Meta-causal/code-withStyleAttack/64947.log

/home/yuqian_fu
here1
here2
{'gpu': '0', 'data': 'sketch', '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//sketch/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/sketch_train.hdf5 torch.Size([3531, 3, 227, 227]) torch.Size([3531])
--------------------------CA_multiple--------------------------
---------------------------16 factors-----------------
randm: True
randn: True
n: 3
randm: False
/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_val.hdf5 torch.Size([398, 3, 227, 227]) torch.Size([398])
-------------------------------------loading pretrain weights----------------------------------
Epoch 1, weight, value: tensor([[-0.0208, -0.0220,  0.0164,  ...,  0.0019,  0.0060,  0.0055],
        [ 0.0100,  0.0072,  0.0083,  ..., -0.0045, -0.0209, -0.0166],
        [ 0.0103,  0.0061,  0.0040,  ..., -0.0083, -0.0102,  0.0187],
        ...,
        [ 0.0074,  0.0004, -0.0085,  ..., -0.0055,  0.0045,  0.0045],
        [ 0.0123, -0.0002, -0.0053,  ...,  0.0055, -0.0196, -0.0166],
        [-0.0206,  0.0139,  0.0016,  ..., -0.0135,  0.0219, -0.0163]],
       device='cuda:0'), grad: None
Epoch 1, bias, value: tensor([-0.0178,  0.0114,  0.0058, -0.0179,  0.0187,  0.0121, -0.0002],
       device='cuda:0'), grad: None
588
0.01
changing lr
---------------------saving model at epoch 0----------------------------------------------------
epoch 0, time 808.00, cls_loss 7.4242 cls_loss_mapping 1.5280 cls_loss_causal 1.6920 re_mapping 0.2361 re_causal 0.2355 /// teacc 58.79 lr 0.00999497
Epoch 2, weight, value: tensor([[-0.0031, -0.0020,  0.0128,  ...,  0.0158,  0.0300,  0.0269],
        [ 0.1318,  0.1076,  0.0753,  ..., -0.0284, -0.0293, -0.0291],
        [-0.0311, -0.0348,  0.0314,  ...,  0.0709,  0.0060,  0.0512],
        ...,
        [-0.1643, -0.1202, -0.1104,  ...,  0.0602,  0.0656,  0.0368],
        [ 0.1364,  0.0869,  0.0833,  ..., -0.0678, -0.0868, -0.0781],
        [-0.0035,  0.0195, -0.0150,  ..., -0.0955, -0.0366, -0.0639]],
       device='cuda:0'), grad: tensor([[ 1.4246e-01,  4.9591e-02,  4.4464e-02,  ...,  4.6326e-02,
          4.4983e-02,  1.6815e-02],
        [-4.0009e-02, -1.2909e-02, -7.5607e-03,  ..., -1.5297e-02,
         -1.6342e-02, -6.8398e-03],
        [-9.7168e-02, -3.3356e-02, -3.5645e-02,  ..., -3.4088e-02,
         -2.8824e-02, -9.9335e-03],
        ...,
        [-3.7323e-02, -1.3939e-02, -1.0986e-02,  ..., -6.9771e-03,
         -8.4991e-03, -2.9697e-03],
        [ 1.8387e-03,  6.6328e-04,  5.3406e-04,  ...,  6.3944e-04,
          6.3848e-04,  2.6083e-04],
        [ 1.7262e-04,  5.7131e-05,  5.0813e-05,  ...,  5.4508e-05,
          4.8101e-05,  1.6570e-05]], device='cuda:0')
Epoch 2, bias, value: tensor([-0.0211, -0.0330,  0.0198,  0.0471, -0.0429,  0.0361,  0.0054],
       device='cuda:0'), grad: tensor([ 0.3125, -0.1070, -0.2090,  0.0684, -0.0692,  0.0041,  0.0004],
       device='cuda:0')
588
0.009994965332706574
changing lr
---------------------saving model at epoch 1----------------------------------------------------
epoch 1, time 814.99, cls_loss 0.9246 cls_loss_mapping 0.9319 cls_loss_causal 1.3704 re_mapping 0.0927 re_causal 0.0920 /// teacc 63.57 lr 0.00997987
Epoch 3, weight, value: tensor([[ 0.0241,  0.0225,  0.0199,  ...,  0.0375,  0.0633,  0.0597],
        [ 0.0934,  0.0721,  0.0373,  ..., -0.0349, -0.0409, -0.0381],
        [-0.0102, -0.0210,  0.0503,  ...,  0.0576, -0.0119,  0.0319],
        ...,
        [-0.1723, -0.1231, -0.1008,  ...,  0.0570,  0.0629,  0.0307],
        [ 0.1450,  0.0950,  0.0944,  ..., -0.0800, -0.0971, -0.0869],
        [-0.0020,  0.0219, -0.0180,  ..., -0.0981, -0.0372, -0.0616]],
       device='cuda:0'), grad: tensor([[-4.3732e-02, -1.3214e-02, -7.5569e-03,  ..., -2.2293e-02,
         -2.5269e-02, -1.8829e-02],
        [ 4.2152e-03,  1.6422e-03,  1.0777e-03,  ...,  1.3571e-03,
          1.5516e-03,  1.1444e-03],
        [ 1.9958e-02,  5.8060e-03,  3.2330e-03,  ...,  1.0559e-02,
          1.1971e-02,  8.9111e-03],
        ...,
        [ 1.5152e-02,  4.4861e-03,  2.5291e-03,  ...,  8.0414e-03,
          9.1019e-03,  6.8016e-03],
        [ 2.3627e-04,  6.9618e-05,  3.9160e-05,  ...,  1.2517e-04,
          1.4174e-04,  1.0562e-04],
        [ 2.0713e-05,  4.4778e-06,  2.8610e-06,  ...,  1.4633e-05,
          1.6212e-05,  1.1973e-05]], device='cuda:0')
Epoch 3, bias, value: tensor([ 0.0072, -0.0282,  0.0243,  0.0413, -0.0554,  0.0151,  0.0064],
       device='cuda:0'), grad: tensor([-1.1499e-01,  9.4757e-03,  5.3375e-02,  1.1086e-02,  4.0375e-02,
         6.2943e-04,  6.2585e-05], device='cuda:0')
588
0.009979871469976196
changing lr
---------------------saving model at epoch 2----------------------------------------------------
epoch 2, time 817.05, cls_loss 0.5564 cls_loss_mapping 0.5713 cls_loss_causal 1.1146 re_mapping 0.0838 re_causal 0.0829 /// teacc 74.37 lr 0.00995475
Epoch 4, weight, value: tensor([[ 0.0254,  0.0233,  0.0175,  ...,  0.0372,  0.0729,  0.0651],
        [ 0.0889,  0.0631,  0.0286,  ..., -0.0262, -0.0341, -0.0294],
        [-0.0140, -0.0197,  0.0500,  ...,  0.0424, -0.0304,  0.0150],
        ...,
        [-0.1693, -0.1204, -0.0955,  ...,  0.0552,  0.0656,  0.0328],
        [ 0.1554,  0.1025,  0.1018,  ..., -0.0833, -0.1015, -0.0903],
        [-0.0044,  0.0212, -0.0197,  ..., -0.0954, -0.0348, -0.0581]],
       device='cuda:0'), grad: tensor([[-9.0551e-04, -1.7595e-04, -1.8907e-04,  ..., -2.0266e-04,
         -2.7013e-04, -2.0099e-04],
        [ 8.4281e-05,  1.9148e-05,  1.9908e-05,  ...,  1.7017e-05,
          2.3603e-05,  1.7703e-05],
        [ 5.8413e-05,  1.0587e-05,  1.1586e-05,  ...,  1.3597e-05,
          1.7911e-05,  1.3277e-05],
        ...,
        [ 6.3276e-04,  1.1927e-04,  1.2910e-04,  ...,  1.4400e-04,
          1.9073e-04,  1.4174e-04],
        [ 1.1668e-05,  2.5332e-06,  2.6636e-06,  ...,  2.4624e-06,
          3.3677e-06,  2.5183e-06],
        [ 3.7141e-06,  7.7859e-07,  8.2701e-07,  ...,  7.9721e-07,
          1.0841e-06,  8.0839e-07]], device='cuda:0')
Epoch 4, bias, value: tensor([ 0.0101, -0.0139,  0.0060,  0.0420, -0.0526,  0.0196, -0.0008],
       device='cuda:0'), grad: tensor([-2.4834e-03,  2.2125e-04,  1.6344e-04,  3.0804e-04,  1.7481e-03,
         3.1203e-05,  1.0014e-05], device='cuda:0')
588
0.009954748808839675
changing lr
---------------------saving model at epoch 3----------------------------------------------------
epoch 3, time 816.59, cls_loss 0.4075 cls_loss_mapping 0.3747 cls_loss_causal 0.9471 re_mapping 0.0793 re_causal 0.0785 /// teacc 82.91 lr 0.00991965
Epoch 5, weight, value: tensor([[ 0.0587,  0.0530,  0.0429,  ...,  0.0455,  0.0862,  0.0758],
        [ 0.0638,  0.0423,  0.0081,  ..., -0.0265, -0.0333, -0.0300],
        [-0.0198, -0.0299,  0.0373,  ...,  0.0311, -0.0395,  0.0039],
        ...,
        [-0.1696, -0.1202, -0.0889,  ...,  0.0488,  0.0549,  0.0276],
        [ 0.1541,  0.1041,  0.1031,  ..., -0.0847, -0.1032, -0.0916],
        [ 0.0019,  0.0266, -0.0159,  ..., -0.0918, -0.0314, -0.0543]],
       device='cuda:0'), grad: tensor([[-0.0616, -0.0126, -0.0163,  ..., -0.0148, -0.0252, -0.0245],
        [ 0.0148,  0.0030,  0.0038,  ...,  0.0036,  0.0058,  0.0057],
        [ 0.0296,  0.0063,  0.0077,  ...,  0.0072,  0.0109,  0.0107],
        ...,
        [ 0.0063,  0.0008,  0.0018,  ...,  0.0015,  0.0047,  0.0045],
        [ 0.0024,  0.0005,  0.0006,  ...,  0.0006,  0.0010,  0.0010],
        [ 0.0011,  0.0002,  0.0003,  ...,  0.0003,  0.0004,  0.0004]],
       device='cuda:0')
Epoch 5, bias, value: tensor([ 0.0308, -0.0303,  0.0234,  0.0388, -0.0666,  0.0088,  0.0053],
       device='cuda:0'), grad: tensor([-0.1545,  0.0386,  0.0775,  0.0186,  0.0105,  0.0064,  0.0030],
       device='cuda:0')
588
0.009919647942993149
changing lr
epoch 4, time 813.70, cls_loss 0.3134 cls_loss_mapping 0.2826 cls_loss_causal 0.8661 re_mapping 0.0756 re_causal 0.0750 /// teacc 79.65 lr 0.00987464
Epoch 6, weight, value: tensor([[ 0.0543,  0.0522,  0.0399,  ...,  0.0406,  0.0876,  0.0772],
        [ 0.0528,  0.0340,  0.0004,  ..., -0.0232, -0.0285, -0.0257],
        [-0.0175, -0.0272,  0.0388,  ...,  0.0211, -0.0511, -0.0094],
        ...,
        [-0.1690, -0.1242, -0.0903,  ...,  0.0546,  0.0608,  0.0338],
        [ 0.1551,  0.1075,  0.1070,  ..., -0.0853, -0.1042, -0.0924],
        [ 0.0127,  0.0348, -0.0096,  ..., -0.0896, -0.0303, -0.0522]],
       device='cuda:0'), grad: tensor([[ 2.3499e-03,  1.0033e-03,  9.9850e-04,  ...,  1.0929e-03,
          1.1692e-03,  1.2064e-03],
        [-2.9125e-03, -1.2484e-03, -1.2417e-03,  ..., -1.3580e-03,
         -1.4515e-03, -1.4982e-03],
        [-1.9181e-04, -7.5772e-06, -1.7881e-05,  ..., -2.9355e-05,
         -5.0336e-05, -5.4300e-05],
        ...,
        [ 9.0778e-05,  1.0014e-05,  1.3940e-05,  ...,  1.8895e-05,
          2.7582e-05,  2.9400e-05],
        [ 2.1681e-06,  8.0466e-07,  7.9907e-07,  ...,  8.4005e-07,
          9.4809e-07,  9.8534e-07],
        [ 2.9460e-05,  1.1928e-05,  1.1489e-05,  ...,  1.1235e-05,
          1.2793e-05,  1.3262e-05]], device='cuda:0')
Epoch 6, bias, value: tensor([ 2.3044e-02, -3.1049e-02,  1.9318e-02,  4.3537e-02, -6.0893e-02,
        -9.9219e-05,  1.5851e-02], device='cuda:0'), grad: tensor([ 4.4403e-03, -5.4855e-03, -5.7840e-04,  1.3084e-03,  2.5487e-04,
         4.4256e-06,  5.7220e-05], device='cuda:0')
588
0.009874639560909117
changing lr
epoch 5, time 816.31, cls_loss 0.2576 cls_loss_mapping 0.2196 cls_loss_causal 0.7947 re_mapping 0.0733 re_causal 0.0730 /// teacc 68.09 lr 0.00981981
Epoch 7, weight, value: tensor([[ 0.0654,  0.0580,  0.0449,  ...,  0.0393,  0.0916,  0.0802],
        [ 0.0489,  0.0304, -0.0033,  ..., -0.0207, -0.0244, -0.0211],
        [-0.0205, -0.0309,  0.0346,  ...,  0.0150, -0.0583, -0.0180],
        ...,
        [-0.1776, -0.1272, -0.0928,  ...,  0.0570,  0.0628,  0.0360],
        [ 0.1572,  0.1097,  0.1087,  ..., -0.0837, -0.1025, -0.0906],
        [ 0.0154,  0.0374, -0.0072,  ..., -0.0894, -0.0337, -0.0543]],
       device='cuda:0'), grad: tensor([[ 2.8763e-03,  1.4992e-03,  1.5965e-03,  ...,  4.7517e-04,
          7.0667e-04,  7.4673e-04],
        [-2.7885e-03, -1.4744e-03, -1.5659e-03,  ..., -4.5371e-04,
         -6.7568e-04, -7.1335e-04],
        [-4.0770e-05, -1.8165e-05, -1.9968e-05,  ..., -1.0021e-05,
         -1.5117e-05, -1.5944e-05],
        ...,
        [-4.1306e-05, -2.6934e-06, -8.1509e-06,  ..., -1.0774e-05,
         -1.5028e-05, -1.6659e-05],
        [ 1.3821e-06,  5.4017e-07,  5.9605e-07,  ...,  2.9802e-07,
          4.5076e-07,  4.8056e-07],
        [-8.7321e-06, -5.0887e-06, -4.1500e-06,  ..., -3.6508e-07,
         -9.2760e-07, -1.0021e-06]], device='cuda:0')
Epoch 7, bias, value: tensor([ 0.0437, -0.0316,  0.0258,  0.0446, -0.0881,  0.0026,  0.0124],
       device='cuda:0'), grad: tensor([ 4.6577e-03, -4.4403e-03, -8.0109e-05, -2.6338e-06, -1.3149e-04,
         2.9169e-06, -1.2100e-05], device='cuda:0')
588
0.009819814303479266
changing lr
epoch 6, time 818.25, cls_loss 0.1873 cls_loss_mapping 0.1654 cls_loss_causal 0.7497 re_mapping 0.0694 re_causal 0.0692 /// teacc 82.66 lr 0.00975528
Epoch 8, weight, value: tensor([[ 0.0525,  0.0526,  0.0366,  ...,  0.0360,  0.0902,  0.0794],
        [ 0.0595,  0.0406,  0.0080,  ..., -0.0211, -0.0242, -0.0207],
        [-0.0158, -0.0296,  0.0351,  ...,  0.0148, -0.0563, -0.0175],
        ...,
        [-0.1736, -0.1281, -0.0912,  ...,  0.0596,  0.0646,  0.0378],
        [ 0.1559,  0.1095,  0.1088,  ..., -0.0824, -0.1009, -0.0893],
        [ 0.0144,  0.0344, -0.0106,  ..., -0.0869, -0.0334, -0.0533]],
       device='cuda:0'), grad: tensor([[-0.0395, -0.0124, -0.0165,  ..., -0.0090, -0.0090, -0.0092],
        [ 0.0177,  0.0055,  0.0073,  ...,  0.0041,  0.0041,  0.0042],
        [ 0.0131,  0.0041,  0.0054,  ...,  0.0029,  0.0029,  0.0030],
        ...,
        [ 0.0060,  0.0019,  0.0027,  ...,  0.0015,  0.0015,  0.0015],
        [ 0.0008,  0.0002,  0.0003,  ...,  0.0002,  0.0002,  0.0002],
        [ 0.0016,  0.0005,  0.0006,  ...,  0.0004,  0.0004,  0.0004]],
       device='cuda:0')
Epoch 8, bias, value: tensor([ 0.0272, -0.0320,  0.0361,  0.0317, -0.0745,  0.0015,  0.0192],
       device='cuda:0'), grad: tensor([-0.0837,  0.0378,  0.0279,  0.0006,  0.0122,  0.0018,  0.0034],
       device='cuda:0')
588
0.009755282581475767
changing lr
---------------------saving model at epoch 7----------------------------------------------------
epoch 7, time 819.90, cls_loss 0.1601 cls_loss_mapping 0.1472 cls_loss_causal 0.7095 re_mapping 0.0678 re_causal 0.0680 /// teacc 87.19 lr 0.00968117
Epoch 9, weight, value: tensor([[ 0.0682,  0.0640,  0.0488,  ...,  0.0328,  0.0870,  0.0772],
        [ 0.0578,  0.0391,  0.0079,  ..., -0.0195, -0.0213, -0.0182],
        [-0.0168, -0.0309,  0.0323,  ...,  0.0155, -0.0549, -0.0170],
        ...,
        [-0.1874, -0.1370, -0.0998,  ...,  0.0593,  0.0642,  0.0377],
        [ 0.1556,  0.1098,  0.1089,  ..., -0.0811, -0.0990, -0.0878],
        [ 0.0138,  0.0337, -0.0127,  ..., -0.0853, -0.0336, -0.0529]],
       device='cuda:0'), grad: tensor([[ 2.3198e-04,  4.4197e-05,  4.7863e-05,  ...,  8.1241e-05,
          7.5579e-05,  8.2672e-05],
        [ 1.6642e-04,  2.4870e-05,  2.9132e-05,  ...,  3.4600e-05,
          3.7372e-05,  3.7581e-05],
        [-8.1396e-04, -1.6522e-04, -1.7905e-04,  ..., -2.9278e-04,
         -2.5034e-04, -2.8634e-04],
        ...,
        [ 2.6441e-04,  6.7830e-05,  7.0870e-05,  ...,  1.3185e-04,
          9.2685e-05,  1.1873e-04],
        [ 6.4850e-05,  1.1407e-05,  1.2659e-05,  ...,  1.8835e-05,
          1.8939e-05,  1.9878e-05],
        [ 6.4135e-05,  1.2577e-05,  1.3813e-05,  ...,  1.9103e-05,
          1.8761e-05,  2.0012e-05]], device='cuda:0')
Epoch 9, bias, value: tensor([ 0.0404, -0.0317,  0.0415,  0.0336, -0.0923,  0.0013,  0.0160],
       device='cuda:0'), grad: tensor([ 5.8842e-04,  5.0735e-04, -1.9464e-03,  5.7876e-05,  4.4751e-04,
         1.7881e-04,  1.6797e-04], device='cuda:0')
588
0.009681174353198686
changing lr
---------------------saving model at epoch 8----------------------------------------------------
