/home/yuqian_fu here1 here2 {'gpu': '0', 'data': 'photo', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 5, 'factor_num': 16, 'epochs': 70, 'nbatch': 100, 'batchsize': 6, 'lr': 0.01, 'lr_scheduler': 'cosine', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} stride: 5 /data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_train.hdf5 torch.Size([1499, 3, 227, 227]) torch.Size([1499]) --------------------------CA_multiple-------------------------- ---------------------------16 factors----------------- randm: True randn: True n: 3 randm: False /data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_val.hdf5 torch.Size([171, 3, 227, 227]) torch.Size([171]) -------------------------------------loading pretrain weights---------------------------------- Epoch 1, weight, value: tensor([[ 1.3781e-02, -2.1813e-02, -1.7643e-02, ..., -3.4786e-03, 7.0747e-03, 1.6051e-02], [ 1.6921e-02, 1.6796e-02, 3.9765e-03, ..., 6.6386e-03, 2.4196e-03, 9.7383e-03], [ 1.8894e-03, -6.7549e-03, 2.1032e-02, ..., -1.9916e-02, -1.0781e-02, 1.7924e-02], ..., [ 4.6711e-04, -7.8218e-03, 8.8535e-03, ..., -1.7695e-02, -1.9742e-02, -5.5186e-03], [-1.4236e-02, 1.5676e-02, -6.0038e-03, ..., 1.4177e-02, 1.0025e-02, -3.0311e-03], [-1.4947e-02, 1.6332e-02, 1.3555e-02, ..., 1.0778e-02, 9.8178e-05, -1.3844e-02]], device='cuda:0'), grad: None Epoch 1, bias, value: tensor([ 0.0165, -0.0171, 0.0209, -0.0155, 0.0020, -0.0164, 0.0005], device='cuda:0'), grad: None 249 0.01 changing lr ---------------------saving model at epoch 0---------------------------------------------------- epoch 0, time 348.21, cls_loss 9.8684 cls_loss_mapping 1.6022 cls_loss_causal 1.9054 re_mapping 1.0159 re_causal 1.0189 /// teacc 69.59 lr 0.00999497 Epoch 2, weight, value: tensor([[ 0.0774, 0.0641, 0.0080, ..., -0.0354, -0.0471, -0.0399], [-0.0260, -0.0458, -0.0257, ..., -0.0312, -0.0105, -0.0277], [-0.0029, -0.0232, 0.0263, ..., -0.0459, -0.0319, -0.0043], ..., [ 0.0248, 0.0389, 0.0264, ..., 0.0109, -0.0060, 0.0109], [-0.0430, -0.0148, -0.0446, ..., -0.0595, -0.0504, -0.0497], [ 0.0564, 0.0909, 0.0901, ..., 0.0745, 0.0518, 0.0574]], device='cuda:0'), grad: tensor([[-1.0231e-02, -3.4180e-03, -1.1459e-02, ..., 7.9918e-04, -4.7612e-04, -1.6251e-03], [ 1.5697e-03, 1.3151e-03, 1.7748e-03, ..., 1.2712e-03, 1.7195e-03, 1.3046e-03], [ 7.8049e-03, 2.6760e-03, 9.8877e-03, ..., 4.4975e-03, 4.5662e-03, 3.6087e-03], ..., [ 2.2011e-03, 1.6279e-03, 2.6894e-03, ..., 2.0332e-03, 2.4071e-03, 1.8253e-03], [ 1.2942e-05, 4.3437e-06, 1.5251e-05, ..., 2.3097e-06, 3.3155e-06, 3.5837e-06], [-3.1967e-03, -2.8076e-03, -3.6583e-03, ..., -2.7180e-03, -3.6221e-03, -2.7809e-03]], device='cuda:0') Epoch 2, bias, value: tensor([ 0.0328, -0.0842, 0.2049, 0.0465, 0.0064, -0.1213, -0.0942], device='cuda:0'), grad: tensor([-0.0866, 0.0029, 0.1262, -0.0544, 0.0152, 0.0001, -0.0035], device='cuda:0') 249 0.009994965332706574 changing lr ---------------------saving model at epoch 1---------------------------------------------------- epoch 1, time 347.28, cls_loss 3.2468 cls_loss_mapping 1.3975 cls_loss_causal 1.8708 re_mapping 0.7421 re_causal 0.7474 /// teacc 80.70 lr 0.00997987 Epoch 3, weight, value: tensor([[ 0.0785, 0.0747, 0.0011, ..., -0.0510, -0.0911, -0.0519], [-0.0322, -0.0574, -0.0140, ..., -0.0477, -0.0259, -0.0421], [-0.0056, -0.0181, 0.0213, ..., -0.0507, -0.0635, -0.0161], ..., [-0.0080, 0.0068, -0.0124, ..., -0.0106, -0.0226, -0.0205], [-0.0517, -0.0187, -0.0551, ..., -0.0376, -0.0179, -0.0239], [ 0.0674, 0.1059, 0.1026, ..., 0.1048, 0.0980, 0.0961]], device='cuda:0'), grad: tensor([[ 3.9101e-05, 3.8743e-05, 1.3423e-04, ..., 1.6487e-04, 1.4436e-04, 1.5497e-04], [ 1.8263e-03, 1.4849e-03, 3.7289e-03, ..., 2.7828e-03, 2.1725e-03, 2.8553e-03], [ 4.3368e-04, 9.7847e-04, 3.6945e-03, ..., 7.3967e-03, 7.4463e-03, 7.1487e-03], ..., [-2.4109e-03, -1.7910e-03, -5.5389e-03, ..., -3.6659e-03, -2.2316e-03, -3.3226e-03], [ 1.4687e-04, 1.4591e-04, 7.2622e-04, ..., 9.2077e-04, 7.4911e-04, 8.0156e-04], [ 3.0994e-06, 2.9132e-06, 1.0602e-05, ..., 1.2219e-05, 1.0259e-05, 1.1235e-05]], device='cuda:0') Epoch 3, bias, value: tensor([-0.0016, -0.0449, 0.2193, 0.0254, 0.0386, -0.1846, -0.0613], device='cuda:0'), grad: tensor([ 0.0033, 0.0715, 0.1316, -0.1288, -0.0958, 0.0180, 0.0003], device='cuda:0') 249 0.009979871469976196 changing lr epoch 2, time 346.20, cls_loss 3.2564 cls_loss_mapping 1.4641 cls_loss_causal 1.8684 re_mapping 0.6795 re_causal 0.6891 /// teacc 63.16 lr 0.00995475 Epoch 4, weight, value: tensor([[ 0.1029, 0.0873, 0.0093, ..., -0.0784, -0.1080, -0.0719], [-0.0437, -0.0644, -0.0257, ..., -0.0425, -0.0232, -0.0373], [-0.0310, -0.0245, -0.0142, ..., -0.0414, -0.0979, -0.0245], ..., [-0.0239, -0.0071, -0.0055, ..., -0.0321, -0.0321, -0.0325], [-0.0364, -0.0142, -0.0472, ..., -0.0315, -0.0031, -0.0247], [ 0.0723, 0.1110, 0.0884, ..., 0.1055, 0.1089, 0.1103]], device='cuda:0'), grad: tensor([[-7.9393e-04, -1.0939e-03, -4.1733e-03, ..., -1.1044e-03, -8.3113e-04, -1.1721e-03], [ 2.8372e-04, 3.9291e-04, 2.0332e-03, ..., 5.7697e-04, 4.7493e-04, 6.3467e-04], [ 1.4377e-04, 1.9968e-04, 1.2112e-03, ..., 3.5334e-04, 3.0017e-04, 3.9363e-04], ..., [ 4.0245e-04, 5.5695e-04, 2.8667e-03, ..., 8.1253e-04, 6.6805e-04, 8.9312e-04], [ 6.9797e-05, 9.6500e-05, 5.6171e-04, ..., 1.6296e-04, 1.3769e-04, 1.8096e-04], [-1.5214e-05, 1.3635e-06, 7.7248e-05, ..., 1.8284e-05, -2.4140e-06, 2.1279e-05]], device='cuda:0') Epoch 4, bias, value: tensor([-0.0196, 0.1142, 0.2746, 0.0413, -0.0684, -0.2400, -0.1112], device='cuda:0'), grad: tensor([-0.0831, 0.0550, 0.0363, -0.1049, 0.0771, 0.0164, 0.0032], device='cuda:0') 249 0.009954748808839675 changing lr epoch 3, time 349.42, cls_loss 2.7298 cls_loss_mapping 1.5867 cls_loss_causal 2.0432 re_mapping 0.5943 re_causal 0.6005 /// teacc 77.78 lr 0.00991965 Epoch 5, weight, value: tensor([[ 0.1081, 0.0900, 0.0151, ..., -0.0868, -0.1245, -0.0882], [-0.0605, -0.0741, -0.0279, ..., -0.0441, -0.0276, -0.0324], [-0.0019, -0.0298, 0.0355, ..., -0.0761, -0.1008, -0.0539], ..., [-0.0150, -0.0192, -0.0049, ..., -0.0336, -0.0156, -0.0245], [-0.0533, -0.0477, -0.0658, ..., -0.0337, -0.0129, -0.0378], [ 0.1000, 0.1785, 0.0972, ..., 0.1097, 0.1032, 0.1183]], device='cuda:0'), grad: tensor([[-1.2827e-03, -1.2982e-04, -2.0485e-03, ..., -5.8632e-03, -4.2610e-03, -4.8866e-03], [ 7.7248e-04, 4.5151e-05, 1.0691e-03, ..., 3.7918e-03, 3.3112e-03, 3.3703e-03], [ 1.0872e-04, 1.4193e-05, 1.6677e-04, ..., 4.7517e-04, 3.9291e-04, 4.0770e-04], ..., [ 5.5408e-04, 1.0329e-04, 8.8072e-04, ..., 2.1973e-03, 1.8730e-03, 1.8682e-03], [ 1.1146e-05, 1.7807e-06, 1.7256e-05, ..., 4.6402e-05, 3.9756e-05, 3.9905e-05], [ 1.4134e-05, 2.3544e-06, 1.9610e-05, ..., 5.8502e-05, 6.0886e-05, 5.3585e-05]], device='cuda:0') Epoch 5, bias, value: tensor([ 0.0686, 0.2019, 0.1705, 0.0255, 0.1121, -0.3466, -0.2412], device='cuda:0'), grad: tensor([-0.1044, 0.0511, 0.0085, -0.0029, 0.0458, 0.0009, 0.0010], device='cuda:0') 249 0.009919647942993149 changing lr epoch 4, time 346.66, cls_loss 3.5903 cls_loss_mapping 1.6275 cls_loss_causal 2.0572 re_mapping 0.5120 re_causal 0.5143 /// teacc 35.09 lr 0.00987464 Epoch 6, weight, value: tensor([[ 0.0966, 0.1068, 0.0187, ..., -0.0904, -0.1378, -0.0916], [-0.0324, -0.0771, 0.0030, ..., -0.0663, -0.0303, -0.0512], [ 0.0144, -0.0241, 0.0409, ..., -0.0741, -0.0996, -0.0597], ..., [-0.0197, -0.0390, -0.0097, ..., -0.0399, -0.0355, -0.0279], [-0.0444, -0.0515, -0.0625, ..., -0.0262, 0.0110, -0.0326], [ 0.0648, 0.2060, 0.0403, ..., 0.1140, 0.0686, 0.1289]], device='cuda:0'), grad: tensor([[ 3.0975e-03, 6.9201e-05, 4.6806e-03, ..., 4.5848e-04, 1.0166e-03, 3.1519e-04], [ 1.7853e-02, 2.6822e-04, 2.4368e-02, ..., 1.7824e-03, 5.9357e-03, 9.9754e-04], [-3.5362e-03, -1.0973e-04, -6.5269e-03, ..., -1.0843e-03, 3.8576e-04, 3.3927e-04], ..., [-1.7624e-02, -7.7367e-05, -1.9379e-02, ..., 6.7770e-05, -6.5308e-03, 1.6078e-05], [-3.2139e-03, -2.4211e-04, -8.5297e-03, ..., -1.7824e-03, -1.9932e-03, -2.1458e-03], [ 1.1911e-03, 2.1294e-05, 1.6680e-03, ..., 1.2493e-04, 4.1246e-04, 9.4652e-05]], device='cuda:0') Epoch 6, bias, value: tensor([ 0.0241, 0.2085, 0.0226, 0.1522, 0.1168, -0.1597, -0.3738], device='cuda:0'), grad: tensor([ 0.0638, 0.2766, -0.0494, 0.0591, -0.0909, -0.2791, 0.0196], device='cuda:0') 249 0.009874639560909117 changing lr epoch 5, time 347.35, cls_loss 3.8974 cls_loss_mapping 1.7399 cls_loss_causal 2.0707 re_mapping 0.4239 re_causal 0.4300 /// teacc 49.12 lr 0.00981981 Epoch 7, weight, value: tensor([[ 0.0807, 0.1011, 0.0078, ..., -0.0864, -0.1360, -0.0869], [-0.0310, -0.0750, 0.0037, ..., -0.0671, -0.0333, -0.0504], [ 0.0193, -0.0341, 0.0472, ..., -0.1011, -0.0952, -0.0832], ..., [-0.0081, -0.0604, 0.0070, ..., -0.0446, -0.0309, -0.0428], [-0.0455, -0.0510, -0.0497, ..., -0.0187, 0.0106, -0.0391], [ 0.0654, 0.2450, 0.0348, ..., 0.1289, 0.0681, 0.1527]], device='cuda:0'), grad: tensor([[ 2.0275e-03, 4.0388e-04, 5.3253e-03, ..., 4.7417e-03, 5.2872e-03, 4.8218e-03], [-2.3365e-03, -5.3227e-05, -5.9280e-03, ..., -1.7691e-03, -9.7513e-04, -1.3580e-03], [ 9.0122e-04, 1.6940e-04, 2.2984e-03, ..., 1.9093e-03, 2.0332e-03, 1.8892e-03], ..., [ 1.4031e-04, 2.3878e-04, 8.2350e-04, ..., -4.3373e-03, -4.6272e-03, -4.6959e-03], [-1.5945e-03, -9.1267e-04, -4.8599e-03, ..., -2.5234e-03, -3.9253e-03, -2.6951e-03], [ 3.6925e-05, 6.7800e-06, 9.6202e-05, ..., 7.4565e-05, 8.3208e-05, 7.4804e-05]], device='cuda:0') Epoch 7, bias, value: tensor([-0.0282, 0.2684, 0.0592, 0.0523, 0.3811, -0.0482, -0.6935], device='cuda:0'), grad: tensor([ 0.1420, -0.1486, 0.0656, 0.0565, 0.0802, -0.1981, 0.0025], device='cuda:0') 249 0.009819814303479266 changing lr epoch 6, time 348.88, cls_loss 2.1611 cls_loss_mapping 1.5626 cls_loss_causal 2.0639 re_mapping 0.4027 re_causal 0.4165 /// teacc 53.80 lr 0.00975528 Epoch 8, weight, value: tensor([[ 0.0680, 0.0850, -0.0052, ..., -0.0881, -0.1487, -0.0975], [-0.0428, -0.0737, 0.0020, ..., -0.0586, -0.0247, -0.0390], [ 0.0264, -0.0334, 0.0473, ..., -0.1011, -0.1059, -0.0839], ..., [-0.0076, -0.0591, 0.0107, ..., -0.0418, -0.0316, -0.0460], [-0.0439, -0.0489, -0.0498, ..., -0.0385, -0.0017, -0.0530], [ 0.0812, 0.2544, 0.0455, ..., 0.1305, 0.0698, 0.1522]], device='cuda:0'), grad: tensor([[ 5.3692e-04, 2.7537e-05, 2.0752e-03, ..., -8.0633e-04, -4.9686e-04, -8.4734e-04], [-4.8578e-05, 6.2227e-05, -8.9824e-05, ..., 7.8249e-04, 6.9737e-06, 7.3862e-04], [ 3.7956e-04, 1.3542e-04, 1.1082e-03, ..., 8.0729e-04, 7.6675e-04, 7.0953e-04], ..., [-6.8426e-04, -2.0921e-05, -2.8839e-03, ..., -1.2426e-03, -8.3065e-04, -1.3056e-03], [-5.6648e-04, -3.2020e-04, -1.4095e-03, ..., -2.0278e-04, -7.4446e-05, 1.4007e-04], [ 1.3745e-04, 3.2336e-05, 4.5538e-04, ..., 2.6965e-04, 2.5702e-04, 2.5105e-04]], device='cuda:0') Epoch 8, bias, value: tensor([ 0.1659, 0.0559, 0.0831, -0.0525, 0.2321, 0.0564, -0.5498], device='cuda:0'), grad: tensor([ 0.0200, 0.0316, 0.0316, 0.0189, -0.0889, -0.0259, 0.0126], device='cuda:0') 249 0.009755282581475767 changing lr epoch 7, time 347.51, cls_loss 1.8054 cls_loss_mapping 1.4719 cls_loss_causal 1.9194 re_mapping 0.3440 re_causal 0.3578 /// teacc 60.23 lr 0.00968117 Epoch 9, weight, value: tensor([[ 0.0677, 0.0792, 0.0004, ..., -0.0899, -0.1448, -0.1001], [-0.0454, -0.0682, -0.0072, ..., -0.0556, -0.0266, -0.0361], [ 0.0299, -0.0330, 0.0590, ..., -0.0957, -0.0870, -0.0778], ..., [-0.0150, -0.0534, 0.0009, ..., -0.0312, -0.0406, -0.0365], [-0.0395, -0.0464, -0.0477, ..., -0.0495, -0.0057, -0.0608], [ 0.0828, 0.2522, 0.0470, ..., 0.1284, 0.0701, 0.1498]], device='cuda:0'), grad: tensor([[ 9.4354e-05, 2.9802e-08, 5.6028e-04, ..., 2.4009e-04, 2.0373e-04, 1.9014e-04], [ 3.6687e-05, 0.0000e+00, 8.0585e-05, ..., 9.3341e-05, 6.4671e-05, 7.3910e-05], [ 8.9228e-05, 0.0000e+00, 5.1832e-04, ..., 2.2709e-04, 1.9145e-04, 1.7989e-04], ..., [ 1.3673e-04, 0.0000e+00, 8.4925e-04, ..., 3.4785e-04, 2.9922e-04, 2.7561e-04], [-4.1175e-04, 0.0000e+00, -2.3327e-03, ..., -1.0481e-03, -8.7738e-04, -8.2970e-04], [ 1.2487e-05, -2.9802e-08, 7.4208e-05, ..., 3.1859e-05, 2.6986e-05, 2.5228e-05]], device='cuda:0') Epoch 9, bias, value: tensor([ 0.2562, 0.0744, 0.0247, -0.0490, 0.3360, -0.0729, -0.5780], device='cuda:0'), grad: tensor([ 