diff --git "a/weights/best_model_metadata.json" "b/weights/best_model_metadata.json"
--- "a/weights/best_model_metadata.json"
+++ "b/weights/best_model_metadata.json"
@@ -1,316 +1,46 @@
 {
-  "epoch": 9,
+  "epoch": 0,
   "optimizer_state_dict": {
     "state": {
       "0": {
-        "step": "tensor(12520.)",
-        "exp_avg": "tensor([[-6.3960e-05, -6.0342e-06,  2.5991e-05,  ...,  7.9031e-06,\n         -1.0817e-05,  1.0993e-06],\n        [ 8.8662e-06, -1.3754e-05, -6.6039e-06,  ..., -7.5104e-06,\n          6.7735e-06,  1.1299e-05],\n        [ 2.0843e-05, -5.0735e-05, -6.4116e-05,  ...,  4.3576e-05,\n          6.4431e-05, -3.8258e-05],\n        ...,\n        [-3.0679e-06, -6.5986e-05, -5.2217e-05,  ...,  1.6762e-06,\n          6.4565e-05, -6.0786e-05],\n        [-5.6052e-45, -5.6052e-45,  5.6052e-45,  ..., -5.6052e-45,\n          5.6052e-45, -5.6052e-45],\n        [ 1.9304e-05,  9.2353e-05, -7.0835e-05,  ...,  7.7714e-05,\n          4.0903e-05, -2.3749e-05]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[1.4227e-08, 9.8451e-09, 5.8651e-09,  ..., 7.5263e-09, 6.9033e-09,\n         6.9766e-09],\n        [5.8534e-09, 5.5103e-09, 3.1861e-09,  ..., 2.9842e-09, 3.1072e-09,\n         2.1654e-09],\n        [1.4612e-08, 1.2444e-08, 1.6592e-08,  ..., 1.1548e-08, 9.1652e-09,\n         8.7804e-09],\n        ...,\n        [1.3218e-08, 1.5949e-08, 2.4706e-08,  ..., 1.1718e-08, 9.9901e-09,\n         1.0700e-08],\n        [5.9810e-16, 3.1201e-15, 9.4819e-16,  ..., 5.5587e-18, 2.9919e-15,\n         1.7631e-17],\n        [1.3356e-08, 1.3326e-08, 1.5134e-08,  ..., 1.3426e-08, 7.8721e-09,\n         6.7177e-09]], device='cuda:0')"
+        "step": "tensor(1252.)",
+        "exp_avg": "tensor([[ 2.1732e-04, -2.4132e-04,  1.1563e-04,  ..., -5.8336e-05,\n          1.2537e-04,  1.3779e-04],\n        [ 1.7667e-04, -3.6072e-04, -1.2474e-05,  ...,  2.0924e-04,\n          2.1620e-04,  4.1487e-05],\n        [-3.0079e-04,  8.5631e-05,  7.0483e-05,  ..., -3.7193e-05,\n         -1.1348e-04, -1.5361e-04],\n        ...,\n        [-8.2470e-04,  1.0081e-03, -1.7612e-04,  ..., -2.3494e-04,\n         -3.9981e-04, -2.6347e-04],\n        [ 1.2842e-04,  3.7175e-05,  7.0022e-05,  ..., -1.4157e-04,\n         -3.5385e-05, -7.2958e-05],\n        [ 2.2581e-04, -1.7839e-04, -5.9781e-05,  ...,  2.7143e-04,\n          1.1676e-04,  8.6820e-05]], device='cuda:0')",
+        "exp_avg_sq": "tensor([[6.1722e-07, 9.9729e-07, 1.2621e-07,  ..., 1.0472e-07, 2.6292e-07,\n         1.6593e-07],\n        [8.7335e-07, 1.2442e-06, 1.2775e-07,  ..., 8.8972e-08, 4.3964e-07,\n         2.0230e-07],\n        [4.5250e-07, 1.2431e-06, 1.3188e-07,  ..., 1.4839e-07, 1.1970e-07,\n         1.2886e-07],\n        ...,\n        [5.4923e-07, 1.1917e-06, 1.1266e-07,  ..., 1.5080e-07, 1.6123e-07,\n         1.1640e-07],\n        [7.2196e-07, 8.9995e-07, 9.6138e-08,  ..., 1.3409e-07, 1.3516e-07,\n         8.6947e-08],\n        [3.4820e-07, 5.3155e-07, 8.0998e-08,  ..., 1.3389e-07, 1.0982e-07,\n         8.2062e-08]], device='cuda:0')"
       },
       "1": {
-        "step": "tensor(12520.)",
-        "exp_avg": "tensor([-9.0691e-04,  7.2409e-04,  3.0193e-03,  ...,  1.0748e-03,\n         5.6052e-45,  3.6514e-03], device='cuda:0')",
-        "exp_avg_sq": "tensor([1.4690e-05, 6.2727e-06, 2.1278e-05,  ..., 2.5137e-05, 2.0251e-11,\n        2.1901e-05], device='cuda:0')"
+        "step": "tensor(1252.)",
+        "exp_avg": "tensor([ 7.0912e-03,  6.0797e-03, -2.2731e-03, -1.1188e-03, -1.7205e-04,\n        -9.0995e-03, -2.1475e-03, -2.6487e-05, -1.2467e-03, -2.5011e-04,\n        -2.9646e-04, -3.2116e-04,  4.7426e-03,  7.0674e-03, -3.0869e-03,\n         3.5444e-03, -2.3375e-03,  2.5352e-03,  2.4416e-03,  1.8826e-03,\n        -4.4655e-03, -2.7397e-03, -1.9701e-03, -2.7799e-04, -5.0051e-03,\n         6.0273e-03,  9.2500e-03,  3.9571e-03,  1.8661e-03, -4.9580e-03,\n        -4.0891e-03, -1.3336e-02, -1.8346e-03, -1.0091e-03, -1.0081e-03,\n         2.2795e-03, -3.0143e-04, -1.0540e-03, -1.3617e-03,  7.0397e-03,\n        -3.7204e-03,  7.1450e-03, -6.5819e-03,  3.1580e-03, -5.3148e-03,\n        -2.6685e-03,  4.9089e-03, -3.5287e-03, -5.4789e-03,  8.7574e-05,\n        -2.4788e-03,  2.4027e-03, -1.2638e-03, -8.4531e-04, -3.2514e-03,\n        -3.0138e-03,  2.1474e-03,  2.5653e-04, -1.6644e-03,  3.1347e-03,\n        -7.8441e-03,  6.6605e-03, -8.5155e-03, -5.3620e-04, -5.2802e-03,\n        -1.6341e-04,  5.6512e-03,  4.7745e-03, -1.0558e-02, -2.7493e-03,\n         1.7077e-03, -5.8065e-03, -4.4423e-03, -5.5764e-03, -4.1995e-03,\n         3.2314e-03, -2.9026e-03, -1.7812e-03, -1.3303e-02,  5.8747e-03,\n         5.3411e-03, -2.8759e-03,  7.2511e-04, -6.4654e-03, -7.9242e-03,\n         1.3007e-02, -3.9319e-03,  6.3910e-03,  9.0574e-04, -7.7051e-04,\n        -3.2434e-04,  2.1786e-04,  3.7087e-03, -2.8836e-03, -3.6037e-03,\n        -7.8597e-03,  1.8171e-04,  8.5504e-04,  3.2590e-04, -2.9010e-04,\n         3.4104e-03,  2.4339e-03, -5.3353e-04,  1.0830e-02, -3.8072e-03,\n        -1.2378e-03,  5.0279e-03,  3.9110e-03, -2.9825e-03,  2.4288e-03,\n         3.0165e-03, -5.5090e-04,  3.5609e-03,  2.6118e-03, -4.1643e-04,\n         4.7672e-03, -3.9978e-04,  6.7410e-03, -7.2266e-03,  3.1271e-03,\n         2.0010e-03, -2.5916e-03, -9.0442e-03, -4.2935e-03, -3.3065e-03,\n        -4.8675e-04, -8.3347e-04, -7.9866e-04, -7.1479e-03,  1.8182e-03,\n         7.2755e-03,  2.2139e-03, -8.0096e-03,  6.8266e-03,  2.1929e-03,\n        -2.6154e-04,  1.7752e-03, -1.2123e-03, -3.3109e-04, -1.5690e-03,\n        -7.0882e-04, -3.8521e-03,  6.8386e-03,  2.3643e-03,  2.3162e-03,\n         2.1250e-03,  3.8087e-04,  2.1138e-03, -3.7927e-03,  5.0401e-03,\n        -2.1434e-03,  2.7185e-03,  8.8836e-03, -2.2127e-03,  7.1977e-03,\n         6.3743e-03,  9.9269e-04, -4.1267e-03,  1.0365e-03, -1.9980e-03,\n         4.8892e-03, -3.2718e-03, -1.1687e-02,  1.4169e-04, -9.5171e-04,\n        -2.7085e-03, -4.0704e-03, -2.1806e-04,  8.2463e-04, -2.4732e-03,\n        -1.7975e-03, -5.0172e-03,  1.1143e-03,  7.1312e-03, -3.8062e-03,\n        -5.0800e-03, -2.5988e-03, -5.6488e-03,  5.5590e-03, -5.1282e-04,\n        -1.3679e-03,  2.4185e-03,  1.4071e-03, -3.8554e-04, -1.4484e-03,\n         1.0441e-03,  1.4888e-03, -6.7347e-04, -9.9482e-04,  2.7858e-03,\n        -1.6250e-03,  9.5575e-03,  1.4734e-03, -6.8645e-03,  5.3222e-03,\n         9.7656e-04,  4.2416e-03, -5.1792e-04,  9.2710e-04,  5.7750e-03,\n        -7.9313e-03,  5.8195e-03, -1.4147e-03, -4.8664e-03,  2.2078e-03,\n         2.5382e-03, -1.5230e-04,  4.3010e-03, -6.6360e-04,  4.3867e-03,\n        -2.6316e-04,  4.4627e-03, -1.8241e-03, -4.5507e-03,  4.7632e-03,\n        -5.0898e-03, -6.4845e-03, -2.9958e-03,  5.4260e-03,  3.3777e-03,\n        -5.0020e-03,  5.4927e-03, -1.5026e-03, -5.9193e-03,  2.2837e-03,\n        -1.2810e-03,  2.6591e-03,  7.4449e-03, -6.0431e-03, -4.8712e-03,\n        -2.3451e-03, -1.0393e-03,  4.2036e-03,  6.1362e-03, -8.8334e-04,\n         3.3778e-03,  3.7556e-04, -1.8910e-03,  3.6084e-03,  2.9121e-03,\n         3.1919e-03, -1.4548e-03,  1.4200e-03,  7.5153e-03,  4.1686e-03,\n        -1.8493e-03,  2.3320e-03, -3.0653e-03,  5.1124e-03, -3.6398e-03,\n        -2.4768e-03,  4.5905e-04, -1.2641e-03, -5.7746e-03,  6.1046e-03,\n        -6.3901e-03, -2.5523e-04, -8.4340e-03,  5.5724e-05, -2.3499e-03,\n        -2.7910e-03, -5.7836e-03,  7.4894e-05, -3.2481e-03,  3.9726e-03,\n         9.2060e-03, -9.2005e-04,  1.4906e-05,  2.1199e-03, -1.8445e-03,\n        -1.7948e-03, -3.5234e-03,  2.0883e-03, -5.0710e-04, -2.5717e-04,\n         4.3037e-03,  4.4966e-03, -7.7924e-03,  8.6677e-03, -1.8408e-03,\n        -6.1555e-03, -7.7540e-03,  7.9262e-03, -2.6324e-03, -4.2165e-03,\n        -5.1994e-03, -3.3296e-04, -1.4055e-03,  3.8356e-03,  7.1755e-04,\n         7.0362e-03,  9.2058e-05,  2.9546e-03,  6.3822e-03, -5.6428e-03,\n         6.6616e-06,  1.9882e-03, -6.6812e-03,  8.3475e-04, -9.3955e-04,\n        -4.1878e-03, -7.2331e-04,  2.2562e-03, -4.1220e-03, -3.4375e-03,\n        -5.5896e-03, -1.4944e-03,  3.9882e-03, -4.3493e-03,  6.8208e-04,\n        -1.3324e-03, -2.2253e-03,  2.3026e-03, -2.7456e-03, -1.8855e-04,\n         3.8285e-04,  9.8479e-04,  8.9096e-04, -4.9162e-04,  2.7539e-03,\n        -3.9258e-03,  3.0915e-03, -2.2254e-03, -1.7285e-04, -3.6156e-04,\n         4.0246e-04, -1.2701e-03,  3.6221e-03, -1.6354e-03,  5.2486e-03,\n         2.8508e-04, -3.7546e-03, -2.0131e-03,  1.8658e-03, -3.4003e-03,\n        -1.1812e-04,  3.5204e-04, -3.0869e-04,  5.7886e-05, -6.8708e-04,\n        -1.8284e-03, -2.0498e-05,  7.5628e-03, -2.7515e-03,  1.6203e-03,\n        -2.8042e-03,  1.7185e-03, -2.4798e-03, -2.6479e-03,  4.7997e-03,\n        -1.9016e-03, -1.7169e-03, -2.3570e-04, -4.6680e-03,  3.2771e-03,\n        -7.5168e-03,  3.7136e-03, -1.9205e-03,  1.9161e-03,  1.9826e-03,\n        -2.6310e-03, -3.8434e-03, -1.6103e-03,  2.9747e-03,  3.0572e-03,\n         1.7101e-04, -6.1491e-03,  4.8488e-03,  9.6064e-04, -8.0626e-04,\n         3.5130e-03, -4.1602e-04, -1.7930e-03,  4.2207e-04,  1.7642e-04,\n        -4.2503e-03,  2.1579e-03,  7.2297e-03, -1.8326e-03, -1.7935e-03,\n        -5.4874e-03,  7.1566e-03,  9.0907e-04, -5.6886e-04, -3.1875e-03,\n        -1.6131e-03, -1.7008e-03, -6.5143e-03,  8.3601e-03, -7.9876e-04,\n         6.4246e-04,  1.5012e-03, -1.2257e-03,  3.5219e-03, -3.7269e-03,\n        -2.1768e-03,  2.5210e-03,  2.6302e-03,  1.0311e-03,  2.0144e-03,\n         3.4861e-03,  1.9706e-04,  1.0585e-03, -4.7431e-03, -2.8216e-03,\n         3.5322e-03, -1.5365e-03,  1.5808e-03,  3.1544e-03, -2.5791e-03,\n        -4.0344e-03, -1.3644e-03,  2.4535e-04,  1.8168e-03, -1.0563e-03,\n         4.3836e-03,  1.2257e-03,  1.5054e-03,  4.7464e-03, -5.1708e-03,\n        -4.4317e-03,  8.9603e-04,  1.6012e-03, -3.1157e-03, -2.4605e-03,\n         1.2364e-03, -2.2034e-03,  1.4938e-05,  1.2810e-03, -7.4490e-03,\n        -1.2025e-03,  2.5513e-03,  1.4516e-03,  4.1075e-03,  1.3590e-03,\n         5.7191e-04, -5.7842e-03,  8.8870e-04,  1.4588e-03,  2.1709e-03,\n        -9.8725e-04,  3.1786e-03, -3.4432e-03, -4.9858e-03,  5.2325e-04,\n        -9.2348e-04,  2.0494e-03,  4.6476e-03, -2.5771e-03, -8.4505e-04,\n         1.8759e-03,  6.2245e-03,  5.6467e-03,  3.8403e-03,  7.7850e-03,\n        -1.3934e-03, -1.0567e-02, -4.6038e-03, -9.9003e-04,  3.1682e-03,\n         5.6918e-03,  1.1739e-04,  4.0534e-03,  5.6161e-03,  5.1312e-03,\n         1.2442e-04,  4.3313e-03,  9.0238e-04, -4.8393e-03, -6.0563e-03,\n         8.0289e-03, -7.2311e-04, -4.6000e-03, -9.5827e-04,  1.3059e-03,\n         1.0847e-03, -1.5818e-03, -2.0929e-04,  4.9184e-03,  5.2542e-04,\n        -1.1754e-03, -3.4989e-03,  3.5434e-03, -1.2947e-03,  9.8844e-04,\n         2.5973e-03, -3.8419e-03,  7.2049e-03,  2.6424e-03,  3.0021e-03,\n         3.5133e-03,  3.4665e-03,  2.6124e-04,  1.3058e-02,  2.2194e-03,\n         8.3854e-03,  2.7876e-03, -4.8036e-04,  1.2251e-04, -3.9441e-03,\n        -2.4125e-03, -1.4467e-02,  1.5160e-03,  7.4386e-03, -6.7232e-04,\n        -2.3544e-04, -1.3008e-04, -3.7047e-03,  1.1277e-03, -1.1985e-02,\n         1.3106e-04,  6.7225e-03], device='cuda:0')",
+        "exp_avg_sq": "tensor([2.9447e-04, 3.7317e-04, 3.0769e-04, 2.6471e-04, 2.3692e-04, 3.5314e-04,\n        2.1902e-04, 1.7906e-04, 2.1044e-04, 1.8192e-04, 2.1482e-04, 7.0766e-05,\n        2.9461e-04, 4.9959e-04, 2.0153e-04, 2.6225e-04, 1.9483e-04, 1.9536e-04,\n        2.5183e-04, 2.7501e-04, 2.1652e-04, 2.5576e-04, 3.0764e-04, 1.7653e-04,\n        4.6414e-04, 2.6101e-04, 2.4140e-04, 3.4430e-04, 1.7731e-04, 1.6433e-04,\n        2.6430e-04, 3.7979e-04, 2.1596e-04, 3.1081e-04, 3.7860e-04, 1.8742e-04,\n        3.6558e-04, 2.5975e-04, 1.4181e-04, 3.4011e-04, 2.9547e-04, 2.9818e-04,\n        2.1602e-04, 1.9651e-04, 2.8383e-04, 1.8665e-04, 2.1635e-04, 2.1205e-04,\n        1.6542e-04, 2.4786e-04, 7.4485e-05, 2.4915e-04, 2.5787e-04, 4.0114e-04,\n        1.8783e-04, 1.6829e-04, 1.8226e-04, 1.7397e-04, 1.8009e-04, 2.4305e-04,\n        2.5268e-04, 2.0510e-04, 1.9335e-04, 4.6825e-05, 2.3267e-04, 2.9784e-04,\n        3.1968e-04, 3.3130e-04, 2.9275e-04, 2.5763e-04, 3.5427e-04, 2.9196e-04,\n        2.5550e-04, 2.9405e-04, 2.2513e-04, 3.1311e-04, 2.9520e-04, 2.1316e-04,\n        2.6201e-04, 2.3898e-04, 3.5285e-04, 1.6689e-04, 1.1688e-04, 3.1497e-04,\n        2.7398e-04, 2.7224e-04, 3.9858e-04, 3.3022e-04, 3.1581e-05, 2.1734e-04,\n        2.1575e-04, 3.0422e-04, 2.3974e-04, 2.0378e-04, 1.8246e-04, 2.2434e-04,\n        2.8697e-04, 8.9937e-05, 2.6815e-04, 1.4957e-04, 2.0106e-04, 2.4685e-04,\n        2.0805e-04, 2.6720e-04, 2.6514e-04, 2.5647e-04, 2.1405e-04, 2.5724e-04,\n        2.5247e-04, 2.3343e-04, 1.7841e-04, 2.4671e-04, 2.3688e-04, 4.2637e-04,\n        1.9855e-04, 2.0871e-04, 2.7769e-04, 2.8097e-04, 1.8511e-04, 2.8667e-04,\n        2.1273e-04, 2.6418e-04, 2.1942e-04, 2.1062e-04, 1.4386e-04, 2.2347e-04,\n        3.0800e-04, 2.4416e-04, 2.4733e-04, 1.5518e-04, 2.2140e-04, 2.5184e-04,\n        2.5104e-04, 2.2683e-04, 2.1679e-04, 2.9015e-04, 2.6186e-04, 2.5157e-04,\n        2.6475e-04, 1.8260e-04, 3.2228e-04, 1.7744e-04, 2.9755e-04, 2.0920e-04,\n        2.3092e-04, 3.1963e-04, 2.3765e-04, 1.6940e-04, 3.5729e-04, 2.9008e-04,\n        2.6877e-04, 1.7373e-04, 2.9628e-04, 1.6400e-04, 2.5480e-04, 2.1375e-04,\n        2.7598e-04, 2.4039e-04, 2.9941e-04, 3.1055e-04, 4.2870e-04, 2.1161e-04,\n        2.1921e-04, 3.6805e-04, 1.3545e-04, 2.0458e-04, 3.6451e-04, 2.1763e-04,\n        2.1920e-04, 2.2399e-04, 2.1450e-04, 2.7571e-04, 2.4447e-04, 3.1175e-04,\n        1.9950e-04, 3.2276e-04, 2.2346e-04, 2.0164e-04, 2.0836e-04, 3.0490e-04,\n        5.9661e-04, 2.8911e-04, 1.5643e-04, 2.4104e-04, 3.6681e-04, 2.0591e-04,\n        3.2098e-04, 2.5081e-04, 2.1218e-04, 2.2976e-04, 2.5976e-04, 2.9883e-04,\n        2.4033e-04, 2.8229e-04, 2.7729e-04, 2.4350e-04, 2.2417e-04, 1.7924e-04,\n        2.8155e-04, 1.9334e-04, 2.9187e-04, 1.4910e-04, 1.8579e-04, 2.5475e-04,\n        1.7988e-04, 2.6229e-04, 2.7908e-04, 2.1525e-04, 2.1232e-04, 3.4535e-04,\n        2.0509e-04, 2.0875e-04, 2.6197e-04, 2.0895e-04, 3.1091e-04, 2.2663e-04,\n        3.0048e-04, 1.3750e-04, 1.4466e-04, 2.5741e-04, 2.6456e-04, 4.9971e-04,\n        2.5545e-04, 1.5469e-04, 2.6366e-04, 2.3993e-04, 1.9428e-04, 2.2426e-04,\n        2.2476e-04, 2.1300e-04, 1.8702e-04, 3.4476e-04, 1.6692e-04, 2.0065e-04,\n        1.9138e-04, 1.9048e-04, 1.9580e-04, 3.6866e-04, 2.9075e-04, 2.0295e-04,\n        2.8478e-04, 2.4457e-04, 2.2804e-04, 1.5773e-04, 2.2214e-04, 3.8855e-04,\n        2.2255e-04, 2.2339e-04, 1.1732e-04, 2.4928e-04, 2.5387e-04, 1.7706e-04,\n        2.4940e-04, 2.6220e-04, 2.5909e-04, 2.1448e-04, 2.5817e-04, 2.6417e-04,\n        2.6663e-04, 2.7899e-04, 2.5987e-04, 1.6427e-04, 2.3980e-04, 2.4583e-04,\n        2.1500e-04, 2.5805e-04, 2.3381e-04, 2.8162e-04, 2.3576e-04, 2.4813e-04,\n        2.2833e-04, 1.6486e-04, 2.6447e-04, 2.3271e-04, 2.5335e-04, 1.4849e-04,\n        1.9227e-04, 1.6196e-04, 2.6197e-04, 2.0174e-04, 2.5599e-04, 2.2340e-04,\n        2.6386e-04, 2.1278e-04, 2.4264e-04, 3.0364e-04, 2.7725e-04, 2.1635e-04,\n        2.1511e-04, 1.2012e-04, 2.2719e-04, 4.1056e-04, 3.2885e-04, 2.2142e-04,\n        2.8487e-04, 2.0930e-04, 3.6140e-04, 2.1923e-04, 2.7760e-04, 1.5732e-04,\n        2.2037e-04, 1.1846e-04, 2.0022e-04, 2.4561e-04, 2.4144e-04, 2.9667e-04,\n        2.6137e-04, 1.5714e-04, 2.7525e-04, 2.4622e-04, 1.2938e-04, 2.0910e-04,\n        2.5203e-04, 1.9712e-04, 1.8866e-04, 2.4993e-04, 1.6358e-04, 2.2596e-04,\n        1.7336e-04, 2.8070e-04, 1.8984e-04, 2.1483e-04, 3.4745e-04, 5.0253e-04,\n        2.3227e-04, 2.1675e-04, 2.4302e-04, 2.8549e-04, 1.8574e-04, 2.1854e-04,\n        2.1350e-04, 2.9515e-04, 2.6781e-04, 2.3115e-04, 2.0208e-04, 2.7666e-04,\n        1.8419e-04, 1.8013e-04, 3.1309e-04, 2.5983e-04, 2.3039e-04, 2.4238e-04,\n        2.2123e-04, 2.6165e-04, 3.0882e-04, 2.7453e-04, 2.3319e-04, 2.5678e-04,\n        2.2072e-04, 2.0204e-04, 2.3164e-04, 1.9436e-04, 1.6204e-04, 2.5961e-04,\n        3.0348e-04, 2.1758e-04, 1.6803e-04, 2.9268e-04, 2.3556e-04, 1.6266e-04,\n        2.9910e-04, 1.6006e-04, 2.9268e-04, 1.7687e-04, 2.9815e-04, 3.4582e-04,\n        2.1822e-04, 2.1002e-04, 2.6928e-04, 1.6722e-04, 4.2373e-04, 2.2255e-04,\n        2.5697e-04, 1.8589e-04, 2.8008e-04, 4.1086e-04, 2.6616e-04, 2.3654e-04,\n        2.9476e-04, 1.3487e-04, 2.1656e-04, 2.2312e-04, 4.9007e-04, 1.7495e-04,\n        1.7244e-04, 1.7352e-04, 1.9052e-04, 4.1044e-04, 2.6560e-04, 3.4738e-04,\n        2.9212e-04, 1.4712e-04, 2.8646e-04, 2.7514e-04, 2.3274e-04, 2.4003e-04,\n        3.1161e-04, 1.7938e-04, 2.0162e-04, 2.7117e-04, 2.5148e-04, 3.8533e-04,\n        2.1721e-04, 2.0932e-04, 2.0759e-04, 3.8381e-04, 2.1099e-04, 1.9811e-04,\n        2.0648e-04, 1.9183e-04, 2.6736e-04, 2.6825e-04, 2.4064e-04, 3.0336e-04,\n        2.8484e-04, 1.8200e-04, 2.9309e-04, 3.1992e-04, 1.4031e-04, 3.3981e-04,\n        1.9319e-04, 2.5011e-04, 2.1276e-04, 2.3967e-04, 2.6501e-04, 2.9038e-04,\n        2.3775e-04, 2.3574e-04, 2.3021e-04, 3.6044e-04, 1.9675e-04, 2.4050e-04,\n        1.9979e-04, 1.3166e-04, 2.2467e-04, 2.0088e-04, 2.2093e-04, 2.4106e-04,\n        2.4315e-04, 2.2263e-04, 2.8154e-04, 2.4672e-04, 3.0070e-04, 2.3575e-04,\n        3.2954e-04, 2.0570e-04, 1.9560e-04, 1.7775e-04, 2.6143e-04, 1.7553e-04,\n        2.6129e-04, 2.3124e-04, 2.8916e-04, 1.6066e-04, 3.2773e-04, 1.9670e-04,\n        2.1148e-04, 1.6714e-04, 2.3379e-04, 1.7128e-04, 2.0169e-04, 1.8986e-04,\n        1.4896e-04, 2.1704e-04, 3.1252e-04, 2.1416e-04, 3.4009e-04, 2.8807e-04,\n        3.5600e-04, 2.5140e-04, 4.1171e-04, 1.6049e-04, 1.9790e-04, 2.7214e-04,\n        2.2382e-04, 2.5065e-04, 1.2592e-04, 2.2314e-04, 3.1649e-04, 2.5426e-04,\n        3.4588e-04, 3.3344e-04, 2.5931e-04, 1.9958e-04, 2.1405e-04, 2.5943e-04,\n        3.7225e-04, 3.8034e-04, 2.4375e-04, 2.0339e-04, 2.2120e-04, 1.9557e-04,\n        2.3890e-04, 2.9015e-04, 2.8212e-04, 3.0234e-04, 1.7472e-04, 2.2272e-04,\n        2.5042e-04, 9.4893e-05, 3.0954e-04, 3.1036e-04, 1.9061e-04, 1.9346e-04,\n        2.1624e-04, 2.6488e-04, 1.9919e-04, 2.4612e-04, 2.5312e-04, 2.5250e-04,\n        2.5765e-04, 2.4213e-04], device='cuda:0')"
       },
       "2": {
-        "step": "tensor(12520.)",
-        "exp_avg": "tensor([[-2.7164e-05, -1.9900e-07,  9.5126e-07,  ...,  9.4927e-06,\n          5.6052e-45, -9.6058e-07],\n        [-2.7432e-05,  5.1023e-07, -2.3293e-06,  ...,  2.8908e-07,\n         -5.6052e-45, -2.5127e-06],\n        [-3.1435e-07, -3.2044e-07, -1.2302e-07,  ...,  2.4341e-06,\n         -5.6052e-45, -1.9378e-07],\n        ...,\n        [ 5.8933e-07,  1.5870e-07, -7.7837e-07,  ...,  9.6177e-07,\n          5.6052e-45,  2.0668e-06],\n        [ 3.8947e-06, -6.5624e-09, -9.0244e-06,  ..., -9.2989e-07,\n         -5.6052e-45,  6.2904e-06],\n        [-1.1632e-05, -7.0983e-06,  8.5217e-06,  ...,  1.5592e-06,\n          5.6052e-45,  2.1561e-06]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[1.9624e-09, 2.2615e-10, 1.6196e-09,  ..., 6.6825e-09, 4.7686e-16,\n         6.7964e-10],\n        [2.0070e-09, 7.9523e-10, 4.2056e-10,  ..., 2.2196e-10, 2.6458e-14,\n         4.2702e-09],\n        [2.3410e-10, 3.6115e-10, 3.2898e-10,  ..., 6.0066e-10, 1.8233e-13,\n         1.4097e-09],\n        ...,\n        [4.7294e-11, 4.8290e-11, 5.5586e-10,  ..., 7.2295e-11, 1.8310e-14,\n         1.7603e-11],\n        [5.4413e-10, 2.2035e-11, 1.9991e-10,  ..., 3.0986e-10, 3.6935e-17,\n         4.8808e-10],\n        [3.3296e-09, 3.1063e-10, 6.0098e-09,  ..., 1.7746e-09, 3.2723e-15,\n         2.6791e-09]], device='cuda:0')"
+        "step": "tensor(1252.)",