epoch 8, time 816.90, cls_loss 0.0972 cls_loss_mapping 0.1247 cls_loss_causal 0.6614 re_mapping 0.0653 re_causal 0.0658 /// teacc 87.69 lr 0.00959764
Epoch 10, weight, value: tensor([[ 0.0649,  0.0594,  0.0458,  ...,  0.0358,  0.0910,  0.0805],
        [ 0.0608,  0.0405,  0.0098,  ..., -0.0244, -0.0258, -0.0222],
        [-0.0181, -0.0319,  0.0302,  ...,  0.0187, -0.0479, -0.0123],
        ...,
        [-0.1769, -0.1286, -0.0939,  ...,  0.0563,  0.0585,  0.0340],
        [ 0.1486,  0.1067,  0.1055,  ..., -0.0797, -0.0975, -0.0865],
        [ 0.0117,  0.0320, -0.0136,  ..., -0.0836, -0.0337, -0.0524]],
       device='cuda:0'), grad: tensor([[ 1.3912e-04,  1.3083e-05,  3.1382e-05,  ...,  1.8790e-05,
          2.7537e-05,  2.7105e-05],
        [ 1.4760e-05,  3.1404e-06,  4.3884e-06,  ...,  3.0473e-06,
          4.1239e-06,  3.8967e-06],
        [-2.3866e-04, -2.5272e-05, -5.5760e-05,  ..., -3.4124e-05,
         -4.9949e-05, -4.8786e-05],
        ...,
        [ 6.2108e-05,  7.3090e-06,  1.5132e-05,  ...,  8.9258e-06,
          1.3202e-05,  1.2845e-05],
        [ 3.4384e-06,  5.6624e-07,  8.8289e-07,  ...,  4.7311e-07,
          6.9290e-07,  6.6683e-07],
        [ 1.0289e-05,  1.9372e-07,  2.1122e-06,  ...,  1.8664e-06,
          2.6971e-06,  2.6338e-06]], device='cuda:0')
Epoch 10, bias, value: tensor([ 0.0402, -0.0282,  0.0481,  0.0325, -0.0892, -0.0080,  0.0130],
       device='cuda:0'), grad: tensor([ 3.9959e-04,  3.6925e-05, -6.7711e-04,  2.5898e-05,  1.7405e-04,
         9.0674e-06,  3.2365e-05], device='cuda:0')
588
0.009597638862757255
changing lr
---------------------saving model at epoch 9----------------------------------------------------
epoch 9, time 819.66, cls_loss 0.0605 cls_loss_mapping 0.0894 cls_loss_causal 0.6396 re_mapping 0.0631 re_causal 0.0640 /// teacc 87.94 lr 0.00950484
Epoch 11, weight, value: tensor([[ 0.0652,  0.0586,  0.0463,  ...,  0.0348,  0.0899,  0.0799],
        [ 0.0594,  0.0395,  0.0094,  ..., -0.0237, -0.0246, -0.0211],
        [-0.0221, -0.0334,  0.0274,  ...,  0.0183, -0.0476, -0.0128],
        ...,
        [-0.1754, -0.1277, -0.0936,  ...,  0.0539,  0.0566,  0.0324],
        [ 0.1493,  0.1067,  0.1054,  ..., -0.0776, -0.0948, -0.0839],
        [ 0.0100,  0.0321, -0.0137,  ..., -0.0817, -0.0339, -0.0521]],
       device='cuda:0'), grad: tensor([[ 4.1342e-04,  6.2287e-05,  5.8174e-05,  ...,  1.6308e-04,
          2.1768e-04,  2.0707e-04],
        [ 5.8487e-06,  1.2424e-06,  1.2908e-06,  ...,  1.1362e-06,
          1.2964e-06,  1.3020e-06],
        [-1.8632e-04, -3.6508e-05, -4.0621e-05,  ..., -2.4974e-05,
         -3.2604e-05, -3.2872e-05],
        ...,
        [-3.2640e-04, -4.5002e-05, -3.9011e-05,  ..., -1.5128e-04,
         -2.0230e-04, -1.9169e-04],
        [ 1.6093e-05,  3.6769e-06,  3.7700e-06,  ...,  2.2780e-06,
          3.0100e-06,  3.0380e-06],
        [-4.5374e-06, -2.0377e-06, -1.6317e-06,  ..., -6.0350e-07,
         -9.1456e-07, -9.5554e-07]], device='cuda:0')
Epoch 11, bias, value: tensor([ 0.0427, -0.0284,  0.0403,  0.0371, -0.0876, -0.0022,  0.0062],
       device='cuda:0'), grad: tensor([ 1.0977e-03,  1.4700e-05, -4.6778e-04,  2.0730e-04, -8.7881e-04,
         3.9816e-05, -1.0781e-05], device='cuda:0')
588
0.009504844339512096
changing lr
epoch 10, time 821.27, cls_loss 0.0561 cls_loss_mapping 0.0788 cls_loss_causal 0.6070 re_mapping 0.0611 re_causal 0.0623 /// teacc 83.92 lr 0.00940298
Epoch 12, weight, value: tensor([[ 0.0616,  0.0580,  0.0463,  ...,  0.0337,  0.0886,  0.0786],
        [ 0.0605,  0.0394,  0.0097,  ..., -0.0220, -0.0226, -0.0193],
        [-0.0253, -0.0349,  0.0243,  ...,  0.0168, -0.0476, -0.0138],
        ...,
        [-0.1707, -0.1254, -0.0913,  ...,  0.0518,  0.0543,  0.0307],
        [ 0.1450,  0.1049,  0.1035,  ..., -0.0758, -0.0927, -0.0822],
        [ 0.0142,  0.0328, -0.0124,  ..., -0.0793, -0.0332, -0.0505]],
       device='cuda:0'), grad: tensor([[-1.8275e-04, -1.1110e-04, -1.1367e-04,  ..., -4.3213e-05,
         -6.9678e-05, -6.4313e-05],
        [ 1.5283e-04,  8.8334e-05,  9.0718e-05,  ...,  3.4511e-05,
          5.3734e-05,  5.0426e-05],
        [-3.7491e-05, -1.5181e-06, -3.3714e-06,  ..., -3.9898e-06,
         -3.8147e-06, -6.1728e-06],
        ...,
        [ 1.9655e-05,  6.2659e-06,  6.9514e-06,  ...,  2.5034e-06,
          3.4645e-06,  3.9563e-06],
        [-3.6597e-05, -1.8775e-05, -1.9938e-05,  ..., -9.4250e-07,
         -1.6708e-06, -2.5108e-06],
        [ 5.4181e-05,  2.5660e-05,  2.7269e-05,  ...,  5.6326e-06,
          9.1419e-06,  9.7007e-06]], device='cuda:0')
Epoch 12, bias, value: tensor([ 0.0346, -0.0208,  0.0361,  0.0345, -0.0834, -0.0068,  0.0138],
       device='cuda:0'), grad: tensor([-2.5177e-04,  2.2542e-04, -1.1235e-04,  6.1154e-05,  4.2081e-05,
        -5.0873e-05,  8.6904e-05], device='cuda:0')
588
0.009402977659283692
changing lr
---------------------saving model at epoch 11----------------------------------------------------
epoch 11, time 821.31, cls_loss 0.0570 cls_loss_mapping 0.0703 cls_loss_causal 0.5692 re_mapping 0.0586 re_causal 0.0600 /// teacc 90.45 lr 0.00929224
Epoch 13, weight, value: tensor([[ 0.0644,  0.0581,  0.0467,  ...,  0.0352,  0.0904,  0.0803],
        [ 0.0619,  0.0420,  0.0125,  ..., -0.0235, -0.0242, -0.0208],
        [-0.0252, -0.0338,  0.0236,  ...,  0.0161, -0.0469, -0.0139],
        ...,
        [-0.1647, -0.1242, -0.0901,  ...,  0.0510,  0.0538,  0.0308],
        [ 0.1384,  0.1016,  0.1000,  ..., -0.0744, -0.0911, -0.0808],
        [ 0.0093,  0.0302, -0.0140,  ..., -0.0776, -0.0332, -0.0500]],
       device='cuda:0'), grad: tensor([[ 8.0392e-06, -8.7544e-08,  2.9430e-07,  ...,  4.3623e-06,
          5.4911e-06,  5.5321e-06],
        [ 1.2279e-05,  9.0711e-07,  1.6913e-06,  ...,  7.1265e-06,
          8.9929e-06,  8.9481e-06],
        [-3.6621e-04, -2.8491e-05, -4.9472e-05,  ..., -2.0599e-04,
         -2.6464e-04, -2.6321e-04],
        ...,
        [ 3.5000e-04,  3.4899e-05,  5.5104e-05,  ...,  1.8859e-04,
          2.4211e-04,  2.4068e-04],
        [-2.7314e-05, -1.5214e-05, -1.6838e-05,  ..., -3.7253e-07,
         -7.0594e-07, -7.6555e-07],
        [ 1.8537e-05,  7.1526e-06,  8.2403e-06,  ...,  3.8929e-06,
          5.1036e-06,  5.0962e-06]], device='cuda:0')
Epoch 13, bias, value: tensor([ 0.0464, -0.0207,  0.0333,  0.0331, -0.0736, -0.0147,  0.0041],
       device='cuda:0'), grad: tensor([ 3.1501e-05,  4.5478e-05, -1.3533e-03,  1.9565e-05,  1.2522e-03,
        -3.3081e-05,  3.8832e-05], device='cuda:0')
588
0.009292243968009333
changing lr
---------------------saving model at epoch 12----------------------------------------------------
epoch 12, time 825.67, cls_loss 0.0419 cls_loss_mapping 0.0680 cls_loss_causal 0.5785 re_mapping 0.0564 re_causal 0.0583 /// teacc 90.95 lr 0.00917287
Epoch 14, weight, value: tensor([[ 0.0640,  0.0587,  0.0481,  ...,  0.0346,  0.0896,  0.0796],
        [ 0.0618,  0.0423,  0.0138,  ..., -0.0229, -0.0235, -0.0202],
        [-0.0253, -0.0344,  0.0214,  ...,  0.0166, -0.0444, -0.0121],
        ...,
        [-0.1670, -0.1243, -0.0911,  ...,  0.0478,  0.0499,  0.0273],
        [ 0.1387,  0.0999,  0.0983,  ..., -0.0718, -0.0881, -0.0780],
        [ 0.0112,  0.0304, -0.0132,  ..., -0.0752, -0.0325, -0.0486]],
       device='cuda:0'), grad: tensor([[ 2.0409e-03,  9.6655e-04,  1.0223e-03,  ...,  8.1360e-05,
          2.6798e-04,  2.6321e-04],
        [ 1.8731e-05,  3.2540e-06,  3.3826e-06,  ...,  1.0967e-05,
          1.3873e-05,  1.2882e-05],
        [-3.7594e-03, -1.2169e-03, -1.4153e-03,  ..., -4.4370e-04,
         -8.7452e-04, -8.0872e-04],
        ...,
        [ 1.4982e-03,  2.1255e-04,  3.4595e-04,  ...,  2.7776e-04,
          4.8876e-04,  4.3702e-04],
        [ 9.2164e-06,  2.5686e-06,  2.9355e-06,  ...,  1.6699e-06,
          2.7679e-06,  2.5965e-06],
        [ 5.2750e-05,  9.8124e-06,  1.0490e-05,  ...,  2.8387e-05,
          3.6836e-05,  3.4153e-05]], device='cuda:0')
Epoch 14, bias, value: tensor([ 0.0433, -0.0196,  0.0364,  0.0285, -0.0805, -0.0071,  0.0064],
       device='cuda:0'), grad: tensor([ 3.7994e-03,  5.9754e-05, -8.5068e-03,  3.9625e-04,  4.0627e-03,
         2.2739e-05,  1.6463e-04], device='cuda:0')
588
0.009172866268606516
changing lr
epoch 13, time 817.11, cls_loss 0.0311 cls_loss_mapping 0.0553 cls_loss_causal 0.5720 re_mapping 0.0535 re_causal 0.0557 /// teacc 90.70 lr 0.00904508
Epoch 15, weight, value: tensor([[ 0.0683,  0.0621,  0.0518,  ...,  0.0324,  0.0866,  0.0768],
        [ 0.0620,  0.0419,  0.0141,  ..., -0.0221, -0.0225, -0.0192],
        [-0.0247, -0.0339,  0.0208,  ...,  0.0148, -0.0450, -0.0135],
        ...,
        [-0.1660, -0.1239, -0.0918,  ...,  0.0477,  0.0502,  0.0282],
        [ 0.1343,  0.0975,  0.0958,  ..., -0.0702, -0.0863, -0.0765],
        [ 0.0075,  0.0282, -0.0144,  ..., -0.0735, -0.0320, -0.0477]],
       device='cuda:0'), grad: tensor([[ 3.0651e-03,  5.9986e-04,  5.0879e-04,  ...,  5.8985e-04,
          8.3113e-04,  7.6103e-04],
        [ 4.5419e-04,  8.9407e-05,  8.0764e-05,  ...,  8.5473e-05,
          1.2624e-04,  1.1367e-04],
        [-8.7509e-03, -1.7099e-03, -1.4582e-03,  ..., -1.6956e-03,
         -2.3994e-03, -2.1935e-03],
        ...,
        [ 2.7809e-03,  5.3740e-04,  4.3988e-04,  ...,  5.4169e-04,
          7.4339e-04,  6.8808e-04],
        [ 1.7011e-04,  3.3975e-05,  2.9579e-05,  ...,  3.2157e-05,
          4.6343e-05,  4.2140e-05],
        [ 4.3726e-04,  8.6069e-05,  7.3731e-05,  ...,  8.3447e-05,
          1.1867e-04,  1.0842e-04]], device='cuda:0')
Epoch 15, bias, value: tensor([ 4.2851e-02, -1.4903e-02,  3.4590e-02,  3.1522e-02, -7.6456e-02,
        -1.0301e-02,  1.8977e-05], device='cuda:0'), grad: tensor([ 0.0080,  0.0012, -0.0229,  0.0049,  0.0072,  0.0004,  0.0011],
       device='cuda:0')
588
0.00904508497187474
changing lr
---------------------saving model at epoch 14----------------------------------------------------
epoch 14, time 826.21, cls_loss 0.0221 cls_loss_mapping 0.0496 cls_loss_causal 0.5465 re_mapping 0.0512 re_causal 0.0539 /// teacc 91.71 lr 0.00890916
Epoch 16, weight, value: tensor([[ 0.0712,  0.0653,  0.0555,  ...,  0.0318,  0.0849,  0.0754],
        [ 0.0595,  0.0405,  0.0135,  ..., -0.0222, -0.0225, -0.0193],
        [-0.0267, -0.0348,  0.0188,  ...,  0.0135, -0.0448, -0.0141],
        ...,
        [-0.1610, -0.1225, -0.0914,  ...,  0.0478,  0.0508,  0.0292],
        [ 0.1298,  0.0948,  0.0930,  ..., -0.0687, -0.0847, -0.0750],
        [ 0.0085,  0.0278, -0.0139,  ..., -0.0716, -0.0310, -0.0463]],
       device='cuda:0'), grad: tensor([[ 1.7147e-03,  3.0088e-04,  3.9959e-04,  ...,  4.1032e-04,
          6.5947e-04,  6.2037e-04],
        [ 1.4715e-05,  4.4033e-06,  4.8503e-06,  ...,  2.9411e-06,
          4.1239e-06,  4.0755e-06],
        [-2.1495e-06,  1.5786e-06,  2.2314e-06,  ..., -4.2655e-06,
         -2.8461e-06, -3.6471e-06],
        ...,
        [-1.7195e-03, -2.9278e-04, -3.9339e-04,  ..., -4.1509e-04,
         -6.6853e-04, -6.2752e-04],
        [ 2.4036e-05,  1.1437e-05,  1.1727e-05,  ...,  2.2370e-06,
          3.2187e-06,  3.4068e-06],
        [-5.2631e-05, -2.8491e-05, -2.8670e-05,  ..., -3.0771e-06,
         -4.1500e-06, -4.8578e-06]], device='cuda:0')
Epoch 16, bias, value: tensor([ 0.0440, -0.0177,  0.0296,  0.0272, -0.0665, -0.0132,  0.0038],
       device='cuda:0'), grad: tensor([ 4.1809e-03,  3.2425e-05, -1.2070e-05,  5.0902e-05, -4.2076e-03,
         4.3541e-05, -8.7917e-05], device='cuda:0')
588
0.008909157412340152
changing lr
epoch 15, time 813.34, cls_loss 0.0266 cls_loss_mapping 0.0508 cls_loss_causal 0.5462 re_mapping 0.0484 re_causal 0.0513 /// teacc 91.71 lr 0.00876536
Epoch 17, weight, value: tensor([[ 0.0715,  0.0649,  0.0553,  ...,  0.0318,  0.0838,  0.0750],
        [ 0.0587,  0.0413,  0.0148,  ..., -0.0222, -0.0224, -0.0192],
        [-0.0274, -0.0342,  0.0180,  ...,  0.0116, -0.0450, -0.0154],
        ...,
        [-0.1618, -0.1227, -0.0923,  ...,  0.0472,  0.0501,  0.0288],
        [ 0.1293,  0.0936,  0.0919,  ..., -0.0669, -0.0824, -0.0730],
        [ 0.0088,  0.0271, -0.0136,  ..., -0.0700, -0.0305, -0.0454]],
       device='cuda:0'), grad: tensor([[ 8.0541e-06,  3.9004e-06,  2.7288e-06,  ...,  2.1141e-06,
          1.3644e-06,  1.4128e-06],
        [-5.5343e-05, -2.2486e-05, -2.2516e-05,  ..., -4.5486e-06,
         -6.7763e-06, -7.1824e-06],
        [ 1.2174e-05,  2.0973e-06,  3.1088e-06,  ...,  2.5313e-06,
          3.2149e-06,  3.2037e-06],
        ...,
        [ 4.1649e-06,  2.9355e-06,  3.1441e-06,  ..., -4.7032e-07,
         -3.0920e-07, -2.5053e-07],
        [ 5.8673e-06,  1.9930e-06,  2.1309e-06,  ...,  5.8115e-07,
          9.0804e-07,  9.4622e-07],
        [ 2.0936e-05,  7.8082e-06,  8.0466e-06,  ...,  1.8356e-06,
          2.7772e-06,  2.9653e-06]], device='cuda:0')
Epoch 17, bias, value: tensor([ 0.0460, -0.0192,  0.0244,  0.0282, -0.0692, -0.0082,  0.0049],
       device='cuda:0'), grad: tensor([ 1.7703e-05, -1.0329e-04,  3.1352e-05,  2.3097e-06, -1.1222e-06,
         1.2226e-05,  4.0591e-05], device='cuda:0')
588
0.00876535733001806
changing lr
---------------------saving model at epoch 16----------------------------------------------------
epoch 16, time 816.99, cls_loss 0.0211 cls_loss_mapping 0.0483 cls_loss_causal 0.5618 re_mapping 0.0473 re_causal 0.0506 /// teacc 91.96 lr 0.00861397
Epoch 18, weight, value: tensor([[ 0.0775,  0.0697,  0.0607,  ...,  0.0303,  0.0809,  0.0726],
        [ 0.0547,  0.0376,  0.0118,  ..., -0.0215, -0.0212, -0.0183],
        [-0.0281, -0.0353,  0.0158,  ...,  0.0125, -0.0429, -0.0142],
        ...,
        [-0.1592, -0.1212, -0.0916,  ...,  0.0467,  0.0500,  0.0292],
        [ 0.1258,  0.0916,  0.0898,  ..., -0.0655, -0.0808, -0.0716],
        [ 0.0085,  0.0268, -0.0133,  ..., -0.0684, -0.0301, -0.0447]],
       device='cuda:0'), grad: tensor([[ 1.4365e-04,  5.3078e-05,  5.5641e-05,  ...,  8.5086e-06,
          1.4983e-05,  1.8373e-05],
        [ 4.0591e-05,  1.3851e-05,  1.5192e-05,  ...,  4.5113e-06,
          6.4410e-06,  7.2829e-06],
        [ 1.2779e-04,  3.4571e-05,  3.9786e-05,  ...,  2.5526e-05,
          2.9683e-05,  3.1978e-05],
        ...,
        [ 1.4191e-03,  5.2643e-04,  5.5504e-04,  ...,  8.7798e-05,