0.0390, -0.0083, 0.0350, 0.0175, 0.0630, -0.1512, 0.0052], device='cuda:0') 249 0.009681174353198686 changing lr epoch 8, time 344.34, cls_loss 1.1558 cls_loss_mapping 1.3434 cls_loss_causal 1.8095 re_mapping 0.3223 re_causal 0.3378 /// teacc 71.35 lr 0.00959764 Epoch 10, weight, value: tensor([[ 0.0637, 0.0797, -0.0172, ..., -0.0936, -0.1668, -0.1064], [-0.0445, -0.0665, -0.0026, ..., -0.0505, -0.0229, -0.0326], [ 0.0233, -0.0360, 0.0487, ..., -0.1026, -0.0940, -0.0814], ..., [-0.0145, -0.0573, 0.0077, ..., -0.0440, -0.0493, -0.0489], [-0.0382, -0.0424, -0.0457, ..., -0.0357, 0.0062, -0.0461], [ 0.0888, 0.2514, 0.0565, ..., 0.1306, 0.0868, 0.1515]], device='cuda:0'), grad: tensor([[-1.7223e-03, 3.1531e-05, -6.3858e-03, ..., -2.5387e-03, -4.1237e-03, -2.4738e-03], [ 3.2926e-04, 4.4525e-05, 6.4516e-04, ..., -1.1349e-03, -1.8263e-03, -1.7786e-03], [ 6.6805e-04, 3.7193e-05, 2.1992e-03, ..., 1.1168e-03, 1.1148e-03, 6.0797e-04], ..., [-7.1383e-04, -2.6226e-04, -6.6233e-04, ..., -3.1734e-04, 3.4962e-03, 3.0727e-03], [ 7.5817e-04, 8.6546e-05, 2.1133e-03, ..., 1.5717e-03, 4.4703e-04, 1.1426e-04], [ 1.2231e-04, 9.2536e-06, 3.8886e-04, ..., 2.2089e-04, 1.8620e-04, 1.0014e-04]], device='cuda:0') Epoch 10, bias, value: tensor([ 0.1999, 0.0502, -0.0396, 0.0345, 0.3870, -0.0554, -0.5850], device='cuda:0'), grad: tensor([-0.1238, 0.0103, 0.0461, 0.0386, -0.0331, 0.0534, 0.0084], device='cuda:0') 249 0.009597638862757255 changing lr epoch 9, time 346.82, cls_loss 1.0152 cls_loss_mapping 1.2447 cls_loss_causal 1.7866 re_mapping 0.2933 re_causal 0.2910 /// teacc 71.93 lr 0.00950484 Epoch 11, weight, value: tensor([[ 6.1647e-02, 7.1985e-02, -6.6718e-03, ..., -8.4501e-02, -1.5132e-01, -9.8548e-02], [-4.4163e-02, -6.3831e-02, -2.0746e-05, ..., -5.3221e-02, -2.4077e-02, -3.4561e-02], [ 2.2785e-02, -3.1338e-02, 4.7170e-02, ..., -1.0755e-01, -1.0055e-01, -8.6465e-02], ..., [-1.1486e-02, -5.7438e-02, 8.9108e-03, ..., -4.2754e-02, -4.6038e-02, -4.9284e-02], [-3.6297e-02, -4.0168e-02, -5.0431e-02, ..., -3.7683e-02, 1.5160e-04, -4.6720e-02], [ 8.4497e-02, 2.4673e-01, 4.9078e-02, ..., 1.2770e-01, 8.1287e-02, 1.4858e-01]], device='cuda:0'), grad: tensor([[-7.1973e-06, -1.4484e-05, -4.7159e-04, ..., -8.3542e-04, -3.4523e-04, -3.3140e-04], [ 5.9009e-05, 1.4454e-06, 4.6563e-04, ..., 1.5724e-04, 2.6941e-04, 1.0693e-04], [ 7.9334e-05, 3.0696e-06, 6.5899e-04, ..., 2.7585e-04, 3.8648e-04, 1.6868e-04], ..., [-2.1803e-04, 7.0184e-06, -1.3599e-03, ..., 1.3018e-04, -7.1812e-04, -1.1444e-04], [ 2.7567e-05, 1.3560e-06, 2.3830e-04, ..., 1.1498e-04, 1.4412e-04, 6.7711e-05], [ 1.1414e-05, 3.4273e-07, 9.2506e-05, ..., 3.5554e-05, 5.5403e-05, 2.3544e-05]], device='cuda:0') Epoch 11, bias, value: tensor([ 0.3634, -0.0167, -0.1066, 0.0363, 0.4038, -0.0840, -0.6044], device='cuda:0'), grad: tensor([-0.0597, 0.0193, 0.0304, 0.0157, -0.0216, 0.0118, 0.0040], device='cuda:0') 249 0.009504844339512096 changing lr ---------------------saving model at epoch 10---------------------------------------------------- epoch 10, time 346.95, cls_loss 0.7606 cls_loss_mapping 1.2027 cls_loss_causal 1.7721 re_mapping 0.2745 re_causal 0.2954 /// teacc 83.04 lr 0.00940298 Epoch 12, weight, value: tensor([[ 0.0608, 0.0733, -0.0066, ..., -0.0808, -0.1450, -0.0952], [-0.0455, -0.0656, -0.0015, ..., -0.0516, -0.0244, -0.0344], [ 0.0261, -0.0287, 0.0484, ..., -0.1044, -0.0978, -0.0841], ..., [-0.0148, -0.0566, 0.0036, ..., -0.0439, -0.0561, -0.0477], [-0.0346, -0.0400, -0.0446, ..., -0.0398, 0.0050, -0.0485], [ 0.0834, 0.2440, 0.0476, ..., 0.1261, 0.0795, 0.1467]], device='cuda:0'), grad: tensor([[ 2.9125e-03, 2.1839e-03, 1.1055e-02, ..., 5.8022e-03, 1.0353e-02, 4.9248e-03], [ 6.7568e-04, 3.6812e-04, 2.3785e-03, ..., 8.6689e-04, 1.7233e-03, 8.5974e-04], [ 2.8276e-04, 1.4174e-04, 1.0948e-03, ..., 5.7745e-04, 9.5844e-04, 4.5371e-04], ..., [-4.9820e-03, -3.3283e-03, -1.8692e-02, ..., -9.2621e-03, -1.6571e-02, -7.9346e-03], [ 2.0456e-04, 1.1301e-04, 7.2622e-04, ..., 2.7800e-04, 5.3978e-04, 2.6870e-04], [ 5.4210e-05, 3.0264e-05, 2.0385e-04, ..., 9.9778e-05, 1.7345e-04, 8.3029e-05]], device='cuda:0') Epoch 12, bias, value: tensor([ 0.2281, 0.0016, -0.0989, 0.1680, 0.3630, -0.0823, -0.5875], device='cuda:0'), grad: tensor([ 0.1564, 0.0358, 0.0146, 0.0443, -0.2646, 0.0108, 0.0028], device='cuda:0') 249 0.009402977659283692 changing lr epoch 11, time 348.26, cls_loss 0.8984 cls_loss_mapping 1.2257 cls_loss_causal 1.7599 re_mapping 0.2440 re_causal 0.2828 /// teacc 64.91 lr 0.00929224 Epoch 13, weight, value: tensor([[ 0.0547, 0.0676, -0.0145, ..., -0.0803, -0.1415, -0.0967], [-0.0441, -0.0585, -0.0022, ..., -0.0485, -0.0291, -0.0319], [ 0.0192, -0.0320, 0.0357, ..., -0.1065, -0.1117, -0.0865], ..., [-0.0133, -0.0545, 0.0126, ..., -0.0470, -0.0490, -0.0475], [-0.0264, -0.0354, -0.0327, ..., -0.0376, 0.0110, -0.0460], [ 0.0860, 0.2429, 0.0513, ..., 0.1244, 0.0811, 0.1448]], device='cuda:0'), grad: tensor([[ 8.0948e-03, 4.7188e-03, 2.4841e-02, ..., 6.6223e-03, 2.3666e-02, 5.8098e-03], [-1.4984e-02, -7.6065e-03, -5.0842e-02, ..., -1.6510e-02, -5.0995e-02, -1.5388e-02], [ 3.6883e-04, 9.5189e-05, 1.7767e-03, ..., 6.6376e-04, 1.8787e-03, 6.5994e-04], ..., [ 4.9591e-03, 2.2087e-03, 1.6663e-02, ..., 5.1651e-03, 1.6617e-02, 4.8523e-03], [-4.7541e-04, 1.4246e-04, -1.5488e-03, ..., 4.7541e-04, -9.6083e-04, 4.7874e-04], [ 5.4026e-04, 1.0842e-04, 2.3575e-03, ..., 7.8869e-04, 2.4357e-03, 7.8535e-04]], device='cuda:0') Epoch 13, bias, value: tensor([ 0.2046, 0.0119, -0.1103, 0.1370, 0.3917, -0.0697, -0.5732], device='cuda:0'), grad: tensor([ 0.1096, -0.2756, 0.0134, 0.0499, 0.0839, 0.0035, 0.0153], device='cuda:0') 249 0.009292243968009333 changing lr epoch 12, time 346.66, cls_loss 0.8350 cls_loss_mapping 1.2619 cls_loss_causal 1.8637 re_mapping 0.2095 re_causal 0.2569 /// teacc 75.44 lr 0.00917287 Epoch 14, weight, value: tensor([[ 0.0415, 0.0553, -0.0194, ..., -0.0846, -0.1425, -0.1021], [-0.0380, -0.0485, -0.0086, ..., -0.0332, -0.0208, -0.0135], [ 0.0273, -0.0239, 0.0513, ..., -0.1055, -0.0992, -0.0865], ..., [-0.0156, -0.0573, 0.0202, ..., -0.0573, -0.0553, -0.0609], [-0.0235, -0.0368, -0.0384, ..., -0.0439, -0.0034, -0.0513], [ 0.0841, 0.2409, 0.0416, ..., 0.1204, 0.0700, 0.1413]], device='cuda:0'), grad: tensor([[ 5.5084e-03, 3.4237e-03, 2.1042e-02, ..., 5.4054e-03, 2.2049e-02, 3.4885e-03], [ 2.6642e-02, 8.3694e-03, 8.7708e-02, ..., 2.2125e-02, 9.0027e-02, 1.5335e-02], [-2.5024e-03, -1.3599e-03, -6.5117e-03, ..., -8.6164e-04, -6.2943e-03, -9.2697e-04], ..., [-2.9495e-02, -1.0429e-02, -1.0193e-01, ..., -2.6520e-02, -1.0535e-01, -1.7776e-02], [-4.4012e-04, -1.0741e-04, -1.1673e-03, ..., -3.8886e-04, -1.3380e-03, -3.0255e-04], [ 2.0817e-05, 5.5730e-06, 5.6982e-05, ..., 1.8135e-05, 6.4135e-05, 1.3933e-05]], device='cuda:0') Epoch 14, bias, value: tensor([ 0.3738, -0.1402, -0.0957, 0.1669, 0.4431, -0.1405, -0.6153], device='cuda:0'), grad: tensor([ 7.6782e-02, 2.4097e-01, -1.0109e-02, 2.5806e-03, -3.0664e-01, -3.6335e-03, 1.7524e-04], device='cuda:0') 249 0.009172866268606516 changing lr epoch 13, time 350.62, cls_loss 0.6228 cls_loss_mapping 1.0631 cls_loss_causal 1.7189 re_mapping 0.2041 re_causal 0.2671 /// teacc 82.46 lr 0.00904508 Epoch 15, weight, value: tensor([[ 0.0415, 0.0524, -0.0151, ..., -0.0856, -0.1345, -0.1026], [-0.0382, -0.0479, -0.0106, ..., -0.0319, -0.0248, -0.0121], [ 0.0226, -0.0257, 0.0433, ..., -0.1051, -0.1057, -0.0868], ..., [-0.0088, -0.0529, 0.0357, ..., -0.0457, -0.0336, -0.0496], [-0.0274, -0.0357, -0.0453, ..., -0.0479, -0.0161, -0.0550], [ 0.0828, 0.2384, 0.0393, ..., 0.1179, 0.0668, 0.1387]], device='cuda:0'), grad: tensor([[ 1.0033e-03, 3.1304e-04, 4.0283e-03, ..., 3.7632e-03, 6.0463e-03, 3.6373e-03], [ 4.2419e-03, 8.4877e-04, 2.1423e-02, ..., 8.8348e-03, 1.9821e-02, 5.6648e-03], [ 1.6034e-04, 7.2122e-05, 6.4850e-04, ..., 5.4169e-04, 8.7976e-04, 5.1785e-04], ..., [-5.4092e-03, -7.6914e-04, -2.6001e-02, ..., -1.4404e-02, -2.8580e-02, -1.1177e-02], [-8.4305e-04, -7.4387e-04, -3.4904e-03, ..., -1.8578e-03, -3.1853e-03, -1.6642e-03], [ 1.1188e-04, 4.2140e-05, 4.5061e-04, ..., 4.0054e-04, 6.4659e-04, 3.8576e-04]], device='cuda:0') Epoch 15, bias, value: tensor([ 0.3647, -0.0599, -0.0534, 0.0527, 0.4296, -0.1354, -0.6058], device='cuda:0'), grad: tensor([ 0.0326, 0.1567, 0.0054, 0.0238, -0.1915, -0.0308, 0.0037], device='cuda:0') 249 0.00904508497187474 changing lr epoch 14, time 345.66, cls_loss 0.4806 cls_loss_mapping 0.9650 cls_loss_causal 1.6150 re_mapping 0.1894 re_causal 0.2416 /// teacc 77.19 lr 0.00890916 Epoch 16, weight, value: tensor([[ 0.0310, 0.0436, -0.0326, ..., -0.0893, -0.1463, -0.1070], [-0.0336, -0.0450, -0.0082, ..., -0.0312, -0.0237, -0.0113], [ 0.0246, -0.0232, 0.0485, ..., -0.1049, -0.0998, -0.0867], ..., [-0.0121, -0.0519, 0.0299, ..., -0.0440, -0.0457, -0.0475], [-0.0270, -0.0393, -0.0407, ..., -0.0581, -0.0167, -0.0640], [ 0.0877, 0.2420, 0.0476, ..., 0.1297, 0.0832, 0.1521]], device='cuda:0'), grad: tensor([[-2.5749e-03, -2.3155e-03, -2.6566e-02, ..., -1.3832e-02, -3.0853e-02, -1.3977e-02], [ 1.1644e-03, 1.1311e-03, 1.2085e-02, ..., 7.2594e-03, 1.5259e-02, 7.4463e-03], [ 1.4566e-05, 1.0222e-05, 8.0287e-05, ..., 4.5151e-05, 8.7559e-05, 4.4107e-05], ..., [ 1.1015e-03, 9.5701e-04, 1.2695e-02, ..., 5.4970e-03, 1.3535e-02, 5.4626e-03], [ 1.5274e-06, 1.0803e-06, 7.8306e-06, ..., 5.0291e-06, 8.9854e-06, 4.9099e-06], [ 1.0610e-04, 7.4029e-05, 5.2881e-04, ..., 3.1948e-04, 5.9080e-04, 3.1304e-04]], device='cuda:0') Epoch 16, bias, value: tensor([ 0.3305, -0.0721, -0.0274, 0.0394, 0.4979, -0.1767, -0.5990], device='cuda:0'), grad: tensor([-2.5024e-01, 1.2463e-01, 6.4611e-04, 1.0529e-02, 1.1017e-01, 6.5565e-05, 4.2992e-03], device='cuda:0') 249 0.008909157412340152 changing lr epoch 15, time 350.72, cls_loss 0.4993 cls_loss_mapping 0.9868 cls_loss_causal 1.6746 re_mapping 0.1866 re_causal 0.2383 /// teacc 80.70 lr 0.00876536 Epoch 17, weight, value: tensor([[ 0.0314, 0.0435, -0.0261, ..., -0.0824, -0.1354, -0.1003], [-0.0328, -0.0435, -0.0109, ..., -0.0273, -0.0238, -0.0068], [ 0.0247, -0.0233, 0.0477, ..., -0.1042, -0.1016, -0.0861], ..., [-0.0157, -0.0528, 0.0222, ..., -0.0466, -0.0491, -0.0507], [-0.0192, -0.0351, -0.0167, ..., -0.0552, 0.0035, -0.0613], [ 0.0835, 0.2381, 0.0352, ..., 0.1240, 0.0684, 0.1463]], device='cuda:0'), grad: tensor([[-2.4104e-04, -2.0754e-04, -8.3399e-04, ..., -1.1355e-04, -6.2180e-04, -2.6512e-04], [ 2.5183e-05, 2.0832e-05, 1.2672e-04, ..., 7.1526e-05, 1.3828e-04, 7.4565e-05], [ 5.3018e-05, 3.4392e-05, 4.8375e-04, ..., 4.2129e-04, 6.4516e-04, 3.6454e-04], ..., [ 1.0633e-04, 1.0949e-04, -1.2743e-04, ..., -6.1083e-04, -5.7030e-04, -3.9434e-04], [ 7.9200e-06, 6.3106e-06, 4.0919e-05, ..., 2.1562e-05, 4.3303e-05, 2.2501e-05], [ 1.1288e-05, 8.6278e-06, 6.8188e-05, ..., 4.3750e-05, 7.8440e-05, 4.2111e-05]], device='cuda:0') Epoch 17, bias, value: tensor([ 0.4034, -0.0693, -0.0067, -0.0020, 0.4321, -0.1292, -0.6354], device='cuda:0'), grad: tensor([-0.0068, 0.0015, 0.0075, 0.0033, -0.0069, 0.0005, 0.0009], device='cuda:0') 249 0.00876535733001806 changing lr ---------------------saving model at epoch 16---------------------------------------------------- epoch 16, time 349.48, cls_loss 0.4967 cls_loss_mapping 0.9364 cls_loss_causal 1.6148 re_mapping 0.1727 re_causal 0.2309 /// teacc 86.55 lr 0.00861397 Epoch 18, weight, value: tensor([[ 0.0292, 0.0414, -0.0274, ..., -0.0836, -0.1345, -0.1007], [-0.0355, -0.0440, -0.0144, ..., -0.0241, -0.0254, -0.0043], [ 0.0262, -0.0213, 0.0525, ..., -0.0980, -0.0898, -0.0800], ..., [-0.0131, -0.0515, 0.0331, ..., -0.0464, -0.0410, -0.0511], [-0.0188, -0.0325, -0.0242, ..., -0.0596, -0.0079, -0.0654], [ 0.0825, 0.2355, 0.0323, ..., 0.1215, 0.0647, 0.1435]], device='cuda:0'), grad: tensor([[-1.1539e-04, 5.9366e-04, -4.9591e-03, ..., 2.3193e-03, -3.9482e-03, 1.3437e-03], [ 1.4591e-04, 1.9825e-04, 1.5459e-03, ..., 2.2554e-04, 1.3132e-03, 2.3484e-04], [ 6.3598e-05, 2.2009e-05, 8.8978e-04, ..., 1.7858e-04, 8.3733e-04, 1.6987e-04], ..., [-5.5456e-04, -8.7023e-04, -4.5128e-03, ..., -3.4504e-03, -4.6577e-03, -2.6188e-03], [ 2.9516e-04, 2.9728e-05, 4.5509e-03, ..., 4.1389e-04, 4.1580e-03, 5.2357e-04], [ 1.5050e-05, 