/home/yuqian_fu here1 here2 {'gpu': '0', 'data': 'mnist', 'ntr': None, 'translate': None, 'autoaug': 'CA_multiple', 'n': 3, 'stride': 3, 'factor_num': 14, 'epochs': 500, 'nbatch': 100, 'batchsize': 32, 'lr': 0.0001, 'lr_scheduler': 'Step', 'svroot': '/data/work-gcp-europe-west4-a/yuqian_fu/datasets/SingleSourceDG/saved-digit/CA_multiple_14fa_all_ep500_lr1e-4_lr_schedulerStep0.8_bs32_lamCa_1_lamRe_1_cls1_adt2_EW2_100_rmTrue_rnTrue_str3_WithStyleAttackExp1', 'clsadapt': True, 'lambda_causal': 1.0, 'lambda_re': 1.0, 'randm': True, 'randn': True, 'network': 'resnet18'} stride: 3 --------------------------CA_multiple-------------------------- ---------------------------14 factors----------------- randm: True randn: True n: 3 randm: False Epoch 1, weight, value: tensor([[ 0.0271, 0.0278, 0.0225, ..., -0.0296, -0.0126, 0.0133], [ 0.0281, -0.0014, 0.0019, ..., 0.0049, -0.0001, -0.0167], [ 0.0096, -0.0095, -0.0208, ..., -0.0112, 0.0086, 0.0155], ..., [ 0.0241, 0.0234, -0.0310, ..., -0.0023, 0.0109, -0.0108], [-0.0104, 0.0021, 0.0085, ..., -0.0012, -0.0071, -0.0235], [ 0.0238, -0.0081, 0.0041, ..., -0.0056, 0.0007, 0.0226]], device='cuda:0'), grad: None Epoch 1, bias, value: tensor([ 0.0091, 0.0009, 0.0029, 0.0177, 0.0243, 0.0294, -0.0206, 0.0056, -0.0307, -0.0182], device='cuda:0'), grad: None 100 0.0001 changing lr ---------------------saving model at epoch 0---------------------------------------------------- epoch 0, time 223.22, cls_loss 1.6736 cls_loss_mapping 2.0181 cls_loss_causal 2.2459 re_mapping 0.0816 re_causal 0.0818 /// teacc 73.46 lr 0.00010000 Epoch 2, weight, value: tensor([[ 0.0191, 0.0323, 0.0264, ..., -0.0359, -0.0128, 0.0174], [ 0.0360, -0.0041, 0.0007, ..., 0.0067, -0.0003, -0.0264], [ 0.0048, -0.0081, -0.0186, ..., -0.0090, 0.0080, 0.0122], ..., [ 0.0229, 0.0276, -0.0377, ..., -0.0113, 0.0093, -0.0189], [-0.0084, -0.0019, 0.0107, ..., 0.0022, -0.0053, -0.0228], [ 0.0194, -0.0109, -0.0004, ..., -0.0055, -0.0014, 0.0232]], device='cuda:0'), grad: tensor([[ 2.0035e-02, -1.3054e-02, -1.5045e-02, ..., 7.2479e-03, 1.3113e-04, -1.1559e-02], [-6.2988e-02, 2.7943e-03, -1.8673e-03, ..., -4.6967e-02, 8.8736e-06, -1.4580e-02], [ 2.7679e-02, 3.3607e-03, 1.9638e-02, ..., 2.5070e-02, 1.5184e-05, 1.3123e-02], ..., [-2.3865e-02, -1.8097e-02, 7.0152e-03, ..., -5.5695e-03, 3.0845e-05, 2.3499e-03], [-3.2921e-03, 7.3509e-03, -9.6130e-04, ..., -2.1133e-03, 1.0669e-04, 1.1459e-02], [ 5.2582e-02, 1.3435e-02, 3.2013e-02, ..., 7.6172e-02, 9.8348e-05, 3.1250e-02]], device='cuda:0') Epoch 2, bias, value: tensor([ 0.0069, 0.0036, 0.0026, 0.0171, 0.0236, 0.0295, -0.0215, 0.0064, -0.0314, -0.0176], device='cuda:0'), grad: tensor([ 0.0031, -0.0215, 0.0242, 0.0185, -0.0174, -0.0472, -0.0401, -0.0387, 0.0231, 0.0958], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 1---------------------------------------------------- epoch 1, time 220.40, cls_loss 0.5839 cls_loss_mapping 0.8899 cls_loss_causal 1.9196 re_mapping 0.2132 re_causal 0.2487 /// teacc 88.40 lr 0.00010000 Epoch 3, weight, value: tensor([[ 1.5019e-02, 3.5892e-02, 2.8696e-02, ..., -3.9882e-02, -1.5826e-02, 1.8511e-02], [ 3.8934e-02, -5.4322e-03, -1.9550e-03, ..., 8.0060e-03, -4.0296e-04, -2.8752e-02], [ 5.3168e-05, -6.9198e-03, -1.8001e-02, ..., -1.0121e-02, 6.0515e-04, 9.4711e-03], ..., [ 2.0476e-02, 3.2725e-02, -4.1894e-02, ..., -1.5440e-02, 8.8445e-03, -2.1740e-02], [-7.3172e-03, -5.6156e-03, 1.3275e-02, ..., 5.6482e-03, -1.0707e-02, -2.2726e-02], [ 1.8082e-02, -1.1282e-02, -1.0400e-03, ..., -9.5011e-03, -4.7546e-03, 2.3836e-02]], device='cuda:0'), grad: tensor([[-0.0201, -0.0018, -0.0141, ..., -0.0006, -0.0001, -0.0125], [-0.0225, 0.0020, 0.0056, ..., -0.0228, -0.0046, -0.0055], [-0.0170, 0.0011, -0.0056, ..., -0.0105, -0.0029, 0.0010], ..., [ 0.0154, -0.0052, 0.0065, ..., 0.0234, 0.0011, 0.0092], [ 0.0093, -0.0120, -0.0126, ..., -0.0094, 0.0024, -0.0142], [-0.0535, 0.0031, -0.0219, ..., -0.0405, -0.0099, -0.0302]], device='cuda:0') Epoch 3, bias, value: tensor([ 0.0060, 0.0037, 0.0024, 0.0170, 0.0235, 0.0311, -0.0226, 0.0060, -0.0307, -0.0176], device='cuda:0'), grad: tensor([-0.0033, -0.0064, -0.0138, 0.0236, 0.0052, 0.0347, 0.0149, 0.0178, -0.0105, -0.0623], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 2---------------------------------------------------- epoch 2, time 221.40, cls_loss 0.3490 cls_loss_mapping 0.5111 cls_loss_causal 1.6934 re_mapping 0.1598 re_causal 0.2406 /// teacc 93.28 lr 0.00010000 Epoch 4, weight, value: tensor([[ 0.0128, 0.0385, 0.0302, ..., -0.0418, -0.0186, 0.0193], [ 0.0411, -0.0065, -0.0034, ..., 0.0094, 0.0028, -0.0307], [-0.0016, -0.0057, -0.0179, ..., -0.0114, -0.0039, 0.0077], ..., [ 0.0176, 0.0352, -0.0448, ..., -0.0192, 0.0066, -0.0247], [-0.0074, -0.0083, 0.0139, ..., 0.0072, -0.0168, -0.0230], [ 0.0178, -0.0127, 0.0003, ..., -0.0107, -0.0075, 0.0257]], device='cuda:0'), grad: tensor([[ 2.1667e-03, -1.6556e-03, -8.8406e-04, ..., 2.3785e-03, -8.8811e-05, 1.1215e-03], [-3.7903e-02, 9.2793e-04, -2.2621e-03, ..., -2.9175e-02, -5.3635e-03, -1.1063e-03], [ 1.2840e-02, 2.3148e-02, 3.6743e-02, ..., 8.3160e-03, 1.0786e-03, 1.8982e-02], ..., [ 8.4782e-04, -2.5757e-02, -1.1475e-02, ..., 5.2986e-03, -8.3566e-05, 2.4052e-03], [ 2.3956e-03, 2.5101e-03, -2.2717e-03, ..., 9.8228e-05, 4.2915e-03, -4.1389e-03], [ 2.4929e-03, 7.8659e-03, -1.2722e-03, ..., 2.6436e-03, 1.9798e-03, 3.5858e-03]], device='cuda:0') Epoch 4, bias, value: tensor([ 0.0061, 0.0039, 0.0026, 0.0171, 0.0231, 0.0317, -0.0228, 0.0055, -0.0311, -0.0173], device='cuda:0'), grad: tensor([ 0.0012, -0.0161, 0.0436, -0.0147, -0.0163, 0.0077, 0.0087, -0.0279, 0.0087, 0.0052], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 3---------------------------------------------------- epoch 3, time 226.06, cls_loss 0.2774 cls_loss_mapping 0.3965 cls_loss_causal 1.5685 re_mapping 0.1232 re_causal 0.2210 /// teacc 94.53 lr 0.00010000 Epoch 5, weight, value: tensor([[ 0.0110, 0.0411, 0.0317, ..., -0.0440, -0.0202, 0.0198], [ 0.0435, -0.0077, -0.0047, ..., 0.0119, 0.0073, -0.0326], [-0.0040, -0.0042, -0.0183, ..., -0.0126, -0.0063, 0.0055], ..., [ 0.0148, 0.0371, -0.0472, ..., -0.0220, 0.0069, -0.0272], [-0.0070, -0.0103, 0.0147, ..., 0.0082, -0.0222, -0.0230], [ 0.0170, -0.0142, 0.0010, ..., -0.0117, -0.0113, 0.0266]], device='cuda:0'), grad: tensor([[ 0.0053, -0.0012, 0.0075, ..., 0.0070, 0.0017, 0.0064], [ 0.0100, 0.0007, 0.0122, ..., 0.0086, 0.0056, 0.0073], [ 0.0088, 0.0049, 0.0098, ..., 0.0096, 0.0011, 0.0059], ..., [ 0.0021, -0.0108, 0.0007, ..., 0.0030, 0.0003, -0.0003], [-0.0210, -0.0032, -0.0249, ..., -0.0383, -0.0084, -0.0205], [-0.0124, 0.0045, -0.0043, ..., -0.0104, 0.0005, 0.0046]], device='cuda:0') Epoch 5, bias, value: tensor([ 0.0063, 0.0044, 0.0029, 0.0172, 0.0232, 0.0316, -0.0234, 0.0052, -0.0313, -0.0174], device='cuda:0'), grad: tensor([ 0.0081, 0.0140, 0.0146, 0.0090, 0.0138, -0.0153, 0.0077, -0.0058, -0.0338, -0.0123], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 4---------------------------------------------------- epoch 4, time 220.46, cls_loss 0.1985 cls_loss_mapping 0.2859 cls_loss_causal 1.4268 re_mapping 0.1054 re_causal 0.2123 /// teacc 95.74 lr 0.00010000 Epoch 6, weight, value: tensor([[ 0.0102, 0.0432, 0.0323, ..., -0.0454, -0.0223, 0.0199], [ 0.0447, -0.0086, -0.0058, ..., 0.0129, 0.0089, -0.0346], [-0.0060, -0.0034, -0.0182, ..., -0.0142, -0.0080, 0.0036], ..., [ 0.0133, 0.0393, -0.0497, ..., -0.0241, 0.0056, -0.0296], [-0.0066, -0.0124, 0.0154, ..., 0.0097, -0.0244, -0.0235], [ 0.0167, -0.0154, 0.0021, ..., -0.0120, -0.0112, 0.0287]], device='cuda:0'), grad: tensor([[ 1.4601e-03, -1.3130e-02, -2.2583e-02, ..., 1.3094e-03, -4.4799e-04, -2.4612e-02], [-1.5251e-02, -1.5140e-04, 8.5640e-04, ..., -1.1124e-02, -4.9400e-04, 4.7040e-04], [ 3.2482e-03, 7.5912e-03, 1.2321e-02, ..., 3.3531e-03, 5.0640e-04, 1.4908e-02], ..., [-1.5125e-03, -8.2092e-03, -1.5507e-03, ..., -2.8706e-03, 8.3327e-05, -1.2074e-03], [ 1.1047e-02, 2.0676e-03, 3.0403e-03, ..., 7.8735e-03, 1.4143e-03, 5.1842e-03], [ 2.3499e-03, 5.7602e-03, 3.4561e-03, ..., 1.1539e-03, 2.4486e-04, 3.8681e-03]], device='cuda:0') Epoch 6, bias, value: tensor([ 0.0066, 0.0043, 0.0030, 0.0170, 0.0233, 0.0313, -0.0236, 0.0054, -0.0313, -0.0171], device='cuda:0'), grad: tensor([-0.0211, -0.0063, 0.0145, 0.0088, 0.0003, -0.0048, 0.0039, -0.0125, 0.0109, 0.0064], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 5---------------------------------------------------- epoch 5, time 222.11, cls_loss 0.1691 cls_loss_mapping 0.2370 cls_loss_causal 1.3333 re_mapping 0.0893 re_causal 0.1940 /// teacc 96.15 lr 0.00010000 Epoch 7, weight, value: tensor([[ 0.0096, 0.0449, 0.0328, ..., -0.0469, -0.0243, 0.0205], [ 0.0462, -0.0079, -0.0069, ..., 0.0142, 0.0102, -0.0361], [-0.0077, -0.0027, -0.0187, ..., -0.0156, -0.0081, 0.0016], ..., [ 0.0117, 0.0412, -0.0513, ..., -0.0260, 0.0059, -0.0314], [-0.0066, -0.0143, 0.0159, ..., 0.0108, -0.0281, -0.0238], [ 0.0155, -0.0170, 0.0025, ..., -0.0129, -0.0132, 0.0298]], device='cuda:0'), grad: tensor([[-4.7989e-03, 2.9755e-02, 1.6830e-02, ..., 7.0429e-04, 5.4896e-05, 1.9928e-02], [-5.1689e-03, -8.4877e-04, 4.2572e-03, ..., -9.7275e-03, -3.8452e-03, 4.0092e-03], [ 3.7861e-03, 1.0368e-02, 8.0414e-03, ..., 2.2278e-03, 3.1161e-04, 2.1763e-03], ..., [ 2.1324e-03, 4.6730e-03, 6.3324e-03, ..., 1.4000e-03, 2.8062e-04, 5.9605e-04], [ 1.7262e-04, 5.6038e-03, 8.4991e-03, ..., 3.6087e-03, 4.4785e-03, 3.9101e-03], [ 4.1866e-04, -3.2837e-02, -2.7374e-02, ..., -1.5717e-03, -2.7522e-05, -2.7679e-02]], device='cuda:0') Epoch 7, bias, value: tensor([ 0.0068, 0.0043, 0.0028, 0.0169, 0.0232, 0.0313, -0.0237, 0.0057, -0.0310, -0.0174], device='cuda:0'), grad: tensor([ 0.0241, -0.0030, 0.0233, -0.0246, -0.0155, -0.0007, 0.0012, 0.0155, 0.0117, -0.0320], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 6---------------------------------------------------- epoch 6, time 221.10, cls_loss 0.1594 cls_loss_mapping 0.2062 cls_loss_causal 1.3062 re_mapping 0.0763 re_causal 0.1682 /// teacc 96.48 lr 0.00010000 Epoch 8, weight, value: tensor([[ 9.5645e-03, 4.6679e-02, 3.3759e-02, ..., -4.8533e-02, -2.6581e-02, 2.0594e-02], [ 4.7984e-02, -6.7997e-03, -7.5239e-03, ..., 1.5814e-02, 1.2059e-02, -3.6902e-02], [-9.2853e-03, -2.6466e-03, -1.9201e-02, ..., -1.6556e-02, -7.1986e-03, 5.3548e-05], ..., [ 1.0127e-02, 4.3045e-02, -5.3172e-02, ..., -2.7381e-02, 3.9840e-03, -3.3275e-02], [-6.3788e-03, -1.6351e-02, 1.7163e-02, ..., 1.1576e-02, -2.9850e-02, -2.3237e-02], [ 1.3993e-02, -1.7653e-02, 2.8311e-03, ..., -1.3822e-02, -1.4593e-02, 3.1092e-02]], device='cuda:0'), grad: tensor([[ 5.3930e-04, -1.5097e-03, -5.8174e-04, ..., 1.1902e-03, 1.3375e-04, -1.5438e-04], [ 2.4242e-03, 2.9621e-03, 2.5196e-03, ..., 3.6716e-03, -5.2005e-05, 1.4477e-03], [ 1.0633e-03, 1.8406e-03, 1.4019e-03, ..., 1.6718e-03, 2.4527e-05, 9.9277e-04], ..., [ 2.4433e-03, -4.7455e-03, 3.3417e-03, ..., 3.3054e-03, 2.6658e-05, 4.9782e-03], [-8.4839e-03, 2.2907e-03, -6.9160e-03, ..., -1.0422e-02, 4.7255e-04, -9.0551e-04], [ 8.1921e-04, -5.2147e-03, -5.4474e-03, ..., -2.7847e-03, 2.2471e-04, -8.9340e-03]], device='cuda:0') Epoch 8, bias, value: tensor([ 0.0071, 0.0049, 0.0026, 0.0170, 0.0233, 0.0308, -0.0240, 0.0054, -0.0309, -0.0173], device='cuda:0'), grad: tensor([ 0.0004, 0.0062, 0.0017, 0.0059, 0.0057, -0.0002, -0.0008, 0.0009, -0.0055, -0.0144], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 7---------------------------------------------------- epoch 7, time 220.63, cls_loss 0.1296 cls_loss_mapping 0.1668 cls_loss_causal 1.2145 re_mapping 0.0685 re_causal 0.1564 /// teacc 96.98 lr 0.00010000 Epoch 9, weight, value: tensor([[ 0.0090, 0.0485, 0.0343, ..., -0.0498, -0.0278, 0.0206], [ 0.0487, -0.0064, -0.0080, ..., 0.0163, 0.0147, -0.0386], [-0.0105, -0.0028, -0.0199, ..., -0.0175, -0.0079, -0.0011], ..., [ 0.0092, 0.0445, -0.0551, ..., -0.0282, 0.0038, -0.0348], [-0.0064, -0.0180, 0.0177, ..., 0.0123, -0.0317, -0.0236], [ 0.0138, -0.0185, 0.0036, ..., -0.0137, -0.0161, 0.0322]], device='cuda:0'), grad: tensor([[ 4.2367e-04, 4.2796e-04, -5.7459e-05, ..., 1.0729e-03, 1.3173e-04, 1.1797e-03], [-3.7937e-03, 4.6444e-04, 3.4308e-04, ..., -2.3613e-03, -8.3447e-04, 5.1880e-04], [ 6.2990e-04, -2.7370e-03, -1.7395e-03, ..., 1.1911e-03, -9.4986e-04, -8.8596e-04], ..., [ 1.0519e-03, -7.4625e-05, 1.2188e-03, ..., 1.6823e-03, 9.6226e-04, 1.5535e-03], [ 1.1902e-03, 9.8610e-04, 1.7996e-03, ..., 5.0354e-04, 5.4073e-04, 2.3155e-03], [ 4.9782e-04, 4.2114e-03, 5.4502e-04, ..., 1.0979e-02, -1.5008e-04, 2.7657e-03]], device='cuda:0') Epoch 9, bias, value: tensor([ 0.0072, 0.0048, 0.0022, 0.0171, 0.0233, 0.0306, -0.0240, 0.0057, -0.0308, -0.0171], device='cuda:0'), grad: tensor([ 0.0021, -0.0009, -0.0110, -0.0023, -0.0227, 0.0058, 0.0010, 0.0061, 0.0062, 0.0157], device='cuda:0') 100 0.0001 changing lr epoch 8, time 219.99, cls_loss 0.1144 cls_loss_mapping 0.1485 cls_loss_causal 1.1824 re_mapping 0.0658 re_causal 0.1525 /// teacc 96.50 lr 0.00010000 Epoch 10, weight, value: tensor([[ 0.0083, 0.0493, 0.0348, ..., -0.0516, -0.0292, 0.0210], [ 0.0496, -0.0067, -0.0091, ..., 0.0169, 0.0169, -0.0400], [-0.0112, -0.0026, -0.0206, ..., -0.0180, -0.0079, -0.0019], ..., [ 0.0081, 0.0461, -0.0564, ..., -0.0293, 0.0026, -0.0361], [-0.0059, -0.0189, 0.0180, ..., 0.0134, -0.0330, -0.0242], [ 0.0130, -0.0192, 0.0042, ..., -0.0143, -0.0175, 0.0331]], device='cuda:0'), grad: tensor([[ 3.9196e-04, -8.0967e-04, -2.7847e-04, ..., 1.8787e-03, -4.2945e-05, 4.9686e-04], [-1.0902e-04, 3.3712e-04, 7.0238e-04, ..., 4.1270e-04, -8.4996e-05, 5.3406e-04], [ 5.8594e-03, 1.2207e-02, 7.1411e-03, ..., 2.6016e-03, 1.2375e-02, 1.0185e-03], ..., [-4.7760e-03, -1.2985e-02, -4.4022e-03, ..., 1.5850e-03, -1.2268e-02, 4.8923e-04], [ 2.0294e-03, 1.9526e-04, -4.1008e-03, ..., -5.0201e-03, 1.3628e-03, -1.4772e-03], [ 1.1253e-03, -2.6989e-03, -3.3021e-04, ..., -1.6678e-02, 7.8773e-04, 1.5488e-03]], device='cuda:0') Epoch 10, bias, value: tensor([ 0.0070, 0.0049, 0.0022, 0.0174, 0.0232, 0.0306, -0.0243, 0.0057, -0.0307, -0.0173], device='cuda:0'), grad: tensor([ 0.0013, 0.0016, 0.0327, -0.0082, 0.0449, -0.0026, -0.0027, -0.0279, -0.0060, -0.0330], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 9---------------------------------------------------- epoch 9, time 220.70, cls_loss 0.1127 cls_loss_mapping 0.1464 cls_loss_causal 1.1127 re_mapping 0.0605 re_causal 0.1393 /// teacc 97.29 lr 0.00010000 Epoch 11, weight, value: tensor([[ 0.0076, 0.0506, 0.0353, ..., -0.0530, -0.0307, 0.0217], [ 0.0500, -0.0070, -0.0104, ..., 0.0172, 0.0171, -0.0415], [-0.0123, -0.0018, -0.0207, ..., -0.0185, -0.0075, -0.0026], ..., [ 0.0074, 0.0470, -0.0578, ..., -0.0300, 0.0033, -0.0372], [-0.0059, -0.0200, 0.0187, ..., 0.0141, -0.0346, -0.0247], [ 0.0125, -0.0205, 0.0052, ..., -0.0148, -0.0168, 0.0342]], device='cuda:0'), grad: tensor([[-6.7902e-04, 3.5667e-03, 9.3174e-04, ..., 5.9080e-04, 5.4240e-05, 4.0016e-03], [ 1.6699e-03, 9.3126e-04, 4.1604e-04, ..., 1.9464e-03, 9.5367e-05, 3.6311e-04], [ 3.6359e-04, -7.3099e-04, 1.8120e-03, ..., 1.6575e-03, -1.2457e-04, 2.0657e-03], ..., [ 1.0335e-04, -2.3132e-02, -2.3289e-03, ..., -2.6569e-03, 4.6134e-04, -1.1581e-02], [ 1.5011e-03, 1.9627e-03, -2.9202e-03, ..., -2.2507e-03, 1.5557e-04, -3.5152e-03], [ 2.3675e-04, 1.0742e-02, 3.9597e-03, ..., 2.7390e-03, 1.4015e-05, 8.4457e-03]], device='cuda:0') Epoch 11, bias, value: tensor([ 0.0071, 0.0045, 0.0025, 0.0171, 0.0231, 0.0308, -0.0242, 0.0058, -0.0306, -0.0172], device='cuda:0'), grad: tensor([ 0.0059, 0.0036, 0.0040, 0.0004, 0.0058, 0.0013, -0.0023, -0.0410, -0.0020, 0.0243], device='cuda:0') 100 0.0001 changing lr epoch 10, time 219.69, cls_loss 0.1089 cls_loss_mapping 0.1431 cls_loss_causal 1.1138 re_mapping 0.0560 re_causal 0.1323 /// teacc 97.06 lr 0.00010000 Epoch 12, weight, value: tensor([[ 0.0077, 0.0518, 0.0356, ..., -0.0537, -0.0312, 0.0216], [ 0.0506, -0.0070, -0.0111, ..., 0.0176, 0.0193, -0.0429], [-0.0134, -0.0019, -0.0214, ..., -0.0194, -0.0083, -0.0037], ..., [ 0.0065, 0.0486, -0.0593, ..., -0.0311, 0.0032, -0.0380], [-0.0055, -0.0220, 0.0193, ..., 0.0151, -0.0354, -0.0247], [ 0.0122, -0.0215, 0.0052, ..., -0.0152, -0.0179, 0.0346]], device='cuda:0'), grad: tensor([[-0.0003, -0.0039, -0.0028, ..., 0.0002, 0.0001, -0.0018], [-0.0016, -0.0002, 0.0006, ..., -0.0008, -0.0003, 0.0003], [ 0.0009, 0.0016, 0.0038, ..., 0.0002, 0.0001, 0.0014], ..., [ 0.0008, 0.0009, 0.0046, ..., 0.0003, 0.0002, 0.0028], [ 0.0011, 0.0010, 0.0049, ..., 0.0004, 0.0003, 0.0026], [ 0.0004, -0.0004, -0.0016, ..., -0.0005, -0.0002, -0.0031]], device='cuda:0') Epoch 12, bias, value: tensor([ 0.0072, 0.0044, 0.0026, 0.0175, 0.0231, 0.0305, -0.0242, 0.0059, -0.0305, -0.0176], device='cuda:0'), grad: tensor([-4.0779e-03, 5.7295e-06, 4.3793e-03, -1.5244e-02, 1.0633e-03, -9.2363e-04, 2.3613e-03, 7.5340e-03, 7.4158e-03, -2.5177e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 11---------------------------------------------------- epoch 11, time 220.20, cls_loss 0.0997 cls_loss_mapping 0.1288 cls_loss_causal 1.0871 re_mapping 0.0517 re_causal 0.1222 /// teacc 97.69 lr 0.00010000 Epoch 13, weight, value: tensor([[ 0.0073, 0.0526, 0.0364, ..., -0.0547, -0.0317, 0.0218], [ 0.0517, -0.0070, -0.0118, ..., 0.0185, 0.0220, -0.0439], [-0.0148, -0.0023, -0.0219, ..., -0.0201, -0.0085, -0.0048], ..., [ 0.0060, 0.0496, -0.0604, ..., -0.0317, 0.0039, -0.0396], [-0.0053, -0.0234, 0.0195, ..., 0.0158, -0.0373, -0.0247], [ 0.0121, -0.0217, 0.0063, ..., -0.0155, -0.0180, 0.0363]], device='cuda:0'), grad: tensor([[ 0.0070, 0.0002, 0.0089, ..., 0.0008, 0.0001, 0.0064], [ 0.0012, 0.0012, 0.0007, ..., 0.0004, 0.0001, 0.0010], [ 0.0008, -0.0001, 0.0007, ..., 0.0004, -0.0005, 0.0007], ..., [-0.0053, -0.0071, -0.0018, ..., 0.0003, 0.0002, -0.0041], [-0.0077, 0.0008, 0.0077, ..., -0.0116, -0.0051, -0.0032], [ 0.0052, 0.0049, 0.0050, ..., 0.0007, 0.0002, 0.0056]], device='cuda:0') Epoch 13, bias, value: tensor([ 0.0074, 0.0046, 0.0023, 0.0171, 0.0231, 0.0301, -0.0242, 0.0059, -0.0303, -0.0171], device='cuda:0'), grad: tensor([ 0.0065, 0.0027, -0.0007, -0.0220, -0.0017, 0.0090, 0.0061, -0.0085, -0.0006, 0.0091], device='cuda:0') 100 0.0001 changing lr epoch 12, time 221.66, cls_loss 0.0806 cls_loss_mapping 0.1054 cls_loss_causal 1.0424 re_mapping 0.0504 re_causal 0.1206 /// teacc 97.65 lr 0.00010000 Epoch 14, weight, value: tensor([[ 0.0070, 0.0535, 0.0368, ..., -0.0556, -0.0326, 0.0221], [ 0.0523, -0.0063, -0.0123, ..., 0.0188, 0.0229, -0.0454], [-0.0150, -0.0025, -0.0219, ..., -0.0206, -0.0076, -0.0047], ..., [ 0.0051, 0.0506, -0.0622, ..., -0.0327, 0.0042, -0.0408], [-0.0053, -0.0244, 0.0199, ..., 0.0165, -0.0383, -0.0247], [ 0.0117, -0.0224, 0.0068, ..., -0.0159, -0.0191, 0.0370]], device='cuda:0'), grad: tensor([[-9.2506e-04, -1.1635e-03, -6.9332e-04, ..., 2.3746e-04, 2.9787e-05, -6.1417e-04], [ 9.6893e-04, 1.1892e-03, 1.0309e-03, ..., 9.8038e-04, -9.0182e-05, 7.1526e-04], [ 8.8024e-04, 9.7275e-04, -2.8327e-05, ..., -2.7537e-04, -1.9813e-04, 4.1389e-04], ..., [ 1.0471e-03, -5.1079e-03, 1.0364e-05, ..., 1.8263e-03, 3.2485e-05, 8.4114e-04], [ 2.6360e-03, 2.3403e-03, 2.8839e-03, ..., 2.4719e-03, 5.4359e-04, 2.9182e-03], [-4.7684e-03, -1.0042e-03, -6.2065e-03, ..., -2.3708e-03, 3.6985e-05, -5.0354e-03]], device='cuda:0') Epoch 14, bias, value: tensor([ 0.0073, 0.0045, 0.0026, 0.0173, 0.0232, 0.0298, -0.0243, 0.0058, -0.0302, -0.0173], device='cuda:0'), grad: tensor([-0.0007, 0.0032, 0.0016, 0.0061, -0.0024, -0.0009, 0.0003, -0.0083, 0.0086, -0.0075], device='cuda:0') 100 0.0001 changing lr epoch 13, time 219.62, cls_loss 0.0816 cls_loss_mapping 0.1040 cls_loss_causal 1.0737 re_mapping 0.0457 re_causal 0.1136 /// teacc 97.55 lr 0.00010000 Epoch 15, weight, value: tensor([[ 0.0067, 0.0543, 0.0372, ..., -0.0563, -0.0337, 0.0223], [ 0.0533, -0.0060, -0.0130, ..., 0.0194, 0.0232, -0.0467], [-0.0158, -0.0027, -0.0224, ..., -0.0211, -0.0064, -0.0057], ..., [ 0.0042, 0.0513, -0.0634, ..., -0.0343, 0.0050, -0.0415], [-0.0051, -0.0251, 0.0202, ..., 0.0170, -0.0395, -0.0248], [ 0.0109, -0.0228, 0.0082, ..., -0.0163, -0.0202, 0.0386]], device='cuda:0'), grad: tensor([[ 1.4293e-04, 1.9598e-04, 1.1196e-03, ..., 9.8050e-05, 7.2479e-05, 1.2302e-03], [-3.0708e-04, 2.3723e-04, 2.9993e-04, ..., -2.1541e-04, -1.0449e-04, 1.4913e-04], [ 1.1468e-04, 2.3162e-04, 4.6110e-04, ..., -5.3525e-05, 4.9978e-05, 2.6751e-04], ..., [-6.3002e-05, -1.8501e-03, 9.9182e-04, ..., 9.4771e-05, -1.0914e-04, 1.2093e-03], [ 3.6502e-04, 3.5429e-04, 1.9035e-03, ..., 5.9307e-05, 1.1194e-04, 7.9918e-04], [ 5.1403e-04, 1.8489e-04, -4.7398e-04, ..., 9.0933e-04, 3.6526e-04, -2.7199e-03]], device='cuda:0') Epoch 15, bias, value: tensor([ 0.0075, 0.0043, 0.0026, 0.0170, 0.0233, 0.0298, -0.0246, 0.0059, -0.0300, -0.0170], device='cuda:0'), grad: tensor([ 1.3027e-03, 4.9686e-04, -1.8072e-04, -3.7308e-03, -1.5554e-03, 3.6621e-04, 2.4527e-05, -1.4362e-03, 2.7523e-03, 1.9627e-03], device='cuda:0') 100 0.0001 changing lr epoch 14, time 220.42, cls_loss 0.0733 cls_loss_mapping 0.0957 cls_loss_causal 0.9991 re_mapping 0.0436 re_causal 0.1045 /// teacc 97.41 lr 0.00010000 Epoch 16, weight, value: tensor([[ 0.0067, 0.0553, 0.0374, ..., -0.0572, -0.0349, 0.0224], [ 0.0541, -0.0060, -0.0134, ..., 0.0199, 0.0238, -0.0478], [-0.0164, -0.0030, -0.0227, ..., -0.0219, -0.0061, -0.0061], ..., [ 0.0035, 0.0523, -0.0648, ..., -0.0351, 0.0050, -0.0426], [-0.0051, -0.0259, 0.0206, ..., 0.0177, -0.0403, -0.0249], [ 0.0106, -0.0233, 0.0086, ..., -0.0167, -0.0209, 0.0396]], device='cuda:0'), grad: tensor([[ 2.1982e-04, -7.0524e-04, -5.6458e-04, ..., 3.3593e-04, 6.4522e-06, -4.7755e-04], [ 7.4482e-04, 4.0197e-04, 1.3170e-03, ..., 4.9353e-04, 4.0078e-04, 1.3733e-03], [ 1.2219e-04, 1.3423e-04, 1.3447e-04, ..., 1.1021e-04, 4.5337e-06, 1.6093e-04], ..., [ 3.1137e-04, -2.1400e-03, 9.5606e-04, ..., 3.8457e-04, -7.2765e-04, 9.2149e-05], [ 6.2828e-03, 2.0838e-04, 1.0574e-02, ..., 6.7101e-03, 1.3466e-03, 1.0597e-02], [-1.7300e-05, 1.1148e-03, -3.2825e-03, ..., 1.1212e-04, 7.1335e-04, -1.5526e-03]], device='cuda:0') Epoch 16, bias, value: tensor([ 0.0075, 0.0042, 0.0027, 0.0172, 0.0233, 0.0296, -0.0248, 0.0059, -0.0298, -0.0173], device='cuda:0'), grad: tensor([-0.0005, 0.0019, 0.0002, -0.0092, 0.0021, 0.0296, -0.0307, -0.0017, 0.0108, -0.0026], device='cuda:0') 100 0.0001 changing lr epoch 15, time 220.13, cls_loss 0.0584 cls_loss_mapping 0.0774 cls_loss_causal 0.9628 re_mapping 0.0421 re_causal 0.1043 /// teacc 97.49 lr 0.00010000 Epoch 17, weight, value: tensor([[ 0.0063, 0.0563, 0.0378, ..., -0.0581, -0.0361, 0.0227], [ 0.0547, -0.0062, -0.0134, ..., 0.0208, 0.0256, -0.0488], [-0.0172, -0.0033, -0.0231, ..., -0.0224, -0.0070, -0.0066], ..., [ 0.0030, 0.0531, -0.0657, ..., -0.0355, 0.0054, -0.0435], [-0.0049, -0.0262, 0.0209, ..., 0.0180, -0.0416, -0.0251], [ 0.0103, -0.0238, 0.0093, ..., -0.0172, -0.0219, 0.0406]], device='cuda:0'), grad: tensor([[ 5.8794e-04, 4.2558e-05, 2.7442e-04, ..., 4.2367e-04, 5.1081e-05, 4.0245e-04], [ 1.0933e-02, -6.7651e-05, 2.2621e-03, ..., 6.9389e-03, 3.3903e-04, 3.5973e-03], [ 1.0300e-03, 2.2531e-04, 2.4700e-04, ..., 5.6505e-04, -3.1185e-04, 4.2701e-04], ..., [ 3.7313e-04, -8.0681e-04, 4.6039e-04, ..., 4.2105e-04, 1.1581e-04, 5.7983e-04], [-1.2350e-03, 1.3733e-04, -1.1473e-03, ..., -1.3304e-03, 1.4567e-04, -1.6260e-03], [ 2.9877e-05, 1.2922e-04, -3.0231e-03, ..., 1.2369e-03, 1.1597e-03, -4.2419e-03]], device='cuda:0') Epoch 17, bias, value: tensor([ 0.0076, 0.0045, 0.0025, 0.0171, 0.0234, 0.0295, -0.0250, 0.0062, -0.0299, -0.0173], device='cuda:0'), grad: tensor([ 7.7820e-04, 1.0666e-02, -2.1696e-04, 3.1586e-03, -2.0084e-03, 4.2076e-03, -1.6663e-02, 3.8457e-04, 1.8179e-05, -3.2973e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 16---------------------------------------------------- epoch 16, time 220.42, cls_loss 0.0681 cls_loss_mapping 0.0863 cls_loss_causal 0.9701 re_mapping 0.0395 re_causal 0.0956 /// teacc 97.89 lr 0.00010000 Epoch 18, weight, value: tensor([[ 0.0058, 0.0570, 0.0378, ..., -0.0590, -0.0369, 0.0225], [ 0.0553, -0.0061, -0.0142, ..., 0.0214, 0.0267, -0.0502], [-0.0180, -0.0035, -0.0230, ..., -0.0226, -0.0075, -0.0070], ..., [ 0.0023, 0.0543, -0.0668, ..., -0.0364, 0.0057, -0.0446], [-0.0047, -0.0268, 0.0210, ..., 0.0182, -0.0422, -0.0255], [ 0.0096, -0.0247, 0.0097, ..., -0.0173, -0.0220, 0.0418]], device='cuda:0'), grad: tensor([[-1.6606e-04, -2.2507e-03, -2.5892e-04, ..., 1.1331e-04, 3.0205e-05, -9.4175e-04], [-1.0264e-04, 1.7176e-03, 3.6168e-04, ..., -3.0780e-04, 2.7394e-04, 9.9599e-05], [-3.8624e-04, 2.9297e-03, 5.4169e-04, ..., 6.2399e-07, 9.6321e-04, 1.1337e-04], ..., [ 2.6369e-04, 5.0774e-03, 2.3155e-03, ..., 1.6797e-04, 1.3809e-03, 1.4105e-03], [ 5.1165e-04, 1.4029e-03, 7.2908e-04, ..., 3.1853e-04, 2.9731e-04, 1.0271e-03], [ 1.1438e-04, -1.8034e-03, -2.6817e-03, ..., -1.7774e-04, 4.6402e-05, -2.5082e-03]], device='cuda:0') Epoch 18, bias, value: tensor([ 0.0074, 0.0043, 0.0027, 0.0170, 0.0231, 0.0298, -0.0250, 0.0064, -0.0299, -0.0172], device='cuda:0'), grad: tensor([-0.0031, 0.0063, 0.0029, -0.0224, 0.0013, 0.0007, 0.0010, 0.0132, 0.0042, -0.0042], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 17---------------------------------------------------- epoch 17, time 219.83, cls_loss 0.0489 cls_loss_mapping 0.0716 cls_loss_causal 0.9451 re_mapping 0.0377 re_causal 0.0944 /// teacc 98.13 lr 0.00010000 