          1.5748e-04,  1.9002e-04],
        [-5.7945e-03, -2.1477e-03, -2.2564e-03,  ..., -3.5357e-04,
         -6.3562e-04, -7.6866e-04],
        [ 3.9864e-03,  1.4734e-03,  1.5469e-03,  ...,  2.4414e-04,
          4.3726e-04,  5.2881e-04]], device='cuda:0')
Epoch 18, bias, value: tensor([ 0.0497, -0.0176,  0.0255,  0.0224, -0.0677, -0.0103,  0.0047],
       device='cuda:0'), grad: tensor([ 2.6631e-04,  7.9513e-05,  2.7847e-04,  1.0753e-04,  2.6379e-03,
        -1.0796e-02,  7.4348e-03], device='cuda:0')
588
0.008613974319136962
changing lr
epoch 17, time 818.76, cls_loss 0.0146 cls_loss_mapping 0.0373 cls_loss_causal 0.5409 re_mapping 0.0453 re_causal 0.0490 /// teacc 91.71 lr 0.00845531
Epoch 19, weight, value: tensor([[ 0.0752,  0.0691,  0.0600,  ...,  0.0290,  0.0787,  0.0707],
        [ 0.0547,  0.0370,  0.0118,  ..., -0.0211, -0.0202, -0.0175],
        [-0.0266, -0.0350,  0.0150,  ...,  0.0123, -0.0417, -0.0137],
        ...,
        [-0.1582, -0.1196, -0.0907,  ...,  0.0456,  0.0486,  0.0283],
        [ 0.1242,  0.0905,  0.0887,  ..., -0.0641, -0.0790, -0.0700],
        [ 0.0071,  0.0258, -0.0134,  ..., -0.0670, -0.0298, -0.0441]],
       device='cuda:0'), grad: tensor([[-1.0096e-06, -1.2126e-06, -1.1018e-06,  ...,  3.4366e-07,
          2.8312e-07,  1.9837e-07],
        [ 7.3351e-06,  5.9605e-07,  1.2675e-06,  ...,  3.3565e-06,
          4.1500e-06,  4.0606e-06],
        [-5.8636e-06,  4.5262e-07, -1.9465e-07,  ..., -4.3362e-06,
         -5.1446e-06, -5.1446e-06],
        ...,
        [ 3.3751e-06,  1.1064e-06,  1.2377e-06,  ...,  1.1427e-06,
          1.4827e-06,  1.4231e-06],
        [ 3.0696e-06,  1.0030e-06,  1.0589e-06,  ...,  3.2783e-07,
          5.2527e-07,  5.0385e-07],
        [-3.3528e-06, -1.4165e-06, -1.3765e-06,  ...,  1.2759e-07,
         -2.5146e-08,  1.8626e-09]], device='cuda:0')
Epoch 19, bias, value: tensor([ 0.0439, -0.0119,  0.0294,  0.0249, -0.0726, -0.0098,  0.0027],
       device='cuda:0'), grad: tensor([ 5.2713e-07,  1.9491e-05, -1.6972e-05, -8.5086e-06,  4.9025e-06,
         5.8562e-06, -5.3048e-06], device='cuda:0')
588
0.008455313244934327
changing lr
---------------------saving model at epoch 18----------------------------------------------------
epoch 18, time 814.80, cls_loss 0.0140 cls_loss_mapping 0.0346 cls_loss_causal 0.5277 re_mapping 0.0424 re_causal 0.0464 /// teacc 92.46 lr 0.00828969
Epoch 20, weight, value: tensor([[ 0.0781,  0.0697,  0.0611,  ...,  0.0294,  0.0788,  0.0709],
        [ 0.0527,  0.0364,  0.0118,  ..., -0.0212, -0.0206, -0.0179],
        [-0.0272, -0.0350,  0.0138,  ...,  0.0120, -0.0409, -0.0136],
        ...,
        [-0.1548, -0.1181, -0.0897,  ...,  0.0448,  0.0481,  0.0283],
        [ 0.1220,  0.0894,  0.0875,  ..., -0.0628, -0.0776, -0.0687],
        [ 0.0063,  0.0250, -0.0135,  ..., -0.0657, -0.0294, -0.0434]],
       device='cuda:0'), grad: tensor([[ 4.7296e-05,  1.2435e-05,  1.3180e-05,  ...,  8.9556e-06,
          1.2085e-05,  1.2808e-05],
        [ 2.1249e-05,  3.6787e-06,  5.5023e-06,  ...,  7.8008e-06,
          8.6352e-06,  9.1121e-06],
        [-9.3341e-05, -2.8506e-05, -2.8655e-05,  ..., -1.3091e-05,
         -1.9684e-05, -2.0817e-05],
        ...,
        [ 2.8729e-05,  9.0674e-06,  9.4250e-06,  ...,  3.9823e-06,
          4.9956e-06,  5.6550e-06],
        [ 8.0541e-06,  2.3283e-06,  2.3935e-06,  ...,  1.2824e-06,
          1.7788e-06,  1.9064e-06],
        [ 5.2862e-06,  1.3188e-06,  1.4612e-06,  ...,  1.1791e-06,
          1.5628e-06,  1.6280e-06]], device='cuda:0')
Epoch 20, bias, value: tensor([ 0.0500, -0.0137,  0.0276,  0.0195, -0.0668, -0.0109,  0.0007],
       device='cuda:0'), grad: tensor([ 1.1319e-04,  5.6565e-05, -2.1183e-04, -5.3734e-05,  6.3956e-05,
         1.8641e-05,  1.2890e-05], device='cuda:0')
588
0.008289693629698565
changing lr
epoch 19, time 817.70, cls_loss 0.0107 cls_loss_mapping 0.0303 cls_loss_causal 0.4927 re_mapping 0.0397 re_causal 0.0443 /// teacc 91.96 lr 0.00811745
Epoch 21, weight, value: tensor([[ 0.0791,  0.0704,  0.0623,  ...,  0.0290,  0.0777,  0.0699],
        [ 0.0500,  0.0352,  0.0110,  ..., -0.0210, -0.0206, -0.0179],
        [-0.0248, -0.0341,  0.0136,  ...,  0.0118, -0.0396, -0.0130],
        ...,
        [-0.1541, -0.1172, -0.0894,  ...,  0.0437,  0.0465,  0.0273],
        [ 0.1196,  0.0874,  0.0854,  ..., -0.0615, -0.0760, -0.0673],
        [ 0.0064,  0.0247, -0.0130,  ..., -0.0642, -0.0286, -0.0423]],
       device='cuda:0'), grad: tensor([[-3.9160e-05, -1.7390e-05, -2.0787e-05,  ..., -1.6809e-05,
         -2.4855e-05, -2.4080e-05],
        [ 3.4142e-04,  8.4937e-05,  9.1851e-05,  ...,  4.8280e-05,
          8.7559e-05,  8.2612e-05],
        [-1.4208e-05,  3.8259e-06,  3.5446e-06,  ..., -7.3127e-06,
         -8.6874e-06, -8.7619e-06],
        ...,
        [ 1.3396e-05,  3.8035e-06,  4.4145e-06,  ...,  3.3602e-06,
          5.0776e-06,  4.9211e-06],
        [-2.9411e-06, -3.5223e-06, -4.0568e-06,  ...,  6.3144e-07,
          1.1791e-06,  1.0673e-06],
        [-3.4285e-04, -7.9155e-05, -8.3864e-05,  ..., -4.0293e-05,
         -7.7784e-05, -7.2777e-05]], device='cuda:0')
Epoch 21, bias, value: tensor([ 0.0515, -0.0168,  0.0326,  0.0178, -0.0700, -0.0094,  0.0005],
       device='cuda:0'), grad: tensor([-7.2360e-05,  8.0204e-04, -5.6326e-05,  1.1629e-04,  3.0488e-05,
         1.3709e-06, -8.2064e-04], device='cuda:0')
588
0.00811744900929367
changing lr
---------------------saving model at epoch 20----------------------------------------------------
epoch 20, time 819.64, cls_loss 0.0094 cls_loss_mapping 0.0289 cls_loss_causal 0.4921 re_mapping 0.0375 re_causal 0.0426 /// teacc 93.97 lr 0.00793893
Epoch 22, weight, value: tensor([[ 0.0755,  0.0697,  0.0617,  ...,  0.0272,  0.0746,  0.0672],
        [ 0.0486,  0.0338,  0.0103,  ..., -0.0204, -0.0197, -0.0172],
        [-0.0257, -0.0337,  0.0127,  ...,  0.0108, -0.0398, -0.0138],
        ...,
        [-0.1503, -0.1157, -0.0884,  ...,  0.0436,  0.0470,  0.0282],
        [ 0.1188,  0.0868,  0.0848,  ..., -0.0600, -0.0743, -0.0659],
        [ 0.0069,  0.0244, -0.0126,  ..., -0.0629, -0.0281, -0.0414]],
       device='cuda:0'), grad: tensor([[-1.3173e-04, -6.0469e-05, -6.1989e-05,  ..., -9.6112e-06,
         -2.5392e-05, -2.3678e-05],
        [-4.2051e-05, -3.4384e-06, -4.9137e-06,  ..., -1.7628e-05,
         -2.1622e-05, -2.1964e-05],
        [ 2.7984e-05,  1.1958e-05,  1.2219e-05,  ...,  3.5465e-06,
          6.8322e-06,  6.5304e-06],
        ...,
        [ 9.0480e-05,  3.0965e-05,  3.3319e-05,  ...,  1.5631e-05,
          2.5705e-05,  2.5153e-05],
        [ 1.5885e-05,  5.9344e-06,  6.1505e-06,  ...,  2.3320e-06,
          4.1239e-06,  3.9600e-06],
        [ 3.3468e-05,  1.1489e-05,  1.2018e-05,  ...,  5.7481e-06,
          9.4101e-06,  9.1642e-06]], device='cuda:0')
Epoch 22, bias, value: tensor([ 0.0416, -0.0155,  0.0282,  0.0205, -0.0621, -0.0081,  0.0015],
       device='cuda:0'), grad: tensor([-2.2805e-04, -1.2910e-04,  5.2780e-05,  7.2978e-06,  1.9133e-04,
         3.2693e-05,  7.3075e-05], device='cuda:0')
588
0.007938926261462368
changing lr
epoch 21, time 817.08, cls_loss 0.0062 cls_loss_mapping 0.0266 cls_loss_causal 0.4959 re_mapping 0.0337 re_causal 0.0396 /// teacc 92.46 lr 0.00775448
Epoch 23, weight, value: tensor([[ 0.0774,  0.0699,  0.0624,  ...,  0.0271,  0.0741,  0.0668],
        [ 0.0476,  0.0338,  0.0106,  ..., -0.0205, -0.0197, -0.0172],
        [-0.0258, -0.0335,  0.0118,  ...,  0.0105, -0.0391, -0.0137],
        ...,
        [-0.1503, -0.1148, -0.0882,  ...,  0.0425,  0.0455,  0.0272],
        [ 0.1166,  0.0855,  0.0835,  ..., -0.0589, -0.0731, -0.0647],
        [ 0.0065,  0.0236, -0.0126,  ..., -0.0617, -0.0276, -0.0406]],
       device='cuda:0'), grad: tensor([[-2.8193e-05, -1.7360e-05, -2.0295e-05,  ...,  4.8876e-06,
          2.8685e-07,  0.0000e+00],
        [ 3.0443e-05,  5.6103e-06,  8.6799e-06,  ...,  1.1146e-05,
          1.4178e-05,  1.3806e-05],
        [ 5.9158e-06,  1.6000e-06,  2.3283e-06,  ...,  3.2876e-06,
          3.8818e-06,  3.7644e-06],
        ...,
        [ 2.7075e-05,  8.2180e-06,  1.0900e-05,  ...,  4.9397e-06,
          8.1584e-06,  8.0317e-06],
        [ 1.0602e-05,  2.5760e-06,  3.6433e-06,  ...,  3.0547e-06,
          4.2059e-06,  4.1127e-06],
        [ 1.5199e-05,  4.4219e-06,  5.9418e-06,  ...,  3.2336e-06,
          5.0031e-06,  4.9137e-06]], device='cuda:0')
Epoch 23, bias, value: tensor([ 0.0455, -0.0160,  0.0273,  0.0216, -0.0653, -0.0090,  0.0019],
       device='cuda:0'), grad: tensor([-3.4958e-05,  7.6354e-05,  1.3001e-05, -1.7130e-04,  5.8532e-05,
         2.4796e-05,  3.3438e-05], device='cuda:0')
588
0.007754484907260515
changing lr
epoch 22, time 813.28, cls_loss 0.0129 cls_loss_mapping 0.0309 cls_loss_causal 0.4926 re_mapping 0.0328 re_causal 0.0391 /// teacc 92.21 lr 0.00756450
Epoch 24, weight, value: tensor([[ 0.0758,  0.0697,  0.0622,  ...,  0.0260,  0.0721,  0.0649],
        [ 0.0471,  0.0335,  0.0109,  ..., -0.0200, -0.0191, -0.0167],
        [-0.0247, -0.0333,  0.0110,  ...,  0.0097, -0.0386, -0.0138],
        ...,
        [-0.1486, -0.1136, -0.0874,  ...,  0.0423,  0.0453,  0.0273],
        [ 0.1150,  0.0843,  0.0823,  ..., -0.0578, -0.0717, -0.0635],
        [ 0.0080,  0.0234, -0.0121,  ..., -0.0597, -0.0263, -0.0389]],
       device='cuda:0'), grad: tensor([[-2.2335e-03, -7.6771e-04, -8.5497e-04,  ..., -2.6250e-04,
         -6.4802e-04, -6.1846e-04],
        [ 5.7936e-04,  2.0504e-04,  2.0194e-04,  ...,  7.2360e-05,
          1.7726e-04,  1.7560e-04],
        [ 1.0309e-03,  3.1829e-04,  3.8576e-04,  ...,  1.1569e-04,
          2.8825e-04,  2.7323e-04],
        ...,
        [ 5.9754e-05,  5.4240e-05,  6.3658e-05,  ...,  6.6273e-06,
          1.5646e-05,  7.1041e-06],
        [ 7.4506e-05,  2.3946e-05,  2.7284e-05,  ...,  8.6576e-06,
          2.1428e-05,  2.0623e-05],
        [ 3.7026e-04,  1.2684e-04,  1.2946e-04,  ...,  4.5449e-05,
          1.1164e-04,  1.1009e-04]], device='cuda:0')
Epoch 24, bias, value: tensor([ 0.0413, -0.0152,  0.0281,  0.0178, -0.0638, -0.0087,  0.0063],
       device='cuda:0'), grad: tensor([-4.5090e-03,  1.1473e-03,  2.1763e-03,  2.4605e-04,  3.8803e-05,
         1.5461e-04,  7.4577e-04], device='cuda:0')
588
0.007564496387029534
changing lr
epoch 23, time 838.63, cls_loss 0.0068 cls_loss_mapping 0.0268 cls_loss_causal 0.4974 re_mapping 0.0309 re_causal 0.0379 /// teacc 93.47 lr 0.00736934
Epoch 25, weight, value: tensor([[ 0.0772,  0.0701,  0.0627,  ...,  0.0253,  0.0711,  0.0641],
        [ 0.0474,  0.0329,  0.0109,  ..., -0.0192, -0.0180, -0.0156],
        [-0.0240, -0.0325,  0.0112,  ...,  0.0092, -0.0379, -0.0136],
        ...,
        [-0.1461, -0.1120, -0.0863,  ...,  0.0420,  0.0448,  0.0272],
        [ 0.1120,  0.0828,  0.0807,  ..., -0.0569, -0.0707, -0.0627],
        [ 0.0052,  0.0222, -0.0129,  ..., -0.0591, -0.0268, -0.0390]],
       device='cuda:0'), grad: tensor([[-5.1880e-04, -1.9383e-04, -1.9336e-04,  ..., -4.0913e-04,
         -5.2977e-04, -5.5504e-04],
        [ 2.3937e-04,  4.2588e-05,  5.4538e-05,  ...,  1.3030e-04,
          1.8048e-04,  1.8859e-04],
        [ 6.7043e-04,  5.6118e-05,  1.1438e-04,  ...,  1.1480e-04,
          2.1446e-04,  2.1780e-04],
        ...,
        [-7.0429e-04,  3.7283e-05, -5.2094e-05,  ...,  6.7055e-05,
         -1.4283e-05, -4.5113e-06],
        [ 8.7440e-05,  1.3962e-05,  1.9893e-05,  ...,  2.1994e-05,
          3.5554e-05,  3.6418e-05],
        [ 1.3733e-04,  2.6837e-05,  3.4779e-05,  ...,  4.2856e-05,
          6.5029e-05,  6.6936e-05]], device='cuda:0')
Epoch 25, bias, value: tensor([ 0.0444, -0.0117,  0.0283,  0.0172, -0.0612, -0.0112, -0.0002],
       device='cuda:0'), grad: tensor([-0.0020,  0.0009,  0.0019,  0.0003, -0.0017,  0.0002,  0.0004],
       device='cuda:0')
588
0.007369343312364995
changing lr
epoch 24, time 812.60, cls_loss 0.0083 cls_loss_mapping 0.0233 cls_loss_causal 0.4944 re_mapping 0.0281 re_causal 0.0354 /// teacc 92.46 lr 0.00716942
Epoch 26, weight, value: tensor([[ 0.0793,  0.0713,  0.0640,  ...,  0.0250,  0.0705,  0.0636],
        [ 0.0455,  0.0320,  0.0104,  ..., -0.0190, -0.0178, -0.0154],
        [-0.0262, -0.0326,  0.0101,  ...,  0.0075, -0.0389, -0.0151],
        ...,
        [-0.1435, -0.1111, -0.0858,  ...,  0.0423,  0.0453,  0.0279],
        [ 0.1096,  0.0815,  0.0795,  ..., -0.0560, -0.0697, -0.0618],
        [ 0.0047,  0.0217, -0.0128,  ..., -0.0581, -0.0266, -0.0386]],
       device='cuda:0'), grad: tensor([[-1.9336e-04, -8.9705e-05, -1.1009e-04,  ..., -4.8935e-05,
         -6.3360e-05, -5.6565e-05],
        [ 1.6642e-04,  6.4254e-05,  7.9572e-05,  ...,  4.3362e-05,
          5.6803e-05,  5.0813e-05],
        [-5.6922e-06, -6.2399e-08, -2.3283e-07,  ..., -2.2445e-07,
         -1.1520e-06, -1.0673e-06],
        ...,
        [ 2.4468e-05,  1.0535e-05,  1.3068e-05,  ...,  6.4000e-06,
          8.0764e-06,  7.2680e-06],
        [ 2.3529e-05,  9.8720e-06,  1.1981e-05,  ...,  5.3756e-06,
          7.2233e-06,  6.5342e-06],
        [-4.8280e-06, -1.2685e-06, -8.4285e-07,  ...,  8.8289e-07,
          5.6252e-07,  1.9465e-07]], device='cuda:0')
Epoch 26, bias, value: tensor([ 0.0483, -0.0136,  0.0214,  0.0208, -0.0566, -0.0131, -0.0017],
       device='cuda:0'), grad: tensor([-5.5456e-04,  4.6062e-04, -7.7561e-06, -2.8417e-05,  6.7294e-05,
         6.2883e-05, -7.5717e-07], device='cuda:0')
588
0.0071694186955877925
changing lr
epoch 25, time 815.09, cls_loss 0.0045 cls_loss_mapping 0.0209 cls_loss_causal 0.4697 re_mapping 0.0255 re_causal 0.0334 /// teacc 90.95 lr 0.00696513
Epoch 27, weight, value: tensor([[ 0.0799,  0.0712,  0.0642,  ...,  0.0250,  0.0700,  0.0632],
        [ 0.0423,  0.0301,  0.0090,  ..., -0.0189, -0.0177, -0.0154],
        [-0.0227, -0.0313,  0.0106,  ...,  0.0076, -0.0376, -0.0143],
        ...,
        [-0.1438, -0.1101, -0.0854,  ...,  0.0410,  0.0437,  0.0268],
        [ 0.1093,  0.0812,  0.0792,  ..., -0.0550, -0.0684, -0.0607],
        [ 0.0045,  0.0212, -0.0128,  ..., -0.0570, -0.0263, -0.0380]],
       device='cuda:0'), grad: tensor([[ 7.5102e-04,  2.4223e-04,  2.5082e-04,  ...,  8.1003e-05,
          1.7059e-04,  1.6427e-04],