3.7774e-06, 2.1791e-04, ..., 3.6508e-05, 2.0361e-04, 3.6627e-05]], device='cuda:0') Epoch 18, bias, value: tensor([ 0.4269, 0.0323, -0.0444, -0.0081, 0.4233, -0.1873, -0.6500], device='cuda:0'), grad: tensor([-0.0424, 0.0184, 0.0097, 0.0242, -0.0605, 0.0482, 0.0024], device='cuda:0') 249 0.008613974319136962 changing lr epoch 17, time 349.89, cls_loss 0.4311 cls_loss_mapping 0.8407 cls_loss_causal 1.5558 re_mapping 0.1541 re_causal 0.2379 /// teacc 77.78 lr 0.00845531 Epoch 19, weight, value: tensor([[ 0.0282, 0.0388, -0.0318, ..., -0.0847, -0.1366, -0.1021], [-0.0396, -0.0462, -0.0243, ..., -0.0225, -0.0308, -0.0026], [ 0.0270, -0.0203, 0.0469, ..., -0.1002, -0.0983, -0.0844], ..., [-0.0153, -0.0498, 0.0313, ..., -0.0474, -0.0472, -0.0527], [-0.0148, -0.0298, -0.0123, ..., -0.0554, 0.0059, -0.0606], [ 0.0825, 0.2338, 0.0333, ..., 0.1194, 0.0642, 0.1420]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.8238e-06, ..., 3.3155e-06, 5.3868e-06, 3.4422e-06], [ 0.0000e+00, 0.0000e+00, -5.2392e-05, ..., -6.1572e-05, -1.0008e-04, -6.3956e-05], [-0.0000e+00, -0.0000e+00, 4.7684e-07, ..., 5.7369e-07, 9.2387e-07, 5.9605e-07], ..., [ 0.0000e+00, 0.0000e+00, 4.4107e-05, ..., 5.1796e-05, 8.4221e-05, 5.3823e-05], [ 0.0000e+00, 0.0000e+00, 2.0340e-06, ..., 2.3842e-06, 3.8743e-06, 2.4736e-06], [ 0.0000e+00, 0.0000e+00, 9.5367e-07, ..., 1.1250e-06, 1.8254e-06, 1.1697e-06]], device='cuda:0') Epoch 19, bias, value: tensor([ 0.3902, -0.0056, -0.0767, -0.0025, 0.4757, -0.1395, -0.6489], device='cuda:0'), grad: tensor([ 7.1049e-05, -1.3189e-03, 1.2085e-05, 4.8608e-05, 1.1101e-03, 5.1111e-05, 2.4080e-05], device='cuda:0') 249 0.008455313244934327 changing lr epoch 18, time 349.75, cls_loss 0.2844 cls_loss_mapping 0.8367 cls_loss_causal 1.5332 re_mapping 0.1430 re_causal 0.2201 /// teacc 86.55 lr 0.00828969 Epoch 20, weight, value: tensor([[ 0.0257, 0.0360, -0.0373, ..., -0.0822, -0.1377, -0.0995], [-0.0379, -0.0446, -0.0158, ..., -0.0208, -0.0206, -0.0008], [ 0.0269, -0.0205, 0.0475, ..., -0.1011, -0.0969, -0.0853], ..., [-0.0122, -0.0477, 0.0364, ..., -0.0440, -0.0426, -0.0496], [-0.0130, -0.0280, -0.0092, ..., -0.0557, 0.0070, -0.0598], [ 0.0787, 0.2307, 0.0253, ..., 0.1169, 0.0551, 0.1387]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 3.0565e-04, ..., 6.5416e-06, 1.8394e-04, 6.3702e-06], [ 7.4506e-09, 7.4506e-09, -9.8419e-04, ..., -1.6794e-05, -5.8889e-04, -1.6853e-05], [ 7.4506e-09, 7.4506e-09, 2.3067e-05, ..., 1.3933e-06, 1.4633e-05, 1.2591e-06], ..., [ 1.4901e-08, 7.4506e-09, 6.2418e-04, ..., 1.3180e-05, 3.7551e-04, 1.2860e-05], [-2.9802e-08, -2.2352e-08, 1.2152e-05, ..., 1.8477e-06, 8.6427e-06, 1.6168e-06], [ 0.0000e+00, 0.0000e+00, 5.0813e-06, ..., 1.1027e-06, 3.8818e-06, 9.6112e-07]], device='cuda:0') Epoch 20, bias, value: tensor([ 0.4344, 0.0165, -0.0767, -0.0469, 0.4770, -0.1534, -0.6582], device='cuda:0'), grad: tensor([ 0.0161, -0.0517, 0.0012, 0.0008, 0.0328, 0.0006, 0.0003], device='cuda:0') 249 0.008289693629698565 changing lr epoch 19, time 345.25, cls_loss 0.4219 cls_loss_mapping 0.7627 cls_loss_causal 1.4024 re_mapping 0.1221 re_causal 0.1921 /// teacc 85.38 lr 0.00811745 Epoch 21, weight, value: tensor([[ 0.0246, 0.0335, -0.0398, ..., -0.0794, -0.1339, -0.0970], [-0.0375, -0.0443, -0.0158, ..., -0.0173, -0.0177, 0.0023], [ 0.0239, -0.0216, 0.0428, ..., -0.1032, -0.1000, -0.0871], ..., [-0.0065, -0.0431, 0.0462, ..., -0.0464, -0.0408, -0.0515], [-0.0135, -0.0270, -0.0098, ..., -0.0579, 0.0038, -0.0615], [ 0.0774, 0.2280, 0.0237, ..., 0.1151, 0.0532, 0.1367]], device='cuda:0'), grad: tensor([[ 3.1769e-05, -1.3721e-04, -6.9380e-05, ..., 1.2875e-04, -1.4472e-04, 1.0723e-04], [-6.3133e-04, -3.1090e-04, -3.1681e-03, ..., -6.6280e-04, -3.2272e-03, -6.7329e-04], [ 1.6415e-04, 7.4446e-05, 1.1692e-03, ..., 2.9182e-04, 1.1387e-03, 2.0289e-04], ..., [ 2.4462e-04, 2.8396e-04, 3.3855e-04, ..., -3.3498e-04, 5.1117e-04, -5.7593e-06], [ 1.1869e-05, 5.7518e-06, 7.9215e-05, ..., 1.9401e-05, 7.8201e-05, 1.4484e-05], [ 2.7493e-06, 1.6987e-06, 2.1398e-05, ..., 7.3612e-06, 2.2352e-05, 5.3942e-06]], device='cuda:0') Epoch 21, bias, value: tensor([ 0.3833, 0.0472, -0.0680, -0.0074, 0.4733, -0.1728, -0.6628], device='cuda:0'), grad: tensor([-0.0019, -0.0153, 0.0072, 0.0116, -0.0022, 0.0005, 0.0001], device='cuda:0') 249 0.00811744900929367 changing lr epoch 20, time 346.79, cls_loss 0.2886 cls_loss_mapping 0.6949 cls_loss_causal 1.3418 re_mapping 0.1059 re_causal 0.1685 /// teacc 84.80 lr 0.00793893 Epoch 22, weight, value: tensor([[ 0.0233, 0.0313, -0.0416, ..., -0.0790, -0.1330, -0.0969], [-0.0378, -0.0435, -0.0128, ..., -0.0185, -0.0118, 0.0010], [ 0.0220, -0.0224, 0.0384, ..., -0.1022, -0.1030, -0.0863], ..., [-0.0045, -0.0410, 0.0448, ..., -0.0452, -0.0463, -0.0504], [-0.0108, -0.0260, -0.0016, ..., -0.0557, 0.0125, -0.0590], [ 0.0770, 0.2260, 0.0238, ..., 0.1138, 0.0528, 0.1352]], device='cuda:0'), grad: tensor([[-1.5383e-03, -1.8444e-03, -2.1469e-02, ..., -1.0399e-02, -2.5879e-02, -9.4986e-03], [ 2.7609e-04, 3.4833e-04, 4.5853e-03, ..., 2.2602e-03, 5.6686e-03, 2.0962e-03], [ 7.6443e-06, 6.0983e-06, 1.0163e-04, ..., 5.5999e-05, 1.0902e-04, 4.1693e-05], ..., [ 1.1730e-03, 1.4095e-03, 1.6357e-02, ..., 7.8964e-03, 1.9699e-02, 7.2136e-03], [ 3.9965e-05, 3.9756e-05, 2.0468e-04, ..., 8.4817e-05, 1.8144e-04, 6.2823e-05], [ 1.2055e-05, 1.1958e-05, 6.2764e-05, ..., 2.6688e-05, 5.6267e-05, 1.9848e-05]], device='cuda:0') Epoch 22, bias, value: tensor([ 0.3787, 0.0097, -0.0376, 0.0187, 0.4699, -0.1906, -0.6559], device='cuda:0'), grad: tensor([-0.2013, 0.0428, 0.0012, 0.0016, 0.1532, 0.0020, 0.0006], device='cuda:0') 249 0.007938926261462368 changing lr epoch 21, time 346.47, cls_loss 0.2398 cls_loss_mapping 0.5935 cls_loss_causal 1.1991 re_mapping 0.0881 re_causal 0.1494 /// teacc 82.46 lr 0.00775448 Epoch 23, weight, value: tensor([[ 0.0253, 0.0321, -0.0354, ..., -0.0819, -0.1262, -0.0989], [-0.0420, -0.0463, -0.0222, ..., -0.0196, -0.0175, -0.0004], [ 0.0206, -0.0235, 0.0366, ..., -0.1000, -0.1013, -0.0838], ..., [-0.0007, -0.0374, 0.0553, ..., -0.0389, -0.0385, -0.0454], [-0.0104, -0.0250, -0.0011, ..., -0.0555, 0.0127, -0.0587], [ 0.0764, 0.2240, 0.0234, ..., 0.1123, 0.0518, 0.1337]], device='cuda:0'), grad: tensor([[-5.4389e-05, -7.5281e-05, 5.6076e-04, ..., 1.3840e-04, 6.6090e-04, 8.7142e-05], [ 2.1443e-05, 2.7061e-05, 3.8767e-04, ..., 8.2433e-05, 3.6097e-04, 7.1824e-05], [ 5.1558e-06, 3.9563e-06, 4.6968e-05, ..., 9.5740e-06, 4.2051e-05, 7.6294e-06], ..., [ 2.5108e-05, 4.2379e-05, -1.0281e-03, ..., -2.3723e-04, -1.0948e-03, -1.7190e-04], [ 7.1526e-07, 4.7684e-07, 8.8960e-06, ..., 1.8738e-06, 8.3223e-06, 1.4454e-06], [ 5.3644e-07, 4.7684e-07, 3.5278e-06, ..., 6.8918e-07, 2.9393e-06, 5.9232e-07]], device='cuda:0') Epoch 23, bias, value: tensor([ 0.3388, -0.0143, -0.0064, 0.0110, 0.5262, -0.2086, -0.6536], device='cuda:0'), grad: tensor([ 4.0855e-03, 3.0880e-03, 3.2306e-04, 1.3936e-04, -7.7248e-03, 6.1810e-05, 2.4408e-05], device='cuda:0') 249 0.007754484907260515 changing lr epoch 22, time 343.89, cls_loss 0.2083 cls_loss_mapping 0.5953 cls_loss_causal 1.3084 re_mapping 0.0874 re_causal 0.1513 /// teacc 80.12 lr 0.00756450 Epoch 24, weight, value: tensor([[ 0.0259, 0.0315, -0.0369, ..., -0.0825, -0.1337, -0.1016], [-0.0418, -0.0450, -0.0224, ..., -0.0180, -0.0164, 0.0007], [ 0.0223, -0.0211, 0.0504, ..., -0.0949, -0.0819, -0.0779], ..., [-0.0006, -0.0368, 0.0494, ..., -0.0380, -0.0434, -0.0445], [-0.0110, -0.0256, -0.0048, ..., -0.0579, 0.0065, -0.0612], [ 0.0736, 0.2195, 0.0208, ..., 0.1090, 0.0525, 0.1319]], device='cuda:0'), grad: tensor([[-9.3132e-08, -1.9372e-07, -6.5938e-07, ..., -2.3469e-07, -7.3388e-07, -2.3097e-07], [ 7.4506e-09, 1.4901e-08, 4.8429e-08, ..., 1.8626e-08, 5.5879e-08, 1.8626e-08], [ 1.1176e-08, 7.4506e-09, 5.2154e-08, ..., 2.2352e-08, 7.0781e-08, 2.6077e-08], ..., [ 8.5682e-08, 1.7509e-07, 6.0350e-07, ..., 2.1607e-07, 6.7055e-07, 2.1234e-07], [-1.1176e-08, -7.4506e-09, -5.2154e-08, ..., -1.8626e-08, -6.7055e-08, -2.6077e-08], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 24, bias, value: tensor([ 0.3312, -0.0065, 0.0246, 0.0131, 0.5065, -0.2245, -0.6513], device='cuda:0'), grad: tensor([-2.1271e-06, 1.5646e-07, 1.3411e-07, 0.0000e+00, 1.9483e-06, -1.1921e-07, 0.0000e+00], device='cuda:0') 249 0.007564496387029534 changing lr ---------------------saving model at epoch 23---------------------------------------------------- epoch 23, time 350.19, cls_loss 0.2486 cls_loss_mapping 0.5770 cls_loss_causal 1.2395 re_mapping 0.0702 re_causal 0.1356 /// teacc 87.72 lr 0.00736934 Epoch 25, weight, value: tensor([[ 0.0267, 0.0314, -0.0344, ..., -0.0778, -0.1292, -0.0977], [-0.0408, -0.0440, -0.0236, ..., -0.0215, -0.0203, -0.0027], [ 0.0218, -0.0228, 0.0440, ..., -0.0951, -0.0880, -0.0785], ..., [-0.0041, -0.0360, 0.0495, ..., -0.0334, -0.0376, -0.0394], [-0.0122, -0.0253, -0.0072, ..., -0.0591, 0.0014, -0.0627], [ 0.0729, 0.2174, 0.0201, ..., 0.1076, 0.0512, 0.1303]], device='cuda:0'), grad: tensor([[ 2.5064e-05, 6.3181e-06, 5.1546e-04, ..., 1.9991e-04, 5.3024e-04, 2.0480e-04], [ 5.9462e-04, 3.4310e-06, 1.1429e-02, ..., 1.4524e-03, 8.9111e-03, 1.6298e-03], [ 1.6734e-05, 6.5640e-06, 2.9111e-04, ..., 1.4937e-04, 3.1781e-04, 1.5056e-04], ..., [-6.4468e-04, -1.8567e-05, -1.2383e-02, ..., -1.8587e-03, -9.9106e-03, -2.0428e-03], [ 3.2559e-06, 7.9721e-07, 5.6028e-05, ..., 2.0042e-05, 5.3853e-05, 2.0519e-05], [ 1.3672e-06, 3.1665e-07, 2.4319e-05, ..., 8.2403e-06, 2.2992e-05, 8.4564e-06]], device='cuda:0') Epoch 25, bias, value: tensor([ 0.3551, -0.0608, 0.0083, -0.0027, 0.5596, -0.2145, -0.6520], device='cuda:0'), grad: tensor([ 0.0057, 0.1221, 0.0036, 0.0008, -0.1331, 0.0007, 0.0003], device='cuda:0') 249 0.007369343312364995 changing lr ---------------------saving model at epoch 24---------------------------------------------------- epoch 24, time 349.95, cls_loss 0.1266 cls_loss_mapping 0.4742 cls_loss_causal 1.1183 re_mapping 0.0595 re_causal 0.1259 /// teacc 88.30 lr 0.00716942 Epoch 26, weight, value: tensor([[ 2.6065e-02, 3.1333e-02, -3.2591e-02, ..., -7.8665e-02, -1.2607e-01, -9.7870e-02], [-4.1028e-02, -4.3888e-02, -2.8106e-02, ..., -2.2541e-02, -2.3828e-02, -3.8757e-03], [ 2.0624e-02, -2.3310e-02, 3.7576e-02, ..., -9.6356e-02, -9.4617e-02, -8.0039e-02], ..., [-3.9162e-03, -3.5971e-02, 4.8947e-02, ..., -3.0996e-02, -3.8739e-02, -3.7762e-02], [-1.0823e-02, -2.4368e-02, -2.1525e-04, ..., -5.7820e-02, 8.0542e-03, -6.0702e-02], [ 7.2154e-02, 2.1539e-01, 1.9567e-02, ..., 1.0641e-01, 5.0375e-02, 1.2895e-01]], device='cuda:0'), grad: tensor([[ 8.8072e-04, 5.1594e-04, 4.9973e-03, ..., 1.3981e-03, 6.2180e-03, 2.2316e-03], [-1.3784e-07, -5.2899e-07, -1.6801e-06, ..., 4.0233e-07, -1.3262e-06, 4.5076e-07], [ 1.2982e-04, 7.1049e-05, 7.2861e-04, ..., 1.9443e-04, 9.0170e-04, 3.2187e-04], ..., [-1.8549e-03, -1.0757e-03, -1.0483e-02, ..., -2.9087e-03, -1.3023e-02, -4.6730e-03], [ 6.5744e-05, 3.9637e-05, 3.7694e-04, ..., 1.0800e-04, 4.7064e-04, 1.6940e-04], [ 1.5274e-07, 1.1176e-07, 1.0990e-06, ..., 3.8370e-07, 1.4305e-06, 5.2527e-07]], device='cuda:0') Epoch 26, bias, value: tensor([ 0.3469, -0.1051, -0.0202, 0.0171, 0.5930, -0.1889, -0.6496], device='cuda:0'), grad: tensor([ 9.0256e-03, -6.0312e-06, 1.2798e-03, 7.8125e-03, -1.8814e-02, 6.9237e-04, 2.4401e-06], device='cuda:0') 249 0.0071694186955877925 changing lr epoch 25, time 346.90, cls_loss 0.1057 cls_loss_mapping 0.4313 cls_loss_causal 1.1527 re_mapping 0.0601 re_causal 0.1291 /// teacc 87.72 lr 0.00696513 Epoch 27, weight, value: tensor([[ 0.0240, 0.0297, -0.0393, ..., -0.0773, -0.1310, -0.0966], [-0.0409, -0.0436, -0.0290, ..., -0.0234, -0.0260, -0.0050], [ 0.0212, -0.0224, 0.0399, ..., -0.0959, -0.0926, -0.0799], ..., [-0.0022, -0.0348, 0.0558, ..., -0.0309, -0.0303, -0.0371], [-0.0113, -0.0246, -0.0006, ..., -0.0579, 0.0076, -0.0603], [ 0.0715, 0.2134, 0.0192, ..., 0.1053, 0.0496, 0.1276]], device='cuda:0'), grad: tensor([[ 4.4703e-08, 6.7055e-08, 8.5682e-07, ..., 2.0489e-07, 6.3702e-07, 2.1979e-07], [-2.1234e-07, -9.6112e-07, -1.2949e-05, ..., -8.9407e-08, -6.5267e-06, -3.5390e-07], [ 9.3132e-08, 9.6858e-08, 9.4250e-07, ..., 6.2585e-07, 1.1213e-06, 6.5938e-07], ..., [ 