+        "exp_avg": "tensor([ 3.4184e-03,  2.7278e-03, -2.8081e-03, -3.7930e-04, -9.7510e-05,\n        -5.8738e-03, -3.5282e-04,  8.3442e-04, -1.2154e-04, -1.4573e-05,\n        -7.2452e-04, -5.6543e-03,  2.9131e-03,  4.6260e-03, -3.5767e-03,\n         3.2509e-03, -1.7123e-03,  1.8318e-03,  2.6752e-03,  1.0468e-03,\n        -1.7199e-03, -9.7654e-04, -4.8754e-04,  3.3165e-04, -4.9488e-03,\n         3.2316e-03,  5.2018e-03,  3.1327e-03,  5.5856e-04, -6.2779e-03,\n        -2.9570e-03, -4.3141e-03, -2.1774e-03, -8.0216e-04, -1.7261e-03,\n         1.9607e-03,  6.1659e-04, -1.7345e-03, -8.7239e-04,  2.6505e-03,\n        -1.9765e-03,  3.8879e-03, -5.6384e-03,  2.5214e-03, -2.8107e-03,\n        -5.8875e-04,  3.6451e-03, -3.3639e-03, -1.0893e-02,  4.5255e-04,\n        -1.0662e-02,  2.5433e-03, -2.1356e-03, -1.4398e-03, -3.8268e-03,\n        -2.1625e-03,  1.3507e-03,  1.1380e-03, -1.7401e-03,  2.9767e-03,\n        -5.9497e-03,  4.2108e-03, -9.8701e-03, -1.4374e-02, -4.0546e-03,\n        -5.2263e-04,  2.0749e-03,  2.9863e-03, -4.8733e-03, -3.3797e-03,\n         1.5042e-04, -4.2552e-03, -2.0624e-03, -5.4418e-03, -2.4114e-03,\n         4.1112e-03, -7.0112e-04, -2.8896e-03, -8.6019e-03,  7.4896e-03,\n         2.8952e-03, -5.7825e-03,  1.1835e-03, -2.2995e-03, -6.0511e-03,\n         6.4554e-03, -2.5340e-03,  3.3304e-03,  9.3410e-24, -4.2280e-04,\n        -4.2691e-03,  1.9652e-03,  3.7844e-03, -3.1400e-03, -3.8535e-03,\n        -6.0053e-03, -1.4311e-03,  3.9327e-03,  1.1910e-03, -1.3046e-03,\n         2.9090e-03,  5.5334e-04, -1.6887e-03,  1.0477e-02, -4.0227e-03,\n        -1.0192e-03,  2.6573e-03,  1.7726e-03, -1.3745e-03,  1.7060e-03,\n         2.8598e-03, -1.4818e-03,  3.3605e-03,  7.5695e-04, -8.2858e-04,\n         3.1565e-03,  1.7958e-03,  4.4471e-03, -6.0539e-03,  1.8275e-03,\n         1.9078e-03, -9.9236e-04, -6.8147e-03, -3.6253e-03, -5.9307e-03,\n        -1.5895e-03, -6.1906e-04, -3.9996e-04, -3.5084e-03,  2.0870e-03,\n         3.9427e-03,  1.8046e-03, -9.7865e-03,  7.4494e-03,  3.2588e-03,\n         1.3857e-04,  1.2492e-03, -2.3269e-03, -5.7836e-04, -2.5859e-03,\n        -1.1721e-03, -4.7796e-03,  4.5991e-03,  4.0910e-03,  3.1960e-03,\n         1.6916e-03, -2.0877e-03,  4.2816e-04, -1.2278e-03,  5.7702e-03,\n        -2.1622e-03,  4.3127e-03,  5.3876e-03, -2.1955e-03,  4.1135e-03,\n         4.2805e-03,  4.7634e-04, -2.3241e-03,  2.9025e-04, -2.4137e-03,\n         2.1759e-03, -1.7161e-03, -1.0870e-02,  3.1825e-04, -8.1508e-04,\n        -2.3249e-03, -2.5426e-03,  8.8758e-04,  9.8898e-04, -3.9451e-03,\n        -1.0564e-03, -2.8019e-03,  1.0875e-03,  3.1644e-03, -5.1027e-03,\n        -1.8086e-03, -3.1807e-03, -6.8280e-03,  5.5181e-03, -2.6287e-03,\n        -1.6519e-03,  1.7492e-03,  8.1751e-04, -2.9324e-03, -1.3657e-03,\n         4.1177e-04,  3.4169e-04, -1.5535e-03,  7.0645e-04,  2.8915e-03,\n        -2.1121e-03,  6.9193e-03,  1.0620e-03, -5.3412e-03,  3.1544e-03,\n         1.6713e-04,  3.7178e-03,  2.4542e-04,  7.8458e-04,  3.8568e-03,\n        -3.0821e-03,  9.5915e-03, -1.5608e-03, -3.7024e-03,  2.0204e-03,\n         1.9235e-03, -1.9786e-04,  2.6449e-03, -5.6376e-04,  1.9634e-03,\n         4.2661e-04,  4.1795e-03, -1.9989e-03, -3.4974e-03,  3.4327e-03,\n        -3.0713e-03, -3.0610e-03, -4.2239e-03,  8.4635e-03,  8.4894e-04,\n        -3.4265e-03,  5.5908e-03, -6.1383e-04, -6.0085e-03, -2.0861e-03,\n        -1.0919e-03,  6.7907e-04,  5.1914e-03, -6.1674e-03, -3.4714e-03,\n        -1.8022e-03,  2.0030e-04,  4.3040e-03,  7.1921e-03, -3.2461e-05,\n         1.2077e-03, -3.2985e-04, -3.6048e-06,  3.1595e-03,  2.3927e-03,\n         2.5950e-03, -1.8790e-04,  1.1631e-03,  6.6310e-03,  4.0112e-03,\n        -2.5934e-04,  2.1013e-03, -2.1607e-03,  7.7924e-03, -1.9381e-03,\n        -1.6783e-03, -2.0762e-04, -1.1051e-03, -1.7680e-03,  3.6529e-03,\n        -8.4169e-03,  1.6156e-04, -6.3086e-03, -2.6614e-04,  7.0708e-04,\n        -3.6475e-03, -8.1781e-03,  7.7157e-04, -2.2509e-03,  3.5293e-03,\n         9.6958e-03, -2.7979e-03, -4.2822e-04,  1.7815e-03, -1.2651e-03,\n        -1.5099e-03, -3.5986e-03,  4.2769e-04, -9.5290e-04, -2.3294e-04,\n         9.9919e-03,  3.6278e-03, -6.3853e-03,  8.4500e-03, -2.1056e-03,\n        -3.9677e-03, -5.5748e-03,  1.0708e-02, -1.2806e-03, -4.4282e-03,\n        -3.9635e-03, -1.0118e-03, -1.6587e-03,  3.4588e-03,  5.7744e-04,\n         3.8342e-03, -7.2854e-04,  1.3008e-03,  3.9695e-03, -1.9321e-03,\n         6.5650e-04,  1.5987e-03, -9.8732e-03,  1.9962e-03, -1.6643e-03,\n        -2.7277e-03, -3.1865e-03,  2.3558e-03, -2.1843e-03, -2.1195e-03,\n        -3.9005e-03, -2.1801e-03,  3.3427e-03, -2.4026e-03,  2.8996e-04,\n        -2.6050e-03, -2.3256e-03,  6.1841e-04, -1.9165e-03,  3.8676e-04,\n         8.1380e-04,  2.2026e-03,  2.2067e-03, -9.1111e-04,  2.0519e-03,\n        -2.7911e-03,  2.1624e-03, -1.9052e-03,  2.6590e-04, -1.7145e-03,\n         3.5091e-04, -5.3331e-04,  1.4208e-03, -1.2433e-03,  7.0732e-03,\n         2.7985e-04, -2.8940e-03, -3.2422e-03,  1.4449e-03, -2.3747e-03,\n        -6.8386e-04,  1.1747e-03, -5.4422e-04, -4.8600e-04, -3.2378e-04,\n        -6.7249e-04, -2.1946e-04,  7.3253e-03, -1.2538e-03,  2.3842e-03,\n        -2.3401e-03, -5.3134e-05, -1.6592e-03, -1.4258e-03,  5.6559e-03,\n        -2.6921e-03, -2.3515e-03,  2.0870e-03, -2.5880e-03,  1.0508e-03,\n        -7.8608e-03,  4.5714e-03, -2.1934e-03,  6.4511e-04,  1.5744e-03,\n        -2.0696e-03, -2.0573e-03, -2.3358e-03,  3.0994e-03,  1.9378e-03,\n         7.4438e-05, -3.9586e-03,  4.2065e-03,  2.8482e-04, -1.8266e-03,\n         9.0880e-04, -9.8569e-04, -2.0896e-03,  1.1372e-03, -7.6679e-05,\n        -3.0742e-03,  6.8614e-04,  8.5280e-03, -1.0057e-03, -3.8477e-03,\n        -7.4526e-03,  4.5727e-03,  1.0802e-03, -4.7644e-05, -2.7261e-03,\n        -4.9294e-03, -4.3644e-04, -5.4936e-03,  6.1642e-03, -1.3486e-03,\n         3.3867e-04,  2.8851e-03, -7.5389e-04,  2.2820e-03, -2.3219e-03,\n         1.9179e-04,  3.7242e-03,  2.1815e-03,  5.2975e-04,  1.3971e-03,\n         2.3667e-03,  3.7373e-04,  4.2356e-04, -2.1948e-03, -3.3116e-03,\n         3.7259e-03, -4.7227e-04,  1.9993e-03,  1.3587e-03, -3.2313e-03,\n        -4.7099e-03, -1.3288e-03, -5.2207e-04, -1.1481e-04, -2.8355e-04,\n         5.9829e-03, -1.2657e-04,  8.9951e-04,  1.1720e-02, -2.6923e-03,\n        -5.7581e-03,  1.1089e-03,  1.7833e-03, -4.4792e-03, -7.0931e-04,\n         3.6109e-04, -1.5927e-03, -1.3835e-04,  8.8109e-04, -3.3068e-03,\n        -2.6530e-04,  2.1425e-03,  4.5240e-04,  1.1654e-02,  1.9390e-03,\n         1.8187e-03, -6.5379e-03,  2.2464e-03,  9.8832e-04,  4.1459e-03,\n        -9.2557e-04,  2.2297e-03, -2.1348e-03, -4.3717e-03, -2.4845e-04,\n        -1.0719e-03,  1.2796e-03,  4.5842e-03, -2.7041e-03, -2.1959e-03,\n         3.0345e-03,  3.1722e-03,  4.3686e-03,  5.7541e-03,  4.7023e-03,\n        -1.8333e-03, -8.9308e-03, -4.5921e-03, -8.9954e-04,  1.8643e-03,\n         4.4605e-03,  1.5317e-04,  5.4704e-03,  4.0192e-03,  3.7174e-03,\n        -7.3009e-05,  3.2739e-03,  1.4126e-03, -2.3728e-03, -4.6553e-03,\n         4.9287e-03, -1.8731e-04, -2.3757e-03, -1.3770e-03,  5.6579e-04,\n         2.0683e-03, -4.7602e-03,  1.3428e-03,  2.3574e-03,  1.2990e-04,\n        -4.4217e-04, -2.0124e-03,  2.9274e-03, -6.2379e-04,  8.5436e-04,\n         1.5156e-03, -1.8395e-03,  3.0566e-03, -4.1310e-04,  2.4849e-03,\n         3.2003e-03,  4.2600e-03, -1.3337e-04,  8.7607e-03,  1.2232e-03,\n         5.4495e-03,  1.4905e-03, -5.0086e-04, -8.5565e-05, -2.1486e-03,\n        -2.1281e-03, -7.0873e-03,  7.3244e-04,  8.8575e-03, -4.5805e-04,\n         6.6243e-04,  1.8176e-04, -2.2098e-03,  5.9491e-04, -7.4454e-03,\n        -3.3656e-05,  4.0627e-03], device='cuda:0')",
+        "exp_avg_sq": "tensor([3.9600e-05, 6.4381e-05, 4.8072e-05, 7.3210e-05, 5.8307e-05, 5.3678e-05,\n        3.3623e-05, 4.5860e-05, 6.6022e-05, 6.5272e-05, 5.1552e-05, 1.6738e-03,\n        5.2756e-05, 6.7838e-05, 1.2087e-04, 4.4350e-05, 1.1770e-04, 1.1252e-04,\n        9.6078e-05, 4.4455e-05, 5.1236e-05, 5.0755e-05, 5.2306e-05, 2.5199e-05,\n        7.2600e-05, 3.9548e-05, 7.0392e-05, 6.6353e-05, 8.1098e-05, 1.3693e-04,\n        5.3729e-05, 5.8445e-05, 3.8029e-05, 3.3524e-05, 8.6110e-05, 5.3107e-05,\n        7.8084e-05, 1.0151e-04, 6.6539e-05, 3.3441e-05, 3.6012e-05, 6.6207e-05,\n        5.9212e-05, 6.1578e-05, 4.6035e-05, 4.0579e-05, 6.2152e-05, 1.1118e-04,\n        1.5289e-04, 3.4514e-05, 6.2806e-04, 1.8387e-04, 6.7518e-05, 4.5748e-05,\n        9.9997e-05, 6.7684e-05, 1.0945e-04, 7.7932e-05, 1.1758e-04, 1.1141e-04,\n        1.0643e-04, 4.6503e-05, 1.5542e-04, 5.3542e-03, 7.3961e-05, 8.1964e-05,\n        6.8237e-05, 4.8917e-05, 2.8945e-05, 1.1846e-04, 4.3508e-05, 9.3928e-05,\n        6.4102e-05, 6.5294e-05, 6.3835e-05, 1.5793e-04, 3.6490e-05, 1.2651e-04,\n        8.1779e-05, 1.2541e-04, 7.8363e-05, 3.5845e-04, 4.1844e-04, 6.2583e-05,\n        7.9322e-05, 4.8535e-05, 5.5934e-05, 9.8612e-05, 3.3677e-04, 1.4057e-04,\n        4.2436e-05, 1.0143e-04, 1.3029e-04, 6.4883e-05, 1.6558e-04, 5.1201e-05,\n        4.1590e-05, 6.5359e-04, 4.6331e-05, 9.3253e-05, 1.1327e-04, 3.7158e-05,\n        5.5202e-05, 1.3847e-04, 4.9189e-05, 3.4487e-05, 4.1969e-05, 3.4024e-05,\n        4.2403e-05, 7.0675e-05, 9.9015e-05, 6.3951e-05, 4.4300e-05, 4.6244e-05,\n        5.7491e-05, 6.9334e-05, 4.3085e-05, 4.1878e-05, 4.6723e-05, 4.4373e-05,\n        9.5254e-05, 4.3977e-05, 7.7592e-05, 1.1402e-04, 9.7114e-05, 3.4156e-05,\n        3.1799e-05, 2.5546e-05, 4.0581e-05, 1.3313e-04, 4.6832e-05, 7.4642e-05,\n        1.6675e-04, 1.5696e-04, 1.9662e-04, 1.2953e-04, 7.6995e-05, 7.9739e-05,\n        6.8174e-05, 1.1657e-04, 5.0580e-05, 9.7171e-05, 5.0731e-05, 9.1917e-05,\n        5.1409e-05, 6.6388e-05, 1.3097e-04, 3.2235e-05, 5.1695e-05, 7.5243e-05,\n        7.2492e-05, 1.6994e-04, 5.8420e-05, 9.5364e-05, 4.1267e-05, 5.1751e-05,\n        6.2857e-05, 3.1941e-05, 9.0395e-05, 5.2793e-05, 5.2425e-05, 4.1051e-05,\n        1.4754e-04, 5.2407e-05, 3.2314e-05, 8.6912e-05, 4.6841e-05, 6.0142e-05,\n        6.2479e-05, 1.5286e-04, 5.5130e-05, 6.3721e-05, 8.8627e-05, 6.0204e-05,\n        1.6831e-04, 6.3566e-05, 5.8835e-05, 1.4341e-04, 7.0422e-05, 1.0191e-04,\n        6.5399e-05, 7.5551e-05, 1.0321e-04, 1.9792e-04, 6.3548e-05, 3.7266e-05,\n        5.0458e-05, 1.7176e-04, 5.3855e-05, 6.5435e-05, 4.2406e-05, 7.1112e-05,\n        1.0644e-04, 4.4769e-05, 6.3218e-05, 3.7588e-05, 1.1528e-04, 5.0218e-05,\n        3.4320e-05, 4.0212e-05, 7.1630e-05, 2.3345e-04, 1.0160e-04, 5.7749e-05,\n        6.5414e-05, 7.7594e-05, 6.8500e-05, 8.7619e-05, 2.7256e-05, 3.0500e-05,\n        4.7310e-05, 1.5539e-04, 7.5878e-05, 5.2446e-05, 5.9303e-05, 5.8812e-05,\n        3.8864e-05, 8.8109e-05, 1.5847e-04, 4.4937e-05, 6.9018e-05, 6.3549e-05,\n        7.0232e-05, 1.0525e-04, 1.0612e-04, 2.1898e-05, 5.4272e-05, 6.6958e-05,\n        5.5488e-05, 4.7423e-05, 2.6731e-05, 4.2092e-05, 1.1603e-04, 1.1945e-04,\n        9.9042e-05, 4.9842e-05, 1.1139e-04, 2.1082e-04, 4.2075e-05, 3.6654e-05,\n        5.6222e-05, 2.1926e-05, 4.7523e-05, 6.8757e-05, 4.5510e-05, 6.2872e-05,\n        3.4599e-05, 8.1056e-05, 1.6585e-04, 5.0132e-05, 1.0609e-04, 8.7736e-05,\n        4.5175e-05, 6.0189e-05, 5.0129e-05, 7.2468e-05, 2.9374e-05, 1.1445e-04,\n        4.2141e-05, 6.4726e-05, 8.2706e-05, 1.1134e-04, 8.2270e-05, 3.5101e-05,\n        1.2360e-04, 1.1101e-04, 7.6243e-05, 2.6201e-05, 6.9142e-05, 8.7068e-05,\n        5.0850e-05, 5.8678e-05, 7.8472e-05, 5.6535e-05, 5.1873e-05, 3.5081e-04,\n        9.1862e-05, 8.6318e-05, 9.7763e-05, 8.8791e-05, 7.3812e-05, 6.6584e-05,\n        2.7497e-04, 7.2667e-05, 7.9735e-05, 7.9080e-05, 4.1959e-05, 7.7589e-05,\n        1.4417e-04, 2.1497e-04, 4.3115e-05, 8.6184e-05, 4.9091e-05, 2.4555e-05,\n        6.9723e-05, 4.4433e-05, 8.1440e-05, 1.4370e-04, 3.3483e-05, 1.8678e-04,\n        5.2311e-05, 3.3757e-04, 3.5898e-05, 4.0793e-05, 3.8237e-05, 7.2866e-05,\n        1.0356e-04, 9.9319e-05, 2.8778e-05, 5.0912e-05, 1.2630e-04, 6.1348e-05,\n        2.9934e-05, 7.3993e-05, 3.4409e-05, 3.8406e-05, 1.1915e-04, 1.0946e-04,\n        7.9104e-05, 7.8837e-05, 9.2321e-05, 9.0655e-05, 6.2748e-05, 8.3371e-05,\n        1.8207e-04, 2.5476e-05, 5.4799e-05, 5.5890e-05, 6.8994e-05, 1.0420e-04,\n        5.1943e-05, 1.0004e-04, 1.2829e-04, 5.1255e-05, 5.0546e-05, 4.9478e-05,\n        3.7562e-05, 6.2637e-05, 3.7131e-05, 7.6964e-05, 1.0186e-04, 7.6213e-05,\n        8.2444e-05, 3.8991e-05, 2.9221e-05, 1.1639e-04, 1.1740e-04, 4.2628e-05,\n        4.2042e-05, 1.2221e-04, 5.6198e-05, 1.1700e-04, 1.7684e-04, 3.1066e-05,\n        4.0916e-05, 1.3970e-04, 8.1107e-05, 4.8749e-05, 1.1592e-04, 4.6130e-05,\n        8.5357e-05, 3.1174e-05, 3.6626e-05, 5.8383e-05, 4.2261e-05, 1.7329e-05,\n        3.4454e-05, 1.4393e-04, 7.9968e-05, 5.0011e-05, 7.5636e-05, 8.6797e-05,\n        3.5984e-05, 1.0899e-04, 1.1936e-04, 1.3840e-04, 3.6028e-05, 1.1247e-04,\n        4.4691e-05, 1.2191e-04, 1.6792e-04, 4.7551e-05, 1.2493e-04, 1.0150e-04,\n        1.1795e-04, 1.9566e-04, 6.4463e-05, 6.6510e-05, 9.6799e-05, 6.3704e-05,\n        5.0194e-05, 4.1368e-04, 3.9959e-05, 5.9155e-05, 8.5753e-05, 6.9573e-05,\n        3.8492e-05, 1.5175e-04, 1.2351e-04, 2.3835e-05, 7.4126e-05, 5.0864e-05,\n        1.1487e-04, 3.7643e-05, 5.1128e-05, 1.2279e-04, 3.2085e-05, 9.8932e-05,\n        4.4380e-05, 1.1838e-04, 1.2962e-04, 1.0192e-04, 3.6317e-05, 5.4738e-05,\n        5.0185e-05, 9.4576e-05, 6.7493e-05, 6.2044e-05, 3.4287e-04, 8.8081e-05,\n        9.4533e-05, 1.4711e-04, 7.8249e-05, 9.6513e-05, 3.7715e-05, 1.0235e-04,\n        5.6602e-05, 1.0007e-04, 8.3447e-05, 4.5039e-05, 2.8902e-05, 7.1463e-05,\n        5.6537e-05, 5.7301e-04, 6.0649e-05, 8.6062e-05, 1.5073e-04, 6.3094e-05,\n        9.3308e-05, 1.4244e-04, 1.5843e-04, 5.0681e-05, 5.2643e-05, 6.1280e-05,\n        6.1522e-05, 1.0554e-04, 9.7306e-05, 9.9238e-05, 6.0242e-05, 1.5380e-04,\n        2.1723e-04, 3.8714e-05, 5.0054e-05, 1.3878e-04, 6.4378e-05, 1.4719e-04,\n        6.3152e-05, 1.0181e-04, 8.8848e-05, 4.8031e-05, 6.6484e-05, 1.8200e-04,\n        1.0909e-04, 7.4998e-05, 7.3979e-05, 1.2405e-04, 7.9479e-05, 1.0843e-04,\n        6.2691e-05, 8.6563e-05, 1.2653e-04, 4.7224e-05, 2.3525e-05, 4.2260e-05,\n        8.6819e-05, 5.7005e-05, 2.1489e-04, 1.0874e-04, 4.7308e-05, 5.4752e-05,\n        4.1583e-05, 8.1988e-05, 8.0489e-05, 2.4834e-05, 4.5631e-05, 7.1088e-05,\n        5.5007e-05, 3.7490e-05, 4.9179e-05, 6.5010e-05, 9.8970e-05, 6.6629e-05,\n        4.9302e-05, 8.1281e-05, 3.8743e-05, 8.2108e-05, 4.7178e-05, 5.1445e-05,\n        2.9104e-05, 1.4734e-05, 7.3063e-05, 4.4081e-05, 5.5845e-05, 1.2687e-04,\n        3.0537e-05, 5.5340e-05, 9.5501e-05, 5.7617e-05, 9.7333e-05, 7.5231e-05,\n        9.8795e-05, 7.6599e-05], device='cuda:0')"
       },
       "3": {
-        "step": "tensor(12520.)",
-        "exp_avg": "tensor([-2.3097e-05,  9.5465e-05, -8.4887e-05,  1.8510e-04,  1.3161e-04,\n        -2.0138e-05, -1.4139e-04,  3.0738e-05, -1.0147e-04,  1.1279e-04,\n         5.3551e-04, -5.0594e-05,  2.7870e-05, -2.8186e-05, -1.8885e-04,\n         1.8148e-04, -1.8156e-05, -1.8771e-05,  1.0264e-04,  4.3141e-05,\n         1.0310e-05, -9.1166e-05,  7.8901e-05, -3.3479e-06,  5.7091e-05,\n         1.1384e-04, -7.9914e-05,  1.9652e-04, -2.5717e-05,  3.2720e-06,\n        -6.0814e-06,  4.1105e-06,  4.4543e-05, -1.6229e-05, -3.3145e-05,\n         1.2382e-04,  3.2658e-05, -1.1043e-04,  1.5420e-04, -3.8682e-05,\n        -1.1848e-05,  5.8904e-05,  1.2257e-05, -9.3945e-06,  1.8034e-04,\n         9.1726e-05, -5.6052e-45, -5.7588e-05, -4.4043e-05,  1.0230e-04,\n        -1.8961e-05,  7.8207e-05,  1.1652e-04, -2.0387e-04,  1.1119e-04,\n        -1.0507e-04, -9.7519e-05, -1.4766e-05,  6.2325e-05,  9.5417e-05,\n         1.9512e-04,  1.9354e-04, -7.6378e-05,  1.7317e-05, -7.5884e-05,\n        -3.8450e-05,  4.8033e-05,  9.6934e-05,  1.7354e-04,  5.6695e-05,\n        -1.0674e-04, -9.1333e-05,  1.1460e-04, -1.6287e-04,  1.4066e-04,\n        -3.2969e-04, -1.7241e-04, -8.8311e-05,  7.6578e-05,  2.0743e-05,\n        -3.3686e-04, -9.5934e-05,  9.1859e-05,  9.0062e-05, -4.4188e-05,\n         5.8586e-04, -2.1020e-06, -1.5128e-04,  6.2619e-05,  1.3123e-04,\n        -1.0706e-04,  3.3948e-04, -9.1203e-05,  2.3025e-05,  6.6483e-05,\n        -3.3932e-05,  2.9560e-04,  3.8678e-04,  8.5746e-05, -1.4657e-04,\n         3.5971e-04,  7.8842e-05, -3.0828e-05,  1.7587e-04, -2.1526e-05,\n         5.8349e-05, -8.9016e-05, -1.5025e-04,  6.3920e-06, -3.5173e-05,\n        -1.8645e-05,  6.7288e-05, -8.4056e-05,  1.0504e-04, -4.8835e-04,\n         9.3352e-05,  2.5125e-04, -2.2435e-04,  2.8625e-04,  2.9043e-04,\n         1.4384e-04, -1.1013e-04, -7.8196e-04, -3.0329e-05, -1.2110e-04,\n        -5.3806e-05, -5.4022e-05,  2.8966e-04,  2.5352e-04, -2.1548e-05,\n        -1.6928e-04,  3.4110e-05, -8.5611e-05, -2.3438e-04,  1.0973e-04,\n         2.1551e-05,  3.3693e-04, -1.3017e-04, -1.4229e-04, -1.7197e-04,\n         1.2198e-04,  3.2057e-04,  3.5783e-05, -9.1998e-04, -9.7898e-05,\n        -4.7958e-05, -2.3403e-04,  1.7019e-04, -1.2190e-04, -1.2435e-04,\n        -2.1264e-04, -7.0273e-05, -4.7520e-05, -3.4855e-05,  1.7900e-04,\n         3.7459e-05, -1.8914e-05,  1.6363e-04,  2.8416e-04, -9.4851e-06,\n        -6.3028e-05, -6.1105e-05, -7.1743e-05,  1.0626e-04,  4.5321e-05,\n         3.9030e-06, -1.0642e-07,  8.2713e-05,  1.6905e-04,  1.7992e-04,\n         1.3156e-04,  1.2906e-04,  8.0033e-05,  2.7912e-04,  1.1921e-05,\n         3.6470e-04,  2.8856e-05, -1.0755e-04, -2.6323e-04,  1.1200e-04,\n         1.0333e-05, -5.7287e-05, -1.6896e-04,  1.3798e-04, -3.5929e-04,\n         8.5352e-06,  9.7074e-05, -7.6636e-05,  7.1150e-05, -1.5857e-04,\n        -4.2728e-05,  1.3065e-04,  2.5683e-04,  1.0638e-04, -1.5501e-04,\n         2.9864e-05, -7.6376e-04, -1.4754e-04,  1.8833e-04, -1.5678e-04,\n        -2.7757e-04, -2.0727e-05, -8.6255e-05, -5.9114e-06,  1.3423e-04,\n        -1.0213e-04,  3.3842e-05,  4.7509e-05, -6.0057e-05,  5.8698e-05,\n         1.2088e-04,  1.6870e-05,  5.6052e-45,  2.6568e-04, -2.7662e-05,\n         8.2736e-05,  1.2345e-05,  8.8218e-05, -8.5139e-06, -3.0057e-04,\n         1.6387e-04, -6.1195e-05, -3.4784e-04,  2.3584e-05,  6.3180e-05,\n        -3.1708e-04,  3.9810e-04, -2.6866e-06, -4.5599e-05,  8.1166e-07,\n        -2.4455e-04, -6.2089e-05, -9.8747e-05,  6.3951e-05, -5.1049e-04,\n        -1.6604e-04, -2.6183e-04,  1.9005e-04, -5.2653e-05,  1.1844e-04,\n        -1.5078e-04,  1.0557e-04, -5.6954e-05,  1.8391e-05,  1.1196e-05,\n        -1.3285e-04, -1.6921e-04, -6.2150e-05, -6.3824e-05,  2.3192e-04,\n        -1.4176e-05, -5.5748e-05,  3.4016e-05, -4.3338e-05,  2.5997e-04,\n        -1.2537e-04, -5.1138e-05, -4.3932e-05, -1.2502e-04,  3.3399e-05,\n        -4.4549e-04,  9.2137e-05, -7.2620e-05,  1.0123e-04,  2.1431e-05,\n         2.0540e-05,  1.9755e-04, -6.9849e-05,  7.0310e-05, -9.9446e-05,\n         1.3434e-05, -1.0451e-04, -7.7842e-05,  1.7967e-04, -6.7510e-05,\n         1.2798e-04,  2.3808e-04,  1.1131e-04,  8.8340e-05, -5.5274e-06,\n         3.2821e-05, -1.6958e-04,  3.7256e-05,  3.1093e-05,  1.2233e-04,\n        -2.5047e-04,  3.2305e-05,  5.0378e-05,  1.0068e-04, -1.5496e-04,\n        -3.9147e-05, -3.3405e-04,  4.4631e-04, -1.5522e-04,  1.4786e-05,\n        -1.7125e-05,  1.8812e-04,  2.0777e-05,  8.9404e-05, -6.4116e-05,\n         6.1663e-06,  1.7845e-04, -2.6063e-04, -7.2249e-05, -9.1835e-05,\n        -2.9728e-05,  1.1194e-06, -5.4563e-04, -1.3660e-04,  1.0344e-04,\n        -2.5862e-04,  1.8876e-04,  2.7539e-04, -2.9561e-04,  1.7747e-04,\n         4.3308e-05, -5.8373e-05,  2.6540e-05, -1.6754e-04, -5.1464e-05,\n         5.2062e-05, -6.8142e-05,  1.2304e-05, -1.4105e-04, -1.4604e-05,\n         2.8377e-06, -3.8877e-05, -5.3468e-05,  1.5921e-04, -1.0798e-04,\n        -1.3162e-04, -1.5323e-04,  8.4145e-05,  1.0660e-04, -5.6187e-04,\n        -3.3303e-05, -6.9691e-05, -1.6655e-05, -4.8823e-05, -2.5034e-05,\n         5.9345e-05, -1.3957e-05,  6.7960e-05,  1.4905e-04,  2.0097e-04,\n         3.5805e-04,  1.8035e-04, -5.3022e-06, -3.9822e-05,  5.5282e-04,\n         1.3070e-04,  1.3386e-04,  3.2436e-04,  1.5350e-04,  2.6154e-05,\n         1.4979e-05, -3.8418e-05, -1.9032e-04,  7.4158e-05,  6.3868e-05,\n        -6.4798e-05,  9.5889e-06, -7.5351e-05,  5.8865e-06, -1.4279e-05,\n         1.6290e-04,  1.6383e-04,  4.6368e-06, -1.1138e-04,  2.8730e-04,\n         9.0168e-05,  3.9317e-04, -7.7682e-05, -2.2131e-04, -5.3471e-05,\n         7.2889e-05, -4.3777e-05,  1.9659e-04, -2.1230e-04, -3.5902e-04,\n         9.0394e-06,  6.1946e-05, -2.6224e-04, -2.8862e-04,  4.7006e-04,\n         1.3675e-05,  9.7516e-05, -1.8979e-05, -5.3750e-05,  1.3230e-04,\n         2.3560e-06,  8.6359e-05,  2.6651e-05, -1.6681e-05, -1.2589e-05,\n        -1.4659e-04, -5.5320e-05,  1.7960e-05, -6.4628e-05, -1.9698e-04,\n         2.0112e-04,  1.1762e-04, -4.8798e-05,  2.2323e-04,  2.1938e-04,\n        -9.1546e-05, -1.2086e-04, -1.0658e-04, -1.1210e-04,  1.3857e-04,\n        -2.4781e-04,  3.3030e-04,  3.5094e-05,  2.1677e-04,  8.2244e-05,\n        -1.7420e-04,  2.6636e-04,  7.7121e-05,  3.5095e-04, -1.4248e-05,\n         2.7391e-06,  4.9044e-06,  6.8081e-05,  4.7585e-05,  3.2127e-04,\n        -6.0759e-05, -7.9097e-05,  1.6906e-04, -2.3544e-05, -2.9171e-05,\n        -1.6745e-04,  1.5603e-04,  4.6529e-05, -1.1040e-04,  3.6129e-05,\n        -1.3145e-04,  5.1521e-05, -1.4818e-04, -4.8436e-05,  1.4861e-04,\n        -2.1237e-05,  2.7486e-04,  1.0794e-04, -1.8279e-05,  4.2193e-05,\n        -1.3471e-04,  7.8152e-05,  1.3767e-04,  6.3685e-05, -1.7406e-04,\n        -1.4160e-04,  5.6353e-05, -1.4992e-05, -1.9234e-05,  2.8758e-05,\n        -2.9798e-04,  2.7565e-05,  6.8054e-05, -1.4859e-05, -1.6320e-04,\n        -2.6463e-05,  1.3570e-04, -1.1376e-04,  1.6880e-04,  5.2559e-04,\n        -7.0014e-05, -1.8083e-04, -6.5876e-05, -4.1378e-05,  5.8489e-05,\n        -1.6953e-04,  3.5184e-05, -1.0442e-04, -2.4437e-05,  2.0582e-04,\n         1.6997e-04, -5.2732e-05, -1.0792e-06, -5.5454e-05, -8.3733e-05,\n        -1.9213e-05, -8.9905e-05,  7.2450e-05,  1.7878e-04,  1.3550e-04,\n        -1.9496e-05,  1.1908e-04, -3.5069e-04,  2.0520e-04, -3.6586e-04,\n        -1.0590e-04,  6.5640e-05, -7.6636e-05, -8.0065e-05,  1.2526e-04,\n        -5.9940e-05,  2.7638e-04, -1.4376e-04,  1.3936e-05, -5.2534e-05,\n         4.7914e-04,  1.6016e-05,  1.3879e-04,  2.0653e-04,  5.4986e-05,\n        -1.4737e-06,  7.5473e-05,  1.4020e-04, -9.5014e-05,  3.4207e-05,\n         9.7805e-05,  2.9363e-04], device='cuda:0')",