        [-2.1534e-03, -4.9210e-04, -5.3501e-04,  ..., -4.0579e-04,
         -7.5388e-04, -7.0095e-04],
        [ 6.5565e-04,  1.3733e-04,  1.5342e-04,  ...,  1.4067e-04,
          2.4140e-04,  2.2197e-04],
        ...,
        [ 4.6134e-04,  1.2887e-04,  1.3399e-04,  ...,  6.0886e-05,
          1.3816e-04,  1.3185e-04],
        [ 1.4603e-04,  4.7207e-05,  4.9174e-05,  ...,  1.7181e-05,
          3.4273e-05,  3.2783e-05],
        [-2.2185e-04, -1.3363e-04, -1.3316e-04,  ...,  1.8224e-05,
          2.9132e-05,  2.1964e-05]], device='cuda:0')
Epoch 27, bias, value: tensor([ 0.0511, -0.0171,  0.0282,  0.0184, -0.0614, -0.0114, -0.0023],
       device='cuda:0'), grad: tensor([ 0.0015, -0.0049,  0.0015,  0.0008,  0.0010,  0.0003, -0.0003],
       device='cuda:0')
588
0.0069651251582696205
changing lr
epoch 26, time 811.56, cls_loss 0.0048 cls_loss_mapping 0.0252 cls_loss_causal 0.4728 re_mapping 0.0231 re_causal 0.0307 /// teacc 92.21 lr 0.00675687
Epoch 28, weight, value: tensor([[ 0.0781,  0.0708,  0.0640,  ...,  0.0244,  0.0687,  0.0620],
        [ 0.0419,  0.0297,  0.0090,  ..., -0.0184, -0.0171, -0.0149],
        [-0.0219, -0.0308,  0.0105,  ...,  0.0071, -0.0371, -0.0143],
        ...,
        [-0.1406, -0.1084, -0.0842,  ...,  0.0407,  0.0434,  0.0268],
        [ 0.1079,  0.0803,  0.0783,  ..., -0.0541, -0.0674, -0.0597],
        [ 0.0038,  0.0204, -0.0132,  ..., -0.0560, -0.0258, -0.0374]],
       device='cuda:0'), grad: tensor([[ 2.9993e-04,  5.0426e-05,  7.1168e-05,  ...,  6.6698e-05,
          1.1802e-04,  1.0926e-04],
        [ 6.8903e-05,  8.4564e-06,  1.6898e-05,  ...,  2.1636e-05,
          3.6865e-05,  3.4124e-05],
        [-2.8992e-04, -4.8310e-05, -7.9393e-05,  ..., -7.9989e-05,
         -1.5891e-04, -1.5199e-04],
        ...,
        [-1.2755e-04, -2.4274e-05, -2.2009e-05,  ..., -1.2144e-05,
         -2.3648e-05, -2.0996e-05],
        [ 3.1620e-05,  5.1260e-06,  8.0764e-06,  ...,  8.0988e-06,
          1.3381e-05,  1.2286e-05],
        [ 5.0277e-05,  7.9572e-06,  1.3076e-05,  ...,  1.3307e-05,
          2.3946e-05,  2.2426e-05]], device='cuda:0')
Epoch 28, bias, value: tensor([ 0.0468, -0.0160,  0.0291,  0.0164, -0.0566, -0.0112, -0.0030],
       device='cuda:0'), grad: tensor([ 8.6927e-04,  2.4223e-04, -1.1597e-03,  3.6865e-05, -2.4700e-04,
         9.1076e-05,  1.6832e-04], device='cuda:0')
588
0.006756874120406716
changing lr
epoch 27, time 812.10, cls_loss 0.0076 cls_loss_mapping 0.0221 cls_loss_causal 0.4726 re_mapping 0.0227 re_causal 0.0314 /// teacc 93.22 lr 0.00654508
Epoch 29, weight, value: tensor([[ 0.0767,  0.0703,  0.0636,  ...,  0.0237,  0.0673,  0.0607],
        [ 0.0415,  0.0293,  0.0090,  ..., -0.0181, -0.0167, -0.0145],
        [-0.0233, -0.0309,  0.0096,  ...,  0.0065, -0.0370, -0.0145],
        ...,
        [-0.1376, -0.1069, -0.0831,  ...,  0.0406,  0.0437,  0.0272],
        [ 0.1074,  0.0792,  0.0773,  ..., -0.0532, -0.0663, -0.0588],
        [ 0.0036,  0.0202, -0.0129,  ..., -0.0553, -0.0257, -0.0370]],
       device='cuda:0'), grad: tensor([[ 4.3094e-05,  5.2899e-06,  8.3968e-06,  ...,  3.5278e-06,
          6.4224e-06,  6.6943e-06],
        [ 1.5283e-04,  2.0862e-05,  3.1590e-05,  ...,  1.8626e-05,
          2.8476e-05,  2.9594e-05],
        [ 3.7730e-05,  5.0962e-06,  7.6368e-06,  ...,  5.5395e-06,
          8.8438e-06,  8.7544e-06],
        ...,
        [-1.3523e-06,  7.4506e-08,  1.4901e-07,  ..., -4.0233e-06,
         -7.0371e-06, -5.8599e-06],
        [ 4.2766e-05,  6.2548e-06,  9.1791e-06,  ...,  2.9728e-06,
          5.7705e-06,  6.0350e-06],
        [ 9.9599e-05,  1.4491e-05,  2.1368e-05,  ...,  5.8301e-06,
          1.1764e-05,  1.2614e-05]], device='cuda:0')
Epoch 29, bias, value: tensor([ 0.0437, -0.0154,  0.0251,  0.0160, -0.0519, -0.0080, -0.0043],
       device='cuda:0'), grad: tensor([ 1.0687e-04,  3.7146e-04,  9.4533e-05, -9.0218e-04, -1.3970e-05,
         1.0341e-04,  2.3937e-04], device='cuda:0')
588
0.00654508497187474
changing lr
epoch 28, time 813.77, cls_loss 0.0030 cls_loss_mapping 0.0186 cls_loss_causal 0.4803 re_mapping 0.0193 re_causal 0.0285 /// teacc 93.72 lr 0.00633018
Epoch 30, weight, value: tensor([[ 0.0781,  0.0712,  0.0646,  ...,  0.0232,  0.0664,  0.0599],
        [ 0.0420,  0.0290,  0.0091,  ..., -0.0173, -0.0157, -0.0134],
        [-0.0223, -0.0306,  0.0091,  ...,  0.0059, -0.0366, -0.0144],
        ...,
        [-0.1383, -0.1062, -0.0829,  ...,  0.0398,  0.0426,  0.0264],
        [ 0.1040,  0.0778,  0.0758,  ..., -0.0526, -0.0658, -0.0584],
        [ 0.0040,  0.0198, -0.0128,  ..., -0.0544, -0.0254, -0.0364]],
       device='cuda:0'), grad: tensor([[-4.5508e-05, -2.1264e-05, -2.1949e-05,  ..., -2.3674e-06,
         -5.2936e-06, -4.8615e-06],
        [-2.9489e-05, -6.8396e-06, -6.5006e-06,  ..., -4.3102e-06,
         -1.1213e-05, -1.0550e-05],
        [-2.6766e-06, -2.5257e-06, -2.3134e-06,  ...,  2.0117e-07,
          1.0058e-07, -2.6077e-08],
        ...,
        [ 3.0696e-05,  1.1295e-05,  1.2062e-05,  ...,  3.9376e-06,
          6.7912e-06,  6.2697e-06],
        [ 1.0431e-05,  3.5726e-06,  3.8259e-06,  ...,  1.7378e-06,
          2.6803e-06,  2.4978e-06],
        [ 4.9114e-05,  1.3009e-05,  1.3947e-05,  ...,  9.2536e-06,
          1.6272e-05,  1.5147e-05]], device='cuda:0')
Epoch 30, bias, value: tensor([ 0.0453, -0.0113,  0.0267,  0.0166, -0.0566, -0.0127, -0.0029],
       device='cuda:0'), grad: tensor([-7.7009e-05, -7.4744e-05, -4.2170e-06, -4.6015e-05,  6.2525e-05,
         2.2516e-05,  1.1718e-04], device='cuda:0')
588
0.006330184227833378
changing lr
epoch 29, time 812.07, cls_loss 0.0059 cls_loss_mapping 0.0181 cls_loss_causal 0.4874 re_mapping 0.0183 re_causal 0.0279 /// teacc 93.97 lr 0.00611260
Epoch 31, weight, value: tensor([[ 0.0769,  0.0709,  0.0645,  ...,  0.0224,  0.0651,  0.0587],
        [ 0.0414,  0.0288,  0.0094,  ..., -0.0171, -0.0156, -0.0134],
        [-0.0222, -0.0306,  0.0084,  ...,  0.0058, -0.0357, -0.0140],
        ...,
        [-0.1354, -0.1048, -0.0819,  ...,  0.0395,  0.0425,  0.0266],
        [ 0.1018,  0.0766,  0.0746,  ..., -0.0519, -0.0650, -0.0577],
        [ 0.0035,  0.0195, -0.0127,  ..., -0.0537, -0.0253, -0.0361]],
       device='cuda:0'), grad: tensor([[ 3.9139e-03,  5.4932e-04,  3.8695e-04,  ...,  3.5524e-04,
          7.6008e-04,  8.3113e-04],
        [-7.1466e-05, -4.0025e-05, -4.5657e-05,  ..., -2.9542e-06,
         -7.9945e-06, -5.8375e-06],
        [ 1.5774e-03,  2.2268e-04,  1.6809e-04,  ...,  1.9121e-04,
          3.5977e-04,  3.7146e-04],
        ...,
        [-5.7487e-03, -7.9298e-04, -5.6410e-04,  ..., -5.7840e-04,
         -1.1806e-03, -1.2665e-03],
        [ 4.4435e-05,  7.5325e-06,  6.5975e-06,  ...,  4.9621e-06,
          9.7528e-06,  9.9167e-06],
        [ 1.0777e-04,  2.2024e-05,  2.0474e-05,  ...,  1.0796e-05,
          2.1920e-05,  2.2098e-05]], device='cuda:0')
Epoch 31, bias, value: tensor([ 0.0428, -0.0124,  0.0279,  0.0186, -0.0526, -0.0146, -0.0046],
       device='cuda:0'), grad: tensor([ 1.0109e-02, -8.3566e-05,  4.1885e-03,  4.4632e-04, -1.5038e-02,
         1.1331e-04,  2.6035e-04], device='cuda:0')
588
0.006112604669781575
changing lr
epoch 30, time 810.53, cls_loss 0.0028 cls_loss_mapping 0.0157 cls_loss_causal 0.4478 re_mapping 0.0176 re_causal 0.0272 /// teacc 89.95 lr 0.00589278
Epoch 32, weight, value: tensor([[ 0.0778,  0.0712,  0.0649,  ...,  0.0224,  0.0645,  0.0582],
        [ 0.0394,  0.0280,  0.0090,  ..., -0.0171, -0.0157, -0.0135],
        [-0.0223, -0.0304,  0.0079,  ...,  0.0056, -0.0352, -0.0139],
        ...,
        [-0.1338, -0.1039, -0.0814,  ...,  0.0390,  0.0420,  0.0264],
        [ 0.1008,  0.0758,  0.0739,  ..., -0.0511, -0.0640, -0.0568],
        [ 0.0047,  0.0196, -0.0122,  ..., -0.0528, -0.0248, -0.0354]],
       device='cuda:0'), grad: tensor([[ 1.5440e-03,  2.1601e-04,  3.2449e-04,  ...,  5.2166e-04,
          7.2908e-04,  7.3481e-04],
        [ 8.5890e-05,  3.0577e-05,  2.9802e-05,  ...,  9.7379e-06,
          1.2994e-05,  1.3396e-05],
        [ 1.4953e-05,  3.4757e-06,  2.5313e-06,  ...,  4.3176e-06,
          5.6326e-06,  5.4464e-06],
        ...,
        [-1.4734e-03, -1.8299e-04, -2.9349e-04,  ..., -5.2118e-04,
         -7.2861e-04, -7.3338e-04],
        [ 5.3227e-05,  2.0131e-05,  2.0579e-05,  ...,  4.8876e-06,
          7.8529e-06,  7.9647e-06],
        [-1.9717e-04, -8.7500e-05, -8.7082e-05,  ..., -3.6918e-06,
         -1.1519e-05, -1.1921e-05]], device='cuda:0')
Epoch 32, bias, value: tensor([ 0.0444, -0.0149,  0.0270,  0.0153, -0.0508, -0.0138, -0.0022],
       device='cuda:0'), grad: tensor([ 4.4670e-03,  1.7548e-04,  4.0948e-05, -9.1374e-05, -4.3602e-03,
         1.0496e-04, -3.3760e-04], device='cuda:0')
588
0.005892784473993186
changing lr
epoch 31, time 813.01, cls_loss 0.0045 cls_loss_mapping 0.0195 cls_loss_causal 0.4568 re_mapping 0.0153 re_causal 0.0249 /// teacc 93.72 lr 0.00567117
Epoch 33, weight, value: tensor([[ 0.0805,  0.0723,  0.0664,  ...,  0.0227,  0.0647,  0.0585],
        [ 0.0396,  0.0276,  0.0089,  ..., -0.0171, -0.0156, -0.0135],
        [-0.0233, -0.0305,  0.0070,  ...,  0.0051, -0.0351, -0.0142],
        ...,
        [-0.1332, -0.1033, -0.0812,  ...,  0.0383,  0.0413,  0.0259],
        [ 0.0997,  0.0750,  0.0731,  ..., -0.0504, -0.0631, -0.0561],
        [ 0.0031,  0.0189, -0.0125,  ..., -0.0522, -0.0249, -0.0353]],
       device='cuda:0'), grad: tensor([[-1.4362e-03, -3.2687e-04, -3.6669e-04,  ..., -3.9530e-04,
         -6.0272e-04, -6.3896e-04],
        [ 8.1539e-05,  1.5169e-05,  1.7136e-05,  ...,  1.8433e-05,
          3.4511e-05,  3.5882e-05],
        [ 2.3198e-04,  6.6042e-05,  6.8784e-05,  ...,  7.9215e-05,
          1.0860e-04,  1.1557e-04],
        ...,
        [ 2.8539e-04,  5.8323e-05,  6.8307e-05,  ...,  6.3658e-05,
          1.0639e-04,  1.1194e-04],
        [ 3.8475e-05,  8.2403e-06,  9.5218e-06,  ...,  7.7859e-06,
          1.3344e-05,  1.3851e-05],
        [ 1.0186e-04,  2.2292e-05,  2.5555e-05,  ...,  2.1353e-05,
          3.5822e-05,  3.7313e-05]], device='cuda:0')
Epoch 33, bias, value: tensor([ 0.0503, -0.0127,  0.0242,  0.0137, -0.0515, -0.0136, -0.0056],
       device='cuda:0'), grad: tensor([-3.3913e-03,  1.8573e-04,  5.0020e-04,  1.6565e-03,  7.0381e-04,
         9.3997e-05,  2.4843e-04], device='cuda:0')
588
0.00567116632908828
changing lr
epoch 32, time 813.20, cls_loss 0.0044 cls_loss_mapping 0.0188 cls_loss_causal 0.4588 re_mapping 0.0156 re_causal 0.0258 /// teacc 91.96 lr 0.00544820
Epoch 34, weight, value: tensor([[ 0.0806,  0.0727,  0.0669,  ...,  0.0224,  0.0639,  0.0578],
        [ 0.0380,  0.0269,  0.0086,  ..., -0.0171, -0.0157, -0.0136],
        [-0.0225, -0.0304,  0.0066,  ...,  0.0050, -0.0343, -0.0138],
        ...,
        [-0.1327, -0.1026, -0.0809,  ...,  0.0376,  0.0405,  0.0254],
        [ 0.0984,  0.0742,  0.0722,  ..., -0.0497, -0.0624, -0.0554],
        [ 0.0033,  0.0186, -0.0124,  ..., -0.0514, -0.0246, -0.0348]],
       device='cuda:0'), grad: tensor([[ 7.3204e-03,  2.9449e-03,  3.4027e-03,  ...,  8.6594e-04,
          1.3466e-03,  1.3685e-03],
        [-8.4152e-03, -3.3875e-03, -3.9139e-03,  ..., -9.9945e-04,
         -1.5469e-03, -1.5726e-03],
        [ 2.3198e-04,  7.9691e-05,  9.4295e-05,  ...,  2.4244e-05,
          4.5002e-05,  4.4882e-05],
        ...,
        [ 2.0623e-05,  3.5644e-05,  3.6538e-05,  ...,  1.2584e-05,
          2.5295e-06,  4.0717e-06],
        [ 2.9469e-04,  1.1790e-04,  1.3626e-04,  ...,  3.4988e-05,
          5.4449e-05,  5.5283e-05],
        [ 2.4796e-04,  9.5606e-05,  1.1057e-04,  ...,  2.9609e-05,
          4.6521e-05,  4.6968e-05]], device='cuda:0')
Epoch 34, bias, value: tensor([ 0.0493, -0.0150,  0.0264,  0.0159, -0.0530, -0.0139, -0.0048],
       device='cuda:0'), grad: tensor([ 1.3618e-02, -1.5640e-02,  4.8470e-04,  5.6648e-04, -6.4909e-05,
         5.5075e-04,  4.7946e-04], device='cuda:0')
588
0.00544819654451717
changing lr
epoch 33, time 818.78, cls_loss 0.0023 cls_loss_mapping 0.0129 cls_loss_causal 0.4522 re_mapping 0.0148 re_causal 0.0254 /// teacc 91.96 lr 0.00522432
Epoch 35, weight, value: tensor([[ 0.0796,  0.0724,  0.0668,  ...,  0.0219,  0.0628,  0.0569],
        [ 0.0381,  0.0267,  0.0087,  ..., -0.0168, -0.0154, -0.0133],
        [-0.0228, -0.0304,  0.0060,  ...,  0.0050, -0.0337, -0.0136],
        ...,
        [-0.1316, -0.1018, -0.0804,  ...,  0.0372,  0.0401,  0.0253],
        [ 0.0989,  0.0742,  0.0721,  ..., -0.0490, -0.0614, -0.0546],
        [ 0.0030,  0.0181, -0.0125,  ..., -0.0507, -0.0242, -0.0343]],
       device='cuda:0'), grad: tensor([[ 2.1183e-04,  5.8502e-05,  6.9916e-05,  ...,  1.9699e-05,
          5.2303e-05,  4.9174e-05],
        [-8.1635e-04, -2.5177e-04, -3.0231e-04,  ..., -5.6922e-05,
         -1.7893e-04, -1.6677e-04],
        [ 1.2612e-04,  2.9176e-05,  3.7014e-05,  ...,  2.0146e-05,
          3.7521e-05,  3.6120e-05],
        ...,
        [-2.9594e-05,  7.0855e-06,  7.8827e-06,  ..., -1.1660e-05,
         -1.6466e-05, -1.6406e-05],
        [ 7.7128e-05,  2.2039e-05,  2.6911e-05,  ...,  6.6794e-06,
          1.7837e-05,  1.6809e-05],
        [ 3.9768e-04,  1.2183e-04,  1.4615e-04,  ...,  2.8759e-05,
          8.8274e-05,  8.2433e-05]], device='cuda:0')
Epoch 35, bias, value: tensor([ 0.0468, -0.0135,  0.0256,  0.0136, -0.0524, -0.0108, -0.0047],
       device='cuda:0'), grad: tensor([ 4.9591e-04, -1.8606e-03,  3.1376e-04,  7.2539e-05, -1.0914e-04,
         1.8167e-04,  9.0694e-04], device='cuda:0')
588
0.005224324151752577
changing lr
epoch 34, time 809.73, cls_loss 0.0030 cls_loss_mapping 0.0131 cls_loss_causal 0.4321 re_mapping 0.0133 re_causal 0.0234 /// teacc 93.72 lr 0.00500000
Epoch 36, weight, value: tensor([[ 0.0792,  0.0726,  0.0671,  ...,  0.0215,  0.0620,  0.0561],
        [ 0.0384,  0.0266,  0.0090,  ..., -0.0166, -0.0150, -0.0129],
        [-0.0234, -0.0303,  0.0054,  ...,  0.0045, -0.0336, -0.0138],
        ...,
        [-0.1299, -0.1010, -0.0799,  ...,  0.0368,  0.0399,  0.0252],
        [ 0.0972,  0.0733,  0.0713,  ..., -0.0484, -0.0608, -0.0540],
        [ 0.0034,  0.0177, -0.0125,  ..., -0.0499, -0.0239, -0.0338]],
       device='cuda:0'), grad: tensor([[-3.4839e-05, -1.7911e-05, -1.8388e-05,  ..., -9.1717e-06,