2.7940e-07, 9.5367e-07, 1.2428e-05, ..., 7.0408e-07, 6.9104e-06, 9.6858e-07], [ 1.2256e-06, 1.0133e-06, 8.2999e-06, ..., 8.7470e-06, 1.3277e-05, 9.1121e-06], [ 1.0058e-07, 8.1956e-08, 6.8918e-07, ..., 7.0781e-07, 1.0803e-06, 7.3761e-07]], device='cuda:0') Epoch 27, bias, value: tensor([ 0.3509, -0.1067, -0.0124, 0.0109, 0.5675, -0.1701, -0.6468], device='cuda:0'), grad: tensor([ 1.0535e-05, -1.7989e-04, 9.7156e-06, -8.5473e-05, 1.6975e-04, 6.9559e-05, 5.8375e-06], device='cuda:0') 249 0.0069651251582696205 changing lr ---------------------saving model at epoch 26---------------------------------------------------- epoch 26, time 350.28, cls_loss 0.1538 cls_loss_mapping 0.4250 cls_loss_causal 1.0776 re_mapping 0.0632 re_causal 0.1255 /// teacc 90.64 lr 0.00675687 Epoch 28, weight, value: tensor([[ 0.0238, 0.0302, -0.0388, ..., -0.0751, -0.1280, -0.0947], [-0.0399, -0.0428, -0.0256, ..., -0.0245, -0.0237, -0.0060], [ 0.0207, -0.0227, 0.0378, ..., -0.0960, -0.0952, -0.0803], ..., [-0.0025, -0.0350, 0.0536, ..., -0.0307, -0.0333, -0.0368], [-0.0113, -0.0245, -0.0016, ..., -0.0580, 0.0047, -0.0603], [ 0.0711, 0.2119, 0.0192, ..., 0.1043, 0.0492, 0.1265]], device='cuda:0'), grad: tensor([[ 4.2057e-04, 4.4632e-04, 1.9722e-03, ..., 1.6460e-03, 2.2316e-03, 1.5688e-03], [-2.1400e-03, -2.7809e-03, -1.2062e-02, ..., -2.9259e-03, -8.4991e-03, -3.2768e-03], [ 1.3411e-04, 1.1194e-04, 5.8365e-04, ..., 5.7316e-04, 7.2050e-04, 5.3406e-04], ..., [ 1.4725e-03, 2.1286e-03, 9.0027e-03, ..., 2.7800e-04, 4.9706e-03, 7.7105e-04], [ 3.4392e-05, 3.2216e-05, 1.6570e-04, ..., 1.0234e-04, 1.6093e-04, 9.8109e-05], [ 1.6198e-05, 1.4491e-05, 7.2718e-05, ..., 6.4850e-05, 8.5056e-05, 6.0856e-05]], device='cuda:0') Epoch 28, bias, value: tensor([ 0.3468, -0.1148, -0.0048, 0.0030, 0.5668, -0.1701, -0.6335], device='cuda:0'), grad: tensor([ 0.0176, -0.1155, 0.0043, 0.0018, 0.0899, 0.0013, 0.0006], device='cuda:0') 249 0.006756874120406716 changing lr epoch 27, time 347.40, cls_loss 0.1100 cls_loss_mapping 0.3891 cls_loss_causal 1.0223 re_mapping 0.0498 re_causal 0.1108 /// teacc 86.55 lr 0.00654508 Epoch 29, weight, value: tensor([[ 0.0251, 0.0313, -0.0354, ..., -0.0745, -0.1234, -0.0937], [-0.0389, -0.0416, -0.0225, ..., -0.0243, -0.0201, -0.0057], [ 0.0203, -0.0227, 0.0378, ..., -0.0959, -0.0940, -0.0799], ..., [-0.0031, -0.0347, 0.0522, ..., -0.0271, -0.0322, -0.0340], [-0.0118, -0.0255, -0.0050, ..., -0.0594, -0.0014, -0.0617], [ 0.0705, 0.2101, 0.0188, ..., 0.1032, 0.0484, 0.1253]], device='cuda:0'), grad: tensor([[ 9.3341e-05, 2.8419e-04, -1.9245e-03, ..., -8.4925e-04, -1.9855e-03, -7.6866e-04], [ 8.4639e-06, 6.5193e-07, 1.0949e-04, ..., 4.6611e-05, 1.0580e-04, 4.3750e-05], [-4.2582e-04, -4.3154e-04, -9.9754e-04, ..., -3.8195e-04, -7.8344e-04, -4.0030e-04], ..., [ 3.0875e-04, 1.4150e-04, 2.6627e-03, ..., 1.1206e-03, 2.5196e-03, 1.0653e-03], [ 7.2010e-06, 3.9637e-06, 5.4628e-05, ..., 2.2888e-05, 5.1260e-05, 2.1860e-05], [ 2.4959e-06, 8.6799e-07, 2.4691e-05, ..., 1.0438e-05, 2.3559e-05, 9.8720e-06]], device='cuda:0') Epoch 29, bias, value: tensor([ 0.3258, -0.1182, -0.0050, -0.0046, 0.5885, -0.1621, -0.6312], device='cuda:0'), grad: tensor([-0.0291, 0.0014, -0.0080, 0.0009, 0.0337, 0.0007, 0.0003], device='cuda:0') 249 0.00654508497187474 changing lr ---------------------saving model at epoch 28---------------------------------------------------- epoch 28, time 350.09, cls_loss 0.0846 cls_loss_mapping 0.3395 cls_loss_causal 1.0291 re_mapping 0.0492 re_causal 0.0999 /// teacc 91.23 lr 0.00633018 Epoch 30, weight, value: tensor([[ 0.0245, 0.0303, -0.0354, ..., -0.0729, -0.1203, -0.0921], [-0.0388, -0.0409, -0.0223, ..., -0.0253, -0.0204, -0.0068], [ 0.0205, -0.0223, 0.0394, ..., -0.0949, -0.0916, -0.0788], ..., [-0.0039, -0.0353, 0.0485, ..., -0.0274, -0.0367, -0.0344], [-0.0116, -0.0252, -0.0048, ..., -0.0589, -0.0014, -0.0611], [ 0.0707, 0.2090, 0.0198, ..., 0.1023, 0.0487, 0.1243]], device='cuda:0'), grad: tensor([[ 1.7681e-03, 4.8485e-03, 2.3468e-02, ..., 1.2390e-02, 3.2959e-02, 1.2566e-02], [ 7.0073e-06, 1.9208e-05, 9.2983e-05, ..., 4.9084e-05, 1.3053e-04, 4.9770e-05], [ 2.7776e-05, 7.6175e-05, 3.6860e-04, ..., 1.9467e-04, 5.1785e-04, 1.9753e-04], ..., [-1.8711e-03, -5.1308e-03, -2.4826e-02, ..., -1.3115e-02, -3.4882e-02, -1.3306e-02], [ 6.5006e-07, 1.8254e-06, 8.8066e-06, ..., 4.6603e-06, 1.2383e-05, 4.7237e-06], [ 8.1956e-08, 2.2165e-07, 1.0766e-06, ..., 5.6811e-07, 1.5125e-06, 5.7742e-07]], device='cuda:0') Epoch 30, bias, value: tensor([ 0.3038, -0.1232, 0.0082, -0.0091, 0.5983, -0.1606, -0.6240], device='cuda:0'), grad: tensor([ 7.3120e-02, 2.8968e-04, 1.1492e-03, 2.8114e-03, -7.7393e-02, 2.7552e-05, 3.3565e-06], device='cuda:0') 249 0.006330184227833378 changing lr epoch 29, time 347.50, cls_loss 0.0666 cls_loss_mapping 0.3109 cls_loss_causal 1.0087 re_mapping 0.0451 re_causal 0.1122 /// teacc 88.89 lr 0.00611260 Epoch 31, weight, value: tensor([[ 0.0241, 0.0298, -0.0359, ..., -0.0724, -0.1209, -0.0915], [-0.0386, -0.0409, -0.0225, ..., -0.0254, -0.0200, -0.0071], [ 0.0206, -0.0218, 0.0393, ..., -0.0944, -0.0911, -0.0786], ..., [-0.0044, -0.0350, 0.0488, ..., -0.0263, -0.0352, -0.0333], [-0.0110, -0.0247, -0.0016, ..., -0.0572, 0.0027, -0.0592], [ 0.0708, 0.2078, 0.0205, ..., 0.1016, 0.0490, 0.1235]], device='cuda:0'), grad: tensor([[ 2.9802e-08, 3.3528e-07, 2.7847e-06, ..., 2.5500e-06, 4.0308e-06, 2.2575e-06], [ 8.9407e-08, 4.4703e-08, 4.9360e-07, ..., 1.6578e-07, 4.1164e-07, 1.8440e-07], [ 6.7055e-08, 6.3330e-08, 6.0722e-07, ..., 3.5018e-07, 6.5938e-07, 3.3528e-07], ..., [ 1.2107e-07, -2.7753e-07, -2.0042e-06, ..., -2.3581e-06, -3.4366e-06, -1.9912e-06], [-3.6135e-07, -2.0303e-07, -2.2780e-06, ..., -8.9034e-07, -2.0526e-06, -9.7975e-07], [ 4.6566e-08, 2.7940e-08, 2.9802e-07, ..., 1.2666e-07, 2.7753e-07, 1.3597e-07]], device='cuda:0') Epoch 31, bias, value: tensor([ 0.3060, -0.1185, 0.0095, -0.0269, 0.5951, -0.1523, -0.6192], device='cuda:0'), grad: tensor([ 5.0515e-05, 3.3751e-06, 6.9812e-06, 1.1921e-06, -4.6074e-05, -1.8597e-05, 2.5593e-06], device='cuda:0') 249 0.006112604669781575 changing lr epoch 30, time 348.46, cls_loss 0.0678 cls_loss_mapping 0.3041 cls_loss_causal 1.0036 re_mapping 0.0422 re_causal 0.0991 /// teacc 87.72 lr 0.00589278 Epoch 32, weight, value: tensor([[ 0.0240, 0.0295, -0.0348, ..., -0.0705, -0.1179, -0.0899], [-0.0389, -0.0411, -0.0259, ..., -0.0262, -0.0234, -0.0080], [ 0.0193, -0.0227, 0.0352, ..., -0.0947, -0.0946, -0.0793], ..., [-0.0031, -0.0331, 0.0529, ..., -0.0269, -0.0323, -0.0332], [-0.0111, -0.0248, -0.0027, ..., -0.0571, 0.0015, -0.0591], [ 0.0704, 0.2064, 0.0204, ..., 0.1008, 0.0485, 0.1226]], device='cuda:0'), grad: tensor([[ 5.2303e-06, 1.8626e-09, 7.0512e-05, ..., 1.5959e-05, 4.7475e-05, 1.5870e-05], [ 1.6205e-07, 3.7253e-09, 2.1849e-06, ..., 4.0792e-07, 1.4659e-06, 4.2841e-07], [ 3.4273e-07, 0.0000e+00, 4.6194e-06, ..., 1.0468e-06, 3.1106e-06, 1.0412e-06], ..., [-6.1430e-06, -5.5879e-09, -8.2850e-05, ..., -1.8671e-05, -5.5790e-05, -1.8582e-05], [ 6.1467e-08, 0.0000e+00, 8.2888e-07, ..., 1.8813e-07, 5.5879e-07, 1.8626e-07], [ 1.0431e-07, 0.0000e+00, 1.4119e-06, ..., 3.2037e-07, 9.5181e-07, 3.1851e-07]], device='cuda:0') Epoch 32, bias, value: tensor([ 0.2986, -0.1310, 0.0074, -0.0110, 0.5969, -0.1544, -0.6129], device='cuda:0'), grad: tensor([ 8.8167e-04, 2.6435e-05, 5.7757e-05, 4.0114e-05, -1.0347e-03, 1.0364e-05, 1.7658e-05], device='cuda:0') 249 0.005892784473993186 changing lr epoch 31, time 347.30, cls_loss 0.0762 cls_loss_mapping 0.2870 cls_loss_causal 0.9118 re_mapping 0.0449 re_causal 0.0952 /// teacc 86.55 lr 0.00567117 Epoch 33, weight, value: tensor([[ 0.0234, 0.0290, -0.0348, ..., -0.0716, -0.1188, -0.0907], [-0.0386, -0.0408, -0.0268, ..., -0.0260, -0.0243, -0.0082], [ 0.0210, -0.0213, 0.0404, ..., -0.0924, -0.0888, -0.0774], ..., [-0.0047, -0.0339, 0.0466, ..., -0.0284, -0.0371, -0.0344], [-0.0111, -0.0246, -0.0016, ..., -0.0565, 0.0029, -0.0583], [ 0.0698, 0.2048, 0.0199, ..., 0.0998, 0.0478, 0.1215]], device='cuda:0'), grad: tensor([[ 1.2469e-04, 3.1799e-05, 1.4935e-03, ..., 7.4530e-04, 1.5926e-03, 7.6580e-04], [ 1.1921e-04, 3.4094e-05, 1.4572e-03, ..., 7.0524e-04, 1.5268e-03, 7.2336e-04], [ 1.2279e-04, 3.2246e-05, 1.4772e-03, ..., 7.3195e-04, 1.5697e-03, 7.5197e-04], ..., [-4.2677e-04, -1.1748e-04, -5.1804e-03, ..., -2.5349e-03, -5.4626e-03, -2.6016e-03], [ 2.8372e-05, 9.4175e-06, 3.5691e-04, ..., 1.6570e-04, 3.6502e-04, 1.6952e-04], [ 9.0078e-06, 2.6394e-06, 1.1051e-04, ..., 5.3346e-05, 1.1563e-04, 5.4717e-05]], device='cuda:0') Epoch 33, bias, value: tensor([ 0.3002, -0.1302, 0.0179, 0.0061, 0.5638, -0.1528, -0.6115], device='cuda:0'), grad: tensor([ 0.0169, 0.0172, 0.0169, 0.0035, -0.0601, 0.0044, 0.0013], device='cuda:0') 249 0.00567116632908828 changing lr epoch 32, time 344.88, cls_loss 0.0540 cls_loss_mapping 0.2908 cls_loss_causal 0.9360 re_mapping 0.0471 re_causal 0.0968 /// teacc 91.23 lr 0.00544820 Epoch 34, weight, value: tensor([[ 0.0228, 0.0284, -0.0369, ..., -0.0712, -0.1194, -0.0903], [-0.0384, -0.0407, -0.0253, ..., -0.0263, -0.0227, -0.0082], [ 0.0209, -0.0210, 0.0396, ..., -0.0922, -0.0895, -0.0773], ..., [-0.0041, -0.0327, 0.0500, ..., -0.0258, -0.0327, -0.0321], [-0.0109, -0.0245, -0.0009, ..., -0.0557, 0.0040, -0.0574], [ 0.0694, 0.2034, 0.0198, ..., 0.0990, 0.0474, 0.1206]], device='cuda:0'), grad: tensor([[-2.0489e-07, -1.2107e-07, -1.1250e-06, ..., -2.6077e-07, -7.8417e-07, -2.5891e-07], [ 2.7940e-08, 2.0489e-08, 1.2480e-07, ..., 3.3528e-08, 8.9407e-08, 3.3528e-08], [-1.8999e-07, -1.6950e-07, -4.6194e-07, ..., -1.8440e-07, -3.6508e-07, -1.8626e-07], ..., [ 3.1851e-07, 2.3097e-07, 1.3150e-06, ..., 3.6135e-07, 9.4995e-07, 3.6322e-07], [ 2.4214e-08, 2.0489e-08, 7.0781e-08, ..., 2.4214e-08, 5.4017e-08, 2.4214e-08], [ 1.6764e-08, 1.4901e-08, 5.5879e-08, ..., 1.8626e-08, 4.2841e-08, 1.8626e-08]], device='cuda:0') Epoch 34, bias, value: tensor([ 0.2845, -0.1277, 0.0171, -0.0108, 0.5882, -0.1503, -0.6074], device='cuda:0'), grad: tensor([-8.8513e-06, 9.9279e-07, -3.8035e-06, 1.6205e-07, 1.0483e-05, 5.7369e-07, 4.5635e-07], device='cuda:0') 249 0.00544819654451717 changing lr epoch 33, time 347.79, cls_loss 0.0434 cls_loss_mapping 0.2676 cls_loss_causal 0.8887 re_mapping 0.0472 re_causal 0.0963 /// teacc 90.06 lr 0.00522432 Epoch 35, weight, value: tensor([[ 0.0221, 0.0274, -0.0383, ..., -0.0715, -0.1193, -0.0896], [-0.0379, -0.0402, -0.0244, ..., -0.0261, -0.0218, -0.0081], [ 0.0206, -0.0210, 0.0386, ..., -0.0917, -0.0898, -0.0771], ..., [-0.0036, -0.0317, 0.0507, ..., -0.0256, -0.0323, -0.0328], [-0.0106, -0.0242, 0.0007, ..., -0.0543, 0.0056, -0.0559], [ 0.0689, 0.2020, 0.0196, ..., 0.0982, 0.0469, 0.1197]], device='cuda:0'), grad: tensor([[ 3.2410e-07, 2.4214e-07, 5.4874e-06, ..., 4.0196e-06, 6.5789e-06, 2.8349e-06], [ 1.8626e-09, 1.8626e-09, 2.6077e-08, ..., 1.6764e-08, 2.7940e-08, 1.1176e-08], [ 3.3528e-08, 2.4214e-08, 5.6624e-07, ..., 4.1351e-07, 6.7614e-07, 2.9057e-07], ..., [-3.6322e-07, -2.7195e-07, -6.1877e-06, ..., -4.5486e-06, -7.4320e-06, -3.2056e-06], [-3.7253e-09, -0.0000e+00, -1.8626e-09, ..., 1.6764e-08, 1.4901e-08, 1.1176e-08], [ 3.7253e-09, 1.8626e-09, 4.4703e-08, ..., 3.1665e-08, 5.2154e-08, 2.2352e-08]], device='cuda:0') Epoch 35, bias, value: tensor([ 0.2687, -0.1285, 0.0204, -0.0102, 0.5901, -0.1419, -0.6049], device='cuda:0'), grad: tensor([ 5.6118e-05, 2.5332e-07, 5.7742e-06, 5.8673e-07, -6.3360e-05, 7.8231e-08, 4.5076e-07], device='cuda:0') 249 0.005224324151752577 changing lr ---------------------saving model at epoch 34---------------------------------------------------- epoch 34, time 349.39, cls_loss 0.0430 cls_loss_mapping 0.2367 cls_loss_causal 0.8582 re_mapping 0.0450 re_causal 0.0985 /// teacc 92.40 lr 0.00500000 Epoch 36, weight, value: tensor([[ 0.0217, 0.0271, -0.0392, ..., -0.0714, -0.1198, -0.0895], [-0.0401, -0.0425, -0.0308, ..., -0.0291, -0.0304, -0.0152], [ 0.0203, -0.0210, 0.0371, ..., -0.0920, -0.0909, -0.0774], ..., [-0.0033, -0.0313, 0.0507, ..., -0.0255, -0.0322, -0.0326], [-0.0104, -0.0240, 0.0017, ..., -0.0535, 0.0067, -0.0551], [ 0.0708, 0.2032, 0.0263, ..., 0.1006, 0.0557, 0.1260]], device='cuda:0'), grad: tensor([[ 