Epoch 19, weight, value: tensor([[ 0.0056, 0.0584, 0.0385, ..., -0.0598, -0.0374, 0.0231], [ 0.0556, -0.0063, -0.0146, ..., 0.0213, 0.0277, -0.0511], [-0.0189, -0.0038, -0.0234, ..., -0.0231, -0.0075, -0.0080], ..., [ 0.0018, 0.0550, -0.0680, ..., -0.0369, 0.0069, -0.0456], [-0.0041, -0.0277, 0.0213, ..., 0.0190, -0.0429, -0.0256], [ 0.0091, -0.0252, 0.0102, ..., -0.0176, -0.0224, 0.0425]], device='cuda:0'), grad: tensor([[ 1.1051e-04, -2.0874e-04, -1.0979e-04, ..., 1.3673e-04, 4.6015e-05, -6.8069e-05], [-2.0638e-03, -9.9421e-05, 4.3869e-05, ..., -1.7042e-03, -6.4707e-04, 5.6326e-05], [ 7.4768e-04, 2.9773e-05, 1.3733e-04, ..., 8.0013e-04, 6.0987e-04, 1.9515e-04], ..., [ 1.4973e-04, -1.7166e-04, 2.4748e-04, ..., 2.0123e-04, 5.3167e-05, 2.1780e-04], [ 1.5160e-02, 1.7381e-04, 1.0971e-02, ..., 2.0920e-02, 1.3718e-02, 1.0208e-02], [ 3.6979e-04, -2.7463e-05, -1.4043e-04, ..., 5.1165e-04, 3.5214e-04, 5.3465e-05]], device='cuda:0') Epoch 19, bias, value: tensor([ 0.0080, 0.0041, 0.0025, 0.0170, 0.0232, 0.0298, -0.0253, 0.0066, -0.0298, -0.0174], device='cuda:0'), grad: tensor([-6.8367e-05, -1.7681e-03, 7.6818e-04, 4.3416e-04, 3.6407e-04, -2.9572e-02, 6.0987e-04, 2.7752e-04, 2.9022e-02, -6.4850e-05], device='cuda:0') 100 0.0001 changing lr epoch 18, time 219.10, cls_loss 0.0536 cls_loss_mapping 0.0717 cls_loss_causal 0.9451 re_mapping 0.0348 re_causal 0.0893 /// teacc 98.08 lr 0.00010000 Epoch 20, weight, value: tensor([[ 0.0055, 0.0591, 0.0385, ..., -0.0605, -0.0389, 0.0230], [ 0.0562, -0.0061, -0.0152, ..., 0.0215, 0.0289, -0.0518], [-0.0198, -0.0042, -0.0237, ..., -0.0235, -0.0078, -0.0085], ..., [ 0.0009, 0.0560, -0.0693, ..., -0.0377, 0.0073, -0.0461], [-0.0038, -0.0285, 0.0214, ..., 0.0195, -0.0442, -0.0259], [ 0.0094, -0.0254, 0.0110, ..., -0.0177, -0.0222, 0.0437]], device='cuda:0'), grad: tensor([[ 4.0591e-05, -2.1148e-04, -9.0837e-05, ..., 2.0587e-04, 1.8328e-05, -9.5546e-05], [-2.4147e-03, 1.0036e-05, -9.7215e-05, ..., -2.6913e-03, -2.0161e-03, 3.1084e-05], [ 1.7223e-03, 4.3726e-04, 2.2066e-04, ..., 2.5406e-03, 1.5469e-03, 6.5684e-05], ..., [ 2.0754e-04, -1.1021e-04, 2.1553e-04, ..., 2.2411e-04, 1.0091e-04, 2.4438e-04], [-3.0231e-03, 6.5863e-05, -2.1629e-03, ..., -2.6226e-03, 1.1224e-04, -3.9673e-03], [ 1.1224e-04, -1.6436e-05, -7.6151e-04, ..., 2.0301e-04, 1.9401e-05, -9.8324e-04]], device='cuda:0') Epoch 20, bias, value: tensor([ 0.0079, 0.0039, 0.0026, 0.0168, 0.0234, 0.0295, -0.0255, 0.0067, -0.0298, -0.0170], device='cuda:0'), grad: tensor([ 0.0001, -0.0035, 0.0035, 0.0005, -0.0154, 0.0029, 0.0163, 0.0005, -0.0042, -0.0006], device='cuda:0') 100 0.0001 changing lr epoch 19, time 219.26, cls_loss 0.0496 cls_loss_mapping 0.0627 cls_loss_causal 0.9332 re_mapping 0.0336 re_causal 0.0876 /// teacc 98.05 lr 0.00010000 Epoch 21, weight, value: tensor([[ 0.0058, 0.0597, 0.0387, ..., -0.0612, -0.0399, 0.0231], [ 0.0569, -0.0062, -0.0156, ..., 0.0219, 0.0299, -0.0524], [-0.0207, -0.0047, -0.0242, ..., -0.0241, -0.0073, -0.0092], ..., [ 0.0002, 0.0567, -0.0705, ..., -0.0381, 0.0070, -0.0472], [-0.0038, -0.0295, 0.0215, ..., 0.0199, -0.0455, -0.0261], [ 0.0086, -0.0252, 0.0117, ..., -0.0183, -0.0229, 0.0444]], device='cuda:0'), grad: tensor([[ 1.1665e-04, 6.6936e-05, 1.3566e-04, ..., 6.4433e-05, 4.7892e-05, 8.9705e-05], [-3.3951e-03, 1.2290e-04, 2.7990e-04, ..., -2.9411e-03, -9.0170e-04, 8.5652e-05], [-1.5503e-02, -2.0523e-03, 2.6627e-03, ..., -1.3382e-02, 1.7557e-03, 9.4509e-04], ..., [ 3.3665e-04, -2.7409e-03, -1.6785e-03, ..., 7.9393e-04, -6.5565e-04, -2.2964e-03], [ 1.6739e-02, 1.9207e-03, -4.5609e-04, ..., 1.4687e-02, 8.9788e-04, 1.8358e-04], [ 6.4564e-04, 2.1019e-03, 5.5237e-03, ..., 3.7932e-04, 2.9430e-03, 3.1948e-03]], device='cuda:0') Epoch 21, bias, value: tensor([ 0.0077, 0.0038, 0.0024, 0.0171, 0.0234, 0.0298, -0.0252, 0.0064, -0.0301, -0.0168], device='cuda:0'), grad: tensor([ 0.0004, -0.0020, -0.0226, 0.0039, -0.0104, -0.0001, 0.0004, -0.0071, 0.0267, 0.0109], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 20---------------------------------------------------- epoch 20, time 219.80, cls_loss 0.0453 cls_loss_mapping 0.0621 cls_loss_causal 0.9226 re_mapping 0.0327 re_causal 0.0852 /// teacc 98.40 lr 0.00010000 Epoch 22, weight, value: tensor([[ 0.0056, 0.0604, 0.0388, ..., -0.0619, -0.0400, 0.0234], [ 0.0575, -0.0059, -0.0155, ..., 0.0227, 0.0304, -0.0530], [-0.0211, -0.0047, -0.0245, ..., -0.0246, -0.0076, -0.0098], ..., [-0.0004, 0.0575, -0.0717, ..., -0.0393, 0.0077, -0.0480], [-0.0035, -0.0300, 0.0217, ..., 0.0205, -0.0456, -0.0258], [ 0.0084, -0.0261, 0.0124, ..., -0.0186, -0.0231, 0.0453]], device='cuda:0'), grad: tensor([[-4.5985e-05, -6.9046e-04, -3.7909e-04, ..., 1.9145e-04, 2.6543e-06, -3.5024e-04], [-2.0170e-04, 5.9932e-05, 1.0216e-04, ..., -4.5627e-05, -3.0935e-05, 7.9393e-05], [ 1.9825e-04, 3.2425e-04, 2.2137e-04, ..., 2.4307e-04, -1.8924e-05, 2.0266e-04], ..., [ 8.8751e-05, -2.0337e-04, 1.0478e-04, ..., 1.3888e-04, 1.1943e-05, 9.8467e-05], [ 4.4727e-04, 1.3149e-04, 1.7762e-04, ..., 5.2977e-04, 1.7032e-05, 1.2341e-03], [ 1.6820e-04, 2.4343e-04, 2.5139e-03, ..., 4.7836e-03, 4.0084e-06, 2.3632e-03]], device='cuda:0') Epoch 22, bias, value: tensor([ 0.0077, 0.0040, 0.0026, 0.0167, 0.0234, 0.0295, -0.0252, 0.0066, -0.0299, -0.0168], device='cuda:0'), grad: tensor([-7.6818e-04, 6.9678e-05, 4.7684e-04, 6.8188e-04, -9.1553e-03, 4.3607e-04, -2.4433e-03, 9.4771e-05, 1.2293e-03, 9.3842e-03], device='cuda:0') 100 0.0001 changing lr epoch 21, time 218.49, cls_loss 0.0501 cls_loss_mapping 0.0673 cls_loss_causal 0.8941 re_mapping 0.0318 re_causal 0.0784 /// teacc 98.07 lr 0.00010000 Epoch 23, weight, value: tensor([[ 0.0055, 0.0611, 0.0393, ..., -0.0624, -0.0404, 0.0241], [ 0.0580, -0.0059, -0.0157, ..., 0.0230, 0.0310, -0.0537], [-0.0213, -0.0052, -0.0252, ..., -0.0251, -0.0074, -0.0109], ..., [-0.0010, 0.0587, -0.0726, ..., -0.0400, 0.0074, -0.0488], [-0.0035, -0.0305, 0.0218, ..., 0.0210, -0.0465, -0.0261], [ 0.0077, -0.0267, 0.0126, ..., -0.0195, -0.0236, 0.0456]], device='cuda:0'), grad: tensor([[ 4.9305e-04, 1.2808e-03, 5.1403e-04, ..., 6.4087e-04, 3.5584e-05, 1.3647e-03], [-5.2929e-04, 1.5032e-04, 1.4496e-04, ..., -4.6825e-04, -2.1553e-04, 1.2052e-04], [ 4.8709e-04, 3.5524e-05, 5.8460e-04, ..., 1.0735e-04, -2.2745e-04, 4.8494e-04], ..., [ 2.9516e-04, -1.2426e-03, 2.2507e-04, ..., 2.0361e-04, 1.9252e-04, 2.1100e-04], [ 1.1044e-03, 2.8062e-04, 1.5516e-03, ..., 2.0099e-04, 1.5306e-04, 1.3542e-03], [ 1.1617e-04, 1.0508e-04, 4.7177e-05, ..., 1.1814e-04, 4.0948e-05, -9.6202e-05]], device='cuda:0') Epoch 23, bias, value: tensor([ 0.0081, 0.0041, 0.0023, 0.0165, 0.0232, 0.0302, -0.0251, 0.0067, -0.0298, -0.0174], device='cuda:0'), grad: tensor([ 0.0023, -0.0002, 0.0003, -0.0031, 0.0002, 0.0013, -0.0024, -0.0013, 0.0026, 0.0005], device='cuda:0') 100 0.0001 changing lr epoch 22, time 218.35, cls_loss 0.0504 cls_loss_mapping 0.0670 cls_loss_causal 0.8771 re_mapping 0.0303 re_causal 0.0771 /// teacc 98.16 lr 0.00010000 Epoch 24, weight, value: tensor([[ 0.0051, 0.0611, 0.0390, ..., -0.0632, -0.0408, 0.0235], [ 0.0586, -0.0060, -0.0163, ..., 0.0230, 0.0322, -0.0546], [-0.0221, -0.0051, -0.0253, ..., -0.0254, -0.0077, -0.0111], ..., [-0.0015, 0.0595, -0.0738, ..., -0.0407, 0.0073, -0.0499], [-0.0035, -0.0305, 0.0221, ..., 0.0213, -0.0472, -0.0262], [ 0.0070, -0.0269, 0.0132, ..., -0.0197, -0.0232, 0.0466]], device='cuda:0'), grad: tensor([[-3.4118e-04, -8.9645e-04, -2.0123e-04, ..., 1.4699e-04, 8.8662e-06, 1.3888e-04], [-8.0109e-04, 3.9220e-04, 7.1406e-05, ..., -5.9652e-04, -8.5413e-05, 5.3406e-05], [ 2.1875e-04, 5.2977e-04, 2.3663e-04, ..., 2.7204e-04, -2.5049e-05, 1.7333e-04], ..., [ 1.6189e-04, 2.8687e-03, 1.5569e-04, ..., 2.8553e-03, 4.5031e-05, 2.6917e-04], [ 1.7428e-04, 9.2649e-04, -1.7080e-03, ..., -1.9574e-04, 6.3121e-05, -9.7132e-04], [ 1.5223e-04, 2.2697e-03, 1.3363e-04, ..., 1.2646e-03, 5.6893e-05, -2.6298e-04]], device='cuda:0') Epoch 24, bias, value: tensor([ 0.0072, 0.0038, 0.0025, 0.0168, 0.0233, 0.0303, -0.0252, 0.0066, -0.0298, -0.0170], device='cuda:0'), grad: tensor([-1.0109e-03, -1.7083e-04, 1.0939e-03, 1.4820e-03, -1.2550e-02, 3.3927e-04, -2.3469e-05, 7.5531e-03, -4.8733e-04, 3.7708e-03], device='cuda:0') 100 0.0001 changing lr epoch 23, time 219.35, cls_loss 0.0496 cls_loss_mapping 0.0664 cls_loss_causal 0.8789 re_mapping 0.0292 re_causal 0.0773 /// teacc 98.13 lr 0.00010000 Epoch 25, weight, value: tensor([[ 0.0054, 0.0615, 0.0394, ..., -0.0639, -0.0413, 0.0239], [ 0.0593, -0.0056, -0.0164, ..., 0.0237, 0.0332, -0.0548], [-0.0227, -0.0055, -0.0257, ..., -0.0262, -0.0076, -0.0118], ..., [-0.0024, 0.0608, -0.0746, ..., -0.0412, 0.0087, -0.0506], [-0.0036, -0.0314, 0.0223, ..., 0.0215, -0.0483, -0.0262], [ 0.0065, -0.0274, 0.0135, ..., -0.0201, -0.0241, 0.0470]], device='cuda:0'), grad: tensor([[ 2.5773e-04, -2.8515e-04, -1.2553e-04, ..., 1.9753e-04, 3.1918e-05, -2.6393e-04], [ 2.7390e-03, 1.0118e-05, 1.3244e-04, ..., 2.4414e-03, 1.0872e-03, 7.9441e-04], [ 1.3554e-04, 3.7968e-05, 1.0228e-04, ..., -3.5000e-04, -6.9714e-04, 8.9407e-05], ..., [ 1.1563e-04, -1.6689e-05, 9.6083e-05, ..., 1.2267e-04, 1.2243e-04, 1.0562e-04], [-7.4539e-03, 5.1409e-05, -6.0409e-05, ..., -5.7449e-03, -1.5554e-03, -2.6073e-03], [ 2.7585e-04, 1.6236e-04, -1.0881e-03, ..., -3.6335e-04, -6.9559e-05, -1.4200e-03]], device='cuda:0') Epoch 25, bias, value: tensor([ 0.0069, 0.0040, 0.0019, 0.0167, 0.0234, 0.0302, -0.0249, 0.0073, -0.0297, -0.0174], device='cuda:0'), grad: tensor([-3.2961e-05, 3.4847e-03, -2.0485e-03, -3.2568e-04, 4.3335e-03, 3.2005e-03, -1.7290e-03, 6.8903e-04, -5.4932e-03, -2.0771e-03], device='cuda:0') 100 0.0001 changing lr epoch 24, time 218.92, cls_loss 0.0489 cls_loss_mapping 0.0642 cls_loss_causal 0.8523 re_mapping 0.0289 re_causal 0.0755 /// teacc 98.18 lr 0.00010000 Epoch 26, weight, value: tensor([[ 0.0057, 0.0623, 0.0403, ..., -0.0645, -0.0410, 0.0244], [ 0.0597, -0.0055, -0.0171, ..., 0.0240, 0.0339, -0.0561], [-0.0233, -0.0057, -0.0260, ..., -0.0267, -0.0066, -0.0123], ..., [-0.0032, 0.0613, -0.0757, ..., -0.0415, 0.0080, -0.0511], [-0.0035, -0.0320, 0.0225, ..., 0.0220, -0.0492, -0.0266], [ 0.0063, -0.0284, 0.0139, ..., -0.0200, -0.0230, 0.0477]], device='cuda:0'), grad: tensor([[ 4.1342e-04, 1.8346e-04, 7.9536e-04, ..., 1.0500e-03, 8.3637e-04, 6.2943e-04], [-7.8773e-04, 4.5395e-04, 2.0790e-04, ..., -3.2711e-04, -6.1870e-05, 1.8167e-04], [ 1.4954e-03, 5.5933e-04, 9.4528e-03, ..., 1.2751e-03, 5.8823e-03, 4.7607e-03], ..., [ 2.0564e-04, -2.8324e-03, 4.4799e-04, ..., -1.3232e-04, -1.1339e-03, 2.8324e-04], [ 3.8090e-03, 1.3046e-03, 2.3289e-03, ..., 3.5496e-03, 6.2180e-03, 1.3952e-03], [ 2.1076e-04, 7.9441e-04, 1.5936e-03, ..., 1.9474e-03, 4.0388e-04, 2.0862e-04]], device='cuda:0') Epoch 26, bias, value: tensor([ 0.0078, 0.0037, 0.0025, 0.0166, 0.0232, 0.0300, -0.0250, 0.0069, -0.0297, -0.0174], device='cuda:0'), grad: tensor([ 2.2564e-03, 9.7561e-04, 2.4323e-02, -2.2736e-02, -8.3506e-05, -1.1223e-02, -2.0866e-03, -6.9542e-03, 1.0719e-02, 4.8141e-03], device='cuda:0') 100 0.0001 changing lr epoch 25, time 219.04, cls_loss 0.0429 cls_loss_mapping 0.0577 cls_loss_causal 0.8779 re_mapping 0.0278 re_causal 0.0738 /// teacc 98.31 lr 0.00010000 Epoch 27, weight, value: tensor([[ 0.0051, 0.0629, 0.0405, ..., -0.0655, -0.0417, 0.0247], [ 0.0607, -0.0057, -0.0170, ..., 0.0246, 0.0348, -0.0565], [-0.0245, -0.0057, -0.0264, ..., -0.0274, -0.0077, -0.0129], ..., [-0.0036, 0.0619, -0.0770, ..., -0.0419, 0.0088, -0.0518], [-0.0038, -0.0329, 0.0221, ..., 0.0222, -0.0505, -0.0272], [ 0.0056, -0.0286, 0.0143, ..., -0.0204, -0.0237, 0.0483]], device='cuda:0'), grad: tensor([[ 1.0714e-05, 1.8954e-04, 2.3139e-04, ..., 1.3053e-05, 9.5814e-06, 3.1638e-04], [-3.6049e-04, 6.8724e-05, 8.7544e-06, ..., -1.9324e-04, -1.0598e-04, -1.8775e-06], [ 4.5180e-05, -2.4343e-04, -1.5521e-04, ..., 2.1398e-05, -3.5137e-05, 1.3745e-04], ..., [ 7.7069e-05, -1.7910e-03, 5.6076e-04, ..., 6.7651e-05, -2.8476e-05, -1.4293e-04], [ 2.0251e-05, 1.1104e-04, 1.3793e-04, ..., -9.6619e-05, 5.4181e-05, 2.0134e-04], [ 9.9242e-05, 1.0691e-03, -1.7738e-03, ..., 7.6175e-05, -4.6641e-06, -1.4668e-03]], device='cuda:0') Epoch 27, bias, value: tensor([ 0.0077, 0.0040, 0.0024, 0.0167, 0.0236, 0.0305, -0.0252, 0.0067, -0.0302, -0.0173], device='cuda:0'), grad: tensor([ 4.5276e-04, -1.5342e-04, -4.5562e-04, 1.2884e-03, 3.6907e-04, 2.5654e-04, 1.0781e-05, -1.7262e-03, 2.4211e-04, -2.8563e-04], device='cuda:0') 100 0.0001 changing lr epoch 26, time 218.59, cls_loss 0.0436 cls_loss_mapping 0.0513 cls_loss_causal 0.8530 re_mapping 0.0279 re_causal 0.0697 /// teacc 98.14 lr 0.00010000 Epoch 28, weight, value: tensor([[ 0.0046, 0.0633, 0.0408, ..., -0.0665, -0.0427, 0.0250], [ 0.0611, -0.0056, -0.0174, ..., 0.0250, 0.0352, -0.0572], [-0.0248, -0.0059, -0.0266, ..., -0.0278, -0.0068, -0.0136], ..., [-0.0042, 0.0627, -0.0780, ..., -0.0429, 0.0086, -0.0525], [-0.0034, -0.0336, 0.0226, ..., 0.0231, -0.0512, -0.0273], [ 0.0046, -0.0292, 0.0150, ..., -0.0211, -0.0242, 0.0488]], device='cuda:0'), grad: tensor([[ 5.3263e-04, 1.0365e-04, 3.3069e-04, ..., 2.9778e-04, 7.8231e-06, 3.8052e-04], [ 3.3712e-04, 5.7161e-05, 1.8036e-04, ..., 6.1274e-04, 1.0198e-04, 9.8884e-05], [ 1.1663e-03, 2.9147e-05, 8.0442e-04, ..., 1.4949e-04, 7.5054e-04, 8.9765e-05], ..., [-1.5581e-04, -3.8586e-03, -7.3433e-04, ..., -5.2404e-04, 9.2834e-06, -2.1648e-03], [ 4.4060e-03, 3.0732e-04, 3.6812e-03, ..., 2.7981e-03, 3.4356e-04, 4.6730e-03], [ 2.0733e-03, 4.5319e-03, 1.6689e-03, ..., 5.6190e-03, 1.0215e-05, 3.5381e-03]], device='cuda:0') Epoch 28, bias, value: tensor([ 0.0075, 0.0039, 0.0031, 0.0164, 0.0236, 0.0306, -0.0253, 0.0065, -0.0299, -0.0176], device='cuda:0'), grad: tensor([ 0.0006, 0.0011, 0.0027, -0.0029, -0.0068, 0.0031, -0.0106, -0.0065, 0.0063, 0.0131], device='cuda:0') 100 0.0001 changing lr epoch 27, time 218.74, cls_loss 0.0394 cls_loss_mapping 0.0515 cls_loss_causal 0.8938 re_mapping 0.0268 re_causal 0.0723 /// teacc 98.38 lr 0.00010000 Epoch 29, weight, value: tensor([[ 0.0045, 0.0643, 0.0412, ..., -0.0673, -0.0435, 0.0254], [ 0.0613, -0.0063, -0.0180, ..., 0.0251, 0.0359, -0.0577], [-0.0258, -0.0062, -0.0269, ..., -0.0286, -0.0075, -0.0136], ..., [-0.0042, 0.0637, -0.0786, ..., -0.0431, 0.0094, -0.0533], [-0.0029, -0.0347, 0.0226, ..., 0.0242, -0.0516, -0.0276], [ 0.0040, -0.0291, 0.0150, ..., -0.0217, -0.0247, 0.0493]], device='cuda:0'), grad: tensor([[ 1.1379e-04, -1.2283e-03, -1.2350e-03, ..., -2.2805e-04, 6.3181e-05, -1.0834e-03], [-2.5902e-03, -1.5068e-04, -4.2129e-04, ..., -1.1053e-03, -1.3523e-03, 3.5548e-04], [ 1.4172e-03, 6.7520e-04, 5.2738e-04, ..., 6.2418e-04, 1.2420e-05, 7.2813e-04], ..., [ 4.0674e-04, 8.8930e-05, 3.2926e-04, ..., 1.8537e-04, 2.1052e-04, 3.8338e-04], [ 1.4410e-03, 2.1482e-04, 5.8174e-04, ..., 9.0981e-04, 4.3321e-04, 4.8709e-04], [-6.9737e-06, -4.3422e-05, -3.6716e-03, ..., 6.1703e-04, 1.1218e-04, -3.2043e-03]], device='cuda:0') Epoch 29, bias, value: tensor([ 0.0077, 0.0035, 0.0028, 0.0168, 0.0234, 0.0304, -0.0256, 0.0066, -0.0297, -0.0174], device='cuda:0'), grad: tensor([-2.9716e-03, -3.2349e-03, 2.4872e-03, 3.4313e-03, -3.2429e-06, 1.8711e-03, -2.1114e-03, 1.0519e-03, 1.9484e-03, -2.4719e-03], device='cuda:0') 100 0.0001 changing lr epoch 28, time 218.66, cls_loss 0.0316 cls_loss_mapping 0.0431 cls_loss_causal 0.7947 re_mapping 0.0261 re_causal 0.0710 /// teacc 98.39 lr 0.00010000 Epoch 30, weight, value: tensor([[ 0.0040, 0.0645, 0.0415, ..., -0.0676, -0.0439, 0.0254], [ 0.0618, -0.0052, -0.0181, ..., 0.0255, 0.0362, -0.0585], [-0.0257, -0.0064, -0.0268, ..., -0.0289, -0.0076, -0.0135], ..., [-0.0049, 0.0640, -0.0797, ..., -0.0437, 0.0097, -0.0542], [-0.0035, -0.0356, 0.0227, ..., 0.0242, -0.0527, -0.0278], [ 0.0040, -0.0288, 0.0149, ..., -0.0223, -0.0248, 0.0496]], device='cuda:0'), grad: tensor([[-3.5137e-05, -1.0848e-04, -7.3195e-05, ..., 3.4660e-05, 4.0054e-05, -8.4460e-05], [ 2.9579e-05, -1.2308e-05, 2.2918e-05, ..., 1.5807e-04, 5.3972e-05, 4.7356e-05], [-4.3464e-04, -1.3262e-05, 1.2398e-05, ..., -2.4378e-04, -5.8413e-04, 2.7508e-05], ..., [ 8.1062e-05, -1.7896e-05, 4.3213e-05, ..., 2.1577e-04, 1.5163e-04, 5.0724e-05], [-6.5446e-05, 2.3305e-05, -6.3419e-04, ..., -1.0853e-03, -4.7684e-05, -6.3086e-04], [ 5.7399e-05, 4.3839e-05, 1.8799e-04, ..., 3.3832e-04, 1.0180e-04, 1.3852e-04]], device='cuda:0') Epoch 30, bias, value: tensor([ 0.0074, 0.0034, 0.0029, 0.0167, 0.0236, 0.0310, -0.0258, 0.0066, -0.0299, -0.0175], device='cuda:0'), grad: tensor([-4.0770e-05, 3.1471e-04, -2.3918e-03, 5.7364e-04, -3.2353e-04, 5.0926e-04, 2.5368e-04, 5.4216e-04, 6.5565e-05, 4.9639e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 29---------------------------------------------------- epoch 29, time 219.74, cls_loss 0.0331 cls_loss_mapping 0.0461 cls_loss_causal 0.8036 re_mapping 0.0244 re_causal 0.0668 /// teacc 98.42 lr 0.00010000 Epoch 31, weight, value: tensor([[ 0.0035, 0.0650, 0.0417, ..., -0.0685, -0.0443, 0.0255], [ 0.0628, -0.0056, -0.0189, ..., 0.0268, 0.0374, -0.0592], [-0.0262, -0.0061, -0.0268, ..., -0.0298, -0.0079, -0.0138], ..., [-0.0057, 0.0648, -0.0807, ..., -0.0448, 0.0100, -0.0551], [-0.0034, -0.0360, 0.0233, ..., 0.0246, -0.0533, -0.0278], [ 0.0030, -0.0296, 0.0153, ..., -0.0229, -0.0250, 0.0502]], device='cuda:0'), grad: tensor([[-9.3460e-05, -9.6619e-05, -5.5283e-05, ..., -3.2157e-05, 8.0541e-06, -1.4734e-04], [-1.9574e-04, 1.3494e-04, 1.4567e-04, ..., -1.3435e-04, -5.5492e-05, 5.2303e-05], [ 1.0145e-04, 2.1725e-03, 1.7262e-03, ..., 1.0467e-04, -1.0812e-04, 3.0518e-05], ..., [ 5.5969e-05, -5.1613e-03, -3.7727e-03, ..., 1.0347e-04, 5.2720e-05, 3.4660e-05], [-2.0027e-04, 3.5429e-04, -3.5346e-05, ..., -4.4584e-04, 3.8564e-05, -2.1338e-04], [ 1.0240e-04, 2.0337e-04, 2.9659e-04, ..., 1.5402e-03, 1.0677e-05, 1.0681e-03]], device='cuda:0') Epoch 31, bias, value: tensor([ 0.0074, 0.0034, 0.0031, 0.0164, 0.0234, 0.0312, -0.0258, 0.0068, -0.0295, -0.0179], device='cuda:0'), grad: tensor([-6.3896e-05, 4.3225e-04, 6.6223e-03, 6.5002e-03, -2.4605e-03, 6.2895e-04, 2.0218e-04, -1.5556e-02, 5.1355e-04, 3.1776e-03], device='cuda:0') 100 0.0001 changing lr epoch 30, time 218.43, cls_loss 0.0352 cls_loss_mapping 0.0457 cls_loss_causal 0.8238 re_mapping 0.0239 re_causal 0.0642 /// teacc 98.36 lr 0.00010000 Epoch 32, weight, value: tensor([[ 0.0034, 0.0656, 0.0420, ..., -0.0692, -0.0448, 0.0259], [ 0.0630, -0.0055, -0.0194, ..., 0.0267, 0.0378, -0.0604], [-0.0267, -0.0064, -0.0270, ..., -0.0303, -0.0076, -0.0139], ..., [-0.0070, 0.0654, -0.0820, ..., -0.0455, 0.0095, -0.0563], [-0.0030, -0.0361, 0.0235, ..., 0.0252, -0.0538, -0.0280], [ 0.0029, -0.0298, 0.0158, ..., -0.0232, -0.0253, 0.0510]], device='cuda:0'), grad: tensor([[ 8.2627e-06, -8.9049e-05, -8.9228e-05, ..., 1.4074e-05, 9.9838e-06, -9.8944e-05], [-2.9488e-03, 1.8612e-05, 1.5028e-05, ..., -3.6716e-03, -3.6812e-03, 1.2040e-05], [ 2.7924e-03, 8.8453e-05, 1.9640e-05, ..., 3.5038e-03, 3.4485e-03, 3.1948e-05], ..., [ 3.8207e-05, -1.4210e-04, 1.9002e-04, ..., 3.9577e-05, 3.2485e-05, 1.3947e-04], [ 1.0353e-04, 2.9549e-05, 1.5903e-04, ..., 9.7036e-05, 7.0989e-05, 1.6773e-04], [ 4.3809e-05, -2.9638e-05, -4.8089e-04, ..., 5.7928e-07, 8.7172e-06, -3.7050e-04]], device='cuda:0') Epoch 32, bias, value: tensor([ 0.0074, 0.0030, 0.0032, 0.0168, 0.0234, 0.0310, -0.0259, 0.0065, -0.0293, -0.0178], device='cuda:0'), grad: tensor([-1.3328e-04, -4.5815e-03, 4.3068e-03, 2.7716e-05, 3.1161e-04, 9.2328e-05, 4.7088e-05, 9.3162e-05, 3.2449e-04, -4.8804e-04], device='cuda:0') 100 0.0001 changing lr epoch 31, time 218.53, cls_loss 0.0307 cls_loss_mapping 0.0406 cls_loss_causal 0.7918 re_mapping 0.0234 re_causal 0.0632 /// teacc 98.26 lr 0.00010000 Epoch 33, weight, value: tensor([[ 0.0032, 0.0660, 0.0422, ..., -0.0698, -0.0454, 0.0260], [ 0.0638, -0.0061, -0.0199, ..., 0.0271, 0.0393, -0.0611], [-0.0277, -0.0066, -0.0269, ..., -0.0309, -0.0080, -0.0142], ..., [-0.0073, 0.0660, -0.0828, ..., -0.0456, 0.0101, -0.0568], [-0.0028, -0.0368, 0.0237, ..., 0.0258, -0.0547, -0.0278], [ 0.0025, -0.0298, 0.0165, ..., -0.0238, -0.0255, 0.0516]], device='cuda:0'), grad: tensor([[ 2.4274e-05, -1.8358e-04, -1.2338e-04, ..., 2.2754e-05, 6.1579e-06, -2.0647e-04], [-2.0885e-04, 9.9465e-06, 5.4032e-05, ..., 1.3065e-04, -1.2302e-04, 3.4302e-05], [ 7.6175e-05, 2.2054e-05, 9.0003e-05, ..., 6.0141e-05, 1.7554e-05, 9.6381e-05], ..., [ 1.3494e-04, -3.3522e-04, 6.8486e-05, ..., 5.1677e-05, 6.8665e-05, 5.1945e-05], [ 1.9300e-04, 2.6286e-05, 3.2663e-04, ..., 8.6248e-05, 6.4015e-05, 3.8505e-04], [-1.7285e-05, 2.9612e-04, -2.4166e-03, ..., 3.8838e-04, 2.0516e-04, -1.7834e-03]], device='cuda:0') Epoch 33, bias, value: tensor([ 0.0075, 0.0031, 0.0032, 0.0163, 0.0234, 0.0308, -0.0258, 0.0069, -0.0292, -0.0177], device='cuda:0'), grad: tensor([-2.8419e-04, 6.2287e-05, 1.1539e-04, 1.8969e-03, -1.8969e-03, 3.6329e-05, 3.0994e-05, -1.5354e-04, 4.5967e-04, -2.6608e-04], device='cuda:0') 100 0.0001 changing lr epoch 32, time 218.64, cls_loss 0.0322 cls_loss_mapping 0.0418 cls_loss_causal 0.8160 re_mapping 0.0227 re_causal 0.0610 /// teacc 98.24 lr 0.00010000 Epoch 34, weight, value: tensor([[ 0.0031, 0.0662, 0.0421, ..., -0.0705, -0.0463, 0.0252], [ 0.0640, -0.0061, -0.0207, ..., 0.0273, 0.0400, -0.0619], [-0.0279, -0.0070, -0.0272, ..., -0.0315, -0.0083, -0.0149], ..., [-0.0079, 0.0670, -0.0836, ..., -0.0461, 0.0093, -0.0572], [-0.0029, -0.0373, 0.0235, ..., 0.0262, -0.0554, -0.0285], [ 0.0017, -0.0303, 0.0172, ..., -0.0240, -0.0259, 0.0528]], device='cuda:0'), grad: tensor([[ 6.0558e-05, -1.8165e-05, 8.3745e-05, ..., 2.9728e-05, 7.1377e-06, 2.0340e-05], [-9.8884e-05, 4.9055e-05, 1.8671e-05, ..., -6.3837e-05, -4.3064e-05, 9.1866e-06], [ 8.0645e-05, 1.5748e-04, 2.7537e-04, ..., 1.5366e-04, 2.0668e-05, 1.2469e-04], ..., [ 3.6478e-05, -8.7690e-04, 1.8287e-04, ..., 8.3327e-05, -3.5107e-05, 1.3268e-04], [-9.4557e-04, 6.4850e-05, -2.5349e-03, ..., -3.7718e-04, 1.7598e-05, -1.5869e-03], [ 4.9257e-04, 1.6963e-04, 1.3065e-03, ..., 2.9993e-04, 1.7449e-05, 8.1110e-04]], device='cuda:0') Epoch 34, bias, value: tensor([ 0.0070, 0.0029, 0.0031, 0.0167, 0.0233, 0.0310, -0.0255, 0.0069, -0.0296, -0.0177], device='cuda:0'), grad: tensor([ 1.2600e-04, 9.1136e-05, 7.2098e-04, 2.4080e-04, 9.2804e-05, 2.1470e-04, 2.4533e-04, -1.1606e-03, -2.5730e-03, 2.0027e-03], device='cuda:0') 100 0.0001 changing lr epoch 33, time 218.62, cls_loss 0.0356 cls_loss_mapping 0.0402 cls_loss_causal 0.7848 re_mapping 0.0237 re_causal 0.0591 /// teacc 98.37 lr 0.00010000 Epoch 35, weight, value: tensor([[ 0.0025, 0.0667, 0.0423, ..., -0.0713, -0.0475, 0.0252], [ 0.0645, -0.0067, -0.0207, ..., 0.0275, 0.0415, -0.0626], [-0.0287, -0.0073, -0.0273, ..., -0.0322, -0.0090, -0.0154], ..., [-0.0085, 0.0682, -0.0849, ..., -0.0469, 0.0099, -0.0581], [-0.0021, -0.0380, 0.0237, ..., 0.0271, -0.0559, -0.0282], [ 0.0014, -0.0307, 0.0176, ..., -0.0247, -0.0265, 0.0536]], device='cuda:0'), grad: tensor([[ 4.2367e-04, 2.9063e-04, 1.2755e-04, ..., 3.2449e-04, 2.8834e-06, 4.5490e-04], [ 4.0889e-05, 2.9445e-04, 2.5392e-04, ..., 1.0020e-04, 4.5747e-06, 2.0993e-04], [ 2.3574e-05, -1.8921e-03, 4.1664e-05, ..., 2.9624e-05, -3.6812e-04, 2.7224e-05], ..., [ 1.9684e-05, 3.5686e-03, 2.0123e-03, ..., 5.1022e-04, 3.3069e-04, 1.7958e-03], [-1.0133e-04, 1.8060e-04, 1.2374e-04, ..., -1.5974e-04, 3.2093e-06, 2.0695e-04], [-1.4573e-05, -2.6340e-03, -2.8706e-03, ..., -5.9128e-04, 3.2019e-06, -2.6112e-03]], device='cuda:0') Epoch 35, bias, value: tensor([ 0.0068, 0.0029, 0.0030, 0.0167, 0.0238, 0.0309, -0.0262, 0.0070, -0.0292, -0.0177], device='cuda:0'), grad: tensor([ 0.0010, 0.0013, -0.0041, 0.0005, 0.0009, -0.0002, -0.0007, 0.0137, 0.0004, -0.0128], device='cuda:0') 100 0.0001 changing lr epoch 34, time 218.48, cls_loss 0.0307 cls_loss_mapping 0.0392 cls_loss_causal 0.7970 re_mapping 0.0231 re_causal 0.0635 /// teacc 98.41 lr 0.00010000 Epoch 36, weight, value: tensor([[ 0.0026, 0.0678, 0.0428, ..., -0.0720, -0.0482, 0.0257], [ 0.0649, -0.0071, -0.0212, ..., 0.0278, 0.0420, -0.0631], [-0.0292, -0.0075, -0.0277, ..., -0.0328, -0.0093, -0.0160], ..., [-0.0089, 0.0684, -0.0862, ..., -0.0474, 0.0109, -0.0589], [-0.0022, -0.0385, 0.0238, ..., 0.0274, -0.0566, -0.0284], [ 0.0008, -0.0305, 0.0178, ..., -0.0253, -0.0263, 0.0536]], device='cuda:0'), grad: tensor([[ 3.9965e-05, 1.2743e-04, 2.9847e-05, ..., 1.9360e-04, 3.8967e-06, 3.5226e-05], [-2.7478e-05, 2.0429e-05, 5.0247e-05, ..., 7.5400e-05, -1.7613e-05, 1.3098e-05], [ 2.2352e-05, 2.2620e-05, 1.7941e-05, ..., 7.3791e-05, -1.6633e-06, 5.0515e-06], ..., [ 1.5542e-05, -8.1003e-05, 5.7817e-05, ..., 1.0389e-04, 4.1239e-06, 1.6868e-05], [ 8.5294e-05, 2.4036e-05, 1.6952e-04, ..., 1.7142e-04, 1.0967e-05, 1.0186e-04], [ 4.7386e-05, 6.0201e-05, 8.1730e-04, ..., 1.5917e-03, 2.8573e-06, -3.0309e-05]], device='cuda:0') Epoch 36, bias, value: tensor([ 0.0074, 0.0027, 0.0029, 0.0166, 0.0238, 0.0318, -0.0261, 0.0064, -0.0295, -0.0177], device='cuda:0'), grad: tensor([ 3.3855e-04, 1.8120e-04, 8.0913e-06, -6.4373e-06, -5.7449e-03, -5.1594e-04, 2.1839e-03, 1.2255e-04, 4.3869e-04, 2.9964e-03], device='cuda:0') 100 0.0001 changing lr epoch 35, time 218.10, cls_loss 0.0274 cls_loss_mapping 0.0356 cls_loss_causal 0.7697 re_mapping 0.0220 re_causal 0.0590 /// teacc 98.36 lr 0.00010000 Epoch 37, weight, value: tensor([[ 0.0025, 0.0683, 0.0430, ..., -0.0727, -0.0484, 0.0260], [ 0.0656, -0.0065, -0.0216, ..., 