         -1.5274e-05, -1.3106e-05],
        [-4.1604e-05, -1.1638e-05, -1.3761e-05,  ..., -7.8678e-06,
         -1.2323e-05, -1.2226e-05],
        [-2.9266e-05, -5.5470e-06, -6.1207e-06,  ..., -4.1202e-06,
         -8.7023e-06, -8.1137e-06],
        ...,
        [ 3.9935e-05,  1.6361e-05,  1.7539e-05,  ...,  9.4622e-06,
          1.5900e-05,  1.4067e-05],
        [ 9.3356e-06,  2.7418e-06,  3.0492e-06,  ...,  1.7481e-06,
          2.9411e-06,  2.8014e-06],
        [ 2.9460e-05,  9.2387e-06,  1.0267e-05,  ...,  5.3309e-06,
          9.0525e-06,  8.5756e-06]], device='cuda:0')
Epoch 36, bias, value: tensor([ 0.0453, -0.0114,  0.0233,  0.0128, -0.0499, -0.0123, -0.0032],
       device='cuda:0'), grad: tensor([-7.5638e-05, -9.5069e-05, -7.8142e-05,  6.7353e-05,  8.9824e-05,
         2.2471e-05,  6.9141e-05], device='cuda:0')
588
0.005000000000000003
changing lr
epoch 35, time 808.54, cls_loss 0.0037 cls_loss_mapping 0.0152 cls_loss_causal 0.4379 re_mapping 0.0126 re_causal 0.0233 /// teacc 92.71 lr 0.00477568
Epoch 37, weight, value: tensor([[ 0.0787,  0.0727,  0.0673,  ...,  0.0212,  0.0613,  0.0555],
        [ 0.0376,  0.0263,  0.0089,  ..., -0.0166, -0.0151, -0.0130],
        [-0.0220, -0.0300,  0.0053,  ...,  0.0042, -0.0331, -0.0135],
        ...,
        [-0.1292, -0.1005, -0.0797,  ...,  0.0367,  0.0396,  0.0252],
        [ 0.0957,  0.0724,  0.0704,  ..., -0.0479, -0.0602, -0.0536],
        [ 0.0029,  0.0175, -0.0124,  ..., -0.0495, -0.0241, -0.0338]],
       device='cuda:0'), grad: tensor([[ 1.1081e-04,  9.8497e-06,  2.6390e-05,  ...,  5.8085e-05,
          7.8619e-05,  7.5281e-05],
        [ 2.9993e-04,  3.1650e-05,  7.5817e-05,  ...,  1.5366e-04,
          2.0933e-04,  2.0015e-04],
        [ 6.7055e-05,  7.0333e-06,  1.7270e-05,  ...,  3.8087e-05,
          5.1022e-05,  4.8697e-05],
        ...,
        [ 3.2514e-05,  4.0419e-06,  8.8885e-06,  ...,  1.5900e-05,
          2.1756e-05,  2.0742e-05],
        [ 2.3872e-05,  2.6692e-06,  6.0759e-06,  ...,  1.1623e-05,
          1.6004e-05,  1.5318e-05],
        [ 9.1270e-06, -7.8883e-07,  1.5451e-06,  ...,  7.1563e-06,
          9.1866e-06,  8.8215e-06]], device='cuda:0')
Epoch 37, bias, value: tensor([ 0.0436, -0.0122,  0.0267,  0.0142, -0.0493, -0.0133, -0.0051],
       device='cuda:0'), grad: tensor([ 3.1781e-04,  8.4877e-04,  1.8942e-04, -1.5469e-03,  9.0361e-05,
         6.7234e-05,  3.1173e-05], device='cuda:0')
588
0.004775675848247429
changing lr
epoch 36, time 818.63, cls_loss 0.0022 cls_loss_mapping 0.0121 cls_loss_causal 0.4217 re_mapping 0.0124 re_causal 0.0234 /// teacc 93.47 lr 0.00455180
Epoch 38, weight, value: tensor([[ 0.0800,  0.0731,  0.0679,  ...,  0.0211,  0.0610,  0.0552],
        [ 0.0368,  0.0258,  0.0087,  ..., -0.0164, -0.0150, -0.0129],
        [-0.0223, -0.0300,  0.0047,  ...,  0.0039, -0.0329, -0.0136],
        ...,
        [-0.1282, -0.0996, -0.0792,  ...,  0.0364,  0.0393,  0.0251],
        [ 0.0949,  0.0719,  0.0699,  ..., -0.0473, -0.0595, -0.0529],
        [ 0.0025,  0.0170, -0.0125,  ..., -0.0490, -0.0238, -0.0334]],
       device='cuda:0'), grad: tensor([[-7.7605e-05, -4.7684e-05, -4.8161e-05,  ..., -6.0908e-07,
         -3.0603e-06, -4.1500e-06],
        [ 4.1366e-05,  2.1577e-05,  2.2545e-05,  ...,  2.4587e-06,
          4.7572e-06,  4.7386e-06],
        [-4.3094e-05, -7.4506e-06, -1.1407e-05,  ..., -1.2629e-05,
         -1.9446e-05, -1.6555e-05],
        ...,
        [ 9.5442e-06,  3.5148e-06,  4.1686e-06,  ...,  1.9278e-06,
          2.9113e-06,  2.4792e-06],
        [ 1.3210e-05,  6.4522e-06,  6.8471e-06,  ...,  8.5309e-07,
          1.6466e-06,  1.6224e-06],
        [ 2.2411e-05,  8.8289e-06,  1.0006e-05,  ...,  2.2650e-06,
          4.0941e-06,  3.7327e-06]], device='cuda:0')
Epoch 38, bias, value: tensor([ 0.0465, -0.0125,  0.0255,  0.0122, -0.0490, -0.0131, -0.0052],
       device='cuda:0'), grad: tensor([-8.8215e-05,  6.2108e-05, -1.2684e-04,  6.6817e-05,  2.1145e-05,
         2.1160e-05,  4.3809e-05], device='cuda:0')
588
0.004551803455482836
changing lr
epoch 37, time 812.52, cls_loss 0.0024 cls_loss_mapping 0.0128 cls_loss_causal 0.4274 re_mapping 0.0113 re_causal 0.0217 /// teacc 93.97 lr 0.00432883
Epoch 39, weight, value: tensor([[ 0.0801,  0.0732,  0.0682,  ...,  0.0209,  0.0604,  0.0547],
        [ 0.0360,  0.0253,  0.0084,  ..., -0.0163, -0.0149, -0.0129],
        [-0.0223, -0.0298,  0.0044,  ...,  0.0036, -0.0327, -0.0137],
        ...,
        [-0.1261, -0.0988, -0.0787,  ...,  0.0363,  0.0394,  0.0253],
        [ 0.0938,  0.0713,  0.0693,  ..., -0.0468, -0.0590, -0.0524],
        [ 0.0012,  0.0164, -0.0128,  ..., -0.0486, -0.0240, -0.0334]],
       device='cuda:0'), grad: tensor([[ 1.3208e-03,  2.7132e-04,  2.7966e-04,  ...,  2.1148e-04,
          3.4261e-04,  3.1471e-04],
        [ 4.4847e-04,  6.2883e-05,  8.6010e-05,  ...,  1.6391e-04,
          2.0814e-04,  1.9217e-04],
        [-2.8467e-04, -5.8651e-05, -8.5235e-05,  ..., -9.0823e-06,
         -5.7757e-05, -7.6950e-05],
        ...,
        [-1.1253e-03, -2.3699e-04, -2.0397e-04,  ..., -1.2493e-04,
         -2.4390e-04, -2.0444e-04],
        [ 1.0937e-04,  1.8716e-05,  2.3216e-05,  ...,  2.8864e-05,
          4.0889e-05,  3.8743e-05],
        [ 3.8415e-05, -2.9765e-06,  2.0452e-06,  ...,  1.7777e-05,
          2.2992e-05,  2.1994e-05]], device='cuda:0')
Epoch 39, bias, value: tensor([ 0.0464, -0.0130,  0.0248,  0.0131, -0.0451, -0.0139, -0.0080],
       device='cuda:0'), grad: tensor([ 0.0036,  0.0013, -0.0009, -0.0013, -0.0031,  0.0003,  0.0002],
       device='cuda:0')
588
0.004328833670911726
changing lr
epoch 38, time 816.59, cls_loss 0.0036 cls_loss_mapping 0.0166 cls_loss_causal 0.4544 re_mapping 0.0109 re_causal 0.0217 /// teacc 92.96 lr 0.00410722
Epoch 40, weight, value: tensor([[ 0.0799,  0.0733,  0.0683,  ...,  0.0205,  0.0596,  0.0540],
        [ 0.0372,  0.0255,  0.0089,  ..., -0.0160, -0.0144, -0.0124],
        [-0.0221, -0.0297,  0.0042,  ...,  0.0036, -0.0322, -0.0135],
        ...,
        [-0.1258, -0.0986, -0.0787,  ...,  0.0360,  0.0391,  0.0252],
        [ 0.0927,  0.0707,  0.0687,  ..., -0.0464, -0.0585, -0.0520],
        [ 0.0008,  0.0161, -0.0128,  ..., -0.0483, -0.0240, -0.0333]],
       device='cuda:0'), grad: tensor([[ 9.0182e-05,  7.5512e-06,  4.5821e-06,  ...,  1.2167e-05,
          2.6450e-05,  2.8521e-05],
        [ 5.2541e-05,  8.3819e-06,  1.0535e-05,  ...,  1.2502e-05,
          2.0817e-05,  1.9684e-05],
        [-3.5405e-04, -8.3625e-05, -8.7738e-05,  ..., -1.2740e-05,
         -4.4912e-05, -5.0306e-05],
        ...,
        [ 9.5487e-05,  4.2170e-05,  3.9607e-05,  ..., -3.0175e-05,
         -3.2872e-05, -2.7359e-05],
        [ 2.5019e-05,  5.5432e-06,  7.0781e-06,  ...,  4.3400e-06,
          7.1600e-06,  6.7949e-06],
        [ 4.0144e-05,  8.8736e-06,  1.1116e-05,  ...,  5.5581e-06,
          9.4995e-06,  9.3132e-06]], device='cuda:0')
Epoch 40, bias, value: tensor([ 0.0456, -0.0094,  0.0255,  0.0118, -0.0453, -0.0149, -0.0090],
       device='cuda:0'), grad: tensor([ 2.6417e-04,  1.5450e-04, -8.4782e-04,  1.2887e-04,  1.3530e-04,
         6.4671e-05,  1.0115e-04], device='cuda:0')
588
0.0041072155260068206
changing lr
epoch 39, time 810.16, cls_loss 0.0023 cls_loss_mapping 0.0103 cls_loss_causal 0.4366 re_mapping 0.0109 re_causal 0.0218 /// teacc 92.96 lr 0.00388740
Epoch 41, weight, value: tensor([[ 0.0813,  0.0739,  0.0691,  ...,  0.0206,  0.0597,  0.0541],
        [ 0.0351,  0.0247,  0.0081,  ..., -0.0162, -0.0148, -0.0127],
        [-0.0220, -0.0297,  0.0037,  ...,  0.0036, -0.0318, -0.0133],
        ...,
        [-0.1254, -0.0982, -0.0785,  ...,  0.0355,  0.0385,  0.0248],
        [ 0.0919,  0.0703,  0.0683,  ..., -0.0460, -0.0580, -0.0516],
        [ 0.0014,  0.0162, -0.0125,  ..., -0.0478, -0.0238, -0.0330]],
       device='cuda:0'), grad: tensor([[ 3.0786e-05,  6.2510e-06,  6.6385e-06,  ...,  6.4000e-06,
          1.0177e-05,  9.1493e-06],
        [ 5.1297e-06,  7.0129e-07,  9.4157e-07,  ...,  2.5406e-06,
          3.5912e-06,  3.3006e-06],
        [-1.6153e-04, -3.6359e-05, -4.1485e-05,  ..., -3.7223e-05,
         -5.7846e-05, -5.2512e-05],
        ...,
        [ 4.1604e-05,  9.7454e-06,  1.1414e-05,  ...,  4.6752e-06,
          9.4995e-06,  8.5384e-06],
        [ 5.7332e-06,  1.5143e-06,  1.7388e-06,  ...,  9.4809e-07,
          1.5926e-06,  1.4519e-06],
        [ 1.7926e-05,  4.1164e-06,  4.8168e-06,  ...,  3.5632e-06,
          5.8189e-06,  5.2750e-06]], device='cuda:0')
Epoch 41, bias, value: tensor([ 0.0489, -0.0132,  0.0257,  0.0117, -0.0461, -0.0150, -0.0077],
       device='cuda:0'), grad: tensor([ 8.4579e-05,  1.6034e-05, -4.2963e-04,  1.6403e-04,  1.0425e-04,
         1.4126e-05,  4.6611e-05], device='cuda:0')
588
0.0038873953302184317
changing lr
epoch 40, time 817.72, cls_loss 0.0031 cls_loss_mapping 0.0110 cls_loss_causal 0.4158 re_mapping 0.0106 re_causal 0.0213 /// teacc 93.72 lr 0.00366982
Epoch 42, weight, value: tensor([[ 0.0811,  0.0739,  0.0692,  ...,  0.0203,  0.0591,  0.0535],
        [ 0.0352,  0.0245,  0.0082,  ..., -0.0161, -0.0146, -0.0125],
        [-0.0223, -0.0296,  0.0034,  ...,  0.0033, -0.0318, -0.0134],
        ...,
        [-0.1249, -0.0977, -0.0782,  ...,  0.0354,  0.0383,  0.0248],
        [ 0.0923,  0.0700,  0.0680,  ..., -0.0454, -0.0573, -0.0509],
        [ 0.0008,  0.0158, -0.0126,  ..., -0.0474, -0.0237, -0.0328]],
       device='cuda:0'), grad: tensor([[-5.3596e-04, -1.5938e-04, -1.5640e-04,  ..., -1.0812e-04,
         -1.7965e-04, -1.6940e-04],
        [ 2.6524e-05,  8.6874e-06,  8.8885e-06,  ...,  5.1670e-06,
          8.2627e-06,  7.8455e-06],
        [ 3.5048e-04,  1.0979e-04,  1.0973e-04,  ...,  6.7711e-05,
          1.1122e-04,  1.0532e-04],
        ...,
        [ 5.9277e-05,  2.0459e-05,  2.0370e-05,  ...,  9.4101e-06,
          1.5780e-05,  1.4931e-05],
        [-2.8157e-04, -1.4019e-04, -1.5092e-04,  ..., -6.7241e-06,
         -9.8944e-06, -1.2219e-05],
        [ 2.2864e-04,  1.0979e-04,  1.1772e-04,  ...,  8.9556e-06,
          1.3918e-05,  1.5289e-05]], device='cuda:0')
Epoch 42, bias, value: tensor([ 0.0479, -0.0116,  0.0243,  0.0106, -0.0460, -0.0122, -0.0089],
       device='cuda:0'), grad: tensor([-1.2474e-03,  5.8174e-05,  7.9107e-04,  3.3116e-04,  1.2589e-04,
        -4.0770e-04,  3.4833e-04], device='cuda:0')
588
0.003669815772166629
changing lr
epoch 41, time 817.28, cls_loss 0.0020 cls_loss_mapping 0.0085 cls_loss_causal 0.4236 re_mapping 0.0098 re_causal 0.0200 /// teacc 92.71 lr 0.00345492
Epoch 43, weight, value: tensor([[ 0.0810,  0.0739,  0.0693,  ...,  0.0202,  0.0586,  0.0532],
        [ 0.0355,  0.0243,  0.0082,  ..., -0.0159, -0.0144, -0.0124],
        [-0.0226, -0.0296,  0.0030,  ...,  0.0031, -0.0316, -0.0135],
        ...,
        [-0.1245, -0.0972, -0.0780,  ...,  0.0350,  0.0379,  0.0244],
        [ 0.0916,  0.0696,  0.0677,  ..., -0.0450, -0.0569, -0.0506],
        [ 0.0011,  0.0156, -0.0125,  ..., -0.0469, -0.0234, -0.0324]],
       device='cuda:0'), grad: tensor([[ 5.7995e-05, -2.0489e-07,  3.4031e-06,  ...,  3.5316e-05,
          2.9922e-05,  3.0935e-05],
        [ 1.4651e-04,  1.6674e-05,  2.3216e-05,  ...,  5.4210e-05,
          5.6028e-05,  5.6356e-05],
        [ 2.0862e-04,  3.3289e-05,  4.1366e-05,  ...,  6.0529e-05,
          7.0274e-05,  6.9916e-05],
        ...,
        [ 1.3411e-04,  1.4596e-05,  2.4140e-05,  ...,  5.3406e-05,
          5.9366e-05,  6.0856e-05],
        [ 5.0902e-05,  5.4762e-06,  8.8289e-06,  ...,  2.0519e-05,
          2.2277e-05,  2.2873e-05],
        [-3.6061e-05, -1.9625e-05, -1.4506e-05,  ...,  1.0744e-05,
          6.2138e-06,  7.1488e-06]], device='cuda:0')
Epoch 43, bias, value: tensor([ 0.0477, -0.0095,  0.0231,  0.0105, -0.0466, -0.0127, -0.0083],
       device='cuda:0'), grad: tensor([ 1.6332e-04,  4.1175e-04,  5.6458e-04, -1.6212e-03,  3.8433e-04,
         1.4436e-04, -4.7505e-05], device='cuda:0')
588
0.0034549150281252667
changing lr
epoch 42, time 812.58, cls_loss 0.0022 cls_loss_mapping 0.0107 cls_loss_causal 0.4164 re_mapping 0.0095 re_causal 0.0194 /// teacc 93.47 lr 0.00324313
Epoch 44, weight, value: tensor([[ 0.0800,  0.0737,  0.0691,  ...,  0.0197,  0.0578,  0.0524],
        [ 0.0340,  0.0236,  0.0077,  ..., -0.0159, -0.0144, -0.0124],
        [-0.0221, -0.0294,  0.0029,  ...,  0.0031, -0.0312, -0.0132],
        ...,
        [-0.1233, -0.0966, -0.0776,  ...,  0.0348,  0.0378,  0.0244],
        [ 0.0914,  0.0695,  0.0676,  ..., -0.0447, -0.0565, -0.0502],
        [ 0.0014,  0.0155, -0.0124,  ..., -0.0465, -0.0231, -0.0320]],
       device='cuda:0'), grad: tensor([[-3.1042e-04, -1.0139e-04, -1.1402e-04,  ..., -6.8963e-05,
         -1.1438e-04, -1.0639e-04],
        [ 5.1528e-05,  1.2212e-05,  1.5438e-05,  ...,  1.2323e-05,
          2.0087e-05,  1.9476e-05],
        [ 3.5435e-05,  9.5963e-06,  1.1921e-05,  ...,  7.0035e-06,
          1.2219e-05,  1.1384e-05],
        ...,
        [ 4.0025e-05,  1.4938e-05,  1.6630e-05,  ...,  5.5954e-06,
          1.2115e-05,  1.0990e-05],
        [-9.4846e-06, -7.3239e-06, -6.9775e-06,  ...,  7.0781e-07,
          1.2163e-06,  1.0170e-06],
        [ 8.6725e-05,  4.3094e-05,  4.1217e-05,  ...,  2.2948e-05,
          3.1769e-05,  2.8774e-05]], device='cuda:0')
Epoch 44, bias, value: tensor([ 0.0451, -0.0115,  0.0242,  0.0110, -0.0447, -0.0125, -0.0075],
       device='cuda:0'), grad: tensor([-7.4959e-04,  1.3137e-04,  9.0301e-05,  2.6941e-04,  9.0837e-05,
        -4.8429e-06,  1.7333e-04], device='cuda:0')
588
0.0032431258795932905
changing lr
epoch 43, time 804.92, cls_loss 0.0019 cls_loss_mapping 0.0109 cls_loss_causal 0.4262 re_mapping 0.0091 re_causal 0.0198 /// teacc 93.47 lr 0.00303487
Epoch 45, weight, value: tensor([[ 0.0801,  0.0737,  0.0692,  ...,  0.0195,  0.0574,  0.0520],
        [ 0.0328,  0.0230,  0.0073,  ..., -0.0160, -0.0145, -0.0125],
        [-0.0228, -0.0296,  0.0025,  ...,  0.0029, -0.0311, -0.0133],
        ...,
        [-0.1231, -0.0964, -0.0774,  ...,  0.0345,  0.0374,  0.0242],
        [ 0.0925,  0.0699,  0.0679,  ..., -0.0442, -0.0558, -0.0496],
        [ 0.0019,  0.0154, -0.0123,  ..., -0.0459, -0.0228, -0.0316]],