2.0303e-07, 3.7253e-09, 3.2932e-06, ..., 1.4752e-06, 3.1497e-06, 9.6299e-07], [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 3.7253e-09, 7.4506e-09, 1.8626e-09], [ 5.5879e-09, -1.6764e-08, 3.4459e-07, ..., 1.5274e-07, 3.3528e-07, 9.6858e-08], ..., [-2.3469e-07, -3.7253e-09, -3.8296e-06, ..., -1.7155e-06, -3.6620e-06, -1.1194e-06], [ 1.8626e-08, 1.4901e-08, 8.7544e-08, ..., 4.0978e-08, 7.8231e-08, 2.9802e-08], [ 3.7253e-09, 0.0000e+00, 3.7253e-08, ..., 1.6764e-08, 3.5390e-08, 1.1176e-08]], device='cuda:0') Epoch 36, bias, value: tensor([ 0.2641, -0.1327, 0.0109, -0.0064, 0.5864, -0.1347, -0.5939], device='cuda:0'), grad: tensor([ 2.8789e-05, 7.4506e-08, 3.0044e-06, 4.7125e-07, -3.3468e-05, 7.7486e-07, 3.1851e-07], device='cuda:0') 249 0.005000000000000003 changing lr epoch 35, time 348.58, cls_loss 0.0310 cls_loss_mapping 0.2276 cls_loss_causal 0.8888 re_mapping 0.0457 re_causal 0.1002 /// teacc 91.23 lr 0.00477568 Epoch 37, weight, value: tensor([[ 0.0220, 0.0273, -0.0368, ..., -0.0704, -0.1163, -0.0880], [-0.0400, -0.0422, -0.0303, ..., -0.0289, -0.0301, -0.0155], [ 0.0201, -0.0209, 0.0377, ..., -0.0911, -0.0898, -0.0768], ..., [-0.0037, -0.0315, 0.0476, ..., -0.0260, -0.0347, -0.0332], [-0.0104, -0.0239, 0.0009, ..., -0.0536, 0.0057, -0.0551], [ 0.0705, 0.2020, 0.0262, ..., 0.1000, 0.0555, 0.1256]], device='cuda:0'), grad: tensor([[ 3.0566e-06, 4.5821e-06, 3.4451e-05, ..., 3.8564e-05, 5.6267e-05, 3.7909e-05], [-8.5682e-08, -3.3528e-08, 6.5193e-08, ..., 5.1409e-07, 5.1036e-07, 4.9546e-07], [ 5.9605e-08, 1.0431e-07, 9.4622e-07, ..., 1.1232e-06, 1.6093e-06, 1.1008e-06], ..., [ 5.9232e-06, 8.9034e-06, 6.6996e-05, ..., 7.5161e-05, 1.0961e-04, 7.3850e-05], [ 3.1758e-06, 2.3935e-06, 1.6183e-05, ..., 1.4700e-05, 2.1026e-05, 1.3165e-05], [ 7.0781e-07, 7.9162e-07, 5.7667e-06, ..., 6.0536e-06, 8.8066e-06, 5.8040e-06]], device='cuda:0') Epoch 37, bias, value: tensor([ 0.2656, -0.1274, 0.0195, -0.0015, 0.5698, -0.1398, -0.5924], device='cuda:0'), grad: tensor([ 3.1233e-04, 7.2457e-07, 8.8885e-06, -1.0853e-03, 6.0797e-04, 1.0645e-04, 4.7445e-05], device='cuda:0') 249 0.004775675848247429 changing lr ---------------------saving model at epoch 36---------------------------------------------------- epoch 36, time 353.79, cls_loss 0.0258 cls_loss_mapping 0.2135 cls_loss_causal 0.8457 re_mapping 0.0440 re_causal 0.0889 /// teacc 92.98 lr 0.00455180 Epoch 38, weight, value: tensor([[ 0.0216, 0.0268, -0.0387, ..., -0.0706, -0.1180, -0.0882], [-0.0392, -0.0414, -0.0269, ..., -0.0278, -0.0259, -0.0145], [ 0.0201, -0.0205, 0.0382, ..., -0.0908, -0.0885, -0.0765], ..., [-0.0037, -0.0314, 0.0478, ..., -0.0250, -0.0340, -0.0322], [-0.0105, -0.0239, -0.0002, ..., -0.0537, 0.0045, -0.0552], [ 0.0700, 0.2008, 0.0257, ..., 0.0992, 0.0547, 0.1247]], device='cuda:0'), grad: tensor([[ 1.7226e-04, 2.0072e-05, 1.4858e-03, ..., 5.2261e-04, 1.4391e-03, 5.3930e-04], [ 3.0667e-05, 3.6322e-06, 2.6464e-04, ..., 9.3102e-05, 2.5630e-04, 9.6083e-05], [ 5.8323e-05, 6.7316e-06, 5.0306e-04, ..., 1.7703e-04, 4.8757e-04, 1.8275e-04], ..., [ 1.9684e-03, 2.2709e-04, 1.6953e-02, ..., 5.9662e-03, 1.6418e-02, 6.1569e-03], [-2.4223e-03, -2.7990e-04, -2.0889e-02, ..., -7.3471e-03, -2.0233e-02, -7.5836e-03], [ 2.8446e-05, 3.3136e-06, 2.4533e-04, ..., 8.6308e-05, 2.3758e-04, 8.9049e-05]], device='cuda:0') Epoch 38, bias, value: tensor([ 0.2553, -0.1169, 0.0185, -0.0010, 0.5754, -0.1466, -0.5910], device='cuda:0'), grad: tensor([ 0.0071, 0.0013, 0.0024, 0.0068, 0.0804, -0.0991, 0.0012], device='cuda:0') 249 0.004551803455482836 changing lr epoch 37, time 345.70, cls_loss 0.0392 cls_loss_mapping 0.2031 cls_loss_causal 0.7827 re_mapping 0.0426 re_causal 0.0847 /// teacc 91.81 lr 0.00432883 Epoch 39, weight, value: tensor([[ 0.0215, 0.0267, -0.0387, ..., -0.0700, -0.1176, -0.0875], [-0.0390, -0.0412, -0.0273, ..., -0.0277, -0.0263, -0.0145], [ 0.0200, -0.0202, 0.0388, ..., -0.0900, -0.0872, -0.0757], ..., [-0.0041, -0.0317, 0.0457, ..., -0.0263, -0.0372, -0.0337], [-0.0103, -0.0238, 0.0006, ..., -0.0531, 0.0052, -0.0546], [ 0.0696, 0.1996, 0.0253, ..., 0.0986, 0.0543, 0.1239]], device='cuda:0'), grad: tensor([[-5.1069e-04, -5.7316e-04, -1.7967e-03, ..., -6.3467e-04, -1.9684e-03, -6.0940e-04], [-2.6338e-06, -2.0098e-06, -2.4334e-05, ..., -3.8091e-06, -2.3484e-05, -4.0159e-06], [ 3.3885e-05, 1.1325e-04, 6.0272e-04, ..., 9.6262e-05, 7.1335e-04, 1.5843e-04], ..., [ 4.6396e-04, 4.4703e-04, 1.1797e-03, ..., 5.2500e-04, 1.2379e-03, 4.4084e-04], [-1.1455e-07, -2.5146e-08, -4.9360e-07, ..., -8.6613e-08, -5.4948e-07, -1.2945e-07], [ 1.5736e-05, 1.5169e-05, 3.9428e-05, ..., 1.7777e-05, 4.1455e-05, 1.4901e-05]], device='cuda:0') Epoch 39, bias, value: tensor([ 0.2524, -0.1184, 0.0241, 0.0039, 0.5624, -0.1418, -0.5888], device='cuda:0'), grad: tensor([-5.0392e-03, -7.8619e-05, 2.1935e-03, 2.6729e-06, 2.8324e-03, -8.0094e-07, 9.3877e-05], device='cuda:0') 249 0.004328833670911726 changing lr ---------------------saving model at epoch 38---------------------------------------------------- epoch 38, time 348.97, cls_loss 0.0203 cls_loss_mapping 0.1966 cls_loss_causal 0.8061 re_mapping 0.0423 re_causal 0.0850 /// teacc 94.74 lr 0.00410722 Epoch 40, weight, value: tensor([[ 0.0215, 0.0266, -0.0386, ..., -0.0695, -0.1169, -0.0868], [-0.0386, -0.0406, -0.0256, ..., -0.0275, -0.0248, -0.0143], [ 0.0196, -0.0203, 0.0365, ..., -0.0902, -0.0895, -0.0764], ..., [-0.0039, -0.0316, 0.0467, ..., -0.0254, -0.0348, -0.0325], [-0.0105, -0.0238, -0.0011, ..., -0.0535, 0.0031, -0.0552], [ 0.0690, 0.1984, 0.0247, ..., 0.0979, 0.0535, 0.1230]], device='cuda:0'), grad: tensor([[-7.0781e-08, -1.9558e-08, -9.7137e-07, ..., -4.9174e-07, -1.0999e-06, -4.7404e-07], [ 1.1176e-08, 9.3132e-10, 9.7789e-08, ..., 2.5146e-08, 1.0058e-07, 2.6077e-08], [ 1.3039e-08, 2.7940e-09, 1.3132e-07, ..., 4.0978e-08, 1.3690e-07, 4.0978e-08], ..., [ 7.5437e-08, 1.6764e-08, 8.1304e-07, ..., 2.9150e-07, 8.6334e-07, 2.8964e-07], [-1.0058e-07, -1.3970e-08, -7.5065e-07, ..., -4.4703e-08, -6.9104e-07, -6.6124e-08], [ 3.4459e-08, 5.5879e-09, 3.0361e-07, ..., 6.6124e-08, 3.0268e-07, 6.9849e-08]], device='cuda:0') Epoch 40, bias, value: tensor([ 0.2518, -0.1152, 0.0227, 0.0056, 0.5624, -0.1456, -0.5879], device='cuda:0'), grad: tensor([-6.8396e-06, 5.5786e-07, 8.0466e-07, 2.2724e-06, 5.1595e-06, -3.6880e-06, 1.7229e-06], device='cuda:0') 249 0.0041072155260068206 changing lr epoch 39, time 348.75, cls_loss 0.0249 cls_loss_mapping 0.1840 cls_loss_causal 0.8273 re_mapping 0.0398 re_causal 0.0809 /// teacc 90.64 lr 0.00388740 Epoch 41, weight, value: tensor([[ 0.0218, 0.0268, -0.0371, ..., -0.0690, -0.1148, -0.0860], [-0.0383, -0.0403, -0.0250, ..., -0.0273, -0.0241, -0.0142], [ 0.0195, -0.0203, 0.0359, ..., -0.0899, -0.0896, -0.0763], ..., [-0.0044, -0.0319, 0.0448, ..., -0.0254, -0.0368, -0.0328], [-0.0102, -0.0233, 0.0007, ..., -0.0526, 0.0053, -0.0540], [ 0.0687, 0.1974, 0.0243, ..., 0.0973, 0.0529, 0.1223]], device='cuda:0'), grad: tensor([[2.2352e-06, 6.1840e-07, 1.8418e-05, ..., 1.3851e-05, 2.5854e-05, 1.2666e-05], [2.2817e-07, 7.8231e-08, 1.3914e-06, ..., 8.5030e-07, 1.7053e-06, 8.2236e-07], [1.6764e-05, 5.5805e-06, 7.7784e-05, ..., 3.8773e-05, 8.3685e-05, 3.9458e-05], ..., [8.1360e-06, 2.7195e-06, 3.8117e-05, ..., 1.9133e-05, 4.1217e-05, 1.9446e-05], [4.0419e-06, 1.3923e-06, 2.1443e-05, ..., 1.1943e-05, 2.4766e-05, 1.1817e-05], [3.3919e-06, 1.2238e-06, 1.7583e-05, ..., 9.4399e-06, 1.9893e-05, 9.4622e-06]], device='cuda:0') Epoch 41, bias, value: tensor([ 2.4937e-01, -1.1424e-01, 2.2541e-02, 4.9674e-04, 5.6182e-01, -1.4016e-01, -5.8597e-01], device='cuda:0'), grad: tensor([ 1.0884e-04, 8.0839e-06, 3.7861e-04, -8.8835e-04, 1.8752e-04, 1.1289e-04, 9.2149e-05], device='cuda:0') 249 0.0038873953302184317 changing lr epoch 40, time 347.58, cls_loss 0.0184 cls_loss_mapping 0.1822 cls_loss_causal 0.8379 re_mapping 0.0405 re_causal 0.0762 /// teacc 93.57 lr 0.00366982 Epoch 42, weight, value: tensor([[ 0.0217, 0.0267, -0.0374, ..., -0.0689, -0.1149, -0.0858], [-0.0381, -0.0401, -0.0244, ..., -0.0270, -0.0234, -0.0139], [ 0.0194, -0.0202, 0.0359, ..., -0.0894, -0.0891, -0.0759], ..., [-0.0035, -0.0310, 0.0476, ..., -0.0241, -0.0331, -0.0314], [-0.0109, -0.0238, -0.0021, ..., -0.0534, 0.0022, -0.0548], [ 0.0683, 0.1964, 0.0239, ..., 0.0967, 0.0524, 0.1216]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, 1.3970e-08, ..., 4.6566e-09, 1.3970e-08, 4.6566e-09], [-1.8626e-09, -2.7940e-09, -2.7008e-08, ..., -8.3819e-09, -2.6077e-08, -8.3819e-09], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, 3.7253e-09, 9.3132e-10], ..., [ 1.2107e-08, 5.5879e-09, 9.8720e-08, ..., 3.0734e-08, 1.0617e-07, 4.3772e-08], [-1.5832e-08, -6.5193e-09, -1.2200e-07, ..., -3.8184e-08, -1.3225e-07, -5.5879e-08], [ 1.8626e-09, 9.3132e-10, 1.7695e-08, ..., 5.5879e-09, 1.9558e-08, 8.3819e-09]], device='cuda:0') Epoch 42, bias, value: tensor([ 2.4822e-01, -1.1041e-01, 2.3572e-02, -2.5686e-04, 5.6161e-01, -1.4442e-01, -5.8441e-01], device='cuda:0'), grad: tensor([ 1.2759e-07, -2.8498e-07, 1.4901e-08, 7.6368e-08, 5.7649e-07, -5.9418e-07, 9.1270e-08], device='cuda:0') 249 0.003669815772166629 changing lr epoch 41, time 349.76, cls_loss 0.0273 cls_loss_mapping 0.1802 cls_loss_causal 0.8285 re_mapping 0.0382 re_causal 0.0830 /// teacc 91.81 lr 0.00345492 Epoch 43, weight, value: tensor([[ 0.0215, 0.0264, -0.0379, ..., -0.0689, -0.1151, -0.0857], [-0.0373, -0.0394, -0.0207, ..., -0.0266, -0.0203, -0.0134], [ 0.0193, -0.0201, 0.0359, ..., -0.0890, -0.0885, -0.0755], ..., [-0.0041, -0.0313, 0.0442, ..., -0.0241, -0.0355, -0.0316], [-0.0105, -0.0235, 0.0002, ..., -0.0525, 0.0048, -0.0537], [ 0.0679, 0.1955, 0.0237, ..., 0.0962, 0.0519, 0.1210]], device='cuda:0'), grad: tensor([[ 1.7416e-07, 8.6613e-08, 3.6266e-06, ..., 8.9873e-07, 3.3863e-06, 7.4785e-07], [-5.5879e-09, -0.0000e+00, -8.6613e-08, ..., -4.6566e-09, -6.0536e-08, -2.7940e-09], [ 3.7253e-09, 9.3132e-10, 1.0710e-07, ..., 2.6077e-08, 1.1269e-07, 3.3528e-08], ..., [-1.6857e-07, -8.7544e-08, -3.5223e-06, ..., -8.9034e-07, -3.2913e-06, -7.2736e-07], [-5.5879e-09, 9.3132e-10, -2.1793e-07, ..., -5.3085e-08, -2.4587e-07, -8.1956e-08], [ 2.7940e-09, 0.0000e+00, 8.1025e-08, ..., 1.9558e-08, 8.7544e-08, 2.7940e-08]], device='cuda:0') Epoch 43, bias, value: tensor([ 0.2411, -0.1017, 0.0264, -0.0056, 0.5517, -0.1354, -0.5827], device='cuda:0'), grad: tensor([ 2.3931e-05, -6.6031e-07, 6.8638e-07, 8.9407e-08, -2.3156e-05, -1.3653e-06, 5.1130e-07], device='cuda:0') 249 0.0034549150281252667 changing lr epoch 42, time 349.43, cls_loss 0.0260 cls_loss_mapping 0.1887 cls_loss_causal 0.7472 re_mapping 0.0344 re_causal 0.0714 /// teacc 92.40 lr 0.00324313 Epoch 44, weight, value: tensor([[ 0.0212, 0.0261, -0.0394, ..., -0.0695, -0.1166, -0.0861], [-0.0372, -0.0394, -0.0214, ..., -0.0270, -0.0215, -0.0139], [ 0.0194, -0.0199, 0.0365, ..., -0.0884, -0.0874, -0.0750], ..., [-0.0035, -0.0308, 0.0474, ..., -0.0223, -0.0309, -0.0298], [-0.0106, -0.0235, -0.0005, ..., -0.0525, 0.0040, -0.0537], [ 0.0676, 0.1946, 0.0233, ..., 0.0957, 0.0514, 0.1203]], device='cuda:0'), grad: tensor([[-1.3959e-04, -9.1136e-05, -1.1940e-03, ..., -1.2386e-04, -1.0395e-03, -1.3590e-04], [ 1.1966e-05, 1.1869e-05, 1.3447e-04, ..., 1.7852e-05, 1.3018e-04, 1.9625e-05], [ 2.1867e-06, 1.9129e-06, 2.2575e-05, ..., 2.8014e-06, 2.1234e-05, 3.0827e-06], ..., [ 1.2386e-04, 7.6175e-05, 1.0223e-03, ..., 1.0151e-04, 8.7500e-04, 1.1128e-04], [ 5.2806e-07, 4.9081e-07, 5.6736e-06, ..., 7.2829e-07, 5.4091e-06, 8.0094e-07], [ 8.0373e-07, 5.4482e-07, 7.0147e-06, ..., 7.4971e-07, 6.1691e-06, 8.2050e-07]], device='cuda:0') Epoch 44, bias, value: tensor([ 2.2749e-01, -1.0423e-01, 3.1431e-02, 1.9333e-04, 5.5837e-01, -1.3793e-01, -5.8138e-01], device='cuda:0'), grad: tensor([-5.3062e-03, 6.4611e-04, 1.0616e-04, 1.3657e-05, 4.4861e-03, 2.6956e-05, 3.1412e-05], device='cuda:0') 249 0.0032431258795932905 changing lr epoch 43, time 347.80, cls_loss 0.0220 cls_loss_mapping 0.1586 cls_loss_causal 0.7720 re_mapping 0.0327 re_causal 0.0697 /// teacc 92.98 lr 0.00303487 Epoch 45, weight, value: tensor([[ 0.0212, 0.0260, -0.0388, ..., -0.0691, -0.1155, -0.0856], [-0.0370, -0.0391, -0.0209, ..., -0.0267, -0.0210, -0.0137], [ 