0.0288, 0.0421, -0.0634], [-0.0297, -0.0078, -0.0277, ..., -0.0334, -0.0087, -0.0162], ..., [-0.0097, 0.0696, -0.0862, ..., -0.0486, 0.0109, -0.0591], [-0.0021, -0.0394, 0.0239, ..., 0.0277, -0.0568, -0.0286], [ 0.0002, -0.0313, 0.0178, ..., -0.0260, -0.0266, 0.0541]], device='cuda:0'), grad: tensor([[-1.6451e-05, -2.3377e-04, 1.1355e-05, ..., -1.0777e-04, 8.7380e-05, -2.8682e-04], [ 2.1946e-04, 2.0254e-04, 2.2805e-04, ..., 2.0730e-04, 9.6738e-05, 6.2361e-06], [-1.6470e-03, -2.9370e-05, -8.6212e-03, ..., 4.2468e-05, -6.3820e-03, 9.6709e-06], ..., [-2.0635e-04, -1.2732e-03, -4.9210e-04, ..., 1.4198e-04, -2.0098e-06, 5.2452e-05], [ 1.7905e-04, 8.5771e-05, 1.3089e-04, ..., 2.1112e-04, 3.2932e-05, 4.8310e-05], [ 7.9012e-04, 6.9332e-04, 1.0748e-03, ..., 1.9798e-03, 9.8765e-05, 2.3007e-04]], device='cuda:0') Epoch 37, bias, value: tensor([ 0.0074, 0.0030, 0.0028, 0.0166, 0.0236, 0.0317, -0.0259, 0.0069, -0.0296, -0.0180], device='cuda:0'), grad: tensor([-7.6413e-05, 1.0233e-03, -1.6785e-02, 1.6418e-02, -4.9248e-03, 6.7854e-04, 2.6450e-05, -1.9798e-03, 6.5660e-04, 4.9744e-03], device='cuda:0') 100 0.0001 changing lr epoch 36, time 219.42, cls_loss 0.0297 cls_loss_mapping 0.0412 cls_loss_causal 0.7410 re_mapping 0.0203 re_causal 0.0544 /// teacc 98.37 lr 0.00010000 Epoch 38, weight, value: tensor([[ 0.0021, 0.0687, 0.0430, ..., -0.0734, -0.0503, 0.0260], [ 0.0662, -0.0068, -0.0221, ..., 0.0294, 0.0422, -0.0641], [-0.0298, -0.0082, -0.0272, ..., -0.0340, -0.0083, -0.0165], ..., [-0.0104, 0.0705, -0.0867, ..., -0.0493, 0.0112, -0.0596], [-0.0018, -0.0398, 0.0239, ..., 0.0283, -0.0572, -0.0285], [-0.0003, -0.0315, 0.0183, ..., -0.0269, -0.0257, 0.0548]], device='cuda:0'), grad: tensor([[-4.3821e-04, -1.0414e-03, -7.4291e-04, ..., -1.4734e-04, 3.1143e-06, -5.2309e-04], [ 2.6211e-05, 2.1353e-05, 5.8860e-06, ..., 5.8115e-05, 2.6405e-05, 6.1803e-06], [ 3.4183e-05, 3.6716e-04, 5.6237e-05, ..., 5.9396e-05, -5.4777e-05, 5.2154e-05], ..., [ 1.6227e-05, -4.7588e-04, 5.1886e-05, ..., 1.1265e-04, 2.1428e-05, 7.1824e-05], [-7.9036e-05, 7.8619e-05, 7.1704e-05, ..., 1.5306e-04, 2.4110e-05, 1.3530e-04], [ 1.6108e-05, 2.4176e-04, -1.7190e-04, ..., -3.3112e-03, 3.7819e-05, -3.8218e-04]], device='cuda:0') Epoch 38, bias, value: tensor([ 0.0073, 0.0028, 0.0031, 0.0166, 0.0238, 0.0318, -0.0263, 0.0069, -0.0296, -0.0181], device='cuda:0'), grad: tensor([-0.0019, 0.0003, 0.0002, 0.0007, 0.0045, -0.0004, 0.0020, -0.0004, 0.0005, -0.0055], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 37---------------------------------------------------- epoch 37, time 219.45, cls_loss 0.0286 cls_loss_mapping 0.0346 cls_loss_causal 0.7683 re_mapping 0.0214 re_causal 0.0582 /// teacc 98.52 lr 0.00010000 Epoch 39, weight, value: tensor([[ 0.0019, 0.0691, 0.0431, ..., -0.0738, -0.0508, 0.0259], [ 0.0672, -0.0072, -0.0213, ..., 0.0302, 0.0437, -0.0641], [-0.0304, -0.0085, -0.0273, ..., -0.0345, -0.0093, -0.0166], ..., [-0.0109, 0.0718, -0.0875, ..., -0.0501, 0.0112, -0.0602], [-0.0027, -0.0404, 0.0237, ..., 0.0282, -0.0580, -0.0289], [-0.0002, -0.0321, 0.0189, ..., -0.0269, -0.0254, 0.0556]], device='cuda:0'), grad: tensor([[ 4.1336e-05, -6.5947e-04, -5.8031e-04, ..., -4.5204e-04, 1.5453e-05, -9.8515e-04], [-7.1585e-05, 2.7306e-06, 1.7464e-05, ..., -3.6657e-05, 2.9966e-05, 2.5719e-05], [ 1.6019e-05, 7.4744e-05, 6.0052e-05, ..., 5.5790e-05, -1.5545e-04, 1.0175e-04], ..., [ 2.9862e-05, 4.9323e-05, 8.3566e-05, ..., 1.4782e-04, 5.5462e-05, 1.2022e-04], [ 4.2021e-05, 4.3809e-05, 9.5308e-05, ..., 9.2983e-05, 9.5144e-06, 1.5080e-04], [ 2.7731e-05, 3.0732e-04, 5.1111e-06, ..., 2.3484e-04, 1.6198e-05, 1.5485e-04]], device='cuda:0') Epoch 39, bias, value: tensor([ 0.0070, 0.0034, 0.0032, 0.0167, 0.0236, 0.0320, -0.0268, 0.0071, -0.0302, -0.0178], device='cuda:0'), grad: tensor([-1.6642e-03, 8.8453e-05, -1.3864e-04, 7.1859e-04, -1.2898e-04, 5.7459e-04, -5.7602e-04, 3.5071e-04, 3.0899e-04, 4.6635e-04], device='cuda:0') 100 0.0001 changing lr epoch 38, time 218.12, cls_loss 0.0238 cls_loss_mapping 0.0290 cls_loss_causal 0.7601 re_mapping 0.0200 re_causal 0.0557 /// teacc 98.32 lr 0.00010000 Epoch 40, weight, value: tensor([[ 0.0007, 0.0692, 0.0437, ..., -0.0752, -0.0516, 0.0259], [ 0.0668, -0.0073, -0.0216, ..., 0.0301, 0.0438, -0.0648], [-0.0306, -0.0088, -0.0276, ..., -0.0348, -0.0096, -0.0169], ..., [-0.0119, 0.0728, -0.0884, ..., -0.0510, 0.0111, -0.0611], [-0.0024, -0.0407, 0.0239, ..., 0.0288, -0.0587, -0.0288], [-0.0007, -0.0327, 0.0188, ..., -0.0272, -0.0257, 0.0558]], device='cuda:0'), grad: tensor([[ 3.0398e-05, -2.0429e-05, 7.2908e-04, ..., 1.4687e-04, 2.9374e-06, 6.4230e-04], [-6.2799e-04, 1.2871e-06, 4.5866e-05, ..., -5.1355e-04, -1.7190e-04, 3.1263e-05], [ 1.7118e-04, 6.6906e-06, 3.0732e-04, ..., 1.4853e-04, 4.1008e-05, 2.6202e-04], ..., [ 5.0664e-05, -6.0499e-06, 3.9577e-05, ..., 3.3826e-05, 1.0937e-05, 3.2306e-05], [ 6.3896e-04, 2.0396e-06, 3.4561e-03, ..., 8.9884e-04, 2.9922e-05, 3.0060e-03], [ 9.0897e-05, 6.0014e-06, -4.5319e-03, ..., -7.8773e-04, 6.4075e-06, -4.0207e-03]], device='cuda:0') Epoch 40, bias, value: tensor([ 0.0071, 0.0029, 0.0032, 0.0171, 0.0236, 0.0319, -0.0261, 0.0071, -0.0302, -0.0182], device='cuda:0'), grad: tensor([ 0.0008, -0.0005, 0.0002, -0.0034, 0.0002, 0.0027, 0.0003, 0.0002, 0.0042, -0.0047], device='cuda:0') 100 0.0001 changing lr epoch 39, time 218.26, cls_loss 0.0254 cls_loss_mapping 0.0353 cls_loss_causal 0.7439 re_mapping 0.0200 re_causal 0.0565 /// teacc 98.34 lr 0.00010000 Epoch 41, weight, value: tensor([[ 0.0004, 0.0694, 0.0438, ..., -0.0758, -0.0519, 0.0258], [ 0.0671, -0.0058, -0.0218, ..., 0.0300, 0.0446, -0.0653], [-0.0312, -0.0091, -0.0277, ..., -0.0352, -0.0099, -0.0172], ..., [-0.0133, 0.0732, -0.0890, ..., -0.0514, 0.0109, -0.0615], [-0.0021, -0.0414, 0.0238, ..., 0.0292, -0.0589, -0.0291], [-0.0015, -0.0337, 0.0195, ..., -0.0276, -0.0256, 0.0569]], device='cuda:0'), grad: tensor([[ 1.6623e-03, 2.2755e-03, -2.2149e-04, ..., 2.9755e-03, 4.3400e-06, -1.4400e-03], [-2.0504e-04, 5.5218e-04, 1.3268e-04, ..., -2.8586e-04, -1.5366e-04, 2.2089e-04], [ 1.7369e-04, -1.9002e-04, 1.3673e-04, ..., 1.5283e-04, 2.2128e-05, 8.7738e-05], ..., [-1.1139e-03, -2.1496e-03, -4.1097e-05, ..., -1.3262e-05, 1.5900e-05, -5.7697e-04], [-1.8425e-03, -2.5978e-03, -4.3440e-04, ..., -3.4637e-03, 2.8044e-05, 1.1530e-03], [ 5.9700e-04, 1.3371e-03, -5.6171e-04, ..., -7.4863e-05, 1.3992e-05, -9.6917e-05]], device='cuda:0') Epoch 41, bias, value: tensor([ 0.0066, 0.0031, 0.0033, 0.0177, 0.0234, 0.0315, -0.0257, 0.0067, -0.0303, -0.0181], device='cuda:0'), grad: tensor([ 0.0027, 0.0008, -0.0011, 0.0003, 0.0016, 0.0003, 0.0008, -0.0035, -0.0035, 0.0016], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 40---------------------------------------------------- epoch 40, time 219.06, cls_loss 0.0203 cls_loss_mapping 0.0270 cls_loss_causal 0.7160 re_mapping 0.0194 re_causal 0.0542 /// teacc 98.54 lr 0.00010000 Epoch 42, weight, value: tensor([[ 0.0002, 0.0698, 0.0439, ..., -0.0765, -0.0522, 0.0256], [ 0.0681, -0.0053, -0.0219, ..., 0.0308, 0.0452, -0.0657], [-0.0314, -0.0092, -0.0278, ..., -0.0357, -0.0100, -0.0173], ..., [-0.0140, 0.0736, -0.0898, ..., -0.0525, 0.0110, -0.0623], [-0.0024, -0.0421, 0.0237, ..., 0.0293, -0.0603, -0.0293], [-0.0023, -0.0341, 0.0199, ..., -0.0280, -0.0260, 0.0576]], device='cuda:0'), grad: tensor([[-3.6061e-05, -3.3927e-04, -6.3598e-05, ..., 1.3605e-05, 1.1340e-05, -1.8716e-04], [ 2.7370e-04, 6.5982e-05, 1.5354e-03, ..., 7.5519e-05, 2.5654e-04, 1.2517e-05], [-5.7650e-04, 2.3899e-03, -2.1992e-03, ..., -1.3363e-04, -5.3024e-04, 3.3796e-05], ..., [ 1.5318e-04, -2.8458e-03, 3.0017e-04, ..., 6.1691e-05, 9.1791e-05, 1.4022e-05], [-6.3038e-04, 8.7678e-05, -3.5453e-04, ..., -7.0953e-04, 5.3316e-05, 2.8446e-05], [ 1.0175e-04, 2.6917e-04, 2.0421e-04, ..., 7.8440e-05, 5.5045e-05, 6.8784e-05]], device='cuda:0') Epoch 42, bias, value: tensor([ 0.0065, 0.0037, 0.0034, 0.0174, 0.0235, 0.0318, -0.0256, 0.0066, -0.0309, -0.0179], device='cuda:0'), grad: tensor([-0.0006, 0.0046, -0.0022, 0.0011, 0.0003, 0.0002, 0.0001, -0.0045, -0.0003, 0.0012], device='cuda:0') 100 0.0001 changing lr epoch 41, time 217.77, cls_loss 0.0242 cls_loss_mapping 0.0313 cls_loss_causal 0.7690 re_mapping 0.0187 re_causal 0.0536 /// teacc 98.39 lr 0.00010000 Epoch 43, weight, value: tensor([[-0.0003, 0.0705, 0.0442, ..., -0.0770, -0.0524, 0.0258], [ 0.0681, -0.0058, -0.0223, ..., 0.0304, 0.0455, -0.0661], [-0.0312, -0.0098, -0.0278, ..., -0.0353, -0.0096, -0.0176], ..., [-0.0141, 0.0743, -0.0905, ..., -0.0530, 0.0113, -0.0629], [-0.0021, -0.0422, 0.0239, ..., 0.0300, -0.0609, -0.0294], [-0.0025, -0.0342, 0.0200, ..., -0.0286, -0.0262, 0.0578]], device='cuda:0'), grad: tensor([[ 1.6403e-04, 4.2707e-05, 8.7559e-05, ..., 1.3316e-04, 3.0790e-06, 5.4091e-06], [ 1.0222e-04, 2.7448e-05, 8.5115e-05, ..., 1.3530e-04, -7.5810e-06, 1.2055e-05], [ 1.8430e-04, -8.0585e-05, 2.4408e-05, ..., 1.8322e-04, -1.8448e-05, 6.3851e-06], ..., [-3.8445e-06, -1.6379e-04, 2.7955e-05, ..., 8.7321e-05, 4.2394e-06, 7.8678e-06], [ 1.2070e-04, 1.1027e-04, -8.0395e-04, ..., -1.5335e-03, 3.9935e-06, -4.1986e-04], [ 1.6892e-04, 1.3876e-04, -1.7178e-04, ..., 1.1339e-03, 1.8388e-05, -1.7107e-04]], device='cuda:0') Epoch 43, bias, value: tensor([ 0.0068, 0.0032, 0.0034, 0.0173, 0.0240, 0.0316, -0.0262, 0.0067, -0.0306, -0.0179], device='cuda:0'), grad: tensor([ 3.9792e-04, 3.2473e-04, -9.0301e-05, 1.5354e-03, -6.2180e-04, 6.3181e-04, -3.3607e-03, -1.8671e-05, -3.3212e-04, 1.5345e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 42---------------------------------------------------- epoch 42, time 217.79, cls_loss 0.0241 cls_loss_mapping 0.0309 cls_loss_causal 0.7237 re_mapping 0.0188 re_causal 0.0513 /// teacc 98.58 lr 0.00010000 Epoch 44, weight, value: tensor([[-0.0007, 0.0715, 0.0444, ..., -0.0777, -0.0531, 0.0257], [ 0.0690, -0.0062, -0.0220, ..., 0.0311, 0.0460, -0.0665], [-0.0323, -0.0102, -0.0281, ..., -0.0361, -0.0099, -0.0180], ..., [-0.0146, 0.0750, -0.0909, ..., -0.0537, 0.0111, -0.0630], [-0.0022, -0.0428, 0.0237, ..., 0.0303, -0.0619, -0.0297], [-0.0035, -0.0344, 0.0201, ..., -0.0294, -0.0264, 0.0583]], device='cuda:0'), grad: tensor([[ 3.9482e-04, -1.2755e-04, 6.7329e-04, ..., 1.1170e-04, 1.2368e-05, 9.6560e-05], [ 6.8605e-05, 6.0424e-06, 8.4996e-05, ..., 5.1528e-05, -7.6741e-06, 4.6611e-05], [ 5.5599e-04, 4.3392e-05, 9.5940e-04, ..., 2.9874e-04, 1.5274e-05, 1.7571e-04], ..., [ 9.4414e-05, 2.6926e-05, 1.8644e-04, ..., 1.3864e-04, 1.2755e-05, 1.4472e-04], [-4.7569e-03, -2.0087e-04, 6.6376e-04, ..., -3.5419e-03, 1.0639e-04, -2.3537e-03], [ 1.5545e-04, -5.4687e-05, -5.4455e-04, ..., -1.0706e-05, 7.0110e-06, -5.9223e-04]], device='cuda:0') Epoch 44, bias, value: tensor([ 0.0070, 0.0034, 0.0033, 0.0171, 0.0239, 0.0326, -0.0261, 0.0065, -0.0309, -0.0182], device='cuda:0'), grad: tensor([ 0.0009, 0.0002, 0.0016, -0.0037, 0.0004, 0.0043, 0.0001, 0.0005, -0.0041, -0.0004], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 43---------------------------------------------------- epoch 43, time 217.75, cls_loss 0.0242 cls_loss_mapping 0.0309 cls_loss_causal 0.7113 re_mapping 0.0189 re_causal 0.0505 /// teacc 98.64 lr 0.00010000 Epoch 45, weight, value: tensor([[-0.0021, 0.0717, 0.0450, ..., -0.0787, -0.0539, 0.0257], [ 0.0698, -0.0058, -0.0223, ..., 0.0315, 0.0463, -0.0666], [-0.0323, -0.0107, -0.0282, ..., -0.0360, -0.0097, -0.0182], ..., [-0.0154, 0.0760, -0.0918, ..., -0.0546, 0.0116, -0.0637], [-0.0024, -0.0434, 0.0242, ..., 0.0307, -0.0633, -0.0294], [-0.0040, -0.0349, 0.0208, ..., -0.0299, -0.0269, 0.0592]], device='cuda:0'), grad: tensor([[ 9.9763e-06, -7.0214e-05, -4.3809e-05, ..., 8.0407e-05, 5.6066e-06, -5.5522e-05], [ 9.1270e-08, 4.2059e-06, 6.7204e-06, ..., 6.2466e-05, -2.3935e-07, 5.1036e-06], [ 5.1670e-06, 1.1794e-05, 1.1558e-06, ..., 8.6308e-05, 5.5991e-06, 1.1981e-05], ..., [ 2.8219e-06, -2.9981e-05, 4.8168e-06, ..., 3.4571e-05, 1.4566e-05, 6.3069e-06], [ 5.4270e-05, 2.0176e-05, 2.8268e-05, ..., 1.3888e-04, 3.2723e-05, 7.4267e-05], [ 6.4895e-06, 3.0607e-05, -6.2287e-05, ..., 9.6440e-05, 3.1263e-05, -6.9559e-05]], device='cuda:0') Epoch 45, bias, value: tensor([ 0.0068, 0.0033, 0.0034, 0.0168, 0.0246, 0.0323, -0.0260, 0.0069, -0.0308, -0.0187], device='cuda:0'), grad: tensor([ 2.1830e-05, 1.0622e-04, 1.1861e-04, 2.0683e-05, -6.6109e-03, 2.7895e-04, 5.6572e-03, 3.1590e-05, 2.7728e-04, 9.7513e-05], device='cuda:0') 100 0.0001 changing lr epoch 44, time 217.05, cls_loss 0.0227 cls_loss_mapping 0.0265 cls_loss_causal 0.7031 re_mapping 0.0179 re_causal 0.0505 /// teacc 98.39 lr 0.00010000 Epoch 46, weight, value: tensor([[-0.0031, 0.0720, 0.0453, ..., -0.0793, -0.0552, 0.0255], [ 0.0702, -0.0061, -0.0226, ..., 0.0318, 0.0465, -0.0670], [-0.0330, -0.0108, -0.0285, ..., -0.0364, -0.0096, -0.0187], ..., [-0.0148, 0.0763, -0.0928, ..., -0.0545, 0.0126, -0.0639], [-0.0027, -0.0439, 0.0243, ..., 0.0306, -0.0647, -0.0295], [-0.0043, -0.0351, 0.0215, ..., -0.0301, -0.0273, 0.0601]], device='cuda:0'), grad: tensor([[-9.3699e-05, -1.5771e-04, 4.3362e-05, ..., 3.6061e-05, 7.5884e-06, 1.0085e-04], [-2.4706e-05, 2.0728e-05, 4.1038e-05, ..., -5.0962e-05, -3.2157e-05, 2.3067e-05], [ 1.4770e-04, 1.1049e-05, 1.8990e-04, ..., 3.4958e-05, -2.8443e-04, 4.0203e-05], ..., [ 1.3876e-04, -8.1182e-05, 2.3961e-04, ..., 1.2279e-04, 1.8895e-04, 2.0278e-04], [-3.1948e-04, 2.5094e-05, -2.9659e-04, ..., -4.5109e-04, 1.3255e-05, -1.9896e-04], [ 2.4092e-04, 4.8041e-05, -1.6689e-06, ..., 1.6201e-04, 1.2256e-05, -9.1195e-05]], device='cuda:0') Epoch 46, bias, value: tensor([ 0.0067, 0.0031, 0.0032, 0.0170, 0.0242, 0.0328, -0.0260, 0.0070, -0.0312, -0.0183], device='cuda:0'), grad: tensor([-7.9125e-06, 2.9355e-05, -2.5481e-05, 9.0218e-04, 1.8322e-04, -1.1930e-03, 1.3006e-04, 5.7697e-04, -7.7152e-04, 1.7452e-04], device='cuda:0') 100 0.0001 changing lr epoch 45, time 216.97, cls_loss 0.0265 cls_loss_mapping 0.0345 cls_loss_causal 0.7272 re_mapping 0.0184 re_causal 0.0503 /// teacc 98.41 lr 0.00010000 Epoch 47, weight, value: tensor([[-0.0036, 0.0726, 0.0454, ..., -0.0800, -0.0557, 0.0250], [ 0.0705, -0.0072, -0.0231, ..., 0.0319, 0.0469, -0.0675], [-0.0329, -0.0106, -0.0278, ..., -0.0371, -0.0086, -0.0188], ..., [-0.0143, 0.0770, -0.0941, ..., -0.0542, 0.0126, -0.0647], [-0.0020, -0.0447, 0.0244, ..., 0.0317, -0.0658, -0.0289], [-0.0044, -0.0345, 0.0220, ..., -0.0308, -0.0282, 0.0609]], device='cuda:0'), grad: tensor([[ 1.3731e-05, -3.6992e-06, 7.3314e-06, ..., 2.9981e-05, 1.4082e-05, 6.9141e-06], [ 4.3917e-04, -6.3516e-07, 7.6056e-04, ..., 4.1275e-03, 1.9321e-03, 1.9484e-03], [ 2.8819e-05, 2.2769e-04, 7.6592e-05, ..., 1.1462e-04, 5.4300e-05, 1.7732e-05], ..., [ 1.1271e-04, -4.1676e-04, 7.0751e-05, ..., 3.9244e-04, 1.8013e-04, 1.8811e-04], [ 6.9737e-05, 1.3813e-05, 3.6627e-05, ..., 7.1347e-05, 3.4034e-05, 4.0054e-05], [-7.6962e-04, -4.7028e-05, -1.1253e-03, ..., -5.3978e-03, -2.6531e-03, -2.6207e-03]], device='cuda:0') Epoch 47, bias, value: tensor([ 0.0062, 0.0026, 0.0042, 0.0163, 0.0244, 0.0328, -0.0269, 0.0068, -0.0308, -0.0177], device='cuda:0'), grad: tensor([ 5.7906e-05, 8.2474e-03, 3.2401e-04, 3.4976e-04, 1.3905e-03, -1.0359e-04, 7.1108e-05, 3.6979e-04, 2.2626e-04, -1.0925e-02], device='cuda:0') 100 0.0001 changing lr epoch 46, time 217.48, cls_loss 0.0184 cls_loss_mapping 0.0263 cls_loss_causal 0.7013 re_mapping 0.0177 re_causal 0.0504 /// teacc 98.45 lr 0.00010000 Epoch 48, weight, value: tensor([[-0.0034, 0.0732, 0.0463, ..., -0.0804, -0.0566, 0.0258], [ 0.0708, -0.0069, -0.0231, ..., 0.0326, 0.0476, -0.0680], [-0.0334, -0.0110, -0.0280, ..., -0.0378, -0.0084, -0.0188], ..., [-0.0148, 0.0778, -0.0949, ..., -0.0555, 0.0122, -0.0653], [-0.0024, -0.0450, 0.0243, ..., 0.0317, -0.0681, -0.0290], [-0.0045, -0.0350, 0.0219, ..., -0.0314, -0.0287, 0.0610]], device='cuda:0'), grad: tensor([[ 3.1567e-04, 1.3709e-04, 4.5896e-05, ..., 1.1444e-04, 5.1856e-06, 7.6830e-05], [-4.6372e-05, 1.8731e-05, 2.8417e-05, ..., 4.1239e-06, -6.7335e-07, 1.0237e-05], [-1.7762e-05, 1.9163e-05, 1.7226e-05, ..., 2.5421e-05, -5.3138e-05, 9.7007e-06], ..., [ 1.7107e-05, -1.4901e-04, 1.1146e-05, ..., 2.3648e-05, -3.5405e-05, 5.0329e-06], [ 6.4015e-05, 2.8819e-05, 5.2214e-05, ..., 8.7082e-05, 2.2888e-05, 4.7803e-05], [ 2.2799e-05, 6.9261e-05, 4.6968e-04, ..., 1.1606e-03, 5.5283e-05, 1.3041e-04]], device='cuda:0') Epoch 48, bias, value: tensor([ 0.0068, 0.0029, 0.0041, 0.0164, 0.0251, 0.0326, -0.0269, 0.0066, -0.0311, -0.0180], device='cuda:0'), grad: tensor([ 3.6931e-04, 1.1069e-04, -4.9233e-05, 2.0361e-04, -2.0313e-03, -3.6806e-05, -4.2701e-04, -2.3913e-04, 1.9526e-04, 1.9035e-03], device='cuda:0') 100 0.0001 changing lr epoch 47, time 216.97, cls_loss 0.0158 cls_loss_mapping 0.0218 cls_loss_causal 0.7106 re_mapping 0.0174 re_causal 0.0497 /// teacc 98.54 lr 0.00010000 Epoch 49, weight, value: tensor([[-0.0029, 0.0739, 0.0466, ..., -0.0809, -0.0570, 0.0260], [ 0.0708, -0.0076, -0.0231, ..., 0.0329, 0.0483, -0.0685], [-0.0340, -0.0111, -0.0283, ..., -0.0384, -0.0084, -0.0190], ..., [-0.0148, 0.0789, -0.0952, ..., -0.0554, 0.0119, -0.0658], [-0.0023, -0.0455, 0.0244, ..., 0.0321, -0.0684, -0.0292], [-0.0051, -0.0355, 0.0223, ..., -0.0317, -0.0289, 0.0615]], device='cuda:0'), grad: tensor([[ 5.5194e-05, 3.0041e-05, 3.7313e-05, ..., 4.7714e-05, 9.9987e-06, 5.0664e-05], [-1.7524e-04, -4.5151e-05, 4.4644e-05, ..., -2.7999e-05, -1.1241e-04, 1.4521e-05], [ 7.4267e-05, 3.7044e-05, 2.4602e-05, ..., 7.3373e-05, 2.9132e-05, 1.2510e-05], ..., [ 1.8135e-05, -1.5945e-03, -4.3464e-04, ..., 7.0870e-05, -5.6058e-05, -5.5504e-04], [ 5.5656e-06, 4.5627e-05, 5.7161e-05, ..., -6.4313e-05, 4.9621e-06, 1.4973e-04], [ 1.7381e-04, 1.4324e-03, 8.6594e-04, ..., 6.5041e-04, 7.8201e-05, 6.6280e-04]], device='cuda:0') Epoch 49, bias, value: tensor([ 0.0071, 0.0026, 0.0041, 0.0161, 0.0247, 0.0328, -0.0268, 0.0069, -0.0311, -0.0180], device='cuda:0'), grad: tensor([ 1.5402e-04, -5.0128e-05, 4.0203e-05, 6.3515e-04, -1.1225e-03, -5.4693e-04, -4.0483e-04, -1.9951e-03, 1.9336e-04, 3.0975e-03], device='cuda:0') 100 0.0001 changing lr epoch 48, time 217.41, cls_loss 0.0186 cls_loss_mapping 0.0244 cls_loss_causal 0.6991 re_mapping 0.0162 re_causal 0.0460 /// teacc 98.57 lr 0.00010000 Epoch 50, weight, value: tensor([[-0.0030, 0.0747, 0.0466, ..., -0.0816, -0.0569, 0.0259], [ 0.0712, -0.0077, -0.0233, ..., 0.0333, 0.0493, -0.0689], [-0.0344, -0.0114, -0.0282, ..., -0.0390, -0.0079, -0.0191], ..., [-0.0154, 0.0795, -0.0958, ..., -0.0560, 0.0117, -0.0663], [-0.0020, -0.0460, 0.0243, ..., 0.0326, -0.0688, -0.0296], [-0.0057, -0.0360, 0.0228, ..., -0.0322, -0.0297, 0.0623]], device='cuda:0'), grad: tensor([[ 6.4850e-05, -4.9993e-06, 2.7046e-05, ..., 4.6968e-05, 1.5080e-05, 2.6837e-05], [-1.4663e-04, 3.7700e-05, 2.5511e-05, ..., -1.9968e-04, -7.0930e-05, 2.0057e-05], [ 8.3208e-05, 2.4283e-04, 3.3528e-05, ..., 8.3685e-05, 8.4162e-05, 9.5367e-06], ..., [ 4.3005e-05, -1.1797e-03, -2.6989e-04, ..., 4.5419e-05, -1.8084e-04, 1.8567e-05], [ 3.4881e-04, 3.3170e-05, 2.1935e-04, ..., 1.3721e-04, 1.3769e-04, 1.9288e-04], [ 8.1241e-05, 8.9407e-05, 8.7857e-05, ..., 5.9247e-05, 3.6210e-05, 5.7369e-05]], device='cuda:0') Epoch 50, bias, value: tensor([ 0.0069, 0.0027, 0.0048, 0.0160, 0.0246, 0.0328, -0.0265, 0.0067, -0.0313, -0.0182], device='cuda:0'), grad: tensor([ 1.1390e-04, -2.4930e-05, 7.5102e-04, 2.6703e-03, 9.5248e-05, -1.2064e-03, -1.0401e-04, -3.2883e-03, 5.8270e-04, 4.1199e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 49---------------------------------------------------- epoch 49, time 217.63, cls_loss 0.0230 cls_loss_mapping 0.0288 cls_loss_causal 0.7104 re_mapping 0.0164 re_causal 0.0465 /// teacc 98.75 lr 0.00010000 Epoch 51, weight, value: tensor([[-0.0041, 0.0750, 0.0473, ..., -0.0827, -0.0594, 0.0258], [ 0.0714, -0.0069, -0.0243, ..., 0.0332, 0.0504, -0.0691], [-0.0341, -0.0124, -0.0285, ..., -0.0384, -0.0081, -0.0198], ..., [-0.0148, 0.0802, -0.0960, ..., -0.0560, 0.0123, -0.0668], [-0.0022, -0.0464, 0.0246, ..., 0.0331, -0.0699, -0.0298], [-0.0056, -0.0364, 0.0234, ..., -0.0327, -0.0303, 0.0634]], device='cuda:0'), grad: tensor([[-7.7784e-05, -6.6459e-05, -1.2422e-04, ..., 1.7035e-04, 5.6550e-06, -1.5748e-04], [-3.3528e-06, 1.8239e-04, 3.8773e-05, ..., 1.0747e-06, -2.8312e-05, 2.8566e-05], [ 4.6253e-05, 1.4794e-04, 1.4722e-04, ..., 4.5180e-05, -1.6270e-06, 1.8120e-05], ..., [ 3.2336e-05, -1.2178e-03, -5.4121e-04, ..., 3.1471e-05, 9.9018e-06, -1.6618e-04], [-1.0699e-05, -1.4797e-05, -1.4460e-04, ..., 5.2899e-06, 1.9029e-05, 4.0472e-05], [ 1.8612e-05, 1.0624e-03, 5.6887e-04, ..., 1.1921e-04, 3.5614e-06, 2.5177e-04]], device='cuda:0') Epoch 51, bias, value: tensor([ 0.0068, 0.0024, 0.0044, 0.0157, 0.0242, 0.0327, -0.0264, 0.0077, -0.0314, -0.0178], device='cuda:0'), grad: tensor([-4.6939e-05, 2.6488e-04, 7.3528e-04, 2.2638e-04, -4.7073e-03, 1.0324e-04, 4.0016e-03, -2.3308e-03, -5.5933e-04, 2.3117e-03], device='cuda:0') 100 0.0001 changing lr epoch 50, time 217.45, cls_loss 0.0172 cls_loss_mapping 0.0253 cls_loss_causal 0.7170 re_mapping 0.0156 re_causal 0.0456 /// teacc 98.49 lr 0.00010000 Epoch 52, weight, value: tensor([[-0.0042, 0.0755, 0.0474, ..., -0.0832, -0.0600, 0.0260], [ 0.0721, -0.0067, -0.0250, ..., 0.0340, 0.0517, -0.0700], [-0.0352, -0.0127, -0.0285, ..., -0.0392, -0.0089, -0.0200], ..., [-0.0159, 0.0810, -0.0966, ..., -0.0574, 0.0121, -0.0674], [-0.0022, -0.0473, 0.0253, ..., 0.0339, -0.0710, -0.0294], [-0.0061, -0.0369, 0.0236, ..., -0.0334, -0.0300, 0.0636]], device='cuda:0'), grad: tensor([[ 2.2864e-04, 2.2564e-03, -1.4141e-05, ..., 2.4602e-05, 4.4751e-04, 3.1686e-04], [-6.8426e-05, 1.7989e-04, 3.4750e-05, ..., -5.0068e-05, -2.1815e-05, 5.0366e-05], [ 7.3850e-05, 1.0931e-04, 2.0236e-05, ..., 5.2780e-05, 5.8919e-05, 2.9698e-05], ..., [-3.7766e-04, -5.1804e-03, 1.1235e-05, ..., -4.0054e-05, -9.0361e-04, -6.6090e-04], [ 3.7932e-04, 2.4605e-04, 7.0035e-05, ..., 7.9811e-05, 3.6597e-04, 1.7750e-04], [ 1.8597e-04, 1.9741e-03, -1.4091e-04, ..., -8.7440e-05, 3.3545e-04, 1.4007e-04]], device='cuda:0') Epoch 52, bias, value: tensor([ 0.0069, 0.0026, 0.0042, 0.0158, 0.0242, 0.0322, -0.0260, 0.0075, -0.0309, -0.0180], device='cuda:0'), grad: tensor([ 0.0033, 0.0002, 0.0003, 0.0002, 0.0004, -0.0019, 0.0013, -0.0073, 0.0010, 0.0026], device='cuda:0') 100 0.0001 changing lr epoch 51, time 216.88, cls_loss 0.0178 cls_loss_mapping 0.0227 cls_loss_causal 0.7162 re_mapping 0.0165 re_causal 0.0473 /// teacc 98.68 lr 0.00010000 Epoch 53, weight, value: tensor([[-0.0043, 0.0759, 0.0476, ..., -0.0838, -0.0603, 0.0261], [ 0.0727, -0.0073, -0.0252, ..., 0.0345, 0.0520, -0.0708], [-0.0357, -0.0129, -0.0285, ..., -0.0398, -0.0088, -0.0201], ..., [-0.0160, 0.0819, -0.0972, ..., -0.0579, 0.0120, -0.0685], [-0.0019, -0.0479, 0.0254, ..., 0.0343, -0.0711, -0.0294], [-0.0062, -0.0369, 0.0240, ..., -0.0337, -0.0301, 0.0646]], device='cuda:0'), grad: tensor([[ 1.1206e-04, 8.6486e-05, 2.0218e-04, ..., 1.5068e-04, 6.2585e-06, 2.8038e-04], [-6.2943e-05, 9.2909e-06, 3.9488e-05, ..., -5.6595e-05, -2.3648e-05, 3.2097e-05], [ 1.2660e-04, 1.5274e-05, 2.4509e-04, ..., 8.1286e-06, -6.9499e-05, 1.5914e-04], ..., [ 4.9233e-05, -2.2447e-04, 6.1274e-05, ..., 5.3883e-05, -1.8239e-04, 5.3525e-05], [-2.3532e-04, 7.8157e-06, -1.5867e-04, ..., -3.2592e-04, 2.1413e-05, -1.2732e-04], [ 7.5769e-04, 6.1214e-05, 6.0463e-03, ..., 2.6631e-04, 5.8353e-05, 4.4098e-03]], device='cuda:0') Epoch 53, bias, value: tensor([ 0.0065, 0.0026, 0.0041, 0.0154, 0.0244, 0.0325, -0.0264, 0.0076, -0.0310, -0.0175], device='cuda:0'), grad: tensor([ 7.2098e-04, 9.5367e-06, 3.9697e-05, -8.7509e-03, 1.9002e-04, 1.3380e-03, -6.5184e-04, -3.0541e-04, -3.7456e-04, 7.7820e-03], device='cuda:0') 100 0.0001 changing lr epoch 52, time 217.10, cls_loss 0.0146 cls_loss_mapping 0.0207 cls_loss_causal 0.6966 re_mapping 0.0164 re_causal 0.0463 /// teacc 98.66 lr 0.00010000 Epoch 54, weight, value: tensor([[-0.0048, 0.0761, 0.0478, ..., -0.0846, -0.0605, 0.0261], [ 0.0732, -0.0070, -0.0255, ..., 0.0350, 0.0527, -0.0712], [-0.0362, -0.0132, -0.0287, ..., -0.0404, -0.0092, -0.0205], ..., [-0.0163, 0.0826, -0.0980, ..., -0.0582, 0.0121, -0.0694], [-0.0019, -0.0484, 0.0253, ..., 0.0343, -0.0714, -0.0299], [-0.0067, -0.0374, 0.0239, ..., -0.0339, -0.0305, 0.0650]], device='cuda:0'), grad: tensor([[-2.5153e-05, -4.3106e-04, -2.5734e-05, ..., 2.9132e-06, 6.7614e-07, -5.6982e-04], [ 2.1219e-05, 1.8871e-04, 5.0552e-06, ..., 8.9347e-05, -5.9903e-06, 3.0845e-05], [ 2.6405e-05, 6.7949e-05, 1.0088e-05, ..., 2.2948e-05, 2.4904e-06, 3.4273e-05], ..., [ 2.4691e-05, -5.0163e-04, 1.9282e-05, ..., -2.1148e-04, 2.7120e-06, 5.6982e-05], [ 9.9540e-05, 6.4313e-05, 1.6972e-05, ..., 1.0896e-04, 2.6450e-06, 7.3791e-05], [ 3.0324e-05, 6.9559e-05, -7.3910e-06, ..., 7.8440e-05, 1.5078e-06, 2.5272e-05]], device='cuda:0') Epoch 54, bias, value: tensor([ 0.0064, 0.0029, 0.0038, 0.0159, 0.0240, 0.0326, -0.0261, 0.0075, -0.0315, -0.0174], device='cuda:0'), grad: tensor([-9.4986e-04, 4.6253e-04, 1.4639e-04, 1.3599e-03, -3.4273e-05, -1.1873e-03, 7.6580e-04, -1.0948e-03, 3.0923e-04, 2.2280e-04], device='cuda:0') 100 0.0001 changing lr epoch 53, time 216.68, cls_loss 0.0171 cls_loss_mapping 0.0225 cls_loss_causal 0.6922 re_mapping 0.0162 re_causal 0.0455 /// teacc 98.58 lr 0.00010000 Epoch 55, weight, value: tensor([[-0.0058, 0.0763, 0.0480, ..., -0.0850, -0.0611, 0.0266], [ 