-        "exp_avg_sq": "tensor([1.8256e-07, 2.1425e-07, 1.8096e-07, 3.1086e-07, 1.5436e-07, 2.5972e-07,\n        1.4867e-07, 2.7652e-07, 1.7006e-07, 1.7839e-07, 3.9224e-07, 2.4319e-07,\n        1.0806e-07, 2.2200e-07, 1.0709e-07, 2.9199e-07, 2.3372e-07, 3.3276e-07,\n        2.1700e-07, 1.4956e-07, 1.4182e-07, 2.2349e-07, 3.9689e-07, 2.1595e-07,\n        3.0088e-07, 3.0177e-07, 3.0431e-07, 4.0482e-07, 1.2525e-07, 1.2114e-07,\n        3.0681e-07, 3.4064e-07, 3.0594e-07, 4.1693e-07, 1.4970e-07, 2.5318e-07,\n        2.2596e-07, 4.0247e-07, 3.4422e-07, 2.9641e-07, 4.3696e-07, 4.2409e-07,\n        3.1674e-07, 1.5198e-07, 3.0235e-07, 3.3255e-07, 8.9036e-10, 1.9861e-07,\n        3.8588e-07, 2.1598e-07, 2.0388e-07, 1.7505e-07, 3.5890e-07, 2.2420e-07,\n        2.3949e-07, 4.0970e-07, 3.1663e-07, 2.9832e-07, 3.6877e-07, 3.3916e-07,\n        2.4072e-07, 3.1500e-07, 2.7670e-07, 2.9994e-07, 2.0455e-07, 1.4815e-07,\n        2.6749e-07, 2.0407e-07, 2.9794e-07, 1.6954e-07, 2.6729e-07, 1.1179e-07,\n        2.4321e-07, 2.9745e-07, 2.0795e-07, 2.8625e-07, 3.4843e-07, 3.0315e-07,\n        3.7308e-07, 2.8408e-07, 2.0312e-07, 2.0092e-07, 2.3537e-07, 2.5766e-07,\n        2.0972e-07, 1.7113e-07, 2.6095e-07, 4.3421e-07, 2.0788e-07, 2.9926e-07,\n        2.1054e-07, 3.3935e-07, 1.9373e-07, 3.1144e-07, 1.0913e-07, 3.9414e-07,\n        3.6177e-07, 4.0089e-07, 4.9945e-07, 1.9958e-07, 3.4982e-07, 2.5925e-07,\n        3.0261e-07, 2.3794e-07, 1.6194e-07, 1.4839e-07, 3.0522e-07, 1.7743e-07,\n        2.0349e-07, 1.8151e-07, 7.4737e-08, 3.7157e-07, 1.5902e-07, 3.3497e-07,\n        2.8562e-07, 3.2404e-07, 2.5718e-07, 3.6781e-07, 2.0814e-07, 1.8925e-07,\n        3.3441e-07, 2.5547e-07, 5.9334e-07, 3.0809e-07, 3.7081e-07, 1.1751e-07,\n        4.0517e-07, 1.8862e-07, 2.1909e-07, 1.4162e-07, 1.1979e-07, 1.7339e-07,\n        2.1070e-07, 1.3662e-07, 1.5900e-07, 2.3966e-07, 3.8422e-07, 2.7259e-07,\n        2.4649e-07, 4.2156e-07, 3.7968e-07, 3.0587e-07, 1.6132e-07, 6.2563e-07,\n        2.5627e-07, 2.9693e-07, 3.4808e-07, 2.3952e-07, 2.3560e-07, 4.3328e-07,\n        3.5887e-07, 3.1825e-07, 1.2514e-07, 3.5611e-07, 1.9743e-07, 9.0684e-08,\n        3.1427e-07, 2.8987e-07, 1.5878e-07, 1.8582e-07, 2.4428e-07, 1.9156e-07,\n        2.1997e-07, 2.2396e-07, 2.8756e-07, 2.1900e-07, 4.5996e-07, 3.1181e-07,\n        3.7335e-07, 1.8851e-07, 3.9430e-07, 1.5814e-07, 7.1152e-08, 3.3474e-07,\n        2.2523e-07, 2.2151e-07, 3.3801e-07, 2.8776e-07, 2.3862e-07, 1.6238e-07,\n        1.5302e-07, 2.3336e-07, 1.9019e-07, 1.3591e-07, 3.5657e-07, 3.3209e-07,\n        3.1976e-07, 2.7448e-07, 2.9121e-07, 2.9496e-07, 3.1630e-07, 4.0501e-07,\n        3.6675e-07, 2.9018e-07, 4.0665e-07, 3.0313e-07, 4.4880e-07, 1.6705e-07,\n        1.4267e-07, 3.9530e-07, 4.0331e-07, 2.1055e-07, 3.1743e-07, 4.3339e-07,\n        1.8589e-07, 2.1722e-07, 1.3684e-07, 4.7503e-07, 1.2302e-07, 1.0313e-07,\n        2.4157e-07, 2.2642e-07, 6.4589e-10, 2.7511e-07, 1.5819e-07, 2.6258e-07,\n        2.6209e-07, 1.4568e-07, 2.8857e-07, 2.8944e-07, 2.2921e-07, 1.9293e-07,\n        1.8708e-07, 2.5509e-07, 3.2730e-07, 3.7669e-07, 4.1516e-07, 2.1599e-07,\n        3.2565e-07, 2.1794e-07, 2.0155e-07, 2.9262e-07, 3.0381e-07, 2.8204e-07,\n        2.9562e-07, 1.8101e-07, 3.2179e-07, 2.3236e-07, 2.7288e-07, 3.1693e-07,\n        1.4705e-07, 3.3113e-07, 2.3241e-07, 2.2905e-07, 1.1740e-07, 2.0800e-07,\n        3.1191e-07, 1.8555e-07, 1.7903e-07, 2.9601e-07, 2.2307e-07, 3.4995e-07,\n        3.1301e-07, 3.5587e-07, 3.8555e-07, 2.8922e-07, 2.2781e-07, 2.0783e-07,\n        3.7546e-07, 2.3805e-07, 3.5527e-07, 3.5037e-07, 2.8838e-07, 2.6903e-07,\n        9.8683e-08, 2.0528e-07, 2.8616e-07, 2.7834e-07, 4.4921e-07, 3.2371e-07,\n        2.9430e-07, 3.0961e-07, 1.9665e-07, 4.7703e-07, 1.3199e-07, 1.4511e-07,\n        1.6614e-07, 3.3142e-07, 2.3435e-07, 1.3907e-07, 3.2215e-07, 2.1930e-07,\n        2.4806e-07, 2.3886e-07, 2.7431e-07, 3.8069e-07, 1.5978e-07, 2.5274e-07,\n        2.3241e-07, 3.9467e-07, 3.3011e-07, 2.9075e-07, 3.8669e-07, 1.6693e-07,\n        1.0705e-07, 3.2581e-07, 2.2301e-07, 2.4748e-07, 3.3591e-07, 2.5318e-07,\n        1.7251e-07, 2.2753e-07, 3.6364e-07, 3.8079e-07, 2.1385e-07, 1.7824e-07,\n        2.1364e-07, 4.4789e-07, 1.8411e-07, 3.2213e-07, 2.4491e-07, 1.7791e-07,\n        1.3328e-07, 2.6314e-07, 3.6003e-07, 3.1611e-07, 1.6867e-07, 3.3040e-07,\n        1.0868e-07, 2.9242e-07, 2.3903e-07, 1.7826e-07, 4.1388e-07, 1.8012e-07,\n        1.6430e-07, 3.0997e-07, 2.7268e-07, 1.8449e-07, 4.0652e-07, 2.5353e-07,\n        3.5883e-07, 3.4857e-07, 2.4792e-07, 1.9784e-07, 3.7018e-07, 2.7227e-07,\n        3.7122e-07, 1.7250e-07, 9.2883e-08, 2.7772e-07, 2.9366e-07, 2.6413e-07,\n        2.6411e-07, 9.2334e-08, 1.7285e-07, 1.8836e-07, 1.3916e-07, 1.0605e-07,\n        1.9253e-07, 2.9649e-07, 2.1877e-07, 2.1372e-07, 2.4435e-07, 2.7261e-07,\n        3.4448e-07, 1.5889e-07, 2.6325e-07, 3.0022e-07, 2.6954e-07, 2.2011e-07,\n        2.7123e-07, 1.1561e-07, 2.6991e-07, 3.4243e-07, 3.4209e-07, 2.0396e-07,\n        2.4188e-08, 3.6758e-07, 1.5635e-07, 3.6905e-07, 2.6259e-07, 3.8687e-07,\n        4.7996e-07, 1.8078e-07, 2.1399e-07, 2.5266e-07, 1.4999e-07, 2.1441e-07,\n        1.4788e-07, 2.1741e-07, 9.1990e-08, 2.6614e-07, 3.7700e-07, 3.6182e-07,\n        3.6236e-07, 2.6526e-07, 3.3645e-07, 1.1933e-07, 2.2411e-07, 3.3755e-07,\n        2.1433e-07, 2.4179e-07, 3.1439e-07, 3.4441e-07, 1.0257e-07, 1.6522e-07,\n        1.9009e-07, 2.2387e-07, 2.0718e-07, 2.9718e-07, 4.3104e-07, 1.5074e-07,\n        3.9462e-07, 1.9716e-07, 4.2145e-07, 2.9546e-07, 2.5878e-07, 4.3282e-07,\n        2.6977e-07, 3.2924e-07, 3.1842e-07, 2.2467e-07, 2.6467e-07, 3.6254e-07,\n        3.0816e-07, 1.7705e-07, 3.1104e-07, 2.2346e-07, 2.1714e-07, 1.9056e-07,\n        1.3208e-07, 2.5582e-07, 2.1585e-07, 1.4291e-07, 4.6411e-07, 2.5803e-07,\n        1.5700e-07, 2.7100e-07, 1.7117e-07, 3.7821e-07, 3.3190e-07, 2.8111e-07,\n        1.8160e-07, 2.7962e-07, 1.7832e-07, 1.5373e-07, 2.8135e-07, 1.4392e-07,\n        1.2374e-07, 1.5946e-07, 3.9163e-07, 2.9864e-07, 1.5609e-07, 1.6107e-07,\n        3.6032e-07, 2.3995e-07, 1.0812e-07, 3.3769e-07, 2.4657e-07, 3.6158e-07,\n        3.8523e-07, 1.2528e-07, 2.2048e-07, 3.0105e-07, 4.2159e-07, 3.1306e-07,\n        1.5309e-07, 2.6877e-07, 1.0997e-07, 2.1401e-07, 2.8960e-07, 2.2950e-07,\n        2.6709e-07, 3.3779e-07, 4.1425e-07, 9.7348e-08, 1.9150e-07, 3.3333e-07,\n        1.7449e-07, 2.9185e-07, 4.2516e-07, 1.4202e-07, 2.8163e-07, 2.7556e-07,\n        2.6623e-07, 1.8766e-07, 2.1143e-07, 3.4654e-07, 1.7227e-07, 2.8115e-07,\n        4.5815e-07, 2.0989e-07, 1.9896e-07, 2.8303e-07, 3.1401e-07, 3.3674e-07,\n        9.0994e-08, 3.1324e-07, 3.4415e-07, 3.7619e-07, 3.6785e-07, 3.0748e-07,\n        2.2138e-07, 3.8593e-07, 2.5920e-07, 1.2004e-07, 2.4800e-07, 4.2623e-07,\n        1.6435e-07, 2.9217e-07, 4.2009e-07, 2.9591e-07, 2.2528e-07, 3.0166e-07,\n        2.8132e-07, 2.9005e-07, 1.2789e-07, 2.5149e-07, 1.3761e-07, 8.1217e-08,\n        1.5388e-07, 2.6906e-07], device='cuda:0')"
+        "step": "tensor(1252.)",
+        "exp_avg": "tensor([ 3.1904e-03,  2.4222e-03, -1.7243e-03, -2.1361e-04, -4.5648e-04,\n        -4.4914e-03, -4.0826e-04,  6.0857e-05, -1.0633e-03, -1.8701e-04,\n        -1.0094e-04, -1.0730e-03,  2.5487e-03,  3.7071e-03, -2.5732e-03,\n         1.7053e-03, -1.6072e-03,  1.2401e-03,  1.9455e-03,  9.8183e-04,\n        -2.3986e-03, -1.3021e-03, -1.2740e-03,  7.8719e-05, -2.2588e-03,\n         2.4844e-03,  4.7591e-03,  2.3537e-03,  8.7274e-04, -2.7813e-03,\n        -2.1584e-03, -5.7121e-03, -1.0514e-03, -4.9135e-04, -5.0863e-04,\n         1.0160e-03,  3.5551e-04, -8.2458e-04, -6.1719e-04,  3.0669e-03,\n        -1.7525e-03,  3.7165e-03, -3.5400e-03,  1.5227e-03, -2.3531e-03,\n        -9.3174e-04,  2.7090e-03, -2.8537e-03, -5.2754e-03,  3.5283e-04,\n        -3.1249e-03,  1.8325e-03, -8.7579e-04, -9.8658e-04, -2.1943e-03,\n        -1.6022e-03,  1.6063e-03, -1.4201e-04, -8.7773e-04,  1.5954e-03,\n        -3.4693e-03,  3.3627e-03, -5.9400e-03, -1.4886e-03, -2.8561e-03,\n         4.9963e-06,  2.6523e-03,  2.1751e-03, -4.5067e-03, -1.6471e-03,\n         6.1238e-04, -3.0726e-03, -2.2776e-03, -3.0961e-03, -2.6491e-03,\n         2.5174e-03, -1.6137e-03, -1.1758e-03, -7.0937e-03,  3.0721e-03,\n         2.4804e-03, -2.9634e-03,  7.5499e-04, -2.5208e-03, -4.3277e-03,\n         6.2486e-03, -1.4583e-03,  2.5355e-03,  6.1928e-25, -4.4319e-04,\n        -9.2588e-05,  6.1980e-04,  2.4074e-03, -1.4264e-03, -2.6702e-03,\n        -5.2027e-03,  3.2811e-04,  7.5429e-04,  4.3014e-04, -2.3400e-04,\n         2.2783e-03,  1.3519e-03, -4.6419e-04,  6.4521e-03, -2.5697e-03,\n        -3.7380e-04,  2.5154e-03,  1.7723e-03, -1.1818e-03,  1.6516e-03,\n         1.4826e-03, -5.6420e-04,  2.0438e-03,  1.1187e-03, -2.9898e-04,\n         2.6698e-03,  7.7897e-05,  3.9188e-03, -4.1652e-03,  1.3913e-03,\n         1.2945e-03, -9.8420e-04, -4.8178e-03, -2.4465e-03, -2.1221e-03,\n        -4.8332e-04, -5.2057e-04, -4.2964e-05, -3.0201e-03,  1.2106e-03,\n         4.0763e-03,  1.4035e-03, -4.7324e-03,  4.2647e-03,  1.7459e-03,\n        -1.6737e-04,  6.8106e-04, -9.5752e-04,  6.2907e-05, -8.4935e-04,\n        -2.0021e-05, -1.6985e-03,  4.1308e-03,  1.9211e-03,  2.1120e-03,\n         1.1449e-03,  7.1857e-05,  1.0149e-03, -1.5457e-03,  3.3767e-03,\n        -1.4126e-03,  2.2736e-03,  4.6065e-03, -1.1290e-03,  3.3845e-03,\n         2.9009e-03,  5.9909e-04, -1.7268e-03,  8.4962e-04, -9.9145e-04,\n         2.3474e-03, -1.6903e-03, -6.8522e-03,  1.5492e-06, -4.3454e-04,\n        -1.6407e-03, -2.2012e-03, -9.6707e-05,  6.2157e-04, -1.7944e-03,\n        -9.5820e-04, -3.2028e-03,  1.0146e-03,  3.0503e-03, -2.6575e-03,\n        -2.3976e-03, -1.2744e-03, -3.7374e-03,  3.2818e-03, -7.9136e-04,\n        -6.7694e-04,  1.3035e-03,  6.8708e-04, -7.5005e-04, -7.9721e-04,\n         3.2199e-04,  8.4443e-04, -1.0358e-03, -5.8625e-04,  2.2493e-03,\n        -1.0940e-03,  4.7334e-03,  7.9526e-04, -3.5127e-03,  2.7730e-03,\n         4.6681e-04,  2.8145e-03,  2.0011e-05,  5.2325e-04,  2.8647e-03,\n        -3.4098e-03,  4.3481e-03, -1.3079e-03, -2.6441e-03,  1.3388e-03,\n         1.3526e-03, -1.6097e-04,  2.0493e-03, -6.5562e-05,  1.9000e-03,\n         1.1330e-04,  3.2703e-03, -1.0130e-03, -2.7398e-03,  2.4459e-03,\n        -2.3422e-03, -3.1243e-03, -2.2579e-03,  3.9420e-03,  1.1702e-03,\n        -2.5664e-03,  2.7142e-03, -5.0034e-04, -3.5694e-03,  1.1598e-03,\n        -5.5907e-04,  1.2753e-03,  4.0660e-03, -2.8114e-03, -2.4260e-03,\n        -1.1035e-03,  8.5328e-05,  2.8403e-03,  4.3773e-03, -5.6567e-04,\n         1.1592e-03, -2.0618e-04, -6.2220e-04,  1.5435e-03,  1.5132e-03,\n         1.4089e-03, -7.0188e-04,  7.6053e-04,  4.6536e-03,  2.1584e-03,\n        -5.8872e-04,  1.2909e-03, -1.6966e-03,  3.5526e-03, -1.4303e-03,\n        -1.1565e-03,  5.5262e-05, -7.4679e-04, -2.2140e-03,  2.8577e-03,\n        -3.3754e-03, -7.0951e-05, -4.3856e-03, -4.9340e-05, -6.7614e-04,\n        -1.5506e-03, -3.6110e-03,  1.7703e-04, -1.5703e-03,  2.7254e-03,\n         5.3880e-03, -8.5066e-04,  8.1779e-05,  1.4173e-03, -9.7033e-04,\n        -9.8494e-04, -1.5660e-03,  9.2168e-04, -2.5521e-04, -2.5823e-04,\n         4.1380e-03,  2.6260e-03, -4.7231e-03,  5.2908e-03, -8.5965e-04,\n        -3.1564e-03, -4.0660e-03,  5.4622e-03, -1.2705e-03, -1.9472e-03,\n        -2.9671e-03, -2.9607e-04, -8.8360e-04,  1.9839e-03,  4.1254e-04,\n         3.4229e-03, -3.6705e-04,  1.1736e-03,  3.2248e-03, -2.5280e-03,\n        -6.7139e-05,  1.2895e-03, -5.3288e-03,  3.5936e-04, -6.0826e-04,\n        -2.0248e-03, -6.7692e-04,  1.4171e-03, -1.7383e-03, -1.5708e-03,\n        -3.3928e-03, -9.3082e-04,  2.1560e-03, -2.1778e-03,  5.4869e-04,\n        -1.0189e-03, -1.1810e-03,  1.3729e-03, -1.1063e-03, -2.6538e-05,\n         2.2007e-04,  1.0640e-03,  1.0277e-03, -3.3726e-04,  1.3428e-03,\n        -2.1437e-03,  1.9025e-03, -1.4138e-03, -2.9612e-04, -6.8967e-04,\n         1.9308e-04, -4.1767e-04,  1.5898e-03, -8.7928e-04,  3.3775e-03,\n         4.6539e-04, -1.9738e-03, -1.6794e-03,  1.2778e-03, -1.5326e-03,\n         1.7231e-04,  3.0655e-04, -1.3434e-04,  3.9599e-05, -3.4258e-05,\n        -7.2063e-04, -3.3310e-04,  4.8141e-03, -1.1483e-03,  6.6917e-04,\n        -1.9422e-03,  1.0457e-03, -1.0133e-03, -9.4841e-04,  2.5923e-03,\n        -1.2265e-03, -1.2267e-03,  3.9142e-04, -2.3726e-03,  1.5940e-03,\n        -5.2374e-03,  2.7246e-03, -1.0910e-03,  1.1526e-03,  1.1178e-03,\n        -1.0941e-03, -1.5307e-03, -1.0303e-03,  1.7788e-03,  1.5859e-03,\n         2.2983e-05, -3.0915e-03,  2.6508e-03,  6.0091e-04, -6.1271e-04,\n         1.8174e-03, -4.5975e-04, -1.0870e-03,  1.7214e-04, -1.6099e-04,\n        -2.1872e-03,  1.1419e-03,  4.6945e-03, -1.2438e-03, -1.4843e-03,\n        -3.4550e-03,  4.5226e-03,  7.1138e-04, -9.7154e-06, -1.8886e-03,\n        -1.9993e-03, -7.4591e-04, -3.9161e-03,  4.0808e-03, -5.0838e-04,\n         6.4355e-04,  8.6135e-04, -7.4710e-04,  2.0459e-03, -2.2105e-03,\n        -8.9224e-04,  1.0878e-03,  1.9240e-03,  4.8070e-04,  1.0789e-03,\n         1.6584e-03,  2.7715e-04,  7.2161e-04, -1.8812e-03, -1.6259e-03,\n         2.0904e-03, -6.4178e-04,  1.3211e-03,  1.5467e-03, -2.5006e-03,\n        -2.2606e-03, -3.9100e-04,  1.4261e-04,  6.3983e-04, -3.4760e-04,\n         2.9336e-03,  7.7692e-04,  8.6507e-04,  4.4178e-03, -2.9295e-03,\n        -2.5126e-03,  9.8705e-04,  9.5787e-04, -1.8633e-03, -1.3545e-03,\n         4.3228e-04, -9.5638e-04,  1.4392e-05,  3.5035e-04, -2.8349e-03,\n        -1.6866e-04,  8.5039e-04,  8.0130e-04,  4.2779e-03,  1.0363e-03,\n         6.0787e-04, -3.8142e-03,  1.2560e-03,  9.5090e-04,  2.0486e-03,\n        -5.7823e-04,  1.6073e-03, -1.0749e-03, -2.6016e-03,  3.6830e-04,\n        -9.2381e-04,  1.2310e-03,  2.7715e-03, -1.7159e-03, -8.0291e-04,\n         2.0933e-03,  3.4137e-03,  2.8626e-03,  2.5903e-03,  3.9098e-03,\n        -9.3130e-04, -6.1518e-03, -3.2061e-03, -4.6206e-04,  1.5748e-03,\n         2.9637e-03, -5.8531e-04,  2.9134e-03,  2.9580e-03,  2.9258e-03,\n        -3.8473e-05,  2.2790e-03,  4.3666e-04, -2.6951e-03, -2.7725e-03,\n         4.4092e-03, -7.0286e-04, -1.7469e-03, -2.5611e-04,  8.0352e-04,\n         6.7673e-04, -1.5469e-03,  9.2937e-06,  2.2508e-03, -9.9972e-05,\n        -2.6990e-04, -1.5148e-03,  2.0812e-03, -8.6283e-04,  3.5961e-04,\n         1.0026e-03, -1.6160e-03,  3.0064e-03,  1.1928e-03,  1.5060e-03,\n         1.7966e-03,  2.4652e-03,  3.1205e-04,  6.5354e-03,  1.1178e-03,\n         3.9190e-03,  1.1978e-03, -3.0784e-04, -3.6838e-05, -1.5580e-03,\n        -1.4839e-03, -6.2756e-03,  5.6554e-04,  4.2337e-03, -4.0199e-04,\n         4.6458e-05, -8.8214e-05, -1.4594e-03,  7.4806e-04, -6.1363e-03,\n         1.9009e-04,  3.3268e-03], device='cuda:0')",
+        "exp_avg_sq": "tensor([3.2608e-05, 5.1813e-05, 3.6478e-05, 4.9759e-05, 3.1374e-05, 3.2858e-05,\n        2.3110e-05, 2.6065e-05, 3.7883e-05, 3.0333e-05, 3.0979e-05, 7.9943e-05,\n        3.7628e-05, 5.2990e-05, 5.4106e-05, 2.9069e-05, 4.4855e-05, 3.9952e-05,\n        4.9734e-05, 3.5846e-05, 2.9616e-05, 3.4745e-05, 3.7715e-05, 1.9398e-05,\n        5.5174e-05, 2.8122e-05, 3.4693e-05, 5.4598e-05, 3.5685e-05, 3.6675e-05,\n        3.3971e-05, 4.6584e-05, 2.2161e-05, 2.9817e-05, 5.2569e-05, 2.6219e-05,\n        5.0869e-05, 5.6757e-05, 2.5216e-05, 2.8672e-05, 2.7959e-05, 4.2479e-05,\n        3.1481e-05, 3.1807e-05, 3.0899e-05, 2.5489e-05, 3.3023e-05, 4.9718e-05,\n        4.8909e-05, 2.6523e-05, 5.2396e-05, 6.7479e-05, 3.6356e-05, 4.3392e-05,\n        3.5922e-05, 2.7994e-05, 4.0730e-05, 3.5743e-05, 4.1670e-05, 4.9743e-05,\n        4.1944e-05, 2.7888e-05, 5.0559e-05, 1.1012e-04, 4.3299e-05, 5.2707e-05,\n        4.4338e-05, 3.3324e-05, 2.7188e-05, 5.1808e-05, 3.5586e-05, 4.5448e-05,\n        3.5480e-05, 3.9800e-05, 3.4946e-05, 7.0627e-05, 3.3953e-05, 5.0293e-05,\n        3.8729e-05, 5.1157e-05, 3.6394e-05, 7.3038e-05, 5.8686e-05, 4.0220e-05,\n        4.3622e-05, 3.4566e-05, 4.7022e-05, 5.6091e-05, 5.0546e-06, 5.6641e-05,\n        2.2846e-05, 6.3658e-05, 4.8345e-05, 3.4071e-05, 5.2106e-05, 3.4783e-05,\n        2.9782e-05, 5.9789e-05, 3.6178e-05, 2.9054e-05, 4.7738e-05, 2.9361e-05,\n        2.6725e-05, 5.7036e-05, 3.5507e-05, 2.3917e-05, 2.8500e-05, 2.4112e-05,\n        2.8923e-05, 3.7104e-05, 3.3309e-05, 3.5095e-05, 3.6185e-05, 4.0533e-05,\n        3.2158e-05, 3.5747e-05, 3.5395e-05, 3.7217e-05, 2.5518e-05, 3.6521e-05,\n        4.3698e-05, 3.4422e-05, 3.8187e-05, 4.8232e-05, 3.3016e-05, 2.5286e-05,\n        2.8268e-05, 2.1387e-05, 2.8860e-05, 4.0245e-05, 2.7977e-05, 4.1323e-05,\n        6.1664e-05, 5.6791e-05, 6.3904e-05, 5.7324e-05, 3.8974e-05, 4.2554e-05,\n        3.9765e-05, 3.9540e-05, 3.4547e-05, 2.9966e-05, 4.2294e-05, 4.4849e-05,\n        2.9022e-05, 3.7501e-05, 5.1115e-05, 1.7921e-05, 3.4351e-05, 4.1571e-05,\n        4.1932e-05, 5.1901e-05, 4.3911e-05, 3.8497e-05, 3.1480e-05, 3.0066e-05,\n        3.7794e-05, 2.9852e-05, 4.5148e-05, 3.4660e-05, 3.5515e-05, 2.5521e-05,\n        4.9242e-05, 4.5122e-05, 1.5869e-05, 4.0670e-05, 4.7479e-05, 3.7994e-05,\n        3.4011e-05, 5.2619e-05, 2.7742e-05, 3.1974e-05, 4.4370e-05, 4.3103e-05,\n        5.7954e-05, 3.8967e-05, 2.8050e-05, 4.7580e-05, 3.4764e-05, 5.8400e-05,\n        5.3260e-05, 4.4203e-05, 3.5991e-05, 6.4280e-05, 4.9932e-05, 2.0936e-05,\n        3.9473e-05, 6.2270e-05, 3.2312e-05, 3.9472e-05, 2.5771e-05, 4.7181e-05,\n        4.3664e-05, 3.4987e-05, 3.7633e-05, 2.7576e-05, 4.9282e-05, 2.4964e-05,\n        2.5539e-05, 2.7012e-05, 4.4453e-05, 4.9104e-05, 4.0547e-05, 3.4927e-05,\n        2.9960e-05, 4.3335e-05, 3.9049e-05, 3.9952e-05, 2.0383e-05, 3.2433e-05,\n        2.6951e-05, 4.9201e-05, 3.9310e-05, 3.3337e-05, 4.0379e-05, 3.2107e-05,\n        3.3901e-05, 3.5042e-05, 4.3775e-05, 3.6807e-05, 3.2625e-05, 5.5724e-05,\n        3.5240e-05, 3.2468e-05, 4.3511e-05, 2.1100e-05, 3.0426e-05, 3.5376e-05,\n        3.0155e-05, 2.8411e-05, 1.7705e-05, 3.0617e-05, 4.3398e-05, 4.9885e-05,\n        4.5708e-05, 2.5958e-05, 4.3901e-05, 8.4682e-05, 3.2613e-05, 2.5753e-05,\n        3.1085e-05, 2.1591e-05, 2.7057e-05, 2.9445e-05, 2.7500e-05, 3.5485e-05,\n        2.5766e-05, 4.0222e-05, 3.4392e-05, 2.9452e-05, 5.2441e-05, 3.6463e-05,\n        3.2552e-05, 4.0690e-05, 3.1536e-05, 3.0109e-05, 2.4839e-05, 5.1441e-05,\n        3.7413e-05, 4.1314e-05, 4.2541e-05, 3.6868e-05, 4.4045e-05, 2.9618e-05,\n        4.3116e-05, 4.8569e-05, 3.4432e-05, 2.5326e-05, 4.0064e-05, 4.1969e-05,\n        2.9393e-05, 2.6801e-05, 4.2183e-05, 3.3915e-05, 2.9240e-05, 6.8680e-05,\n        4.3725e-05, 3.7071e-05, 4.7470e-05, 3.4592e-05, 3.5597e-05, 3.5783e-05,\n        7.6772e-05, 4.1478e-05, 4.2449e-05, 4.4364e-05, 3.4347e-05, 3.5186e-05,\n        5.1583e-05, 4.2614e-05, 2.6735e-05, 5.9264e-05, 4.3715e-05, 2.2254e-05,\n        3.5565e-05, 2.5955e-05, 4.8387e-05, 5.9097e-05, 1.7540e-05, 4.1614e-05,\n        3.0456e-05, 5.1736e-05, 2.2741e-05, 2.6639e-05, 2.9400e-05, 5.0960e-05,\n        4.7542e-05, 3.7676e-05, 2.9634e-05, 3.0276e-05, 3.5328e-05, 3.4354e-05,\n        2.6617e-05, 2.8736e-05, 2.1163e-05, 2.8214e-05, 4.9795e-05, 4.2116e-05,\n        3.3335e-05, 4.2205e-05, 3.8198e-05, 4.1428e-05, 5.1545e-05, 6.5996e-05,\n        7.0261e-05, 1.9427e-05, 3.1066e-05, 3.5937e-05, 3.4066e-05, 4.2430e-05,\n        3.0496e-05, 5.9409e-05, 6.0696e-05, 3.3119e-05, 2.2896e-05, 3.4432e-05,\n        2.5491e-05, 2.9343e-05, 3.2953e-05, 4.1691e-05, 4.4710e-05, 4.7396e-05,\n        4.2963e-05, 2.3772e-05, 3.0405e-05, 6.2058e-05, 5.2358e-05, 2.5863e-05,\n        2.7585e-05, 3.7039e-05, 3.6992e-05, 4.5142e-05, 4.5037e-05, 2.9929e-05,\n        3.3757e-05, 4.9805e-05, 2.7740e-05, 4.0557e-05, 5.2582e-05, 2.3903e-05,\n        5.4078e-05, 1.5692e-05, 2.6962e-05, 3.0930e-05, 3.7187e-05, 2.0318e-05,\n        2.5082e-05, 4.2672e-05, 4.6621e-05, 2.8251e-05, 5.3580e-05, 4.2044e-05,\n        2.7176e-05, 3.4785e-05, 5.6138e-05, 6.9396e-05, 3.1785e-05, 4.1883e-05,\n        3.1659e-05, 3.5620e-05, 5.9935e-05, 3.4060e-05, 8.8174e-05, 3.9377e-05,\n        3.6814e-05, 5.3285e-05, 2.9641e-05, 5.4779e-05, 4.1094e-05, 3.8774e-05,\n        3.9281e-05, 6.3158e-05, 2.8784e-05, 3.7206e-05, 4.1169e-05, 3.4387e-05,\n        2.5929e-05, 4.3367e-05, 4.8873e-05, 2.0964e-05, 4.4807e-05, 3.9749e-05,\n        4.5384e-05, 2.0908e-05, 2.5394e-05, 5.6192e-05, 2.5171e-05, 3.7346e-05,\n        3.0626e-05, 4.5672e-05, 4.6783e-05, 4.3141e-05, 2.6629e-05, 4.0067e-05,\n        3.5018e-05, 3.7252e-05, 4.5220e-05, 4.0715e-05, 5.8007e-05, 5.8535e-05,\n        4.1670e-05, 5.8711e-05, 3.9437e-05, 3.8816e-05, 2.8924e-05, 6.0359e-05,\n        3.6461e-05, 4.1827e-05, 3.9043e-05, 3.3750e-05, 2.1728e-05, 3.5902e-05,\n        2.2062e-05, 8.1242e-05, 3.4953e-05, 4.1082e-05, 5.2174e-05, 4.1625e-05,\n        4.5753e-05, 5.9839e-05, 6.8913e-05, 3.0984e-05, 3.7476e-05, 2.9579e-05,\n        3.5499e-05, 4.9814e-05, 4.2719e-05, 4.0008e-05, 4.2896e-05, 5.0278e-05,\n        7.9120e-05, 2.7155e-05, 3.8322e-05, 4.2573e-05, 3.7037e-05, 4.8976e-05,\n        3.0148e-05, 3.6619e-05, 3.9858e-05, 2.2120e-05, 2.9387e-05, 5.5581e-05,\n        3.6904e-05, 3.2654e-05, 5.0724e-05, 4.8904e-05, 4.3524e-05, 4.9652e-05,\n        4.8339e-05, 3.7918e-05, 7.3630e-05, 2.4555e-05, 1.6924e-05, 3.0256e-05,\n        4.0265e-05, 3.2645e-05, 3.8231e-05, 4.3055e-05, 3.7927e-05, 3.3574e-05,\n        3.2205e-05, 4.2003e-05, 4.3956e-05, 1.9243e-05, 2.6357e-05, 3.7517e-05,\n        4.3266e-05, 2.8992e-05, 3.1765e-05, 3.0849e-05, 4.1076e-05, 3.0590e-05,\n        3.0444e-05, 4.5449e-05, 2.6051e-05, 5.0734e-05, 2.5179e-05, 2.8450e-05,\n        2.5363e-05, 8.9012e-06, 4.2007e-05, 3.3325e-05, 2.9209e-05, 3.8279e-05,\n        2.3830e-05, 3.8488e-05, 4.2143e-05, 3.4694e-05, 5.2841e-05, 3.7961e-05,\n        4.8133e-05, 3.6130e-05], device='cuda:0')"
       },
       "4": {
-        "step": "tensor(12520.)",
-        "exp_avg": "tensor([[-6.2645e-06,  2.7620e-05, -6.3752e-06,  ..., -2.2096e-07,\n          1.1970e-05,  8.1451e-06],\n        [-5.5997e-06, -2.0422e-05,  5.4662e-06,  ..., -2.0322e-06,\n          3.8756e-06, -6.0927e-06],\n        [-1.9815e-06, -3.1985e-05,  1.1588e-05,  ...,  5.0815e-07,\n         -2.6333e-05,  1.2598e-06],\n        ...,\n        [-3.2060e-05, -3.0144e-06,  4.5311e-06,  ..., -1.0798e-06,\n         -5.0732e-05, -9.6518e-06],\n        [ 7.6311e-06, -1.6229e-06, -7.7973e-06,  ..., -1.1816e-06,\n         -6.2412e-05,  9.8432e-06],\n        [-1.5716e-05, -9.9179e-05, -1.7823e-05,  ..., -3.2049e-06,\n          6.9133e-06, -3.0023e-06]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[2.4652e-09, 7.9788e-10, 8.8448e-10,  ..., 5.6255e-11, 3.4621e-10,\n         2.6910e-09],\n        [4.6996e-09, 2.7530e-09, 3.1170e-09,  ..., 6.2600e-11, 5.3921e-10,\n         4.2599e-09],\n        [3.0225e-09, 2.1399e-09, 1.4565e-09,  ..., 1.4477e-10, 1.0545e-09,\n         2.2051e-09],\n        ...,\n        [1.8733e-08, 1.7064e-09, 4.4232e-09,  ..., 8.1151e-11, 2.0680e-09,\n         4.5656e-09],\n        [4.0706e-09, 1.4898e-09, 3.2587e-09,  ..., 7.6016e-11, 3.5762e-09,\n         5.8001e-09],\n        [5.5760e-09, 3.0821e-09, 5.4045e-09,  ..., 1.3457e-10, 5.7061e-10,\n         6.1333e-09]], device='cuda:0')"
-      },