0.0193, -0.0198, 0.0367, ..., -0.0879, -0.0868, -0.0746], ..., [-0.0036, -0.0307, 0.0466, ..., -0.0223, -0.0314, -0.0298], [-0.0106, -0.0234, -0.0010, ..., -0.0529, 0.0031, -0.0540], [ 0.0673, 0.1938, 0.0231, ..., 0.0952, 0.0510, 0.1198]], device='cuda:0'), grad: tensor([[ 4.1910e-08, 6.5193e-09, 2.7753e-07, ..., 1.2107e-08, 2.2911e-07, 4.1910e-08], [ 6.3330e-08, 9.3132e-09, 4.2841e-07, ..., 1.8626e-08, 3.5483e-07, 6.6124e-08], [ 7.0781e-08, 7.4506e-09, 4.8056e-07, ..., 1.6764e-08, 3.9861e-07, 7.2643e-08], ..., [ 1.8440e-07, 2.7008e-08, 1.2359e-06, ..., 5.4017e-08, 1.0226e-06, 1.8999e-07], [-6.0163e-07, -8.5682e-08, -4.0382e-06, ..., -1.6857e-07, -3.3397e-06, -6.1560e-07], [ 1.5274e-07, 2.2352e-08, 1.0235e-06, ..., 4.5635e-08, 8.4843e-07, 1.5832e-07]], device='cuda:0') Epoch 45, bias, value: tensor([ 0.2288, -0.1013, 0.0338, 0.0020, 0.5523, -0.1420, -0.5797], device='cuda:0'), grad: tensor([ 1.2945e-06, 2.0079e-06, 2.1942e-06, 2.7083e-06, 5.8040e-06, -1.8835e-05, 4.7907e-06], device='cuda:0') 249 0.0030348748417303863 changing lr epoch 44, time 348.82, cls_loss 0.0162 cls_loss_mapping 0.1627 cls_loss_causal 0.7833 re_mapping 0.0324 re_causal 0.0694 /// teacc 92.40 lr 0.00283058 Epoch 46, weight, value: tensor([[ 0.0214, 0.0261, -0.0369, ..., -0.0681, -0.1128, -0.0845], [-0.0369, -0.0390, -0.0210, ..., -0.0267, -0.0211, -0.0138], [ 0.0192, -0.0197, 0.0366, ..., -0.0876, -0.0864, -0.0743], ..., [-0.0037, -0.0307, 0.0451, ..., -0.0229, -0.0332, -0.0304], [-0.0105, -0.0233, -0.0010, ..., -0.0527, 0.0031, -0.0538], [ 0.0670, 0.1930, 0.0228, ..., 0.0948, 0.0506, 0.1193]], device='cuda:0'), grad: tensor([[-4.6976e-06, -1.3476e-06, -2.8268e-05, ..., -5.4240e-06, -2.3216e-05, -6.0126e-06], [ 9.7230e-07, 2.8871e-07, 5.8413e-06, ..., 1.1381e-06, 4.8243e-06, 1.2657e-06], [ 1.6019e-07, 6.9384e-08, 9.8627e-07, ..., 2.4773e-07, 8.8522e-07, 2.7753e-07], ..., [ 3.2540e-06, 8.7917e-07, 1.9431e-05, ..., 3.5278e-06, 1.5780e-05, 3.9414e-06], [-8.3353e-08, -5.3085e-08, -3.2783e-07, ..., -4.0978e-08, -3.4273e-07, -1.0384e-07], [ 3.0082e-07, 1.1595e-07, 1.7677e-06, ..., 3.8277e-07, 1.5311e-06, 4.4005e-07]], device='cuda:0') Epoch 46, bias, value: tensor([ 0.2330, -0.1023, 0.0352, 0.0026, 0.5458, -0.1421, -0.5782], device='cuda:0'), grad: tensor([-1.0890e-04, 2.2680e-05, 4.4368e-06, 2.9523e-06, 7.2658e-05, -9.5367e-07, 7.2941e-06], device='cuda:0') 249 0.0028305813044122124 changing lr epoch 45, time 347.20, cls_loss 0.0141 cls_loss_mapping 0.1489 cls_loss_causal 0.7867 re_mapping 0.0319 re_causal 0.0649 /// teacc 91.81 lr 0.00263066 Epoch 47, weight, value: tensor([[ 0.0213, 0.0260, -0.0365, ..., -0.0677, -0.1122, -0.0841], [-0.0368, -0.0389, -0.0212, ..., -0.0267, -0.0213, -0.0138], [ 0.0191, -0.0197, 0.0361, ..., -0.0874, -0.0865, -0.0742], ..., [-0.0037, -0.0305, 0.0450, ..., -0.0228, -0.0331, -0.0303], [-0.0104, -0.0231, -0.0005, ..., -0.0524, 0.0036, -0.0535], [ 0.0668, 0.1923, 0.0226, ..., 0.0944, 0.0503, 0.1188]], device='cuda:0'), grad: tensor([[ 8.6725e-06, 5.0217e-06, 4.3958e-05, ..., 5.4464e-06, 3.4899e-05, 6.7502e-06], [ 7.4506e-07, 4.2841e-07, 3.7719e-06, ..., 4.8149e-07, 3.0138e-06, 5.9186e-07], [ 1.7090e-07, 9.6858e-08, 8.6566e-07, ..., 1.2107e-07, 7.0408e-07, 1.4529e-07], ..., [-9.7156e-06, -5.6326e-06, -4.9174e-05, ..., -5.9083e-06, -3.8832e-05, -7.3835e-06], [-1.1781e-07, -2.3283e-08, -5.2340e-07, ..., -1.5181e-07, -5.5414e-07, -1.5413e-07], [ 1.7742e-07, 7.4971e-08, 8.8243e-07, ..., 2.3749e-07, 8.6799e-07, 2.5053e-07]], device='cuda:0') Epoch 47, bias, value: tensor([ 0.2331, -0.1031, 0.0329, 0.0037, 0.5448, -0.1405, -0.5769], device='cuda:0'), grad: tensor([ 1.4853e-04, 1.2763e-05, 2.9393e-06, 2.2631e-07, -1.6594e-04, -1.7229e-06, 3.0603e-06], device='cuda:0') 249 0.0026306566876350096 changing lr epoch 46, time 351.40, cls_loss 0.0152 cls_loss_mapping 0.1497 cls_loss_causal 0.7486 re_mapping 0.0335 re_causal 0.0688 /// teacc 93.57 lr 0.00243550 Epoch 48, weight, value: tensor([[ 0.0212, 0.0259, -0.0366, ..., -0.0673, -0.1120, -0.0837], [-0.0368, -0.0388, -0.0214, ..., -0.0268, -0.0216, -0.0139], [ 0.0191, -0.0196, 0.0359, ..., -0.0873, -0.0864, -0.0741], ..., [-0.0035, -0.0303, 0.0454, ..., -0.0227, -0.0324, -0.0301], [-0.0103, -0.0231, -0.0005, ..., -0.0522, 0.0037, -0.0533], [ 0.0665, 0.1917, 0.0224, ..., 0.0940, 0.0500, 0.1183]], device='cuda:0'), grad: tensor([[-3.7253e-08, -2.8871e-08, -4.2934e-07, ..., -1.4808e-07, -5.2014e-07, -1.5087e-07], [ 3.4925e-08, 2.1886e-08, 3.6834e-07, ..., 1.2806e-07, 4.2701e-07, 1.3644e-07], [ 6.6590e-08, 2.9802e-08, 4.0792e-07, ..., 1.6857e-07, 4.2701e-07, 1.7509e-07], ..., [-6.9849e-08, -3.3993e-08, -4.3446e-07, ..., -2.0536e-07, -4.4284e-07, -2.0489e-07], [-2.5611e-08, 1.3970e-09, -1.7043e-07, ..., -7.9162e-09, -1.7602e-07, -4.1910e-08], [ 1.8161e-08, 4.6566e-09, 1.5087e-07, ..., 3.3993e-08, 1.6671e-07, 4.8429e-08]], device='cuda:0') Epoch 48, bias, value: tensor([ 0.2353, -0.1035, 0.0320, 0.0043, 0.5411, -0.1397, -0.5755], device='cuda:0'), grad: tensor([-2.6878e-06, 2.3711e-06, 2.2128e-06, 5.7975e-07, -2.6841e-06, -5.8161e-07, 7.8138e-07], device='cuda:0') 249 0.0024355036129704724 changing lr epoch 47, time 348.26, cls_loss 0.0167 cls_loss_mapping 0.1378 cls_loss_causal 0.7175 re_mapping 0.0324 re_causal 0.0644 /// teacc 92.40 lr 0.00224552 Epoch 49, weight, value: tensor([[ 0.0212, 0.0258, -0.0364, ..., -0.0671, -0.1116, -0.0834], [-0.0367, -0.0387, -0.0215, ..., -0.0267, -0.0217, -0.0139], [ 0.0190, -0.0195, 0.0362, ..., -0.0869, -0.0856, -0.0737], ..., [-0.0036, -0.0302, 0.0452, ..., -0.0227, -0.0324, -0.0301], [-0.0103, -0.0230, -0.0005, ..., -0.0521, 0.0036, -0.0532], [ 0.0663, 0.1911, 0.0223, ..., 0.0937, 0.0498, 0.1180]], device='cuda:0'), grad: tensor([[ 1.5041e-07, 1.8626e-08, 2.8182e-06, ..., 2.5146e-07, 2.3022e-06, 2.9523e-07], [-4.1649e-06, -3.5809e-07, -8.1778e-05, ..., -5.4538e-06, -6.5923e-05, -7.3649e-06], [-3.5390e-07, -2.3004e-07, -1.1837e-06, ..., -6.8778e-07, -1.3318e-06, -6.9663e-07], ..., [ 4.3176e-06, 5.4855e-07, 7.9632e-05, ..., 5.8115e-06, 6.4492e-05, 7.6815e-06], [ 1.8626e-08, 1.0710e-08, 1.4761e-07, ..., 3.2596e-08, 1.3225e-07, 3.3993e-08], [ 1.1176e-08, 6.0536e-09, 1.0571e-07, ..., 1.8626e-08, 9.1735e-08, 2.0023e-08]], device='cuda:0') Epoch 49, bias, value: tensor([ 0.2346, -0.1037, 0.0338, 0.0039, 0.5387, -0.1391, -0.5740], device='cuda:0'), grad: tensor([ 2.5690e-05, -7.4625e-04, -8.5533e-06, 1.1157e-06, 7.2432e-04, 1.2517e-06, 9.1363e-07], device='cuda:0') 249 0.00224551509273949 changing lr epoch 48, time 349.54, cls_loss 0.0157 cls_loss_mapping 0.1420 cls_loss_causal 0.6908 re_mapping 0.0317 re_causal 0.0606 /// teacc 92.98 lr 0.00206107 Epoch 50, weight, value: tensor([[ 0.0211, 0.0257, -0.0366, ..., -0.0671, -0.1117, -0.0833], [-0.0365, -0.0386, -0.0207, ..., -0.0266, -0.0210, -0.0138], [ 0.0189, -0.0195, 0.0356, ..., -0.0868, -0.0859, -0.0738], ..., [-0.0037, -0.0303, 0.0446, ..., -0.0229, -0.0328, -0.0302], [-0.0102, -0.0229, -0.0003, ..., -0.0517, 0.0038, -0.0527], [ 0.0661, 0.1906, 0.0221, ..., 0.0934, 0.0496, 0.1176]], device='cuda:0'), grad: tensor([[ 9.1374e-05, 2.3050e-07, 1.6365e-03, ..., 1.0624e-03, 2.1915e-03, 1.0624e-03], [ 1.9059e-05, 2.4680e-08, 3.4165e-04, ..., 2.2185e-04, 4.5753e-04, 2.2185e-04], [ 1.7571e-04, 4.6566e-09, 3.1509e-03, ..., 2.0466e-03, 4.2229e-03, 2.0466e-03], ..., [ 3.8236e-05, -2.6356e-07, 6.8855e-04, ..., 4.4870e-04, 9.2459e-04, 4.4870e-04], [ 2.2948e-05, 0.0000e+00, 4.1151e-04, ..., 2.6751e-04, 5.5122e-04, 2.6751e-04], [ 1.2010e-05, 2.3283e-09, 2.1529e-04, ..., 1.3983e-04, 2.8849e-04, 1.3983e-04]], device='cuda:0') Epoch 50, bias, value: tensor([ 0.2324, -0.1012, 0.0313, 0.0066, 0.5357, -0.1379, -0.5729], device='cuda:0'), grad: tensor([ 0.0201, 0.0042, 0.0387, -0.0791, 0.0085, 0.0051, 0.0026], device='cuda:0') 249 0.002061073738537637 changing lr epoch 49, time 349.14, cls_loss 0.0116 cls_loss_mapping 0.1371 cls_loss_causal 0.6570 re_mapping 0.0313 re_causal 0.0632 /// teacc 90.06 lr 0.00188255 Epoch 51, weight, value: tensor([[ 0.0212, 0.0257, -0.0357, ..., -0.0667, -0.1105, -0.0829], [-0.0363, -0.0384, -0.0197, ..., -0.0265, -0.0200, -0.0137], [ 0.0188, -0.0195, 0.0353, ..., -0.0867, -0.0859, -0.0736], ..., [-0.0039, -0.0303, 0.0433, ..., -0.0230, -0.0339, -0.0304], [-0.0103, -0.0229, -0.0008, ..., -0.0518, 0.0032, -0.0528], [ 0.0659, 0.1901, 0.0220, ..., 0.0932, 0.0494, 0.1173]], device='cuda:0'), grad: tensor([[ 2.1094e-07, 1.3877e-07, 1.2722e-06, ..., 5.4948e-07, 1.3569e-06, 4.4703e-07], [ 4.6566e-10, -4.6566e-09, -1.8626e-08, ..., 3.8650e-08, 1.1642e-08, 1.9092e-08], [ 5.5879e-09, 2.3283e-09, 3.3993e-08, ..., 2.4680e-08, 4.3306e-08, 1.9558e-08], ..., [-1.2573e-07, -1.0757e-07, -7.4692e-07, ..., -1.7555e-07, -6.9290e-07, -1.4016e-07], [ 4.6082e-06, 1.4938e-06, 2.7388e-05, ..., 2.1830e-05, 3.6180e-05, 1.7256e-05], [ 1.1269e-07, 3.7253e-08, 6.7614e-07, ..., 5.3132e-07, 8.8988e-07, 4.2561e-07]], device='cuda:0') Epoch 51, bias, value: tensor([ 0.2339, -0.0990, 0.0304, 0.0086, 0.5305, -0.1386, -0.5717], device='cuda:0'), grad: tensor([ 5.8785e-06, -1.5087e-07, 1.8068e-07, -1.5676e-04, -2.9448e-06, 1.5032e-04, 3.7197e-06], device='cuda:0') 249 0.0018825509907063344 changing lr epoch 50, time 347.08, cls_loss 0.0089 cls_loss_mapping 0.1230 cls_loss_causal 0.6582 re_mapping 0.0311 re_causal 0.0606 /// teacc 92.98 lr 0.00171031 Epoch 52, weight, value: tensor([[ 0.0211, 0.0257, -0.0356, ..., -0.0665, -0.1102, -0.0827], [-0.0362, -0.0383, -0.0197, ..., -0.0264, -0.0200, -0.0137], [ 0.0188, -0.0195, 0.0351, ..., -0.0865, -0.0858, -0.0735], ..., [-0.0038, -0.0302, 0.0433, ..., -0.0230, -0.0337, -0.0303], [-0.0103, -0.0229, -0.0008, ..., -0.0516, 0.0032, -0.0526], [ 0.0658, 0.1896, 0.0219, ..., 0.0929, 0.0492, 0.1170]], device='cuda:0'), grad: tensor([[ 5.9232e-06, 1.7181e-05, 1.1671e-04, ..., 4.6760e-05, 1.3542e-04, 5.2303e-05], [ 3.1404e-06, 8.9407e-06, 6.0827e-05, ..., 2.4393e-05, 7.0512e-05, 2.7254e-05], [-1.4760e-05, -4.3690e-05, -2.9564e-04, ..., -1.1837e-04, -3.4332e-04, -1.3256e-04], ..., [ 2.0899e-06, 5.6662e-06, 3.8713e-05, ..., 1.5587e-05, 4.4852e-05, 1.7360e-05], [ 3.9935e-06, 8.7619e-06, 6.1870e-05, ..., 2.5079e-05, 7.0989e-05, 2.7627e-05], [ 1.7686e-06, 3.5763e-06, 2.5600e-05, ..., 1.0423e-05, 2.9281e-05, 1.1422e-05]], device='cuda:0') Epoch 52, bias, value: tensor([ 0.2331, -0.0991, 0.0303, 0.0088, 0.5297, -0.1382, -0.5705], device='cuda:0'), grad: tensor([ 9.2030e-04, 4.7851e-04, -2.3384e-03, -1.7226e-05, 3.0231e-04, 4.6611e-04, 1.9002e-04], device='cuda:0') 249 0.0017103063703014388 changing lr epoch 51, time 350.62, cls_loss 0.0133 cls_loss_mapping 0.1291 cls_loss_causal 0.6928 re_mapping 0.0306 re_causal 0.0583 /// teacc 92.40 lr 0.00154469 Epoch 53, weight, value: tensor([[ 0.0209, 0.0255, -0.0364, ..., -0.0666, -0.1109, -0.0828], [-0.0361, -0.0383, -0.0199, ..., -0.0264, -0.0202, -0.0137], [ 0.0189, -0.0192, 0.0361, ..., -0.0861, -0.0845, -0.0731], ..., [-0.0038, -0.0301, 0.0433, ..., -0.0228, -0.0335, -0.0301], [-0.0102, -0.0228, -0.0007, ..., -0.0515, 0.0033, -0.0525], [ 0.0656, 0.1892, 0.0218, ..., 0.0927, 0.0490, 0.1167]], device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.8626e-09, 1.2247e-07, ..., 1.2107e-08, 9.7323e-08, 9.7789e-09], [-1.0245e-08, -2.7940e-09, -2.2631e-07, ..., -2.4680e-08, -1.8207e-07, -1.9558e-08], [ 9.3132e-10, 4.6566e-10, 1.0710e-08, ..., 9.3132e-10, 8.8476e-09, 1.3970e-09], ..., [ 1.4901e-08, 4.6566e-09, 2.0303e-07, ..., 1.9092e-08, 1.6438e-07, 2.3283e-08], [-1.4435e-08, -4.6566e-09, -1.4110e-07, ..., -9.7789e-09, -1.1316e-07, -1.9092e-08], [ 2.3283e-09, 9.3132e-10, 2.3749e-08, ..., 1.8626e-09, 1.9092e-08, 3.2596e-09]], device='cuda:0') Epoch 53, bias, value: tensor([ 0.2295, -0.0996, 0.0344, 0.0079, 0.5294, -0.1378, -0.5696], device='cuda:0'), grad: tensor([ 8.1165e-07, -1.5292e-06, 4.8894e-08, 4.0513e-08, 1.1073e-06, -5.8394e-07, 1.0803e-07], device='cuda:0') 249 0.0015446867550656784 changing lr epoch 52, time 348.17, cls_loss 0.0145 cls_loss_mapping 0.1320 cls_loss_causal 0.7023 re_mapping 0.0281 re_causal 0.0541 /// teacc 91.81 lr 0.00138603 Epoch 54, weight, value: tensor([[ 0.0208, 