0.0737, -0.0078, -0.0259, ..., 0.0353, 0.0537, -0.0721], [-0.0370, -0.0131, -0.0289, ..., -0.0411, -0.0098, -0.0209], ..., [-0.0164, 0.0837, -0.0989, ..., -0.0577, 0.0122, -0.0700], [-0.0016, -0.0489, 0.0258, ..., 0.0348, -0.0710, -0.0297], [-0.0068, -0.0381, 0.0240, ..., -0.0349, -0.0309, 0.0650]], device='cuda:0'), grad: tensor([[ 1.3700e-06, -1.9562e-04, -6.9857e-05, ..., 1.9759e-05, 3.6597e-05, -7.3835e-06], [-2.0707e-04, 2.7232e-06, 2.5295e-06, ..., -3.2282e-04, -4.4256e-05, 7.6368e-06], [ 2.2665e-05, 3.5256e-05, 1.2226e-05, ..., 2.5883e-05, 1.4953e-05, 1.5706e-05], ..., [ 4.9949e-05, 2.9489e-05, 1.7309e-04, ..., 3.2139e-04, 8.3029e-05, 2.1482e-04], [ 2.9653e-05, 3.3081e-05, -1.5087e-07, ..., 4.4316e-05, 1.8731e-05, 2.0027e-05], [ 2.3663e-05, -2.5183e-05, -1.8644e-04, ..., 8.9943e-05, 6.6817e-05, -1.8764e-04]], device='cuda:0') Epoch 55, bias, value: tensor([ 0.0065, 0.0028, 0.0036, 0.0159, 0.0241, 0.0329, -0.0267, 0.0082, -0.0310, -0.0182], device='cuda:0'), grad: tensor([ 8.6129e-05, -3.1328e-04, 1.3995e-04, 1.5926e-04, -1.9951e-03, 2.5249e-04, 1.5712e-04, 1.0376e-03, 1.4603e-04, 3.2949e-04], device='cuda:0') 100 0.0001 changing lr epoch 54, time 217.25, cls_loss 0.0172 cls_loss_mapping 0.0246 cls_loss_causal 0.6925 re_mapping 0.0164 re_causal 0.0448 /// teacc 98.74 lr 0.00010000 Epoch 56, weight, value: tensor([[-0.0057, 0.0773, 0.0481, ..., -0.0854, -0.0617, 0.0266], [ 0.0737, -0.0080, -0.0266, ..., 0.0351, 0.0537, -0.0735], [-0.0368, -0.0134, -0.0282, ..., -0.0410, -0.0090, -0.0201], ..., [-0.0171, 0.0843, -0.1000, ..., -0.0584, 0.0122, -0.0701], [-0.0012, -0.0494, 0.0260, ..., 0.0355, -0.0714, -0.0297], [-0.0068, -0.0387, 0.0246, ..., -0.0348, -0.0309, 0.0661]], device='cuda:0'), grad: tensor([[ 9.7454e-06, -1.2624e-04, -1.7032e-05, ..., 1.5736e-05, 2.8424e-06, -1.6451e-05], [-4.3571e-05, 1.6421e-05, 8.8215e-06, ..., -3.5256e-05, -2.0832e-05, 8.6576e-06], [ 1.2405e-05, -1.5467e-05, 3.4183e-05, ..., 1.0386e-05, -2.2411e-05, 4.4376e-05], ..., [ 2.1636e-05, -5.6219e-04, -1.0401e-04, ..., -5.9783e-05, 2.2262e-05, -4.5866e-05], [-6.4373e-05, 5.4032e-05, -4.8965e-05, ..., -6.9201e-05, 4.4331e-06, 9.1940e-06], [ 4.8399e-05, 4.9925e-04, 4.9770e-05, ..., 9.8228e-05, 3.0901e-06, -1.5414e-04]], device='cuda:0') Epoch 56, bias, value: tensor([ 0.0070, 0.0023, 0.0048, 0.0155, 0.0231, 0.0333, -0.0264, 0.0077, -0.0309, -0.0181], device='cuda:0'), grad: tensor([-1.6797e-04, 2.4661e-06, -3.0935e-05, 1.9833e-05, 1.1218e-04, 1.8322e-04, -1.4246e-04, -9.7847e-04, 4.2766e-05, 9.5892e-04], device='cuda:0') 100 0.0001 changing lr epoch 55, time 217.64, cls_loss 0.0152 cls_loss_mapping 0.0221 cls_loss_causal 0.6769 re_mapping 0.0147 re_causal 0.0434 /// teacc 98.39 lr 0.00010000 Epoch 57, weight, value: tensor([[-0.0061, 0.0778, 0.0483, ..., -0.0865, -0.0623, 0.0266], [ 0.0744, -0.0077, -0.0269, ..., 0.0359, 0.0547, -0.0740], [-0.0370, -0.0135, -0.0283, ..., -0.0411, -0.0090, -0.0203], ..., [-0.0180, 0.0850, -0.1004, ..., -0.0596, 0.0120, -0.0706], [-0.0017, -0.0502, 0.0257, ..., 0.0354, -0.0721, -0.0303], [-0.0074, -0.0391, 0.0242, ..., -0.0353, -0.0311, 0.0658]], device='cuda:0'), grad: tensor([[-3.7700e-05, -5.6887e-04, -3.5238e-04, ..., -1.5700e-04, 7.9628e-07, -5.5027e-04], [-5.8651e-05, 6.4336e-06, 2.8871e-06, ..., -3.9399e-05, -3.4660e-05, 4.5933e-06], [ 2.7224e-05, 4.3750e-05, 1.4402e-05, ..., 2.7642e-05, 1.0014e-05, 3.1263e-05], ..., [ 1.2755e-05, -1.4015e-05, 1.3880e-05, ..., 2.2650e-05, 5.7071e-06, 1.6466e-05], [ 1.9923e-05, 2.1070e-05, -5.4762e-06, ..., 1.4119e-05, 1.2733e-05, 4.1693e-05], [-1.4650e-06, 2.9340e-05, -7.1824e-05, ..., 1.9699e-05, 1.0617e-06, -5.6028e-05]], device='cuda:0') Epoch 57, bias, value: tensor([ 0.0069, 0.0026, 0.0054, 0.0164, 0.0233, 0.0330, -0.0264, 0.0075, -0.0316, -0.0188], device='cuda:0'), grad: tensor([-1.1654e-03, -5.0247e-05, 1.2130e-04, 3.6454e-04, 6.3419e-05, -2.0337e-04, 8.6641e-04, 3.7849e-05, 8.4996e-05, -1.1998e-04], device='cuda:0') 100 0.0001 changing lr epoch 56, time 216.82, cls_loss 0.0140 cls_loss_mapping 0.0178 cls_loss_causal 0.7035 re_mapping 0.0155 re_causal 0.0441 /// teacc 98.67 lr 0.00010000 Epoch 58, weight, value: tensor([[-0.0063, 0.0782, 0.0482, ..., -0.0870, -0.0625, 0.0265], [ 0.0747, -0.0080, -0.0270, ..., 0.0361, 0.0550, -0.0744], [-0.0371, -0.0136, -0.0285, ..., -0.0413, -0.0092, -0.0204], ..., [-0.0182, 0.0856, -0.1010, ..., -0.0602, 0.0129, -0.0713], [-0.0018, -0.0506, 0.0254, ..., 0.0353, -0.0725, -0.0307], [-0.0073, -0.0397, 0.0251, ..., -0.0356, -0.0310, 0.0673]], device='cuda:0'), grad: tensor([[ 7.4245e-06, -7.7039e-06, -4.6119e-06, ..., 1.8656e-05, 4.4703e-07, -1.4104e-05], [ 1.2207e-04, 8.3968e-06, 1.5805e-06, ..., 3.3545e-04, -2.0280e-05, 7.0035e-07], [ 1.5795e-05, 2.3327e-03, 1.8024e-04, ..., 3.9756e-05, 4.1485e-04, 5.2676e-06], ..., [ 1.4886e-05, -2.4052e-03, -1.7929e-04, ..., 8.6129e-05, -4.1842e-04, 1.2545e-06], [ 1.2887e-04, 2.2784e-05, 2.0951e-05, ..., 2.2483e-04, 1.0483e-05, 2.3425e-05], [ 1.3240e-05, 8.8453e-05, 3.8654e-05, ..., 6.5660e-04, 1.5451e-06, -1.2599e-05]], device='cuda:0') Epoch 58, bias, value: tensor([ 0.0068, 0.0023, 0.0054, 0.0163, 0.0229, 0.0334, -0.0262, 0.0080, -0.0321, -0.0186], device='cuda:0'), grad: tensor([ 4.0978e-05, 4.1986e-04, 1.2688e-02, 2.4700e-04, -3.2578e-03, 4.8327e-04, -1.5581e-04, -1.2627e-02, 4.1580e-04, 1.7433e-03], device='cuda:0') 100 0.0001 changing lr epoch 57, time 217.10, cls_loss 0.0141 cls_loss_mapping 0.0213 cls_loss_causal 0.7167 re_mapping 0.0146 re_causal 0.0437 /// teacc 98.74 lr 0.00010000 Epoch 59, weight, value: tensor([[-0.0062, 0.0788, 0.0485, ..., -0.0881, -0.0632, 0.0266], [ 0.0745, -0.0083, -0.0274, ..., 0.0359, 0.0554, -0.0753], [-0.0369, -0.0139, -0.0282, ..., -0.0411, -0.0092, -0.0201], ..., [-0.0185, 0.0860, -0.1017, ..., -0.0607, 0.0126, -0.0722], [-0.0019, -0.0511, 0.0253, ..., 0.0355, -0.0730, -0.0310], [-0.0076, -0.0401, 0.0256, ..., -0.0363, -0.0312, 0.0678]], device='cuda:0'), grad: tensor([[-1.0264e-04, -7.8321e-05, -1.4067e-04, ..., 9.2089e-06, 2.2516e-05, -8.8334e-05], [-3.2991e-05, 2.6911e-05, 5.8040e-06, ..., 4.9025e-05, 3.4022e-04, 5.6848e-06], [ 3.0845e-05, 2.8706e-04, 3.5793e-05, ..., -6.6638e-05, -2.7418e-04, 3.1292e-05], ..., [ 1.2018e-05, -7.9012e-04, 5.2303e-06, ..., -5.0694e-05, -2.7752e-04, -4.3809e-05], [ 2.9564e-05, 6.5088e-05, 3.9190e-05, ..., 1.8075e-05, 9.6560e-06, 3.4600e-05], [ 8.2031e-06, 8.7380e-05, -2.1964e-05, ..., 1.9699e-05, 8.3074e-06, -1.8343e-05]], device='cuda:0') Epoch 59, bias, value: tensor([ 0.0071, 0.0019, 0.0055, 0.0161, 0.0241, 0.0343, -0.0266, 0.0077, -0.0326, -0.0190], device='cuda:0'), grad: tensor([-0.0003, 0.0010, -0.0006, 0.0002, 0.0003, 0.0003, 0.0001, -0.0014, 0.0002, 0.0002], device='cuda:0') 100 0.0001 changing lr epoch 58, time 216.93, cls_loss 0.0141 cls_loss_mapping 0.0184 cls_loss_causal 0.6714 re_mapping 0.0145 re_causal 0.0426 /// teacc 98.66 lr 0.00010000 Epoch 60, weight, value: tensor([[-0.0059, 0.0800, 0.0488, ..., -0.0887, -0.0636, 0.0264], [ 0.0749, -0.0084, -0.0276, ..., 0.0366, 0.0560, -0.0762], [-0.0377, -0.0142, -0.0287, ..., -0.0420, -0.0095, -0.0203], ..., [-0.0190, 0.0856, -0.1028, ..., -0.0615, 0.0127, -0.0728], [-0.0015, -0.0508, 0.0254, ..., 0.0360, -0.0734, -0.0311], [-0.0079, -0.0397, 0.0267, ..., -0.0365, -0.0312, 0.0688]], device='cuda:0'), grad: tensor([[ 2.7549e-06, -4.5598e-05, -2.8729e-05, ..., 2.9355e-06, 1.2685e-06, -3.1739e-05], [-1.2183e-04, 7.7724e-05, 3.3434e-06, ..., -1.0091e-04, -8.9645e-05, 7.8604e-07], [ 1.0264e-04, 1.5706e-05, 2.6245e-06, ..., 1.3316e-04, 5.3167e-05, 8.5831e-06], ..., [-1.1331e-04, -2.0719e-04, 1.0490e-05, ..., -2.5153e-04, 2.9560e-06, 5.5209e-06], [ 6.9797e-05, 4.7773e-05, 5.2035e-05, ..., 7.9691e-05, 1.6987e-05, 2.7657e-05], [ 1.8865e-05, 3.5524e-05, -1.0617e-06, ..., 2.4870e-05, 2.7437e-06, -1.1690e-05]], device='cuda:0') Epoch 60, bias, value: tensor([ 0.0079, 0.0020, 0.0051, 0.0161, 0.0241, 0.0340, -0.0268, 0.0065, -0.0323, -0.0181], device='cuda:0'), grad: tensor([-9.3937e-05, -4.9174e-07, 1.3185e-04, 1.3161e-03, 1.7309e-04, -1.3561e-03, 8.8573e-05, -6.1941e-04, 2.7013e-04, 8.9347e-05], device='cuda:0') 100 0.0001 changing lr epoch 59, time 217.49, cls_loss 0.0158 cls_loss_mapping 0.0236 cls_loss_causal 0.7000 re_mapping 0.0148 re_causal 0.0413 /// teacc 98.57 lr 0.00010000 Epoch 61, weight, value: tensor([[-0.0064, 0.0803, 0.0485, ..., -0.0896, -0.0641, 0.0264], [ 0.0757, -0.0085, -0.0272, ..., 0.0376, 0.0563, -0.0758], [-0.0378, -0.0144, -0.0286, ..., -0.0424, -0.0100, -0.0205], ..., [-0.0191, 0.0865, -0.1037, ..., -0.0619, 0.0148, -0.0736], [-0.0017, -0.0513, 0.0254, ..., 0.0361, -0.0738, -0.0313], [-0.0087, -0.0402, 0.0273, ..., -0.0378, -0.0318, 0.0695]], device='cuda:0'), grad: tensor([[-3.6001e-05, -4.8876e-05, -2.0847e-05, ..., 2.5630e-05, 1.0058e-05, 1.0848e-05], [-1.5214e-05, 1.0318e-04, 1.1158e-04, ..., 1.3828e-04, 9.2864e-05, 5.5492e-05], [ 7.4580e-06, 3.4213e-05, -2.0909e-04, ..., -2.2459e-04, -2.7394e-04, -8.7738e-05], ..., [-2.1189e-05, -8.4686e-04, 6.3241e-05, ..., 5.7906e-05, 9.3997e-05, 4.6194e-05], [ 6.6906e-06, 2.2754e-05, 1.8895e-05, ..., 4.0054e-05, 1.1064e-05, 2.0280e-05], [ 2.4870e-05, 5.4979e-04, -5.8532e-05, ..., 3.5954e-04, 1.7509e-07, -7.2896e-05]], device='cuda:0') Epoch 61, bias, value: tensor([ 0.0075, 0.0022, 0.0052, 0.0156, 0.0244, 0.0335, -0.0263, 0.0078, -0.0329, -0.0184], device='cuda:0'), grad: tensor([ 1.6555e-05, 8.6069e-04, -1.5326e-03, 3.6836e-04, -4.9496e-04, 8.4281e-05, 3.4332e-05, -6.1321e-04, 1.3590e-04, 1.1377e-03], device='cuda:0') 100 0.0001 changing lr epoch 60, time 217.04, cls_loss 0.0131 cls_loss_mapping 0.0191 cls_loss_causal 0.6444 re_mapping 0.0146 re_causal 0.0409 /// teacc 98.56 lr 0.00010000 Epoch 62, weight, value: tensor([[-0.0057, 0.0814, 0.0496, ..., -0.0903, -0.0650, 0.0265], [ 0.0764, -0.0090, -0.0274, ..., 0.0382, 0.0570, -0.0762], [-0.0386, -0.0150, -0.0290, ..., -0.0432, -0.0104, -0.0210], ..., [-0.0193, 0.0874, -0.1044, ..., -0.0625, 0.0144, -0.0744], [-0.0020, -0.0527, 0.0252, ..., 0.0362, -0.0743, -0.0318], [-0.0094, -0.0403, 0.0273, ..., -0.0388, -0.0320, 0.0703]], device='cuda:0'), grad: tensor([[ 7.6115e-05, -3.1024e-05, 1.0109e-04, ..., 7.2300e-05, -1.8366e-06, 1.8024e-04], [ 4.7231e-04, 2.7388e-05, 8.5354e-05, ..., 3.2473e-04, 3.9721e-04, 1.2493e-04], [-3.5000e-04, 2.7984e-05, 7.5996e-05, ..., -1.1760e-04, -4.0269e-04, 1.2982e-04], ..., [ 3.3200e-05, -5.7928e-07, 6.1333e-05, ..., 8.2314e-05, 3.0492e-06, 1.0407e-04], [-2.3574e-05, 2.0832e-05, 3.5822e-05, ..., -6.3181e-05, 1.6242e-06, 9.0718e-05], [-8.9359e-04, -4.6730e-04, -1.7204e-03, ..., -1.0118e-03, 7.9349e-07, -2.9564e-03]], device='cuda:0') Epoch 62, bias, value: tensor([ 0.0074, 0.0025, 0.0047, 0.0162, 0.0246, 0.0335, -0.0264, 0.0077, -0.0333, -0.0184], device='cuda:0'), grad: tensor([ 3.3760e-04, 1.3380e-03, -6.9761e-04, 2.0027e-05, 3.7498e-03, 7.6830e-05, 1.3101e-04, 2.8086e-04, 5.6058e-05, -5.2910e-03], device='cuda:0') 100 0.0001 changing lr epoch 61, time 217.28, cls_loss 0.0139 cls_loss_mapping 0.0198 cls_loss_causal 0.7196 re_mapping 0.0132 re_causal 0.0402 /// teacc 98.61 lr 0.00010000 Epoch 63, weight, value: tensor([[-0.0061, 0.0817, 0.0495, ..., -0.0919, -0.0660, 0.0264], [ 0.0766, -0.0088, -0.0279, ..., 0.0386, 0.0576, -0.0777], [-0.0389, -0.0149, -0.0293, ..., -0.0438, -0.0101, -0.0214], ..., [-0.0199, 0.0882, -0.1050, ..., -0.0635, 0.0141, -0.0752], [-0.0022, -0.0538, 0.0251, ..., 0.0364, -0.0745, -0.0327], [-0.0094, -0.0411, 0.0278, ..., -0.0393, -0.0319, 0.0711]], device='cuda:0'), grad: tensor([[-2.7943e-04, -4.9829e-04, -5.8556e-04, ..., 9.2909e-06, 1.7975e-06, -3.7694e-04], [-6.9666e-04, 1.5702e-06, 3.4496e-06, ..., -3.9101e-04, -5.6934e-04, 3.7476e-06], [-3.1257e-04, 5.3465e-05, 5.9664e-05, ..., 2.9802e-04, 5.3263e-04, 3.7163e-05], ..., [ 3.3885e-05, 8.4788e-06, 1.1921e-05, ..., 2.7940e-05, 1.9848e-05, 1.1757e-05], [ 1.0443e-03, 3.4958e-05, 6.8307e-05, ..., 1.3506e-04, 1.8120e-05, 1.0216e-04], [ 2.1100e-05, 5.4777e-05, 4.6879e-05, ..., 4.5240e-05, 2.7418e-06, 2.7537e-05]], device='cuda:0') Epoch 63, bias, value: tensor([ 0.0069, 0.0023, 0.0047, 0.0165, 0.0247, 0.0333, -0.0252, 0.0077, -0.0336, -0.0187], device='cuda:0'), grad: tensor([-1.0357e-03, -1.1702e-03, -5.3501e-04, 1.4496e-04, -6.0946e-05, -9.3746e-04, 1.5068e-03, 8.7798e-05, 1.8692e-03, 1.3113e-04], device='cuda:0') 100 0.0001 changing lr epoch 62, time 217.46, cls_loss 0.0137 cls_loss_mapping 0.0196 cls_loss_causal 0.6784 re_mapping 0.0147 re_causal 0.0403 /// teacc 98.62 lr 0.00010000 Epoch 64, weight, value: tensor([[-0.0063, 0.0821, 0.0497, ..., -0.0928, -0.0667, 0.0263], [ 0.0774, -0.0088, -0.0282, ..., 0.0393, 0.0582, -0.0785], [-0.0388, -0.0155, -0.0294, ..., -0.0437, -0.0097, -0.0213], ..., [-0.0203, 0.0890, -0.1059, ..., -0.0642, 0.0144, -0.0757], [-0.0020, -0.0538, 0.0249, ..., 0.0368, -0.0756, -0.0326], [-0.0102, -0.0414, 0.0283, ..., -0.0398, -0.0327, 0.0717]], device='cuda:0'), grad: tensor([[ 1.0885e-05, -1.4317e-04, -1.2624e-04, ..., 1.1876e-05, 4.7311e-07, -1.3590e-04], [-5.2154e-05, -5.8077e-06, 9.4175e-06, ..., -3.3021e-05, -2.3827e-05, 5.2825e-06], [ 1.4387e-05, 1.9521e-05, 2.7299e-05, ..., 6.9886e-06, 2.9150e-06, 1.8775e-05], ..., [ 3.1650e-05, -7.1898e-06, 4.3184e-05, ..., 1.3903e-05, 8.9481e-06, 2.5570e-05], [ 2.8163e-05, 5.3719e-06, 6.1333e-05, ..., 1.8656e-05, 8.5235e-06, 6.2883e-05], [-3.3188e-04, 1.2837e-05, -1.1730e-03, ..., -4.1294e-04, 2.5332e-06, -1.0061e-03]], device='cuda:0') Epoch 64, bias, value: tensor([ 0.0069, 0.0026, 0.0053, 0.0162, 0.0246, 0.0337, -0.0256, 0.0078, -0.0342, -0.0188], device='cuda:0'), grad: tensor([-3.3283e-04, -4.1038e-05, 5.9336e-05, 2.8634e-04, 1.2703e-03, 7.3481e-04, 6.8069e-05, 8.5473e-05, 1.0389e-04, -2.2354e-03], device='cuda:0') 100 0.0001 changing lr epoch 63, time 217.58, cls_loss 0.0159 cls_loss_mapping 0.0185 cls_loss_causal 0.6907 re_mapping 0.0142 re_causal 0.0397 /// teacc 98.58 lr 0.00010000 Epoch 65, weight, value: tensor([[-0.0067, 0.0835, 0.0504, ..., -0.0939, -0.0670, 0.0266], [ 0.0786, -0.0071, -0.0291, ..., 0.0400, 0.0591, -0.0796], [-0.0394, -0.0160, -0.0296, ..., -0.0441, -0.0098, -0.0213], ..., [-0.0220, 0.0886, -0.1073, ..., -0.0651, 0.0135, -0.0763], [-0.0007, -0.0532, 0.0262, ..., 0.0384, -0.0760, -0.0321], [-0.0111, -0.0413, 0.0287, ..., -0.0406, -0.0324, 0.0721]], device='cuda:0'), grad: tensor([[ 6.2108e-05, -1.3180e-05, 8.1062e-05, ..., 6.0737e-05, 4.2729e-06, 9.2089e-05], [ 1.8373e-05, 1.9372e-04, 5.7638e-05, ..., -3.5077e-05, 3.9905e-05, 3.6895e-05], [ 3.0947e-04, -4.1628e-04, 4.4227e-04, ..., 1.9014e-04, 1.9521e-05, 3.7313e-04], ..., [-5.9754e-06, 1.9526e-04, 1.0705e-04, ..., 4.7445e-05, -8.5711e-05, 7.5042e-05], [ 7.3433e-04, 1.4782e-04, 8.2493e-04, ..., 4.6515e-04, 2.5123e-05, 7.3671e-04], [-1.2434e-04, 2.4348e-05, -4.8780e-04, ..., -8.9526e-05, 3.8832e-05, -4.0293e-04]], device='cuda:0') Epoch 65, bias, value: tensor([ 0.0076, 0.0037, 0.0051, 0.0154, 0.0246, 0.0334, -0.0268, 0.0068, -0.0326, -0.0187], device='cuda:0'), grad: tensor([ 0.0002, 0.0004, -0.0003, 0.0087, 0.0012, -0.0111, -0.0009, 0.0009, 0.0018, -0.0009], device='cuda:0') 100 0.0001 changing lr epoch 64, time 217.59, cls_loss 0.0151 cls_loss_mapping 0.0191 cls_loss_causal 0.6702 re_mapping 0.0140 re_causal 0.0397 /// teacc 98.72 lr 0.00010000 Epoch 66, weight, value: tensor([[-0.0059, 0.0844, 0.0504, ..., -0.0948, -0.0673, 0.0266], [ 0.0783, -0.0088, -0.0293, ..., 0.0399, 0.0596, -0.0797], [-0.0400, -0.0165, -0.0296, ..., -0.0446, -0.0098, -0.0215], ..., [-0.0209, 0.0892, -0.1088, ..., -0.0652, 0.0136, -0.0766], [-0.0009, -0.0538, 0.0265, ..., 0.0382, -0.0768, -0.0320], [-0.0124, -0.0419, 0.0289, ..., -0.0413, -0.0329, 0.0723]], device='cuda:0'), grad: tensor([[ 8.7738e-05, -6.5029e-05, 2.3276e-05, ..., 8.7082e-05, 1.0896e-06, 1.6510e-05], [-2.5511e-04, -1.3387e-04, -3.3021e-05, ..., -1.3030e-04, -1.2052e-04, 1.1414e-05], [ 7.6473e-05, 1.3256e-04, 3.6538e-05, ..., 6.0320e-05, 2.7046e-05, 9.2313e-06], ..., [ 1.2195e-04, -6.8378e-04, 3.7700e-06, ..., 1.1122e-04, 5.1707e-05, 9.2909e-06], [ 8.0109e-03, 4.5300e-05, 4.1618e-03, ..., 6.7596e-03, 1.1295e-05, 4.8904e-03], [ 1.2696e-04, 1.0109e-04, 2.4376e-03, ..., 2.9640e-03, 2.0027e-05, 9.9659e-04]], device='cuda:0') Epoch 66, bias, value: tensor([ 0.0079, 0.0029, 0.0043, 0.0164, 0.0248, 0.0332, -0.0260, 0.0073, -0.0331, -0.0189], device='cuda:0'), grad: tensor([ 5.0336e-05, -6.1035e-04, 3.8457e-04, 9.7656e-04, -5.9509e-03, 8.2111e-04, -1.2543e-02, -9.3555e-04, 1.1742e-02, 6.0616e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 65---------------------------------------------------- epoch 65, time 218.36, cls_loss 0.0114 cls_loss_mapping 0.0160 cls_loss_causal 0.6463 re_mapping 0.0145 re_causal 0.0400 /// teacc 98.81 lr 0.00010000 Epoch 67, weight, value: tensor([[-0.0064, 0.0848, 0.0500, ..., -0.0959, -0.0678, 0.0265], [ 0.0787, -0.0092, -0.0298, ..., 0.0400, 0.0600, -0.0805], [-0.0405, -0.0169, -0.0296, ..., -0.0453, -0.0100, -0.0215], ..., [-0.0207, 0.0903, -0.1092, ..., -0.0658, 0.0138, -0.0772], [-0.0010, -0.0542, 0.0264, ..., 0.0384, -0.0772, -0.0324], [-0.0123, -0.0421, 0.0296, ..., -0.0421, -0.0330, 0.0728]], device='cuda:0'), grad: tensor([[-4.0829e-05, -3.4600e-05, -9.2387e-05, ..., -8.8885e-06, 8.4005e-07, -1.4961e-04], [-4.6700e-05, -2.5667e-06, -7.4804e-06, ..., -4.0352e-05, -2.0251e-05, 3.1628e-06], [-5.2806e-07, 2.0117e-06, 2.0284e-06, ..., -5.1744e-06, 1.4510e-06, 8.6650e-06], ..., [ 3.6895e-05, 1.3039e-07, 9.8869e-06, ..., 2.7433e-05, 6.1281e-06, 9.8497e-06], [ 1.5616e-04, 1.7822e-05, 1.1545e-04, ..., 1.2434e-04, 8.0615e-06, 1.3888e-04], [ 3.5882e-05, 7.1041e-06, 2.3231e-05, ..., 2.7299e-05, 2.4233e-06, 4.9502e-05]], device='cuda:0') Epoch 67, bias, value: tensor([ 0.0076, 0.0023, 0.0042, 0.0159, 0.0258, 0.0335, -0.0260, 0.0077, -0.0334, -0.0188], device='cuda:0'), grad: tensor([ 8.9705e-05, -4.6015e-05, -5.7369e-05, 1.8806e-03, -1.0958e-03, -2.1915e-03, 1.5533e-04, 7.4983e-05, 3.8362e-04, 8.0585e-04], device='cuda:0') 100 0.0001 changing lr epoch 66, time 217.29, cls_loss 0.0129 cls_loss_mapping 0.0183 cls_loss_causal 0.6694 re_mapping 0.0137 re_causal 0.0393 /// teacc 98.53 lr 0.00010000 Epoch 68, weight, value: tensor([[-0.0067, 0.0847, 0.0504, ..., -0.0963, -0.0683, 0.0259], [ 0.0792, -0.0088, -0.0299, ..., 0.0404, 0.0611, -0.0807], [-0.0406, -0.0175, -0.0297, ..., -0.0455, -0.0099, -0.0220], ..., [-0.0216, 0.0904, -0.1104, ..., -0.0662, 0.0130, -0.0775], [-0.0004, -0.0543, 0.0266, ..., 0.0389, -0.0774, -0.0323], [-0.0130, -0.0413, 0.0303, ..., -0.0427, -0.0342, 0.0739]], device='cuda:0'), grad: tensor([[-1.1879e-04, -4.3607e-04, -4.4179e-04, ..., -1.1015e-04, 4.8727e-06, -5.3358e-04], [-1.5218e-06, 2.1234e-06, 1.0572e-05, ..., -4.4703e-06, -6.7018e-06, 8.8513e-06], [-5.8174e-05, 1.3530e-04, -3.5191e-04, ..., -1.5998e-04, -4.3392e-05, -2.0611e-04], ..., [ 2.7394e-04, 7.0691e-05, 1.4305e-04, ..., 4.0710e-05, 4.3735e-06, 2.2125e-04], [ 3.5810e-04, -1.0556e-04, 6.7711e-05, ..., 3.6907e-04, 4.3821e-04, -9.7632e-05], [ 1.4353e-04, 2.9850e-04, 6.4039e-04, ..., 1.3578e-04, 1.2949e-05, 7.1764e-04]], device='cuda:0') Epoch 68, bias, value: tensor([ 0.0068, 0.0028, 0.0046, 0.0153, 0.0250, 0.0336, -0.0260, 0.0073, -0.0333, -0.0178], device='cuda:0'), grad: tensor([-8.6784e-04, 3.1382e-05, -8.8644e-04, 2.1648e-04, 5.0403e-06, -9.8801e-04, 2.2995e-04, 8.0776e-04, 8.7082e-05, 1.3628e-03], device='cuda:0') 100 0.0001 changing lr epoch 67, time 217.70, cls_loss 0.0099 cls_loss_mapping 0.0130 cls_loss_causal 0.6772 re_mapping 0.0132 re_causal 0.0392 /// teacc 98.64 lr 0.00010000 Epoch 69, weight, value: tensor([[-0.0061, 0.0856, 0.0505, ..., -0.0967, -0.0683, 0.0261], [ 0.0789, -0.0103, -0.0302, ..., 0.0403, 0.0613, -0.0813], [-0.0409, -0.0172, -0.0301, ..., -0.0458, -0.0100, -0.0222], ..., [-0.0211, 0.0915, -0.1106, ..., -0.0665, 0.0129, -0.0779], [-0.0005, -0.0546, 0.0263, ..., 0.0390, -0.0782, -0.0326], [-0.0136, -0.0420, 0.0303, ..., -0.0437, -0.0353, 0.0742]], device='cuda:0'), grad: tensor([[ 7.1079e-06, -5.7220e-06, -5.5790e-05, ..., 4.2468e-06, 5.6205e-07, -7.3850e-05], [ 4.1574e-06, 2.0172e-06, 7.4804e-06, ..., 1.6354e-06, -9.4529e-08, 3.7067e-06], [ 1.2867e-05, 5.4948e-07, 2.1771e-05, ..., 1.7434e-06, -1.0645e-06, 9.2089e-06], ..., [ 3.5688e-06, -7.2718e-05, 8.7246e-06, ..., 1.7844e-06, 3.2922e-07, -8.6352e-06], [ 3.6627e-05, 3.1404e-06, 6.6459e-05, ..., 3.8743e-06, 1.8049e-06, 1.9982e-05], [ 1.4193e-05, 6.3896e-05, 5.2840e-05, ..., 7.6741e-06, 8.7684e-07, 6.5565e-05]], device='cuda:0') Epoch 69, bias, value: tensor([ 0.0071, 0.0018, 0.0049, 0.0155, 0.0256, 0.0345, -0.0262, 0.0077, -0.0337, -0.0186], device='cuda:0'), grad: tensor([-6.5327e-05, 1.7598e-05, 3.3289e-05, -2.3615e-04, 7.6964e-06, 6.0469e-05, -4.7758e-06, -1.5426e-04, 1.1778e-04, 2.2399e-04], device='cuda:0') 100 0.0001 changing lr epoch 68, time 217.55, cls_loss 0.0102 cls_loss_mapping 0.0148 cls_loss_causal 0.6548 re_mapping 0.0137 re_causal 0.0374 /// teacc 98.75 lr 0.00010000 Epoch 70, weight, value: tensor([[-0.0061, 0.0859, 0.0508, ..., -0.0975, -0.0685, 0.0264], [ 0.0794, -0.0104, -0.0310, ..., 0.0408, 0.0621, -0.0818], [-0.0414, -0.0170, -0.0297, ..., -0.0465, -0.0103, -0.0224], ..., [-0.0217, 0.0919, -0.1116, ..., -0.0673, 0.0133, -0.0785], [-0.0004, -0.0549, 0.0266, ..., 0.0396, -0.0791, -0.0326], [-0.0141, -0.0424, 0.0312, ..., -0.0442, -0.0355, 0.0747]], device='cuda:0'), grad: tensor([[ 1.4611e-05, -3.7206e-07, 1.1660e-06, ..., 1.4313e-05, 1.3430e-06, 1.7239e-06], [ 5.8460e-04, 6.6757e-06, 2.5947e-06, ..., 4.4775e-04, -1.4141e-05, 5.1886e-05], [ 3.6955e-05, 8.0466e-06, 3.2187e-06, ..., 3.3587e-05, 1.3188e-05, 1.5497e-06], ..., [ 6.6683e-06, -9.3162e-05, 1.5765e-05, ..., 1.6510e-05, 5.7667e-06, 2.1487e-05], [-9.0361e-04, 3.8929e-06, -1.6287e-05, ..., -6.6948e-04, -2.9117e-05, -5.1916e-05], [ 3.4273e-05, 2.2680e-05, -1.1677e-04, ..., -2.5090e-06, 2.6971e-06, -1.1867e-04]], device='cuda:0') Epoch 70, bias, value: tensor([ 0.0071, 0.0019, 0.0054, 0.0151, 0.0247, 0.0346, -0.0260, 0.0076, -0.0337, -0.0182], device='cuda:0'), grad: tensor([ 1.5244e-05, 5.9795e-04, -9.0078e-06, 2.0039e-04, 1.3351e-04, -1.3447e-04, 2.6369e-04, -9.4414e-05, -8.7166e-04, -1.0192e-04], device='cuda:0') 100 0.0001 changing lr epoch 69, time 217.81, cls_loss 0.0091 cls_loss_mapping 0.0133 cls_loss_causal 0.6313 re_mapping 0.0134 re_causal 0.0387 /// teacc 98.74 lr 0.00010000 Epoch 71, weight, value: tensor([[-0.0069, 0.0862, 0.0505, ..., -0.0986, -0.0698, 0.0260], [ 0.0802, -0.0104, -0.0310, ..., 0.0418, 0.0631, -0.0820], [-0.0417, -0.0172, -0.0298, ..., -0.0470, -0.0102, -0.0225], ..., [-0.0228, 0.0929, -0.1125, ..., -0.0682, 0.0130, -0.0789], [-0.0002, -0.0549, 0.0265, ..., 0.0396, -0.0802, -0.0328], [-0.0141, -0.0429, 0.0318, ..., -0.0443, -0.0359, 0.0756]], device='cuda:0'), grad: tensor([[ 1.2174e-05, -1.6883e-05, 3.7476e-06, ..., 2.1532e-05, 2.6263e-07, 8.5831e-06], [ 1.7667e-04, 2.3925e-04, 6.5845e-07, ..., 1.0449e-04, 6.3837e-05, 6.2957e-07], [ 1.5073e-05, 8.5086e-06, 5.4054e-06, ..., 1.7419e-05, 1.1502e-06, 2.0955e-06], ..., [-3.0017e-04, -3.9744e-04, 9.9279e-07, ..., -1.7309e-04, -1.1367e-04, 8.9640e-07], [ 5.6118e-05, 8.1599e-05, -2.5555e-06, ..., 3.1948e-05, 2.5213e-05, 1.6093e-05], [ 2.8118e-05, 4.0352e-05, -3.1460e-06, ..., 3.4064e-05, 9.3207e-06, 2.2352e-07]], device='cuda:0') Epoch 71, bias, value: tensor([ 0.0060, 0.0025, 0.0055, 0.0147, 0.0244, 0.0354, -0.0262, 0.0075, -0.0338, -0.0179], device='cuda:0'), grad: tensor([ 1.5227e-07, 5.0735e-04, 3.9279e-05, 3.6538e-05, -2.3767e-05, 1.3840e-04, -1.3018e-04, -8.3303e-04, 1.6761e-04, 9.7513e-05], device='cuda:0') 100 0.0001 changing lr epoch 70, time 217.40, cls_loss 0.0126 cls_loss_mapping 0.0166 cls_loss_causal 0.6516 re_mapping 0.0130 re_causal 0.0374 /// teacc 98.71 lr 0.00010000 Epoch 72, weight, value: tensor([[-0.0067, 0.0869, 0.0529, ..., -0.0992, -0.0705, 0.0273], [ 0.0809, -0.0106, -0.0307, ..., 0.0424, 0.0632, -0.0820], [-0.0422, -0.0176, -0.0301, ..., -0.0474, -0.0099, -0.0228], ..., [-0.0225, 0.0936, -0.1124, ..., -0.0690, 0.0130, -0.0793], [-0.0008, -0.0556, 0.0262, ..., 0.0393, -0.0819, -0.0334], [-0.0148, -0.0439, 0.0310, ..., -0.0449, -0.0365, 0.0752]], device='cuda:0'), grad: tensor([[ 9.8497e-06, -3.2037e-06, -6.2250e-06, ..., 1.1481e-05, 2.3060e-06, -5.7779e-06], [-6.1572e-05, -4.6343e-06, 4.9993e-06, ..., -5.3078e-05, -3.4392e-05, 5.6177e-06], [ 1.5140e-05, 2.1495e-06, 4.1164e-06, ..., 1.6630e-05, 6.0350e-06, 1.0896e-06], ..., [ 6.4313e-05, 3.0845e-06, 7.2241e-05, ..., 1.0329e-04, 2.1175e-05, 5.4091e-05], [-6.4135e-05, 5.6773e-06, -8.5356e-07, ..., -6.0588e-05, 9.4250e-06, 1.5572e-05], [ 1.2189e-05, -8.6427e-06, -9.5069e-05, ..., -2.7508e-05, -1.5453e-05, -8.5175e-05]], device='cuda:0') Epoch 72, bias, value: tensor([ 0.0071, 0.0024, 0.0056, 0.0147, 0.0245, 0.0361, -0.0265, 0.0079, -0.0348, -0.0188], device='cuda:0'), grad: tensor([ 7.4040e-08, -6.1214e-05, 3.0249e-05, 1.6212e-05, 4.0233e-05, 3.2842e-05, -2.7925e-05, 3.0589e-04, -7.6711e-05, -2.5964e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 71---------------------------------------------------- epoch 71, time 217.41, cls_loss 0.0109 cls_loss_mapping 0.0170 cls_loss_causal 0.6640 re_mapping 0.0126 re_causal 0.0361 /// teacc 98.87 lr 0.00010000 Epoch 73, weight, value: tensor([[-0.0070, 0.0869, 0.0538, ..., -0.0998, -0.0714, 0.0272], [ 0.0812, -0.0113, -0.0318, ..., 