-      "5": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([[-1.8525e-06,  2.0083e-07,  3.1452e-07,  ...,  7.3335e-07,\n          0.0000e+00,  5.2947e-07],\n        [ 1.7694e-06, -9.4981e-08,  5.4940e-07,  ..., -9.8179e-07,\n          0.0000e+00,  4.1946e-08],\n        [ 1.5789e-07, -1.6118e-07,  1.0848e-06,  ...,  3.7259e-06,\n          0.0000e+00,  2.9067e-06],\n        ...,\n        [-4.1457e-06, -1.7088e-07, -2.2955e-07,  ...,  3.0773e-07,\n          0.0000e+00,  9.7976e-07],\n        [-3.9881e-08, -1.9297e-07, -3.0809e-07,  ..., -6.9249e-07,\n          0.0000e+00,  6.4987e-08],\n        [ 3.2264e-07,  1.2982e-07,  1.1459e-07,  ..., -1.1472e-06,\n          0.0000e+00, -1.5853e-06]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[2.7611e-11, 3.6099e-12, 1.5492e-11,  ..., 2.4540e-12, 0.0000e+00,\n         6.1661e-12],\n        [1.3127e-11, 5.1008e-12, 1.4483e-11,  ..., 2.0982e-11, 0.0000e+00,\n         8.0593e-12],\n        [1.1453e-11, 8.4014e-12, 8.0252e-12,  ..., 5.5600e-11, 0.0000e+00,\n         6.6102e-11],\n        ...,\n        [3.4881e-11, 7.9718e-12, 7.6355e-12,  ..., 1.5760e-12, 0.0000e+00,\n         3.5873e-12],\n        [9.0826e-13, 1.3714e-12, 4.4607e-11,  ..., 6.1792e-11, 0.0000e+00,\n         1.2949e-11],\n        [2.1942e-12, 9.8990e-12, 7.6443e-12,  ..., 2.1875e-10, 0.0000e+00,\n         2.6375e-11]], device='cuda:0')"
-      },
-      "6": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([ 8.4450e-06, -1.1573e-05,  2.1310e-05,  ..., -5.5773e-06,\n         2.5801e-06,  1.5663e-05], device='cuda:0')",
-        "exp_avg_sq": "tensor([2.8729e-09, 2.0459e-09, 2.6377e-09,  ..., 1.2660e-09, 1.7396e-09,\n        1.8250e-09], device='cuda:0')"
-      },
-      "7": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([[ 4.5972e-07,  3.1196e-08,  5.7611e-07,  ...,  8.5632e-08,\n          2.1753e-08, -4.2801e-07],\n        [ 1.6404e-06,  2.7880e-07, -6.3306e-07,  ...,  6.9177e-07,\n          3.6146e-07, -1.0307e-06],\n        [ 4.5100e-07,  5.2791e-07, -1.6244e-07,  ..., -1.9920e-07,\n          9.6318e-07, -1.0921e-06],\n        ...,\n        [-7.4474e-08, -2.5112e-06,  1.7036e-07,  ..., -5.1389e-08,\n         -3.4643e-07,  4.7856e-07],\n        [ 4.3694e-07, -1.3849e-06, -3.8582e-07,  ..., -1.0422e-07,\n          5.3094e-08, -2.7910e-07],\n        [-5.7219e-08,  5.0992e-07,  3.4745e-08,  ...,  7.7025e-07,\n         -2.7766e-08, -1.0441e-06]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[5.2503e-12, 2.0841e-12, 2.7646e-12,  ..., 1.8498e-12, 2.7767e-12,\n         2.9103e-12],\n        [7.1288e-12, 3.6513e-12, 3.8564e-12,  ..., 4.4297e-12, 5.1838e-12,\n         6.0143e-12],\n        [6.6456e-12, 4.2578e-12, 3.7655e-12,  ..., 3.2873e-12, 4.3643e-12,\n         7.3331e-12],\n        ...,\n        [5.9260e-12, 7.8217e-12, 6.7143e-12,  ..., 4.9780e-12, 7.2787e-12,\n         5.6242e-12],\n        [5.6961e-12, 3.5072e-12, 4.5105e-12,  ..., 2.6611e-12, 4.2406e-12,\n         6.4120e-12],\n        [8.5862e-12, 3.6779e-12, 5.7739e-12,  ..., 3.1581e-11, 5.1929e-12,\n         3.1326e-12]], device='cuda:0')"
-      },
-      "32": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([4.5374e-10], device='cuda:0')"
-      },
-      "33": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([-5.6052e-45,  5.6052e-45, -5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([4.1304e-12, 2.0078e-11, 6.1141e-12], device='cuda:0')"
-      },
-      "34": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([-5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45],\n       device='cuda:0')",
-        "exp_avg_sq": "tensor([2.4649e-08, 2.2347e-10, 3.0183e-10, 3.1322e-10, 3.5911e-10, 4.1252e-10,\n        3.2797e-10, 2.4288e-10, 3.0678e-10, 2.7262e-10], device='cuda:0')"
-      },
-      "36": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([[ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [ 5.6052e-45, -5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00, -5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        ...,\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45, -5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [-5.6052e-45,  5.6052e-45, -5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[3.8067e-17, 1.7153e-17, 2.5486e-17,  ..., 1.1141e-16, 0.0000e+00,\n         3.7600e-17],\n        [6.8525e-18, 1.6567e-17, 4.9828e-17,  ..., 5.5088e-17, 0.0000e+00,\n         2.2115e-17],\n        [6.1496e-15, 1.4550e-15, 8.0845e-14,  ..., 2.5919e-14, 0.0000e+00,\n         7.2322e-14],\n        ...,\n        [1.7139e-15, 7.1278e-17, 1.3499e-14,  ..., 7.5328e-15, 0.0000e+00,\n         2.0268e-14],\n        [4.6196e-17, 1.6623e-17, 4.8855e-18,  ..., 1.8139e-16, 0.0000e+00,\n         9.3531e-17],\n        [1.5585e-18, 2.9243e-17, 3.8203e-18,  ..., 9.7949e-18, 0.0000e+00,\n         3.9900e-17]], device='cuda:0')"
-      },
-      "37": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([ 5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([6.7602e-14, 2.3332e-14, 8.7980e-11, 1.1674e-13, 4.2467e-11, 1.4626e-11,\n        2.7982e-12, 3.2110e-12, 8.1145e-11, 2.9692e-13, 4.1254e-11, 5.0198e-12,\n        4.4499e-14, 9.2815e-13, 6.5526e-15, 4.7067e-13, 6.7148e-12, 7.6044e-12,\n        4.9165e-13, 5.9047e-13, 7.2017e-12, 3.3226e-13, 5.0081e-12, 2.9832e-12,\n        2.5213e-12, 2.5120e-11, 4.7747e-12, 4.1576e-12, 3.1639e-12, 6.3060e-11,\n        2.3168e-11, 1.4336e-11, 1.6701e-10, 2.7160e-13, 2.2376e-12, 5.0049e-13,\n        6.9579e-12, 3.0923e-13, 2.8086e-11, 7.4181e-13, 2.7813e-11, 4.7430e-13,\n        1.4774e-12, 2.6333e-11, 4.3718e-14, 1.5957e-12, 2.4685e-12, 2.4613e-12,\n        5.0269e-13, 1.7728e-11, 6.1258e-12, 6.5688e-12, 4.4004e-11, 2.8779e-12,\n        4.8896e-13, 3.5236e-12, 7.0266e-16, 1.7161e-12, 7.5588e-12, 7.4101e-11,\n        9.1360e-12, 8.1054e-13, 2.5836e-14, 2.7593e-12, 3.1355e-12, 4.1473e-11,\n        3.7844e-13, 2.7786e-12, 9.1194e-13, 6.3740e-12, 3.9018e-12, 3.4032e-12,\n        1.0501e-12, 1.5533e-11, 2.1834e-11, 1.2329e-11, 1.4959e-11, 3.6119e-12,\n        1.0802e-12, 2.9868e-14, 3.6992e-12, 1.4372e-12, 3.6963e-13, 2.0814e-12,\n        2.9306e-12, 9.2092e-15, 3.0585e-11, 4.9078e-12, 8.9542e-13, 4.0021e-14,\n        2.8695e-13, 2.9288e-12, 1.6742e-11, 2.6210e-11, 2.4557e-11, 1.8648e-12,\n        2.1988e-11, 9.2180e-12, 6.3504e-12, 9.1528e-13, 2.6961e-12, 1.1315e-13,\n        4.8186e-13, 2.8891e-11, 1.1609e-11, 2.0279e-11, 6.6085e-12, 6.3373e-14,\n        2.7186e-11, 2.2712e-12, 3.9414e-12, 5.4501e-13, 5.7729e-12, 2.8062e-12,\n        2.5978e-12, 1.2283e-10, 8.9703e-14, 4.1454e-12, 7.9854e-13, 3.8382e-11,\n        3.9091e-11, 5.3346e-12, 1.2208e-11, 4.8093e-13, 6.0287e-13, 1.0532e-11,\n        5.6894e-12, 6.4270e-12, 8.6047e-13, 2.0158e-11, 1.5908e-14, 4.5451e-13,\n        1.0958e-10, 9.9906e-12, 1.8822e-13, 2.1680e-12, 6.7894e-12, 2.7022e-12,\n        4.4465e-11, 8.5645e-11, 2.0018e-12, 6.1932e-12, 5.2587e-13, 8.6776e-14,\n        9.5838e-13, 4.3421e-12, 4.5682e-12, 8.6140e-12, 3.2458e-13, 1.8362e-11,\n        1.6433e-13, 4.5208e-13, 1.5083e-11, 6.7263e-12, 5.6788e-12, 3.2804e-11,\n        4.7362e-13, 2.5169e-11, 2.1787e-13, 1.1762e-12, 2.8222e-14, 2.0292e-11,\n        1.8833e-14, 1.5322e-12, 2.6128e-11, 1.4830e-11, 6.9850e-13, 9.1728e-12,\n        7.8077e-14, 4.5716e-13, 1.9595e-12, 1.0872e-11, 2.5805e-11, 9.0863e-12,\n        5.4882e-12, 1.8675e-12, 2.7010e-11, 3.2724e-11, 7.8415e-14, 5.9994e-15,\n        5.9588e-14, 9.7509e-12, 7.8574e-13, 2.9725e-13, 1.0212e-12, 3.8129e-11,\n        2.3737e-12, 4.6780e-14, 8.5004e-13, 1.4081e-12, 1.6620e-11, 9.8251e-14,\n        1.5546e-11, 3.1049e-11, 9.2328e-12, 2.5145e-12, 7.5315e-13, 8.5052e-12,\n        2.3004e-12, 9.6258e-12, 2.5715e-12, 1.5466e-13, 8.0506e-11, 1.5855e-13,\n        5.5186e-12, 1.0335e-14, 6.2226e-12, 2.0699e-11, 5.4415e-14, 1.4146e-12,\n        2.3793e-11, 1.4381e-12, 1.0969e-11, 7.3829e-11, 6.8060e-11, 7.5029e-12,\n        1.1199e-13, 3.4170e-12, 1.9903e-11, 4.0915e-14, 4.6974e-12, 8.2767e-11,\n        2.5684e-12, 4.4970e-12, 1.4890e-11, 1.9154e-12, 7.3287e-13, 6.6907e-12,\n        4.0311e-12, 3.3611e-12, 1.9034e-12, 7.8647e-12, 3.5316e-12, 1.8812e-12,\n        8.1689e-12, 1.2936e-12, 2.6196e-12, 3.2810e-12, 7.9772e-12, 1.4053e-13,\n        3.9491e-11, 1.3648e-12, 3.7057e-13, 1.3470e-11, 4.6216e-12, 6.9149e-14,\n        7.2008e-12, 4.7901e-11, 3.5642e-11, 5.2234e-11, 1.7538e-12, 7.3604e-12,\n        5.5967e-12, 2.4384e-11, 1.9314e-13, 2.4293e-14], device='cuda:0')"
-      },
-      "38": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([ 5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([1.5187e-16, 1.7418e-19, 1.6902e-13, 5.7739e-16, 8.9904e-14, 2.8297e-14,\n        4.9316e-15, 7.2007e-15, 1.8675e-13, 7.3868e-16, 1.4761e-13, 8.5998e-15,\n        2.6909e-16, 1.1597e-15, 2.1058e-17, 1.7048e-15, 1.4910e-14, 2.0922e-14,\n        1.7838e-15, 1.2892e-15, 1.1711e-14, 8.1664e-16, 6.6337e-15, 5.1297e-15,\n        4.7211e-15, 3.1273e-14, 7.3232e-15, 7.4417e-15, 1.8104e-14, 1.8302e-13,\n        2.8596e-14, 2.5165e-14, 3.1561e-13, 3.9863e-16, 3.8927e-15, 1.8965e-15,\n        8.8481e-15, 9.3052e-16, 7.1173e-14, 1.0419e-15, 7.8390e-14, 1.6156e-15,\n        4.3507e-15, 3.4979e-14, 4.3840e-17, 1.7567e-15, 3.8827e-15, 3.5338e-15,\n        1.3736e-15, 4.2210e-14, 8.0917e-15, 1.7268e-14, 1.1906e-13, 4.0773e-15,\n        7.9951e-16, 5.2272e-15, 4.2257e-17, 2.2060e-15, 1.0616e-14, 2.1544e-13,\n        1.5012e-14, 1.2699e-15, 2.5436e-19, 6.1730e-15, 5.6367e-15, 1.1489e-13,\n        9.5783e-16, 3.8587e-15, 1.8567e-15, 8.2341e-15, 6.5675e-15, 6.1851e-15,\n        9.1247e-16, 3.1165e-14, 3.7633e-14, 1.8773e-14, 4.1426e-14, 5.5167e-15,\n        2.8459e-15, 7.7618e-18, 6.5322e-15, 2.0749e-15, 7.1503e-16, 2.6115e-15,\n        4.2895e-15, 2.7359e-17, 4.6596e-14, 1.1949e-14, 2.0194e-15, 7.2043e-18,\n        6.9617e-16, 4.9106e-15, 3.7680e-14, 7.1836e-14, 2.9661e-14, 2.3196e-15,\n        3.0788e-14, 1.3530e-14, 9.0382e-15, 4.0221e-15, 3.7420e-15, 1.6316e-16,\n        8.0276e-16, 5.7792e-14, 3.3770e-14, 3.9143e-14, 9.5567e-15, 2.1369e-16,\n        9.7510e-14, 4.1525e-15, 8.8275e-15, 1.1524e-15, 9.2456e-15, 3.5167e-15,\n        4.8509e-15, 3.1635e-13, 2.1365e-16, 4.1961e-15, 4.1801e-15, 1.2086e-13,\n        6.7483e-14, 7.3396e-15, 1.5462e-14, 4.1012e-15, 1.7197e-15, 2.2589e-14,\n        1.1603e-14, 1.5134e-14, 1.6699e-15, 6.5712e-14, 2.6431e-16, 9.0672e-16,\n        2.9748e-13, 9.0782e-15, 2.6697e-16, 2.9641e-15, 2.3223e-14, 3.8576e-15,\n        1.2637e-13, 2.7545e-13, 3.9806e-15, 1.3201e-14, 2.0041e-15, 3.6607e-16,\n        1.6223e-15, 6.6433e-15, 9.0134e-15, 1.6151e-14, 1.2195e-15, 6.5280e-14,\n        2.5589e-16, 5.2727e-16, 3.2282e-14, 1.3350e-14, 1.2420e-14, 5.8200e-14,\n        6.8779e-16, 2.9191e-14, 3.1846e-16, 1.7646e-15, 1.0686e-16, 3.3560e-14,\n        7.2425e-17, 3.4934e-15, 4.4843e-14, 2.3288e-14, 8.6440e-16, 1.9144e-14,\n        3.0636e-16, 1.5272e-15, 1.8246e-15, 6.3593e-14, 6.3759e-14, 2.7012e-14,\n        1.5966e-14, 4.3335e-15, 5.3243e-14, 8.3904e-14, 3.3399e-16, 5.1200e-18,\n        1.2516e-16, 2.6708e-14, 2.4210e-15, 5.4076e-16, 3.5996e-15, 4.9574e-14,\n        3.9276e-15, 9.8538e-17, 1.6663e-15, 2.5180e-15, 3.7211e-14, 6.2327e-16,\n        2.9014e-14, 1.1292e-13, 1.3526e-14, 7.3219e-15, 1.3521e-15, 1.4623e-14,\n        3.2478e-15, 1.8178e-14, 4.1876e-15, 1.1207e-15, 1.3039e-13, 5.4043e-16,\n        8.4634e-15, 1.9947e-17, 1.4263e-14, 4.7389e-14, 1.8666e-16, 1.5734e-15,\n        4.1937e-14, 1.9534e-15, 1.8864e-14, 1.3659e-13, 1.0918e-13, 1.4110e-14,\n        1.8734e-16, 5.5284e-15, 3.2587e-14, 3.6043e-17, 8.3133e-15, 1.4193e-13,\n        6.1072e-15, 8.2739e-15, 2.1225e-14, 4.3891e-15, 1.7124e-15, 1.0957e-14,\n        4.3003e-15, 7.7113e-15, 3.6258e-15, 2.1638e-14, 5.2819e-15, 3.4504e-15,\n        1.2517e-14, 2.8603e-15, 3.3005e-15, 4.7209e-15, 2.7842e-14, 2.9940e-16,\n        1.0925e-13, 2.1956e-15, 3.5961e-15, 1.4457e-14, 6.4765e-15, 5.1526e-16,\n        1.5431e-14, 1.4876e-13, 5.8184e-14, 1.8268e-13, 2.8620e-15, 1.5188e-14,\n        9.0186e-15, 4.2205e-14, 3.3383e-16, 9.0653e-17], device='cuda:0')"
-      },
-      "39": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([ 5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([3.0956e-16, 1.1820e-17, 2.4193e-13, 5.2110e-16, 1.1031e-13, 3.7018e-14,\n        8.7719e-15, 9.4259e-15, 2.2626e-13, 1.5741e-15, 1.1870e-13, 1.4335e-14,\n        4.2323e-16, 2.8379e-15, 8.5158e-17, 2.1552e-15, 2.0162e-14, 2.2691e-14,\n        1.5979e-15, 2.0595e-15, 2.0703e-14, 1.1366e-15, 1.4558e-14, 9.0316e-15,\n        7.5054e-15, 6.8376e-14, 1.4456e-14, 1.1152e-14, 1.2148e-14, 1.7501e-13,\n        6.0130e-14, 3.9903e-14, 4.4894e-13, 7.0159e-16, 6.7194e-15, 1.9737e-15,\n        1.7937e-14, 1.6914e-15, 7.6171e-14, 2.4868e-15, 7.1744e-14, 1.7815e-15,\n        4.9355e-15, 7.7497e-14, 3.1098e-17, 3.6067e-15, 7.2440e-15, 7.6606e-15,\n        1.5008e-15, 4.9026e-14, 1.8248e-14, 1.5199e-14, 1.1524e-13, 8.0521e-15,\n        1.6599e-15, 1.0777e-14, 1.0359e-16, 4.5878e-15, 2.0350e-14, 2.0156e-13,\n        2.4289e-14, 2.4040e-15, 2.3752e-17, 9.0057e-15, 9.7225e-15, 1.0655e-13,\n        1.2990e-15, 7.9358e-15, 2.9072e-15, 1.8333e-14, 1.1823e-14, 1.0440e-14,\n        2.7417e-15, 4.4578e-14, 6.2820e-14, 3.3854e-14, 4.2167e-14, 1.0731e-14,\n        3.6977e-15, 7.0548e-17, 1.0931e-14, 3.7794e-15, 1.4590e-15, 6.5020e-15,\n        7.4191e-15, 1.2174e-16, 8.2893e-14, 1.4741e-14, 4.0722e-15, 7.6887e-18,\n        8.9701e-16, 8.2245e-15, 4.2952e-14, 6.8075e-14, 6.1317e-14, 4.2915e-15,\n        5.8528e-14, 2.6968e-14, 1.6523e-14, 4.5581e-15, 8.0665e-15, 3.9300e-16,\n        1.6469e-15, 7.5726e-14, 3.4086e-14, 4.7739e-14, 1.9722e-14, 3.7571e-16,\n        6.7503e-14, 6.8223e-15, 1.1958e-14, 2.0580e-15, 1.4780e-14, 6.5223e-15,\n        7.0887e-15, 3.3093e-13, 4.7863e-16, 1.1197e-14, 3.7104e-15, 1.0915e-13,\n        1.0472e-13, 1.5193e-14, 3.4298e-14, 4.2535e-15, 2.0393e-15, 3.1299e-14,\n        1.4219e-14, 1.8227e-14, 2.9983e-15, 5.6726e-14, 3.5792e-16, 1.5369e-15,\n        2.9401e-13, 2.6493e-14, 5.8427e-16, 6.7892e-15, 2.0184e-14, 6.4888e-15,\n        1.1644e-13, 2.2739e-13, 6.5108e-15, 1.8804e-14, 1.5648e-15, 4.4721e-16,\n        3.0549e-15, 1.2460e-14, 1.1651e-14, 2.4019e-14, 1.5490e-15, 5.2865e-14,\n        5.5015e-16, 1.1172e-15, 3.9455e-14, 1.6704e-14, 1.6285e-14, 8.9957e-14,\n        1.4693e-15, 6.6824e-14, 6.8855e-16, 3.2606e-15, 5.4023e-16, 5.8005e-14,\n        2.6327e-16, 4.6268e-15, 7.3734e-14, 3.8475e-14, 2.1500e-15, 2.4961e-14,\n        4.1862e-16, 1.5732e-15, 4.9100e-15, 3.3354e-14, 6.6142e-14, 2.8205e-14,\n        1.4722e-14, 6.0636e-15, 7.6261e-14, 9.3259e-14, 9.3869e-16, 1.8273e-17,\n        2.0721e-16, 2.6260e-14, 2.7228e-15, 1.0536e-15, 3.1008e-15, 1.0398e-13,\n        6.5608e-15, 2.0529e-16, 3.0080e-15, 4.2414e-15, 4.5224e-14, 8.2989e-16,\n        3.9672e-14, 8.7107e-14, 2.5007e-14, 7.6775e-15, 2.3512e-15, 2.0767e-14,\n        6.2183e-15, 2.7599e-14, 7.6892e-15, 1.2547e-15, 2.1578e-13, 9.0545e-16,\n        1.4863e-14, 2.6140e-17, 1.7855e-14, 5.4181e-14, 6.4058e-16, 4.6105e-15,\n        6.7241e-14, 3.4529e-15, 3.1347e-14, 1.9336e-13, 1.8171e-13, 2.0165e-14,\n        4.6955e-16, 9.2508e-15, 5.3024e-14, 1.3607e-16, 1.1422e-14, 2.2608e-13,\n        7.4901e-15, 1.2700e-14, 4.3591e-14, 4.6260e-15, 2.5800e-15, 1.8981e-14,\n        1.1478e-14, 9.8764e-15, 6.1658e-15, 2.1306e-14, 8.2529e-15, 5.5889e-15,\n        2.3410e-14, 4.1332e-15, 6.6861e-15, 9.7498e-15, 2.0074e-14, 6.6284e-16,\n        1.0565e-13, 4.1072e-15, 2.7739e-15, 3.8933e-14, 1.2684e-14, 6.8705e-16,\n        2.1728e-14, 1.2179e-13, 1.0357e-13, 1.4818e-13, 5.4153e-15, 1.8879e-14,\n        1.3282e-14, 6.6565e-14, 8.1795e-16, 8.8488e-17], device='cuda:0')"
-      },
-      "40": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([[-5.6052e-45,  5.6052e-45, -5.6052e-45,  ..., -5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [-5.6052e-45, -5.6052e-45, -5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        ...,\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [-5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[6.8036e-21, 2.7845e-19, 1.9085e-17,  ..., 7.8653e-18, 0.0000e+00,\n         7.9333e-17],\n        [1.8123e-17, 2.1025e-18, 1.8871e-16,  ..., 1.1612e-16, 0.0000e+00,\n         5.0431e-17],\n        [5.7999e-15, 8.9477e-17, 7.7679e-14,  ..., 1.2162e-14, 0.0000e+00,\n         5.5812e-14],\n        ...,\n        [1.3456e-15, 3.7460e-16, 2.3735e-14,  ..., 1.3488e-14, 0.0000e+00,\n         1.2141e-14],\n        [7.7572e-18, 1.7120e-17, 2.8716e-16,  ..., 2.4899e-17, 0.0000e+00,\n         1.3751e-16],\n        [1.4298e-17, 1.8727e-16, 1.2743e-15,  ..., 1.4906e-16, 0.0000e+00,\n         1.2532e-15]], device='cuda:0')"
-      },
-      "41": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([ 5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([6.4470e-16, 1.3765e-14, 2.6147e-11, 1.5243e-14, 5.7434e-12, 8.3110e-12,\n        4.3151e-12, 5.5613e-13, 8.8192e-11, 2.4843e-12, 2.0417e-11, 1.8309e-11,\n        5.4829e-14, 1.7146e-12, 3.5598e-14, 7.1168e-14, 8.2841e-12, 4.6418e-12,\n        9.9192e-13, 2.5180e-13, 1.4616e-12, 1.4562e-13, 9.3085e-12, 4.4165e-12,\n        1.1776e-11, 1.5821e-11, 5.0702e-12, 7.3556e-12, 2.8794e-12, 4.1077e-11,\n        4.3386e-11, 1.4909e-12, 1.7364e-10, 1.5281e-12, 1.8384e-12, 1.6598e-13,\n        6.3454e-12, 1.3472e-13, 4.1410e-12, 1.4662e-12, 1.4158e-11, 1.3421e-13,\n        3.3858e-14, 8.6095e-11, 2.1505e-14, 3.8738e-12, 9.1394e-12, 7.3482e-12,\n        8.3147e-14, 4.8147e-12, 7.2211e-12, 9.3820e-13, 3.0588e-11, 2.0069e-12,\n        9.9352e-13, 2.0543e-11, 6.5780e-13, 1.8757e-12, 3.7982e-11, 3.7374e-11,\n        8.0010e-12, 1.0514e-11, 2.5272e-13, 5.0792e-12, 6.2380e-12, 3.4852e-11,\n        2.1087e-13, 9.1659e-12, 1.9313e-12, 4.4167e-12, 1.6110e-12, 1.7188e-12,\n        4.6064e-12, 4.9407e-11, 1.4934e-11, 4.3673e-11, 9.2591e-12, 1.3173e-12,\n        1.3889e-12, 6.3481e-14, 4.2869e-12, 6.1442e-13, 3.9543e-13, 1.1185e-11,\n        3.2481e-12, 3.2904e-13, 6.5189e-11, 6.5589e-12, 1.8829e-12, 1.5589e-14,\n        1.1811e-13, 4.1768e-12, 7.3458e-12, 2.0223e-12, 7.0228e-11, 2.6837e-13,\n        6.5557e-11, 4.5347e-11, 7.8823e-12, 5.2131e-13, 1.7714e-12, 1.0643e-12,\n        1.1771e-13, 3.3235e-11, 3.8693e-13, 5.5069e-12, 4.3653e-12, 7.3605e-13,\n        4.8007e-12, 2.1745e-11, 5.6031e-12, 3.7666e-13, 2.3362e-11, 2.3881e-12,\n        1.8336e-12, 7.0165e-12, 6.3450e-13, 5.4254e-12, 6.7410e-13, 1.8422e-11,\n        5.0649e-11, 6.4280e-12, 5.7501e-12, 7.4937e-14, 1.3050e-12, 7.3845e-11,\n        5.4200e-13, 4.0525e-12, 1.5345e-12, 4.8567e-12, 5.3847e-14, 1.5432e-12,\n        7.2395e-11, 3.2841e-12, 1.4487e-13, 6.1313e-12, 3.8962e-12, 6.0092e-13,\n        2.0061e-11, 3.5635e-11, 6.4267e-13, 1.7842e-12, 2.5973e-11, 4.3436e-14,\n        9.3879e-13, 5.6760e-12, 1.8163e-12, 2.1373e-11, 3.8714e-13, 1.2511e-11,\n        1.1616e-12, 7.0578e-14, 1.8871e-11, 5.0627e-12, 1.2997e-12, 1.8382e-11,\n        6.6442e-12, 2.5253e-11, 3.6798e-14, 3.2494e-12, 1.4301e-13, 1.7434e-11,\n        1.5953e-14, 2.7917e-11, 3.7036e-12, 1.8556e-11, 1.8980e-12, 6.5761e-12,\n        1.3436e-13, 4.7498e-15, 1.5890e-12, 6.0242e-12, 5.4561e-12, 6.7826e-12,\n        2.5232e-12, 2.1332e-12, 3.0067e-11, 6.7749e-12, 5.6151e-14, 9.9788e-15,\n        6.4753e-13, 6.8810e-12, 8.1312e-14, 7.4658e-13, 3.8487e-13, 8.4649e-11,\n        8.0310e-12, 1.1514e-13, 1.5904e-12, 1.8941e-12, 2.6567e-11, 7.6002e-14,\n        1.1273e-11, 2.8730e-11, 1.8796e-11, 2.2258e-13, 2.6514e-13, 8.2934e-13,\n        8.2395e-12, 2.3246e-11, 1.1899e-11, 1.4033e-13, 5.1835e-11, 6.4631e-15,\n        3.4428e-12, 2.3249e-14, 1.9901e-12, 1.6953e-11, 4.7523e-13, 5.3551e-13,\n        3.1906e-11, 1.0027e-12, 1.2402e-11, 6.8319e-11, 2.2984e-10, 1.7022e-12,\n        2.8141e-13, 1.0538e-11, 5.2185e-12, 6.9709e-15, 4.0920e-12, 9.4407e-11,\n        1.3244e-12, 8.6558e-12, 7.8115e-12, 9.7746e-13, 5.4680e-13, 1.2969e-11,\n        1.6014e-11, 8.0782e-13, 1.5564e-11, 2.3548e-12, 7.4927e-12, 1.0030e-11,\n        8.9276e-12, 4.5397e-12, 2.6013e-12, 5.0583e-12, 1.5273e-12, 2.9417e-13,\n        2.6316e-11, 5.5092e-12, 1.7043e-13, 2.1895e-11, 2.2478e-12, 4.5088e-14,\n        2.0562e-11, 2.5016e-11, 6.2970e-11, 1.9427e-11, 6.0388e-13, 1.4976e-11,\n        1.3641e-12, 2.0007e-11, 2.3748e-13, 1.7140e-12], device='cuda:0')"
-      },
