.1392e-04,  3.7265e-04, -8.0017e-04,  1.7456e-01,
        -2.3951e-04,  1.6956e-04, -9.2336e-05,  2.5890e-04,  1.3173e-04,
        -8.8606e-05, -1.2526e-04,  6.2440e-05,  1.8732e-04,  4.2186e-05,
         3.5938e-04,  4.8213e-04,  6.4233e-05,  4.8544e-01, -6.2647e-05,
         1.2537e-04,  1.0766e-04, -5.1572e-04,  8.2842e-05,  8.3368e-05,
         1.1889e-01,  2.1210e-04, -2.7912e-05,  1.3141e-01,  4.1917e-04,
         1.8024e-04,  1.9667e-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([-6.2725e-04, -6.9478e-05,  1.8004e-05, -8.4534e-05, -4.6412e-06,
        -2.0362e-04, -4.7476e-05, -3.9316e-02, -1.1872e-04, -2.3709e-04,
        -1.2041e-03,  8.5797e-05, -7.6872e-05,  7.8955e-04, -1.2679e-01,
        -1.3389e-04,  8.8616e-05,  2.0675e-04,  1.4029e-04, -1.4872e-05,
         3.6863e-05,  6.4330e-05,  2.9778e-04, -6.5912e-05,  2.4251e-05,
         1.3468e-04,  2.9585e-05, -1.3701e-04, -2.2259e-01,  6.8942e-05,
        -7.7561e-05, -1.0370e-04, -1.3442e-04, -3.3165e-04,  1.2503e-04,
        -3.2530e-05,  1.9824e-04, -1.1088e-03, -3.7117e-04,  8.9257e-06,
        -1.1558e-04, -5.2178e-02, -1.6229e-05,  1.8566e-04,  1.0859e-04,
        -5.5069e-01,  2.1936e-04,  1.8501e-04, -9.8023e-05], device='cuda:0',
       grad_fn=<DivBackward0>)]
Iter #700: [tensor([-3.2050e-03, -5.0459e-05,  4.9745e-04, -1.2383e-04, -7.6459e-05,
         3.3533e-04,  1.4530e-01,  2.8173e-04, -9.6019e-06, -1.4385e-04,
        -1.9143e-03,  1.1556e-01, -1.3925e-04,  3.1492e-04,  2.2986e-01,
        -3.8070e-04,  1.7896e-03, -1.1065e-04, -1.7291e-03, -1.9213e-04,
         1.1546e-04, -1.0567e-04, -2.7878e-04,  4.9724e-01,  1.3949e-04,
        -7.9554e-05, -1.6335e-05,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-9.2818e-04,  5.2175e-02, -1.8602e-04, -3.8964e-04,  2.7994e-01,
        -2.5854e-04,  5.4630e-04,  5.6049e-05,  6.6143e-01,  4.1521e-05,
         3.0041e-04,  1.8909e-04,  3.8189e-04,  7.2705e-05, -1.2890e-04,
         6.5250e-04,  4.3293e-05, -2.6626e-05,  3.5769e-04, -5.5587e-07,
         3.0451e-04, -8.0348e-05, -3.8360e-04, -9.1921e-05,  4.5110e-05,
         2.0349e-05, -1.6946e-04, -1.1046e-04, -9.5940e-05, -2.6675e-04,
         1.3975e-04,  2.1072e-05,  1.5690e-04,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.9217e-03, -4.2085e-02,  2.3790e-04,  9.0182e-05, -9.3556e-02,
        -4.5096e-04, -5.1408e-04, -2.4239e-04, -7.3620e-02,  5.9130e-05,
        -1.2311e-04, -2.5249e-04,  2.4085e-04, -2.6994e-01,  1.7357e-04,
        -1.4728e-04,  9.3410e-05, -4.0212e-01, -2.7772e-04,  5.8138e-05,
         6.4516e-04, -1.1131e-01, -5.1346e-04,  8.3936e-05,  2.4226e-04,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-2.2679e-03, -8.3327e-02,  4.4221e-05,  3.1462e-04, -2.7929e-02,
        -3.3131e-04,  4.2300e-04,  4.0317e-05, -5.4872e-01, -1.1060e-04,
         5.2313e-05, -1.9828e-05,  1.8247e-04, -3.5511e-02, -3.4033e-04,
        -1.8621e-04, -1.9170e-05, -1.5976e-01, -5.6007e-04, -2.0785e-04,
        -3.3863e-06, -2.7163e-04,  1.5416e-04, -1.3842e-01, -4.4960e-04,
        -1.0172e-04,  2.5650e-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([ 1.0363e-03, -5.3666e-01,  4.7503e-05, -2.4990e-04,  8.6874e-04,
        -1.3668e-01,  1.4180e-04, -1.1797e-04, -8.1286e-05,  1.3864e-04,
         1.0527e-03, -6.4126e-03, -2.6156e-04, -2.9789e-04, -9.6849e-02,
        -1.4383e-04,  5.8588e-05,  7.1056e-05, -1.5826e-01,  3.9228e-05,
         2.7905e-04,  4.4572e-05, -6.9919e-05,  7.4472e-05,  4.9858e-05,