       device='cuda:0'), grad: tensor([[-1.1873e-04, -5.6744e-05, -5.7548e-05,  ..., -4.7311e-06,
         -1.7673e-05, -1.6883e-05],
        [-4.3958e-05, -7.3984e-06, -9.5293e-06,  ..., -1.9416e-05,
         -2.5347e-05, -2.5064e-05],
        [ 6.7830e-05,  2.6092e-05,  2.6479e-05,  ...,  5.3681e-06,
          1.3813e-05,  1.3068e-05],
        ...,
        [ 5.3883e-05,  2.4945e-05,  2.5496e-05,  ...,  8.3223e-06,
          1.2457e-05,  1.2666e-05],
        [ 2.9638e-05,  1.0327e-05,  1.0647e-05,  ...,  4.1239e-06,
          7.9423e-06,  7.6257e-06],
        [-1.5184e-05, -6.3851e-06, -4.7423e-06,  ...,  1.8915e-06,
          8.3167e-07,  1.0170e-06]], device='cuda:0')
Epoch 45, bias, value: tensor([ 0.0453, -0.0129,  0.0222,  0.0099, -0.0452, -0.0095, -0.0056],
       device='cuda:0'), grad: tensor([-2.0373e-04, -1.1629e-04,  1.4174e-04,  5.9426e-05,  8.1062e-05,
         6.4790e-05, -2.6986e-05], device='cuda:0')
588
0.0030348748417303863
changing lr
epoch 44, time 804.95, cls_loss 0.0013 cls_loss_mapping 0.0102 cls_loss_causal 0.4315 re_mapping 0.0091 re_causal 0.0199 /// teacc 93.97 lr 0.00283058
Epoch 46, weight, value: tensor([[ 0.0796,  0.0736,  0.0691,  ...,  0.0193,  0.0569,  0.0516],
        [ 0.0330,  0.0230,  0.0074,  ..., -0.0159, -0.0143, -0.0123],
        [-0.0223, -0.0294,  0.0024,  ...,  0.0028, -0.0309, -0.0132],
        ...,
        [-0.1218, -0.0958, -0.0770,  ...,  0.0343,  0.0374,  0.0242],
        [ 0.0917,  0.0694,  0.0674,  ..., -0.0440, -0.0556, -0.0494],
        [ 0.0009,  0.0151, -0.0125,  ..., -0.0457, -0.0228, -0.0316]],
       device='cuda:0'), grad: tensor([[ 8.4102e-05,  1.2226e-05,  1.5602e-05,  ...,  2.7627e-05,
          3.4750e-05,  3.2485e-05],
        [ 4.2558e-05,  6.7316e-06,  8.8438e-06,  ...,  1.4171e-05,
          1.5810e-05,  1.5028e-05],
        [-5.0277e-05, -7.6517e-06, -8.5086e-06,  ..., -2.5943e-05,
         -3.3945e-05, -3.1084e-05],
        ...,
        [-1.1601e-05, -1.3327e-06, -7.1526e-07,  ..., -3.8464e-07,
         -3.6620e-06, -3.3043e-06],
        [ 2.0102e-05,  3.0212e-06,  3.9376e-06,  ...,  6.6571e-06,
          8.0541e-06,  7.5586e-06],
        [ 3.5107e-05,  6.1132e-06,  7.5251e-06,  ...,  1.1280e-05,
          1.3128e-05,  1.2472e-05]], device='cuda:0')
Epoch 46, bias, value: tensor([ 0.0443, -0.0117,  0.0232,  0.0096, -0.0433, -0.0102, -0.0078],
       device='cuda:0'), grad: tensor([ 2.4033e-04,  1.1915e-04, -2.0254e-04, -2.9182e-04, -2.3723e-05,
         5.7220e-05,  1.0157e-04], device='cuda:0')
588
0.0028305813044122124
changing lr
---------------------saving model at epoch 45----------------------------------------------------
epoch 45, time 802.96, cls_loss 0.0019 cls_loss_mapping 0.0101 cls_loss_causal 0.4210 re_mapping 0.0088 re_causal 0.0195 /// teacc 95.23 lr 0.00263066
Epoch 47, weight, value: tensor([[ 0.0792,  0.0736,  0.0691,  ...,  0.0191,  0.0565,  0.0512],
        [ 0.0332,  0.0229,  0.0075,  ..., -0.0157, -0.0141, -0.0122],
        [-0.0215, -0.0292,  0.0023,  ...,  0.0029, -0.0303, -0.0129],
        ...,
        [-0.1216, -0.0955, -0.0769,  ...,  0.0340,  0.0371,  0.0240],
        [ 0.0911,  0.0690,  0.0670,  ..., -0.0437, -0.0553, -0.0491],
        [ 0.0004,  0.0148, -0.0126,  ..., -0.0455, -0.0229, -0.0315]],
       device='cuda:0'), grad: tensor([[-1.8418e-05, -7.2941e-06, -5.7556e-06,  ...,  1.8589e-06,
          5.3830e-07, -3.7253e-08],
        [-7.3433e-05, -2.5347e-05, -2.9951e-05,  ..., -9.6112e-06,
         -1.9461e-05, -1.8835e-05],
        [ 3.6359e-05,  8.6427e-06,  9.7007e-06,  ...,  1.0766e-05,
          1.2040e-05,  1.2040e-05],
        ...,
        [ 5.8889e-05,  1.8105e-05,  2.0280e-05,  ...,  9.3952e-06,
          1.4931e-05,  1.5140e-05],
        [ 1.7866e-05,  5.5432e-06,  6.1654e-06,  ...,  3.0100e-06,
          4.7572e-06,  4.6492e-06],
        [ 2.5898e-05,  7.6815e-06,  8.3670e-06,  ...,  5.1856e-06,
          7.3612e-06,  7.0445e-06]], device='cuda:0')
Epoch 47, bias, value: tensor([ 0.0431, -0.0104,  0.0253,  0.0092, -0.0437, -0.0105, -0.0089],
       device='cuda:0'), grad: tensor([-3.0026e-05, -1.9670e-04,  8.9884e-05, -1.2034e-04,  1.3995e-04,
         4.6521e-05,  7.1287e-05], device='cuda:0')
588
0.0026306566876350096
changing lr
epoch 46, time 796.28, cls_loss 0.0014 cls_loss_mapping 0.0082 cls_loss_causal 0.4007 re_mapping 0.0084 re_causal 0.0181 /// teacc 93.97 lr 0.00243550
Epoch 48, weight, value: tensor([[ 0.0800,  0.0739,  0.0695,  ...,  0.0190,  0.0563,  0.0510],
        [ 0.0325,  0.0226,  0.0074,  ..., -0.0158, -0.0142, -0.0122],
        [-0.0215, -0.0291,  0.0021,  ...,  0.0028, -0.0302, -0.0128],
        ...,
        [-0.1211, -0.0952, -0.0767,  ...,  0.0339,  0.0370,  0.0240],
        [ 0.0904,  0.0686,  0.0666,  ..., -0.0434, -0.0550, -0.0489],
        [ 0.0002,  0.0146, -0.0126,  ..., -0.0452, -0.0228, -0.0314]],
       device='cuda:0'), grad: tensor([[-9.3728e-06, -8.9779e-06, -1.1541e-05,  ..., -3.2373e-06,
         -2.8387e-06, -2.4326e-06],
        [-2.7701e-05, -5.1409e-06, -5.3905e-06,  ..., -1.6578e-06,
         -6.9477e-06, -6.6012e-06],
        [-2.6628e-05, -4.9174e-06, -5.9009e-06,  ..., -5.7742e-07,
         -5.8301e-07, -1.8738e-06],
        ...,
        [ 1.3180e-05,  5.2154e-06,  5.7928e-06,  ...,  8.4750e-07,
         -2.2538e-07, -2.4959e-07],
        [ 1.4670e-05,  4.6529e-06,  5.8711e-06,  ...,  2.4792e-06,
          3.8445e-06,  4.0159e-06],
        [ 2.3812e-05,  6.0014e-06,  7.0669e-06,  ...,  2.8089e-06,
          5.7369e-06,  5.8338e-06]], device='cuda:0')
Epoch 48, bias, value: tensor([ 0.0447, -0.0115,  0.0252,  0.0092, -0.0433, -0.0108, -0.0094],
       device='cuda:0'), grad: tensor([ 3.0175e-06, -8.2195e-05, -6.7174e-05,  3.0354e-05,  2.5079e-05,
         3.1263e-05,  5.9664e-05], device='cuda:0')
588
0.0024355036129704724
changing lr
epoch 47, time 792.94, cls_loss 0.0016 cls_loss_mapping 0.0101 cls_loss_causal 0.4051 re_mapping 0.0084 re_causal 0.0183 /// teacc 94.97 lr 0.00224552
Epoch 49, weight, value: tensor([[ 0.0804,  0.0740,  0.0697,  ...,  0.0189,  0.0560,  0.0508],
        [ 0.0324,  0.0225,  0.0073,  ..., -0.0157, -0.0141, -0.0122],
        [-0.0214, -0.0291,  0.0019,  ...,  0.0027, -0.0300, -0.0128],
        ...,
        [-0.1210, -0.0949, -0.0766,  ...,  0.0338,  0.0368,  0.0239],
        [ 0.0896,  0.0683,  0.0662,  ..., -0.0432, -0.0548, -0.0487],
        [ 0.0002,  0.0144, -0.0126,  ..., -0.0449, -0.0227, -0.0312]],
       device='cuda:0'), grad: tensor([[-4.8965e-05, -4.8161e-05, -3.5852e-05,  ...,  9.9897e-05,
          1.1665e-04,  1.0639e-04],
        [ 3.0613e-04,  4.1783e-05,  6.2168e-05,  ...,  1.5008e-04,
          1.8573e-04,  1.8001e-04],
        [ 1.8191e-04,  2.0936e-05,  4.0591e-05,  ...,  1.1367e-04,
          1.3721e-04,  1.3256e-04],
        ...,
        [-3.9673e-04, -1.5929e-05, -6.0380e-05,  ..., -3.6240e-04,
         -4.4727e-04, -4.2868e-04],
        [ 8.0705e-05,  1.0327e-05,  1.5706e-05,  ...,  3.7074e-05,
          4.5300e-05,  4.4018e-05],
        [ 7.8142e-05,  1.2510e-05,  1.6153e-05,  ...,  2.5809e-05,
          3.1680e-05,  3.1292e-05]], device='cuda:0')
Epoch 49, bias, value: tensor([ 0.0453, -0.0111,  0.0253,  0.0092, -0.0439, -0.0116, -0.0092],
       device='cuda:0'), grad: tensor([ 1.4640e-05,  8.7214e-04,  5.3501e-04, -5.8937e-04, -1.2760e-03,
         2.3127e-04,  2.1350e-04], device='cuda:0')
588
0.00224551509273949
changing lr
epoch 48, time 794.79, cls_loss 0.0011 cls_loss_mapping 0.0088 cls_loss_causal 0.3953 re_mapping 0.0084 re_causal 0.0184 /// teacc 93.47 lr 0.00206107
Epoch 50, weight, value: tensor([[ 0.0810,  0.0742,  0.0699,  ...,  0.0188,  0.0559,  0.0507],
        [ 0.0323,  0.0223,  0.0073,  ..., -0.0156, -0.0140, -0.0120],
        [-0.0219, -0.0292,  0.0016,  ...,  0.0026, -0.0300, -0.0129],
        ...,
        [-0.1206, -0.0946, -0.0764,  ...,  0.0336,  0.0366,  0.0238],
        [ 0.0894,  0.0681,  0.0661,  ..., -0.0430, -0.0545, -0.0484],
        [-0.0002,  0.0142, -0.0127,  ..., -0.0447, -0.0226, -0.0312]],
       device='cuda:0'), grad: tensor([[ 3.0398e-04,  5.4359e-05,  5.4806e-05,  ...,  4.3452e-05,
          8.6904e-05,  8.8394e-05],
        [ 1.2064e-04,  2.2903e-05,  2.6226e-05,  ...,  3.2097e-05,
          4.6730e-05,  4.1187e-05],
        [ 9.8765e-05,  1.6436e-05,  2.0295e-05,  ...,  2.4199e-05,
          3.8147e-05,  3.3438e-05],
        ...,
        [-5.0640e-04, -8.4877e-05, -8.9824e-05,  ..., -8.7798e-05,
         -1.6844e-04, -1.5855e-04],
        [ 4.8965e-05,  8.0839e-06,  9.2164e-06,  ...,  1.0423e-05,
          1.8060e-05,  1.6093e-05],
        [ 2.6956e-05,  4.8652e-06,  5.8636e-06,  ...,  6.7279e-06,
          9.9689e-06,  8.8587e-06]], device='cuda:0')
Epoch 50, bias, value: tensor([ 0.0467, -0.0107,  0.0239,  0.0093, -0.0439, -0.0116, -0.0097],
       device='cuda:0'), grad: tensor([ 8.0776e-04,  3.3808e-04,  2.8300e-04, -2.2209e-04, -1.4238e-03,
         1.4186e-04,  7.4983e-05], device='cuda:0')
588
0.002061073738537637
changing lr
epoch 49, time 791.10, cls_loss 0.0014 cls_loss_mapping 0.0094 cls_loss_causal 0.4036 re_mapping 0.0082 re_causal 0.0180 /// teacc 95.23 lr 0.00188255
Epoch 51, weight, value: tensor([[ 8.0744e-02,  7.4140e-02,  6.9905e-02,  ...,  1.8685e-02,
          5.5529e-02,  5.0404e-02],
        [ 3.1746e-02,  2.2061e-02,  7.1410e-03,  ..., -1.5593e-02,
         -1.3971e-02, -1.2023e-02],
        [-2.2164e-02, -2.9223e-02,  1.3682e-03,  ...,  2.5026e-03,
         -2.9860e-02, -1.2878e-02],
        ...,
        [-1.1973e-01, -9.4191e-02, -7.6101e-02,  ...,  3.3470e-02,
          3.6480e-02,  2.3775e-02],
        [ 8.9263e-02,  6.8041e-02,  6.5975e-02,  ..., -4.2811e-02,
         -5.4277e-02, -4.8199e-02],
        [-3.9018e-05,  1.4104e-02, -1.2632e-02,  ..., -4.4504e-02,
         -2.2524e-02, -3.0979e-02]], device='cuda:0'), grad: tensor([[ 6.5947e-04,  1.4544e-04,  1.5628e-04,  ...,  1.0252e-04,
          1.5032e-04,  1.4842e-04],
        [-2.4295e-04, -9.3803e-06, -1.9312e-05,  ..., -1.9222e-05,
         -4.0501e-05, -4.5538e-05],
        [-2.1343e-03, -4.5037e-04, -4.8351e-04,  ..., -4.4036e-04,
         -6.3276e-04, -5.9366e-04],
        ...,
        [ 8.5306e-04,  1.6356e-04,  1.7834e-04,  ...,  1.7262e-04,
          2.5058e-04,  2.3615e-04],
        [ 2.0075e-04,  3.8862e-05,  4.3094e-05,  ...,  3.9220e-05,
          5.7250e-05,  5.3853e-05],
        [ 3.0589e-04,  5.1260e-05,  5.7101e-05,  ...,  6.2466e-05,
          9.1553e-05,  8.6486e-05]], device='cuda:0')
Epoch 51, bias, value: tensor([ 0.0460, -0.0112,  0.0234,  0.0090, -0.0426, -0.0114, -0.0092],
       device='cuda:0'), grad: tensor([ 0.0016, -0.0008, -0.0052,  0.0009,  0.0022,  0.0005,  0.0008],
       device='cuda:0')
588
0.0018825509907063344
changing lr
epoch 50, time 795.94, cls_loss 0.0011 cls_loss_mapping 0.0068 cls_loss_causal 0.4129 re_mapping 0.0080 re_causal 0.0179 /// teacc 94.97 lr 0.00171031
Epoch 52, weight, value: tensor([[ 0.0813,  0.0744,  0.0702,  ...,  0.0186,  0.0554,  0.0503],
        [ 0.0316,  0.0219,  0.0071,  ..., -0.0156, -0.0139, -0.0120],
        [-0.0220, -0.0292,  0.0012,  ...,  0.0025, -0.0297, -0.0128],
        ...,
        [-0.1198, -0.0940, -0.0760,  ...,  0.0332,  0.0362,  0.0235],
        [ 0.0887,  0.0678,  0.0657,  ..., -0.0427, -0.0541, -0.0481],
        [-0.0004,  0.0139, -0.0127,  ..., -0.0444, -0.0225, -0.0309]],
       device='cuda:0'), grad: tensor([[ 1.1241e-04,  1.8403e-05,  2.2218e-05,  ...,  9.9093e-06,
          1.7971e-05,  1.4402e-05],
        [ 6.7770e-05,  1.4529e-05,  1.5959e-05,  ...,  8.6650e-06,
          1.3560e-05,  1.1742e-05],
        [-4.6802e-04, -1.0204e-04, -1.1444e-04,  ..., -4.8071e-05,
         -8.4698e-05, -7.0930e-05],
        ...,
        [ 1.3340e-04,  3.3885e-05,  3.7134e-05,  ...,  2.2039e-05,
          3.2812e-05,  2.8864e-05],
        [ 3.4034e-05,  7.4208e-06,  8.2627e-06,  ...,  3.6862e-06,
          6.2957e-06,  5.3234e-06],
        [ 7.0393e-05,  1.4640e-05,  1.6451e-05,  ...,  6.6310e-06,
          1.1899e-05,  9.8571e-06]], device='cuda:0')
Epoch 52, bias, value: tensor([ 0.0470, -0.0109,  0.0237,  0.0099, -0.0437, -0.0121, -0.0100],
       device='cuda:0'), grad: tensor([ 2.9802e-04,  1.7095e-04, -1.1902e-03,  1.1951e-04,  3.3593e-04,
         8.6248e-05,  1.7893e-04], device='cuda:0')
588
0.0017103063703014388
changing lr
epoch 51, time 792.04, cls_loss 0.0011 cls_loss_mapping 0.0071 cls_loss_causal 0.3982 re_mapping 0.0078 re_causal 0.0171 /// teacc 94.22 lr 0.00154469
Epoch 53, weight, value: tensor([[ 0.0812,  0.0743,  0.0702,  ...,  0.0185,  0.0551,  0.0501],
        [ 0.0315,  0.0219,  0.0071,  ..., -0.0155, -0.0139, -0.0119],
        [-0.0223, -0.0292,  0.0011,  ...,  0.0024, -0.0296, -0.0128],
        ...,
        [-0.1192, -0.0937, -0.0758,  ...,  0.0331,  0.0361,  0.0235],
        [ 0.0883,  0.0675,  0.0655,  ..., -0.0425, -0.0539, -0.0479],
        [-0.0003,  0.0138, -0.0127,  ..., -0.0441, -0.0224, -0.0308]],
       device='cuda:0'), grad: tensor([[ 1.5008e-04,  5.9873e-05,  6.0588e-05,  ...,  1.9684e-05,
          2.8446e-05,  2.8625e-05],
        [ 7.9155e-05,  2.2858e-05,  2.1115e-05,  ...,  1.0148e-05,
          1.1407e-05,  1.2696e-05],
        [ 2.7150e-05,  3.9190e-05,  3.9279e-05,  ..., -2.3127e-05,
         -2.2963e-05, -2.4423e-05],
        ...,
        [ 4.2111e-05,  2.3544e-05,  2.2352e-05,  ..., -1.0498e-05,
         -1.4298e-05, -1.1802e-05],
        [-1.4181e-03, -6.0606e-04, -6.0272e-04,  ..., -9.5427e-05,
         -1.4818e-04, -1.5831e-04],
        [ 6.4373e-04,  2.7061e-04,  2.6965e-04,  ...,  4.9829e-05,
          7.5281e-05,  7.9691e-05]], device='cuda:0')
Epoch 53, bias, value: tensor([ 0.0467, -0.0108,  0.0230,  0.0096, -0.0428, -0.0121, -0.0098],
       device='cuda:0'), grad: tensor([ 2.8634e-04,  1.4770e-04, -2.5302e-05,  8.5211e-04,  1.3009e-05,
        -2.3708e-03,  1.0977e-03], device='cuda:0')
588
0.0015446867550656784
changing lr
epoch 52, time 795.64, cls_loss 0.0012 cls_loss_mapping 0.0080 cls_loss_causal 0.4011 re_mapping 0.0074 re_causal 0.0164 /// teacc 94.72 lr 0.00138603
Epoch 54, weight, value: tensor([[ 0.0809,  0.0743,  0.0701,  ...,  0.0183,  0.0548,  0.0498],