-      "42": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([ 5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([2.1234e-17, 3.0068e-17, 4.5547e-14, 1.2846e-16, 1.0170e-14, 1.4046e-14,\n        7.2674e-15, 9.0943e-16, 1.5901e-13, 5.3391e-15, 3.2609e-14, 2.9668e-14,\n        1.6347e-16, 2.4224e-15, 1.1064e-16, 1.8819e-16, 1.5626e-14, 1.9387e-14,\n        2.5918e-15, 8.3964e-16, 3.4624e-15, 3.8523e-16, 2.4222e-14, 9.0565e-15,\n        4.2178e-14, 2.4697e-14, 1.0243e-14, 2.2203e-14, 1.1149e-14, 6.7998e-14,\n        8.0174e-14, 3.6646e-15, 3.5007e-13, 1.7042e-15, 3.0545e-15, 2.4772e-16,\n        1.0038e-14, 7.1733e-16, 5.6685e-15, 3.1607e-15, 1.8618e-14, 1.6992e-15,\n        4.4965e-17, 2.8750e-13, 8.8109e-17, 6.8234e-15, 2.5131e-14, 2.9089e-14,\n        2.5332e-16, 7.6496e-15, 1.4032e-14, 1.1750e-15, 5.3433e-14, 3.2346e-15,\n        2.1255e-15, 3.1913e-14, 2.3615e-15, 2.5333e-15, 8.9998e-14, 4.9623e-14,\n        1.1758e-14, 2.8763e-14, 6.7849e-16, 8.0920e-15, 1.5989e-14, 7.6243e-14,\n        6.2181e-16, 1.3499e-14, 3.5758e-15, 4.9398e-15, 2.6293e-15, 3.7866e-15,\n        7.5045e-15, 1.1953e-13, 2.0682e-14, 9.4545e-14, 1.8411e-14, 2.3148e-15,\n        2.7377e-15, 6.7673e-17, 5.2492e-15, 9.2049e-16, 1.2092e-15, 4.4016e-14,\n        4.2216e-15, 1.7892e-15, 1.7341e-13, 2.2926e-14, 4.7421e-15, 9.3584e-17,\n        5.7202e-16, 6.8206e-15, 9.5604e-15, 3.2373e-15, 1.8716e-13, 6.3576e-16,\n        1.6735e-13, 1.1609e-13, 1.4724e-14, 2.0721e-15, 3.4265e-15, 2.2009e-15,\n        3.6207e-16, 6.1844e-14, 1.3836e-15, 7.6293e-15, 5.1177e-15, 1.8950e-15,\n        4.8253e-15, 5.0249e-14, 1.2533e-14, 5.0184e-16, 3.5774e-14, 3.0058e-15,\n        4.6077e-15, 1.1764e-14, 2.3317e-15, 6.0778e-15, 2.5168e-15, 2.5979e-14,\n        9.6308e-14, 1.9065e-14, 6.3542e-15, 1.7208e-16, 3.9849e-15, 1.5007e-13,\n        1.0572e-15, 1.1241e-14, 3.5203e-15, 8.5509e-15, 3.0835e-16, 2.0624e-15,\n        1.3714e-13, 5.3234e-15, 3.7196e-16, 1.2431e-14, 1.0508e-14, 1.3320e-15,\n        3.2183e-14, 5.6518e-14, 1.2912e-15, 2.6943e-15, 4.9650e-14, 2.9003e-16,\n        2.4412e-15, 1.1083e-14, 2.5440e-15, 6.9374e-14, 1.6290e-15, 4.1335e-14,\n        2.2787e-15, 4.3718e-17, 3.8758e-14, 9.6424e-15, 2.4291e-15, 1.9077e-14,\n        1.0499e-14, 3.2375e-14, 5.4211e-17, 6.8756e-15, 1.6884e-15, 3.7948e-14,\n        8.9368e-17, 9.0157e-14, 6.4611e-15, 3.1781e-14, 4.5999e-15, 1.4288e-14,\n        3.0808e-16, 1.3830e-17, 2.2212e-15, 1.4977e-14, 6.8688e-15, 9.4435e-15,\n        5.6448e-15, 4.7166e-15, 6.6270e-14, 1.2626e-14, 1.9852e-16, 2.3864e-17,\n        1.0611e-15, 1.8254e-14, 7.5796e-17, 1.4464e-15, 9.6780e-16, 1.4659e-13,\n        1.0196e-14, 6.2784e-16, 7.0977e-15, 5.0403e-15, 8.1586e-14, 5.6325e-16,\n        1.6887e-14, 6.2074e-14, 3.2935e-14, 6.6749e-16, 2.5664e-16, 1.5828e-15,\n        1.2935e-14, 5.5142e-14, 2.7901e-14, 7.8857e-16, 8.8479e-14, 1.5010e-16,\n        3.7997e-15, 4.3977e-17, 2.5374e-15, 3.5998e-14, 1.9475e-15, 8.3504e-16,\n        9.6844e-14, 1.0437e-15, 1.6487e-14, 1.3950e-13, 6.6441e-13, 2.5022e-15,\n        4.9461e-16, 1.8729e-14, 7.5915e-15, 6.3076e-18, 4.0583e-15, 1.6902e-13,\n        3.4789e-15, 2.7056e-14, 1.4124e-14, 1.5937e-15, 1.4268e-15, 2.7118e-14,\n        2.7697e-14, 1.2526e-15, 1.9774e-14, 4.8577e-15, 1.5566e-14, 2.0737e-14,\n        1.4284e-14, 8.3601e-15, 4.2393e-15, 9.9480e-15, 2.3474e-15, 5.8946e-16,\n        5.4205e-14, 1.3872e-14, 1.6619e-15, 3.0417e-14, 2.3237e-15, 2.9859e-16,\n        5.5991e-14, 6.0185e-14, 1.4853e-13, 3.1719e-14, 9.7684e-16, 3.2288e-14,\n        2.0587e-15, 4.5507e-14, 6.1489e-16, 7.2878e-15], device='cuda:0')"
-      },
-      "43": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([-5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([7.3115e-17, 5.4462e-17, 7.5521e-14, 2.1279e-16, 1.5669e-14, 2.0602e-14,\n        1.2201e-14, 1.9916e-15, 2.4401e-13, 7.2189e-15, 5.7919e-14, 4.4626e-14,\n        1.6331e-16, 5.3547e-15, 1.6431e-16, 5.4033e-16, 2.4894e-14, 1.5102e-14,\n        2.3882e-15, 1.1468e-15, 4.9915e-15, 6.9343e-16, 2.4302e-14, 1.3124e-14,\n        3.4953e-14, 4.0768e-14, 1.4929e-14, 1.8953e-14, 1.0859e-14, 1.1394e-13,\n        1.0749e-13, 4.9827e-15, 4.8385e-13, 3.7130e-15, 6.2632e-15, 8.3331e-16,\n        1.6321e-14, 1.1423e-15, 1.2555e-14, 4.3924e-15, 3.4059e-14, 1.0641e-15,\n        2.0345e-16, 2.4134e-13, 9.5718e-17, 9.1456e-15, 2.3116e-14, 2.2571e-14,\n        4.5785e-16, 1.4209e-14, 2.2460e-14, 2.3802e-15, 7.9168e-14, 5.2026e-15,\n        3.3812e-15, 5.4728e-14, 2.6167e-15, 4.7797e-15, 9.7998e-14, 1.0755e-13,\n        1.9456e-14, 2.4476e-14, 1.1750e-15, 1.4712e-14, 1.8045e-14, 8.6582e-14,\n        7.7119e-16, 2.4474e-14, 5.7945e-15, 1.2996e-14, 4.8676e-15, 5.0789e-15,\n        9.6839e-15, 1.3596e-13, 4.3025e-14, 1.1113e-13, 2.6762e-14, 4.1902e-15,\n        4.7477e-15, 1.4556e-16, 1.1858e-14, 1.9019e-15, 1.6973e-15, 3.2214e-14,\n        7.9529e-15, 1.6008e-15, 1.7996e-13, 2.0660e-14, 6.6256e-15, 1.3284e-16,\n        6.0693e-16, 1.0772e-14, 1.9320e-14, 6.0839e-15, 1.7616e-13, 6.9310e-16,\n        1.6628e-13, 1.1510e-13, 1.7820e-14, 3.3223e-15, 5.3499e-15, 3.8657e-15,\n        3.4383e-16, 8.6125e-14, 1.1522e-15, 1.4429e-14, 1.2351e-14, 2.8312e-15,\n        1.2383e-14, 5.3344e-14, 1.6396e-14, 1.2877e-15, 6.2289e-14, 5.3684e-15,\n        5.9515e-15, 2.0442e-14, 2.4870e-15, 1.3704e-14, 3.3680e-15, 5.2845e-14,\n        1.3833e-13, 2.0152e-14, 1.6598e-14, 1.2038e-15, 4.7986e-15, 1.8841e-13,\n        1.4681e-15, 1.2612e-14, 4.6593e-15, 1.4016e-14, 4.7119e-16, 5.0424e-15,\n        1.8606e-13, 8.8661e-15, 6.4000e-16, 1.6090e-14, 1.2474e-14, 1.4845e-15,\n        5.0322e-14, 8.9860e-14, 2.4147e-15, 5.6877e-15, 6.4036e-14, 5.3083e-16,\n        3.2841e-15, 1.6618e-14, 4.2403e-15, 5.5542e-14, 1.4860e-15, 3.6323e-14,\n        3.8305e-15, 8.5756e-17, 4.8387e-14, 1.1995e-14, 4.4635e-15, 4.7759e-14,\n        1.6141e-14, 6.1128e-14, 2.0181e-16, 9.5804e-15, 2.3039e-15, 5.0961e-14,\n        1.7736e-16, 7.8483e-14, 1.0608e-14, 4.7326e-14, 4.7014e-15, 1.6805e-14,\n        5.0852e-16, 3.8005e-17, 4.1446e-15, 1.8409e-14, 1.5301e-14, 2.0099e-14,\n        7.1505e-15, 6.2556e-15, 8.4551e-14, 1.8859e-14, 7.8269e-16, 5.1650e-17,\n        2.4030e-15, 1.8709e-14, 3.1602e-16, 2.6198e-15, 1.5672e-15, 2.3303e-13,\n        1.9446e-14, 5.8033e-16, 5.8495e-15, 6.5993e-15, 6.5526e-14, 1.0154e-15,\n        2.8395e-14, 7.6541e-14, 4.5763e-14, 6.2262e-16, 9.8517e-16, 2.8472e-15,\n        2.1993e-14, 6.6592e-14, 3.0075e-14, 1.2854e-15, 1.3520e-13, 3.2333e-16,\n        8.9223e-15, 9.5781e-17, 6.3468e-15, 4.1802e-14, 2.4001e-15, 1.7307e-15,\n        9.0554e-14, 2.3077e-15, 3.5731e-14, 1.8333e-13, 6.2338e-13, 4.5761e-15,\n        9.3240e-16, 2.4866e-14, 1.4668e-14, 3.3469e-17, 9.4090e-15, 2.5857e-13,\n        4.4725e-15, 2.5716e-14, 2.3847e-14, 2.1817e-15, 1.9333e-15, 3.6091e-14,\n        4.4652e-14, 2.6046e-15, 4.2446e-14, 7.4593e-15, 1.7757e-14, 2.9745e-14,\n        2.5595e-14, 1.4377e-14, 6.8978e-15, 1.5954e-14, 3.9608e-15, 9.2303e-16,\n        6.8074e-14, 1.6634e-14, 2.2040e-15, 6.0434e-14, 6.0744e-15, 7.1273e-16,\n        5.6861e-14, 6.3539e-14, 1.7325e-13, 5.7116e-14, 1.8168e-15, 3.6252e-14,\n        3.2442e-15, 5.8728e-14, 1.1851e-15, 5.9057e-15], device='cuda:0')"
-      },
-      "44": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([[ 5.6052e-45, -5.6052e-45, -5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45, -5.6052e-45,  ..., -5.6052e-45,\n          0.0000e+00, -5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        ...,\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [-5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[1.8837e-18, 3.1945e-18, 1.2736e-17,  ..., 2.2859e-18, 0.0000e+00,\n         5.0502e-17],\n        [8.1939e-18, 1.0124e-18, 9.2588e-17,  ..., 6.0307e-17, 0.0000e+00,\n         2.7694e-16],\n        [1.5964e-14, 1.9042e-15, 1.2136e-13,  ..., 7.5678e-14, 0.0000e+00,\n         2.0002e-13],\n        ...,\n        [4.2555e-15, 5.1611e-16, 2.9096e-14,  ..., 1.3738e-14, 0.0000e+00,\n         1.4968e-14],\n        [1.8505e-17, 1.1228e-17, 4.1032e-17,  ..., 3.0705e-16, 0.0000e+00,\n         1.8519e-16],\n        [4.0727e-18, 1.7285e-18, 1.5401e-16,  ..., 1.1582e-16, 0.0000e+00,\n         7.8799e-18]], device='cuda:0')"
-      },
-      "45": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([ 5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([7.9241e-15, 1.9546e-14, 9.9506e-11, 3.2360e-15, 3.3004e-11, 2.0608e-11,\n        4.9018e-12, 2.8881e-12, 1.0053e-10, 2.0831e-12, 2.6596e-12, 9.4841e-12,\n        1.1493e-13, 2.1525e-12, 4.6033e-13, 8.3460e-13, 7.5251e-12, 1.8510e-12,\n        2.5899e-13, 8.2761e-14, 7.0879e-12, 3.4270e-13, 9.5433e-12, 1.0143e-12,\n        8.6997e-12, 3.1041e-11, 3.7945e-12, 3.8995e-12, 1.3913e-12, 3.1169e-11,\n        1.0843e-12, 4.2025e-12, 1.6364e-10, 1.9875e-12, 1.0694e-12, 6.1327e-14,\n        1.0731e-11, 6.2050e-14, 6.6116e-12, 2.1062e-12, 5.0076e-12, 1.8663e-13,\n        5.5824e-14, 1.5522e-11, 5.0485e-15, 5.7787e-12, 2.5554e-12, 3.4328e-12,\n        6.5303e-13, 2.4833e-12, 2.0885e-11, 1.3242e-12, 8.1560e-12, 1.5130e-12,\n        8.4114e-13, 4.1184e-11, 2.4309e-13, 3.4977e-12, 4.3590e-12, 7.4955e-11,\n        7.7367e-12, 7.3242e-12, 3.8915e-13, 1.1700e-11, 6.8184e-12, 2.6008e-11,\n        7.5336e-14, 1.4159e-11, 1.0286e-12, 3.4379e-12, 3.5084e-12, 1.0309e-11,\n        4.2182e-13, 2.7717e-11, 2.9710e-11, 2.4216e-11, 1.6982e-12, 1.7866e-12,\n        5.1511e-13, 1.6160e-13, 7.1383e-12, 5.0577e-13, 7.8483e-13, 5.6390e-13,\n        7.1062e-12, 1.8075e-14, 2.1990e-11, 3.1039e-12, 1.6819e-12, 6.6563e-15,\n        4.5052e-14, 3.8531e-12, 1.1434e-11, 7.7605e-12, 2.2687e-11, 1.2984e-12,\n        3.6085e-11, 2.3900e-11, 1.1319e-12, 1.2441e-12, 4.0152e-12, 1.4176e-12,\n        2.5628e-12, 3.7252e-11, 5.4500e-12, 5.0948e-12, 8.0100e-12, 5.5443e-13,\n        5.5716e-12, 8.3933e-13, 2.1454e-13, 1.3549e-12, 1.3561e-12, 3.3914e-12,\n        5.5541e-13, 7.5377e-12, 8.9061e-13, 4.0688e-11, 3.8220e-13, 2.3584e-11,\n        4.3651e-11, 3.4576e-12, 3.0938e-11, 1.8135e-13, 1.5402e-12, 5.6775e-11,\n        3.0904e-12, 1.6233e-12, 1.6523e-12, 5.0673e-12, 1.4036e-13, 2.4254e-12,\n        1.2834e-10, 1.1369e-11, 1.7652e-13, 7.6975e-12, 4.5241e-12, 4.8111e-12,\n        4.1451e-11, 1.7598e-11, 3.5970e-12, 7.9793e-13, 1.1234e-11, 6.5868e-14,\n        4.5539e-14, 8.1656e-12, 5.8563e-12, 6.9893e-12, 1.1226e-13, 1.7867e-12,\n        2.3618e-12, 1.2088e-15, 9.1252e-12, 2.2106e-12, 4.2176e-13, 4.6283e-11,\n        2.7702e-12, 1.1545e-11, 3.2851e-14, 1.3724e-12, 3.1507e-13, 2.8098e-11,\n        5.2491e-14, 1.3954e-11, 4.8732e-11, 8.8889e-12, 1.4111e-12, 7.4843e-12,\n        1.1856e-13, 2.5821e-15, 6.4829e-12, 1.4943e-11, 3.6207e-11, 4.1576e-12,\n        5.0157e-12, 9.4214e-14, 1.3020e-11, 3.1314e-12, 2.6250e-13, 3.1764e-13,\n        7.7968e-13, 7.8124e-12, 3.4825e-14, 2.4092e-13, 1.0801e-12, 1.4236e-10,\n        2.2420e-12, 3.0886e-14, 3.5767e-13, 1.1357e-12, 2.2903e-11, 9.6892e-14,\n        1.6910e-11, 5.4298e-12, 2.9648e-12, 3.4732e-12, 2.9936e-13, 8.6075e-12,\n        6.0521e-12, 9.8921e-12, 8.9928e-12, 4.4497e-14, 4.5309e-11, 1.4277e-14,\n        5.8528e-12, 2.6474e-14, 3.1374e-12, 1.3721e-11, 3.4838e-13, 2.7369e-12,\n        9.4356e-12, 3.3264e-12, 3.1315e-11, 4.5349e-11, 8.1510e-11, 1.1565e-11,\n        1.4909e-12, 1.9397e-12, 5.7276e-11, 3.0898e-13, 5.5186e-12, 7.9992e-11,\n        4.4689e-13, 6.1861e-12, 7.3624e-11, 1.1372e-12, 1.6538e-12, 7.7097e-12,\n        1.2270e-11, 4.2093e-12, 4.7402e-12, 3.5530e-12, 7.1923e-12, 2.6532e-12,\n        8.0408e-12, 1.0540e-11, 1.1985e-12, 8.9123e-13, 5.4950e-12, 1.6481e-13,\n        5.0216e-11, 3.5187e-12, 2.1703e-13, 2.6493e-11, 8.1593e-12, 3.7793e-14,\n        9.8143e-12, 3.9687e-12, 8.8609e-11, 6.2742e-11, 1.3552e-12, 3.8646e-12,\n        1.5629e-12, 2.4470e-11, 4.9858e-13, 1.3616e-13], device='cuda:0')"
-      },
-      "46": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([ 5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([3.2221e-17, 2.3780e-16, 2.2105e-13, 4.4502e-18, 5.7181e-14, 3.8518e-14,\n        8.5349e-15, 9.1733e-15, 2.1897e-13, 5.0435e-15, 5.1417e-15, 1.6252e-14,\n        2.1843e-16, 3.5596e-15, 1.0637e-15, 3.0561e-15, 1.4958e-14, 3.8395e-15,\n        6.3977e-16, 4.3558e-16, 1.0857e-14, 1.7651e-15, 2.7146e-14, 1.6526e-15,\n        2.9408e-14, 3.7856e-14, 7.7684e-15, 7.3837e-15, 7.4168e-15, 6.0152e-14,\n        3.4436e-15, 8.1554e-15, 3.3546e-13, 2.7615e-15, 1.6998e-15, 2.9954e-16,\n        1.7507e-14, 1.1380e-15, 9.9424e-15, 5.3986e-15, 7.2674e-15, 2.8193e-16,\n        1.3108e-16, 2.9988e-14, 9.7396e-18, 1.2368e-14, 3.4146e-15, 1.0136e-14,\n        2.1498e-15, 3.3743e-15, 3.7161e-14, 1.6617e-15, 9.6183e-15, 2.3121e-15,\n        4.3459e-15, 1.1476e-13, 7.9065e-16, 5.3319e-15, 7.0912e-15, 1.5180e-13,\n        1.0942e-14, 1.1974e-14, 1.2795e-15, 1.2299e-14, 1.3653e-14, 4.1547e-14,\n        1.6777e-16, 2.3135e-14, 1.5251e-15, 5.0312e-15, 6.3649e-15, 2.5431e-14,\n        4.9046e-16, 4.7807e-14, 7.5061e-14, 3.6039e-14, 2.3256e-15, 2.3971e-15,\n        6.7792e-16, 1.1010e-16, 1.0003e-14, 6.9985e-16, 1.5180e-15, 8.1027e-16,\n        1.8185e-14, 1.3120e-16, 3.0670e-14, 1.3191e-14, 3.7638e-15, 1.7384e-17,\n        6.3621e-18, 7.0876e-15, 2.0757e-14, 1.1330e-14, 2.7984e-14, 1.2147e-15,\n        6.7419e-14, 3.7497e-14, 2.2319e-15, 4.8152e-15, 8.1164e-15, 2.7867e-15,\n        5.9513e-15, 8.7768e-14, 1.0010e-14, 6.8157e-15, 1.1766e-14, 1.1253e-15,\n        5.0218e-15, 2.0101e-15, 6.8028e-16, 5.4253e-15, 4.8569e-15, 4.5639e-15,\n        6.9823e-16, 2.0864e-14, 2.5707e-15, 1.3948e-13, 1.5823e-15, 5.5167e-14,\n        7.8781e-14, 7.5435e-15, 7.3501e-14, 2.0559e-15, 4.5517e-15, 1.2382e-13,\n        2.7912e-15, 1.8520e-15, 3.6641e-15, 9.9568e-15, 4.3837e-16, 3.8521e-15,\n        4.6215e-13, 1.2873e-14, 4.6035e-16, 1.7221e-14, 1.6966e-14, 9.8406e-15,\n        1.0068e-13, 1.8308e-14, 7.3890e-15, 1.9142e-15, 1.8010e-14, 2.1470e-16,\n        1.5325e-16, 2.7532e-14, 9.3528e-15, 1.0999e-14, 3.1553e-16, 2.6652e-15,\n        7.8695e-15, 1.5046e-17, 1.1709e-14, 3.4353e-15, 6.5927e-16, 6.6015e-14,\n        3.7573e-15, 1.5267e-14, 3.2925e-17, 2.5951e-15, 6.9435e-16, 5.8913e-14,\n        5.5021e-16, 3.1213e-14, 1.1712e-13, 1.1651e-14, 3.0052e-15, 1.6543e-14,\n        4.9977e-16, 9.0270e-17, 1.2791e-14, 6.5730e-14, 8.1586e-14, 5.7642e-15,\n        1.3801e-14, 2.5489e-16, 1.7707e-14, 4.8065e-15, 1.7173e-15, 1.8022e-15,\n        1.9385e-15, 1.9680e-14, 1.1710e-16, 6.0444e-16, 4.8995e-15, 3.9662e-13,\n        2.7069e-15, 1.0599e-16, 5.5688e-16, 3.7629e-15, 6.9629e-14, 4.1602e-16,\n        2.6357e-14, 7.7124e-15, 6.5353e-15, 8.6329e-15, 4.4826e-16, 1.2106e-14,\n        8.2894e-15, 1.3775e-14, 1.7414e-14, 2.2226e-16, 5.2368e-14, 9.3278e-17,\n        9.4145e-15, 8.0152e-17, 4.6441e-15, 2.3519e-14, 2.5581e-15, 5.9930e-15,\n        1.4540e-14, 4.3575e-15, 6.9603e-14, 6.9764e-14, 1.2537e-13, 3.0654e-14,\n        3.4575e-15, 2.6722e-15, 1.1362e-13, 1.1116e-15, 8.0064e-15, 1.9330e-13,\n        1.0896e-15, 1.5530e-14, 2.0358e-13, 1.9662e-15, 7.9374e-15, 9.0728e-15,\n        2.0246e-14, 8.4919e-15, 8.6440e-15, 4.9153e-15, 1.8549e-14, 4.6651e-15,\n        1.1298e-14, 2.1960e-14, 1.8951e-15, 1.4055e-15, 1.0752e-14, 4.4425e-16,\n        1.6073e-13, 5.8103e-15, 9.0563e-16, 3.8933e-14, 8.9763e-15, 9.2670e-17,\n        1.8807e-14, 4.6141e-15, 2.2445e-13, 2.5750e-13, 2.0312e-15, 4.4503e-15,\n        2.6517e-15, 6.0416e-14, 1.2384e-15, 2.6802e-16], device='cuda:0')"
-      },
-      "47": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([ 5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([8.1858e-17, 1.6639e-16, 2.7344e-13, 4.0303e-17, 8.0904e-14, 5.0434e-14,\n        1.4054e-14, 9.6806e-15, 2.6930e-13, 6.0823e-15, 8.4453e-15, 2.3687e-14,\n        4.9485e-16, 7.1055e-15, 2.0272e-15, 2.6819e-15, 2.1727e-14, 6.0245e-15,\n        9.6759e-16, 5.8194e-16, 2.0680e-14, 1.7442e-15, 2.4594e-14, 3.4882e-15,\n        2.7245e-14, 7.8035e-14, 1.1899e-14, 1.0575e-14, 6.1116e-15, 8.8312e-14,\n        3.1024e-15, 1.1024e-14, 4.4564e-13, 4.7860e-15, 3.3271e-15, 4.3636e-16,\n        2.4975e-14, 1.1739e-15, 2.0284e-14, 6.8328e-15, 1.2723e-14, 7.8369e-16,\n        2.3770e-16, 4.7221e-14, 2.5042e-17, 1.2960e-14, 7.3817e-15, 1.2047e-14,\n        2.4212e-15, 7.7067e-15, 6.3008e-14, 2.7911e-15, 2.0380e-14, 4.1406e-15,\n        3.5066e-15, 1.0424e-13, 1.1721e-15, 8.0569e-15, 1.1376e-14, 2.1048e-13,\n        1.7826e-14, 1.6967e-14, 1.5520e-15, 3.3078e-14, 1.9758e-14, 6.1289e-14,\n        3.7028e-16, 3.4984e-14, 3.1772e-15, 1.0662e-14, 1.1249e-14, 3.1870e-14,\n        8.9313e-16, 7.8399e-14, 8.4882e-14, 6.2899e-14, 5.4898e-15, 5.8575e-15,\n        1.7541e-15, 1.7928e-16, 2.1756e-14, 1.4455e-15, 2.4436e-15, 1.8492e-15,\n        1.6146e-14, 3.9640e-16, 6.4260e-14, 1.1268e-14, 6.0600e-15, 2.4191e-17,\n        6.2853e-17, 9.8151e-15, 2.8063e-14, 2.0515e-14, 5.8271e-14, 2.9494e-15,\n        9.3787e-14, 6.0913e-14, 3.0223e-15, 5.1438e-15, 1.2453e-14, 4.5029e-15,\n        8.5421e-15, 9.5424e-14, 1.6047e-14, 1.2753e-14, 2.3025e-14, 2.0861e-15,\n        1.2557e-14, 2.4618e-15, 6.7453e-16, 5.1971e-15, 3.7260e-15, 7.7115e-15,\n        1.9785e-15, 2.1205e-14, 3.5544e-15, 9.9211e-14, 1.9108e-15, 6.7231e-14,\n        1.2303e-13, 1.1390e-14, 7.9602e-14, 3.1206e-15, 5.6612e-15, 1.5170e-13,\n        7.2157e-15, 4.9385e-15, 5.4610e-15, 1.5062e-14, 7.3519e-16, 7.7652e-15,\n        3.2715e-13, 2.8498e-14, 7.3461e-16, 1.8604e-14, 1.3507e-14, 1.0996e-14,\n        1.0254e-13, 4.8244e-14, 1.0600e-14, 2.4725e-15, 2.7765e-14, 4.6268e-16,\n        2.1681e-16, 2.4749e-14, 1.2235e-14, 1.9781e-14, 6.9202e-16, 5.8674e-15,\n        8.1988e-15, 5.8142e-19, 2.3760e-14, 5.3514e-15, 1.3173e-15, 1.1447e-13,\n        7.3570e-15, 2.9697e-14, 4.4526e-17, 4.8173e-15, 1.3454e-15, 7.8570e-14,\n        6.9072e-16, 3.9804e-14, 1.3756e-13, 2.2732e-14, 3.3569e-15, 1.8760e-14,\n        7.5448e-16, 1.0051e-16, 1.5081e-14, 4.4393e-14, 8.9906e-14, 1.2726e-14,\n        1.2458e-14, 3.7392e-16, 3.7139e-14, 9.5609e-15, 1.7711e-15, 1.5703e-15,\n        3.1307e-15, 2.0974e-14, 1.5723e-16, 9.6011e-16, 4.3199e-15, 3.8847e-13,\n        5.5342e-15, 2.1953e-16, 1.2891e-15, 4.4739e-15, 5.7559e-14, 8.5342e-16,\n        4.2175e-14, 1.7165e-14, 7.6742e-15, 1.1292e-14, 1.0010e-15, 2.1668e-14,\n        1.5842e-14, 2.9142e-14, 2.2335e-14, 7.2067e-16, 1.2137e-13, 2.9658e-16,\n        1.3719e-14, 5.9972e-17, 9.7136e-15, 3.6628e-14, 2.1907e-15, 8.4198e-15,\n        2.6353e-14, 7.4431e-15, 8.7313e-14, 1.1890e-13, 2.2969e-13, 2.7633e-14,\n        5.4240e-15, 4.6003e-15, 1.4608e-13, 1.3887e-15, 1.2949e-14, 2.1346e-13,\n        1.6872e-15, 1.9151e-14, 2.0396e-13, 1.9906e-15, 6.0261e-15, 2.2228e-14,\n        3.5790e-14, 1.1408e-14, 1.4117e-14, 1.0162e-14, 1.6229e-14, 8.1271e-15,\n        2.2605e-14, 3.0292e-14, 3.4933e-15, 2.7743e-15, 1.2864e-14, 6.5563e-16,\n        1.2506e-13, 9.9274e-15, 1.6017e-15, 7.4896e-14, 1.9336e-14, 2.3273e-16,\n        2.6912e-14, 9.7578e-15, 2.4297e-13, 1.7713e-13, 4.4225e-15, 9.5167e-15,\n        3.5463e-15, 6.8024e-14, 1.8365e-15, 6.0661e-16], device='cuda:0')"
-      },
-      "48": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([[ 5.6052e-45, -5.6052e-45, -5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [-5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        ...,\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [ 5.6052e-45, -5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45],\n        [ 5.6052e-45, -5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          0.0000e+00,  5.6052e-45]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[1.1352e-17, 2.0626e-18, 3.2239e-17,  ..., 3.7569e-17, 0.0000e+00,\n         5.0946e-18],\n        [2.2881e-18, 9.3955e-18, 1.1877e-16,  ..., 1.6057e-16, 0.0000e+00,\n         2.8070e-16],\n        [1.8980e-14, 3.7107e-16, 7.9353e-14,  ..., 2.0159e-14, 0.0000e+00,\n         1.1383e-13],\n        ...,\n        [2.5054e-15, 1.0374e-16, 2.8864e-14,  ..., 8.5446e-15, 0.0000e+00,\n         9.1856e-15],\n        [3.1177e-16, 3.6359e-19, 9.4639e-18,  ..., 3.9731e-17, 0.0000e+00,\n         4.0726e-17],\n        [1.1707e-18, 6.9542e-19, 2.5897e-16,  ..., 2.1996e-17, 0.0000e+00,\n         1.7356e-16]], device='cuda:0')"
-      },
-      "49": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([ 5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([9.7074e-15, 8.4739e-14, 4.5459e-11, 4.7209e-14, 7.4147e-12, 2.3895e-11,\n        1.1208e-11, 2.5251e-12, 1.1931e-12, 2.4389e-12, 3.1478e-11, 9.2699e-12,\n        1.7307e-12, 6.1868e-12, 2.6861e-14, 1.7546e-12, 4.6655e-12, 2.2068e-12,\n        3.5258e-13, 7.9562e-13, 1.1407e-11, 2.2033e-13, 9.3112e-13, 3.5118e-12,\n        5.8817e-12, 3.0710e-12, 4.8675e-12, 9.9004e-12, 1.7665e-12, 1.7846e-11,\n        2.4188e-11, 2.8251e-11, 1.5341e-10, 4.8359e-13, 6.6895e-12, 1.3319e-13,\n        1.3697e-11, 2.2618e-13, 5.0532e-11, 1.4660e-12, 8.0680e-12, 1.8885e-13,\n        8.3608e-13, 3.2911e-11, 1.4984e-14, 1.6554e-12, 7.0559e-12, 4.8729e-12,\n        9.5765e-13, 9.2834e-12, 1.7130e-11, 5.7221e-12, 1.8691e-11, 1.5239e-12,\n        1.9374e-12, 4.2922e-11, 1.4852e-13, 6.5507e-12, 3.4238e-11, 4.2493e-11,\n        5.4331e-12, 4.8454e-12, 1.9277e-13, 3.3510e-11, 1.5998e-12, 4.4522e-11,\n        5.4091e-14, 1.1303e-11, 6.3632e-12, 6.3911e-12, 1.3135e-11, 2.1166e-11,\n        5.6491e-12, 1.5607e-11, 6.8964e-12, 2.6405e-11, 1.5133e-11, 5.9661e-12,\n        9.1863e-13, 3.5415e-14, 3.1450e-11, 2.2760e-12, 1.8144e-12, 1.2206e-11,\n        2.8732e-12, 4.2165e-13, 6.8155e-11, 2.0992e-12, 2.4031e-12, 1.0226e-13,\n        2.5338e-14, 2.5834e-12, 9.6755e-12, 2.1273e-12, 5.9385e-11, 1.8328e-12,\n        3.5148e-11, 4.8572e-11, 7.0016e-12, 4.2317e-12, 4.2796e-12, 1.8152e-12,\n        8.0959e-13, 1.4026e-11, 8.1052e-12, 5.7312e-12, 8.2805e-12, 6.3858e-13,\n        5.3826e-12, 6.4865e-12, 3.0654e-12, 7.6821e-13, 3.7586e-11, 9.4421e-13,\n        1.1766e-12, 3.7463e-11, 1.2125e-12, 7.7899e-12, 1.4163e-12, 2.3996e-12,\n        8.1095e-11, 6.7266e-12, 3.0222e-11, 5.3976e-13, 3.2111e-13, 9.4800e-11,\n        1.2372e-11, 6.4626e-13, 3.2550e-13, 1.3815e-11, 2.0089e-15, 2.8278e-12,\n        4.9681e-11, 3.1397e-11, 1.2766e-13, 3.1757e-12, 7.6783e-12, 2.5984e-12,\n        2.8098e-11, 2.0552e-11, 3.2856e-12, 2.9949e-12, 1.3961e-11, 2.1596e-14,\n        1.0695e-12, 1.6129e-12, 1.0349e-12, 1.7090e-11, 9.9508e-13, 1.6254e-12,\n        2.2522e-12, 2.9376e-12, 6.6404e-12, 5.0107e-12, 1.9254e-12, 2.2871e-11,\n        5.0864e-13, 7.4981e-11, 2.7005e-14, 1.3875e-12, 5.6270e-13, 1.9278e-12,\n        2.8955e-14, 3.5073e-12, 2.1674e-11, 1.4135e-11, 8.6892e-13, 4.0320e-12,\n        4.0693e-15, 7.0351e-13, 3.2681e-13, 6.0221e-12, 4.8332e-12, 5.9759e-12,\n        3.1472e-12, 7.8831e-13, 2.5789e-11, 1.5031e-11, 1.4833e-13, 3.7917e-14,\n        4.0953e-12, 1.1100e-11, 3.1086e-14, 2.3807e-12, 1.6592e-12, 3.2570e-11,\n        6.5686e-13, 1.6829e-13, 1.7680e-12, 7.6805e-13, 3.2768e-12, 3.8176e-14,\n        7.7533e-12, 2.5366e-12, 1.7931e-11, 2.1839e-12, 1.0009e-12, 7.2528e-12,\n        3.5241e-11, 3.4000e-12, 4.9903e-12, 1.0581e-13, 6.0200e-11, 1.3413e-14,\n        7.4542e-12, 3.0299e-14, 8.1999e-12, 8.1505e-12, 3.4362e-13, 5.0447e-12,\n        4.7551e-11, 2.8192e-12, 3.3529e-12, 4.0090e-11, 1.5893e-10, 2.0186e-12,\n        1.6566e-12, 1.6784e-12, 8.3836e-11, 9.1958e-14, 1.5563e-12, 1.5991e-10,\n        2.1606e-13, 1.2180e-11, 5.7288e-11, 9.9476e-13, 1.7140e-12, 2.5180e-11,\n        6.1308e-12, 1.8264e-12, 3.3422e-11, 7.2023e-12, 4.1228e-13, 2.6615e-12,\n        1.1448e-11, 1.1361e-11, 1.7014e-12, 2.1190e-12, 1.2374e-12, 7.7797e-13,\n        7.0346e-12, 5.2606e-12, 4.2974e-13, 7.2566e-11, 9.6976e-12, 8.9040e-14,\n        7.9561e-12, 3.2749e-11, 4.0929e-11, 3.5675e-11, 1.1271e-12, 3.8958e-12,\n        1.5570e-12, 1.4829e-11, 1.6097e-13, 4.1223e-14], device='cuda:0')"