0.0253, -0.0370, ..., -0.0667, -0.1116, -0.0828], [-0.0361, -0.0382, -0.0199, ..., -0.0264, -0.0202, -0.0137], [ 0.0189, -0.0191, 0.0367, ..., -0.0856, -0.0834, -0.0727], ..., [-0.0038, -0.0300, 0.0433, ..., -0.0229, -0.0335, -0.0302], [-0.0102, -0.0228, -0.0007, ..., -0.0514, 0.0033, -0.0524], [ 0.0655, 0.1888, 0.0217, ..., 0.0925, 0.0489, 0.1165]], device='cuda:0'), grad: tensor([[ 3.2522e-06, 1.6252e-06, 1.9357e-05, ..., 1.2005e-06, 1.5438e-05, 1.0198e-06], [ 4.3004e-07, 1.4878e-07, 2.4345e-06, ..., 1.5367e-07, 1.9297e-06, 1.2107e-07], [ 3.3993e-07, 1.6950e-07, 2.0228e-06, ..., 1.2713e-07, 1.6158e-06, 1.0803e-07], ..., [-4.1947e-06, -2.0079e-06, -2.4796e-05, ..., -1.5376e-06, -1.9774e-05, -1.2927e-06], [ 8.4285e-08, 2.9569e-08, 4.8243e-07, ..., 5.0990e-08, 4.0536e-07, 4.4703e-08], [ 3.1665e-08, 1.3039e-08, 1.8463e-07, ..., 1.7928e-08, 1.5344e-07, 1.5832e-08]], device='cuda:0') Epoch 54, bias, value: tensor([ 0.2269, -0.0993, 0.0366, 0.0082, 0.5280, -0.1374, -0.5688], device='cuda:0'), grad: tensor([ 5.9217e-05, 7.1488e-06, 6.1877e-06, 7.3621e-07, -7.5459e-05, 1.4594e-06, 5.6159e-07], device='cuda:0') 249 0.001386025680863044 changing lr epoch 53, time 348.79, cls_loss 0.0133 cls_loss_mapping 0.1284 cls_loss_causal 0.7003 re_mapping 0.0296 re_causal 0.0593 /// teacc 91.81 lr 0.00123464 Epoch 55, weight, value: tensor([[ 0.0208, 0.0252, -0.0373, ..., -0.0668, -0.1118, -0.0828], [-0.0360, -0.0381, -0.0197, ..., -0.0263, -0.0200, -0.0136], [ 0.0189, -0.0190, 0.0370, ..., -0.0854, -0.0829, -0.0725], ..., [-0.0038, -0.0300, 0.0432, ..., -0.0228, -0.0334, -0.0301], [-0.0102, -0.0228, -0.0010, ..., -0.0515, 0.0028, -0.0524], [ 0.0654, 0.1885, 0.0217, ..., 0.0923, 0.0488, 0.1162]], device='cuda:0'), grad: tensor([[-3.7253e-09, -3.4925e-09, -1.8626e-08, ..., -5.1223e-09, -2.2585e-08, -5.1223e-09], [-7.6834e-09, -4.6566e-09, -5.3551e-08, ..., 2.3283e-09, -5.0291e-08, -3.0268e-09], [ 5.1688e-08, 2.7008e-08, 2.6380e-07, ..., 1.2666e-07, 3.8091e-07, 1.4249e-07], ..., [ 1.4435e-08, 8.8476e-09, 9.0804e-08, ..., 8.8476e-09, 9.7556e-08, 1.7462e-08], [-1.3737e-08, -6.9849e-09, -8.3586e-08, ..., -1.1409e-08, -9.4064e-08, -2.3516e-08], [ 1.1642e-08, 6.5193e-09, 6.9384e-08, ..., 2.3516e-08, 9.2434e-08, 3.0734e-08]], device='cuda:0') Epoch 55, bias, value: tensor([ 0.2246, -0.0989, 0.0374, 0.0097, 0.5275, -0.1382, -0.5680], device='cuda:0'), grad: tensor([-7.2876e-08, -1.2759e-07, 6.0722e-07, -5.7183e-07, 2.2934e-07, -1.8883e-07, 1.6927e-07], device='cuda:0') 249 0.0012346426699819469 changing lr epoch 54, time 350.60, cls_loss 0.0113 cls_loss_mapping 0.1171 cls_loss_causal 0.6751 re_mapping 0.0298 re_causal 0.0561 /// teacc 92.40 lr 0.00109084 Epoch 56, weight, value: tensor([[ 0.0207, 0.0252, -0.0373, ..., -0.0667, -0.1117, -0.0827], [-0.0360, -0.0381, -0.0198, ..., -0.0263, -0.0201, -0.0137], [ 0.0189, -0.0189, 0.0372, ..., -0.0851, -0.0823, -0.0722], ..., [-0.0037, -0.0299, 0.0433, ..., -0.0228, -0.0333, -0.0301], [-0.0102, -0.0228, -0.0010, ..., -0.0514, 0.0028, -0.0524], [ 0.0653, 0.1882, 0.0216, ..., 0.0922, 0.0486, 0.1161]], device='cuda:0'), grad: tensor([[-1.0012e-08, -6.0536e-09, -3.8184e-08, ..., -4.1910e-09, -2.0023e-08, -4.8894e-09], [-1.8626e-09, 0.0000e+00, -2.3516e-08, ..., 0.0000e+00, -1.7462e-08, -4.6566e-10], [ 4.8894e-09, 2.7940e-09, 1.8394e-08, ..., 2.0955e-09, 9.7789e-09, 2.3283e-09], ..., [ 4.6566e-09, 1.8626e-09, 3.2596e-08, ..., 1.3970e-09, 2.1420e-08, 1.8626e-09], [ 4.6566e-10, 2.3283e-10, 1.8626e-09, ..., 2.3283e-10, 1.1642e-09, 4.6566e-10], [ 1.3970e-09, 9.3132e-10, 5.5879e-09, ..., 6.9849e-10, 3.0268e-09, 6.9849e-10]], device='cuda:0') Epoch 56, bias, value: tensor([ 0.2242, -0.0992, 0.0382, 0.0095, 0.5265, -0.1378, -0.5674], device='cuda:0'), grad: tensor([-3.2573e-07, -1.0384e-07, 1.5576e-07, 3.0268e-09, 1.9046e-07, 1.3504e-08, 4.7730e-08], device='cuda:0') 249 0.0010908425876598518 changing lr epoch 55, time 347.23, cls_loss 0.0155 cls_loss_mapping 0.1176 cls_loss_causal 0.7039 re_mapping 0.0292 re_causal 0.0602 /// teacc 91.81 lr 0.00095492 Epoch 57, weight, value: tensor([[ 0.0207, 0.0251, -0.0374, ..., -0.0666, -0.1116, -0.0826], [-0.0359, -0.0380, -0.0198, ..., -0.0263, -0.0201, -0.0137], [ 0.0188, -0.0189, 0.0371, ..., -0.0850, -0.0822, -0.0722], ..., [-0.0037, -0.0298, 0.0435, ..., -0.0227, -0.0330, -0.0300], [-0.0106, -0.0229, -0.0022, ..., -0.0516, 0.0016, -0.0526], [ 0.0652, 0.1880, 0.0215, ..., 0.0920, 0.0486, 0.1159]], device='cuda:0'), grad: tensor([[-3.4750e-05, -3.1918e-05, -1.2326e-04, ..., -5.0396e-05, -1.2255e-04, -5.0694e-05], [ 6.5193e-08, 1.5763e-07, -2.6054e-07, ..., 1.1874e-07, -1.5344e-07, 9.9652e-08], [ 3.3677e-06, 3.0696e-06, 1.2785e-05, ..., 5.1595e-06, 1.2890e-05, 5.2527e-06], ..., [ 2.9519e-05, 2.7031e-05, 1.0395e-04, ..., 4.2289e-05, 1.0300e-04, 4.2528e-05], [ 1.1106e-07, 1.3853e-07, -4.8429e-08, ..., 1.5693e-07, -1.3364e-07, -2.3283e-10], [ 1.0636e-06, 9.4343e-07, 4.4405e-06, ..., 1.7015e-06, 4.5374e-06, 1.8021e-06]], device='cuda:0') Epoch 57, bias, value: tensor([ 0.2236, -0.0991, 0.0380, 0.0114, 0.5264, -0.1394, -0.5667], device='cuda:0'), grad: tensor([-7.5626e-04, -9.3132e-10, 7.8321e-05, 1.4409e-05, 6.3515e-04, 1.3635e-06, 2.6494e-05], device='cuda:0') 249 0.000954915028125264 changing lr epoch 56, time 347.41, cls_loss 0.0119 cls_loss_mapping 0.1096 cls_loss_causal 0.6021 re_mapping 0.0282 re_causal 0.0535 /// teacc 92.98 lr 0.00082713 Epoch 58, weight, value: tensor([[ 0.0207, 0.0251, -0.0369, ..., -0.0663, -0.1110, -0.0823], [-0.0359, -0.0380, -0.0198, ..., -0.0263, -0.0202, -0.0137], [ 0.0188, -0.0189, 0.0372, ..., -0.0849, -0.0819, -0.0721], ..., [-0.0037, -0.0298, 0.0430, ..., -0.0228, -0.0334, -0.0301], [-0.0105, -0.0229, -0.0022, ..., -0.0515, 0.0016, -0.0525], [ 0.0651, 0.1877, 0.0215, ..., 0.0919, 0.0484, 0.1157]], device='cuda:0'), grad: tensor([[ 6.9849e-10, 6.9849e-10, 2.5611e-09, ..., 2.3283e-10, 2.0955e-09, 2.3283e-10], [ 2.0955e-09, 2.3283e-09, 7.6834e-09, ..., 6.9849e-10, 6.5193e-09, 9.3132e-10], [-1.6531e-08, -1.6997e-08, -5.6112e-08, ..., -4.6566e-09, -4.7032e-08, -6.5193e-09], ..., [ 1.8626e-09, 1.8626e-09, 6.5193e-09, ..., 6.9849e-10, 5.5879e-09, 9.3132e-10], [ 7.2177e-09, 7.4506e-09, 2.4680e-08, ..., 2.0955e-09, 2.0722e-08, 3.0268e-09], [ 6.9849e-10, 6.9849e-10, 2.7940e-09, ..., 2.3283e-10, 2.5611e-09, 4.6566e-10]], device='cuda:0') Epoch 58, bias, value: tensor([ 0.2244, -0.0993, 0.0384, 0.0119, 0.5244, -0.1393, -0.5663], device='cuda:0'), grad: tensor([ 5.3551e-09, 1.6298e-08, -1.1828e-07, 3.1665e-08, 1.4901e-08, 5.2154e-08, 6.7521e-09], device='cuda:0') 249 0.0008271337313934874 changing lr epoch 57, time 376.49, cls_loss 0.0116 cls_loss_mapping 0.1136 cls_loss_causal 0.6296 re_mapping 0.0275 re_causal 0.0539 /// teacc 92.40 lr 0.00070776 Epoch 59, weight, value: tensor([[ 0.0208, 0.0253, -0.0363, ..., -0.0659, -0.1101, -0.0819], [-0.0360, -0.0381, -0.0201, ..., -0.0264, -0.0205, -0.0138], [ 0.0188, -0.0189, 0.0371, ..., -0.0849, -0.0819, -0.0720], ..., [-0.0037, -0.0298, 0.0429, ..., -0.0228, -0.0335, -0.0301], [-0.0105, -0.0228, -0.0022, ..., -0.0515, 0.0016, -0.0524], [ 0.0650, 0.1875, 0.0214, ..., 0.0918, 0.0483, 0.1155]], device='cuda:0'), grad: tensor([[ 2.8219e-07, 3.6857e-07, 2.9244e-06, ..., -5.0757e-07, 2.0470e-06, -4.2375e-07], [-8.3912e-07, -7.2038e-07, -7.1563e-06, ..., -8.2701e-07, -6.3777e-06, -9.5321e-07], [-2.9076e-06, -2.4326e-06, -1.1869e-05, ..., -3.7178e-06, -1.2442e-05, -4.3623e-06], ..., [ 2.9034e-07, 2.2585e-07, 2.2016e-06, ..., 9.3365e-07, 2.5388e-06, 9.7509e-07], [ 2.8107e-06, 2.2501e-06, 1.1228e-05, ..., 2.5090e-06, 1.0677e-05, 3.0845e-06], [ 1.0035e-07, 6.4261e-08, 6.9477e-07, ..., 3.0361e-07, 7.8604e-07, 3.1339e-07]], device='cuda:0') Epoch 59, bias, value: tensor([ 0.2261, -0.1007, 0.0381, 0.0114, 0.5241, -0.1388, -0.5660], device='cuda:0'), grad: tensor([ 1.1176e-05, -3.3289e-05, -3.5226e-05, 9.2983e-06, 1.0848e-05, 3.3617e-05, 3.5930e-06], device='cuda:0') 249 0.00070775603199067 changing lr epoch 58, time 348.35, cls_loss 0.0117 cls_loss_mapping 0.1144 cls_loss_causal 0.6328 re_mapping 0.0285 re_causal 0.0533 /// teacc 92.40 lr 0.00059702 Epoch 60, weight, value: tensor([[ 0.0208, 0.0253, -0.0363, ..., -0.0659, -0.1100, -0.0818], [-0.0360, -0.0380, -0.0203, ..., -0.0264, -0.0206, -0.0138], [ 0.0187, -0.0189, 0.0369, ..., -0.0849, -0.0820, -0.0720], ..., [-0.0037, -0.0298, 0.0431, ..., -0.0228, -0.0332, -0.0300], [-0.0106, -0.0228, -0.0023, ..., -0.0515, 0.0015, -0.0524], [ 0.0650, 0.1874, 0.0213, ..., 0.0917, 0.0482, 0.1154]], device='cuda:0'), grad: tensor([[ 1.0012e-08, 7.2177e-09, 2.1653e-08, ..., 3.2596e-09, 1.8394e-08, 4.8894e-09], [ 6.9849e-10, 0.0000e+00, 1.1874e-08, ..., 2.7940e-09, 1.2806e-08, 3.7253e-09], [ 1.8626e-08, 0.0000e+00, 2.6240e-07, ..., 9.2434e-08, 3.1875e-07, 1.0966e-07], ..., [ 3.0268e-09, 0.0000e+00, 5.2387e-08, ..., 1.4668e-08, 5.8906e-08, 1.8394e-08], [-2.4447e-08, 0.0000e+00, -3.6880e-07, ..., -1.1991e-07, -4.3539e-07, -1.4459e-07], [-9.3132e-09, -7.9162e-09, 3.7253e-09, ..., 3.9581e-09, 1.1176e-08, 4.4238e-09]], device='cuda:0') Epoch 60, bias, value: tensor([ 0.2262, -0.1010, 0.0375, 0.0116, 0.5244, -0.1389, -0.5656], device='cuda:0'), grad: tensor([ 5.0524e-08, 7.7067e-08, 1.6429e-06, 9.4064e-08, 3.3434e-07, -2.3246e-06, 1.1711e-07], device='cuda:0') 249 0.0005970223407163104 changing lr epoch 59, time 345.73, cls_loss 0.0119 cls_loss_mapping 0.1103 cls_loss_causal 0.6765 re_mapping 0.0283 re_causal 0.0535 /// teacc 91.81 lr 0.00049516 Epoch 61, weight, value: tensor([[ 0.0208, 0.0252, -0.0365, ..., -0.0659, -0.1102, -0.0818], [-0.0359, -0.0380, -0.0201, ..., -0.0263, -0.0205, -0.0138], [ 0.0187, -0.0189, 0.0370, ..., -0.0848, -0.0819, -0.0719], ..., [-0.0036, -0.0297, 0.0436, ..., -0.0226, -0.0326, -0.0298], [-0.0105, -0.0228, -0.0023, ..., -0.0515, 0.0014, -0.0524], [ 0.0649, 0.1872, 0.0213, ..., 0.0916, 0.0481, 0.1153]], device='cuda:0'), grad: tensor([[ 4.6253e-05, 6.1572e-05, 5.0497e-04, ..., 2.6250e-04, 5.8270e-04, 1.9765e-04], [ 4.1835e-06, 3.6862e-06, 2.5585e-05, ..., 4.8950e-06, 2.0474e-05, 4.6268e-06], [ 2.0102e-05, 1.8865e-05, 1.2565e-04, ..., 3.6746e-05, 1.1319e-04, 3.0190e-05], ..., [-9.2268e-05, -1.0306e-04, -7.8583e-04, ..., -3.2687e-04, -8.1778e-04, -2.5463e-04], [ 1.0476e-05, 8.9481e-06, 6.1095e-05, ..., 9.4324e-06, 4.6521e-05, 9.6112e-06], [ 7.7114e-06, 6.6534e-06, 4.5508e-05, ..., 7.6964e-06, 3.5346e-05, 7.5698e-06]], device='cuda:0') Epoch 61, bias, value: tensor([ 0.2248, -0.1002, 0.0378, 0.0107, 0.5253, -0.1390, -0.5653], device='cuda:0'), grad: tensor([ 0.0040, 0.0001, 0.0008, 0.0001, -0.0057, 0.0003, 0.0003], device='cuda:0') 249 0.0004951556604879052 changing lr epoch 60, time 352.12, cls_loss 0.0114 cls_loss_mapping 0.1141 cls_loss_causal 0.6381 re_mapping 0.0275 re_causal 0.0531 /// teacc 92.98 lr 0.00040236 Epoch 62, weight, value: tensor([[ 0.0208, 0.0252, -0.0365, ..., -0.0659, -0.1102, -0.0818], [-0.0359, -0.0379, -0.0199, ..., -0.0263, -0.0203, -0.0137], [ 0.0187, -0.0188, 0.0369, ..., -0.0848, -0.0819, -0.0719], ..., [-0.0036, -0.0297, 0.0433, ..., -0.0226, -0.0328, -0.0299], [-0.0105, -0.0228, -0.0023, ..., -0.0514, 0.0014, -0.0524], [ 0.0649, 0.1871, 0.0213, ..., 0.0915, 0.0481, 0.1153]], device='cuda:0'), grad: tensor([[ 1.0245e-08, 3.7253e-09, 1.2573e-07, ..., 2.3283e-09, 8.9407e-08, 8.8476e-09], [-1.2107e-08, -4.1910e-09, -1.4901e-07, ..., -1.3970e-09, -1.0338e-07, -8.8476e-09], [ 4.6566e-10, 0.0000e+00, 6.0536e-09, ..., 1.8626e-09, 6.5193e-09, 1.8626e-09], ..., [ 2.7940e-09, 9.3132e-10, 3.4459e-08, ..., 1.8626e-09, 2.6543e-08, 3.7253e-09], [-3.7253e-09, 0.0000e+00, -4.6566e-08, ..., -1.4435e-08, -5.3085e-08, -1.5832e-08], [ 9.3132e-10, 0.0000e+00, 1.4435e-08, ..., 4.1910e-09, 1.6298e-08, 4.6566e-09]], device='cuda:0') Epoch 62, bias, value: tensor([ 0.2244, -0.0993, 0.0373, 0.0113, 0.5244, -0.1389, -0.5651], device='cuda:0'), grad: tensor([ 1.2144e-06, -1.4575e-06, 4.0513e-08, 1.2666e-07, 3.1944e-07, -2.9523e-07, 9.4064e-08], device='cuda:0') 249 0.00040236113724274745 changing lr epoch 61, time 346.83, cls_loss 0.0095 cls_loss_mapping 0.1049 cls_loss_causal 0.6761 re_mapping 0.0273 re_causal 0.0569 /// teacc 92.40 lr 0.00031883 Epoch 63, weight, value: tensor([[ 0.0208, 