        -1.7285e-02, -1.8808e-04, -2.4940e-04, -6.6606e-05, -4.1570e-02,
         7.2000e-05, -2.3116e-05,  3.4120e-04,  2.1840e-04], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.1828e-03, -1.6727e-01, -1.0052e-04, -2.7638e-04,  7.0390e-04,
        -3.0486e-01, -9.2967e-06, -1.8211e-04, -3.0460e-03, -4.6675e-05,
         1.5812e-04, -4.6043e-02, -4.6692e-04, -2.4499e-04, -2.8649e-01,
        -7.4289e-05, -1.7012e-04, -4.6631e-05, -2.4654e-02, -5.7249e-05,
        -6.7474e-05,  4.3820e-05,  2.5312e-05, -1.8050e-04, -8.3069e-05,
        -7.4364e-02, -8.3057e-05, -7.4772e-04, -1.2358e-04, -8.7871e-02,
        -1.8200e-04, -5.5205e-05, -9.3596e-05,  7.5862e-07], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.5574e-03, -4.7960e-01,  1.0135e-05, -6.3871e-05,  3.4636e-04,
        -1.8047e-01,  1.8667e-04, -1.9340e-05, -1.9343e-04, -7.5587e-05,
         9.6472e-04, -1.1555e-02, -4.0681e-04, -2.5321e-04, -1.6777e-01,
         1.9422e-05,  1.5682e-04,  6.6583e-05, -1.0793e-01, -4.5122e-04,
        -6.4313e-06,  1.3838e-04,  1.0452e-04, -3.5163e-05,  4.8303e-05,
        -4.5237e-02, -1.8795e-04,  2.1680e-04,  3.3755e-04,  3.4425e-04,
         4.1570e-05,  2.0704e-04,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-3.8341e-03,  6.0555e-02,  3.2950e-04, -1.4117e-03,  1.7452e-01,
         7.7096e-04, -5.0082e-04, -1.6088e-04, -1.6157e-03, -1.6420e-04,
        -4.7759e-04, -4.2672e-04, -1.8903e-04, -6.2840e-05, -2.1507e-05,
        -2.2599e-04,  7.5891e-06,  5.0829e-01,  3.2226e-04,  9.1037e-04,
         1.6806e-04, -1.4480e-03, -4.8235e-05, -2.2227e-04,  2.4258e-01,
         2.3478e-05,  4.8412e-04, -2.2592e-04,  5.7505e-06,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.2485e-03,  1.1991e-01, -1.3518e-04, -9.4011e-04,  8.6944e-01,
        -2.2695e-04, -5.5708e-04, -5.3181e-04, -9.7381e-04, -3.5677e-04,
         1.8621e-04, -7.6295e-04, -1.2832e-04, -2.6430e-04,  1.6922e-04,
         3.5415e-04,  1.0934e-04, -1.0607e-03,  2.0989e-03, -1.1481e-05,
        -7.3235e-05, -3.9342e-04,  6.3458e-05,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 1.0708e-04,  9.9673e-02,  1.9176e-04, -1.2943e-03,  6.5702e-01,
         1.9920e-04, -2.7159e-05,  3.1394e-04, -1.3677e-03,  2.1595e-04,
         2.7186e-04, -3.3723e-05, -3.6743e-04,  3.1166e-04,  6.4921e-05,
         1.7258e-04,  1.6769e-04,  2.3586e-01, -6.8744e-04,  5.6550e-04,
         5.1628e-04, -6.5608e-05, -2.0042e-04, -3.0379e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([-1.0594e-02,  9.3120e-05, -3.7018e-04, -6.7664e-02, -2.9609e-04,
        -2.0415e-04,  4.1215e-05,  2.1904e-04,  4.0165e-04,  3.2305e-04,
         1.3543e-03,  6.2799e-04, -5.7108e-04, -8.6644e-04, -9.1094e-01,
         1.5654e-03, -6.9511e-04, -7.0833e-05, -1.3716e-03, -6.3051e-04,
         2.7519e-04, -4.7518e-04,  3.5373e-04,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>), tensor([ 2.3991e-03,  5.1799e-04,  1.7081e-04,  1.8512e-01, -4.6008e-04,
         3.3878e-04,  8.9539e-04,  3.0344e-04, -9.5391e-04, -7.0569e-06,
         1.2561e-04,  4.5079e-04, -2.9079e-04, -8.3969e-05,  8.0376e-01,
         1.7359e-04, -3.4582e-05,  2.1151e-04,  1.3601e-03,  6.1333e-04,
        -4.3464e-04,  6.5459e-04, -1.3810e-04,  5.0473e-04,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00,
         0.0000e+00,  0.0000e+00,  0.0000e+00,  0.0000e+00], device='cuda:0',
       grad_fn=<DivBackward0>)]