0.0426, 0.0634, -0.0830], [-0.0425, -0.0180, -0.0307, ..., -0.0477, -0.0095, -0.0232], ..., [-0.0229, 0.0946, -0.1128, ..., -0.0704, 0.0135, -0.0801], [-0.0006, -0.0555, 0.0264, ..., 0.0398, -0.0828, -0.0334], [-0.0151, -0.0442, 0.0311, ..., -0.0451, -0.0373, 0.0765]], device='cuda:0'), grad: tensor([[-1.0394e-05, -3.7611e-05, -2.0280e-05, ..., -2.5451e-05, 7.8510e-07, -4.9055e-05], [-1.4329e-04, 3.8818e-06, -6.0126e-06, ..., -1.4424e-04, -5.3257e-05, -5.5045e-05], [ 5.5842e-06, 1.2495e-05, 3.7551e-06, ..., 4.8056e-06, 8.3214e-07, 2.5835e-06], ..., [ 1.9431e-05, -8.4400e-05, 2.6636e-06, ..., 2.3872e-05, 6.2808e-06, 8.7023e-06], [ 6.5386e-05, 8.8662e-06, 5.5321e-06, ..., 6.2704e-05, 2.3872e-05, 3.3528e-05], [ 3.3885e-05, 2.2873e-05, 6.9961e-06, ..., 4.6581e-05, 9.6411e-06, 1.9372e-05]], device='cuda:0') Epoch 73, bias, value: tensor([ 0.0065, 0.0018, 0.0056, 0.0146, 0.0240, 0.0366, -0.0268, 0.0081, -0.0344, -0.0182], device='cuda:0'), grad: tensor([-1.2529e-04, -1.6105e-04, -5.3570e-06, 4.1306e-05, 1.2541e-04, 7.3135e-05, 2.2769e-05, -2.3019e-04, 1.3590e-04, 1.2326e-04], device='cuda:0') 100 0.0001 changing lr epoch 72, time 216.82, cls_loss 0.0108 cls_loss_mapping 0.0152 cls_loss_causal 0.6341 re_mapping 0.0127 re_causal 0.0352 /// teacc 98.71 lr 0.00010000 Epoch 74, weight, value: tensor([[-0.0074, 0.0874, 0.0539, ..., -0.1007, -0.0720, 0.0272], [ 0.0823, -0.0116, -0.0312, ..., 0.0439, 0.0648, -0.0829], [-0.0431, -0.0183, -0.0310, ..., -0.0482, -0.0100, -0.0233], ..., [-0.0233, 0.0948, -0.1136, ..., -0.0713, 0.0145, -0.0805], [-0.0009, -0.0559, 0.0262, ..., 0.0398, -0.0836, -0.0336], [-0.0154, -0.0439, 0.0316, ..., -0.0460, -0.0369, 0.0768]], device='cuda:0'), grad: tensor([[ 9.8228e-05, -4.9639e-07, 1.1140e-04, ..., 6.6638e-05, 1.7500e-06, 4.6313e-05], [ 3.4552e-06, 4.7609e-06, 1.4238e-05, ..., 8.5607e-06, 3.4183e-05, 1.1042e-05], [ 4.8131e-05, 6.1393e-06, 3.2008e-05, ..., -4.3735e-06, -5.2482e-05, 3.5733e-05], ..., [ 8.7619e-06, -9.7513e-05, 1.2994e-05, ..., 1.0863e-05, 7.2531e-06, 1.2279e-05], [-3.9291e-04, 2.0303e-06, -3.9721e-04, ..., -1.5020e-04, 1.8835e-05, -2.1887e-04], [ 1.7822e-04, 4.2766e-05, 7.4446e-05, ..., -8.3372e-06, -1.1139e-05, 9.0450e-06]], device='cuda:0') Epoch 74, bias, value: tensor([ 0.0065, 0.0027, 0.0054, 0.0147, 0.0241, 0.0363, -0.0269, 0.0080, -0.0350, -0.0182], device='cuda:0'), grad: tensor([ 5.6952e-05, 2.4819e-04, -1.5855e-04, 1.7858e-04, 2.2602e-04, -1.7196e-05, 3.1292e-05, -1.2177e-04, -4.9973e-04, 5.4419e-05], device='cuda:0') 100 0.0001 changing lr epoch 73, time 216.84, cls_loss 0.0106 cls_loss_mapping 0.0136 cls_loss_causal 0.6510 re_mapping 0.0127 re_causal 0.0354 /// teacc 98.71 lr 0.00010000 Epoch 75, weight, value: tensor([[-0.0081, 0.0876, 0.0538, ..., -0.1032, -0.0736, 0.0268], [ 0.0829, -0.0119, -0.0313, ..., 0.0445, 0.0652, -0.0837], [-0.0441, -0.0187, -0.0313, ..., -0.0495, -0.0102, -0.0234], ..., [-0.0236, 0.0958, -0.1143, ..., -0.0723, 0.0149, -0.0811], [-0.0006, -0.0563, 0.0268, ..., 0.0409, -0.0844, -0.0335], [-0.0160, -0.0445, 0.0321, ..., -0.0464, -0.0354, 0.0778]], device='cuda:0'), grad: tensor([[ 2.3139e-04, 9.2030e-05, 1.0246e-04, ..., 2.6774e-04, 3.4180e-07, 2.2006e-04], [-1.0937e-05, 1.4137e-06, 1.5378e-05, ..., -2.8774e-05, -2.2203e-05, 9.5442e-06], [ 1.4804e-05, 8.8988e-07, 2.1607e-05, ..., 1.0617e-05, 4.1462e-06, 1.2010e-05], ..., [ 4.6849e-05, -9.2685e-06, 6.4194e-05, ..., 3.7998e-05, 8.6278e-06, 5.5254e-05], [ 3.7527e-04, 1.5366e-04, 1.7643e-04, ..., 4.2915e-04, 1.8552e-06, 3.5620e-04], [ 6.9141e-05, 1.7673e-05, 1.2338e-04, ..., 5.2959e-05, 1.6727e-06, 1.1003e-04]], device='cuda:0') Epoch 75, bias, value: tensor([ 0.0060, 0.0026, 0.0047, 0.0154, 0.0244, 0.0351, -0.0266, 0.0084, -0.0345, -0.0181], device='cuda:0'), grad: tensor([ 5.8842e-04, 1.5143e-06, 6.8724e-05, -2.9335e-03, 7.7009e-05, -4.3335e-03, 4.8714e-03, 2.4652e-04, 9.7942e-04, 4.3201e-04], device='cuda:0') 100 0.0001 changing lr epoch 74, time 217.00, cls_loss 0.0095 cls_loss_mapping 0.0162 cls_loss_causal 0.6392 re_mapping 0.0128 re_causal 0.0361 /// teacc 98.85 lr 0.00010000 Epoch 76, weight, value: tensor([[-0.0084, 0.0879, 0.0538, ..., -0.1038, -0.0738, 0.0265], [ 0.0834, -0.0118, -0.0318, ..., 0.0456, 0.0662, -0.0840], [-0.0445, -0.0190, -0.0313, ..., -0.0499, -0.0106, -0.0232], ..., [-0.0241, 0.0964, -0.1151, ..., -0.0740, 0.0146, -0.0817], [-0.0004, -0.0567, 0.0272, ..., 0.0410, -0.0847, -0.0336], [-0.0166, -0.0449, 0.0323, ..., -0.0478, -0.0358, 0.0785]], device='cuda:0'), grad: tensor([[ 1.3940e-05, -2.0057e-05, 4.6119e-06, ..., 1.8671e-05, 3.4366e-06, -5.1633e-06], [-2.8685e-06, 3.0901e-06, 5.4203e-06, ..., 7.7933e-06, -1.1008e-06, 7.3053e-06], [ 8.2105e-06, 3.9563e-06, -8.0019e-06, ..., -5.2601e-05, 1.2830e-05, -1.0818e-05], ..., [ 2.1100e-05, -2.0355e-05, 4.2498e-05, ..., 3.1829e-05, 1.0930e-05, 4.6939e-05], [ 3.0726e-05, 3.0138e-06, 4.5061e-05, ..., 4.4256e-05, 7.9796e-06, 4.3958e-05], [ 2.4819e-04, 1.9744e-05, 2.7561e-04, ..., 2.1684e-04, 1.4400e-04, 2.8157e-04]], device='cuda:0') Epoch 76, bias, value: tensor([ 0.0054, 0.0032, 0.0047, 0.0163, 0.0251, 0.0344, -0.0268, 0.0082, -0.0343, -0.0187], device='cuda:0'), grad: tensor([ 2.9460e-05, 5.8413e-05, -2.1577e-04, 4.3392e-04, -5.6458e-04, -1.0576e-03, -4.9084e-05, 1.3280e-04, 1.9884e-04, 1.0309e-03], device='cuda:0') 100 0.0001 changing lr epoch 75, time 217.29, cls_loss 0.0096 cls_loss_mapping 0.0140 cls_loss_causal 0.6486 re_mapping 0.0126 re_causal 0.0357 /// teacc 98.76 lr 0.00010000 Epoch 77, weight, value: tensor([[-0.0085, 0.0886, 0.0543, ..., -0.1046, -0.0745, 0.0267], [ 0.0835, -0.0120, -0.0321, ..., 0.0459, 0.0660, -0.0846], [-0.0447, -0.0196, -0.0316, ..., -0.0501, -0.0103, -0.0234], ..., [-0.0245, 0.0977, -0.1156, ..., -0.0746, 0.0149, -0.0817], [ 0.0003, -0.0572, 0.0276, ..., 0.0416, -0.0849, -0.0337], [-0.0173, -0.0460, 0.0322, ..., -0.0492, -0.0357, 0.0785]], device='cuda:0'), grad: tensor([[ 1.4165e-06, -2.6301e-05, -2.2992e-05, ..., 2.3115e-06, 2.3656e-07, -4.3571e-05], [-1.0222e-05, -8.0606e-07, 2.8722e-06, ..., -1.6570e-05, -8.2701e-06, 1.1288e-06], [ 2.6617e-06, 4.9695e-06, 9.7379e-06, ..., -2.0564e-05, -5.7742e-06, 3.6545e-06], ..., [ 9.0823e-06, -1.4079e-04, 1.2197e-05, ..., 1.7300e-05, 4.8019e-06, 1.1131e-05], [-2.6917e-04, 5.0142e-06, -5.6297e-05, ..., -3.5739e-04, 6.1356e-06, -9.9003e-05], [ 5.2601e-06, 2.6122e-05, 9.3877e-06, ..., 1.0945e-05, -4.8131e-06, 2.5909e-06]], device='cuda:0') Epoch 77, bias, value: tensor([ 0.0058, 0.0027, 0.0044, 0.0160, 0.0260, 0.0346, -0.0267, 0.0089, -0.0340, -0.0197], device='cuda:0'), grad: tensor([-4.1217e-05, -5.1633e-06, -1.2226e-05, 6.2764e-05, 1.9014e-05, 6.3848e-04, 1.8641e-05, -1.8847e-04, -5.4932e-04, 5.7161e-05], device='cuda:0') 100 0.0001 changing lr epoch 76, time 216.57, cls_loss 0.0104 cls_loss_mapping 0.0168 cls_loss_causal 0.6174 re_mapping 0.0129 re_causal 0.0360 /// teacc 98.65 lr 0.00010000 Epoch 78, weight, value: tensor([[-0.0091, 0.0889, 0.0544, ..., -0.1057, -0.0762, 0.0265], [ 0.0843, -0.0121, -0.0326, ..., 0.0463, 0.0677, -0.0853], [-0.0460, -0.0202, -0.0318, ..., -0.0513, -0.0111, -0.0237], ..., [-0.0254, 0.0990, -0.1157, ..., -0.0757, 0.0142, -0.0820], [ 0.0009, -0.0581, 0.0280, ..., 0.0424, -0.0852, -0.0338], [-0.0174, -0.0472, 0.0330, ..., -0.0497, -0.0355, 0.0801]], device='cuda:0'), grad: tensor([[ 4.5188e-06, -6.0024e-07, 3.0637e-05, ..., 1.2532e-05, 9.7416e-07, 2.9087e-05], [-1.1463e-03, -2.5344e-04, -3.0184e-04, ..., -9.2793e-04, -2.8443e-04, -1.7715e-04], [ 8.7246e-06, -4.3362e-06, 3.0976e-06, ..., -1.8300e-06, 3.1292e-07, -3.8683e-05], ..., [ 3.1614e-04, 6.8665e-05, 9.2983e-05, ..., 2.7251e-04, 8.6129e-05, 6.3777e-05], [ 1.1998e-04, 2.4259e-05, 2.3103e-04, ..., 1.4997e-04, 3.0339e-05, 2.0719e-04], [ 5.8842e-04, 1.4615e-04, -8.7214e-04, ..., 1.7083e-04, 1.4830e-04, -8.8787e-04]], device='cuda:0') Epoch 78, bias, value: tensor([ 0.0055, 0.0032, 0.0041, 0.0153, 0.0251, 0.0347, -0.0262, 0.0094, -0.0340, -0.0194], device='cuda:0'), grad: tensor([ 3.9250e-05, -2.1858e-03, -1.3316e-04, 5.0068e-04, 7.9632e-05, 4.1676e-04, -4.9453e-07, 6.8426e-04, 4.3297e-04, 1.6403e-04], device='cuda:0') 100 0.0001 changing lr epoch 77, time 217.60, cls_loss 0.0095 cls_loss_mapping 0.0137 cls_loss_causal 0.6158 re_mapping 0.0123 re_causal 0.0350 /// teacc 98.73 lr 0.00010000 Epoch 79, weight, value: tensor([[-0.0097, 0.0887, 0.0542, ..., -0.1064, -0.0778, 0.0266], [ 0.0837, -0.0125, -0.0338, ..., 0.0456, 0.0672, -0.0860], [-0.0449, -0.0196, -0.0311, ..., -0.0502, -0.0099, -0.0239], ..., [-0.0252, 0.1001, -0.1160, ..., -0.0754, 0.0147, -0.0828], [ 0.0015, -0.0592, 0.0284, ..., 0.0432, -0.0850, -0.0337], [-0.0179, -0.0476, 0.0336, ..., -0.0500, -0.0362, 0.0811]], device='cuda:0'), grad: tensor([[ 2.1711e-05, -7.3425e-06, 4.4316e-05, ..., 5.6177e-05, 2.5565e-07, 5.5999e-05], [-2.2516e-05, 7.8836e-07, 3.2708e-06, ..., -3.6448e-05, -1.9744e-05, 3.2894e-06], [ 1.1593e-05, 2.5511e-05, 2.0906e-05, ..., 2.1711e-05, 2.9318e-06, 2.1636e-05], ..., [ 1.3456e-05, -8.2135e-05, 2.0832e-05, ..., 2.7165e-05, 5.4426e-06, 2.3559e-05], [-6.6578e-05, -2.8297e-05, -1.8561e-04, ..., -1.5283e-04, 4.7982e-06, -2.0170e-04], [ 2.0444e-05, 3.1531e-05, 2.4423e-05, ..., 4.8578e-05, 1.3113e-06, 3.3349e-05]], device='cuda:0') Epoch 79, bias, value: tensor([ 0.0051, 0.0019, 0.0058, 0.0154, 0.0246, 0.0347, -0.0265, 0.0097, -0.0339, -0.0192], device='cuda:0'), grad: tensor([ 1.3912e-04, -3.9041e-05, 9.0420e-05, 1.6725e-04, 3.3438e-05, -4.0792e-06, 3.8654e-05, -3.7044e-05, -5.2691e-04, 1.3828e-04], device='cuda:0') 100 0.0001 changing lr epoch 78, time 217.57, cls_loss 0.0110 cls_loss_mapping 0.0143 cls_loss_causal 0.6278 re_mapping 0.0127 re_causal 0.0334 /// teacc 98.75 lr 0.00010000 Epoch 80, weight, value: tensor([[-0.0100, 0.0893, 0.0541, ..., -0.1078, -0.0785, 0.0264], [ 0.0843, -0.0124, -0.0352, ..., 0.0456, 0.0680, -0.0881], [-0.0456, -0.0198, -0.0315, ..., -0.0512, -0.0107, -0.0245], ..., [-0.0257, 0.1007, -0.1167, ..., -0.0767, 0.0141, -0.0837], [ 0.0009, -0.0604, 0.0286, ..., 0.0429, -0.0855, -0.0340], [-0.0181, -0.0478, 0.0341, ..., -0.0499, -0.0342, 0.0821]], device='cuda:0'), grad: tensor([[ 3.7951e-07, -1.3128e-05, 4.3437e-06, ..., 2.1562e-05, 5.7340e-05, -4.5672e-06], [-2.3112e-05, 3.4012e-06, 1.8254e-06, ..., -3.5912e-05, -1.5378e-05, 1.0030e-06], [ 6.8359e-06, 1.2085e-05, 3.5465e-05, ..., 3.3677e-05, 2.5198e-05, 2.0247e-06], ..., [ 6.5081e-06, -7.9632e-05, 3.0071e-05, ..., 3.2842e-05, -2.0891e-05, -1.4484e-05], [-1.2845e-05, 1.1995e-05, 1.7673e-05, ..., -1.7375e-05, 2.6673e-05, -1.0766e-05], [ 4.9286e-06, 4.4644e-05, 1.1519e-05, ..., 3.2604e-05, 7.8857e-05, 6.8061e-06]], device='cuda:0') Epoch 80, bias, value: tensor([ 0.0048, 0.0018, 0.0056, 0.0154, 0.0250, 0.0343, -0.0243, 0.0093, -0.0347, -0.0191], device='cuda:0'), grad: tensor([ 3.8791e-04, 3.2727e-06, 1.8442e-04, -2.7347e-04, -1.5287e-03, 1.7726e-04, 4.8971e-04, -3.8713e-05, 1.0628e-04, 4.9210e-04], device='cuda:0') 100 0.0001 changing lr epoch 79, time 217.34, cls_loss 0.0108 cls_loss_mapping 0.0158 cls_loss_causal 0.6251 re_mapping 0.0123 re_causal 0.0342 /// teacc 98.72 lr 0.00010000 Epoch 81, weight, value: tensor([[-0.0095, 0.0906, 0.0545, ..., -0.1085, -0.0787, 0.0266], [ 0.0850, -0.0126, -0.0358, ..., 0.0467, 0.0695, -0.0885], [-0.0464, -0.0183, -0.0318, ..., -0.0525, -0.0116, -0.0248], ..., [-0.0261, 0.1003, -0.1172, ..., -0.0774, 0.0137, -0.0843], [-0.0003, -0.0622, 0.0275, ..., 0.0421, -0.0870, -0.0353], [-0.0181, -0.0482, 0.0350, ..., -0.0495, -0.0343, 0.0829]], device='cuda:0'), grad: tensor([[ 7.7903e-05, -1.5050e-06, 5.2266e-06, ..., 9.4891e-05, 2.4796e-07, 3.6675e-06], [ 3.2759e-04, -1.5914e-05, 7.4133e-07, ..., 3.9077e-04, -1.3627e-05, 4.1490e-07], [ 6.3479e-05, 4.7833e-06, 1.7658e-06, ..., 7.8201e-05, 1.2694e-06, 9.2201e-07], ..., [ 1.9237e-05, -1.7869e-04, 9.7871e-05, ..., 2.8521e-05, 9.5218e-06, 8.8394e-05], [ 6.4611e-04, 2.9560e-06, -8.2478e-06, ..., 7.6485e-04, 6.5845e-07, 1.0379e-05], [ 8.9407e-06, -1.0836e-04, -2.8348e-04, ..., 1.5482e-05, 8.3819e-07, -2.6131e-04]], device='cuda:0') Epoch 81, bias, value: tensor([ 0.0052, 0.0024, 0.0058, 0.0151, 0.0246, 0.0343, -0.0240, 0.0087, -0.0357, -0.0185], device='cuda:0'), grad: tensor([ 0.0002, 0.0008, 0.0002, 0.0003, 0.0010, 0.0004, -0.0032, -0.0008, 0.0015, -0.0004], device='cuda:0') 100 0.0001 changing lr epoch 80, time 217.63, cls_loss 0.0111 cls_loss_mapping 0.0158 cls_loss_causal 0.6414 re_mapping 0.0120 re_causal 0.0342 /// teacc 98.79 lr 0.00010000 Epoch 82, weight, value: tensor([[-9.8741e-03, 9.1382e-02, 5.4812e-02, ..., -1.0929e-01, -7.9147e-02, 2.6775e-02], [ 8.5780e-02, -1.1860e-02, -3.6087e-02, ..., 4.7225e-02, 7.1365e-02, -8.8860e-02], [-4.7703e-02, -1.8855e-02, -3.1960e-02, ..., -5.2953e-02, -1.3034e-02, -2.5070e-02], ..., [-2.5541e-02, 1.0060e-01, -1.1791e-01, ..., -7.7870e-02, 1.3934e-02, -8.5318e-02], [ 1.1617e-04, -6.1831e-02, 2.7713e-02, ..., 4.2588e-02, -8.7229e-02, -3.5074e-02], [-1.8341e-02, -4.9051e-02, 3.5217e-02, ..., -5.0428e-02, -3.4711e-02, 8.3518e-02]], device='cuda:0'), grad: tensor([[ 7.0445e-06, -1.6227e-05, 2.1551e-06, ..., 4.8652e-06, 1.6242e-06, -8.4564e-07], [-1.5959e-05, 3.1814e-06, 8.4378e-07, ..., -1.7837e-05, -1.3530e-05, 1.3821e-06], [ 4.3474e-06, -9.6202e-05, 1.0515e-06, ..., 4.3549e-06, 1.3281e-06, 1.2033e-06], ..., [ 5.9269e-06, -1.6248e-04, 5.2676e-06, ..., 7.1302e-06, 4.7274e-06, -6.4313e-05], [ 5.5104e-05, 8.3074e-06, 2.0027e-05, ..., 2.2218e-05, 8.6501e-06, 2.0131e-05], [ 4.5560e-06, 2.3437e-04, -1.3232e-05, ..., 4.9323e-06, 5.9977e-07, 5.3346e-05]], device='cuda:0') Epoch 82, bias, value: tensor([ 0.0055, 0.0031, 0.0052, 0.0148, 0.0253, 0.0340, -0.0248, 0.0091, -0.0353, -0.0190], device='cuda:0'), grad: tensor([-1.2301e-05, -6.9141e-06, -5.9891e-04, 2.2605e-05, 4.0740e-05, 1.3638e-04, -2.0611e-04, 3.5286e-05, 1.0598e-04, 4.8208e-04], device='cuda:0') 100 0.0001 changing lr epoch 81, time 217.35, cls_loss 0.0097 cls_loss_mapping 0.0159 cls_loss_causal 0.6430 re_mapping 0.0126 re_causal 0.0342 /// teacc 98.81 lr 0.00010000 Epoch 83, weight, value: tensor([[-0.0101, 0.0917, 0.0546, ..., -0.1099, -0.0802, 0.0266], [ 0.0855, -0.0111, -0.0363, ..., 0.0463, 0.0706, -0.0892], [-0.0472, -0.0190, -0.0327, ..., -0.0518, -0.0115, -0.0254], ..., [-0.0259, 0.1012, -0.1194, ..., -0.0784, 0.0135, -0.0856], [ 0.0004, -0.0628, 0.0282, ..., 0.0431, -0.0878, -0.0352], [-0.0189, -0.0497, 0.0354, ..., -0.0509, -0.0355, 0.0838]], device='cuda:0'), grad: tensor([[ 2.7612e-05, -1.7613e-05, 1.4925e-04, ..., 1.8120e-05, 1.8060e-05, -1.2212e-05], [ 8.4221e-05, 3.6025e-04, 8.7768e-06, ..., 7.6517e-06, 1.8254e-05, 9.0711e-07], [-1.7762e-05, 1.0520e-05, 3.5334e-04, ..., 3.9935e-06, -8.5533e-06, 7.3947e-07], ..., [-8.3625e-05, -5.2166e-04, 2.1219e-05, ..., 1.2055e-05, -2.3082e-05, 1.4100e-06], [ 4.1038e-05, 1.2636e-05, 1.4596e-05, ..., 6.0499e-05, 9.8497e-06, 6.9924e-06], [ 1.5661e-05, 4.2439e-05, 3.4034e-05, ..., -8.1444e-04, 8.0988e-06, -2.0713e-06]], device='cuda:0') Epoch 83, bias, value: tensor([ 0.0051, 0.0025, 0.0063, 0.0144, 0.0256, 0.0349, -0.0250, 0.0087, -0.0351, -0.0193], device='cuda:0'), grad: tensor([ 0.0006, 0.0005, 0.0011, -0.0021, 0.0037, -0.0044, 0.0042, -0.0006, 0.0002, -0.0032], device='cuda:0') 100 0.0001 changing lr epoch 82, time 217.01, cls_loss 0.0084 cls_loss_mapping 0.0114 cls_loss_causal 0.6572 re_mapping 0.0119 re_causal 0.0344 /// teacc 98.65 lr 0.00010000 Epoch 84, weight, value: tensor([[-1.0296e-02, 9.2249e-02, 5.4699e-02, ..., -1.1042e-01, -8.0827e-02, 2.6515e-02], [ 8.6181e-02, -9.8126e-03, -3.6275e-02, ..., 4.6982e-02, 7.1194e-02, -8.9245e-02], [-4.7421e-02, -1.9725e-02, -3.3079e-02, ..., -5.2103e-02, -1.1407e-02, -2.5273e-02], ..., [-2.6926e-02, 1.0159e-01, -1.1968e-01, ..., -8.0110e-02, 1.3139e-02, -8.6026e-02], [ 4.3944e-05, -6.3522e-02, 2.7892e-02, ..., 4.3039e-02, -8.8442e-02, -3.5265e-02], [-1.9875e-02, -5.0545e-02, 3.5718e-02, ..., -5.0521e-02, -3.5832e-02, 8.4657e-02]], device='cuda:0'), grad: tensor([[ 7.4096e-06, -1.6883e-05, -8.7172e-07, ..., 7.0035e-06, 1.6969e-06, -3.0873e-07], [-8.4734e-04, 5.9977e-07, 1.0896e-06, ..., -9.0361e-04, -1.7090e-03, 1.4454e-06], [ 6.4802e-04, 3.6452e-06, -3.6918e-06, ..., 6.8808e-04, 1.2655e-03, 4.6454e-06], ..., [ 2.0027e-04, 6.1933e-07, 6.7875e-06, ..., 3.0947e-04, 3.9220e-04, 9.4995e-06], [ 4.3929e-05, -7.4273e-08, 7.9930e-05, ..., 3.4839e-05, -4.0457e-06, 7.1824e-05], [ 3.7074e-05, 7.0333e-06, 3.8296e-05, ..., 8.1182e-05, 1.4175e-06, 5.4538e-05]], device='cuda:0') Epoch 84, bias, value: tensor([ 0.0048, 0.0030, 0.0062, 0.0145, 0.0245, 0.0348, -0.0246, 0.0087, -0.0355, -0.0187], device='cuda:0'), grad: tensor([-2.1338e-05, -1.9894e-03, 1.4706e-03, 5.0688e-04, -3.0947e-04, -7.4053e-04, -2.6766e-06, 6.7377e-04, 1.8239e-04, 2.2733e-04], device='cuda:0') 100 0.0001 changing lr epoch 83, time 217.40, cls_loss 0.0072 cls_loss_mapping 0.0099 cls_loss_causal 0.6295 re_mapping 0.0117 re_causal 0.0352 /// teacc 98.79 lr 0.00010000 Epoch 85, weight, value: tensor([[-0.0100, 0.0928, 0.0543, ..., -0.1108, -0.0829, 0.0260], [ 0.0865, -0.0101, -0.0364, ..., 0.0473, 0.0713, -0.0896], [-0.0479, -0.0206, -0.0334, ..., -0.0526, -0.0116, -0.0256], ..., [-0.0274, 0.1022, -0.1208, ..., -0.0806, 0.0136, -0.0862], [ 0.0002, -0.0638, 0.0279, ..., 0.0433, -0.0886, -0.0355], [-0.0203, -0.0505, 0.0364, ..., -0.0508, -0.0353, 0.0853]], device='cuda:0'), grad: tensor([[ 2.6450e-06, -5.2713e-07, 8.3148e-06, ..., 4.6529e-06, 1.4110e-06, 7.6815e-06], [-1.2726e-05, 5.9232e-07, 2.9560e-06, ..., -2.3365e-05, -8.6874e-06, 1.6410e-06], [ 3.3751e-06, 3.2410e-06, 1.9208e-05, ..., 3.7197e-06, 1.6699e-06, 2.1607e-06], ..., [ 1.4424e-05, -1.6782e-06, 4.7445e-05, ..., 1.5102e-05, 6.6720e-06, 2.5392e-05], [ 1.8314e-05, 5.3421e-06, 4.2319e-05, ..., 1.7241e-05, 1.3314e-05, 2.1994e-05], [ 9.2566e-05, 5.5619e-06, 1.4699e-04, ..., 9.3162e-05, 8.1658e-06, 2.7394e-04]], device='cuda:0') Epoch 85, bias, value: tensor([ 0.0046, 0.0027, 0.0058, 0.0147, 0.0243, 0.0349, -0.0248, 0.0089, -0.0356, -0.0182], device='cuda:0'), grad: tensor([ 1.6943e-05, -1.6868e-05, 3.8475e-05, 1.8730e-03, 2.8849e-05, -2.5635e-03, 2.8133e-05, 9.6917e-05, 1.0794e-04, 3.8934e-04], device='cuda:0') 100 0.0001 changing lr epoch 84, time 217.03, cls_loss 0.0080 cls_loss_mapping 0.0105 cls_loss_causal 0.6125 re_mapping 0.0114 re_causal 0.0329 /// teacc 98.83 lr 0.00010000 Epoch 86, weight, value: tensor([[-1.0266e-02, 9.3217e-02, 5.4418e-02, ..., -1.1161e-01, -8.3075e-02, 2.5764e-02], [ 8.7239e-02, -1.0392e-02, -3.5655e-02, ..., 4.8504e-02, 7.1378e-02, -8.9176e-02], [-4.8153e-02, -2.0935e-02, -3.3708e-02, ..., -5.2898e-02, -1.1542e-02, -2.5909e-02], ..., [-2.7471e-02, 1.0312e-01, -1.2112e-01, ..., -8.0917e-02, 1.3959e-02, -8.7156e-02], [-9.8744e-05, -6.4129e-02, 2.8179e-02, ..., 4.3266e-02, -8.8892e-02, -3.5836e-02], [-2.1463e-02, -5.1019e-02, 3.6275e-02, ..., -5.2047e-02, -3.5488e-02, 8.5811e-02]], device='cuda:0'), grad: tensor([[-5.1260e-06, -4.2707e-05, -3.2216e-05, ..., -2.6748e-06, 1.8813e-07, -4.8190e-05], [-8.3353e-07, -1.3439e-06, 2.7250e-06, ..., -1.9316e-06, -1.3784e-06, 3.3118e-06], [ 8.3521e-06, 5.9120e-06, 1.6075e-06, ..., 5.6922e-06, 5.8673e-08, 1.8571e-06], ..., [ 4.1336e-05, 1.1042e-05, 3.9279e-05, ..., 3.2961e-05, 1.0990e-05, 7.0333e-05], [-9.5814e-06, 2.2408e-06, 2.0802e-04, ..., 1.5765e-05, 7.2643e-07, 2.1946e-04], [-1.4946e-05, -1.8612e-05, -3.5238e-04, ..., -7.5161e-05, -2.3410e-05, -4.3917e-04]], device='cuda:0') Epoch 86, bias, value: tensor([ 0.0042, 0.0029, 0.0060, 0.0136, 0.0243, 0.0358, -0.0242, 0.0093, -0.0358, -0.0187], device='cuda:0'), grad: tensor([-6.3181e-05, 2.7180e-05, -1.3137e-04, -1.9848e-04, 3.6025e-04, 2.1350e-04, 9.5546e-05, 2.9778e-04, 5.6601e-04, -1.1673e-03], device='cuda:0') 100 0.0001 changing lr epoch 85, time 217.61, cls_loss 0.0078 cls_loss_mapping 0.0114 cls_loss_causal 0.5988 re_mapping 0.0116 re_causal 0.0325 /// teacc 98.81 lr 0.00010000 Epoch 87, weight, value: tensor([[-0.0104, 0.0943, 0.0554, ..., -0.1120, -0.0833, 0.0262], [ 0.0875, -0.0101, -0.0365, ..., 0.0483, 0.0715, -0.0904], [-0.0489, -0.0214, -0.0342, ..., -0.0538, -0.0115, -0.0264], ..., [-0.0281, 0.1035, -0.1215, ..., -0.0815, 0.0137, -0.0877], [ 0.0009, -0.0635, 0.0284, ..., 0.0441, -0.0892, -0.0356], [-0.0212, -0.0516, 0.0369, ..., -0.0518, -0.0350, 0.0869]], device='cuda:0'), grad: tensor([[ 4.1090e-06, 2.1476e-06, -2.8033e-07, ..., 5.6587e-06, 1.9046e-07, -4.0093e-07], [-7.1153e-06, 2.0694e-06, 5.3691e-07, ..., -5.7854e-06, -2.3060e-06, 2.2352e-07], [ 7.0855e-06, 1.6659e-05, 6.5751e-07, ..., 8.5458e-06, -1.8692e-06, 2.6450e-07], ..., [ 3.4980e-06, -3.9101e-05, 1.3951e-06, ..., 8.7991e-06, 2.9840e-06, 1.6354e-06], [ 2.6047e-05, 2.4103e-06, 4.5866e-05, ..., 4.0442e-05, 2.5835e-06, 1.0294e-04], [ 1.0878e-05, 6.6720e-06, 1.2167e-05, ..., 2.8193e-05, 8.0932e-07, 1.6972e-05]], device='cuda:0') Epoch 87, bias, value: tensor([ 0.0050, 0.0028, 0.0055, 0.0143, 0.0242, 0.0353, -0.0255, 0.0091, -0.0351, -0.0183], device='cuda:0'), grad: tensor([ 1.3091e-05, 3.9637e-06, 5.7556e-07, -3.2425e-05, -3.4750e-05, -1.1396e-04, -4.4238e-07, -5.7444e-06, 1.0848e-04, 6.1452e-05], device='cuda:0') 100 0.0001 changing lr epoch 86, time 217.72, cls_loss 0.0074 cls_loss_mapping 0.0093 cls_loss_causal 0.6044 re_mapping 0.0114 re_causal 0.0332 /// teacc 98.72 lr 0.00010000 Epoch 88, weight, value: tensor([[-0.0104, 0.0951, 0.0553, ..., -0.1133, -0.0849, 0.0259], [ 0.0879, -0.0102, -0.0366, ..., 0.0487, 0.0716, -0.0906], [-0.0487, -0.0213, -0.0330, ..., -0.0535, -0.0108, -0.0265], ..., [-0.0286, 0.1037, -0.1220, ..., -0.0825, 0.0136, -0.0890], [ 0.0015, -0.0629, 0.0287, ..., 0.0449, -0.0898, -0.0348], [-0.0215, -0.0521, 0.0373, ..., -0.0521, -0.0353, 0.0874]], device='cuda:0'), grad: tensor([[ 5.7608e-05, 1.6168e-05, 4.6790e-06, ..., 4.8310e-05, 1.1940e-06, 8.3307e-07], [ 3.2842e-05, 7.1041e-06, 4.2655e-06, ..., 4.2140e-05, -5.9791e-07, 2.4810e-06], [ 5.3197e-05, 1.7490e-06, 3.1441e-05, ..., 3.6120e-05, 1.8492e-05, 4.9593e-07], ..., [ 5.5991e-06, -1.4864e-05, 5.4985e-06, ..., 3.8683e-05, 4.7646e-06, 3.2224e-06], [ 1.0532e-04, 4.0799e-05, -5.6863e-05, ..., 4.8697e-05, 5.1707e-06, -7.4744e-05], [ 4.8816e-05, 8.6576e-06, 6.9320e-05, ..., 1.7290e-03, 1.4789e-06, 1.9038e-04]], device='cuda:0') Epoch 88, bias, value: tensor([ 0.0049, 0.0027, 0.0068, 0.0136, 0.0243, 0.0350, -0.0255, 0.0088, -0.0345, -0.0185], device='cuda:0'), grad: tensor([ 1.3769e-04, 1.2314e-04, 1.6916e-04, -1.6093e-04, -3.6316e-03, 8.6308e-05, -7.8535e-04, 8.0168e-05, 2.3484e-04, 3.7441e-03], device='cuda:0') 100 0.0001 changing lr epoch 87, time 217.26, cls_loss 0.0073 cls_loss_mapping 0.0096 cls_loss_causal 0.5807 re_mapping 0.0113 re_causal 0.0307 /// teacc 98.83 lr 0.00010000 Epoch 89, weight, value: tensor([[-0.0107, 0.0958, 0.0555, ..., -0.1142, -0.0850, 0.0260], [ 0.0885, -0.0102, -0.0369, ..., 0.0491, 0.0727, -0.0909], [-0.0489, -0.0215, -0.0330, ..., -0.0532, -0.0113, -0.0271], ..., [-0.0289, 0.1038, -0.1221, ..., -0.0834, 0.0133, -0.0888], [ 0.0013, -0.0632, 0.0285, ..., 0.0449, -0.0903, -0.0354], [-0.0222, -0.0534, 0.0376, ..., -0.0527, -0.0355, 0.0878]], device='cuda:0'), grad: tensor([[ 6.2920e-06, -9.9465e-07, 1.7565e-06, ..., 7.7859e-06, 2.4168e-07, 1.8403e-06], [-2.2203e-05, 3.1758e-07, 7.6974e-07, ..., -1.0230e-05, -2.3291e-05, 6.6822e-07], [ 4.9919e-05, 4.1910e-07, 3.3751e-06, ..., 7.7307e-05, 1.6123e-05, 9.1409e-07], ..., [ 5.2266e-06, -4.1351e-07, 6.7987e-07, ..., 8.1509e-06, 3.8259e-06, 5.9186e-07], [-1.6168e-05, 2.9430e-06, -3.4958e-05, ..., -2.3022e-05, 1.2945e-06, -1.1824e-05], [ 8.8364e-06, 5.4343e-07, 7.9125e-06, ..., 1.0565e-05, 8.5728e-07, 5.3942e-06]], device='cuda:0') Epoch 89, bias, value: tensor([ 0.0050, 0.0033, 0.0068, 0.0130, 0.0252, 0.0356, -0.0255, 0.0085, -0.0351, -0.0191], device='cuda:0'), grad: tensor([ 1.4596e-05, -4.2409e-05, 1.5056e-04, 3.0577e-05, 3.3951e-04, 3.0667e-05, -5.1451e-04, 1.7881e-05, -5.1022e-05, 2.4498e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 88---------------------------------------------------- epoch 88, time 218.18, cls_loss 0.0058 cls_loss_mapping 0.0090 cls_loss_causal 0.5834 re_mapping 0.0115 re_causal 0.0322 /// teacc 98.89 lr 0.00010000 Epoch 90, weight, value: tensor([[-0.0120, 0.0956, 0.0559, ..., -0.1162, -0.0873, 0.0258], [ 0.0885, -0.0103, -0.0370, ..., 0.0493, 0.0731, -0.0911], [-0.0492, -0.0221, -0.0332, ..., -0.0536, -0.0115, -0.0274], ..., [-0.0286, 0.1045, -0.1226, ..., -0.0833, 0.0133, -0.0895], [ 0.0013, -0.0635, 0.0284, ..., 0.0451, -0.0906, -0.0356], [-0.0226, -0.0538, 0.0377, ..., -0.0529, -0.0357, 0.0882]], device='cuda:0'), grad: tensor([[ 1.0297e-05, -3.5204e-06, -3.4273e-07, ..., 8.7991e-06, 2.1607e-07, 1.8533e-06], [-2.9653e-05, 1.3812e-06, 1.9539e-06, ..., -4.7356e-05, -2.4945e-05, 2.1569e-06], [ 1.1601e-05, 2.6733e-05, 3.3546e-06, ..., 1.2681e-05, 4.2021e-06, 2.8089e-06], ..., [ 4.5419e-05, -3.4332e-05, 2.4214e-05, ..., 4.3333e-05, 9.6709e-06, 2.3663e-05], [ 4.9204e-05, 7.0361e-07, 2.2814e-05, ..., 4.1485e-05, 3.5148e-06, 2.6822e-05], [ 1.8075e-05, 4.2617e-06, 1.3318e-07, ..., 1.5751e-05, 1.7527e-06, 2.2855e-06]], device='cuda:0') Epoch 90, bias, value: tensor([ 0.0045, 0.0029, 0.0065, 0.0131, 0.0251, 0.0358, -0.0249, 0.0090, -0.0353, -0.0192], device='cuda:0'), grad: tensor([ 2.0474e-05, -6.5982e-05, 8.4996e-05, 9.5889e-06, 2.2545e-05, -3.9673e-04, 5.4568e-05, 5.9754e-05, 