0.0252, -0.0365, ..., -0.0659, -0.1101, -0.0817], [-0.0359, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], [ 0.0187, -0.0188, 0.0369, ..., -0.0847, -0.0819, -0.0719], ..., [-0.0036, -0.0297, 0.0434, ..., -0.0226, -0.0328, -0.0298], [-0.0105, -0.0228, -0.0023, ..., -0.0514, 0.0014, -0.0523], [ 0.0648, 0.1870, 0.0212, ..., 0.0915, 0.0481, 0.1152]], device='cuda:0'), grad: tensor([[ 3.4785e-07, 5.5879e-08, 3.0063e-06, ..., 2.0433e-06, 4.2580e-06, 1.8263e-06], [ 6.0536e-09, -3.2596e-09, 4.8429e-08, ..., 5.4948e-08, 9.0804e-08, 4.7497e-08], [ 2.3749e-08, 1.3970e-09, 2.2352e-07, ..., 1.5460e-07, 3.2084e-07, 1.3690e-07], ..., [-4.3260e-07, -6.4727e-08, -3.7365e-06, ..., -2.5649e-06, -5.3197e-06, -2.2911e-06], [ 1.6764e-08, 3.2596e-09, 1.4296e-07, ..., 9.6858e-08, 2.0163e-07, 8.6613e-08], [ 1.5832e-08, 3.2596e-09, 1.3271e-07, ..., 8.9407e-08, 1.8626e-07, 8.0094e-08]], device='cuda:0') Epoch 63, bias, value: tensor([ 0.2242, -0.0994, 0.0373, 0.0114, 0.5243, -0.1388, -0.5648], device='cuda:0'), grad: tensor([ 1.7613e-05, 2.6822e-07, 1.3281e-06, 1.0775e-06, -2.1890e-05, 8.3307e-07, 7.7114e-07], device='cuda:0') 249 0.00031882564680131423 changing lr epoch 62, time 350.05, cls_loss 0.0106 cls_loss_mapping 0.1023 cls_loss_causal 0.6294 re_mapping 0.0283 re_causal 0.0536 /// teacc 94.15 lr 0.00024472 Epoch 64, weight, value: tensor([[ 0.0208, 0.0252, -0.0364, ..., -0.0658, -0.1100, -0.0816], [-0.0359, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], [ 0.0187, -0.0188, 0.0369, ..., -0.0847, -0.0818, -0.0719], ..., [-0.0036, -0.0297, 0.0433, ..., -0.0226, -0.0328, -0.0299], [-0.0105, -0.0228, -0.0023, ..., -0.0514, 0.0014, -0.0523], [ 0.0648, 0.1869, 0.0212, ..., 0.0914, 0.0480, 0.1152]], device='cuda:0'), grad: tensor([[ 6.7167e-06, 5.4836e-06, 5.4657e-05, ..., 2.2575e-05, 6.0171e-05, 2.3410e-05], [-4.5478e-05, -3.7491e-05, -3.7265e-04, ..., -1.5616e-04, -4.1270e-04, -1.6129e-04], [-3.0193e-06, -2.0340e-06, -2.1368e-05, ..., -5.6773e-06, -1.9714e-05, -6.5938e-06], ..., [ 3.9160e-05, 3.2187e-05, 3.1996e-04, ..., 1.3328e-04, 3.5357e-04, 1.3781e-04], [ 1.5711e-06, 1.0915e-06, 1.1586e-05, ..., 3.5428e-06, 1.1206e-05, 3.9823e-06], [ 3.3574e-07, 2.4680e-07, 2.5537e-06, ..., 8.7265e-07, 2.5854e-06, 9.4902e-07]], device='cuda:0') Epoch 64, bias, value: tensor([ 0.2244, -0.0992, 0.0373, 0.0111, 0.5240, -0.1387, -0.5647], device='cuda:0'), grad: tensor([ 3.9625e-04, -2.6951e-03, -1.6081e-04, 3.8087e-05, 2.3155e-03, 8.8155e-05, 1.9044e-05], device='cuda:0') 249 0.0002447174185242325 changing lr epoch 63, time 349.59, cls_loss 0.0104 cls_loss_mapping 0.1057 cls_loss_causal 0.6085 re_mapping 0.0285 re_causal 0.0531 /// teacc 92.40 lr 0.00018019 Epoch 65, weight, value: tensor([[ 0.0208, 0.0252, -0.0364, ..., -0.0658, -0.1100, -0.0816], [-0.0358, -0.0379, -0.0199, ..., -0.0262, -0.0202, -0.0137], [ 0.0187, -0.0188, 0.0369, ..., -0.0847, -0.0817, -0.0718], ..., [-0.0036, -0.0297, 0.0432, ..., -0.0226, -0.0329, -0.0299], [-0.0105, -0.0228, -0.0022, ..., -0.0514, 0.0015, -0.0523], [ 0.0648, 0.1869, 0.0212, ..., 0.0914, 0.0480, 0.1151]], device='cuda:0'), grad: tensor([[-2.6748e-06, -2.0582e-06, -1.2316e-05, ..., -4.2096e-06, -1.1683e-05, -3.3285e-06], [ 5.5740e-07, 4.2049e-07, 2.5891e-06, ..., 9.0618e-07, 2.4661e-06, 7.1293e-07], [ 8.3819e-09, 7.4506e-09, 3.5390e-08, ..., 9.3132e-09, 3.1665e-08, 7.4506e-09], ..., [ 2.0694e-06, 1.5972e-06, 9.5293e-06, ..., 3.2485e-06, 9.0301e-06, 2.5686e-06], [ 1.3970e-09, 9.3132e-10, 5.5879e-09, ..., 1.8626e-09, 5.1223e-09, 1.3970e-09], [ 1.6298e-08, 1.5832e-08, 6.5193e-08, ..., 1.3504e-08, 5.7276e-08, 1.2107e-08]], device='cuda:0') Epoch 65, bias, value: tensor([ 0.2243, -0.0992, 0.0374, 0.0111, 0.5236, -0.1383, -0.5645], device='cuda:0'), grad: tensor([-3.7402e-05, 7.9349e-06, 9.6392e-08, 2.9430e-07, 2.8893e-05, 1.6764e-08, 1.6717e-07], device='cuda:0') 249 0.0001801856965207339 changing lr epoch 64, time 344.76, cls_loss 0.0114 cls_loss_mapping 0.1071 cls_loss_causal 0.6610 re_mapping 0.0270 re_causal 0.0537 /// teacc 92.98 lr 0.00012536 Epoch 66, weight, value: tensor([[ 0.0208, 0.0252, -0.0364, ..., -0.0658, -0.1099, -0.0816], [-0.0358, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], [ 0.0187, -0.0188, 0.0369, ..., -0.0846, -0.0817, -0.0718], ..., [-0.0036, -0.0297, 0.0432, ..., -0.0226, -0.0329, -0.0299], [-0.0105, -0.0228, -0.0022, ..., -0.0513, 0.0015, -0.0523], [ 0.0648, 0.1868, 0.0212, ..., 0.0914, 0.0480, 0.1151]], device='cuda:0'), grad: tensor([[-9.0804e-07, -9.2201e-08, -9.3803e-06, ..., -6.5751e-06, -1.2361e-05, -6.6906e-06], [ 1.5413e-07, 3.9581e-08, 1.1614e-06, ..., 7.3295e-07, 1.4473e-06, 7.4739e-07], [-1.1558e-06, -5.9744e-07, -3.2261e-06, ..., -6.3749e-07, -2.5984e-06, -6.8406e-07], ..., [ 5.5227e-07, 1.1967e-07, 4.5411e-06, ..., 2.9691e-06, 5.7630e-06, 3.0249e-06], [ 5.0338e-07, 2.4168e-07, 1.7425e-06, ..., 5.7649e-07, 1.6401e-06, 6.0070e-07], [ 3.6415e-07, 1.2945e-07, 2.0899e-06, ..., 1.1530e-06, 2.4363e-06, 1.1791e-06]], device='cuda:0') Epoch 66, bias, value: tensor([ 0.2242, -0.0993, 0.0375, 0.0111, 0.5233, -0.1382, -0.5644], device='cuda:0'), grad: tensor([-7.1764e-05, 8.9332e-06, -2.5719e-05, 2.3767e-05, 3.4869e-05, 1.3739e-05, 1.6198e-05], device='cuda:0') 249 0.000125360439090882 changing lr epoch 65, time 347.47, cls_loss 0.0074 cls_loss_mapping 0.1005 cls_loss_causal 0.6283 re_mapping 0.0267 re_causal 0.0555 /// teacc 92.98 lr 0.00008035 Epoch 67, weight, value: tensor([[ 0.0208, 0.0252, -0.0363, ..., -0.0658, -0.1099, -0.0816], [-0.0358, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], [ 0.0187, -0.0188, 0.0369, ..., -0.0846, -0.0817, -0.0718], ..., [-0.0036, -0.0297, 0.0432, ..., -0.0226, -0.0329, -0.0299], [-0.0105, -0.0228, -0.0022, ..., -0.0513, 0.0015, -0.0523], [ 0.0648, 0.1868, 0.0212, ..., 0.0914, 0.0480, 0.1151]], device='cuda:0'), grad: tensor([[-8.9873e-08, -8.8476e-08, -5.7323e-07, ..., -3.3528e-08, -5.2247e-07, -5.7742e-08], [-1.3970e-08, -2.0955e-08, -9.2667e-08, ..., -1.9558e-08, -8.4750e-08, -1.8626e-08], [ 3.7253e-09, 2.7940e-09, 2.0489e-08, ..., 1.3970e-09, 1.7695e-08, 2.3283e-09], ..., [ 9.4529e-08, 9.4529e-08, 5.8860e-07, ..., 4.7032e-08, 5.3411e-07, 6.8918e-08], [-1.9092e-08, -2.7940e-09, -6.6590e-08, ..., -2.7940e-09, -4.8894e-08, -8.8476e-09], [ 2.2817e-08, 1.3039e-08, 1.1083e-07, ..., 9.3132e-09, 9.4995e-08, 1.5367e-08]], device='cuda:0') Epoch 67, bias, value: tensor([ 0.2242, -0.0993, 0.0374, 0.0111, 0.5233, -0.1382, -0.5644], device='cuda:0'), grad: tensor([-2.8703e-06, -4.4191e-07, 8.5682e-08, 8.1025e-08, 2.8610e-06, -9.1735e-08, 4.0047e-07], device='cuda:0') 249 8.03520570068517e-05 changing lr epoch 66, time 350.79, cls_loss 0.0102 cls_loss_mapping 0.1035 cls_loss_causal 0.6503 re_mapping 0.0282 re_causal 0.0544 /// teacc 92.98 lr 0.00004525 Epoch 68, weight, value: tensor([[ 0.0208, 0.0252, -0.0363, ..., -0.0657, -0.1098, -0.0816], [-0.0358, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], [ 0.0187, -0.0188, 0.0369, ..., -0.0846, -0.0817, -0.0718], ..., [-0.0036, -0.0297, 0.0432, ..., -0.0226, -0.0330, -0.0299], [-0.0105, -0.0228, -0.0022, ..., -0.0513, 0.0015, -0.0523], [ 0.0647, 0.1868, 0.0212, ..., 0.0914, 0.0480, 0.1151]], device='cuda:0'), grad: tensor([[ 2.3283e-09, -1.6578e-07, 3.4040e-07, ..., 8.8708e-07, 1.9390e-06, 1.2936e-06], [ 5.5656e-06, 2.6505e-06, 3.2008e-05, ..., 6.8583e-06, 3.1292e-05, 9.4473e-06], [ 1.8822e-06, -2.0154e-06, -5.2452e-06, ..., -6.8061e-06, -1.0513e-05, -6.3330e-06], ..., [ 7.3723e-06, 3.7421e-06, 4.2647e-05, ..., 8.7246e-06, 4.0591e-05, 1.1817e-05], [-2.2918e-05, -8.0541e-06, -1.1629e-04, ..., -1.9357e-05, -1.0854e-04, -2.9698e-05], [ 6.1579e-06, 2.8070e-06, 3.4779e-05, ..., 7.3351e-06, 3.3885e-05, 1.0177e-05]], device='cuda:0') Epoch 68, bias, value: tensor([ 0.2243, -0.0993, 0.0374, 0.0110, 0.5232, -0.1381, -0.5643], device='cuda:0'), grad: tensor([ 4.2133e-06, 1.1146e-04, -1.1331e-04, 4.3392e-05, 1.4925e-04, -3.1281e-04, 1.1742e-04], device='cuda:0') 249 4.5251191160326525e-05 changing lr epoch 67, time 348.82, cls_loss 0.0106 cls_loss_mapping 0.1088 cls_loss_causal 0.6864 re_mapping 0.0277 re_causal 0.0531 /// teacc 94.15 lr 0.00002013 Epoch 69, weight, value: tensor([[ 0.0208, 0.0252, -0.0362, ..., -0.0657, -0.1098, -0.0816], [-0.0358, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], [ 0.0187, -0.0188, 0.0369, ..., -0.0846, -0.0817, -0.0718], ..., [-0.0037, -0.0297, 0.0431, ..., -0.0227, -0.0330, -0.0299], [-0.0105, -0.0228, -0.0022, ..., -0.0513, 0.0015, -0.0523], [ 0.0647, 0.1868, 0.0212, ..., 0.0914, 0.0480, 0.1150]], device='cuda:0'), grad: tensor([[-1.3970e-09, -1.8626e-09, -8.8476e-09, ..., -2.3283e-09, -8.3819e-09, -2.3283e-09], [ 2.7940e-09, 1.3970e-09, 9.3132e-09, ..., 1.8626e-09, 7.4506e-09, 2.3283e-09], [ 6.0536e-09, 2.7940e-09, 2.1886e-08, ..., 3.7253e-09, 1.7695e-08, 5.1223e-09], ..., [ 1.9558e-08, 9.7789e-09, 7.5437e-08, ..., 1.3970e-08, 6.1933e-08, 1.7229e-08], [-2.1560e-07, -9.8720e-08, -7.9814e-07, ..., -1.4249e-07, -6.4913e-07, -1.8161e-07], [ 3.6322e-08, 1.6764e-08, 1.3597e-07, ..., 2.4214e-08, 1.1083e-07, 3.1199e-08]], device='cuda:0') Epoch 69, bias, value: tensor([ 0.2244, -0.0993, 0.0374, 0.0110, 0.5230, -0.1381, -0.5643], device='cuda:0'), grad: tensor([-7.5437e-08, 1.5367e-08, 6.5193e-08, 1.6261e-06, 2.5705e-07, -2.2799e-06, 3.9907e-07], device='cuda:0') 249 2.0128530023804673e-05 changing lr epoch 68, time 348.35, cls_loss 0.0081 cls_loss_mapping 0.0988 cls_loss_causal 0.6166 re_mapping 0.0275 re_causal 0.0561 /// teacc 92.98 lr 0.00000503 Epoch 70, weight, value: tensor([[ 0.0208, 0.0252, -0.0362, ..., -0.0657, -0.1098, -0.0816], [-0.0358, -0.0379, -0.0199, ..., -0.0262, -0.0203, -0.0137], [ 0.0187, -0.0188, 0.0369, ..., -0.0846, -0.0817, -0.0718], ..., [-0.0036, -0.0297, 0.0431, ..., -0.0226, -0.0330, -0.0299], [-0.0105, -0.0228, -0.0022, ..., -0.0513, 0.0015, -0.0523], [ 0.0647, 0.1868, 0.0212, ..., 0.0914, 0.0480, 0.1150]], device='cuda:0'), grad: tensor([[ 1.6321e-07, 1.0757e-07, 8.1025e-07, ..., 8.5915e-08, 8.3540e-07, 1.4552e-07], [-1.7928e-07, -8.5915e-08, -8.9034e-07, ..., -8.5915e-08, -9.1083e-07, -1.5227e-07], [ 6.4261e-08, 6.8452e-08, 3.1642e-07, ..., 4.1211e-08, 3.3132e-07, 6.4494e-08], ..., [-6.8918e-08, -9.8720e-08, -3.4203e-07, ..., -4.8662e-08, -3.6345e-07, -7.3109e-08], [ 1.4435e-08, 5.8208e-09, 7.5903e-08, ..., 4.6566e-09, 7.7533e-08, 1.0012e-08], [ 4.1910e-09, 3.0268e-09, 1.9558e-08, ..., 3.0268e-09, 2.0023e-08, 4.4238e-09]], device='cuda:0') Epoch 70, bias, value: tensor([ 0.2244, -0.0993, 0.0374, 0.0110, 0.5230, -0.1381, -0.5643], device='cuda:0'), grad: tensor([ 1.9502e-06, -2.0340e-06, 8.4704e-07, 5.7044e-08, -1.0040e-06, 1.7532e-07, 4.5868e-08], device='cuda:0') 249 5.034667293427056e-06 changing lr epoch 69, time 350.06, cls_loss 0.0124 cls_loss_mapping 0.1044 cls_loss_causal 0.6741 re_mapping 0.0288 re_causal 0.0534 /// teacc 91.81 lr 0.00000000 ---------------------saving last model at epoch 69---------------------------------------------------- /home/yuqian_fu {'gpu': '0', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1', 'source_domain': 'photo', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//photo/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1/photo_16factor_best_test_check.csv', 'factor_num': 16, 'epoch': 'best', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'} -------------------------------------loading pretrain weights---------------------------------- loading weight of best randm: False stride: 5 loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best loading weight of best columns: ['photo', 'art_painting', 'cartoon', 'sketch'] /data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670]) /data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/art_painting_test.hdf5 torch.Size([2048, 3, 227, 227]) torch.Size([2048]) /data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/cartoon_test.hdf5 torch.Size([2344, 3, 227, 227]) torch.Size([2344]) /data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929]) photo art_painting cartoon sketch Avg w/o do (original x) 99.461078 55.761719 49.616041 61.28786 55.555206 photo art_painting cartoon sketch Avg do 98.802395 57.617188 50.0 60.269789 55.962325