        [ 0.0313,  0.0217,  0.0071,  ..., -0.0155, -0.0138, -0.0119],
        [-0.0224, -0.0292,  0.0009,  ...,  0.0023, -0.0295, -0.0128],
        ...,
        [-0.1189, -0.0935, -0.0757,  ...,  0.0330,  0.0360,  0.0235],
        [ 0.0875,  0.0672,  0.0651,  ..., -0.0424, -0.0538, -0.0478],
        [ 0.0007,  0.0141, -0.0123,  ..., -0.0439, -0.0222, -0.0305]],
       device='cuda:0'), grad: tensor([[ 5.2261e-04,  1.0097e-04,  1.3101e-04,  ...,  6.6817e-05,
          1.1981e-04,  1.1861e-04],
        [-1.7338e-03, -2.5797e-04, -3.6550e-04,  ..., -4.3845e-04,
         -7.2861e-04, -6.7663e-04],
        [ 2.9945e-04,  2.9072e-05,  5.4806e-05,  ...,  1.1230e-04,
          1.8954e-04,  1.6987e-04],
        ...,
        [ 2.6298e-04,  4.7296e-05,  6.0678e-05,  ...,  4.7565e-05,
          7.9811e-05,  7.6592e-05],
        [ 6.3241e-05,  1.1340e-05,  1.4648e-05,  ...,  1.3143e-05,
          2.2039e-05,  2.0698e-05],
        [ 3.9071e-05,  6.1700e-07,  3.9712e-06,  ...,  1.5177e-05,
          2.3156e-05,  2.1845e-05]], device='cuda:0')
Epoch 54, bias, value: tensor([ 0.0460, -0.0107,  0.0226,  0.0094, -0.0424, -0.0133, -0.0077],
       device='cuda:0'), grad: tensor([ 0.0013, -0.0054,  0.0011,  0.0019,  0.0007,  0.0002,  0.0001],
       device='cuda:0')
588
0.001386025680863044
changing lr
epoch 53, time 793.18, cls_loss 0.0015 cls_loss_mapping 0.0074 cls_loss_causal 0.4159 re_mapping 0.0073 re_causal 0.0167 /// teacc 95.23 lr 0.00123464
Epoch 55, weight, value: tensor([[ 0.0808,  0.0742,  0.0700,  ...,  0.0183,  0.0547,  0.0497],
        [ 0.0314,  0.0217,  0.0072,  ..., -0.0154, -0.0137, -0.0118],
        [-0.0221, -0.0291,  0.0009,  ...,  0.0024, -0.0293, -0.0127],
        ...,
        [-0.1188, -0.0934, -0.0757,  ...,  0.0329,  0.0359,  0.0234],
        [ 0.0875,  0.0670,  0.0650,  ..., -0.0422, -0.0536, -0.0476],
        [ 0.0005,  0.0141, -0.0122,  ..., -0.0438, -0.0222, -0.0305]],
       device='cuda:0'), grad: tensor([[ 9.7215e-05,  3.1769e-05,  3.2514e-05,  ...,  6.7428e-06,
          1.2338e-05,  1.2912e-05],
        [-1.1349e-04, -5.8800e-05, -5.8651e-05,  ..., -9.7454e-06,
         -1.0319e-05, -1.2718e-05],
        [ 3.9130e-05,  6.7838e-06,  8.2403e-06,  ...,  3.8221e-06,
          1.0595e-05,  9.8199e-06],
        ...,
        [-1.5461e-04, -1.5020e-05, -2.1026e-05,  ..., -1.6153e-05,
         -4.4107e-05, -4.0203e-05],
        [ 4.2558e-05,  7.0110e-06,  8.5458e-06,  ...,  4.8019e-06,
          1.1727e-05,  1.0826e-05],
        [ 5.9634e-05,  1.9446e-05,  2.0742e-05,  ...,  6.4783e-06,
          1.2368e-05,  1.2167e-05]], device='cuda:0')
Epoch 55, bias, value: tensor([ 0.0458, -0.0104,  0.0234,  0.0088, -0.0427, -0.0127, -0.0083],
       device='cuda:0'), grad: tensor([ 1.7858e-04, -1.4818e-04,  1.0729e-04,  6.5923e-05, -4.4537e-04,
         1.1533e-04,  1.2648e-04], device='cuda:0')
588
0.0012346426699819469
changing lr
epoch 54, time 790.30, cls_loss 0.0012 cls_loss_mapping 0.0074 cls_loss_causal 0.4004 re_mapping 0.0072 re_causal 0.0161 /// teacc 91.46 lr 0.00109084
Epoch 56, weight, value: tensor([[ 0.0806,  0.0741,  0.0700,  ...,  0.0182,  0.0546,  0.0496],
        [ 0.0316,  0.0217,  0.0073,  ..., -0.0154, -0.0137, -0.0117],
        [-0.0220, -0.0291,  0.0008,  ...,  0.0023, -0.0292, -0.0126],
        ...,
        [-0.1187, -0.0933, -0.0756,  ...,  0.0328,  0.0357,  0.0233],
        [ 0.0872,  0.0669,  0.0649,  ..., -0.0421, -0.0535, -0.0475],
        [ 0.0006,  0.0140, -0.0122,  ..., -0.0436, -0.0221, -0.0304]],
       device='cuda:0'), grad: tensor([[ 1.4591e-04,  3.2336e-05,  3.3647e-05,  ...,  1.9282e-05,
          3.5226e-05,  3.2991e-05],
        [-2.3139e-04, -5.2631e-05, -5.8651e-05,  ..., -1.4164e-05,
         -5.5045e-05, -5.0008e-05],
        [ 1.2636e-04,  2.8759e-05,  2.8357e-05,  ...,  2.2590e-05,
          2.8297e-05,  2.6703e-05],
        ...,
        [ 5.3108e-05,  1.2226e-05,  1.3150e-05,  ...,  4.5523e-06,
          1.2390e-05,  1.1265e-05],
        [ 3.8326e-05,  8.5011e-06,  9.1940e-06,  ...,  3.8184e-06,
          9.3132e-06,  8.6427e-06],
        [ 6.7830e-05,  1.5043e-05,  1.6212e-05,  ...,  7.2420e-06,
          1.6958e-05,  1.5751e-05]], device='cuda:0')
Epoch 56, bias, value: tensor([ 0.0456, -0.0096,  0.0236,  0.0082, -0.0428, -0.0130, -0.0080],
       device='cuda:0'), grad: tensor([ 0.0004, -0.0008,  0.0003, -0.0005,  0.0002,  0.0001,  0.0002],
       device='cuda:0')
588
0.0010908425876598518
changing lr
epoch 55, time 791.04, cls_loss 0.0013 cls_loss_mapping 0.0064 cls_loss_causal 0.4006 re_mapping 0.0070 re_causal 0.0157 /// teacc 92.96 lr 0.00095492
Epoch 57, weight, value: tensor([[ 0.0808,  0.0742,  0.0701,  ...,  0.0182,  0.0545,  0.0495],
        [ 0.0315,  0.0217,  0.0073,  ..., -0.0154, -0.0136, -0.0117],
        [-0.0221, -0.0291,  0.0007,  ...,  0.0023, -0.0292, -0.0127],
        ...,
        [-0.1184, -0.0931, -0.0755,  ...,  0.0327,  0.0357,  0.0233],
        [ 0.0870,  0.0668,  0.0647,  ..., -0.0420, -0.0534, -0.0474],
        [ 0.0003,  0.0139, -0.0123,  ..., -0.0435, -0.0222, -0.0304]],
       device='cuda:0'), grad: tensor([[-1.2696e-05, -2.6584e-05, -2.5883e-05,  ...,  2.2784e-05,
          1.7941e-05,  1.9476e-05],
        [ 3.1024e-05, -3.7383e-06, -1.1204e-06,  ...,  2.1696e-05,
          2.3231e-05,  2.0608e-05],
        [ 7.4089e-05,  8.7768e-06,  1.0565e-05,  ...,  2.9594e-05,
          3.3528e-05,  3.1263e-05],
        ...,
        [ 1.1605e-04,  3.8505e-05,  4.2140e-05,  ...,  2.1189e-05,
          3.2544e-05,  2.7776e-05],
        [ 2.7940e-05,  4.8168e-06,  5.5768e-06,  ...,  9.3654e-06,
          1.1064e-05,  1.0468e-05],
        [ 4.4197e-05,  7.8306e-06,  8.4937e-06,  ...,  1.2912e-05,
          1.5251e-05,  1.4573e-05]], device='cuda:0')
Epoch 57, bias, value: tensor([ 0.0457, -0.0097,  0.0232,  0.0087, -0.0423, -0.0132, -0.0086],
       device='cuda:0'), grad: tensor([ 2.4304e-05,  1.1349e-04,  2.1267e-04, -8.1396e-04,  2.6894e-04,
         7.5221e-05,  1.1951e-04], device='cuda:0')
588
0.000954915028125264
changing lr
epoch 56, time 792.00, cls_loss 0.0014 cls_loss_mapping 0.0058 cls_loss_causal 0.3947 re_mapping 0.0070 re_causal 0.0153 /// teacc 93.97 lr 0.00082713
Epoch 58, weight, value: tensor([[ 0.0812,  0.0743,  0.0703,  ...,  0.0182,  0.0545,  0.0495],
        [ 0.0314,  0.0216,  0.0072,  ..., -0.0153, -0.0136, -0.0117],
        [-0.0222, -0.0291,  0.0006,  ...,  0.0023, -0.0291, -0.0126],
        ...,
        [-0.1183, -0.0930, -0.0755,  ...,  0.0326,  0.0356,  0.0232],
        [ 0.0867,  0.0667,  0.0646,  ..., -0.0420, -0.0533, -0.0474],
        [ 0.0002,  0.0138, -0.0123,  ..., -0.0435, -0.0221, -0.0303]],
       device='cuda:0'), grad: tensor([[ 2.2018e-04,  4.3094e-05,  4.2945e-05,  ...,  3.9011e-05,
          6.7592e-05,  6.3181e-05],
        [ 8.5890e-05,  1.3866e-05,  1.4685e-05,  ...,  2.1249e-05,
          3.2306e-05,  3.1620e-05],
        [-4.9263e-05, -8.6352e-06, -7.9125e-06,  ..., -1.0364e-05,
         -1.6361e-05, -1.6630e-05],
        ...,
        [-2.1291e-04, -1.9759e-05, -2.5615e-05,  ..., -6.2048e-05,
         -1.0628e-04, -9.5904e-05],
        [ 2.1315e-04,  7.1287e-05,  6.4909e-05,  ...,  1.3031e-05,
          1.5482e-05,  2.2084e-05],
        [-3.4785e-04, -1.2279e-04, -1.1086e-04,  ..., -1.4894e-05,
         -1.4380e-05, -2.6733e-05]], device='cuda:0')
Epoch 58, bias, value: tensor([ 0.0465, -0.0097,  0.0230,  0.0087, -0.0424, -0.0135, -0.0087],
       device='cuda:0'), grad: tensor([ 0.0006,  0.0002, -0.0001,  0.0002, -0.0007,  0.0004, -0.0007],
       device='cuda:0')
588
0.0008271337313934874
changing lr
epoch 57, time 785.47, cls_loss 0.0014 cls_loss_mapping 0.0066 cls_loss_causal 0.3959 re_mapping 0.0070 re_causal 0.0155 /// teacc 93.72 lr 0.00070776
Epoch 59, weight, value: tensor([[ 8.1379e-02,  7.4439e-02,  7.0417e-02,  ...,  1.8152e-02,
          5.4398e-02,  4.9417e-02],
        [ 3.1237e-02,  2.1533e-02,  7.2049e-03,  ..., -1.5339e-02,
         -1.3612e-02, -1.1715e-02],
        [-2.2203e-02, -2.9107e-02,  5.2069e-04,  ...,  2.2342e-03,
         -2.9068e-02, -1.2622e-02],
        ...,
        [-1.1797e-01, -9.2928e-02, -7.5440e-02,  ...,  3.2604e-02,
          3.5572e-02,  2.3227e-02],
        [ 8.6427e-02,  6.6554e-02,  6.4479e-02,  ..., -4.1890e-02,
         -5.3248e-02, -4.7302e-02],
        [ 3.9185e-05,  1.3749e-02, -1.2354e-02,  ..., -4.3392e-02,
         -2.2147e-02, -3.0321e-02]], device='cuda:0'), grad: tensor([[-1.1235e-04, -3.5435e-05, -3.2008e-05,  ..., -6.7614e-06,
         -2.2739e-05, -2.2516e-05],
        [ 2.6822e-07, -1.0356e-05, -1.0908e-05,  ...,  5.3830e-06,
          3.4962e-06,  6.2659e-06],
        [ 2.7850e-05,  5.6587e-06,  7.6592e-06,  ...,  5.3123e-06,
          9.2909e-06,  9.0823e-06],
        ...,
        [ 8.6054e-07,  1.7315e-05,  9.5293e-06,  ..., -1.4305e-05,
         -1.3761e-05, -1.5251e-05],
        [ 2.3812e-05,  5.7258e-06,  6.6236e-06,  ...,  3.6899e-06,
          7.1824e-06,  6.9141e-06],
        [ 2.9907e-05,  8.1360e-06,  9.1270e-06,  ...,  3.8054e-06,
          8.0764e-06,  7.7859e-06]], device='cuda:0')
Epoch 59, bias, value: tensor([ 0.0468, -0.0098,  0.0229,  0.0086, -0.0417, -0.0137, -0.0092],
       device='cuda:0'), grad: tensor([-2.5082e-04,  2.2411e-05,  7.2360e-05,  6.9439e-05, -4.3809e-05,
         5.9098e-05,  7.1585e-05], device='cuda:0')
588
0.00070775603199067
changing lr
epoch 58, time 790.13, cls_loss 0.0015 cls_loss_mapping 0.0079 cls_loss_causal 0.3930 re_mapping 0.0070 re_causal 0.0153 /// teacc 92.71 lr 0.00059702
Epoch 60, weight, value: tensor([[ 0.0813,  0.0744,  0.0704,  ...,  0.0181,  0.0543,  0.0493],
        [ 0.0312,  0.0215,  0.0072,  ..., -0.0153, -0.0136, -0.0117],
        [-0.0220, -0.0291,  0.0005,  ...,  0.0022, -0.0290, -0.0126],
        ...,
        [-0.1179, -0.0928, -0.0754,  ...,  0.0325,  0.0355,  0.0232],
        [ 0.0865,  0.0665,  0.0644,  ..., -0.0418, -0.0531, -0.0472],
        [-0.0002,  0.0137, -0.0124,  ..., -0.0433, -0.0222, -0.0303]],
       device='cuda:0'), grad: tensor([[-1.3614e-04, -4.7654e-05, -4.9800e-05,  ..., -9.0376e-06,
         -2.8431e-05, -3.0428e-05],
        [ 2.2963e-05, -1.4622e-07,  5.2713e-07,  ...,  7.8231e-06,
          9.9093e-06,  9.8273e-06],
        [ 2.0003e-04,  4.9949e-05,  5.4389e-05,  ...,  3.1143e-05,
          4.8310e-05,  4.9323e-05],
        ...,
        [-3.0428e-05,  3.3360e-06,  2.5891e-06,  ..., -6.1467e-06,
         -7.1004e-06, -7.6443e-06],
        [ 1.8820e-05,  4.6901e-06,  5.1148e-06,  ...,  3.0342e-06,
          4.5076e-06,  4.5747e-06],
        [ 3.3736e-05,  9.0003e-06,  9.7901e-06,  ...,  5.9679e-06,
          8.0764e-06,  8.0466e-06]], device='cuda:0')
Epoch 60, bias, value: tensor([ 0.0465, -0.0098,  0.0233,  0.0084, -0.0418, -0.0132, -0.0096],
       device='cuda:0'), grad: tensor([-2.8539e-04,  7.4804e-05,  4.8685e-04, -2.9206e-04, -1.1086e-04,
         4.5896e-05,  8.0407e-05], device='cuda:0')
588
0.0005970223407163104
changing lr
epoch 59, time 787.39, cls_loss 0.0013 cls_loss_mapping 0.0065 cls_loss_causal 0.4018 re_mapping 0.0070 re_causal 0.0152 /// teacc 91.46 lr 0.00049516
Epoch 61, weight, value: tensor([[ 0.0813,  0.0745,  0.0705,  ...,  0.0181,  0.0542,  0.0493],
        [ 0.0313,  0.0215,  0.0072,  ..., -0.0153, -0.0135, -0.0116],
        [-0.0220, -0.0291,  0.0005,  ...,  0.0022, -0.0289, -0.0125],
        ...,
        [-0.1179, -0.0928, -0.0754,  ...,  0.0324,  0.0354,  0.0231],
        [ 0.0865,  0.0664,  0.0644,  ..., -0.0417, -0.0530, -0.0471],
        [-0.0003,  0.0136, -0.0124,  ..., -0.0433, -0.0222, -0.0303]],
       device='cuda:0'), grad: tensor([[ 6.6161e-05,  2.5108e-05,  2.2292e-05,  ..., -2.0247e-06,
          2.7250e-06,  2.9895e-06],
        [ 1.3210e-05,  1.6913e-05,  1.7300e-05,  ..., -6.8434e-06,
         -1.0744e-05, -8.6725e-06],
        [ 5.3674e-05,  3.2604e-05,  3.2365e-05,  ...,  5.1893e-06,
          6.1728e-06,  6.3814e-06],
        ...,
        [-1.4462e-05,  5.9456e-06,  9.0003e-06,  ...,  2.1271e-06,
         -2.9467e-06, -3.6284e-06],
        [-7.7200e-04, -3.9530e-04, -4.0317e-04,  ..., -4.7147e-05,
         -6.9857e-05, -7.2420e-05],
        [ 4.9162e-04,  2.4366e-04,  2.4891e-04,  ...,  3.4362e-05,
          5.1647e-05,  5.2601e-05]], device='cuda:0')
Epoch 61, bias, value: tensor([ 0.0466, -0.0095,  0.0235,  0.0083, -0.0421, -0.0130, -0.0101],
       device='cuda:0'), grad: tensor([ 1.3614e-04, -1.2711e-05,  6.5029e-05,  2.9325e-04, -7.4446e-05,
        -1.1940e-03,  7.8773e-04], device='cuda:0')
588
0.0004951556604879052
changing lr
epoch 60, time 786.59, cls_loss 0.0013 cls_loss_mapping 0.0053 cls_loss_causal 0.4019 re_mapping 0.0070 re_causal 0.0155 /// teacc 93.22 lr 0.00040236
Epoch 62, weight, value: tensor([[ 0.0815,  0.0745,  0.0705,  ...,  0.0181,  0.0542,  0.0493],
        [ 0.0313,  0.0215,  0.0073,  ..., -0.0153, -0.0135, -0.0116],
        [-0.0220, -0.0291,  0.0004,  ...,  0.0022, -0.0289, -0.0125],
        ...,
        [-0.1179, -0.0927, -0.0754,  ...,  0.0324,  0.0353,  0.0230],
        [ 0.0863,  0.0664,  0.0643,  ..., -0.0417, -0.0530, -0.0471],
        [-0.0004,  0.0135, -0.0125,  ..., -0.0433, -0.0222, -0.0303]],
       device='cuda:0'), grad: tensor([[-1.4246e-04, -1.9640e-05, -3.0264e-05,  ..., -6.9857e-05,
         -1.2910e-04, -1.1581e-04],
        [-1.0908e-05, -6.3218e-06, -5.1782e-06,  ...,  2.3730e-06,
          3.9525e-06,  3.3639e-06],
        [ 2.5630e-05,  3.5446e-06,  5.0627e-06,  ...,  1.0781e-05,
          2.0429e-05,  1.8463e-05],
        ...,
        [ 3.6359e-05,  7.1079e-06,  9.3430e-06,  ...,  1.7747e-05,
          3.2604e-05,  2.9057e-05],
        [ 1.3679e-05,  2.3674e-06,  3.1535e-06,  ...,  5.2191e-06,
          9.7379e-06,  8.8066e-06],
        [ 2.7865e-05,  5.3830e-06,  6.8247e-06,  ...,  1.0662e-05,
          1.9848e-05,  1.7881e-05]], device='cuda:0')
Epoch 62, bias, value: tensor([ 0.0468, -0.0092,  0.0235,  0.0086, -0.0424, -0.0132, -0.0102],
       device='cuda:0'), grad: tensor([-5.7840e-04,  9.1270e-07,  9.4593e-05,  1.9586e-04,  1.4210e-04,
         4.8250e-05,  9.6798e-05], device='cuda:0')
588
0.00040236113724274745