1.6046e-04, 5.0098e-05], device='cuda:0') 100 0.0001 changing lr epoch 89, time 217.02, cls_loss 0.0056 cls_loss_mapping 0.0090 cls_loss_causal 0.5937 re_mapping 0.0112 re_causal 0.0314 /// teacc 98.82 lr 0.00010000 Epoch 91, weight, value: tensor([[-0.0122, 0.0965, 0.0567, ..., -0.1159, -0.0872, 0.0268], [ 0.0891, -0.0095, -0.0375, ..., 0.0500, 0.0736, -0.0920], [-0.0493, -0.0224, -0.0335, ..., -0.0538, -0.0117, -0.0277], ..., [-0.0294, 0.1044, -0.1235, ..., -0.0847, 0.0127, -0.0899], [ 0.0013, -0.0637, 0.0283, ..., 0.0451, -0.0907, -0.0360], [-0.0227, -0.0543, 0.0383, ..., -0.0525, -0.0354, 0.0892]], device='cuda:0'), grad: tensor([[ 2.7996e-06, -3.0417e-06, 3.4366e-07, ..., 5.7667e-06, 8.5868e-07, 2.9569e-07], [-6.1572e-05, 9.1866e-06, 4.3847e-06, ..., -1.1069e-04, -3.7163e-05, 3.0026e-06], [ 6.0678e-05, -1.4710e-04, 1.2502e-05, ..., 1.0681e-04, 3.1948e-05, 1.9707e-06], ..., [ 8.0317e-06, 1.2553e-04, 6.9477e-06, ..., 1.7300e-05, 1.6987e-06, 4.4331e-06], [ 4.4793e-05, 4.8503e-06, 5.3436e-05, ..., 6.4850e-05, 2.5518e-06, 3.9250e-05], [ 2.7061e-05, 2.5667e-06, 2.6584e-05, ..., 4.7445e-05, 3.7253e-07, 1.6823e-05]], device='cuda:0') Epoch 91, bias, value: tensor([ 0.0053, 0.0035, 0.0065, 0.0133, 0.0247, 0.0356, -0.0251, 0.0083, -0.0355, -0.0188], device='cuda:0'), grad: tensor([ 8.7842e-06, -2.0802e-04, -6.1464e-04, 7.8499e-05, -8.2910e-05, -2.5034e-04, -1.9506e-05, 8.4162e-04, 1.6427e-04, 8.2910e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 90---------------------------------------------------- epoch 90, time 217.68, cls_loss 0.0084 cls_loss_mapping 0.0124 cls_loss_causal 0.6120 re_mapping 0.0107 re_causal 0.0299 /// teacc 98.90 lr 0.00010000 Epoch 92, weight, value: tensor([[-0.0124, 0.0976, 0.0570, ..., -0.1167, -0.0875, 0.0270], [ 0.0891, -0.0101, -0.0378, ..., 0.0497, 0.0736, -0.0921], [-0.0489, -0.0230, -0.0335, ..., -0.0530, -0.0117, -0.0281], ..., [-0.0301, 0.1058, -0.1246, ..., -0.0851, 0.0143, -0.0903], [ 0.0018, -0.0643, 0.0284, ..., 0.0454, -0.0914, -0.0362], [-0.0239, -0.0549, 0.0389, ..., -0.0524, -0.0363, 0.0896]], device='cuda:0'), grad: tensor([[ 3.6173e-06, -1.8895e-05, -7.5903e-07, ..., 1.5525e-06, 1.7673e-05, -2.3752e-05], [-1.3880e-05, 7.1712e-06, 5.0068e-06, ..., -2.3186e-05, -1.4953e-05, 2.7381e-06], [ 9.7454e-06, 7.4655e-06, -7.2360e-05, ..., 1.6600e-05, -9.6321e-05, 5.7481e-06], ..., [ 3.0156e-06, -2.2423e-04, -7.9155e-05, ..., 4.8801e-06, 2.6152e-05, -5.2780e-05], [-5.4449e-05, 2.8834e-05, 1.4842e-05, ..., -6.3837e-05, 3.4515e-06, 9.4250e-06], [ 8.1733e-06, 1.7118e-04, 9.1016e-05, ..., 1.4223e-05, 1.7688e-05, 4.6641e-05]], device='cuda:0') Epoch 92, bias, value: tensor([ 0.0056, 0.0026, 0.0071, 0.0130, 0.0245, 0.0357, -0.0257, 0.0086, -0.0349, -0.0186], device='cuda:0'), grad: tensor([-7.8455e-06, -5.0850e-06, -1.9073e-04, 6.1929e-05, 1.3039e-05, 5.5194e-05, 6.0260e-05, -3.5691e-04, -2.5854e-06, 3.7289e-04], device='cuda:0') 100 0.0001 changing lr epoch 91, time 216.99, cls_loss 0.0086 cls_loss_mapping 0.0121 cls_loss_causal 0.6155 re_mapping 0.0109 re_causal 0.0293 /// teacc 98.84 lr 0.00010000 Epoch 93, weight, value: tensor([[-0.0123, 0.0979, 0.0571, ..., -0.1177, -0.0874, 0.0263], [ 0.0894, -0.0103, -0.0381, ..., 0.0502, 0.0748, -0.0927], [-0.0494, -0.0240, -0.0337, ..., -0.0537, -0.0121, -0.0286], ..., [-0.0298, 0.1067, -0.1251, ..., -0.0853, 0.0136, -0.0895], [ 0.0020, -0.0654, 0.0284, ..., 0.0459, -0.0917, -0.0364], [-0.0247, -0.0545, 0.0396, ..., -0.0531, -0.0367, 0.0910]], device='cuda:0'), grad: tensor([[ 2.0131e-05, 1.0468e-05, 4.7013e-06, ..., 1.4156e-05, 1.9036e-06, 5.6922e-06], [ 8.1211e-06, 2.7999e-05, 3.4142e-06, ..., -2.8424e-06, -1.8580e-06, 3.8929e-07], [ 1.5661e-05, 4.4778e-06, 1.2569e-05, ..., 1.4246e-05, -2.9970e-06, 7.5670e-07], ..., [-9.2760e-06, -5.8353e-05, 5.2750e-06, ..., 5.9605e-06, 1.2824e-06, 7.9907e-07], [-9.5367e-07, 2.0768e-06, 6.1356e-06, ..., -2.8446e-05, 1.7285e-06, 6.1840e-07], [ 1.2249e-05, 2.4885e-05, 6.1169e-06, ..., 6.2510e-06, 9.2480e-07, 1.9968e-06]], device='cuda:0') Epoch 93, bias, value: tensor([ 0.0046, 0.0026, 0.0060, 0.0128, 0.0243, 0.0353, -0.0256, 0.0096, -0.0345, -0.0180], device='cuda:0'), grad: tensor([ 4.9710e-05, 6.1989e-05, 2.8670e-05, -1.9989e-03, 7.0743e-06, 1.9064e-03, -3.5554e-05, -8.0168e-05, -8.9630e-06, 6.9201e-05], device='cuda:0') 100 0.0001 changing lr epoch 92, time 217.13, cls_loss 0.0060 cls_loss_mapping 0.0106 cls_loss_causal 0.5980 re_mapping 0.0103 re_causal 0.0308 /// teacc 98.83 lr 0.00010000 Epoch 94, weight, value: tensor([[-0.0127, 0.0982, 0.0573, ..., -0.1186, -0.0872, 0.0265], [ 0.0897, -0.0105, -0.0384, ..., 0.0503, 0.0751, -0.0930], [-0.0500, -0.0244, -0.0337, ..., -0.0540, -0.0124, -0.0287], ..., [-0.0302, 0.1079, -0.1255, ..., -0.0862, 0.0133, -0.0906], [ 0.0022, -0.0659, 0.0286, ..., 0.0466, -0.0921, -0.0360], [-0.0251, -0.0555, 0.0397, ..., -0.0547, -0.0373, 0.0903]], device='cuda:0'), grad: tensor([[ 1.7300e-05, -3.9153e-06, 6.3069e-06, ..., 3.9153e-06, 7.5437e-06, -7.2643e-06], [-2.6315e-05, 3.1479e-07, 2.3134e-06, ..., -3.9160e-05, -1.2182e-05, 1.2666e-06], [ 4.5061e-05, 8.2999e-06, 3.4332e-05, ..., 1.3210e-05, -1.2420e-05, 3.8277e-07], ..., [ 2.0579e-05, -3.6061e-06, 2.5079e-05, ..., 2.3678e-05, 8.9183e-06, 2.1160e-05], [-1.5700e-04, 3.6731e-06, -1.2851e-04, ..., -3.1978e-05, 5.7705e-06, 9.3877e-06], [ 4.1258e-07, -1.2264e-05, -7.6473e-05, ..., 2.2016e-06, 8.6380e-07, -8.5115e-05]], device='cuda:0') Epoch 94, bias, value: tensor([ 0.0045, 0.0025, 0.0057, 0.0136, 0.0249, 0.0350, -0.0250, 0.0097, -0.0342, -0.0193], device='cuda:0'), grad: tensor([ 1.5998e-04, -2.8551e-05, -1.9407e-04, 3.1137e-04, 9.2566e-05, 3.8534e-05, -1.2405e-06, 6.1631e-05, -2.9635e-04, -1.4400e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 93---------------------------------------------------- epoch 93, time 217.87, cls_loss 0.0058 cls_loss_mapping 0.0095 cls_loss_causal 0.6052 re_mapping 0.0102 re_causal 0.0312 /// teacc 98.95 lr 0.00010000 Epoch 95, weight, value: tensor([[-0.0131, 0.0986, 0.0572, ..., -0.1193, -0.0876, 0.0266], [ 0.0901, -0.0106, -0.0385, ..., 0.0507, 0.0753, -0.0933], [-0.0497, -0.0248, -0.0339, ..., -0.0538, -0.0119, -0.0290], ..., [-0.0305, 0.1086, -0.1261, ..., -0.0867, 0.0130, -0.0914], [ 0.0017, -0.0667, 0.0279, ..., 0.0459, -0.0925, -0.0375], [-0.0252, -0.0561, 0.0404, ..., -0.0549, -0.0374, 0.0910]], device='cuda:0'), grad: tensor([[ 1.2359e-06, -1.8075e-05, -2.8446e-05, ..., 1.2051e-06, -1.1466e-05, -2.5585e-05], [-8.4829e-04, -5.2881e-04, 9.2620e-07, ..., -8.3494e-04, -2.0754e-04, 6.9709e-07], [ 2.0400e-05, 1.7479e-05, 4.5486e-06, ..., 1.8254e-05, 3.0234e-05, 3.6974e-06], ..., [ 6.2704e-04, 3.7241e-04, 1.6522e-06, ..., 6.1798e-04, 1.3912e-04, 1.8086e-06], [ 1.4037e-05, 1.1742e-05, 5.4017e-06, ..., 1.4015e-05, 5.1744e-06, 1.0625e-05], [ 1.5724e-04, 1.1396e-04, 9.4995e-06, ..., 1.5748e-04, 2.8402e-05, 9.8497e-06]], device='cuda:0') Epoch 95, bias, value: tensor([ 0.0044, 0.0025, 0.0062, 0.0135, 0.0251, 0.0351, -0.0249, 0.0098, -0.0350, -0.0193], device='cuda:0'), grad: tensor([-8.0526e-05, -2.0885e-03, 6.7115e-05, 4.6015e-05, 6.6102e-05, -2.6584e-05, 4.0114e-05, 1.5030e-03, 5.9217e-05, 4.1389e-04], device='cuda:0') 100 0.0001 changing lr epoch 94, time 217.29, cls_loss 0.0072 cls_loss_mapping 0.0108 cls_loss_causal 0.6246 re_mapping 0.0104 re_causal 0.0296 /// teacc 98.93 lr 0.00010000 Epoch 96, weight, value: tensor([[-0.0131, 0.1002, 0.0583, ..., -0.1192, -0.0877, 0.0274], [ 0.0910, -0.0100, -0.0389, ..., 0.0514, 0.0757, -0.0946], [-0.0501, -0.0252, -0.0343, ..., -0.0541, -0.0122, -0.0296], ..., [-0.0317, 0.1091, -0.1273, ..., -0.0881, 0.0130, -0.0921], [ 0.0020, -0.0678, 0.0281, ..., 0.0465, -0.0916, -0.0376], [-0.0261, -0.0568, 0.0407, ..., -0.0556, -0.0382, 0.0916]], device='cuda:0'), grad: tensor([[ 1.0338e-06, -3.8594e-06, -1.0096e-06, ..., 1.3225e-06, 4.6100e-08, -2.2911e-06], [ 1.9297e-06, 8.2031e-06, 3.8221e-06, ..., 5.0142e-06, -5.4389e-07, 3.5111e-07], [ 8.5160e-06, 5.7757e-05, 9.4995e-06, ..., 5.4464e-06, 2.3795e-07, 7.9209e-07], ..., [ 1.5851e-06, -1.3125e-04, -8.2795e-07, ..., 4.0494e-06, 2.3236e-07, 6.2631e-07], [-5.9186e-07, 5.2340e-06, -4.1835e-06, ..., -2.5444e-06, 1.0710e-08, -7.6229e-07], [ 1.5935e-06, 1.3046e-05, 3.4999e-06, ..., 1.5959e-05, 1.0664e-07, 2.0713e-06]], device='cuda:0') Epoch 96, bias, value: tensor([ 0.0050, 0.0028, 0.0062, 0.0130, 0.0248, 0.0354, -0.0250, 0.0097, -0.0351, -0.0193], device='cuda:0'), grad: tensor([-1.3672e-06, 2.2426e-05, 8.3208e-05, 4.3422e-05, -6.3360e-05, -1.0595e-05, -1.3649e-05, -1.3638e-04, 1.7837e-05, 5.8651e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 95---------------------------------------------------- epoch 95, time 218.73, cls_loss 0.0065 cls_loss_mapping 0.0098 cls_loss_causal 0.5547 re_mapping 0.0104 re_causal 0.0286 /// teacc 98.98 lr 0.00010000 Epoch 97, weight, value: tensor([[-0.0155, 0.0994, 0.0587, ..., -0.1206, -0.0905, 0.0265], [ 0.0913, -0.0097, -0.0397, ..., 0.0514, 0.0765, -0.0953], [-0.0507, -0.0249, -0.0348, ..., -0.0545, -0.0125, -0.0296], ..., [-0.0324, 0.1091, -0.1277, ..., -0.0887, 0.0128, -0.0926], [ 0.0022, -0.0684, 0.0283, ..., 0.0468, -0.0932, -0.0377], [-0.0268, -0.0572, 0.0412, ..., -0.0565, -0.0383, 0.0919]], device='cuda:0'), grad: tensor([[ 1.3597e-06, -3.5688e-06, -4.3064e-06, ..., 1.3448e-06, 1.1176e-07, -5.6587e-06], [ 1.2137e-05, 1.0186e-04, 4.6846e-07, ..., -2.0444e-05, -4.6156e-06, 3.8790e-07], [ 7.1377e-06, 1.5646e-05, 3.7551e-06, ..., 4.8243e-06, 1.4715e-07, 5.7463e-07], ..., [-1.4448e-04, -6.0987e-04, 1.7118e-06, ..., 2.2296e-06, 3.6415e-07, 1.6298e-06], [ 1.4395e-05, 8.6904e-05, -1.7360e-06, ..., -4.8317e-06, 2.4140e-06, 7.5661e-06], [ 3.6001e-05, 1.4269e-04, -8.1882e-06, ..., 3.6117e-06, -1.7099e-06, -1.3180e-05]], device='cuda:0') Epoch 97, bias, value: tensor([ 0.0040, 0.0029, 0.0062, 0.0128, 0.0258, 0.0357, -0.0240, 0.0094, -0.0352, -0.0201], device='cuda:0'), grad: tensor([ 5.7071e-06, 3.3998e-04, 6.6645e-06, 8.5926e-04, 6.2644e-05, 1.6272e-05, 2.9698e-05, -2.0905e-03, 3.0065e-04, 4.6921e-04], device='cuda:0') 100 0.0001 changing lr epoch 96, time 217.34, cls_loss 0.0059 cls_loss_mapping 0.0083 cls_loss_causal 0.5662 re_mapping 0.0103 re_causal 0.0302 /// teacc 98.83 lr 0.00010000 Epoch 98, weight, value: tensor([[-0.0161, 0.0994, 0.0589, ..., -0.1221, -0.0910, 0.0263], [ 0.0914, -0.0110, -0.0408, ..., 0.0520, 0.0758, -0.0959], [-0.0504, -0.0252, -0.0340, ..., -0.0540, -0.0113, -0.0299], ..., [-0.0316, 0.1106, -0.1280, ..., -0.0887, 0.0144, -0.0932], [ 0.0019, -0.0689, 0.0285, ..., 0.0465, -0.0929, -0.0378], [-0.0277, -0.0576, 0.0412, ..., -0.0571, -0.0395, 0.0923]], device='cuda:0'), grad: tensor([[ 2.2771e-07, 4.1444e-07, 1.6158e-07, ..., 2.7567e-07, 1.4296e-07, 2.1281e-07], [-8.3493e-07, 3.3528e-07, 1.0384e-07, ..., -1.1558e-06, -1.8487e-07, -1.1362e-07], [ 5.4203e-07, -3.2857e-06, 4.3493e-07, ..., 2.1188e-07, -1.6652e-06, 1.0617e-07], ..., [ 5.1642e-07, -1.9483e-06, 7.9488e-07, ..., 9.5554e-07, 4.8522e-07, 1.2852e-06], [-1.0934e-06, 8.8476e-08, 4.4703e-07, ..., -1.0803e-06, 1.0990e-07, 1.2312e-06], [ 1.1576e-06, 2.8927e-06, -1.0701e-06, ..., 1.9446e-06, 6.1048e-07, -5.6475e-06]], device='cuda:0') Epoch 98, bias, value: tensor([ 0.0037, 0.0023, 0.0073, 0.0125, 0.0253, 0.0356, -0.0238, 0.0110, -0.0358, -0.0206], device='cuda:0'), grad: tensor([ 2.0582e-06, -3.8650e-08, -1.2361e-05, -2.4755e-06, 8.6753e-07, 2.1346e-06, 1.4901e-07, 2.4103e-06, 4.3027e-07, 6.8434e-06], device='cuda:0') 100 0.0001 changing lr epoch 97, time 218.27, cls_loss 0.0059 cls_loss_mapping 0.0100 cls_loss_causal 0.6173 re_mapping 0.0104 re_causal 0.0301 /// teacc 98.86 lr 0.00010000 Epoch 99, weight, value: tensor([[-0.0164, 0.0995, 0.0589, ..., -0.1237, -0.0912, 0.0263], [ 0.0919, -0.0116, -0.0411, ..., 0.0525, 0.0768, -0.0962], [-0.0514, -0.0254, -0.0337, ..., -0.0547, -0.0119, -0.0297], ..., [-0.0316, 0.1114, -0.1285, ..., -0.0890, 0.0142, -0.0939], [ 0.0015, -0.0697, 0.0282, ..., 0.0466, -0.0937, -0.0381], [-0.0283, -0.0581, 0.0414, ..., -0.0580, -0.0393, 0.0922]], device='cuda:0'), grad: tensor([[-4.4331e-07, -2.7895e-05, -8.8960e-06, ..., 1.2666e-06, 2.0070e-07, -1.0841e-06], [-5.6401e-06, -3.1777e-06, 2.4103e-06, ..., -1.8571e-06, -3.2969e-06, 2.7381e-06], [ 1.9632e-06, 5.3532e-06, 2.0228e-06, ..., 2.8443e-06, 3.0315e-07, 1.5367e-06], ..., [ 5.7109e-06, 3.0734e-06, 3.6657e-06, ..., 9.2760e-06, 2.5034e-06, 7.5139e-06], [-1.1340e-05, 1.6857e-06, -4.9965e-07, ..., -2.5705e-06, 9.4343e-07, 1.3039e-05], [-1.6969e-06, 3.0305e-06, -5.9158e-05, ..., -9.0742e-04, -3.0473e-06, -1.2264e-03]], device='cuda:0') Epoch 99, bias, value: tensor([ 0.0033, 0.0022, 0.0071, 0.0125, 0.0263, 0.0360, -0.0234, 0.0113, -0.0362, -0.0218], device='cuda:0'), grad: tensor([-6.5744e-05, 3.3267e-06, 5.8152e-06, 1.7151e-05, 4.1771e-03, 1.3836e-05, 6.9022e-05, 3.3945e-05, 3.1084e-05, -4.2877e-03], device='cuda:0') 100 0.0001 changing lr epoch 98, time 217.43, cls_loss 0.0044 cls_loss_mapping 0.0072 cls_loss_causal 0.5994 re_mapping 0.0102 re_causal 0.0298 /// teacc 98.85 lr 0.00010000 Epoch 100, weight, value: tensor([[-0.0168, 0.0996, 0.0589, ..., -0.1244, -0.0918, 0.0261], [ 0.0919, -0.0116, -0.0413, ..., 0.0528, 0.0769, -0.0964], [-0.0515, -0.0255, -0.0337, ..., -0.0550, -0.0120, -0.0294], ..., [-0.0320, 0.1118, -0.1293, ..., -0.0898, 0.0139, -0.0947], [ 0.0017, -0.0703, 0.0281, ..., 0.0471, -0.0938, -0.0384], [-0.0283, -0.0585, 0.0419, ..., -0.0583, -0.0397, 0.0928]], device='cuda:0'), grad: tensor([[ 1.1727e-05, 4.3362e-06, 6.1132e-06, ..., 1.6242e-05, 1.6391e-07, 1.7472e-06], [-1.3781e-03, -1.5345e-03, 4.6939e-07, ..., -8.9502e-04, -2.5947e-06, 3.0966e-07], [ 6.9477e-06, 5.6028e-06, 1.9148e-06, ..., 6.7689e-06, 3.6927e-07, 5.6671e-07], ..., [ 1.1168e-03, 1.2169e-03, 1.5132e-05, ..., 7.2432e-04, 8.9174e-07, 2.1741e-05], [ 9.6411e-06, 4.1686e-06, 6.0834e-06, ..., 1.1824e-05, 5.1642e-07, 4.9621e-06], [ 1.5175e-04, 1.9717e-04, -1.6063e-05, ..., 9.8825e-05, 2.9989e-07, -2.7046e-05]], device='cuda:0') Epoch 100, bias, value: tensor([ 0.0029, 0.0023, 0.0071, 0.0126, 0.0264, 0.0363, -0.0235, 0.0111, -0.0361, -0.0218], device='cuda:0'), grad: tensor([ 3.6180e-05, -3.8548e-03, 1.8761e-05, 5.5671e-05, 2.7323e-04, -3.1382e-05, -7.4267e-05, 3.1204e-03, 3.0518e-05, 4.2391e-04], device='cuda:0') 100 0.0001 changing lr epoch 99, time 221.61, cls_loss 0.0072 cls_loss_mapping 0.0106 cls_loss_causal 0.5654 re_mapping 0.0104 re_causal 0.0290 /// teacc 98.89 lr 0.00010000 Epoch 101, weight, value: tensor([[-0.0169, 0.1004, 0.0600, ..., -0.1255, -0.0919, 0.0267], [ 0.0918, -0.0112, -0.0430, ..., 0.0519, 0.0771, -0.0991], [-0.0523, -0.0260, -0.0333, ..., -0.0561, -0.0120, -0.0298], ..., [-0.0332, 0.1117, -0.1308, ..., -0.0922, 0.0137, -0.0952], [ 0.0015, -0.0704, 0.0277, ..., 0.0471, -0.0945, -0.0388], [-0.0267, -0.0590, 0.0445, ..., -0.0555, -0.0399, 0.0947]], device='cuda:0'), grad: tensor([[ 1.9558e-06, -3.2913e-06, 3.9022e-07, ..., 3.2410e-06, 2.9989e-07, 9.9838e-07], [ 6.3777e-06, 2.6394e-06, 5.2489e-06, ..., 1.2763e-05, -8.8476e-08, 3.0342e-06], [ 5.7230e-07, 7.6834e-07, 8.3586e-07, ..., 1.4193e-06, 4.1910e-09, 5.8208e-07], ..., [ 2.7176e-06, -9.7007e-06, 2.7083e-06, ..., 5.8487e-06, 5.0943e-07, 3.4384e-06], [-7.2680e-06, 7.3994e-07, -1.7434e-06, ..., -1.1712e-05, 6.9663e-07, 3.6061e-06], [ 4.3325e-06, 4.1835e-06, -5.5274e-07, ..., 7.5698e-06, 2.2491e-07, -2.0787e-06]], device='cuda:0') Epoch 101, bias, value: tensor([ 0.0035, 0.0014, 0.0072, 0.0124, 0.0255, 0.0366, -0.0230, 0.0100, -0.0369, -0.0196], device='cuda:0'), grad: tensor([ 1.2908e-06, 2.8968e-05, -1.9670e-06, 2.0757e-05, -1.0289e-05, -7.7546e-05, 4.5121e-05, -2.0474e-05, -5.3942e-06, 1.9670e-05], device='cuda:0') 100 0.0001 changing lr epoch 100, time 220.01, cls_loss 0.0048 cls_loss_mapping 0.0064 cls_loss_causal 0.6021 re_mapping 0.0101 re_causal 0.0298 /// teacc 98.72 lr 0.00010000 Epoch 102, weight, value: tensor([[-0.0170, 0.1011, 0.0603, ..., -0.1261, -0.0923, 0.0266], [ 0.0928, -0.0111, -0.0426, ..., 0.0528, 0.0773, -0.0991], [-0.0530, -0.0265, -0.0337, ..., -0.0568, -0.0122, -0.0301], ..., [-0.0337, 0.1122, -0.1315, ..., -0.0932, 0.0139, -0.0955], [ 0.0010, -0.0707, 0.0270, ..., 0.0466, -0.0946, -0.0394], [-0.0274, -0.0594, 0.0450, ..., -0.0559, -0.0400, 0.0954]], device='cuda:0'), grad: tensor([[ 4.9686e-07, -7.3537e-06, -2.5947e-06, ..., 2.2501e-06, 2.1886e-08, -3.2000e-06], [ 2.7418e-04, 2.7241e-07, 2.5658e-07, ..., 9.8515e-04, -2.7427e-07, 2.0349e-07], [ 8.1398e-07, 5.9884e-07, 3.6415e-07, ..., 2.2352e-06, 5.1688e-08, 2.5332e-07], ..., [ 4.6119e-06, 1.5087e-07, 6.2771e-07, ..., 1.6108e-05, 8.8010e-08, 7.2224e-07], [ 5.0664e-06, 2.5472e-07, 1.1679e-06, ..., 1.6883e-05, 3.4459e-08, 1.1679e-06], [ 4.5151e-06, 1.3215e-06, -6.5984e-07, ..., 1.7345e-05, 3.2131e-08, -3.4049e-06]], device='cuda:0') Epoch 102, bias, value: tensor([ 0.0038, 0.0019, 0.0067, 0.0126, 0.0258, 0.0369, -0.0236, 0.0099, -0.0378, -0.0195], device='cuda:0'), grad: tensor([-8.3670e-06, 1.1654e-03, 3.8296e-06, -2.7437e-06, -1.2550e-03, 8.3074e-06, 2.5705e-05, 2.0385e-05, 2.2292e-05, 1.9699e-05], device='cuda:0') 100 0.0001 changing lr epoch 101, time 219.74, cls_loss 0.0046 cls_loss_mapping 0.0077 cls_loss_causal 0.5650 re_mapping 0.0098 re_causal 0.0287 /// teacc 98.82 lr 0.00010000 Epoch 103, weight, value: tensor([[-0.0171, 0.1016, 0.0605, ..., -0.1268, -0.0924, 0.0266], [ 0.0935, -0.0108, -0.0433, ..., 0.0532, 0.0773, -0.0994], [-0.0533, -0.0270, -0.0337, ..., -0.0569, -0.0120, -0.0300], ..., [-0.0342, 0.1127, -0.1309, ..., -0.0948, 0.0139, -0.0959], [ 0.0014, -0.0709, 0.0273, ..., 0.0470, -0.0946, -0.0395], [-0.0289, -0.0603, 0.0449, ..., -0.0565, -0.0403, 0.0959]], device='cuda:0'), grad: tensor([[ 1.7975e-06, 1.3253e-06, 2.4009e-06, ..., 2.1122e-06, 7.9162e-09, 6.3051e-07], [ 9.6671e-07, 1.2880e-06, 7.9125e-06, ..., 1.3910e-05, -1.8720e-07, 5.5432e-06], [ 3.3975e-06, 5.9977e-06, 1.1530e-06, ..., 2.6841e-06, 1.2573e-08, 2.3609e-07], ..., [ 5.5227e-07, -1.5005e-05, 1.4221e-06, ..., 2.6915e-06, 8.0559e-08, 1.0971e-06], [-1.1253e-04, 5.8487e-07, -1.6117e-04, ..., -1.1420e-04, 5.9605e-08, -2.7984e-05], [ 8.4341e-06, 3.3583e-06, 9.7811e-05, ..., 1.9884e-04, 5.8673e-08, 5.7727e-05]], device='cuda:0') Epoch 103, bias, value: tensor([ 0.0039, 0.0017, 0.0070, 0.0126, 0.0260, 0.0366, -0.0236, 0.0105, -0.0376, -0.0202], device='cuda:0'), grad: tensor([ 7.4431e-06, 3.0071e-05, 1.6794e-05, 1.7321e-04, -3.9005e-04, 3.4660e-05, 7.3165e-06, -1.4804e-05, -2.4796e-04, 3.8242e-04], device='cuda:0') 100 0.0001 changing lr epoch 102, time 219.44, cls_loss 0.0048 cls_loss_mapping 0.0085 cls_loss_causal 0.5868 re_mapping 0.0100 re_causal 0.0291 /// teacc 98.92 lr 0.00010000 Epoch 104, weight, value: tensor([[-0.0173, 0.1022, 0.0611, ..., -0.1273, -0.0924, 0.0270], [ 0.0951, -0.0115, -0.0426, ..., 0.0543, 0.0785, -0.0994], [-0.0542, -0.0273, -0.0343, ..., -0.0574, -0.0126, -0.0303], ..., [-0.0354, 0.1146, -0.1313, ..., -0.0953, 0.0133, -0.0961], [ 0.0017, -0.0711, 0.0274, ..., 0.0472, -0.0946, -0.0395], [-0.0299, -0.0627, 0.0447, ..., -0.0575, -0.0407, 0.0960]], device='cuda:0'), grad: tensor([[ 4.6333e-07, -7.4692e-06, -4.6045e-06, ..., 2.9728e-06, 3.2876e-07, -1.7677e-06], [ 1.0706e-05, 2.3216e-05, 1.3523e-06, ..., 1.3873e-05, -4.6939e-07, 5.9791e-07], [ 5.3868e-06, 2.6524e-06, 3.0342e-06, ..., -5.6744e-05, -3.2084e-07, 3.8929e-07], ..., [-1.7226e-05, -4.7892e-05, 2.6375e-06, ..., -2.2694e-05, 2.2491e-07, 1.8198e-06], [-8.4657e-07, 4.7795e-06, 6.5938e-06, ..., -5.8953e-07, -1.4016e-07, 8.1509e-06], [ 1.9325e-07, 7.1116e-06, -1.5110e-05, ..., 2.5965e-06, 1.9511e-07, -1.5840e-05]], device='cuda:0') Epoch 104, bias, value: tensor([ 0.0044, 0.0026, 0.0065, 0.0127, 0.0263, 0.0366, -0.0241, 0.0108, -0.0373, -0.0212], device='cuda:0'), grad: tensor([ 2.3589e-05, 5.8800e-05, -8.2445e-04, 3.7074e-05, 7.0477e-04, 7.8976e-06, 6.1810e-05, -1.1665e-04, 4.1664e-05, 5.0440e-06], device='cuda:0') 100 0.0001 changing lr epoch 103, time 219.90, cls_loss 0.0053 cls_loss_mapping 0.0072 cls_loss_causal 0.6032 re_mapping 0.0099 re_causal 0.0287 /// teacc 98.91 lr 0.00010000 Epoch 105, weight, value: tensor([[-0.0172, 0.1028, 0.0608, ..., -0.1273, -0.0924, 0.0263], [ 0.0963, -0.0116, -0.0424, ..., 0.0556, 0.0796, -0.0996], [-0.0550, -0.0280, -0.0347, ..., -0.0581, -0.0133, -0.0309], ..., [-0.0359, 0.1150, -0.1318, ..., -0.0962, 0.0129, -0.0970], [ 0.0004, -0.0716, 0.0268, ..., 0.0465, -0.0963, -0.0397], [-0.0306, -0.0623, 0.0458, ..., -0.0577, -0.0408, 0.0974]], device='cuda:0'), grad: tensor([[ 1.4668e-06, -3.8603e-07, 1.3269e-05, ..., 1.1837e-06, 2.9709e-07, 1.9908e-05], [ 2.8051e-06, 7.5903e-08, 9.9018e-06, ..., -1.8207e-07, 2.7753e-06, 7.3165e-06], [ 2.2966e-06, 5.4482e-08, 9.5218e-06, ..., 4.6287e-07, 2.6450e-06, 9.6038e-06], ..., [ 9.0897e-06, -1.5507e-07, 2.7344e-05, ..., 9.7603e-07, 4.6641e-06, 8.5011e-06], [ 2.2762e-06, 1.0757e-07, 1.0706e-05, ..., 4.2422e-07, 2.7362e-06, 1.3269e-05], [ 4.1388e-06, 1.9604e-07, -9.3043e-05, ..., 1.0384e-06, 9.0478e-07, -1.5414e-04]], device='cuda:0') Epoch 105, bias, value: tensor([ 0.0040, 0.0034, 0.0062, 0.0123, 0.0261, 0.0378, -0.0248, 0.0106, -0.0387, -0.0205], device='cuda:0'), grad: tensor([ 3.0339e-05, 2.6643e-05, 2.2814e-05, -8.0526e-05, 2.0504e-05, -2.2042e-04, 2.7108e-04, 8.4877e-05, 2.7955e-05, -1.8370e-04], device='cuda:0') 100 0.0001 changing lr epoch 104, time 219.66, cls_loss 0.0041 cls_loss_mapping 0.0064 cls_loss_causal 0.5925 re_mapping 0.0096 re_causal 0.0281 /// teacc 98.91 lr 0.00010000 Epoch 106, weight, value: tensor([[-0.0173, 0.1034, 0.0614, ..., -0.1283, -0.0924, 0.0265], [ 0.0967, -0.0117, -0.0424, ..., 0.0558, 0.0800, -0.0997], [-0.0551, -0.0280, -0.0347, ..., -0.0586, -0.0133, -0.0312], ..., [-0.0361, 0.1154, -0.1323, ..., -0.0966, 0.0132, -0.0980], [ 0.0003, -0.0721, 0.0268, ..., 0.0465, -0.0971, -0.0398], [-0.0307, -0.0626, 0.0458, ..., -0.0582, -0.0414, 0.0977]], device='cuda:0'), grad: tensor([[ 1.3486e-06, -3.2014e-07, 1.3225e-06, ..., 1.3718e-06, 1.3504e-08, 9.7789e-09], [ 6.2678e-07, 2.0536e-07, 1.5395e-06, ..., 8.8662e-07, -2.1001e-07, 1.9558e-07], [ 7.0743e-06, 1.0221e-07, 8.0615e-06, ..., 5.6326e-06, -6.7521e-09, 3.1781e-07], ..., [ 2.5295e-06, -2.5835e-06, 3.2522e-06, ..., 7.5996e-06, -2.9965e-07, 1.2964e-06], [ 2.2631e-06, 1.4598e-07, 1.7183e-06, ..., -3.0547e-06, 6.7987e-08, 2.3395e-06], [ 1.7351e-06, 1.2266e-06, -2.9746e-06, ..., 1.5404e-06, 2.3609e-07, -1.1131e-05]], device='cuda:0') Epoch 106, bias, value: tensor([ 0.0043, 0.0034, 0.0063, 0.0121, 0.0265, 0.0380, -0.0251, 0.0104, -0.0391, -0.0205], device='cuda:0'), grad: tensor([ 5.5991e-06, 6.6496e-06, 2.3454e-05, -7.0453e-05, -7.4580e-06, 1.2569e-05, -1.8254e-07, 1.4909e-05, 1.1951e-05, 3.0063e-06], device='cuda:0') 100 0.0001 changing lr epoch 105, time 219.54, cls_loss 0.0060 cls_loss_mapping 0.0076 cls_loss_causal 0.5872 re_mapping 0.0099 re_causal 0.0296 /// teacc 98.81 lr 0.00010000 Epoch 107, weight, value: tensor([[-0.0177, 0.1038, 0.0618, ..., -0.1294, -0.0925, 0.0266], [ 0.0961, -0.0117, -0.0434, ..., 0.0547, 0.0799, -0.1014], [-0.0553, -0.0285, -0.0345, ..., -0.0587, -0.0129, -0.0306], ..., [-0.0362, 0.1158, -0.1328, ..., -0.0969, 0.0133, -0.0988], [ 0.0013, -0.0723, 0.0277, ..., 0.0484, -0.0970, -0.0388], [-0.0311, -0.0630, 0.0454, ..., -0.0584, -0.0417, 0.0978]], device='cuda:0'), grad: tensor([[ 2.7753e-06, 5.6744e-04, 2.6226e-03, ..., 3.0026e-06, 4.1910e-09, 4.9210e-03], [-2.1304e-07, 1.9446e-06, 2.9318e-06, ..., 4.9174e-06, -1.6298e-07, 5.2564e-06], [ 3.5157e-08, 6.1020e-06, 1.7121e-05, ..., -1.7524e-05, 1.4203e-08, 3.0205e-05], ..., [ 2.1537e-07, -3.7216e-06, 9.0674e-06, ..., 1.5572e-06, 2.8405e-08, 1.6674e-05], [ 2.3302e-06, 9.3551e-07, 5.4538e-06, ..., 6.3591e-06, 6.1933e-08, 4.2357e-06], [ 5.5321e-07, -5.8031e-04, -2.6760e-03, ..., 1.0438e-05, 1.6764e-08, -5.0316e-03]], device='cuda:0') Epoch 107, bias, value: tensor([ 0.0045, 0.0028, 0.0066, 0.0141, 0.0262, 0.0365, -0.0252, 0.0101, -0.0380, -0.0207], device='cuda:0'), grad: tensor([ 8.3618e-03, 2.0325e-05, 1.2279e-05, 3.0547e-05, 2.7061e-05, 2.2352e-05, 1.0967e-05, 2.1651e-05, 2.2978e-05, -8.5220e-03], device='cuda:0') 100 0.0001 changing lr epoch 106, time 219.85, cls_loss 0.0065 cls_loss_mapping 0.0107 cls_loss_causal 0.5779 re_mapping 0.0098 re_causal 0.0289 /// teacc 98.82 lr 0.00010000 Epoch 108, weight, value: tensor([[-0.0176, 0.1046, 0.0618, ..., -0.1298, -0.0926, 0.0264], [ 0.0951, -0.0103, -0.0428, ..., 0.0540, 0.0801, -0.1019], [-0.0557, -0.0289, -0.0351, ..., -0.0596, -0.0129, -0.0310], ..., [-0.0343, 0.1151, -0.1336, ..., -0.0949, 0.0135, -0.0996], [ 0.0019, -0.0721, 0.0280, ..., 0.0492, -0.0974, -0.0385], [-0.0311, -0.0638, 0.0459, ..., -0.0590, -0.0419, 0.0985]], device='cuda:0'), grad: tensor([[ 6.0629e-07, 4.0382e-06, 1.0803e-06, ..., 1.7285e-06, 2.9476e-07, 2.6952e-06], [-7.0155e-05, 8.7693e-06, -1.3545e-05, ..., -3.4183e-05, -4.5389e-05, 5.8673e-07], [ 3.1814e-06, -7.0035e-05, -5.3681e-06, ..., -2.4494e-06, -1.5758e-06, 2.8964e-07], ..., [ 5.3585e-05, -1.9848e-05, 1.1884e-05, ..., 2.9594e-05, 3.5077e-05, -1.6898e-05], [-1.1921e-06, 3.4515e-06, -5.4436e-07, ..., -2.7753e-06, 9.7789e-07, 1.2806e-08], [ 2.0750e-06, 4.4674e-05, 1.6466e-06, ..., 2.1383e-06, 1.1567e-06, 1.4976e-05]], device='cuda:0') Epoch 108, bias, value: tensor([ 0.0043, 0.0026, 0.0061, 0.0131, 0.0267, 0.0372, -0.0261, 