-      },
-      "50": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([ 5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([7.5161e-18, 3.2417e-16, 6.7381e-14, 7.1437e-16, 6.4176e-15, 5.4490e-14,\n        3.1147e-14, 5.5932e-15, 1.1664e-14, 4.2085e-15, 6.7246e-14, 1.6491e-14,\n        7.5282e-15, 1.1251e-14, 1.1000e-16, 6.2764e-15, 7.7828e-15, 4.3851e-15,\n        1.3557e-15, 1.4576e-15, 2.1276e-14, 3.5980e-16, 1.7946e-15, 4.9664e-15,\n        1.1145e-14, 6.3362e-15, 6.2196e-15, 4.3180e-14, 1.1520e-14, 2.1722e-14,\n        3.2215e-14, 6.2595e-14, 2.1653e-13, 5.7028e-16, 2.6025e-14, 1.8488e-16,\n        2.7411e-14, 1.6935e-15, 1.5950e-13, 2.0093e-15, 8.3066e-15, 5.4595e-16,\n        1.4960e-15, 4.9496e-14, 1.3414e-18, 1.9241e-15, 1.6870e-14, 1.0598e-14,\n        4.8612e-15, 1.5498e-14, 3.3978e-14, 1.0972e-14, 3.3345e-14, 1.9363e-15,\n        4.1192e-15, 9.9610e-14, 3.8548e-16, 1.8762e-14, 4.9160e-14, 5.4923e-14,\n        5.8179e-15, 7.0227e-15, 3.1932e-16, 6.2177e-14, 2.4993e-15, 1.2984e-13,\n        1.5160e-16, 1.4347e-14, 1.6139e-14, 7.5905e-15, 3.1037e-14, 4.9519e-14,\n        1.3862e-14, 2.0974e-14, 9.0383e-15, 4.0017e-14, 3.1495e-14, 8.6095e-15,\n        1.2842e-15, 1.0717e-16, 5.2886e-14, 6.5543e-15, 4.8536e-15, 4.4202e-14,\n        3.6407e-15, 1.1200e-15, 1.4523e-13, 2.3290e-15, 4.5804e-15, 1.5834e-16,\n        1.5985e-18, 4.8335e-15, 1.5532e-14, 4.0917e-15, 1.2350e-13, 2.6559e-15,\n        4.6763e-14, 1.0577e-13, 9.7787e-15, 2.4542e-14, 8.0957e-15, 3.1915e-15,\n        1.3863e-15, 2.4524e-14, 1.4708e-14, 8.4338e-15, 1.3881e-14, 1.9359e-15,\n        7.0694e-15, 7.6874e-15, 4.9027e-15, 1.0514e-15, 6.6820e-14, 8.4742e-16,\n        1.2980e-15, 5.4787e-14, 3.0449e-15, 1.3339e-14, 5.0948e-15, 3.9880e-15,\n        1.9991e-13, 9.6664e-15, 6.9496e-14, 4.5320e-15, 9.5460e-16, 2.7071e-13,\n        3.4532e-14, 1.2416e-15, 9.0878e-16, 3.5242e-14, 7.8519e-18, 6.0198e-15,\n        6.9224e-14, 5.5490e-14, 3.3889e-16, 4.8035e-15, 2.5986e-14, 3.0227e-15,\n        5.0702e-14, 2.6918e-14, 6.5423e-15, 5.1314e-15, 1.8199e-14, 1.5389e-16,\n        1.5068e-15, 2.4658e-15, 8.1754e-16, 5.4884e-14, 3.3098e-15, 2.6643e-15,\n        6.6734e-15, 6.7834e-15, 1.0744e-14, 8.8811e-15, 4.2474e-15, 4.7379e-14,\n        8.6664e-16, 1.7680e-13, 3.6993e-17, 1.9283e-15, 1.2641e-15, 4.4244e-15,\n        1.2190e-16, 5.9446e-15, 4.5334e-14, 2.5169e-14, 9.0794e-16, 6.1427e-15,\n        5.3386e-17, 1.7291e-15, 5.1797e-16, 1.3432e-14, 7.0209e-15, 9.3851e-15,\n        5.8898e-15, 1.1520e-15, 3.9713e-14, 2.4310e-14, 1.6556e-15, 1.3808e-16,\n        1.4804e-14, 3.1149e-14, 5.6152e-17, 4.5944e-15, 6.4669e-15, 3.9752e-14,\n        1.5151e-15, 3.7299e-16, 5.5258e-15, 1.3405e-15, 4.9918e-15, 1.9306e-16,\n        9.2315e-15, 3.8055e-15, 3.3350e-14, 4.0148e-15, 1.4613e-15, 1.0548e-14,\n        1.1539e-13, 5.3202e-15, 1.0442e-14, 6.4257e-16, 7.1057e-14, 1.9243e-16,\n        1.8330e-14, 2.5551e-17, 2.6726e-14, 9.3251e-15, 2.2891e-15, 1.4846e-14,\n        1.4192e-13, 4.2735e-15, 8.1576e-15, 4.1608e-14, 2.6323e-13, 3.4532e-15,\n        3.5396e-15, 1.5263e-15, 2.5176e-13, 2.8912e-16, 1.2635e-15, 3.5150e-13,\n        5.6330e-16, 4.0634e-14, 1.2275e-13, 1.6102e-15, 3.2467e-15, 6.5383e-14,\n        6.0400e-15, 2.6158e-15, 6.4990e-14, 1.4492e-14, 7.1409e-16, 3.5298e-15,\n        1.8839e-14, 2.7794e-14, 3.4056e-15, 3.5335e-15, 1.9351e-15, 1.8319e-15,\n        9.0593e-15, 1.0745e-14, 1.9642e-15, 1.6441e-13, 1.5184e-14, 8.0277e-16,\n        1.1993e-14, 7.2284e-14, 5.6527e-14, 7.3144e-14, 1.7943e-15, 6.4641e-15,\n        1.3361e-15, 2.0806e-14, 3.4331e-16, 2.2775e-16], device='cuda:0')"
-      },
-      "51": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([ 5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([7.6379e-17, 3.7910e-16, 1.2107e-13, 4.0372e-16, 1.9721e-14, 6.0826e-14,\n        3.0604e-14, 8.0001e-15, 3.8417e-15, 7.4320e-15, 8.7357e-14, 2.2672e-14,\n        4.7440e-15, 1.7985e-14, 1.3959e-16, 5.4812e-15, 1.3528e-14, 6.3824e-15,\n        1.3710e-15, 2.4901e-15, 3.2214e-14, 6.7475e-16, 3.6235e-15, 1.0647e-14,\n        1.7002e-14, 8.7719e-15, 1.3310e-14, 2.6390e-14, 9.0026e-15, 4.6925e-14,\n        6.3866e-14, 7.3817e-14, 4.1449e-13, 1.2142e-15, 1.9512e-14, 4.2063e-16,\n        3.5068e-14, 1.9142e-15, 1.3818e-13, 4.3924e-15, 2.1632e-14, 6.6947e-16,\n        2.4540e-15, 8.9225e-14, 4.1435e-17, 4.0651e-15, 1.9237e-14, 1.4448e-14,\n        3.4017e-15, 2.6034e-14, 4.7031e-14, 1.3200e-14, 4.8777e-14, 4.1095e-15,\n        5.2877e-15, 1.0790e-13, 6.8194e-16, 1.6696e-14, 8.7767e-14, 1.1088e-13,\n        1.4360e-14, 1.1941e-14, 6.1398e-16, 8.8449e-14, 4.8028e-15, 1.1438e-13,\n        1.8384e-16, 2.9105e-14, 1.7730e-14, 1.7184e-14, 3.6616e-14, 5.7287e-14,\n        1.3561e-14, 4.2172e-14, 2.0171e-14, 6.7424e-14, 4.1020e-14, 1.6767e-14,\n        2.7007e-15, 1.0652e-16, 8.2815e-14, 6.2717e-15, 5.3213e-15, 3.3885e-14,\n        6.9593e-15, 1.4408e-15, 1.8103e-13, 5.7175e-15, 7.8887e-15, 1.2081e-16,\n        1.0556e-16, 7.0599e-15, 2.5661e-14, 6.1505e-15, 1.5272e-13, 4.4895e-15,\n        9.0875e-14, 1.2470e-13, 1.6526e-14, 1.3845e-14, 1.2400e-14, 5.6630e-15,\n        2.5456e-15, 3.8437e-14, 2.3362e-14, 1.5184e-14, 2.3161e-14, 2.4824e-15,\n        1.3503e-14, 1.7209e-14, 9.0586e-15, 2.1674e-15, 9.8422e-14, 2.2708e-15,\n        3.5653e-15, 1.0424e-13, 4.1090e-15, 2.0438e-14, 5.4788e-15, 7.0667e-15,\n        2.1977e-13, 1.9510e-14, 7.8055e-14, 5.3797e-15, 1.2535e-15, 2.4247e-13,\n        3.0464e-14, 2.0201e-15, 1.1802e-15, 3.7769e-14, 2.6220e-17, 8.6423e-15,\n        1.2439e-13, 7.8605e-14, 4.8596e-16, 8.6190e-15, 2.2642e-14, 6.2897e-15,\n        7.1699e-14, 5.3172e-14, 9.4249e-15, 8.9476e-15, 3.5783e-14, 6.4368e-16,\n        3.3266e-15, 4.5405e-15, 2.6092e-15, 4.5339e-14, 2.9314e-15, 4.6406e-15,\n        6.5842e-15, 6.7999e-15, 1.8060e-14, 1.2404e-14, 6.2464e-15, 6.3027e-14,\n        1.5201e-15, 1.9760e-13, 7.5966e-17, 4.3822e-15, 1.8717e-15, 6.0197e-15,\n        2.8033e-16, 1.0996e-14, 5.7294e-14, 3.6597e-14, 2.2991e-15, 1.2608e-14,\n        1.9122e-16, 2.3541e-15, 7.9142e-16, 1.7557e-14, 1.3386e-14, 1.7594e-14,\n        9.1923e-15, 2.4230e-15, 6.7554e-14, 4.1789e-14, 2.0684e-15, 1.8825e-16,\n        1.1562e-14, 3.2297e-14, 1.0999e-16, 6.8538e-15, 5.5994e-15, 9.0526e-14,\n        1.7395e-15, 6.9443e-16, 5.6429e-15, 2.2741e-15, 1.0691e-14, 5.9086e-16,\n        1.9020e-14, 7.4463e-15, 4.4230e-14, 6.8853e-15, 3.2918e-15, 1.8939e-14,\n        8.8149e-14, 1.0601e-14, 1.4218e-14, 1.1883e-15, 1.4972e-13, 4.6770e-16,\n        1.7809e-14, 3.5999e-17, 2.2824e-14, 2.0302e-14, 2.4885e-15, 1.5281e-14,\n        1.2895e-13, 5.9094e-15, 1.0369e-14, 1.1379e-13, 4.1966e-13, 6.3487e-15,\n        4.5560e-15, 3.8610e-15, 2.2251e-13, 4.5176e-16, 3.6409e-15, 4.2719e-13,\n        7.6927e-16, 3.2542e-14, 1.5262e-13, 2.2156e-15, 4.8735e-15, 6.8892e-14,\n        1.7903e-14, 5.3348e-15, 9.1449e-14, 1.9333e-14, 8.9982e-16, 7.4665e-15,\n        3.2703e-14, 3.0842e-14, 5.1475e-15, 6.7264e-15, 3.4670e-15, 2.1725e-15,\n        1.9687e-14, 1.5335e-14, 2.6593e-15, 1.9081e-13, 2.6676e-14, 1.1402e-15,\n        2.1925e-14, 8.8225e-14, 1.0990e-13, 9.9626e-14, 3.1576e-15, 1.0534e-14,\n        3.7910e-15, 4.0610e-14, 4.9390e-16, 1.7100e-16], device='cuda:0')"
-      },
-      "52": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([[-5.6052e-45, -5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n         -5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45, -5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n         -5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45, -5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n         -5.6052e-45,  5.6052e-45],\n        ...,\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        [-5.6052e-45, -5.6052e-45, -5.6052e-45,  ..., -5.6052e-45,\n         -5.6052e-45, -5.6052e-45]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[1.9088e-17, 1.4973e-17, 2.7805e-17,  ..., 1.8096e-18, 8.5794e-19,\n         1.3632e-16],\n        [4.3612e-18, 1.2880e-17, 4.2723e-17,  ..., 1.6876e-16, 1.0080e-17,\n         2.2346e-17],\n        [2.6576e-17, 9.5552e-19, 1.4119e-17,  ..., 8.1545e-17, 8.4221e-19,\n         1.1857e-18],\n        ...,\n        [3.7001e-14, 6.0222e-13, 8.1862e-13,  ..., 1.4011e-12, 7.1514e-13,\n         3.5374e-13],\n        [1.1024e-14, 1.8264e-13, 2.4941e-13,  ..., 4.0306e-13, 2.2700e-13,\n         1.1772e-13],\n        [4.5320e-15, 5.3266e-14, 7.9200e-14,  ..., 1.3029e-13, 5.6265e-14,\n         2.7879e-14]], device='cuda:0')"
-      },
-      "53": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([-5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([1.5030e-16, 1.2033e-16, 3.6158e-16, 4.3927e-17, 1.7953e-16, 2.8655e-16,\n        8.1683e-17, 3.4874e-16, 8.5335e-16, 2.8842e-15, 1.0427e-16, 4.3297e-17,\n        1.2022e-15, 3.2846e-16, 2.2252e-15, 1.3137e-18, 2.2382e-16, 5.0257e-17,\n        5.9053e-16, 1.6287e-15, 1.3915e-16, 3.6341e-19, 6.2304e-18, 3.4628e-17,\n        2.1056e-16, 1.4621e-16, 3.1464e-16, 4.4039e-16, 4.0741e-17, 3.3485e-16,\n        9.2019e-16, 1.0668e-16, 6.3573e-16, 1.0061e-15, 2.2585e-15, 1.3883e-15,\n        3.5670e-16, 3.7955e-16, 6.9729e-16, 1.5652e-16, 2.7827e-17, 4.8357e-16,\n        5.0564e-16, 2.4417e-18, 3.0529e-15, 1.4523e-15, 1.3821e-17, 4.3511e-18,\n        1.6472e-16, 5.1954e-16, 5.6672e-16, 3.5148e-16, 4.0218e-16, 1.8270e-16,\n        1.6649e-15, 6.2191e-16, 9.2419e-17, 1.7225e-15, 1.3198e-17, 3.2820e-16,\n        4.8614e-16, 1.2252e-16, 1.0234e-17, 3.1667e-15, 7.9721e-17, 1.9019e-15,\n        1.1794e-17, 1.9913e-15, 3.6852e-16, 3.7442e-15, 4.4203e-15, 1.0713e-16,\n        1.0662e-15, 2.9442e-15, 2.0987e-15, 1.9514e-16, 5.1911e-15, 1.9877e-15,\n        7.2849e-15, 1.9621e-17, 2.0797e-16, 6.2376e-16, 2.4694e-16, 8.3332e-16,\n        2.6343e-15, 9.6957e-17, 1.9619e-16, 6.3558e-16, 7.5790e-15, 1.8197e-16,\n        1.4372e-15, 1.0229e-17, 5.9640e-15, 5.7337e-15, 9.8354e-16, 8.2919e-16,\n        6.5541e-18, 1.0609e-15, 1.0897e-15, 8.2907e-16, 2.4169e-15, 5.8989e-18,\n        3.1290e-17, 1.2344e-17, 6.3702e-18, 1.0085e-16, 2.5662e-17, 1.4676e-16,\n        5.9641e-16, 1.8918e-15, 2.9047e-16, 1.5169e-16, 8.6972e-17, 4.7852e-16,\n        1.3167e-16, 1.8321e-15, 6.8405e-16, 1.3351e-15, 2.0841e-15, 7.0428e-16,\n        2.8270e-17, 6.6832e-18, 1.3307e-15, 1.7973e-16, 1.2319e-16, 7.0940e-17,\n        2.3100e-15, 3.7267e-16, 5.7320e-17, 2.1158e-15, 9.0282e-17, 2.5541e-16,\n        3.6162e-16, 1.4009e-15, 2.6499e-15, 2.4433e-17, 1.0154e-16, 2.0166e-16,\n        1.4468e-16, 3.7966e-18, 5.9881e-16, 1.6655e-16, 4.8718e-17, 2.1894e-18,\n        4.8273e-16, 7.6731e-16, 1.1443e-15, 5.0164e-16, 8.8946e-17, 3.6526e-16,\n        8.5643e-18, 1.2143e-15, 1.2618e-17, 3.6602e-17, 1.8230e-15, 8.9932e-16,\n        2.9312e-18, 1.8087e-18, 1.2083e-16, 8.0881e-16, 3.6040e-15, 2.2355e-16,\n        2.1654e-16, 1.1455e-16, 3.1717e-16, 3.7409e-16, 3.0559e-16, 5.9733e-16,\n        6.9244e-18, 7.4796e-17, 8.5160e-16, 1.7150e-16, 2.1619e-16, 1.2649e-15,\n        7.9487e-19, 9.1414e-16, 1.1979e-15, 1.0900e-15, 9.2516e-18, 6.0697e-17,\n        6.9439e-17, 4.4130e-16, 1.1301e-18, 2.8590e-17, 2.0601e-16, 2.4062e-15,\n        2.2992e-16, 3.7554e-17, 1.0109e-15, 1.1461e-16, 9.8784e-17, 8.2893e-16,\n        5.8375e-16, 6.9525e-16, 3.7795e-16, 1.0749e-15, 1.2974e-18, 7.8222e-17,\n        5.8334e-16, 6.6346e-16, 1.1030e-17, 1.7091e-15, 8.1652e-17, 9.8791e-16,\n        5.3649e-17, 4.2069e-17, 1.6441e-17, 3.2359e-16, 2.0820e-15, 6.6790e-17,\n        3.4020e-16, 5.4958e-16, 1.3166e-15, 1.3994e-16, 7.2739e-18, 1.5676e-16,\n        9.1442e-17, 8.8781e-17, 4.4403e-17, 9.9208e-17, 2.5303e-17, 6.8817e-15,\n        1.8271e-15, 1.1357e-15, 3.4298e-18, 3.7691e-17, 3.7616e-16, 1.2152e-15,\n        3.6036e-15, 1.1500e-15, 4.0552e-17, 2.2085e-16, 1.2777e-15, 1.1181e-17,\n        7.3198e-16, 5.5802e-16, 1.5095e-15, 3.2981e-16, 2.3810e-16, 4.9127e-17,\n        9.8406e-17, 4.3789e-17, 1.1635e-15, 2.9990e-16, 2.8872e-16, 2.0398e-15,\n        3.5957e-16, 3.4429e-18, 7.8661e-16, 7.9789e-16, 3.3263e-15, 6.8238e-16,\n        9.7932e-16, 5.3557e-16, 8.3972e-19, 4.2418e-16, 2.2978e-30, 5.6272e-32,\n        3.5795e-31, 6.5112e-32, 3.6775e-31, 1.0948e-31, 5.9114e-31, 1.4252e-32,\n        3.2284e-31, 5.0209e-31, 4.6983e-31, 1.2069e-32, 3.0504e-34, 2.5644e-32,\n        1.4422e-31, 1.0778e-32, 2.5042e-31, 1.9182e-32, 2.2839e-31, 5.4957e-32,\n        3.5313e-33, 1.7712e-31, 1.0017e-32, 2.0640e-31, 2.9222e-33, 1.7662e-31,\n        3.9021e-31, 1.6875e-32, 1.0185e-30, 3.4740e-31, 5.6510e-32, 8.1403e-32,\n        1.0426e-31, 6.2391e-31, 5.8241e-32, 4.8005e-32, 1.9514e-32, 1.0576e-31,\n        2.8486e-31, 5.6322e-32, 9.9889e-33, 1.1582e-31, 2.3294e-32, 5.1514e-32,\n        4.8687e-32, 4.3910e-31, 1.7973e-31, 6.0212e-32, 6.4553e-32, 4.0085e-32,\n        8.0251e-32, 2.6591e-31, 2.6820e-31, 3.3614e-31, 1.3690e-31, 3.1483e-32,\n        2.7990e-31, 4.8179e-31, 6.6033e-31, 2.8822e-31, 2.3279e-31, 1.6134e-31,\n        1.1100e-32, 9.9905e-32, 7.5284e-34, 3.8992e-31, 1.0324e-30, 3.2184e-31,\n        8.6953e-33, 5.1672e-32, 3.1419e-31, 1.0506e-31, 1.7039e-31, 1.0248e-32,\n        8.7771e-32, 4.7469e-32, 1.6700e-31, 4.7043e-31, 1.6206e-30, 5.9368e-31,\n        6.4312e-32, 4.8721e-31, 5.9019e-32, 7.8748e-31, 9.6346e-32, 8.8921e-32,\n        7.9163e-31, 2.7524e-31, 1.4262e-31, 1.8249e-32, 3.1875e-32, 1.7352e-31,\n        1.8200e-31, 3.8298e-32, 1.4797e-31, 1.0522e-32, 1.3969e-32, 4.5730e-31,\n        2.4023e-32, 1.3766e-31, 2.0612e-31, 1.4378e-32, 1.2560e-31, 1.1868e-31,\n        4.7636e-32, 6.5361e-32, 4.0790e-32, 6.8090e-31, 2.7658e-31, 4.9247e-32,\n        3.7666e-32, 2.0032e-31, 2.1691e-31, 4.0737e-31, 1.7157e-31, 8.0285e-31,\n        2.2048e-31, 2.9869e-31, 6.9457e-31, 8.5282e-31, 5.1603e-31, 2.7181e-31,\n        3.2067e-31, 1.5703e-31, 3.5956e-32, 3.2210e-32, 1.7887e-30, 2.1014e-33,\n        3.3977e-31, 2.0228e-32, 7.9469e-32, 9.6858e-32, 2.4937e-31, 8.0578e-32,\n        2.1133e-32, 2.3733e-35, 5.4267e-33, 3.1921e-32, 7.1651e-32, 1.8620e-31,\n        1.7708e-32, 4.2068e-31, 1.6117e-30, 1.8194e-31, 5.2056e-31, 6.9364e-32,\n        2.8462e-31, 7.8503e-31, 3.6172e-31, 2.4241e-32, 2.6104e-32, 1.0572e-30,\n        1.8203e-31, 2.0464e-31, 9.4908e-33, 1.3021e-31, 4.5077e-31, 1.4760e-31,\n        7.7344e-34, 2.0358e-31, 5.9918e-32, 5.8322e-32, 3.3106e-31, 3.0603e-32,\n        2.1393e-32, 1.8568e-31, 1.2279e-31, 7.8630e-33, 7.2293e-32, 8.4171e-31,\n        4.9711e-32, 1.7274e-30, 8.7889e-31, 1.2384e-32, 2.8634e-31, 9.0804e-31,\n        8.9851e-31, 3.3965e-32, 4.2621e-32, 7.4448e-32, 5.2262e-32, 2.2482e-31,\n        4.4529e-33, 2.2671e-31, 1.1597e-31, 2.9173e-32, 1.0585e-31, 1.9580e-33,\n        3.6657e-31, 2.2449e-33, 4.5086e-31, 2.5167e-31, 2.7764e-31, 6.3070e-31,\n        5.7491e-31, 2.0839e-32, 1.1747e-30, 1.0670e-31, 2.7191e-31, 5.8718e-32,\n        1.7371e-33, 9.8444e-32, 1.3547e-31, 2.2597e-31, 2.6092e-32, 8.0985e-32,\n        5.3230e-33, 1.2284e-31, 1.8476e-31, 8.0932e-33, 3.6226e-31, 6.8792e-33,\n        2.0772e-31, 7.7433e-32, 3.9628e-31, 1.8137e-31, 1.0331e-32, 8.2072e-32,\n        1.9596e-31, 5.3454e-32, 5.9271e-32, 1.0799e-31, 2.1131e-31, 1.2406e-31,\n        1.6344e-31, 4.7275e-33, 2.1180e-31, 2.2774e-31, 6.0832e-31, 6.1756e-31,\n        1.5357e-32, 1.1305e-31, 1.6125e-32, 1.0779e-31, 2.3072e-33, 2.9735e-32,\n        3.5686e-32, 1.9041e-31, 6.5518e-33, 9.6570e-33, 4.2447e-32, 4.7439e-31,\n        3.3868e-32, 9.7281e-32, 2.5881e-32, 3.2417e-31, 6.6878e-31, 6.9631e-35,\n        6.3654e-31, 4.4801e-31, 6.1304e-31, 2.8852e-33, 1.0090e-31, 6.9146e-31,\n        3.0461e-32, 1.3644e-32, 8.7774e-13, 1.3257e-12, 1.3948e-14, 1.1140e-11,\n        3.0310e-12, 7.6036e-13, 1.9167e-12, 5.4003e-12, 1.0748e-13, 1.3425e-12,\n        8.8795e-12, 2.1071e-14, 6.2435e-13, 2.2625e-12, 1.1229e-11, 2.5434e-13,\n        3.3244e-12, 1.1169e-11, 7.6740e-13, 1.2850e-12, 3.5772e-12, 6.3275e-15,\n        4.7045e-12, 1.4372e-11, 9.0587e-13, 1.5664e-11, 3.3139e-13, 2.5054e-12,\n        2.0597e-12, 9.4364e-12, 2.0146e-12, 2.1854e-14, 3.5533e-13, 2.3128e-14,\n        4.3272e-12, 6.0001e-12, 6.0114e-12, 9.2311e-13, 1.7581e-13, 5.8776e-12,\n        1.1147e-11, 5.5990e-12, 6.6264e-12, 5.1708e-12, 2.2461e-12, 1.4722e-14,\n        4.0249e-12, 1.1462e-12, 5.4942e-12, 3.9154e-12, 2.1914e-11, 4.2620e-12,\n        7.2631e-13, 3.3762e-14, 1.8755e-12, 1.8706e-12, 1.2011e-11, 1.0250e-12,\n        2.8729e-12, 2.9027e-12, 2.9765e-12, 1.1425e-12, 1.0990e-12, 8.0734e-13,\n        3.1215e-13, 2.5147e-14, 6.5295e-12, 2.4350e-12, 3.3112e-12, 4.2946e-12,\n        3.1049e-13, 1.0072e-13, 6.1339e-14, 1.4566e-11, 4.7155e-13, 1.9892e-11,\n        3.8108e-12, 1.1045e-11, 5.0680e-12, 5.1284e-13, 3.2115e-12, 3.7494e-11,\n        1.3118e-12, 8.2886e-12, 3.0112e-13, 1.1266e-11, 4.0708e-12, 2.5800e-13,\n        7.2271e-14, 4.5091e-12, 6.5587e-12, 3.0592e-12, 2.2141e-14, 6.2799e-12,\n        1.6082e-11, 7.9256e-13, 3.3866e-13, 6.5728e-12, 6.1881e-13, 6.2210e-13,\n        8.3281e-13, 3.9140e-12, 1.0479e-13, 5.1375e-13, 1.3881e-12, 1.3860e-11,\n        8.3876e-12, 1.1941e-12, 1.2706e-13, 1.4892e-12, 6.7140e-13, 2.2397e-12,\n        4.1851e-12, 2.3187e-13, 3.2135e-12, 4.4724e-12, 1.1827e-12, 5.5869e-13,\n        7.8950e-12, 1.5476e-13, 5.5090e-13, 3.0284e-11, 5.5404e-12, 1.9130e-11,\n        6.9026e-12, 2.2893e-12, 1.7021e-13, 2.7997e-13, 3.4591e-12, 3.8778e-12,\n        2.1636e-12, 3.0590e-12, 7.6568e-13, 1.2954e-14, 3.3105e-12, 1.2383e-12,\n        1.7781e-12, 8.1996e-12, 9.2527e-13, 1.8089e-13, 1.4636e-15, 6.7807e-12,\n        2.9350e-12, 6.8403e-14, 2.0160e-12, 1.6296e-12, 4.2521e-13, 1.4501e-11,\n        6.1924e-13, 1.1190e-14, 1.9146e-12, 2.1659e-11, 1.1248e-12, 9.1855e-13,\n        8.5893e-12, 2.5763e-13, 9.7622e-14, 1.6231e-12, 1.4121e-15, 2.7524e-12,\n        6.9283e-12, 9.7786e-12, 2.8271e-13, 1.8488e-12, 2.5766e-12, 4.6350e-12,\n        2.1998e-12, 1.0678e-12, 1.7747e-13, 1.1593e-11, 7.1484e-13, 1.4957e-11,\n        9.4621e-12, 7.1391e-13, 2.4101e-12, 6.2597e-13, 3.8636e-14, 7.9300e-13,\n        6.4669e-12, 2.4944e-14, 4.4979e-12, 1.2853e-11, 3.3945e-14, 3.0572e-12,\n        8.4930e-12, 7.9902e-12, 2.1601e-12, 1.2484e-12, 9.7515e-14, 8.2193e-12,\n        1.4984e-13, 2.7828e-12, 1.2613e-11, 1.3055e-11, 2.2725e-12, 3.8277e-12,\n        6.8286e-12, 1.6506e-11, 1.5939e-13, 7.3215e-12, 9.0980e-13, 4.5850e-13,\n        1.6999e-11, 1.1662e-12, 5.3875e-13, 2.0797e-12, 6.7420e-13, 1.2232e-12,\n        1.9007e-12, 6.2310e-13, 3.5893e-12, 2.6925e-12, 8.6406e-12, 6.6996e-13,\n        9.9942e-14, 2.7192e-11, 8.5598e-13, 1.6101e-11, 3.2666e-14, 1.7603e-12,\n        1.3775e-12, 6.1187e-12, 6.9856e-12, 4.1244e-12, 8.9290e-12, 3.1637e-13,\n        5.6616e-12, 7.5024e-13, 3.4891e-11, 2.2587e-13, 8.2396e-13, 3.4299e-12,\n        3.5050e-12, 2.0482e-12, 7.1594e-12, 2.5906e-12, 7.2651e-13, 6.4472e-13,\n        4.3733e-12, 2.6201e-14, 6.3702e-12, 2.4042e-13, 5.9099e-16, 1.1925e-12,\n        6.0651e-12, 8.0081e-14, 1.6027e-12, 1.5621e-13, 2.1099e-13, 1.4163e-12,\n        5.0702e-12, 3.6620e-12, 2.7557e-12, 1.1961e-11, 3.6146e-12, 1.0546e-12],\n       device='cuda:0')"
-      },
-      "54": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([[ 5.6052e-45,  5.6052e-45, -5.6052e-45,  ...,  5.6052e-45,\n         -5.6052e-45, -5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45, -5.6052e-45,  ...,  5.6052e-45,\n         -5.6052e-45, -5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45, -5.6052e-45,  ...,  5.6052e-45,\n         -5.6052e-45, -5.6052e-45],\n        ...,\n        [ 5.6052e-45,  5.6052e-45, -5.6052e-45,  ...,  5.6052e-45,\n         -5.6052e-45, -5.6052e-45],\n        [-5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45, -5.6052e-45,  ...,  5.6052e-45,\n         -5.6052e-45, -5.6052e-45]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[2.4202e-13, 7.9313e-15, 5.5237e-15,  ..., 1.8325e-14, 9.7267e-14,\n         1.2686e-13],\n        [1.2758e-13, 5.8437e-15, 4.1798e-15,  ..., 1.5476e-14, 5.5270e-14,\n         7.2375e-14],\n        [2.3338e-12, 5.2914e-14, 4.6270e-14,  ..., 7.6578e-14, 9.3305e-13,\n         1.1718e-12],\n        ...,\n        [7.3406e-13, 1.8380e-14, 1.3340e-14,  ..., 3.3773e-14, 2.9351e-13,\n         3.8971e-13],\n        [1.8208e-14, 1.3790e-15, 4.2671e-16,  ..., 6.9496e-15, 7.1599e-15,\n         6.6839e-15],\n        [4.1155e-14, 3.3339e-15, 2.5957e-15,  ..., 1.1174e-14, 2.0243e-14,\n         2.1292e-14]], device='cuda:0')"
-      },
-      "55": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([-5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n        -5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45, -5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,\n        -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n        -5.6052e-45,  5.6052e-45, -5.6052e-45, -5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45, -5.6052e-45,  5.6052e-45,\n        -5.6052e-45], device='cuda:0')",