changing lr
---------------------saving model at epoch 61----------------------------------------------------
epoch 61, time 798.02, cls_loss 0.0014 cls_loss_mapping 0.0060 cls_loss_causal 0.3726 re_mapping 0.0069 re_causal 0.0147 /// teacc 95.73 lr 0.00031883
Epoch 63, weight, value: tensor([[ 0.0815,  0.0745,  0.0706,  ...,  0.0181,  0.0542,  0.0492],
        [ 0.0313,  0.0215,  0.0073,  ..., -0.0153, -0.0135, -0.0116],
        [-0.0219, -0.0290,  0.0004,  ...,  0.0022, -0.0288, -0.0125],
        ...,
        [-0.1179, -0.0927, -0.0753,  ...,  0.0323,  0.0353,  0.0230],
        [ 0.0863,  0.0663,  0.0642,  ..., -0.0416, -0.0529, -0.0470],
        [-0.0005,  0.0135, -0.0125,  ..., -0.0432, -0.0222, -0.0303]],
       device='cuda:0'), grad: tensor([[-4.3929e-05, -3.4690e-05, -4.0025e-05,  ..., -3.5614e-06,
         -5.4650e-06, -4.4927e-06],
        [ 5.9128e-05,  2.5079e-05,  2.8685e-05,  ...,  1.2778e-05,
          1.7628e-05,  1.6913e-05],
        [-8.0585e-05, -1.8463e-05, -1.8775e-05,  ..., -5.4836e-05,
         -6.4671e-05, -6.2823e-05],
        ...,
        [-6.9857e-05, -3.1013e-06, -4.7386e-06,  ...,  1.1753e-06,
         -6.4522e-06, -6.5900e-06],
        [ 2.0623e-05,  6.4410e-06,  7.3016e-06,  ...,  4.1425e-06,
          6.3218e-06,  6.1989e-06],
        [ 3.7521e-05,  1.0610e-05,  1.1824e-05,  ...,  1.3597e-05,
          1.7479e-05,  1.6838e-05]], device='cuda:0')
Epoch 63, bias, value: tensor([ 0.0468, -0.0093,  0.0235,  0.0086, -0.0424, -0.0130, -0.0104],
       device='cuda:0'), grad: tensor([-1.0625e-05,  1.1456e-04, -2.4128e-04,  2.0921e-04, -2.1255e-04,
         4.9680e-05,  9.0718e-05], device='cuda:0')
588
0.00031882564680131423
changing lr
epoch 62, time 791.56, cls_loss 0.0014 cls_loss_mapping 0.0056 cls_loss_causal 0.3860 re_mapping 0.0069 re_causal 0.0145 /// teacc 94.97 lr 0.00024472
Epoch 64, weight, value: tensor([[ 0.0815,  0.0746,  0.0706,  ...,  0.0180,  0.0541,  0.0492],
        [ 0.0313,  0.0215,  0.0073,  ..., -0.0153, -0.0135, -0.0116],
        [-0.0220, -0.0290,  0.0004,  ...,  0.0022, -0.0288, -0.0125],
        ...,
        [-0.1178, -0.0926, -0.0753,  ...,  0.0323,  0.0353,  0.0230],
        [ 0.0862,  0.0663,  0.0642,  ..., -0.0416, -0.0529, -0.0470],
        [-0.0006,  0.0135, -0.0125,  ..., -0.0432, -0.0222, -0.0303]],
       device='cuda:0'), grad: tensor([[ 4.8518e-04,  1.4091e-04,  1.4400e-04,  ...,  5.1767e-05,
          1.0324e-04,  1.0949e-04],
        [-1.6241e-03, -4.9543e-04, -4.9257e-04,  ..., -1.2589e-04,
         -2.7442e-04, -3.0184e-04],
        [-9.9719e-05, -6.7130e-06, -1.5736e-05,  ..., -2.8729e-05,
         -4.7058e-05, -3.9220e-05],
        ...,
        [ 1.5032e-04,  4.4167e-05,  4.5151e-05,  ...,  1.7181e-05,
          3.1292e-05,  3.0518e-05],
        [ 7.5758e-05,  2.1458e-05,  2.2143e-05,  ...,  7.0743e-06,
          1.4804e-05,  1.5438e-05],
        [ 2.1255e-04,  7.0274e-05,  7.1526e-05,  ...,  1.7896e-05,
          3.7909e-05,  3.9816e-05]], device='cuda:0')
Epoch 64, bias, value: tensor([ 0.0468, -0.0092,  0.0235,  0.0086, -0.0423, -0.0131, -0.0105],
       device='cuda:0'), grad: tensor([ 0.0011, -0.0037, -0.0003,  0.0019,  0.0003,  0.0002,  0.0005],
       device='cuda:0')
588
0.0002447174185242325
changing lr
epoch 63, time 792.66, cls_loss 0.0014 cls_loss_mapping 0.0055 cls_loss_causal 0.3893 re_mapping 0.0069 re_causal 0.0150 /// teacc 92.96 lr 0.00018019
Epoch 65, weight, value: tensor([[ 0.0815,  0.0746,  0.0706,  ...,  0.0180,  0.0541,  0.0491],
        [ 0.0312,  0.0214,  0.0073,  ..., -0.0153, -0.0135, -0.0116],
        [-0.0220, -0.0290,  0.0003,  ...,  0.0022, -0.0288, -0.0125],
        ...,
        [-0.1177, -0.0926, -0.0753,  ...,  0.0323,  0.0352,  0.0230],
        [ 0.0862,  0.0662,  0.0642,  ..., -0.0416, -0.0529, -0.0470],
        [-0.0006,  0.0135, -0.0125,  ..., -0.0432, -0.0222, -0.0303]],
       device='cuda:0'), grad: tensor([[ 8.9228e-05,  1.2606e-05,  1.9163e-05,  ...,  2.3171e-05,
          3.9995e-05,  3.6895e-05],
        [ 2.4647e-05,  4.6715e-06,  5.2266e-06,  ...,  1.1235e-05,
          1.3925e-05,  1.3247e-05],
        [-1.8505e-06, -1.0058e-06, -1.2983e-06,  ...,  3.5521e-06,
          4.2692e-06,  3.7104e-06],
        ...,
        [-1.0407e-04, -1.2234e-05, -2.0415e-05,  ..., -3.8534e-05,
         -6.0707e-05, -5.6356e-05],
        [ 2.6360e-05,  9.4771e-06,  9.5293e-06,  ...,  3.1702e-06,
          6.7726e-06,  6.0052e-06],
        [-2.6450e-05, -1.1444e-05, -1.0967e-05,  ...,  1.0896e-07,
         -3.2596e-06, -2.4661e-06]], device='cuda:0')
Epoch 65, bias, value: tensor([ 0.0467, -0.0093,  0.0235,  0.0087, -0.0422, -0.0130, -0.0106],
       device='cuda:0'), grad: tensor([ 2.6202e-04,  7.3969e-05,  4.8243e-06, -1.3702e-05, -3.3689e-04,
         5.6893e-05, -4.6700e-05], device='cuda:0')
588
0.0001801856965207339
changing lr
epoch 64, time 789.94, cls_loss 0.0014 cls_loss_mapping 0.0060 cls_loss_causal 0.3751 re_mapping 0.0069 re_causal 0.0144 /// teacc 94.47 lr 0.00012536
Epoch 66, weight, value: tensor([[ 0.0815,  0.0746,  0.0706,  ...,  0.0180,  0.0541,  0.0491],
        [ 0.0312,  0.0214,  0.0072,  ..., -0.0152, -0.0135, -0.0116],
        [-0.0220, -0.0290,  0.0003,  ...,  0.0022, -0.0288, -0.0125],
        ...,
        [-0.1177, -0.0926, -0.0753,  ...,  0.0323,  0.0352,  0.0230],
        [ 0.0862,  0.0662,  0.0642,  ..., -0.0415, -0.0528, -0.0469],
        [-0.0006,  0.0135, -0.0125,  ..., -0.0432, -0.0222, -0.0303]],
       device='cuda:0'), grad: tensor([[ 2.0289e-04,  4.7714e-05,  4.6968e-05,  ...,  2.2978e-05,
          5.6028e-05,  5.4181e-05],
        [-1.6952e-04, -6.3956e-05, -7.1347e-05,  ..., -3.0875e-05,
         -5.2601e-05, -5.0157e-05],
        [-2.3925e-04, -3.6091e-05, -2.8566e-05,  ..., -1.7762e-05,
         -6.2704e-05, -6.1035e-05],
        ...,
        [ 5.8234e-05,  1.5043e-05,  1.5035e-05,  ...,  6.8173e-06,
          1.5691e-05,  1.5207e-05],
        [ 3.3915e-05,  7.8082e-06,  7.6219e-06,  ...,  3.8110e-06,
          9.4622e-06,  9.1419e-06],
        [ 4.7415e-05,  1.1221e-05,  1.1034e-05,  ...,  5.9307e-06,
          1.4342e-05,  1.3761e-05]], device='cuda:0')
Epoch 66, bias, value: tensor([ 0.0466, -0.0093,  0.0235,  0.0087, -0.0421, -0.0130, -0.0106],
       device='cuda:0'), grad: tensor([ 5.1594e-04, -3.5167e-04, -6.7663e-04,  1.6046e-04,  1.4317e-04,
         8.6904e-05,  1.2153e-04], device='cuda:0')
588
0.000125360439090882
changing lr
epoch 65, time 792.08, cls_loss 0.0013 cls_loss_mapping 0.0047 cls_loss_causal 0.4086 re_mapping 0.0070 re_causal 0.0149 /// teacc 94.47 lr 0.00008035
Epoch 67, weight, value: tensor([[ 0.0815,  0.0746,  0.0706,  ...,  0.0180,  0.0541,  0.0491],
        [ 0.0312,  0.0214,  0.0072,  ..., -0.0152, -0.0135, -0.0116],
        [-0.0220, -0.0290,  0.0003,  ...,  0.0022, -0.0288, -0.0125],
        ...,
        [-0.1176, -0.0926, -0.0752,  ...,  0.0323,  0.0352,  0.0230],
        [ 0.0861,  0.0662,  0.0641,  ..., -0.0415, -0.0528, -0.0469],
        [-0.0006,  0.0134, -0.0125,  ..., -0.0432, -0.0222, -0.0303]],
       device='cuda:0'), grad: tensor([[ 8.9169e-05,  1.9014e-05,  2.5526e-05,  ...,  8.1286e-06,
          2.1189e-05,  2.0862e-05],
        [-7.4327e-05, -1.5467e-05, -2.2471e-05,  ..., -1.4305e-06,
         -1.2703e-05, -1.2293e-05],
        [ 6.7241e-06,  6.5006e-07,  7.9162e-07,  ..., -1.1548e-06,
         -6.4261e-07, -3.6508e-07],
        ...,
        [-1.8179e-05,  2.3656e-07, -1.3821e-06,  ..., -5.5097e-06,
         -9.0227e-06, -9.2089e-06],
        [ 1.1787e-05,  3.2131e-06,  3.6694e-06,  ...,  1.5590e-06,
          3.1739e-06,  3.1292e-06],
        [-2.1517e-05, -9.9093e-06, -8.9705e-06,  ..., -1.8366e-06,
         -3.9525e-06, -4.0568e-06]], device='cuda:0')
Epoch 67, bias, value: tensor([ 0.0467, -0.0093,  0.0235,  0.0086, -0.0421, -0.0130, -0.0106],
       device='cuda:0'), grad: tensor([ 2.2435e-04, -1.8561e-04,  1.4521e-05,  1.4886e-05, -6.1214e-05,
         2.7463e-05, -3.4809e-05], device='cuda:0')
588
8.03520570068517e-05
changing lr
epoch 66, time 791.60, cls_loss 0.0012 cls_loss_mapping 0.0048 cls_loss_causal 0.3824 re_mapping 0.0069 re_causal 0.0144 /// teacc 92.71 lr 0.00004525
Epoch 68, weight, value: tensor([[ 0.0815,  0.0746,  0.0706,  ...,  0.0180,  0.0540,  0.0491],
        [ 0.0312,  0.0214,  0.0072,  ..., -0.0152, -0.0135, -0.0116],
        [-0.0220, -0.0290,  0.0003,  ...,  0.0022, -0.0288, -0.0125],
        ...,
        [-0.1176, -0.0926, -0.0752,  ...,  0.0323,  0.0352,  0.0230],
        [ 0.0861,  0.0662,  0.0641,  ..., -0.0415, -0.0528, -0.0469],
        [-0.0006,  0.0134, -0.0125,  ..., -0.0432, -0.0222, -0.0303]],
       device='cuda:0'), grad: tensor([[-3.4499e-04, -1.0622e-04, -1.1998e-04,  ..., -5.3972e-05,
         -8.1837e-05, -8.7082e-05],
        [ 3.6597e-05,  1.3076e-05,  1.3515e-05,  ..., -1.1846e-05,
         -9.7007e-06, -7.8529e-06],
        [ 2.3949e-04,  6.9976e-05,  7.8917e-05,  ...,  4.0054e-05,
          6.0946e-05,  6.2346e-05],
        ...,
        [ 8.7678e-05,  2.6032e-05,  2.8893e-05,  ...,  1.6496e-05,
          2.5570e-05,  2.4974e-05],
        [ 6.9022e-05,  2.1338e-05,  2.3648e-05,  ...,  9.4622e-06,
          1.6078e-05,  1.6108e-05],
        [-8.6308e-05, -3.0845e-05, -3.1888e-05,  ...,  2.0768e-06,
         -9.3579e-06, -7.0520e-06]], device='cuda:0')
Epoch 68, bias, value: tensor([ 0.0467, -0.0093,  0.0235,  0.0086, -0.0420, -0.0130, -0.0106],
       device='cuda:0'), grad: tensor([-7.2241e-04,  5.9903e-05,  5.2357e-04, -2.9445e-05,  1.9670e-04,
         1.4806e-04, -1.7786e-04], device='cuda:0')
588
4.5251191160326525e-05
changing lr
epoch 67, time 790.33, cls_loss 0.0011 cls_loss_mapping 0.0054 cls_loss_causal 0.4041 re_mapping 0.0069 re_causal 0.0150 /// teacc 94.72 lr 0.00002013
Epoch 69, weight, value: tensor([[ 0.0815,  0.0746,  0.0706,  ...,  0.0180,  0.0540,  0.0491],
        [ 0.0312,  0.0214,  0.0072,  ..., -0.0152, -0.0135, -0.0116],
        [-0.0220, -0.0290,  0.0003,  ...,  0.0022, -0.0288, -0.0125],
        ...,
        [-0.1176, -0.0925, -0.0752,  ...,  0.0323,  0.0352,  0.0230],
        [ 0.0861,  0.0662,  0.0641,  ..., -0.0415, -0.0528, -0.0469],
        [-0.0006,  0.0134, -0.0125,  ..., -0.0432, -0.0222, -0.0303]],
       device='cuda:0'), grad: tensor([[ 3.5071e-04,  1.1247e-04,  1.2410e-04,  ...,  2.8133e-05,
          6.1333e-05,  5.7489e-05],
        [-6.8474e-04, -2.2268e-04, -2.4223e-04,  ..., -5.7817e-05,
         -1.1975e-04, -1.1617e-04],
        [-9.4235e-05, -2.9683e-05, -3.3438e-05,  ..., -1.4827e-05,
         -2.4989e-05, -2.4036e-05],
        ...,
        [ 1.2064e-04,  3.9995e-05,  4.2647e-05,  ...,  1.1832e-05,
          2.3708e-05,  2.3514e-05],
        [ 5.0396e-05,  1.6794e-05,  1.8194e-05,  ...,  4.7646e-06,
          9.4473e-06,  9.3356e-06],
        [ 1.3649e-04,  4.4614e-05,  4.8399e-05,  ...,  1.2234e-05,
          2.4438e-05,  2.3961e-05]], device='cuda:0')
Epoch 69, bias, value: tensor([ 0.0467, -0.0093,  0.0234,  0.0086, -0.0420, -0.0131, -0.0106],
       device='cuda:0'), grad: tensor([ 0.0008, -0.0015, -0.0002,  0.0003,  0.0003,  0.0001,  0.0003],
       device='cuda:0')
588
2.0128530023804673e-05
changing lr
epoch 68, time 787.75, cls_loss 0.0012 cls_loss_mapping 0.0053 cls_loss_causal 0.3859 re_mapping 0.0069 re_causal 0.0147 /// teacc 94.47 lr 0.00000503
Epoch 70, weight, value: tensor([[ 0.0815,  0.0746,  0.0706,  ...,  0.0180,  0.0540,  0.0491],
        [ 0.0312,  0.0214,  0.0072,  ..., -0.0152, -0.0135, -0.0116],
        [-0.0220, -0.0290,  0.0003,  ...,  0.0022, -0.0288, -0.0125],
        ...,
        [-0.1176, -0.0925, -0.0752,  ...,  0.0323,  0.0352,  0.0230],
        [ 0.0861,  0.0662,  0.0641,  ..., -0.0415, -0.0528, -0.0469],
        [-0.0006,  0.0134, -0.0125,  ..., -0.0431, -0.0222, -0.0303]],
       device='cuda:0'), grad: tensor([[-1.6153e-04, -5.3585e-05, -5.9724e-05,  ..., -3.2395e-05,
         -5.6535e-05, -5.7220e-05],
        [ 8.6248e-05,  2.2650e-05,  2.6375e-05,  ...,  2.1592e-05,
          3.2693e-05,  3.2932e-05],
        [ 4.7624e-05,  1.4886e-05,  1.6645e-05,  ...,  1.0297e-05,
          1.6779e-05,  1.7002e-05],
        ...,
        [ 4.1991e-05,  1.4745e-05,  1.6540e-05,  ...,  7.4059e-06,
          1.2279e-05,  1.2465e-05],
        [-8.6069e-05, -4.3333e-05, -4.7565e-05,  ..., -1.9930e-06,
         -4.6790e-06, -6.3106e-06],
        [ 7.8261e-05,  3.6418e-05,  4.0323e-05,  ...,  4.9211e-06,
          8.8066e-06,  9.9540e-06]], device='cuda:0')
Epoch 70, bias, value: tensor([ 0.0467, -0.0093,  0.0234,  0.0086, -0.0420, -0.0131, -0.0106],
       device='cuda:0'), grad: tensor([-3.4833e-04,  1.9848e-04,  1.0222e-04, -4.1157e-05,  8.4817e-05,
        -1.1569e-04,  1.1945e-04], device='cuda:0')
588
5.034667293427056e-06
changing lr
epoch 69, time 799.21, cls_loss 0.0015 cls_loss_mapping 0.0062 cls_loss_causal 0.3843 re_mapping 0.0069 re_causal 0.0145 /// teacc 93.47 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//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1', 'source_domain': 'sketch', 'svpath': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-PACS//sketch/CA_multiple_16fa_v2_ep70_lr0.01_cosine_base0.01_bs6_lamCa_1_lamRe1_adt4_cls1_EW2_70_rmTrue_rnTrue_str5_WithStyleAttackExp1/sketch_16factor_last_test_check.csv', 'factor_num': 16, 'epoch': 'last', 'stride': 5, 'eval_mapping': False, 'network': 'resnet18'}
-------------------------------------loading pretrain weights----------------------------------
loading weight of last
randm: False
stride: 5
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
loading weight of last
columns: ['sketch', 'art_painting', 'cartoon', 'photo']
/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/data/PACS/sketch_test.hdf5 torch.Size([3929, 3, 227, 227]) torch.Size([3929])
/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/photo_test.hdf5 torch.Size([1670, 3, 227, 227]) torch.Size([1670])
                        sketch  art_painting    cartoon      photo        Avg
w/o do (original x)  99.338254     43.212891  56.186007  46.467066  48.621988
      sketch  art_painting    cartoon      photo        Avg
do  99.28735     46.630859  60.025597  52.035928  52.897462