-        "exp_avg_sq": "tensor([4.3318e-12, 2.3206e-12, 4.1229e-11, 2.2206e-11, 1.6645e-11, 2.2722e-14,\n        5.5336e-12, 5.7436e-12, 9.9233e-12, 9.9039e-13, 1.8312e-11, 2.4101e-13,\n        1.4516e-11, 8.5051e-15, 1.4284e-11, 3.4878e-14, 7.6588e-13, 6.7675e-12,\n        6.3327e-12, 1.0849e-11, 1.5537e-11, 2.3745e-12, 1.8168e-11, 8.1935e-12,\n        4.5757e-11, 1.2360e-11, 9.8480e-12, 2.2967e-12, 9.1834e-12, 2.2124e-11,\n        1.1586e-11, 4.0399e-11, 4.4965e-12, 5.2049e-13, 1.8659e-12, 2.0023e-12,\n        5.1360e-12, 1.7012e-11, 1.0540e-11, 2.1692e-12, 5.8748e-12, 1.6471e-11,\n        1.7322e-13, 6.3939e-12, 1.1392e-11, 4.0098e-12, 4.2003e-12, 1.3249e-11,\n        4.9506e-11, 2.4688e-12, 1.8801e-11, 9.5968e-12, 4.6835e-12, 3.9734e-12,\n        1.0325e-13, 3.6310e-11, 6.8560e-13, 1.5249e-11, 3.7122e-12, 3.5800e-11,\n        2.0334e-11, 1.3861e-11, 4.3827e-13, 7.3843e-12, 3.1731e-12, 3.4936e-12,\n        3.8417e-12, 3.6398e-11, 3.6439e-13, 1.7270e-11, 1.3060e-13, 1.3989e-11,\n        2.3152e-13, 1.4337e-11, 8.7260e-14, 3.8302e-12, 3.1931e-11, 1.8267e-12,\n        1.3503e-11, 2.5649e-11, 1.1872e-11, 8.3936e-12, 9.6752e-12, 1.3345e-11,\n        1.2086e-11, 1.7045e-11, 5.8510e-12, 8.2138e-12, 1.0020e-11, 5.4994e-13,\n        2.8696e-12, 3.6047e-12, 6.3750e-11, 1.7385e-12, 8.5605e-12, 8.9480e-15,\n        4.5897e-13, 1.4582e-13, 3.2263e-14, 9.9206e-13, 6.3726e-11, 4.5354e-13,\n        6.0353e-12, 5.2973e-11, 4.1712e-12, 1.1056e-12, 1.7367e-12, 4.4622e-13,\n        1.2199e-11, 2.9011e-11, 1.6446e-13, 8.5281e-12, 3.3974e-12, 3.3232e-12,\n        1.0839e-11, 2.5581e-12, 6.3458e-13, 3.5879e-13, 8.2584e-12, 6.4379e-12,\n        1.2670e-13, 9.0529e-12, 6.4039e-13, 6.4557e-11, 2.6482e-11, 6.6465e-11,\n        2.4763e-11, 1.5572e-12, 1.3769e-12, 3.8532e-12, 2.9918e-11, 4.9015e-13,\n        3.0761e-12, 1.3805e-11, 4.8348e-14, 1.9384e-13, 8.5922e-12, 2.0220e-11,\n        9.8950e-12, 2.3853e-12, 4.4393e-12, 1.1657e-11, 6.0657e-12, 5.6236e-12,\n        6.1005e-12, 1.3501e-11, 5.0354e-12, 7.6882e-12, 5.4425e-12, 3.5276e-11,\n        1.9632e-14, 4.0422e-12, 4.9404e-13, 4.0842e-13, 8.5272e-11, 3.7122e-11,\n        1.8495e-12, 3.4675e-11, 7.8726e-12, 4.8381e-11, 9.7606e-13, 9.9970e-13,\n        6.4212e-12, 2.7273e-14, 4.7131e-11, 7.9059e-12, 1.0405e-11, 1.6647e-12,\n        7.4388e-12, 3.4126e-11, 4.6739e-13, 8.0808e-12, 1.0994e-11, 3.0205e-12,\n        3.3515e-11, 8.2178e-12, 6.5513e-12, 2.3063e-13, 5.8891e-12, 1.7741e-12,\n        1.5647e-11, 7.2519e-12, 1.0949e-12, 2.2346e-12, 2.1628e-13, 4.3274e-12,\n        1.8697e-12, 2.7621e-11, 6.8295e-13, 2.6183e-12, 2.2625e-12, 4.9950e-13,\n        1.6347e-12, 1.7746e-11, 6.4331e-12, 5.6453e-12, 1.7214e-11, 9.3888e-12,\n        1.0104e-11, 4.6986e-12, 5.0256e-12, 2.1142e-11, 1.2914e-11, 3.4860e-13,\n        2.2229e-11, 7.5602e-13, 1.7120e-11, 2.2648e-11, 2.3310e-11, 2.2413e-11,\n        6.4210e-12, 4.8598e-12, 2.0751e-11, 8.3251e-13, 1.8989e-11, 8.9948e-12,\n        1.2110e-14, 1.3090e-11, 3.1224e-14, 9.7709e-12, 5.6322e-12, 1.4065e-10,\n        1.6446e-11, 5.0255e-11, 4.5836e-13, 2.2725e-12, 1.0601e-11, 8.1307e-13,\n        2.9280e-12, 1.6324e-13, 5.0406e-12, 1.2879e-11, 1.3523e-11, 2.0289e-12,\n        4.8075e-11, 6.2966e-12, 2.7813e-12, 3.0357e-13, 2.7335e-12, 3.5414e-14,\n        1.5136e-12, 1.1441e-11, 8.3809e-13, 7.6591e-13, 5.5092e-12, 3.4526e-11,\n        5.7661e-12, 1.7953e-11, 8.7878e-12, 4.2853e-12, 3.6218e-11, 8.4204e-13,\n        2.8428e-11, 1.3099e-11, 2.4713e-13, 7.2117e-13], device='cuda:0')"
-      },
-      "56": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([[-5.6052e-45, -5.6052e-45, -5.6052e-45,  ..., -5.6052e-45,\n         -5.6052e-45, -5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        ...,\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[1.0625e-10, 2.3049e-11, 6.3773e-11,  ..., 2.7615e-11, 2.3642e-12,\n         3.6274e-13],\n        [9.5738e-13, 2.4950e-13, 5.6642e-13,  ..., 2.2765e-13, 2.2551e-14,\n         4.4821e-15],\n        [1.3039e-12, 2.7514e-13, 7.7805e-13,  ..., 3.4471e-13, 2.9168e-14,\n         4.1816e-15],\n        ...,\n        [1.0660e-12, 2.2477e-13, 6.4349e-13,  ..., 2.7396e-13, 2.5260e-14,\n         5.2502e-15],\n        [1.3206e-12, 2.8517e-13, 7.9380e-13,  ..., 3.4357e-13, 2.4486e-14,\n         2.4563e-15],\n        [1.1719e-12, 2.4725e-13, 7.3292e-13,  ..., 3.1371e-13, 2.8304e-14,\n         7.1725e-15]], device='cuda:0')"
-      },
-      "57": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([-5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45],\n       device='cuda:0')",
-        "exp_avg_sq": "tensor([2.7387e-09, 2.4830e-11, 3.3537e-11, 3.4802e-11, 3.9901e-11, 4.5835e-11,\n        3.6441e-11, 2.6986e-11, 3.4086e-11, 3.0291e-11], device='cuda:0')"
-      },
-      "58": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([[-5.6052e-45, -5.6052e-45, -5.6052e-45,  ..., -5.6052e-45,\n         -5.6052e-45, -5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        ...,\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[1.0624e-10, 2.3045e-11, 6.3754e-11,  ..., 2.7615e-11, 2.3588e-12,\n         3.5908e-13],\n        [9.5722e-13, 2.4944e-13, 5.6615e-13,  ..., 2.2765e-13, 2.2472e-14,\n         4.4286e-15],\n        [1.3038e-12, 2.7509e-13, 7.7785e-13,  ..., 3.4471e-13, 2.9108e-14,\n         4.1407e-15],\n        ...,\n        [1.0658e-12, 2.2470e-13, 6.4318e-13,  ..., 2.7396e-13, 2.5171e-14,\n         5.1892e-15],\n        [1.3206e-12, 2.8516e-13, 7.9373e-13,  ..., 3.4357e-13, 2.4464e-14,\n         2.4418e-15],\n        [1.1716e-12, 2.4712e-13, 7.3239e-13,  ..., 3.1371e-13, 2.8146e-14,\n         7.0653e-15]], device='cuda:0')"
-      },
-      "59": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([-5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45],\n       device='cuda:0')",
-        "exp_avg_sq": "tensor([2.7387e-09, 2.4830e-11, 3.3537e-11, 3.4802e-11, 3.9901e-11, 4.5835e-11,\n        3.6441e-11, 2.6986e-11, 3.4086e-11, 3.0290e-11], device='cuda:0')"
-      },
-      "60": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([[-5.6052e-45, -5.6052e-45, -5.6052e-45,  ..., -5.6052e-45,\n         -5.6052e-45, -5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        ...,\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45],\n        [ 5.6052e-45,  5.6052e-45,  5.6052e-45,  ...,  5.6052e-45,\n          5.6052e-45,  5.6052e-45]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[1.0625e-10, 2.3049e-11, 6.3773e-11,  ..., 2.7615e-11, 2.3642e-12,\n         3.6274e-13],\n        [9.5738e-13, 2.4950e-13, 5.6642e-13,  ..., 2.2765e-13, 2.2551e-14,\n         4.4821e-15],\n        [1.3039e-12, 2.7514e-13, 7.7805e-13,  ..., 3.4471e-13, 2.9168e-14,\n         4.1816e-15],\n        ...,\n        [1.0660e-12, 2.2477e-13, 6.4349e-13,  ..., 2.7396e-13, 2.5260e-14,\n         5.2502e-15],\n        [1.3206e-12, 2.8517e-13, 7.9380e-13,  ..., 3.4357e-13, 2.4486e-14,\n         2.4563e-15],\n        [1.1719e-12, 2.4725e-13, 7.3292e-13,  ..., 3.1371e-13, 2.8304e-14,\n         7.1725e-15]], device='cuda:0')"
-      },
-      "61": {
-        "step": "tensor(11268.)",
-        "exp_avg": "tensor([-5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,\n         5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45,  5.6052e-45],\n       device='cuda:0')",
-        "exp_avg_sq": "tensor([2.7387e-09, 2.4830e-11, 3.3537e-11, 3.4802e-11, 3.9901e-11, 4.5835e-11,\n        3.6441e-11, 2.6986e-11, 3.4086e-11, 3.0291e-11], device='cuda:0')"
-      },
-      "8": {
-        "step": "tensor(10016.)",
-        "exp_avg": "tensor([[ 5.2391e-07, -9.5890e-08, -1.6794e-07,  ...,  1.0383e-06,\n          0.0000e+00,  1.6587e-07],\n        [-2.9076e-07, -5.4219e-09, -8.0596e-06,  ..., -3.0523e-07,\n          0.0000e+00, -1.1051e-07],\n        [ 3.0308e-07,  1.0616e-08,  2.8124e-07,  ..., -8.6895e-08,\n          0.0000e+00, -2.3080e-07],\n        ...,\n        [-7.1909e-08,  1.7884e-08, -1.0837e-07,  ...,  4.0920e-06,\n          0.0000e+00,  2.0505e-07],\n        [-7.4948e-07,  1.5425e-07,  8.0185e-07,  ...,  2.5868e-06,\n          0.0000e+00,  1.0694e-06],\n        [-2.2005e-07,  6.5157e-08,  1.4630e-06,  ...,  1.3091e-06,\n          0.0000e+00,  4.5262e-06]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[3.7811e-12, 3.4363e-13, 2.4449e-11,  ..., 1.9803e-11, 0.0000e+00,\n         5.4632e-12],\n        [1.7406e-11, 1.2091e-13, 2.4161e-11,  ..., 2.5411e-12, 0.0000e+00,\n         1.0146e-11],\n        [3.4680e-12, 7.8053e-13, 2.0143e-11,  ..., 3.2588e-12, 0.0000e+00,\n         2.1691e-12],\n        ...,\n        [2.2858e-12, 7.1243e-13, 6.1150e-12,  ..., 3.3772e-11, 0.0000e+00,\n         8.1645e-12],\n        [9.0069e-12, 5.4013e-13, 7.2687e-11,  ..., 5.8154e-11, 0.0000e+00,\n         4.7426e-11],\n        [1.5617e-11, 3.0288e-12, 3.6350e-11,  ..., 3.1824e-11, 0.0000e+00,\n         2.1300e-11]], device='cuda:0')"
-      },
-      "9": {
-        "step": "tensor(10016.)",
-        "exp_avg": "tensor([ 3.6428e-06, -1.5594e-05,  4.7698e-06,  ...,  5.7057e-06,\n         5.1658e-06,  3.8920e-05], device='cuda:0')",
-        "exp_avg_sq": "tensor([1.4399e-09, 1.6370e-09, 1.0969e-09,  ..., 1.1355e-09, 1.5990e-09,\n        1.5532e-09], device='cuda:0')"
-      },
-      "10": {
-        "step": "tensor(10016.)",
-        "exp_avg": "tensor([[ 3.9524e-08, -9.2433e-08,  1.0221e-07,  ..., -4.4198e-07,\n          7.6784e-07,  2.0899e-07],\n        [ 3.2126e-07,  3.1007e-07,  2.6399e-07,  ...,  3.7226e-07,\n         -5.8755e-07,  5.4598e-07],\n        [-2.2047e-07,  3.0856e-07,  1.3871e-07,  ..., -1.7447e-07,\n         -9.6529e-07, -2.3211e-07],\n        ...,\n        [-1.7317e-07, -1.6380e-07, -9.4509e-08,  ..., -3.1332e-08,\n          3.2863e-08,  1.8801e-07],\n        [-2.4867e-07, -1.4015e-07, -1.4686e-07,  ...,  6.6397e-07,\n         -3.8798e-07, -2.3568e-07],\n        [ 2.6323e-07, -3.6754e-07, -1.6236e-07,  ...,  4.9175e-07,\n         -6.2566e-07, -2.3958e-07]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[6.7076e-13, 7.1408e-13, 1.1768e-12,  ..., 1.8106e-12, 1.0345e-12,\n         6.2712e-13],\n        [1.3839e-12, 1.6290e-12, 1.3338e-12,  ..., 2.2275e-12, 1.6904e-12,\n         1.4582e-12],\n        [1.5684e-12, 2.2788e-12, 1.8632e-12,  ..., 1.9984e-12, 2.4658e-12,\n         1.2703e-12],\n        ...,\n        [1.9778e-12, 1.1253e-12, 1.7830e-12,  ..., 2.8008e-12, 2.9287e-12,\n         1.3667e-12],\n        [1.6394e-12, 1.3856e-12, 2.5231e-12,  ..., 2.5221e-12, 3.2398e-12,\n         1.3361e-12],\n        [2.2460e-12, 1.5797e-12, 1.4107e-12,  ..., 2.1398e-12, 1.8176e-12,\n         2.2789e-12]], device='cuda:0')"
-      },
-      "11": {
-        "step": "tensor(8764.)",
-        "exp_avg": "tensor([[-4.5763e-07, -5.1912e-09, -6.4911e-08,  ..., -4.0983e-08,\n          0.0000e+00, -3.3231e-07],\n        [-7.6240e-09, -1.2870e-07,  7.0829e-08,  ...,  4.1722e-07,\n          0.0000e+00,  3.3970e-06],\n        [ 1.5098e-07,  8.2674e-10,  4.2594e-07,  ...,  4.2718e-07,\n          0.0000e+00,  3.4151e-07],\n        ...,\n        [ 3.4169e-08,  6.1919e-09, -2.5748e-07,  ...,  1.3139e-07,\n          0.0000e+00, -2.2114e-07],\n        [-2.9416e-07,  4.2277e-07,  9.9381e-07,  ...,  1.4964e-06,\n          0.0000e+00, -3.6192e-07],\n        [-2.7604e-07,  1.4446e-10, -8.1898e-08,  ...,  3.2305e-08,\n          0.0000e+00,  3.6308e-08]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[1.3432e-11, 8.3001e-12, 4.1771e-12,  ..., 4.1365e-12, 0.0000e+00,\n         7.5649e-12],\n        [2.2236e-12, 8.0298e-12, 7.6686e-13,  ..., 2.2507e-12, 0.0000e+00,\n         5.9823e-11],\n        [3.2761e-12, 6.3425e-13, 1.8211e-11,  ..., 7.7775e-11, 0.0000e+00,\n         5.9067e-12],\n        ...,\n        [3.2106e-13, 3.9059e-13, 7.5758e-12,  ..., 2.1005e-12, 0.0000e+00,\n         1.0711e-11],\n        [7.7567e-12, 7.0507e-12, 2.6024e-11,  ..., 7.7393e-11, 0.0000e+00,\n         5.7496e-11],\n        [1.3668e-12, 5.0223e-15, 1.4851e-12,  ..., 7.4668e-12, 0.0000e+00,\n         2.2968e-12]], device='cuda:0')"
-      },
-      "12": {
-        "step": "tensor(8764.)",
-        "exp_avg": "tensor([-3.1734e-06,  4.8233e-06,  4.5898e-06,  ..., -3.0027e-05,\n         3.4156e-07, -4.7405e-06], device='cuda:0')",
-        "exp_avg_sq": "tensor([1.0698e-09, 7.4755e-10, 9.0855e-10,  ..., 1.2911e-09, 1.4461e-09,\n        1.0259e-09], device='cuda:0')"
-      },
-      "13": {
-        "step": "tensor(8764.)",
-        "exp_avg": "tensor([[ 2.2182e-07, -2.0739e-09,  7.5835e-07,  ..., -3.3445e-07,\n          2.1949e-08,  3.0941e-07],\n        [ 1.6808e-07,  1.8245e-07,  3.3214e-07,  ...,  2.7568e-07,\n          2.7083e-07,  4.0770e-07],\n        [-2.0915e-07,  6.5567e-08, -1.1137e-06,  ...,  6.5655e-07,\n          2.2354e-07, -9.9601e-08],\n        ...,\n        [-1.6749e-07,  2.1184e-07, -2.2033e-07,  ..., -6.8793e-09,\n          2.4235e-07, -1.6973e-07],\n        [ 1.4642e-07,  2.4461e-08, -4.6586e-07,  ..., -8.2878e-10,\n         -3.2903e-09,  3.3803e-09],\n        [-1.5389e-07,  9.5534e-08, -1.2836e-07,  ..., -8.0069e-08,\n          2.5866e-07, -2.3752e-07]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[4.5088e-13, 3.1527e-13, 6.2491e-13,  ..., 1.0108e-12, 5.9876e-13,\n         1.5150e-12],\n        [8.1732e-13, 4.4879e-13, 1.9033e-12,  ..., 2.6162e-12, 1.5164e-12,\n         1.9223e-12],\n        [8.9293e-13, 7.2005e-13, 2.2119e-12,  ..., 2.6867e-12, 1.7467e-12,\n         7.9776e-13],\n        ...,\n        [1.0520e-12, 1.0296e-12, 2.2339e-12,  ..., 1.5686e-12, 1.9143e-12,\n         2.7351e-12],\n        [7.8992e-13, 9.9624e-13, 7.6716e-13,  ..., 1.3802e-12, 1.2542e-12,\n         3.2748e-12],\n        [9.0855e-13, 6.6180e-13, 1.1952e-12,  ..., 3.4612e-12, 1.5513e-12,\n         1.2847e-12]], device='cuda:0')"
-      },
-      "14": {
-        "step": "tensor(7512.)",
-        "exp_avg": "tensor([[-9.9144e-08, -3.9194e-08,  6.6145e-06,  ...,  3.6352e-05,\n          0.0000e+00,  4.3041e-08],\n        [ 8.1803e-08,  4.6771e-10,  9.3593e-06,  ..., -1.5705e-07,\n          0.0000e+00,  7.0697e-07],\n        [ 1.4673e-07,  3.3228e-08,  3.6580e-07,  ..., -1.3127e-08,\n          0.0000e+00,  4.7572e-07],\n        ...,\n        [ 5.0273e-07,  1.8915e-09, -3.1544e-08,  ..., -7.1909e-09,\n          0.0000e+00,  1.7110e-07],\n        [-5.6052e-09, -9.2777e-10,  2.9593e-07,  ...,  4.5681e-06,\n          0.0000e+00,  3.3314e-06],\n        [ 4.4287e-08,  5.6212e-09,  1.1217e-05,  ...,  8.2747e-05,\n          0.0000e+00,  6.8293e-08]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[5.2447e-13, 2.2746e-13, 9.2144e-11,  ..., 3.8680e-10, 0.0000e+00,\n         6.9366e-12],\n        [9.2418e-12, 2.7513e-13, 7.3857e-11,  ..., 1.5449e-11, 0.0000e+00,\n         2.8060e-11],\n        [2.1221e-11, 4.9112e-13, 3.5593e-12,  ..., 7.8030e-13, 0.0000e+00,\n         1.2318e-12],\n        ...,\n        [2.3565e-11, 1.7063e-13, 7.0745e-12,  ..., 3.5663e-12, 0.0000e+00,\n         6.4724e-12],\n        [2.8506e-12, 5.2280e-13, 5.3315e-12,  ..., 1.2502e-10, 0.0000e+00,\n         1.3957e-10],\n        [3.5639e-12, 1.4660e-13, 1.5281e-10,  ..., 1.6086e-09, 0.0000e+00,\n         1.5458e-11]], device='cuda:0')"
-      },
-      "15": {
-        "step": "tensor(7512.)",
-        "exp_avg": "tensor([ 3.3027e-05,  2.2930e-05,  3.7146e-06,  ...,  1.8131e-05,\n        -3.2139e-06,  6.8807e-05], device='cuda:0')",
-        "exp_avg_sq": "tensor([2.3151e-09, 2.0855e-09, 1.9548e-09,  ..., 2.8112e-09, 3.1657e-09,\n        3.5868e-09], device='cuda:0')"
-      },
-      "16": {
-        "step": "tensor(7512.)",
-        "exp_avg": "tensor([[-2.1174e-09, -8.2970e-07, -1.1059e-07,  ...,  5.8443e-08,\n         -1.8422e-07, -4.0374e-07],\n        [ 4.0283e-07, -2.6375e-07,  1.5211e-07,  ..., -1.8824e-07,\n         -1.4319e-06,  1.5411e-06],\n        [-1.3496e-07,  5.9090e-07,  5.6911e-08,  ...,  3.6908e-07,\n          3.0063e-07, -6.0522e-07],\n        ...,\n        [ 8.0747e-08, -2.1092e-07, -1.2430e-07,  ..., -2.7408e-07,\n          2.8949e-07,  2.6415e-07],\n        [ 6.3674e-08, -6.9401e-07, -2.4110e-07,  ..., -3.4040e-07,\n         -2.9516e-06,  6.0260e-08],\n        [-3.6565e-07, -1.6630e-08,  6.8491e-08,  ..., -1.8997e-07,\n         -1.0694e-06, -1.1596e-06]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[6.8353e-13, 7.9268e-13, 6.1169e-13,  ..., 4.0575e-13, 9.0677e-13,\n         4.7225e-13],\n        [9.2926e-13, 1.0657e-12, 1.2535e-12,  ..., 2.1452e-12, 2.0690e-12,\n         1.1186e-12],\n        [7.6799e-13, 9.2466e-13, 7.5990e-13,  ..., 7.1637e-13, 2.1274e-12,\n         7.4540e-13],\n        ...,\n        [1.6328e-12, 1.1935e-12, 6.1102e-13,  ..., 1.2074e-12, 7.5510e-13,\n         7.7320e-13],\n        [1.0021e-12, 1.3796e-12, 1.2161e-12,  ..., 1.5376e-12, 6.0830e-12,\n         6.1298e-13],\n        [1.0035e-12, 2.1793e-12, 1.0239e-12,  ..., 7.3054e-13, 1.6506e-12,\n         1.3283e-12]], device='cuda:0')"
-      },
-      "17": {
-        "step": "tensor(6260.)",
-        "exp_avg": "tensor([[ 9.5503e-07, -1.3270e-09, -2.9824e-07,  ...,  5.8676e-08,\n          0.0000e+00, -1.4706e-07],\n        [-8.1059e-08,  1.4618e-07,  1.1680e-08,  ...,  2.0458e-06,\n          0.0000e+00,  2.1162e-07],\n        [ 1.4763e-07,  1.2673e-09,  6.5117e-08,  ..., -6.0637e-08,\n          0.0000e+00, -2.7041e-07],\n        ...,\n        [-4.1496e-08, -2.2324e-07,  1.0280e-07,  ...,  8.3597e-07,\n          0.0000e+00, -9.0796e-08],\n        [-1.2783e-07, -6.0111e-08,  7.7144e-08,  ..., -5.2068e-06,\n          0.0000e+00,  1.2393e-07],\n        [ 1.1435e-07, -8.7357e-09,  1.9023e-06,  ...,  2.4482e-07,\n          0.0000e+00,  8.5112e-08]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[6.9995e-12, 6.4075e-14, 6.3755e-12,  ..., 2.2819e-12, 0.0000e+00,\n         2.2964e-12],\n        [1.0422e-12, 7.2287e-12, 1.5739e-11,  ..., 4.0577e-10, 0.0000e+00,\n         3.0019e-10],\n        [5.9075e-11, 5.2013e-13, 2.8558e-11,  ..., 1.0987e-12, 0.0000e+00,\n         3.4598e-12],\n        ...,\n        [5.3187e-12, 6.8257e-12, 6.3911e-11,  ..., 1.2294e-11, 0.0000e+00,\n         1.1448e-11],\n        [1.1078e-11, 1.3477e-11, 2.5968e-11,  ..., 6.5423e-11, 0.0000e+00,\n         1.0950e-11],\n        [1.8585e-12, 1.4802e-12, 7.0148e-11,  ..., 9.6244e-12, 0.0000e+00,\n         2.7618e-12]], device='cuda:0')"
-      },
-      "18": {
-        "step": "tensor(6260.)",
-        "exp_avg": "tensor([-7.1890e-07, -5.6708e-05, -8.4577e-06,  ..., -1.9044e-05,\n        -2.6201e-05,  1.1328e-05], device='cuda:0')",
-        "exp_avg_sq": "tensor([2.2720e-09, 3.6701e-09, 2.6689e-09,  ..., 1.4793e-09, 2.5373e-09,\n        2.0937e-09], device='cuda:0')"
-      },
-      "19": {
-        "step": "tensor(6260.)",
-        "exp_avg": "tensor([[-2.6417e-07,  4.7218e-08,  1.9952e-08,  ...,  1.2112e-07,\n         -4.0797e-08, -1.0241e-07],\n        [-1.0129e-07,  2.9652e-07, -7.9153e-09,  ..., -7.3317e-08,\n          2.1145e-07, -4.3682e-08],\n        [-9.1283e-07, -1.3733e-07,  3.2584e-07,  ..., -3.1132e-08,\n          1.0390e-07,  4.0456e-08],\n        ...,\n        [-4.9108e-08,  3.5919e-08, -1.1573e-07,  ...,  5.2719e-09,\n         -7.5300e-08,  3.9253e-10],\n        [ 1.8438e-07,  3.0469e-08, -1.7891e-07,  ..., -1.2010e-08,\n         -4.4716e-07, -1.3029e-07],\n        [-2.0279e-08, -8.9886e-08,  4.3602e-08,  ...,  1.6823e-07,\n          2.5753e-07, -6.5618e-08]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[2.9577e-13, 1.0110e-12, 3.0728e-13,  ..., 3.3254e-13, 5.0185e-13,\n         5.0305e-13],\n        [4.8619e-13, 2.3776e-12, 8.1337e-13,  ..., 7.3788e-13, 1.3377e-12,\n         7.5397e-13],\n        [9.2196e-13, 1.8124e-12, 1.3355e-12,  ..., 8.1659e-13, 1.4627e-12,\n         6.2089e-13],\n        ...,\n        [4.7633e-13, 8.1497e-13, 6.6300e-13,  ..., 7.5343e-13, 1.2362e-12,\n         7.8757e-13],\n        [2.6809e-13, 2.0698e-12, 1.0591e-12,  ..., 6.5731e-13, 1.5546e-12,\n         9.0076e-13],\n        [3.8023e-13, 6.2662e-13, 7.4862e-13,  ..., 8.9159e-13, 1.3479e-12,\n         7.8253e-13]], device='cuda:0')"
-      },
-      "20": {
-        "step": "tensor(5008.)",
-        "exp_avg": "tensor([[-6.1845e-07,  1.4413e-07, -1.9318e-08,  ..., -1.1891e-05,\n          0.0000e+00,  6.4408e-09],\n        [-5.6226e-08,  5.8805e-07, -2.1868e-07,  ..., -2.5101e-07,\n          0.0000e+00, -5.6194e-06],\n        [-3.4456e-09, -1.6632e-08, -1.7002e-06,  ..., -5.1660e-07,\n          0.0000e+00, -5.5233e-08],\n        ...,\n        [ 9.0172e-08,  6.4309e-08,  1.4471e-07,  ...,  2.6056e-07,\n          0.0000e+00, -4.0211e-06],\n        [-4.8459e-07,  1.5604e-08,  1.1054e-07,  ..., -1.5318e-07,\n          0.0000e+00, -1.9394e-07],\n        [ 2.0508e-07, -8.9246e-09, -2.7086e-07,  ..., -5.4503e-08,\n          0.0000e+00,  2.0587e-07]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[2.9100e-11, 5.6381e-12, 7.9570e-12,  ..., 1.5905e-10, 0.0000e+00,\n         3.4304e-12],\n        [2.2399e-11, 5.8809e-11, 2.5600e-11,  ..., 2.7774e-11, 0.0000e+00,\n         7.5856e-11],\n        [1.1651e-12, 5.0924e-13, 6.7110e-11,  ..., 1.0249e-11, 0.0000e+00,\n         4.4959e-11],\n        ...,\n        [1.4415e-12, 3.9656e-13, 9.8264e-11,  ..., 1.1758e-10, 0.0000e+00,\n         1.2689e-10],\n        [4.0739e-11, 2.2677e-12, 4.7581e-12,  ..., 6.9080e-12, 0.0000e+00,\n         2.6612e-11],\n        [6.0083e-11, 2.7452e-12, 2.6890e-11,  ..., 1.6219e-11, 0.0000e+00,\n         4.1770e-11]], device='cuda:0')"
-      },
-      "21": {
-        "step": "tensor(5008.)",
-        "exp_avg": "tensor([-3.7956e-05, -5.4031e-06,  7.0594e-06,  ...,  2.7980e-05,\n        -2.8727e-05,  1.1770e-05], device='cuda:0')",
-        "exp_avg_sq": "tensor([5.6596e-09, 4.7770e-09, 3.7082e-09,  ..., 5.6955e-09, 6.1637e-09,\n        4.7482e-09], device='cuda:0')"
-      },
-      "22": {
-        "step": "tensor(5008.)",
-        "exp_avg": "tensor([[-6.8812e-08,  3.2808e-07, -1.0458e-07,  ..., -2.6516e-07,\n          1.2057e-07, -1.0668e-07],\n        [ 3.3064e-07,  3.1144e-07, -1.5562e-07,  ...,  1.8608e-07,\n         -1.5108e-07, -1.7357e-08],\n        [ 6.3645e-07, -5.7058e-08, -1.4178e-07,  ..., -5.1062e-07,\n         -1.1225e-07, -4.3968e-08],\n        ...,\n        [-1.6826e-07, -3.9908e-07,  6.1119e-07,  ...,  2.9228e-07,\n          9.4299e-09,  7.5585e-08],\n        [-5.2122e-08, -9.3836e-08,  9.5097e-08,  ..., -4.7426e-07,\n         -3.3345e-07, -2.4999e-08],\n        [ 1.1571e-06, -4.2969e-07, -3.5850e-08,  ...,  4.9074e-07,\n         -1.7465e-07, -5.1722e-08]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[9.1614e-13, 1.0121e-12, 4.9026e-13,  ..., 2.4729e-12, 3.4346e-13,\n         6.9624e-13],\n        [7.2384e-13, 9.3973e-13, 8.1467e-13,  ..., 1.8977e-12, 7.2797e-13,\n         8.4920e-13],\n        [1.3063e-12, 1.5015e-12, 1.0908e-12,  ..., 3.6295e-12, 3.9936e-13,\n         1.2586e-12],\n        ...,\n        [1.1700e-12, 1.3447e-12, 1.3369e-12,  ..., 1.5498e-12, 6.1317e-13,\n         2.0993e-12],\n        [6.6169e-13, 1.5391e-12, 9.4662e-13,  ..., 1.9525e-12, 1.1179e-12,\n         6.5295e-13],\n        [2.1305e-12, 1.8973e-12, 1.0005e-12,  ..., 2.1663e-12, 8.8708e-13,\n         2.3615e-12]], device='cuda:0')"
-      },
-      "23": {
-        "step": "tensor(3756.)",
-        "exp_avg": "tensor([[ 8.6625e-08, -1.8288e-07,  8.3755e-07,  ...,  3.9635e-06,\n          0.0000e+00,  2.1120e-08],\n        [ 4.0071e-07,  1.8793e-08, -4.0375e-07,  ..., -5.6561e-07,\n          0.0000e+00, -7.3566e-08],\n        [-1.4025e-06,  3.7002e-07,  9.7900e-07,  ...,  1.2052e-06,\n          0.0000e+00, -4.3734e-07],\n        ...,\n        [-3.7614e-06,  5.9165e-08, -1.9186e-06,  ...,  8.2550e-07,\n          0.0000e+00, -3.2771e-07],\n        [-1.5528e-07,  1.0160e-06, -6.5283e-07,  ..., -2.0301e-07,\n          0.0000e+00, -1.9045e-06],\n        [ 1.6437e-07,  5.3580e-09, -2.5795e-06,  ..., -3.8795e-07,\n          0.0000e+00,  3.0473e-07]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[3.8523e-11, 2.1524e-11, 2.5717e-11,  ..., 1.3093e-10, 0.0000e+00,\n         6.7806e-11],\n        [1.7258e-11, 1.5388e-11, 2.2932e-11,  ..., 5.8121e-11, 0.0000e+00,\n         2.5037e-11],\n        [3.2654e-11, 4.8526e-11, 1.0993e-10,  ..., 2.8691e-11, 0.0000e+00,\n         1.3936e-11],\n        ...,\n        [1.8172e-10, 1.3802e-11, 2.6416e-11,  ..., 3.6502e-11, 0.0000e+00,\n         3.8433e-11],\n        [2.1074e-11, 1.9554e-10, 3.4346e-11,  ..., 1.7332e-11, 0.0000e+00,\n         9.0037e-10],\n        [7.1665e-11, 1.8041e-12, 1.1006e-10,  ..., 6.6133e-11, 0.0000e+00,\n         2.7220e-11]], device='cuda:0')"
-      },
-      "24": {
-        "step": "tensor(3756.)",
-        "exp_avg": "tensor([ 4.2499e-05,  1.5117e-06, -1.1170e-05,  ...,  2.8940e-05,\n        -1.3405e-05, -2.5607e-06], device='cuda:0')",
-        "exp_avg_sq": "tensor([1.1779e-08, 1.0526e-08, 9.5232e-09,  ..., 8.4988e-09, 1.1161e-08,\n        1.1279e-08], device='cuda:0')"
-      },
-      "25": {
-        "step": "tensor(3756.)",
-        "exp_avg": "tensor([[ 3.6704e-07, -6.6624e-08,  4.4290e-08,  ..., -3.3636e-07,\n          1.0906e-07, -3.6141e-08],\n        [ 3.8203e-07,  7.1240e-08,  4.9511e-07,  ...,  4.2769e-07,\n         -2.1302e-07, -1.6420e-07],\n        [ 4.0396e-07,  3.3011e-07, -3.1809e-08,  ...,  2.4831e-07,\n         -8.7148e-08, -3.5561e-07],\n        ...,\n        [-4.2670e-07, -3.5578e-07,  9.9700e-07,  ..., -2.2710e-07,\n         -2.2648e-07, -1.0786e-07],\n        [ 5.7559e-09,  3.5136e-07, -1.8028e-07,  ..., -1.0323e-07,\n          3.0361e-08,  1.0445e-07],\n        [ 2.6267e-07, -1.4132e-07, -2.6558e-09,  ..., -1.1966e-06,\n          5.1226e-07,  1.6046e-07]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[1.4248e-12, 7.3370e-13, 9.7379e-13,  ..., 2.6822e-12, 2.3264e-12,\n         1.3054e-12],\n        [2.6117e-12, 1.6776e-12, 2.0095e-12,  ..., 3.0942e-12, 3.7556e-12,\n         4.9571e-12],\n        [3.1881e-12, 2.2160e-12, 2.2749e-12,  ..., 2.7148e-12, 4.6848e-12,\n         5.0733e-12],\n        ...,\n        [2.8834e-12, 2.3681e-12, 2.6422e-12,  ..., 4.3040e-12, 3.3138e-12,\n         4.8472e-12],\n        [3.8981e-12, 1.3963e-12, 2.5805e-12,  ..., 5.9204e-12, 2.8707e-12,\n         4.9756e-12],\n        [2.6075e-12, 1.4737e-12, 2.8133e-12,  ..., 3.1698e-12, 3.8211e-12,\n         2.9316e-12]], device='cuda:0')"
-      },
-      "26": {
-        "step": "tensor(2504.)",
-        "exp_avg": "tensor([[ 2.6818e-07, -5.6154e-07,  1.7891e-06,  ...,  2.9804e-06,\n          0.0000e+00, -9.3856e-07],\n        [ 1.0162e-06,  8.9720e-08,  1.3353e-06,  ...,  6.9238e-07,\n          0.0000e+00,  2.1799e-07],\n        [ 8.5368e-07, -2.2396e-06, -9.7303e-07,  ..., -7.0409e-06,\n          0.0000e+00, -7.1518e-07],\n        ...,\n        [-1.0315e-07,  4.5212e-08,  1.8161e-05,  ...,  1.7887e-04,\n          0.0000e+00, -1.0776e-06],\n        [-4.4393e-07,  1.0571e-06, -3.6268e-07,  ..., -7.2545e-06,\n          0.0000e+00, -2.9153e-06],\n        [-1.3460e-07,  1.5739e-06,  1.3340e-06,  ..., -4.3566e-06,\n          0.0000e+00,  3.0733e-06]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[4.7025e-11, 6.6304e-11, 5.2935e-11,  ..., 8.1728e-11, 0.0000e+00,\n         4.4678e-11],\n        [6.6985e-11, 7.9299e-11, 3.3033e-10,  ..., 3.8502e-11, 0.0000e+00,\n         2.0548e-10],\n        [1.0850e-10, 3.2719e-10, 4.2286e-11,  ..., 2.8572e-10, 0.0000e+00,\n         1.9407e-10],\n        ...,\n        [8.3318e-11, 7.8421e-12, 1.5768e-09,  ..., 6.0843e-09, 0.0000e+00,\n         6.0611e-10],\n        [1.3663e-10, 1.7230e-10, 4.3073e-11,  ..., 2.6909e-10, 0.0000e+00,\n         1.2199e-10],\n        [6.6905e-11, 3.0412e-11, 3.7693e-10,  ..., 2.2863e-10, 0.0000e+00,\n         4.6511e-10]], device='cuda:0')"
-      },
-      "27": {
-        "step": "tensor(2504.)",
-        "exp_avg": "tensor([ 2.4152e-05,  4.0226e-05, -2.2483e-04,  ...,  1.2204e-04,\n        -1.1629e-05, -3.6174e-05], device='cuda:0')",
-        "exp_avg_sq": "tensor([1.9226e-08, 1.8703e-08, 5.1420e-08,  ..., 3.1267e-08, 1.6952e-08,\n        2.9708e-08], device='cuda:0')"
-      },
-      "28": {
-        "step": "tensor(2504.)",
-        "exp_avg": "tensor([[ 7.0014e-08, -2.3550e-07, -1.8928e-07,  ...,  9.9566e-07,\n          3.5126e-08,  8.7351e-07],\n        [ 3.3054e-07, -1.3881e-07,  6.6509e-07,  ...,  2.5783e-06,\n         -2.5351e-07,  2.2954e-07],\n        [-1.7870e-06, -8.3651e-08, -1.6035e-06,  ...,  5.2402e-08,\n          4.0490e-08,  7.3153e-07],\n        ...,\n        [-1.4263e-06,  6.4761e-07, -3.9834e-07,  ...,  4.7154e-07,\n          4.9860e-07,  4.2942e-07],\n        [-1.4211e-07, -3.5462e-07, -1.1245e-06,  ...,  8.8498e-07,\n         -4.1586e-08, -1.3816e-06],\n        [ 6.1833e-07,  2.5424e-07, -6.6536e-07,  ..., -9.7480e-07,\n         -6.8482e-07,  4.6943e-07]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[3.9734e-12, 3.6468e-12, 1.1986e-11,  ..., 7.9659e-12, 4.1398e-12,\n         6.7981e-12],\n        [8.2350e-12, 5.4648e-12, 8.4214e-12,  ..., 1.3570e-11, 9.0190e-12,\n         1.0688e-11],\n        [8.6306e-12, 1.0488e-11, 1.5727e-11,  ..., 1.3385e-11, 8.3640e-12,\n         1.1622e-11],\n        ...,\n        [1.4206e-11, 7.6523e-12, 1.6370e-11,  ..., 1.7749e-11, 1.4635e-11,\n         1.3071e-11],\n        [8.9737e-12, 6.8857e-12, 9.2066e-12,  ..., 1.3603e-11, 9.1160e-12,\n         1.5713e-11],\n        [8.0421e-12, 7.1649e-12, 9.7956e-12,  ..., 1.0800e-11, 1.1423e-11,\n         1.1974e-11]], device='cuda:0')"
-      },
-      "29": {
         "step": "tensor(1252.)",
-        "exp_avg": "tensor([[ 4.6549e-06,  1.0463e-07,  1.1726e-05,  ...,  3.6087e-06,\n          0.0000e+00, -1.1304e-05],\n        [-9.7613e-06,  2.8244e-06,  1.8208e-06,  ..., -1.0593e-06,\n          0.0000e+00,  2.3663e-06],\n        [ 3.5186e-06, -2.2636e-06, -6.4493e-06,  ...,  1.2065e-05,\n          0.0000e+00, -5.4349e-06],\n        ...,\n        [ 2.2149e-06, -7.3551e-07, -1.5728e-05,  ...,  1.5083e-05,\n          0.0000e+00, -1.8443e-06],\n        [ 4.7179e-06, -1.4825e-06, -4.6107e-07,  ..., -3.2699e-06,\n          0.0000e+00,  2.1100e-06],\n        [-3.4079e-06, -8.8482e-07,  7.7006e-06,  ...,  4.0681e-06,\n          0.0000e+00, -1.1189e-06]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[3.1810e-10, 1.4450e-10, 1.0865e-09,  ..., 6.7395e-10, 0.0000e+00,\n         8.9202e-10],\n        [5.3529e-10, 9.5948e-11, 3.6140e-10,  ..., 2.2561e-10, 0.0000e+00,\n         2.8036e-10],\n        [2.4466e-10, 1.3245e-10, 4.2474e-10,  ..., 7.6973e-10, 0.0000e+00,\n         1.4730e-09],\n        ...,\n        [1.7838e-10, 1.8297e-10, 1.4252e-09,  ..., 7.3969e-10, 0.0000e+00,\n         6.2450e-10],\n        [3.8208e-10, 2.1327e-10, 1.6977e-10,  ..., 3.3027e-10, 0.0000e+00,\n         3.0958e-10],\n        [6.4949e-10, 1.5909e-10, 1.8713e-09,  ..., 4.3686e-10, 0.0000e+00,\n         9.1388e-10]], device='cuda:0')"
+        "exp_avg": "tensor([[ 4.3313e-06, -8.5848e-05, -2.3694e-05,  ...,  1.9346e-05,\n          1.4683e-05,  8.5242e-05],\n        [-1.2722e-05,  7.6283e-05,  7.3687e-05,  ..., -1.7570e-05,\n          1.7892e-05,  2.9558e-05],\n        [-7.5209e-05, -3.7764e-05,  8.0535e-06,  ...,  2.3776e-05,\n          4.4740e-06, -1.3045e-04],\n        ...,\n        [-3.9314e-06,  1.1009e-04,  2.6299e-05,  ..., -1.3002e-05,\n          1.9424e-05,  7.0476e-05],\n        [-2.0331e-05, -9.2566e-05,  2.2730e-06,  ...,  2.8280e-05,\n         -1.0114e-04,  1.5327e-05],\n        [-3.3568e-05,  1.8985e-05,  1.4551e-05,  ...,  3.5216e-05,\n         -2.1837e-06,  2.8932e-05]], device='cuda:0')",
+        "exp_avg_sq": "tensor([[3.7971e-08, 6.0170e-08, 6.3390e-08,  ..., 2.8713e-08, 3.3194e-08,\n         4.0259e-08],\n        [5.5826e-08, 1.1562e-07, 9.6918e-08,  ..., 4.6189e-08, 4.7870e-08,\n         7.5119e-08],\n        [6.0282e-08, 1.2140e-07, 9.2586e-08,  ..., 6.3903e-08, 3.8224e-08,\n         9.1918e-08],\n        ...,\n        [6.0442e-08, 9.4329e-08, 9.5867e-08,  ..., 5.0398e-08, 5.7388e-08,\n         8.8877e-08],\n        [5.9988e-08, 9.9616e-08, 1.0130e-07,  ..., 4.4964e-08, 7.0245e-08,\n         8.3955e-08],\n        [6.7426e-08, 9.2967e-08, 1.2939e-07,  ..., 6.5939e-08, 5.5437e-08,\n         8.4152e-08]], device='cuda:0')"
       },
-      "30": {
+      "5": {
         "step": "tensor(1252.)",
-        "exp_avg": "tensor([ 1.0712e-04, -1.0426e-05, -2.6064e-05,  ..., -6.4751e-05,\n        -5.9825e-06, -4.5317e-05], device='cuda:0')",
-        "exp_avg_sq": "tensor([5.7841e-08, 6.2386e-08, 6.5328e-08,  ..., 5.7239e-08, 5.1020e-08,\n        7.6560e-08], device='cuda:0')"
+        "exp_avg": "tensor([[ 1.4430e-05, -1.1706e-05, -4.2505e-06,  ...,  9.4333e-06,\n         -4.7393e-05,  4.2187e-05],\n        [-1.1030e-06,  1.7644e-05,  1.4433e-05,  ..., -9.8050e-06,\n         -2.0379e-05,  1.6822e-05],\n        [-3.5394e-05, -3.3732e-05, -2.7051e-05,  ...,  2.9086e-05,\n          2.9219e-05, -5.8495e-05],\n        ...,\n        [ 1.6078e-05, -9.0171e-05,  1.2413e-05,  ..., -2.1114e-05,\n          1.3468e-05,  1.3303e-05],\n        [-3.7386e-06,  2.5794e-05, -1.3835e-05,  ..., -3.7103e-06,\n         -4.3155e-05, -1.8257e-05],\n        [ 2.2013e-05,  3.9908e-05,  3.2269e-05,  ...,  2.0060e-05,\n         -4.2385e-05, -3.9860e-05]], device='cuda:0')",
+        "exp_avg_sq": "tensor([[1.0855e-08, 1.1875e-08, 1.7739e-08,  ..., 7.9900e-09, 6.0227e-09,\n         1.6394e-08],\n        [1.6176e-08, 2.9787e-08, 2.2523e-08,  ..., 1.2433e-08, 1.4064e-08,\n         1.9748e-08],\n        [2.3372e-08, 3.6078e-08, 2.4778e-08,  ..., 1.9851e-08, 1.1107e-08,\n         4.1721e-08],\n        ...,\n        [1.8227e-08, 2.5559e-08, 3.2876e-08,  ..., 1.1376e-08, 1.5359e-08,\n         2.1241e-08],\n        [1.9527e-08, 4.0396e-08, 2.6205e-08,  ..., 1.6501e-08, 1.6760e-08,\n         2.3153e-08],\n        [1.8697e-08, 2.5146e-08, 2.3862e-08,  ..., 1.5342e-08, 1.0780e-08,\n         2.5310e-08]], device='cuda:0')"
       },
-      "31": {
+      "6": {
         "step": "tensor(1252.)",
-        "exp_avg": "tensor([[-8.7127e-07,  2.3044e-06,  5.4709e-07,  ...,  7.9924e-07,\n         -1.5380e-06, -2.5151e-06],\n        [-2.4811e-06, -9.3771e-07,  4.0047e-06,  ..., -2.0239e-06,\n          5.3400e-07, -1.1437e-06],\n        [-4.2748e-07, -2.8912e-06, -8.8406e-07,  ..., -2.2906e-06,\n          3.9683e-06, -1.3857e-06],\n        ...,\n        [-5.2646e-07, -2.6523e-07,  2.5794e-07,  ...,  1.8322e-06,\n          7.0432e-07, -1.9736e-06],\n        [ 9.9442e-07, -2.8867e-06, -2.5019e-06,  ..., -6.7045e-06,\n         -3.0750e-06, -9.6310e-06],\n        [-7.0356e-07,  3.3788e-06, -1.2132e-06,  ...,  5.8414e-06,\n          4.4418e-06,  6.7221e-06]], device='cuda:0')",
-        "exp_avg_sq": "tensor([[3.7381e-11, 4.7652e-11, 6.1029e-11,  ..., 3.9400e-11, 5.7003e-11,\n         4.7475e-11],\n        [6.2163e-11, 7.9086e-11, 1.2563e-10,  ..., 7.5610e-11, 8.1847e-11,\n         7.7077e-11],\n        [5.6396e-11, 7.7227e-11, 1.3232e-10,  ..., 6.4271e-11, 8.3451e-11,\n         7.6538e-11],\n        ...,\n        [5.1067e-11, 6.8155e-11, 1.2001e-10,  ..., 6.5949e-11, 7.1325e-11,\n         8.0999e-11],\n        [7.6909e-11, 8.6984e-11, 1.4603e-10,  ..., 7.7356e-11, 8.8286e-11,\n         8.7655e-11],\n        [6.5247e-11, 8.7666e-11, 1.1892e-10,  ..., 7.5368e-11, 8.9398e-11,\n         8.1653e-11]], device='cuda:0')"
+        "exp_avg": "tensor([ 0.0004, -0.0004], device='cuda:0')",
+        "exp_avg_sq": "tensor([1.2634e-05, 1.2634e-05], device='cuda:0')"
       }
     },
     "param_groups": [
       {
-        "lr": 0.01,
+        "lr": 0.00975530705321762,
         "name": "shared",
         "betas": [
           0.9,
@@ -328,11 +58,13 @@
         "initial_lr": 0.01,
         "params": [
           0,
-          1
+          1,
+          2,
+          3
         ]
       },
       {
-        "lr": 0.01,
+        "lr": 0.00975530705321762,
         "name": "scale_256",
         "betas": [
           0.9,
@@ -349,13 +81,11 @@
         "decoupled_weight_decay": true,
         "initial_lr": 0.01,
         "params": [
-          2,
-          3,
           4
         ]
       },
       {
-        "lr": 0.01,
+        "lr": 0.00975530705321762,
         "name": "scale_512",
         "betas": [
           0.9,
@@ -372,197 +102,11 @@
         "decoupled_weight_decay": true,
         "initial_lr": 0.01,
         "params": [
-          5,
-          6,
-          7
+          5
         ]
       },
       {
-        "lr": 0.01,
-        "name": "scale_768",
-        "betas": [
-          0.9,
-          0.999
-        ],
-        "eps": 1e-08,
-        "weight_decay": 1e-05,
-        "amsgrad": false,
-        "maximize": false,
-        "foreach": null,
-        "capturable": false,
-        "differentiable": false,
-        "fused": null,
-        "decoupled_weight_decay": true,
-        "initial_lr": 0.01,
-        "params": [
-          8,
-          9,
-          10
-        ]
-      },
-      {
-        "lr": 0.01,
-        "name": "scale_1024",
-        "betas": [
-          0.9,
-          0.999
-        ],
-        "eps": 1e-08,
-        "weight_decay": 1e-05,
-        "amsgrad": false,
-        "maximize": false,
-        "foreach": null,
-        "capturable": false,
-        "differentiable": false,
-        "fused": null,
-        "decoupled_weight_decay": true,
-        "initial_lr": 0.01,
-        "params": [
-          11,
-          12,
-          13
-        ]
-      },
-      {
-        "lr": 0.01,
-        "name": "scale_1280",
-        "betas": [
-          0.9,
-          0.999
-        ],
-        "eps": 1e-08,
-        "weight_decay": 1e-05,
-        "amsgrad": false,
-        "maximize": false,
-        "foreach": null,
-        "capturable": false,
-        "differentiable": false,
-        "fused": null,
-        "decoupled_weight_decay": true,
-        "initial_lr": 0.01,
-        "params": [
-          14,
-          15,
-          16
-        ]
-      },
-      {
-        "lr": 0.01,
-        "name": "scale_1536",
-        "betas": [
-          0.9,
-          0.999
-        ],
-        "eps": 1e-08,
-        "weight_decay": 1e-05,
-        "amsgrad": false,
-        "maximize": false,
-        "foreach": null,
-        "capturable": false,
-        "differentiable": false,
-        "fused": null,
-        "decoupled_weight_decay": true,
-        "initial_lr": 0.01,
-        "params": [
-          17,
-          18,
-          19
-        ]
-      },
-      {
-        "lr": 0.01,
-        "name": "scale_1792",
-        "betas": [
-          0.9,
-          0.999
-        ],
-        "eps": 1e-08,
-        "weight_decay": 1e-05,
-        "amsgrad": false,
-        "maximize": false,
-        "foreach": null,
-        "capturable": false,
-        "differentiable": false,
-        "fused": null,
-        "decoupled_weight_decay": true,
-        "initial_lr": 0.01,
-        "params": [
-          20,
-          21,
-          22
-        ]
-      },
-      {
-        "lr": 0.01,
-        "name": "scale_2048",
-        "betas": [
-          0.9,
-          0.999
-        ],
-        "eps": 1e-08,
-        "weight_decay": 1e-05,
-        "amsgrad": false,
-        "maximize": false,
-        "foreach": null,
-        "capturable": false,
-        "differentiable": false,
-        "fused": null,
-        "decoupled_weight_decay": true,
-        "initial_lr": 0.01,
-        "params": [
-          23,
-          24,
-          25
-        ]
-      },
-      {
-        "lr": 0.01,
-        "name": "scale_2304",
-        "betas": [
-          0.9,
-          0.999
-        ],
-        "eps": 1e-08,
-        "weight_decay": 1e-05,
-        "amsgrad": false,
-        "maximize": false,
-        "foreach": null,
-        "capturable": false,
-        "differentiable": false,
-        "fused": null,
-        "decoupled_weight_decay": true,
-        "initial_lr": 0.01,
-        "params": [
-          26,
-          27,
-          28
-        ]
-      },
-      {
-        "lr": 0.01,
-        "name": "scale_2560",
-        "betas": [
-          0.9,
-          0.999
-        ],
-        "eps": 1e-08,
-        "weight_decay": 1e-05,
-        "amsgrad": false,
-        "maximize": false,
-        "foreach": null,
-        "capturable": false,
-        "differentiable": false,
-        "fused": null,
-        "decoupled_weight_decay": true,
-        "initial_lr": 0.01,
-        "params": [
-          29,
-          30,
-          31
-        ]
-      },
-      {
-        "lr": 0.005,
+        "lr": 0.004877665762479736,
         "name": "fusion",
         "betas": [
           0.9,
@@ -579,119 +123,55 @@
         "decoupled_weight_decay": true,
         "initial_lr": 0.005,
         "params": [
-          32,
-          33,
-          34,
-          35,
-          36,
-          37,
-          38,
-          39,
-          40,
-          41,
-          42,
-          43,
-          44,
-          45,
-          46,
-          47,
-          48,
-          49,
-          50,
-          51,
-          52,
-          53,
-          54,
-          55,
-          56,
-          57,
-          58,
-          59,
-          60,
-          61
+          6
         ]
       }
     ]
   },
   "scheduler_state_dict": {
     "T_0": 10,
-    "T_i": 20,
+    "T_i": 10,
     "T_mult": 2,
     "eta_min": 1e-06,
-    "T_cur": 0,
+    "T_cur": 1,
     "base_lrs": [
-      0.01,
-      0.01,
-      0.01,
-      0.01,
-      0.01,
-      0.01,
-      0.01,
-      0.01,
       0.01,
       0.01,
       0.01,
       0.005
     ],
-    "last_epoch": 10,
+    "last_epoch": 1,
     "_step_count": 0,
     "_is_initial": false,
     "_get_lr_called_within_step": false,
     "_last_lr": [
-      0.01,
-      0.01,
-      0.01,
-      0.01,
-      0.01,
-      0.01,
-      0.01,
-      0.01,
-      0.01,
-      0.01,
-      0.01,
-      0.005
+      0.00975530705321762,
+      0.00975530705321762,
+      0.00975530705321762,
+      0.004877665762479736
     ]
   },
   "metrics": {
-    "best_val_acc": 82.374,
-    "best_epoch": 9,
+    "best_val_acc": 72.376,
+    "best_epoch": 0,
     "scale_accuracies": {
-      "256": 82.374,
-      "512": 82.728,
-      "768": 82.702,
-      "1024": 82.526,
-      "1280": 82.534,
-      "1536": 82.482,
-      "1792": 82.604,
-      "2048": 82.554,
-      "2304": 82.604,
-      "2560": 82.166
+      "256": 71.828,
+      "512": 72.234
     }
   },
   "train_config": {
     "name": "david_training",
-    "run_id": "20251012_050214",
+    "run_id": "20251012_060013",
     "dataset_name": "AbstractPhil/imagenet-clip-features-orderly",
-    "model_variant": "clip_vit_l14",
+    "model_variant": "clip_vit_b16",
     "num_classes": 1000,
-    "preset": "clip_vit_l14_deep",
+    "preset": "small_fast",
     "custom_config_path": null,
     "num_classes_override": null,
     "use_belly_override": null,
     "belly_expand_override": null,
-    "progressive_training_override": true,
-    "scale_warmup_epochs_override": {
-      "256": 0,
-      "512": 1,
-      "768": 2,
-      "1024": 3,
-      "1280": 4,
-      "1536": 5,
-      "1792": 6,
-      "2048": 7,
-      "2304": 8,
-      "2560": 9
-    },
+    "progressive_training_override": false,
+    "scale_warmup_epochs_override": null,
     "num_epochs": 10,
     "batch_size": 1024,
     "learning_rate": 0.01,