0.0107, -0.0377, -0.0208], device='cuda:0'), grad: tensor([ 1.2018e-05, -9.4950e-05, -5.3930e-04, 2.4104e-04, 8.3074e-07, -1.8418e-05, 3.2838e-06, 2.8896e-04, 2.7955e-05, 7.9453e-05], device='cuda:0') 100 0.0001 changing lr epoch 107, time 219.36, cls_loss 0.0056 cls_loss_mapping 0.0071 cls_loss_causal 0.5808 re_mapping 0.0096 re_causal 0.0274 /// teacc 98.85 lr 0.00010000 Epoch 109, weight, value: tensor([[-0.0184, 0.1053, 0.0624, ..., -0.1315, -0.0932, 0.0265], [ 0.0953, -0.0102, -0.0431, ..., 0.0541, 0.0803, -0.1022], [-0.0559, -0.0296, -0.0357, ..., -0.0596, -0.0131, -0.0317], ..., [-0.0346, 0.1154, -0.1353, ..., -0.0954, 0.0135, -0.1012], [ 0.0024, -0.0725, 0.0286, ..., 0.0501, -0.0974, -0.0382], [-0.0313, -0.0638, 0.0472, ..., -0.0593, -0.0420, 0.0997]], device='cuda:0'), grad: tensor([[ 3.1441e-05, 8.4788e-06, -1.5944e-05, ..., 4.3750e-05, 7.9162e-09, 9.5144e-06], [-9.5606e-05, 8.8438e-06, 4.4517e-07, ..., -1.7083e-04, -1.6438e-07, 7.1665e-07], [ 5.2117e-06, 9.3162e-05, 2.6003e-05, ..., 2.8118e-05, 6.2399e-08, 2.6971e-05], ..., [ 8.2031e-06, -1.1736e-04, 4.5309e-07, ..., 1.6093e-05, 3.4925e-08, 5.6531e-07], [ 1.9297e-05, 7.1004e-06, 1.3560e-06, ..., 3.1859e-05, 1.2107e-08, 5.3346e-06], [ 1.4052e-05, 3.8773e-05, 1.1377e-05, ..., 2.4021e-05, 1.3970e-08, 1.7196e-05]], device='cuda:0') Epoch 109, bias, value: tensor([ 0.0044, 0.0025, 0.0061, 0.0123, 0.0266, 0.0368, -0.0261, 0.0106, -0.0371, -0.0201], device='cuda:0'), grad: tensor([ 4.8965e-05, -1.7607e-04, 2.4939e-04, 4.4882e-05, 5.7608e-05, 5.6416e-05, -2.1255e-04, -2.3341e-04, 5.3197e-05, 1.1170e-04], device='cuda:0') 100 0.0001 changing lr epoch 108, time 219.56, cls_loss 0.0046 cls_loss_mapping 0.0074 cls_loss_causal 0.5783 re_mapping 0.0103 re_causal 0.0284 /// teacc 98.85 lr 0.00010000 Epoch 110, weight, value: tensor([[-0.0189, 0.1053, 0.0623, ..., -0.1319, -0.0937, 0.0257], [ 0.0956, -0.0099, -0.0431, ..., 0.0545, 0.0807, -0.1024], [-0.0562, -0.0305, -0.0356, ..., -0.0599, -0.0133, -0.0316], ..., [-0.0348, 0.1163, -0.1351, ..., -0.0960, 0.0134, -0.1004], [ 0.0026, -0.0727, 0.0290, ..., 0.0507, -0.0975, -0.0381], [-0.0316, -0.0653, 0.0475, ..., -0.0597, -0.0423, 0.1006]], device='cuda:0'), grad: tensor([[ 4.5970e-06, -7.9945e-06, -5.0887e-06, ..., 1.2867e-05, 3.3062e-08, -6.7428e-06], [ 6.3218e-06, 2.4319e-05, 1.3728e-06, ..., 2.7046e-05, -9.8627e-07, 5.2061e-07], [ 9.3430e-06, 2.4050e-05, 2.6766e-06, ..., 2.4766e-05, 3.8929e-07, 5.3551e-07], ..., [ 1.6302e-05, 1.0550e-05, 1.0051e-05, ..., 5.1767e-05, 2.0070e-07, 1.3784e-05], [ 2.5019e-05, 2.4378e-05, 2.2560e-05, ..., 4.9859e-05, 6.2771e-07, 1.7956e-05], [ 7.1116e-06, 1.7568e-05, -2.2367e-05, ..., 1.8790e-05, 1.7509e-07, -2.9162e-05]], device='cuda:0') Epoch 110, bias, value: tensor([ 0.0034, 0.0027, 0.0061, 0.0121, 0.0267, 0.0365, -0.0260, 0.0110, -0.0368, -0.0203], device='cuda:0'), grad: tensor([-5.7518e-06, 7.5877e-05, 7.1168e-05, -1.9029e-05, -5.5504e-04, 1.0920e-04, 4.8101e-05, 1.0926e-04, 1.5938e-04, 6.6683e-06], device='cuda:0') 100 0.0001 changing lr epoch 109, time 219.88, cls_loss 0.0057 cls_loss_mapping 0.0080 cls_loss_causal 0.5823 re_mapping 0.0094 re_causal 0.0280 /// teacc 98.81 lr 0.00010000 Epoch 111, weight, value: tensor([[-0.0191, 0.1060, 0.0630, ..., -0.1320, -0.0937, 0.0258], [ 0.0974, -0.0100, -0.0437, ..., 0.0552, 0.0827, -0.1031], [-0.0584, -0.0311, -0.0364, ..., -0.0609, -0.0152, -0.0328], ..., [-0.0351, 0.1168, -0.1366, ..., -0.0967, 0.0132, -0.1023], [ 0.0024, -0.0731, 0.0287, ..., 0.0507, -0.0977, -0.0384], [-0.0319, -0.0656, 0.0486, ..., -0.0598, -0.0426, 0.1018]], device='cuda:0'), grad: tensor([[ 2.6058e-06, 6.0536e-08, 7.6462e-07, ..., 1.9856e-06, 3.3947e-07, 3.9395e-07], [ 6.0955e-07, 3.7951e-07, 1.4948e-07, ..., -7.7952e-07, -5.5879e-07, 7.2177e-08], [ 1.0088e-05, 6.5193e-07, 1.3160e-06, ..., 6.4038e-06, 1.9297e-06, 6.5193e-07], ..., [ 1.0487e-06, -3.6806e-06, 8.4750e-07, ..., 6.0871e-06, 2.7195e-07, 9.2341e-07], [-2.3052e-05, 2.4401e-07, -1.3545e-05, ..., -1.3284e-05, 5.7509e-07, -6.5975e-06], [ 1.4946e-05, 1.3970e-06, 7.1600e-06, ..., 1.1750e-05, 2.1840e-07, 2.2501e-06]], device='cuda:0') Epoch 111, bias, value: tensor([ 0.0036, 0.0033, 0.0051, 0.0125, 0.0268, 0.0364, -0.0265, 0.0105, -0.0374, -0.0195], device='cuda:0'), grad: tensor([ 7.1228e-06, 5.9083e-06, 2.0504e-05, -1.2517e-06, -1.9684e-05, 1.2189e-05, -2.5257e-05, 1.2934e-05, -5.5313e-05, 4.2737e-05], device='cuda:0') 100 0.0001 changing lr epoch 110, time 219.79, cls_loss 0.0053 cls_loss_mapping 0.0084 cls_loss_causal 0.5818 re_mapping 0.0097 re_causal 0.0273 /// teacc 98.92 lr 0.00010000 Epoch 112, weight, value: tensor([[-0.0192, 0.1064, 0.0634, ..., -0.1324, -0.0940, 0.0260], [ 0.0981, -0.0102, -0.0434, ..., 0.0558, 0.0826, -0.1033], [-0.0585, -0.0315, -0.0356, ..., -0.0608, -0.0149, -0.0329], ..., [-0.0353, 0.1170, -0.1374, ..., -0.0976, 0.0136, -0.1033], [ 0.0026, -0.0722, 0.0285, ..., 0.0512, -0.0980, -0.0386], [-0.0321, -0.0665, 0.0494, ..., -0.0605, -0.0429, 0.1025]], device='cuda:0'), grad: tensor([[ 9.2462e-06, 1.2927e-06, -1.7602e-06, ..., 1.0513e-05, 7.1758e-07, -2.1085e-06], [ 1.9614e-06, 2.6412e-06, 7.2829e-07, ..., 3.2298e-06, 5.7695e-07, 3.5809e-07], [ 1.6410e-06, 1.2834e-06, 4.2794e-07, ..., 2.0862e-06, -4.8354e-06, 5.8906e-07], ..., [ 4.9872e-07, -9.7454e-06, 1.4910e-06, ..., 1.1325e-06, 1.2675e-06, 1.2117e-06], [ 1.3001e-05, 3.1590e-06, 5.6922e-06, ..., 1.3441e-05, 1.6503e-06, 9.0674e-06], [ 1.9278e-06, 5.8711e-06, 2.8498e-07, ..., 3.0920e-06, 6.0862e-07, 8.2888e-07]], device='cuda:0') Epoch 112, bias, value: tensor([ 0.0038, 0.0032, 0.0055, 0.0126, 0.0269, 0.0365, -0.0274, 0.0102, -0.0367, -0.0197], device='cuda:0'), grad: tensor([ 1.7300e-05, 1.2845e-05, -1.3098e-05, -2.1122e-06, 3.1024e-05, -3.7074e-05, -5.5254e-05, -5.9046e-06, 3.8177e-05, 1.4044e-05], device='cuda:0') 100 0.0001 changing lr epoch 111, time 219.68, cls_loss 0.0042 cls_loss_mapping 0.0064 cls_loss_causal 0.5716 re_mapping 0.0092 re_causal 0.0261 /// teacc 98.85 lr 0.00010000 Epoch 113, weight, value: tensor([[-0.0194, 0.1066, 0.0642, ..., -0.1330, -0.0941, 0.0264], [ 0.0989, -0.0102, -0.0431, ..., 0.0566, 0.0831, -0.1030], [-0.0588, -0.0319, -0.0351, ..., -0.0606, -0.0147, -0.0330], ..., [-0.0355, 0.1172, -0.1390, ..., -0.0981, 0.0136, -0.1042], [ 0.0024, -0.0727, 0.0281, ..., 0.0509, -0.0985, -0.0393], [-0.0326, -0.0667, 0.0499, ..., -0.0610, -0.0435, 0.1031]], device='cuda:0'), grad: tensor([[ 9.5740e-07, -2.4308e-06, -7.4357e-06, ..., 2.9579e-06, 5.1688e-08, -1.3359e-05], [-1.7555e-07, 5.8487e-07, 4.7609e-06, ..., 8.2143e-07, -1.4016e-07, 4.5411e-06], [ 4.0652e-07, 4.9500e-07, 1.0990e-05, ..., 1.9651e-06, 1.9092e-08, 2.9616e-06], ..., [ 2.7288e-07, -1.6475e-06, 6.2548e-06, ..., 1.3988e-06, 1.8068e-07, 6.5491e-06], [-8.3447e-07, 9.6299e-07, 8.5309e-06, ..., -6.6543e-07, 6.8452e-08, 1.8030e-06], [ 3.6974e-07, 1.1772e-06, 1.1409e-06, ..., 1.8757e-06, -2.9383e-07, -1.0528e-05]], device='cuda:0') Epoch 113, bias, value: tensor([ 0.0042, 0.0034, 0.0058, 0.0132, 0.0268, 0.0361, -0.0273, 0.0099, -0.0372, -0.0198], device='cuda:0'), grad: tensor([-1.5974e-05, 1.5333e-05, 2.9907e-05, -1.0639e-04, 8.8632e-05, 1.6347e-05, -9.0003e-05, 1.5393e-05, 3.4779e-05, 1.2077e-05], device='cuda:0') 100 0.0001 changing lr epoch 112, time 219.36, cls_loss 0.0071 cls_loss_mapping 0.0102 cls_loss_causal 0.5708 re_mapping 0.0101 re_causal 0.0273 /// teacc 98.86 lr 0.00010000 Epoch 114, weight, value: tensor([[-0.0196, 0.1060, 0.0635, ..., -0.1339, -0.0941, 0.0250], [ 0.0991, -0.0104, -0.0437, ..., 0.0572, 0.0833, -0.1035], [-0.0590, -0.0326, -0.0351, ..., -0.0608, -0.0146, -0.0323], ..., [-0.0357, 0.1184, -0.1392, ..., -0.0987, 0.0135, -0.1042], [ 0.0012, -0.0747, 0.0264, ..., 0.0499, -0.0985, -0.0418], [-0.0334, -0.0664, 0.0513, ..., -0.0614, -0.0437, 0.1050]], device='cuda:0'), grad: tensor([[-1.4435e-08, -2.9188e-06, -5.7742e-06, ..., 4.1677e-07, 3.7719e-08, -6.1393e-06], [-1.1455e-06, 5.8115e-06, 1.7183e-07, ..., -1.2843e-06, -2.5472e-07, 1.2526e-07], [ 5.9791e-07, 3.9712e-06, 2.4633e-07, ..., 4.2561e-07, -5.4017e-08, 2.0443e-07], ..., [ 1.0533e-06, -2.3648e-05, 5.2713e-07, ..., 1.4389e-06, 3.3947e-07, 6.3935e-07], [ 6.4308e-07, 5.9605e-07, 1.0198e-06, ..., 1.8068e-07, 3.1199e-07, 1.1856e-06], [ 4.4983e-07, 6.1989e-06, 2.0824e-06, ..., 7.9488e-07, 7.7765e-08, 2.7604e-06]], device='cuda:0') Epoch 114, bias, value: tensor([ 0.0029, 0.0034, 0.0059, 0.0135, 0.0259, 0.0353, -0.0262, 0.0104, -0.0383, -0.0187], device='cuda:0'), grad: tensor([-8.9556e-06, 5.8040e-06, 6.4820e-07, 1.5333e-05, 1.5134e-07, -1.7807e-06, -2.7083e-06, -2.2605e-05, 3.2391e-06, 1.0870e-05], device='cuda:0') 100 0.0001 changing lr epoch 113, time 219.56, cls_loss 0.0049 cls_loss_mapping 0.0064 cls_loss_causal 0.5931 re_mapping 0.0091 re_causal 0.0264 /// teacc 98.98 lr 0.00010000 Epoch 115, weight, value: tensor([[-0.0200, 0.1065, 0.0639, ..., -0.1346, -0.0942, 0.0254], [ 0.0999, -0.0096, -0.0439, ..., 0.0577, 0.0836, -0.1036], [-0.0594, -0.0323, -0.0352, ..., -0.0613, -0.0150, -0.0330], ..., [-0.0361, 0.1180, -0.1396, ..., -0.0991, 0.0139, -0.1050], [ 0.0003, -0.0752, 0.0261, ..., 0.0493, -0.0997, -0.0423], [-0.0339, -0.0671, 0.0517, ..., -0.0621, -0.0440, 0.1052]], device='cuda:0'), grad: tensor([[ 7.8045e-07, -6.8499e-07, -1.4063e-06, ..., 6.5900e-06, 1.9139e-07, -3.5670e-07], [ 3.2485e-05, -4.5635e-08, 3.3557e-05, ..., 1.3614e-04, 1.0863e-05, 5.5760e-05], [-3.3341e-06, 4.6566e-08, 5.2340e-07, ..., -1.3256e-04, 1.5507e-07, 7.4133e-07], ..., [ 2.3581e-06, 3.5483e-07, 5.6550e-06, ..., 2.0191e-05, 2.6356e-07, 6.2995e-06], [-3.1963e-06, 8.6613e-08, 3.1013e-07, ..., -1.9193e-05, 2.6859e-06, 1.1370e-05], [ 1.7300e-05, -8.6613e-07, 3.6899e-06, ..., 2.3082e-05, 6.3479e-06, 1.3009e-05]], device='cuda:0') Epoch 115, bias, value: tensor([ 0.0031, 0.0040, 0.0066, 0.0129, 0.0263, 0.0362, -0.0260, 0.0095, -0.0393, -0.0190], device='cuda:0'), grad: tensor([ 3.0607e-05, 6.5756e-04, -7.4720e-04, 5.4747e-05, 4.7982e-05, -1.3673e-04, 2.1473e-05, 5.7638e-05, 6.6590e-08, 1.2949e-05], device='cuda:0') 100 0.0001 changing lr epoch 114, time 219.59, cls_loss 0.0036 cls_loss_mapping 0.0066 cls_loss_causal 0.6067 re_mapping 0.0088 re_causal 0.0276 /// teacc 98.87 lr 0.00010000 Epoch 116, weight, value: tensor([[-0.0201, 0.1067, 0.0641, ..., -0.1354, -0.0942, 0.0256], [ 0.1001, -0.0093, -0.0437, ..., 0.0579, 0.0837, -0.1037], [-0.0595, -0.0321, -0.0349, ..., -0.0615, -0.0151, -0.0336], ..., [-0.0365, 0.1179, -0.1407, ..., -0.0995, 0.0137, -0.1067], [ 0.0009, -0.0755, 0.0265, ..., 0.0502, -0.0997, -0.0419], [-0.0343, -0.0672, 0.0514, ..., -0.0634, -0.0442, 0.1051]], device='cuda:0'), grad: tensor([[ 3.9535e-07, -3.3174e-06, -1.1399e-06, ..., 1.4165e-06, 8.2050e-07, -1.0170e-06], [-3.0268e-07, 1.0334e-05, 8.0699e-07, ..., 5.5470e-06, -2.6310e-07, 4.0093e-07], [ 4.3726e-07, 2.1607e-06, 1.2442e-06, ..., 1.5106e-06, 7.3947e-07, 4.1351e-07], ..., [ 4.8056e-07, -2.2784e-05, -3.7719e-08, ..., 1.0654e-05, 2.2678e-07, 1.1381e-06], [-3.5055e-06, 4.2515e-07, -6.1616e-06, ..., -4.1649e-06, 1.9874e-06, -7.9162e-07], [ 3.3947e-07, 8.7544e-06, -2.5257e-06, ..., 8.8513e-06, 8.6147e-08, -3.6266e-06]], device='cuda:0') Epoch 116, bias, value: tensor([ 0.0032, 0.0042, 0.0068, 0.0132, 0.0267, 0.0359, -0.0262, 0.0090, -0.0387, -0.0193], device='cuda:0'), grad: tensor([-1.9688e-06, 2.5228e-05, 9.2536e-06, 2.4170e-05, -2.9847e-05, -2.1115e-05, 8.6278e-06, -3.7998e-05, -3.3118e-06, 2.6822e-05], device='cuda:0') 100 0.0001 changing lr epoch 115, time 220.37, cls_loss 0.0049 cls_loss_mapping 0.0067 cls_loss_causal 0.5790 re_mapping 0.0086 re_causal 0.0257 /// teacc 98.89 lr 0.00010000 Epoch 117, weight, value: tensor([[-0.0205, 0.1073, 0.0644, ..., -0.1364, -0.0944, 0.0255], [ 0.1006, -0.0094, -0.0435, ..., 0.0581, 0.0840, -0.1040], [-0.0598, -0.0325, -0.0356, ..., -0.0619, -0.0155, -0.0346], ..., [-0.0367, 0.1187, -0.1411, ..., -0.0997, 0.0143, -0.1074], [ 0.0010, -0.0759, 0.0283, ..., 0.0514, -0.1006, -0.0403], [-0.0351, -0.0681, 0.0513, ..., -0.0638, -0.0446, 0.1055]], device='cuda:0'), grad: tensor([[ 1.2899e-06, 1.8813e-06, 1.6149e-06, ..., 7.6555e-07, 6.3330e-08, 1.7863e-06], [ 1.1316e-07, 1.2657e-06, 3.3677e-05, ..., 7.2643e-08, -8.5216e-08, 5.1081e-05], [ 4.1490e-07, -3.3695e-06, 3.2876e-07, ..., 2.3423e-07, 1.4016e-07, 3.5344e-07], ..., [ 8.4937e-07, -2.6915e-06, 6.3851e-06, ..., 8.8476e-07, 6.7987e-08, 5.6848e-06], [-7.9907e-07, 1.8124e-06, 3.5524e-05, ..., 2.6599e-06, 2.9709e-07, 2.8685e-05], [ 3.5111e-07, 1.2890e-06, -1.7154e-04, ..., -6.9402e-06, 2.7940e-08, -2.2888e-04]], device='cuda:0') Epoch 117, bias, value: tensor([ 0.0031, 0.0042, 0.0065, 0.0129, 0.0270, 0.0356, -0.0269, 0.0094, -0.0375, -0.0197], device='cuda:0'), grad: tensor([ 3.2693e-05, 8.6784e-05, -9.1553e-05, 4.7795e-06, 2.0564e-04, 1.2971e-05, -8.1956e-06, 3.1173e-05, 9.4533e-05, -3.6883e-04], device='cuda:0') 100 0.0001 changing lr epoch 116, time 219.06, cls_loss 0.0044 cls_loss_mapping 0.0067 cls_loss_causal 0.5838 re_mapping 0.0092 re_causal 0.0264 /// teacc 98.85 lr 0.00010000 Epoch 118, weight, value: tensor([[-0.0206, 0.1075, 0.0644, ..., -0.1368, -0.0944, 0.0253], [ 0.1008, -0.0088, -0.0438, ..., 0.0581, 0.0844, -0.1047], [-0.0601, -0.0328, -0.0360, ..., -0.0624, -0.0156, -0.0352], ..., [-0.0370, 0.1186, -0.1418, ..., -0.0998, 0.0141, -0.1084], [ 0.0008, -0.0763, 0.0283, ..., 0.0517, -0.1010, -0.0405], [-0.0354, -0.0685, 0.0523, ..., -0.0637, -0.0447, 0.1069]], device='cuda:0'), grad: tensor([[ 3.9227e-06, -2.6375e-06, 6.1728e-06, ..., 6.0201e-06, 1.9278e-07, 2.0731e-06], [-1.3433e-05, 4.2003e-07, 6.1234e-07, ..., -1.9163e-05, -5.6289e-06, 6.6916e-07], [ 4.8913e-06, 1.6624e-07, 2.0210e-07, ..., 6.3516e-06, 2.0210e-06, 1.3178e-06], ..., [ 7.3295e-07, -8.6194e-07, 6.1188e-07, ..., 2.1439e-06, 1.5367e-07, 5.0990e-07], [-3.7313e-05, 1.3178e-07, -4.5598e-05, ..., -4.7117e-05, -6.8024e-06, -3.0354e-05], [ 6.8024e-06, 2.4196e-06, 1.8537e-05, ..., 3.0816e-05, 3.0082e-07, 1.6078e-05]], device='cuda:0') Epoch 118, bias, value: tensor([ 0.0029, 0.0045, 0.0066, 0.0120, 0.0265, 0.0358, -0.0264, 0.0093, -0.0378, -0.0191], device='cuda:0'), grad: tensor([ 1.1124e-05, -3.0816e-05, -2.0891e-05, 2.6301e-05, -2.2948e-05, 3.4757e-06, 6.9439e-05, 3.2410e-06, -1.0979e-04, 7.0810e-05], device='cuda:0') 100 0.0001 changing lr epoch 117, time 219.96, cls_loss 0.0040 cls_loss_mapping 0.0069 cls_loss_causal 0.5664 re_mapping 0.0090 re_causal 0.0264 /// teacc 98.84 lr 0.00010000 Epoch 119, weight, value: tensor([[-0.0214, 0.1078, 0.0646, ..., -0.1386, -0.0951, 0.0253], [ 0.1011, -0.0093, -0.0444, ..., 0.0586, 0.0846, -0.1049], [-0.0604, -0.0331, -0.0365, ..., -0.0628, -0.0157, -0.0357], ..., [-0.0367, 0.1194, -0.1415, ..., -0.1001, 0.0140, -0.1095], [ 0.0004, -0.0769, 0.0281, ..., 0.0516, -0.1012, -0.0407], [-0.0358, -0.0690, 0.0522, ..., -0.0642, -0.0448, 0.1072]], device='cuda:0'), grad: tensor([[ 4.7088e-06, -5.7276e-08, 6.8285e-06, ..., 4.6976e-06, 2.6962e-07, 2.3786e-06], [ 1.9185e-06, 1.0654e-06, 2.5872e-06, ..., 4.6901e-06, 9.4995e-08, 1.0785e-06], [ 3.8557e-06, 3.0175e-07, 1.2524e-05, ..., 2.9504e-06, 2.2203e-06, 9.7230e-07], ..., [ 1.0356e-06, -1.3113e-05, 2.0098e-06, ..., 2.9095e-06, 2.3749e-07, 1.0580e-06], [-2.3395e-05, 1.3746e-06, -2.1607e-05, ..., -2.1994e-05, 2.8033e-07, -3.2373e-06], [ 4.3921e-06, 9.0003e-06, -2.9374e-06, ..., -1.7393e-04, 6.3796e-08, -1.7858e-04]], device='cuda:0') Epoch 119, bias, value: tensor([ 0.0023, 0.0043, 0.0065, 0.0117, 0.0268, 0.0363, -0.0256, 0.0098, -0.0383, -0.0195], device='cuda:0'), grad: tensor([ 1.7047e-05, 1.5959e-05, 8.4162e-05, -8.6129e-05, 6.5088e-04, 4.6007e-06, 1.4141e-05, -1.1981e-05, -3.0056e-05, -6.5851e-04], device='cuda:0') 100 0.0001 changing lr epoch 118, time 219.59, cls_loss 0.0044 cls_loss_mapping 0.0060 cls_loss_causal 0.6092 re_mapping 0.0087 re_causal 0.0259 /// teacc 98.97 lr 0.00010000 Epoch 120, weight, value: tensor([[-0.0220, 0.1085, 0.0647, ..., -0.1402, -0.0952, 0.0252], [ 0.1013, -0.0097, -0.0456, ..., 0.0587, 0.0847, -0.1051], [-0.0608, -0.0335, -0.0372, ..., -0.0634, -0.0158, -0.0360], ..., [-0.0367, 0.1200, -0.1419, ..., -0.1006, 0.0139, -0.1103], [ 0.0017, -0.0760, 0.0298, ..., 0.0529, -0.1012, -0.0400], [-0.0364, -0.0695, 0.0525, ..., -0.0644, -0.0450, 0.1077]], device='cuda:0'), grad: tensor([[ 7.9768e-07, -8.9034e-07, -1.1083e-07, ..., 3.8892e-06, 1.5367e-08, -1.6484e-07], [-2.4214e-06, 1.0140e-05, 4.3958e-07, ..., 1.5097e-06, -1.2666e-07, 2.6403e-07], [ 5.7230e-07, 9.5367e-07, 6.6170e-07, ..., 1.8969e-05, -2.0210e-07, 1.6438e-07], ..., [ 1.2973e-06, -1.8612e-05, 1.1455e-06, ..., 3.4273e-06, 2.3516e-07, 9.2387e-07], [ 7.7710e-06, 9.3691e-07, 6.2473e-06, ..., 1.4640e-05, 3.6834e-07, 7.3351e-06], [ 1.6633e-06, 5.9530e-06, 1.1288e-06, ..., 8.8364e-06, 6.1002e-08, 4.5262e-07]], device='cuda:0') Epoch 120, bias, value: tensor([ 0.0023, 0.0040, 0.0061, 0.0113, 0.0270, 0.0366, -0.0268, 0.0098, -0.0363, -0.0195], device='cuda:0'), grad: tensor([ 5.0962e-06, 2.4304e-05, 2.8253e-05, -8.3819e-06, -5.0020e-04, -2.0623e-05, 4.2486e-04, -2.1994e-05, 3.9548e-05, 2.8834e-05], device='cuda:0') 100 0.0001 changing lr epoch 119, time 219.88, cls_loss 0.0036 cls_loss_mapping 0.0061 cls_loss_causal 0.5716 re_mapping 0.0086 re_causal 0.0256 /// teacc 98.91 lr 0.00010000 Epoch 121, weight, value: tensor([[-0.0222, 0.1091, 0.0647, ..., -0.1409, -0.0953, 0.0250], [ 0.1017, -0.0097, -0.0458, ..., 0.0589, 0.0849, -0.1052], [-0.0611, -0.0338, -0.0374, ..., -0.0638, -0.0160, -0.0362], ..., [-0.0370, 0.1206, -0.1423, ..., -0.1009, 0.0138, -0.1105], [ 0.0016, -0.0769, 0.0295, ..., 0.0528, -0.1015, -0.0402], [-0.0366, -0.0702, 0.0532, ..., -0.0654, -0.0450, 0.1083]], device='cuda:0'), grad: tensor([[ 5.5879e-07, -2.3544e-06, 2.5472e-07, ..., 6.0629e-07, 1.8114e-07, -5.0897e-07], [-1.4314e-06, 1.8720e-07, 2.3730e-06, ..., -2.0154e-06, -1.3690e-07, 1.2619e-07], [ 9.0674e-06, 4.9593e-07, 1.6928e-05, ..., 1.7434e-06, 7.3984e-06, 2.9104e-07], ..., [ 6.4913e-07, -8.3400e-07, 6.8685e-07, ..., 6.5845e-07, 2.4866e-07, 2.9709e-07], [ 1.0114e-06, 2.9849e-07, 2.8126e-06, ..., -1.1977e-06, 1.3402e-06, 5.4529e-07], [ 7.3249e-07, 1.3020e-06, 5.6345e-07, ..., 1.1306e-06, 2.1886e-07, 3.7346e-07]], device='cuda:0') Epoch 121, bias, value: tensor([ 0.0023, 0.0041, 0.0060, 0.0113, 0.0276, 0.0364, -0.0266, 0.0098, -0.0367, -0.0198], device='cuda:0'), grad: tensor([-2.4326e-06, -2.2119e-07, 2.9728e-05, -2.8282e-05, 4.0000e-07, -4.1813e-05, 3.3528e-05, 1.3784e-06, 3.9376e-06, 3.9265e-06], device='cuda:0') 100 0.0001 changing lr epoch 120, time 219.47, cls_loss 0.0043 cls_loss_mapping 0.0082 cls_loss_causal 0.5748 re_mapping 0.0089 re_causal 0.0253 /// teacc 98.91 lr 0.00010000 Epoch 122, weight, value: tensor([[-0.0222, 0.1102, 0.0650, ..., -0.1412, -0.0953, 0.0252], [ 0.1020, -0.0100, -0.0464, ..., 0.0593, 0.0854, -0.1056], [-0.0612, -0.0346, -0.0375, ..., -0.0640, -0.0162, -0.0366], ..., [-0.0371, 0.1225, -0.1424, ..., -0.1018, 0.0133, -0.1120], [ 0.0016, -0.0793, 0.0295, ..., 0.0531, -0.1016, -0.0403], [-0.0369, -0.0713, 0.0539, ..., -0.0652, -0.0450, 0.1094]], device='cuda:0'), grad: tensor([[ 1.9185e-06, -1.8165e-05, -9.3877e-06, ..., 3.2205e-06, 1.2713e-07, -1.0721e-05], [-2.3562e-07, 2.9672e-06, 2.0862e-06, ..., 3.9935e-06, 8.3353e-07, 4.0885e-07], [-2.6952e-06, 8.5458e-06, -3.8594e-05, ..., -7.5391e-07, 8.5309e-07, -3.4392e-05], ..., [ 2.8592e-07, -7.4625e-05, 6.9523e-07, ..., 7.5512e-06, -3.5226e-05, 7.5996e-07], [-8.1165e-07, 6.7532e-05, 3.4217e-06, ..., -1.5143e-06, 2.9087e-05, 2.3879e-06], [ 4.7591e-07, 4.3549e-06, 1.2822e-05, ..., 8.7470e-06, 8.9174e-07, 1.3933e-05]], device='cuda:0') Epoch 122, bias, value: tensor([ 0.0026, 0.0040, 0.0061, 0.0107, 0.0271, 0.0363, -0.0270, 0.0108, -0.0374, -0.0195], device='cuda:0'), grad: tensor([-7.1883e-05, 2.5645e-05, -1.3089e-04, 9.9778e-05, -3.7849e-05, 2.6390e-05, 6.6981e-06, -2.0576e-04, 2.2829e-04, 5.9545e-05], device='cuda:0') 100 0.0001 changing lr epoch 121, time 219.12, cls_loss 0.0038 cls_loss_mapping 0.0070 cls_loss_causal 0.5679 re_mapping 0.0087 re_causal 0.0257 /// teacc 98.96 lr 0.00010000 Epoch 123, weight, value: tensor([[-0.0224, 0.1109, 0.0653, ..., -0.1418, -0.0953, 0.0253], [ 0.1013, -0.0102, -0.0466, ..., 0.0580, 0.0845, -0.1056], [-0.0598, -0.0349, -0.0378, ..., -0.0618, -0.0152, -0.0368], ..., [-0.0372, 0.1234, -0.1433, ..., -0.1021, 0.0132, -0.1125], [ 0.0014, -0.0803, 0.0297, ..., 0.0531, -0.1018, -0.0402], [-0.0375, -0.0724, 0.0540, ..., -0.0657, -0.0451, 0.1096]], device='cuda:0'), grad: tensor([[ 9.2201e-08, -6.8452e-08, 9.3132e-10, ..., 1.3132e-07, 7.9162e-09, 9.3132e-10], [-8.7731e-07, 3.7812e-07, 7.7300e-08, ..., -1.7649e-06, -6.9384e-08, 4.6100e-08], [ 8.2515e-07, 1.5348e-06, 5.6485e-07, ..., 8.9174e-07, -2.4680e-08, 3.9581e-08], ..., [ 5.3504e-07, -4.7609e-06, 2.1933e-07, ..., 8.5123e-07, 8.5216e-08, 1.4342e-07], [-1.8626e-07, 6.3470e-07, -3.8184e-08, ..., -6.6124e-08, -4.3772e-08, 1.2312e-06], [ 2.6776e-07, 1.4426e-06, -3.0780e-07, ..., 6.8359e-07, 1.0245e-08, -3.9907e-07]], device='cuda:0') Epoch 123, bias, value: tensor([ 0.0028, 0.0029, 0.0077, 0.0117, 0.0266, 0.0359, -0.0269, 0.0108, -0.0376, -0.0196], device='cuda:0'), grad: tensor([ 7.6042e-07, 6.1700e-07, -4.9114e-05, 9.7677e-06, 1.2135e-06, -3.4440e-06, 1.2722e-06, 2.8789e-05, 7.8827e-06, 2.2184e-06], device='cuda:0') 100 0.0001 changing lr epoch 122, time 218.99, cls_loss 0.0045 cls_loss_mapping 0.0079 cls_loss_causal 0.5839 re_mapping 0.0086 re_causal 0.0256 /// teacc 98.93 lr 0.00010000 Epoch 124, weight, value: tensor([[-0.0235, 0.1116, 0.0657, ..., -0.1435, -0.0962, 0.0253], [ 0.1023, -0.0084, -0.0471, ..., 0.0592, 0.0847, -0.1060], [-0.0600, -0.0359, -0.0381, ..., -0.0620, -0.0153, -0.0370], ..., [-0.0388, 0.1231, -0.1432, ..., -0.1043, 0.0131, -0.1127], [ 0.0007, -0.0824, 0.0289, ..., 0.0522, -0.1020, -0.0414], [-0.0381, -0.0744, 0.0529, ..., -0.0665, -0.0455, 0.1091]], device='cuda:0'), grad: tensor([[ 1.4640e-06, -1.6652e-06, -6.6031e-07, ..., 1.6252e-06, -3.9581e-08, -1.5358e-06], [-3.1758e-06, 1.3374e-06, 1.1846e-06, ..., -3.8780e-06, -1.3765e-06, 1.1288e-06], [ 2.9393e-06, 7.5018e-07, 1.3672e-06, ..., 3.3211e-06, 1.5739e-07, 3.0873e-07], ..., [ 1.9222e-06, -7.5288e-06, 5.9642e-06, ..., 3.3937e-06, 4.7637e-07, 6.3218e-06], [-1.1757e-05, 8.1863e-07, -6.3404e-06, ..., -1.3158e-05, 2.3935e-07, 1.0626e-06], [ 8.3912e-07, 3.4235e-06, -1.5721e-05, ..., 1.0453e-05, 1.2992e-07, -1.7345e-05]], device='cuda:0') Epoch 124, bias, value: tensor([ 0.0029, 0.0037, 0.0077, 0.0119, 0.0265, 0.0369, -0.0265, 0.0105, -0.0386, -0.0205], device='cuda:0'), grad: tensor([ 1.7388e-06, -8.1863e-07, 3.5800e-06, 5.1767e-05, -6.8545e-06, -3.5882e-05, 1.4342e-05, 6.0797e-06, -2.8431e-05, -5.4576e-06], device='cuda:0') 100 0.0001 changing lr epoch 123, time 218.59, cls_loss 0.0033 cls_loss_mapping 0.0055 cls_loss_causal 0.5471 re_mapping 0.0088 re_causal 0.0248 /// teacc 98.92 lr 0.00010000 Epoch 125, weight, value: tensor([[-0.0238, 0.1118, 0.0659, ..., -0.1442, -0.0961, 0.0254], [ 0.1025, -0.0085, -0.0475, ..., 0.0595, 0.0848, -0.1064], [-0.0603, -0.0359, -0.0383, ..., -0.0623, -0.0152, -0.0371], ..., [-0.0391, 0.1236, -0.1438, ..., -0.1047, 0.0129, -0.1131], [ 0.0009, -0.0827, 0.0290, ..., 0.0525, -0.1026, -0.0414], [-0.0384, -0.0754, 0.0531, ..., -0.0675, -0.0458, 0.1094]], device='cuda:0'), grad: tensor([[ 1.4948e-07, -3.9535e-07, -2.5425e-07, ..., 1.6205e-07, 2.7940e-08, -4.2655e-07], [-3.2783e-05, 4.5123e-07, 3.6508e-07, ..., -5.5999e-05, -3.1233e-05, 3.1292e-07], [ 1.5482e-05, 1.0291e-07, 1.2945e-06, ..., 2.5585e-05, 1.4320e-05, 5.2154e-08], ..., [ 1.7107e-05, -2.0266e-06, 6.2473e-06, ..., 3.4153e-05, 1.5423e-05, 1.5810e-05], [ 5.0925e-06, 2.6310e-07, 1.1683e-05, ..., 2.7865e-06, 1.3271e-07, 4.4703e-06], [ 2.8545e-07, 6.1374e-07, -1.2413e-05, ..., -2.7522e-05, 6.4261e-08, -7.6711e-05]], device='cuda:0') Epoch 125, bias, value: tensor([ 0.0029, 0.0037, 0.0079, 0.0116, 0.0269, 0.0372, -0.0265, 0.0106, -0.0386, -0.0211], device='cuda:0'), grad: tensor([-5.3318e-07, -5.4687e-05, 2.8312e-05, -3.1829e-05, 1.0687e-04, 5.9977e-06, -1.3318e-07, 5.7817e-05, 3.3587e-05, -1.4532e-04], device='cuda:0') 100 0.0001 changing lr epoch 124, time 218.14, cls_loss 0.0042 cls_loss_mapping 0.0059 cls_loss_causal 0.5615 re_mapping 0.0086 re_causal 0.0247 /// teacc 98.88 lr 0.00010000 Epoch 126, weight, value: tensor([[-0.0246, 0.1120, 0.0663, ..., -0.1458, -0.0966, 0.0256], [ 0.1020, -0.0086, -0.0486, ..., 0.0591, 0.0847, -0.1078], [-0.0594, -0.0363, -0.0387, ..., -0.0618, -0.0149, -0.0373], ..., [-0.0393, 0.1240, -0.1443, ..., -0.1048, 0.0127, -0.1141], [ 0.0024, -0.0824, 0.0283, ..., 0.0539, -0.1030, -0.0423], [-0.0389, -0.0757, 0.0538, ..., -0.0676, -0.0455, 0.1104]], device='cuda:0'), grad: tensor([[ 3.8091e-07, -6.9337e-07, -4.3539e-07, ..., 6.2538e-07, 5.7742e-08, -7.7533e-07], [ 1.8135e-05, 2.9756e-07, 5.1083e-07, ..., 3.5077e-05, 5.0291e-06, 1.6764e-08], [ 1.3262e-06, 2.4773e-07, -5.3318e-07, ..., 2.0117e-06, 1.4529e-07, 5.4948e-08], ..., [ 4.6045e-06, 3.3528e-06, 2.3156e-05, ..., 7.1758e-07, 8.5682e-08, 8.6147e-08], [-2.0638e-05, 1.2433e-07, 1.2824e-06, ..., -4.0710e-05, -5.7295e-06, 8.0094e-08], [ 5.6485e-07, 9.6858e-07, 2.2668e-06, ..., 2.2352e-07, 2.3283e-08, 3.1991e-07]], device='cuda:0') Epoch 126, bias, value: tensor([ 0.0029, 0.0026, 0.0092, 0.0119, 0.0269, 0.0372, -0.0279, 0.0107, -0.0378, -0.0209], device='cuda:0'), grad: tensor([-3.0827e-07, 4.6611e-05, 2.5611e-07, -5.9783e-05, 1.5385e-06, 5.9158e-06, -1.0626e-06, 4.9382e-05, -4.8250e-05, 5.8077e-06], device='cuda:0') 100 0.0001 changing lr epoch 125, time 218.17, cls_loss 0.0039 cls_loss_mapping 0.0057 cls_loss_causal 0.5538 re_mapping 0.0086 re_causal 0.0249 /// teacc 98.92 lr 0.00010000 Epoch 127, weight, value: tensor([[-0.0251, 0.1112, 0.0653, ..., -0.1468, -0.0969, 0.0242], [ 0.1028, -0.0088, -0.0487, ..., 0.0596, 0.0855, -0.1080], [-0.0600, -0.0372, -0.0391, ..., -0.0623, -0.0155, -0.0375], ..., [-0.0396, 0.1247, -0.1449, ..., -0.1051, 0.0123, -0.1148], [ 0.0022, -0.0826, 0.0282, ..., 0.0538, -0.1034, -0.0424], [-0.0396, -0.0753, 0.0548, ..., -0.0678, -0.0457, 0.1119]], device='cuda:0'), grad: tensor([[ 2.4140e-06, -1.9193e-05, -2.0206e-05, ..., 1.9092e-06, 2.7940e-09, -2.2322e-05], [ 6.9849e-08, 1.5851e-06, 7.2550e-07, ..., -2.2305e-07, -6.8452e-08, 1.0151e-06], [ 8.1304e-07, -8.3074e-06, 7.4506e-07, ..., 6.1430e-06, 2.1886e-08, 6.4820e-07], ..., [ 4.6426e-07, 6.7279e-06, 1.1493e-06, ..., 7.0594e-07, 4.3306e-08, 9.2611e-06], [ 7.6741e-07, 2.0415e-06, 2.8722e-06, ..., 7.8091e-07, 8.7079e-08, 2.0023e-06], [ 7.0874e-07, 1.6332e-05, 1.3016e-05, ..., 1.6745e-06, 1.2107e-08, 2.1840e-07]], device='cuda:0') Epoch 127, bias, value: tensor([ 0.0012, 0.0029, 0.0087, 0.0120, 0.0266, 0.0376, -0.0278, 0.0109, -0.0381, -0.0198], device='cuda:0'), grad: tensor([-7.2718e-05, 8.9705e-06, -9.2268e-05, -2.0079e-06, 4.1455e-05, 6.0052e-06, -3.6448e-05, 1.2124e-04, 9.4771e-06, 1.6317e-05], device='cuda:0') 100 0.0001 changing lr epoch 126, time 218.89, cls_loss 0.0037 cls_loss_mapping 0.0046 cls_loss_causal 0.5610 re_mapping 0.0089 re_causal 0.0255 /// teacc 98.86 lr 0.00010000 Epoch 128, weight, value: tensor([[-0.0249, 0.1119, 0.0663, ..., -0.1465, -0.0972, 0.0244], [ 0.1024, -0.0090, -0.0508, ..., 0.0585, 0.0857, -0.1093], [-0.0600, -0.0380, -0.0376, ..., -0.0613, -0.0155, -0.0375], ..., [-0.0398, 0.1255, -0.1445, ..., -0.1053, 0.0121, -0.1138], [ 0.0020, -0.0832, 0.0280, ..., 0.0536, -0.1038, -0.0428], [-0.0400, -0.0764, 0.0542, ..., -0.0683, -0.0458, 0.1118]], device='cuda:0'), grad: tensor([[-3.9302e-06, -7.8261e-05, -1.0943e-04, ..., -4.2468e-05, 4.6566e-10, -1.0145e-04], [ 3.1805e-07, 8.9360e-07, 1.0598e-06, ..., 2.1644e-06, -1.3970e-09, 1.0319e-06], [ 4.7078e-07, 6.8638e-07, -6.9011e-07, ..., 2.1514e-06, 9.3132e-10, 3.5297e-07], ..., [ 1.5041e-06, 2.5854e-05, 2.9709e-06, ..., 8.7917e-05, 1.8626e-09, 3.1572e-06], [ 6.0320e-05, 1.9029e-05, 1.9506e-05, ..., 6.1929e-05, 1.3039e-08, 3.7670e-05], [ 1.0662e-05, 5.3495e-05, 5.6088e-05, ..., 7.2002e-05, 0.0000e+00, 4.7833e-05]], device='cuda:0') Epoch 128, bias, value: tensor([ 0.0016, 0.0020, 0.0094, 0.0121, 0.0262, 0.0380, -0.0275, 0.0118, -0.0384, -0.0206], device='cuda:0'), grad: tensor([-2.0099e-04, 6.1840e-06, 1.5507e-06, 1.3866e-05, -3.1114e-04, -3.9554e-04, 3.7408e-04, 2.0599e-04, 9.9599e-05, 2.0742e-04], device='cuda:0') 100 0.0001 changing lr epoch 127, time 217.69, cls_loss 0.0038 cls_loss_mapping 0.0059 cls_loss_causal 0.5678 re_mapping 0.0080 re_causal 0.0242 /// teacc 98.82 lr 0.00010000 Epoch 129, weight, value: tensor([[-0.0255, 0.1126, 0.0667, ..., -0.1473, -0.0978, 0.0246], [ 0.1028, -0.0096, -0.0518, ..., 0.0590, 0.0860, -0.1096], [-0.0605, -0.0388, -0.0374, ..., -0.0618, -0.0157, -0.0379], ..., [-0.0399, 0.1263, -0.1450, ..., -0.1056, 0.0121, -0.1144], [ 0.0027, -0.0821, 0.0295, ..., 0.0548, -0.1052, -0.0420], [-0.0412, -0.0770, 0.0542, ..., -0.0691, -0.0459, 0.1118]], device='cuda:0'), grad: tensor([[ 1.8510e-07, -3.0873e-07, -1.2550e-07, ..., 3.6694e-07, 7.9721e-07, -4.1444e-08], [-1.3009e-05, 2.4904e-06, -2.5844e-08, ..., -2.1622e-05, 3.7672e-07, 1.3690e-07], [ 2.2110e-06, 6.6590e-07, 1.3527e-07, ..., 3.7793e-06, -1.2226e-05, 3.5623e-08], ..., [ 1.6429e-06, -5.6103e-06, 6.4913e-07, ..., 1.7891e-06, 1.3225e-06, 8.8476e-07], [ 1.2584e-05, 3.0990e-07, 4.8764e-06, ..., 2.3305e-05, 3.6019e-07, 1.0192e-05], [ 3.0990e-07, 1.3523e-06, -2.1309e-06, ..., 1.6671e-06, 3.8906e-07, -3.0492e-06]], device='cuda:0') Epoch 129, bias, value: tensor([ 0.0018, 0.0019, 0.0093, 0.0118, 0.0262, 0.0380, -0.0278, 0.0120, -0.0378, -0.0208], device='cuda:0'), grad: tensor([ 1.7077e-05, -3.5074e-06, -2.6202e-04, 1.8990e-04, 3.2596e-06, -1.7792e-05, 8.9258e-06, 2.4125e-05, 3.1859e-05, 8.0690e-06], device='cuda:0') 100 0.0001 changing lr epoch 128, time 216.95, cls_loss 0.0045 cls_loss_mapping 0.0078 cls_loss_causal 0.5343 re_mapping 0.0086 re_causal 0.0235 /// teacc 98.98 lr 0.00010000 Epoch 130, weight, value: tensor([[-0.0260, 0.1134, 0.0674, ..., -0.1482, -0.0981, 0.0251], [ 0.1046, -0.0096, -0.0511, ..., 0.0607, 0.0884, -0.1087], [-0.0612, -0.0397, -0.0375, ..., -0.0624, -0.0172, -0.0394], ..., [-0.0409, 0.1270, -0.1454, ..., -0.1066, 0.0104, -0.1151], [ 0.0023, -0.0817, 0.0295, ..., 0.0542, -0.1056, -0.0423], [-0.0425, -0.0781, 0.0548, ..., -0.0673, -0.0462, 0.1134]], device='cuda:0'), grad: tensor([[ 7.7067e-08, -4.2953e-06, -2.7511e-06, ..., 2.3469e-07, 1.6904e-07, -4.8503e-06], [-2.1188e-08, 2.9020e-06, 9.4064e-07, ..., 1.2480e-07, 4.4308e-07, 2.9965e-07], [ 2.0373e-07, 7.8827e-06, 3.8370e-06, ..., 8.6054e-07, 1.2387e-06, 8.7544e-07], ..., [ 4.2259e-07, -2.8282e-05, 8.1211e-06, ..., 2.1756e-06, 4.9220e-07, -2.7083e-06], [ 6.1374e-07, 7.1479e-07, 2.3879e-06, ..., 1.1232e-06, 8.3260e-07, 9.2667e-07], [ 1.1967e-07, 8.9109e-06, 2.9784e-06, ..., 2.7902e-06, 2.0228e-06, 5.1335e-06]], device='cuda:0') Epoch 130, bias, value: tensor([ 0.0022, 0.0031, 0.0089, 0.0114, 0.0240, 0.0379, -0.0271, 0.0118, -0.0387, -0.0199], device='cuda:0'), grad: tensor([-9.1046e-06, 4.5240e-05, -2.1458e-05, -3.5763e-05, 4.1425e-06, 1.0140e-05, 2.7083e-06, -3.0190e-05, 7.5623e-06, 2.6688e-05], device='cuda:0') 100 0.0001 changing lr epoch 129, time 217.09, cls_loss 0.0030 cls_loss_mapping 0.0053 cls_loss_causal 0.6039 re_mapping 0.0082 re_causal 0.0251 /// teacc 98.94 lr 0.00010000 Epoch 131, weight, value: tensor([[-0.0270, 0.1138, 0.0676, ..., -0.1489, -0.0989, 0.0252], [ 0.1051, -0.0099, -0.0512, ..., 0.0612, 0.0889, -0.1086], [-0.0615, -0.0404, -0.0375, ..., -0.0625, -0.0173, -0.0399], ..., [-0.0410, 0.1276, -0.1462, ..., -0.1067, 0.0102, -0.1163], [ 0.0020, -0.0824, 0.0292, ..., 0.0539, -0.1058, -0.0426], [-0.0430, -0.0785, 0.0551, ..., -0.0675, -0.0464, 0.1138]], device='cuda:0'), grad: tensor([[ 2.7511e-06, -1.0068e-06, -1.2224e-07, ..., 3.1292e-06, 3.4692e-08, -3.9674e-07], [-1.3188e-05, 9.6392e-07, 2.8014e-05, ..., 3.2205e-06, -3.7462e-05, 2.9713e-05], [ 2.0698e-05, 5.6252e-07, 1.6112e-06, ..., 1.7837e-05, 3.2395e-05, 8.9966e-07], ..., [ 3.9004e-06, -3.4571e-06, 1.6931e-06, ..., 4.2245e-06, 4.8839e-06, 1.3681e-06], [ 1.9848e-04, 2.0908e-07, 1.3411e-06, ..., 2.1315e-04, 1.8883e-07, 3.9011e-05], [-1.8299e-05, 1.6401e-06, -6.7651e-05, ..., -4.7237e-05, 2.2328e-07, -7.1585e-05]], device='cuda:0') Epoch 131, bias, value: tensor([ 0.0021, 0.0031, 0.0088, 0.0116, 0.0239, 0.0382, -0.0269, 0.0120, -0.0392, -0.0200], device='cuda:0'), grad: tensor([ 4.2766e-06, 1.9580e-05, 5.0694e-05, -1.3374e-05, 9.0659e-05, 4.9710e-05, -4.8566e-04, 8.1733e-06, 4.2439e-04, -1.4806e-04], device='cuda:0') 100 0.0001 changing lr epoch 130, time 217.19, cls_loss 0.0041 cls_loss_mapping 0.0070 cls_loss_causal 0.5723 re_mapping 0.0082 re_causal 0.0247 /// teacc 98.85 lr 0.00010000 Epoch 132, weight, value: tensor([[-0.0271, 0.1122, 0.0667, ..., -0.1495, -0.0990, 0.0231], [ 0.1055, -0.0104, -0.0514, ..., 0.0614, 0.0894, -0.1091], [-0.0619, -0.0410, -0.0384, ..., -0.0628, -0.0177, -0.0406], ..., [-0.0411, 0.1283, -0.1466, ..., -0.1066, 0.0098, -0.1173], [ 0.0025, -0.0828, 0.0294, ..., 0.0543, -0.1060, -0.0426], [-0.0435, -0.0772, 0.0564, ..., -0.0678, -0.0466, 0.1163]], device='cuda:0'), grad: tensor([[ 5.5227e-07, -7.4785e-07, 2.5239e-07, ..., 7.8557e-07, 1.2107e-08, -2.4633e-07], [-4.1090e-06, 6.9709e-07, 3.8510e-07, ..., -9.4026e-06, -1.5199e-06, 3.4040e-07], [ 2.1886e-06, 2.4540e-07, 5.0571e-07, ..., 4.3884e-06, 8.1258e-07, 2.5565e-07], ..., [ 8.0094e-07, -2.2799e-06, 1.7285e-06, ..., 1.3858e-06, 1.8207e-07, 2.5909e-06], [ 4.3027e-07, 3.3341e-07, -4.3958e-07, ..., 9.5926e-08, 1.5413e-07, 5.4240e-06], [ 2.1607e-06, 7.6648e-07, 1.9670e-06, ..., 2.8424e-06, 4.6566e-08, 9.3598e-07]], device='cuda:0') Epoch 132, bias, value: tensor([ 8.0441e-05, 2.7230e-03, 8.1190e-03, 1.2262e-02, 2.3700e-02, 3.8310e-02, -2.7228e-02, 1.2495e-02, -3.8886e-02, -1.8500e-02], device='cuda:0'), grad: tensor([ 3.1106e-06, -7.9423e-06, -1.9789e-05, 3.1263e-05, 3.1777e-06, -3.4034e-05, 6.0871e-06, 9.4548e-06, 3.8035e-06, 4.8205e-06], device='cuda:0') 100 0.0001 changing lr epoch 131, time 217.26, cls_loss 0.0040 cls_loss_mapping 0.0060 cls_loss_causal 0.5742 re_mapping 0.0083 re_causal 0.0240 /// teacc 98.91 lr 0.00010000 Epoch 133, weight, value: tensor([[-0.0264, 0.1138, 0.0674, ..., -0.1497, -0.0995, 0.0236], [ 0.1060, -0.0107, -0.0514, ..., 0.0621, 0.0898, -0.1087], [-0.0621, -0.0439, -0.0411, ..., -0.0638, -0.0181, -0.0442], ..., [-0.0413, 0.1290, -0.1472, ..., -0.1072, 0.0095, -0.1181], [ 0.0020, -0.0833, 0.0286, ..., 0.0539, -0.1067, -0.0436], [-0.0440, -0.0780, 0.0565, ..., -0.0684, -0.0473, 0.1163]], device='cuda:0'), grad: tensor([[ 7.2643e-08, -5.7191e-05, -2.0385e-05, ..., 2.1281e-07, 3.7253e-09, -8.0228e-05], [-3.1590e-06, 2.8312e-07, 1.9046e-07, ..., -2.7977e-06, -4.7823e-07, 2.8312e-07], [ 7.0082e-07, 2.1085e-06, 7.2876e-07, ..., 1.3653e-06, 8.8476e-09, 2.7698e-06], ..., [ 1.2852e-06, 1.6570e-05, 6.2510e-06, ..., 9.7007e-06, 1.1874e-07, 2.4512e-05], [ 1.3644e-07, 3.7178e-06, 1.9241e-06, ..., 3.0547e-07, 1.1642e-08, 6.2175e-06], [ 7.6881e-07, 3.1721e-06, -3.0287e-06, ..., 1.3381e-05, 2.9989e-07, -2.9709e-06]], device='cuda:0') Epoch 133, bias, value: tensor([ 0.0010, 0.0027, 0.0065, 0.0127, 0.0241, 0.0400, -0.0275, 0.0125, -0.0397, -0.0189], device='cuda:0'), grad: tensor([-1.7560e-04, -3.2764e-06, 8.3372e-06, 8.3372e-06, -4.1604e-05, 6.7353e-05, 2.6450e-05, 7.0333e-05, 1.3143e-05, 2.6584e-05], device='cuda:0') 100 0.0001 changing lr epoch 132, time 217.19, cls_loss 0.0032 cls_loss_mapping 0.0061 cls_loss_causal 0.5685 re_mapping 0.0084 re_causal 0.0237 /// teacc 98.97 lr 0.00010000 Epoch 134, weight, value: tensor([[-0.0269, 0.1142, 0.0676, ..., -0.1506, -0.0996, 0.0238], [ 0.1065, -0.0108, -0.0515, ..., 0.0628, 0.0902, -0.1089], [-0.0623, -0.0447, -0.0413, ..., -0.0643, -0.0181, -0.0448], ..., [-0.0417, 0.1296, -0.1478, ..., -0.1078, 0.0087, -0.1190], [ 0.0020, -0.0835, 0.0289, ..., 0.0539, -0.1067, -0.0437], [-0.0445, -0.0784, 0.0566, ..., -0.0693, -0.0475, 0.1164]], device='cuda:0'), grad: tensor([[ 2.7800e-07, -3.4068e-06, -5.1893e-06, ..., 6.6590e-07, 3.6741e-07, -9.1642e-06], [-7.6771e-05, 3.1479e-06, 1.1409e-07, ..., -1.9920e-04, -1.6487e-04, 1.4855e-07], [ 6.8665e-05, 6.7940e-07, 2.0256e-07, ..., 1.8036e-04, 1.4806e-04, 2.1188e-07], ..., [ 6.3740e-06, -6.1169e-06, 2.3469e-07, ..., 1.3106e-05, 1.2286e-05, 3.2922e-07], [ 4.1910e-09, 1.2089e-06, -1.4016e-07, ..., 3.3947e-07, 6.1281e-07, 3.3621e-07], [ 2.4913e-07, 1.6131e-06, 2.2724e-06, ..., 6.8359e-07, 3.8277e-07, 3.6061e-06]], device='cuda:0') Epoch 134, bias, value: tensor([ 0.0011, 0.0026, 0.0064, 0.0122, 0.0246, 0.0400, -0.0271, 0.0126, -0.0396, -0.0193], device='cuda:0'), grad: tensor([-1.2517e-05, -5.7459e-04, 5.2214e-04, 6.5453e-06, 4.6901e-06, 1.7080e-06, 7.1190e-06, 3.3140e-05, 3.5297e-06, 7.4692e-06], device='cuda:0') 100 0.0001 changing lr epoch 133, time 217.92, cls_loss 0.0039 cls_loss_mapping 0.0068 cls_loss_causal 0.5475 re_mapping 0.0081 re_causal 0.0236 /// teacc 98.95 lr 0.00010000 Epoch 135, weight, value: tensor([[-0.0271, 0.1147, 0.0681, ..., -0.1512, -0.0998, 0.0241], [ 0.1084, -0.0106, -0.0490, ..., 0.0650, 0.0907, -0.1066], [-0.0629, -0.0450, -0.0414, ..., -0.0653, -0.0188, -0.0456], ..., [-0.0425, 0.1295, -0.1498, ..., -0.1092, 0.0091, -0.1195], [ 0.0015, -0.0841, 0.0282, ..., 0.0539, -0.1060, -0.0444], [-0.0470, -0.0790, 0.0560, ..., -0.0702, -0.0481, 0.1162]], device='cuda:0'), grad: tensor([[ 6.0815e-07, 4.6380e-07, 4.5868e-07, ..., 1.0924e-06, 4.0233e-07, 9.8534e-07], [ 1.8720e-07, 8.4043e-06, 1.2685e-06, ..., 7.6974e-07, 3.7067e-07, 2.3693e-06], [ 2.9663e-07, 3.9876e-05, -3.9116e-07, ..., 3.0734e-07, -1.4538e-06, 2.1327e-07], ..., [ 8.8150e-07, -7.5877e-05, -4.6240e-07, ..., 1.3057e-06, -9.4855e-07, -3.4785e-07], [ 3.4094e-04, 9.0718e-05, 3.1233e-04, ..., 4.1032e-04, 8.3447e-07, 3.8600e-04], [ 1.0291e-06, 1.1154e-05, 3.7905e-07, ..., 4.3809e-06, 1.1288e-06, -3.1246e-07]], device='cuda:0') Epoch 135, bias, value: tensor([ 0.0014, 0.0038, 0.0061, 0.0114, 0.0244, 0.0404, -0.0265, 0.0121, -0.0396, -0.0200], device='cuda:0'), grad: tensor([ 5.9083e-06, 2.9042e-05, 1.3523e-05, 5.7578e-05, -5.3011e-06, -8.1348e-04, 7.4744e-05, -1.4639e-04, 7.5722e-04, 2.7373e-05], device='cuda:0') 100 0.0001 changing lr epoch 134, time 217.31, cls_loss 0.0027 cls_loss_mapping 0.0050 cls_loss_causal 0.5486 re_mapping 0.0082 re_causal 0.0238 /// teacc 98.84 lr 0.00010000 Epoch 136, weight, value: tensor([[-0.0276, 0.1150, 0.0680, ..., -0.1521, -0.1000, 0.0241], [ 0.1089, -0.0106, -0.0495, ..., 0.0652, 0.0910, -0.1075], [-0.0632, -0.0452, -0.0412, ..., -0.0655, -0.0188, -0.0460], ..., [-0.0428, 0.1297, -0.1508, ..., -0.1099, 0.0091, -0.1215], [ 0.0015, -0.0847, 0.0281, ..., 0.0539, -0.1065, -0.0449], [-0.0471, -0.0791, 0.0570, ..., -0.0702, -0.0484, 0.1173]], device='cuda:0'), grad: tensor([[ 4.0978e-06, -7.4739e-07, 2.0534e-05, ..., 1.7658e-05, 9.7789e-09, 1.6615e-05], [-4.5542e-07, 1.8347e-07, 1.2089e-06, ..., 1.3597e-07, -2.4028e-07, 1.3355e-06], [ 4.4703e-07, 7.4040e-08, 1.4529e-07, ..., 1.3169e-06, 1.1642e-08, 1.1474e-06], ..., [ 3.5111e-07, -1.1697e-06, 6.8471e-06, ..., 3.2876e-06, 8.8941e-08, 1.2293e-05], [-6.3956e-05, 1.6158e-07, -3.4881e-04, ..., -2.8896e-04, 4.0047e-08, -2.7919e-04], [ 1.6943e-05, 1.2396e-06, 7.4089e-05, ..., 6.9022e-05, 6.2399e-08, 4.1902e-05]], device='cuda:0') Epoch 136, bias, value: tensor([ 0.0013, 0.0038, 0.0064, 0.0109, 0.0245, 0.0403, -0.0265, 0.0116, -0.0400, -0.0193], device='cuda:0'), grad: tensor([ 3.9220e-05, 2.0694e-06, -1.0081e-05, 1.2226e-05, 2.0802e-05, 4.1747e-04, 2.3201e-05, 2.1100e-05, -6.4754e-04, 1.2147e-04], device='cuda:0') 100 0.0001 changing lr epoch 135, time 217.41, cls_loss 0.0034 cls_loss_mapping 0.0045 cls_loss_causal 0.5399 re_mapping 0.0085 re_causal 0.0240 /// teacc 98.85 lr 0.00010000 Epoch 137, weight, value: tensor([[-0.0278, 0.1151, 0.0682, ..., -0.1528, -0.1000, 0.0242], [ 0.1094, -0.0107, -0.0505, ..., 0.0654, 0.0911, -0.1078], [-0.0635, -0.0455, -0.0416, ..., -0.0658, -0.0190, -0.0461], ..., [-0.0435, 0.1293, -0.1533, ..., -0.1107, 0.0096, -0.1208], [ 0.0014, -0.0851, 0.0280, ..., 0.0540, -0.1069, -0.0451], [-0.0476, -0.0782, 0.0589, ..., -0.0705, -0.0497, 0.1175]], device='cuda:0'), grad: tensor([[ 1.9401e-05, 3.9525e-06, 9.4697e-06, ..., 7.3910e-06, 2.0489e-08, 6.6757e-06], [-2.6226e-04, -7.3314e-05, -1.0854e-04, ..., -1.0622e-04, -1.2062e-05, -7.1883e-05], [ 4.3325e-06, 9.7416e-07, 2.1271e-06, ..., 1.8841e-06, 8.1956e-08, 1.3188e-06], ..., [ 6.6042e-05, 2.6807e-05, 1.6123e-05, ..., 3.3021e-05, 1.0431e-05, 1.1280e-05], [ 1.2022e-04, 2.5049e-05, 5.9277e-05, ..., 4.5061e-05, 4.9872e-07, 4.2379e-05], [ 4.2558e-05, 1.3739e-05, 5.3458e-07, ..., 1.8045e-05, 5.3877e-07, -1.8030e-05]], device='cuda:0') Epoch 137, bias, value: tensor([ 0.0013, 0.0034, 0.0062, 0.0118, 0.0247, 0.0398, -0.0266, 0.0109, -0.0401, -0.0183], device='cuda:0'), grad: tensor([ 3.9309e-05, -5.4216e-04, 9.8124e-06, 7.9051e-06, -4.8280e-06, 3.0160e-05, 1.0924e-06, 1.4102e-04, 2.4533e-04, 7.2002e-05], device='cuda:0') 100 0.0001 changing lr epoch 136, time 217.30, cls_loss 0.0032 cls_loss_mapping 0.0055 cls_loss_causal 0.5481 re_mapping 0.0075 re_causal 0.0227 /// teacc 98.98 lr 0.00010000 Epoch 138, weight, value: tensor([[-0.0283, 0.1154, 0.0684, ..., -0.1536, -0.0999, 0.0243], [ 0.1101, -0.0110, -0.0509, ..., 0.0656, 0.0920, -0.1081], [-0.0636, -0.0457, -0.0403, ..., -0.0659, -0.0191, -0.0448], ..., [-0.0439, 0.1299, -0.1545, ..., -0.1114, 0.0086, -0.1222], [ 0.0013, -0.0854, 0.0278, ..., 0.0541, -0.1069, -0.0451], [-0.0485, -0.0786, 0.0592, ..., -0.0707, -0.0503, 0.1181]], device='cuda:0'), grad: tensor([[ 3.8464e-07, 4.1910e-08, 1.8114e-07, ..., 3.6974e-07, -1.1409e-07, -2.4028e-07], [-1.1809e-05, 2.0996e-05, 1.8226e-06, ..., 3.8631e-06, -3.3379e-06, 6.7195e-07], [ 2.3190e-06, 4.0531e-06, 3.3826e-06, ..., 7.6741e-07, 8.5961e-07, 2.8592e-07], ..., [ 1.2279e-05, -2.7850e-05, 3.0398e-06, ..., 1.6317e-05, 1.4994e-06, 7.9116e-07], [-1.4268e-05, 9.7323e-08, -7.6145e-06, ..., -2.4617e-05, 1.9837e-07, -2.0221e-05], [ 1.7092e-05, 7.4320e-07, 2.1309e-05, ..., 2.9400e-05, 5.0850e-07, 1.3776e-05]], device='cuda:0') Epoch 138, bias, value: tensor([ 0.0013, 0.0034, 0.0070, 0.0120, 0.0248, 0.0400, -0.0266, 0.0104, -0.0406, -0.0183], device='cuda:0'), grad: tensor([ 1.1725e-06, 3.8117e-05, 1.6496e-05, -7.5340e-05, -5.6535e-05, 1.9744e-05, 6.1169e-06, -7.5214e-06, -2.1741e-05, 7.9334e-05], device='cuda:0') 100 0.0001 changing lr epoch 137, time 217.11, cls_loss 0.0046 cls_loss_mapping 0.0070 cls_loss_causal 0.5767 re_mapping 0.0080 re_causal 0.0235 /// teacc 98.92 lr 0.00010000 Epoch 139, weight, value: tensor([[-0.0285, 0.1155, 0.0694, ..., -0.1560, -0.1001, 0.0245], [ 0.1112, -0.0112, -0.0504, ..., 0.0665, 0.0922, -0.1077], [-0.0640, -0.0462, -0.0407, ..., -0.0664, -0.0188, -0.0451], ..., [-0.0445, 0.1311, -0.1542, ..., -0.1121, 0.0085, -0.1219], [ 0.0034, -0.0856, 0.0331, ..., 0.0560, -0.1063, -0.0402], [-0.0522, -0.0799, 0.0554, ..., -0.0707, -0.0521, 0.1158]], device='cuda:0'), grad: tensor([[ 6.5677e-06, 1.0245e-06, 5.8785e-06, ..., 8.3148e-06, 1.0245e-08, 4.3958e-06], [ 3.1898e-07, 8.8988e-07, 1.3988e-06, ..., 1.4454e-06, -1.3737e-07, 1.2890e-06], [ 9.3551e-07, 4.1118e-07, 1.3877e-06, ..., 1.3793e-06, 7.4971e-08, 7.7765e-07], ..., [ 6.8313e-07, -4.7199e-06, 7.7710e-06, ..., 3.1628e-06, 1.3132e-07, 1.1645e-05], [ 4.1444e-07, 1.7043e-07, -4.9025e-06, ..., -4.2059e-06, 6.0862e-07, -2.7139e-06], [ 4.7637e-07, 3.4496e-06, -1.2487e-05, ..., 2.6934e-06, 5.1223e-08, -1.9103e-05]], device='cuda:0') Epoch 139, bias, value: tensor([ 0.0015, 0.0036, 0.0071, 0.0119, 0.0229, 0.0400, -0.0274, 0.0108, -0.0362, -0.0208], device='cuda:0'), grad: tensor([ 3.5256e-05, 7.7784e-06, -7.5400e-05, -2.4419e-06, 3.5554e-05, 1.8522e-05, -1.2554e-05, 2.8431e-05, -2.8927e-06, -3.2395e-05], device='cuda:0') 100 0.0001 changing lr epoch 138, time 217.23, cls_loss 0.0034 cls_loss_mapping 0.0063 cls_loss_causal 0.5555 re_mapping 0.0078 re_causal 0.0228 /// teacc 98.81 lr 0.00010000 Epoch 140, weight, value: tensor([[-0.0288, 0.1160, 0.0697, ..., -0.1570, -0.1003, 0.0245], [ 0.1116, -0.0116, -0.0495, ..., 0.0662, 0.0936, -0.1074], [-0.0643, -0.0465, -0.0409, ..., -0.0664, -0.0189, -0.0453], ..., [-0.0447, 0.1324, -0.1533, ..., -0.1119, 0.0064, -0.1214], [ 0.0033, -0.0865, 0.0331, ..., 0.0560, -0.1068, -0.0402], [-0.0536, -0.0812, 0.0551, ..., -0.0708, -0.0522, 0.1159]], device='cuda:0'), grad: tensor([[ 1.7136e-07, -1.4171e-05, -3.5971e-05, ..., -6.3218e-06, 2.5611e-08, -2.3663e-05], [-3.6089e-07, 9.4390e-07, 1.0207e-06, ..., -2.5984e-07, -2.0489e-07, 8.4704e-07], [ 1.2806e-07, 4.8168e-06, 1.3277e-05, ..., 2.5481e-06, 2.1886e-08, 8.7470e-06], ..., [ 3.6228e-07, -4.0419e-06, 2.2240e-06, ..., 5.1456e-07, 1.0896e-07, 2.5369e-06], [ 4.6305e-06, 1.5898e-06, 2.6956e-05, ..., 1.1235e-05, 6.3796e-08, 4.1813e-05], [-1.4724e-06, 4.0568e-06, -2.0549e-05, ..., -5.9940e-06, 6.1933e-08, -4.3929e-05]], device='cuda:0') Epoch 140, bias, value: tensor([ 0.0015, 0.0030, 0.0076, 0.0124, 0.0225, 0.0396, -0.0270, 0.0116, -0.0363, -0.0211], device='cuda:0'), grad: tensor([-8.1003e-05, 2.6524e-06, 2.9922e-05, 2.6748e-06, 7.6294e-06, 4.0904e-06, 1.7151e-05, -8.1677e-07, 5.1975e-05, -3.4362e-05], device='cuda:0') 100 0.0001 changing lr epoch 139, time 217.21, cls_loss 0.0042 cls_loss_mapping 0.0055 cls_loss_causal 0.5839 re_mapping 0.0083 re_causal 0.0231 /// teacc 98.69 lr 0.00010000 Epoch 141, weight, value: tensor([[-0.0294, 0.1163, 0.0699, ..., -0.1582, -0.1015, 0.0245], [ 0.1112, -0.0119, -0.0499, ..., 0.0658, 0.0929, -0.1079], [-0.0646, -0.0477, -0.0423, ..., -0.0669, -0.0189, -0.0443], ..., [-0.0450, 0.1334, -0.1535, ..., -0.1125, 0.0062, -0.1224], [ 0.0031, -0.0871, 0.0334, ..., 0.0563, -0.1049, -0.0406], [-0.0541, -0.0817, 0.0551, ..., -0.0714, -0.0525, 0.1161]], device='cuda:0'), grad: tensor([[ 3.7253e-07, -1.5572e-05, -2.5071e-06, ..., 4.1816e-07, -1.3888e-05, -8.5309e-06], [-1.3039e-06, 3.6182e-07, 6.4354e-07, ..., 4.0270e-06, 4.3213e-07, 1.1222e-07], [ 1.4007e-06, 1.8589e-06, -6.8173e-07, ..., 6.6310e-07, -1.2845e-05, 3.4412e-07], ..., [ 2.9281e-06, 1.9111e-06, 5.6662e-06, ..., 1.1295e-05, 1.8794e-06, 4.7591e-07], [ 5.7295e-06, 1.5516e-06, 3.8035e-06, ..., 6.1393e-06, 6.9849e-06, 6.8396e-06], [ 8.2003e-07, 2.9374e-06, 1.1744e-06, ..., 5.4277e-06, 2.1644e-06, 4.7637e-07]], device='cuda:0') Epoch 141, bias, value: tensor([ 0.0014, 0.0023, 0.0062, 0.0121, 0.0225, 0.0397, -0.0261, 0.0118, -0.0350, -0.0215], device='cuda:0'), grad: tensor([-2.2560e-05, 9.2268e-05, -3.3321e-03, 4.2558e-04, -3.2216e-05, -4.9233e-05, 2.8908e-05, 2.7657e-03, 9.8765e-05, 2.2471e-05], device='cuda:0') 100 0.0001 changing lr epoch 140, time 217.61, cls_loss 0.0063 cls_loss_mapping 0.0072 cls_loss_causal 0.5785 re_mapping 0.0085 re_causal 0.0238 /// teacc 98.90 lr 0.00010000 Epoch 142, weight, value: tensor([[-0.0297, 0.1171, 0.0702, ..., -0.1590, -0.1013, 0.0247], [ 0.1124, -0.0122, -0.0502, ..., 0.0662, 0.0932, -0.1082], [-0.0652, -0.0504, -0.0442, ..., -0.0679, -0.0193, -0.0443], ..., [-0.0462, 0.1345, -0.1539, ..., -0.1139, 0.0059, -0.1230], [ 0.0028, -0.0875, 0.0359, ..., 0.0560, -0.1041, -0.0382], [-0.0544, -0.0820, 0.0533, ..., -0.0719, -0.0526, 0.1146]], device='cuda:0'), grad: tensor([[ 1.9372e-07, -1.1129e-07, 4.0652e-07, ..., 4.1630e-07, 3.3062e-08, 3.0082e-07], [-1.3504e-08, 3.1851e-07, 3.4971e-07, ..., -1.0915e-05, -2.5555e-06, 2.8918e-07], [ 1.3083e-05, 9.5461e-08, 3.6415e-07, ..., 1.7852e-05, 5.4948e-07, 4.2841e-08], ..., [ 7.4506e-08, -4.9919e-06, 2.5239e-07, ..., 5.1558e-06, 1.1474e-06, 1.8720e-07], [ 4.7451e-07, 6.3796e-08, 5.5954e-06, ..., 2.3320e-06, 2.6263e-07, 5.6773e-06], [ 1.6019e-07, 4.2431e-06, -1.8120e-05, ..., -2.9467e-06, 6.6124e-08, -1.6928e-05]], device='cuda:0') Epoch 142, bias, value: tensor([ 0.0016, 0.0021, 0.0055, 0.0118, 0.0231, 0.0395, -0.0252, 0.0118, -0.0329, -0.0232], device='cuda:0'), grad: tensor([ 1.1930e-06, -1.5527e-05, 6.1333e-05, 6.8173e-06, 1.9342e-05, -4.5300e-06, -6.2048e-05, 1.7779e-06, 1.0975e-05, -1.9431e-05], device='cuda:0') 100 0.0001 changing lr epoch 141, time 217.74, cls_loss 0.0039 cls_loss_mapping 0.0046 cls_loss_causal 0.5521 re_mapping 0.0088 re_causal 0.0244 /// teacc 98.90 lr 0.00010000 Epoch 143, weight, value: tensor([[-0.0298, 0.1175, 0.0703, ..., -0.1592, -0.1014, 0.0248], [ 0.1132, -0.0124, -0.0498, ..., 0.0669, 0.0934, -0.1081], [-0.0652, -0.0491, -0.0471, ..., -0.0702, -0.0211, -0.0445], ..., [-0.0465, 0.1341, -0.1547, ..., -0.1142, 0.0057, -0.1235], [ 0.0026, -0.0881, 0.0356, ..., 0.0556, -0.1042, -0.0384], [-0.0547, -0.0818, 0.0538, ..., -0.0720, -0.0528, 0.1152]], device='cuda:0'), grad: tensor([[ 2.8424e-06, -5.7695e-07, 4.0457e-06, ..., 9.7156e-06, 4.6566e-10, 3.7160e-06], [ 4.1910e-09, 2.6748e-06, 5.9698e-07, ..., 3.6554e-07, -3.6787e-08, 6.6869e-07], [ 2.5295e-06, 5.5414e-07, 4.3586e-06, ..., 8.6948e-06, 4.6566e-09, 4.0643e-06], ..., [-2.4214e-08, -2.3916e-05, 1.4193e-06, ..., 4.7218e-07, 1.8626e-08, 1.6587e-06], [ 4.8392e-06, 5.3346e-06, 1.3806e-05, ..., 1.6183e-05, 4.1910e-09, 1.7002e-05], [ 1.4389e-07, 7.0967e-07, -1.2435e-05, ..., 3.2820e-06, 2.7940e-09, -1.7881e-05]], device='cuda:0') Epoch 143, bias, value: tensor([ 0.0016, 0.0023, 0.0041, 0.0117, 0.0227, 0.0418, -0.0250, 0.0110, -0.0334, -0.0227], device='cuda:0'), grad: tensor([ 1.9073e-05, 6.5342e-06, 1.8761e-05, 2.7806e-05, 1.8422e-06, 4.8399e-04, -5.5647e-04, -3.9190e-05, 5.9187e-05, -2.1219e-05], device='cuda:0') 100 0.0001 changing lr epoch 142, time 217.68, cls_loss 0.0041 cls_loss_mapping 0.0053 cls_loss_causal 0.5561 re_mapping 0.0079 re_causal 0.0230 /// teacc 98.88 lr 0.00010000 Epoch 144, weight, value: tensor([[-0.0305, 0.1180, 0.0706, ..., -0.1600, -0.1030, 0.0248], [ 0.1133, -0.0125, -0.0504, ..., 0.0669, 0.0933, -0.1083], [-0.0650, -0.0491, -0.0469, ..., -0.0701, -0.0211, -0.0447], ..., [-0.0466, 0.1350, -0.1548, ..., -0.1149, 0.0059, -0.1239], [ 0.0027, -0.0886, 0.0356, ..., 0.0559, -0.1042, -0.0385], [-0.0549, -0.0824, 0.0540, ..., -0.0726, -0.0525, 0.1154]], device='cuda:0'), grad: tensor([[ 6.7139e-04, 3.6812e-04, -3.7253e-08, ..., 6.9551e-06, 1.2178e-03, 1.9836e-04], [-1.1265e-05, 3.2410e-07, 9.0338e-08, ..., -1.8269e-05, 6.5612e-07, 2.1746e-07], [ 7.7635e-06, 1.0133e-06, 1.0245e-07, ..., 9.6485e-06, 2.7008e-06, 5.4669e-07], ..., [ 4.2915e-06, -5.8068e-07, 1.5367e-07, ..., 6.7279e-06, 1.4435e-07, 1.0012e-07], [ 7.4971e-07, 1.0421e-06, -1.9390e-06, ..., -4.2170e-06, 1.6969e-06, -2.8200e-06], [ 2.1942e-06, 6.7195e-07, 2.1886e-06, ..., 6.2101e-06, 8.2795e-07, 3.1181e-06]], device='cuda:0') Epoch 144, bias, value: tensor([ 0.0013, 0.0018, 0.0044, 0.0129, 0.0233, 0.0404, -0.0246, 0.0116, -0.0334, -0.0231], device='cuda:0'), grad: tensor([ 2.3117e-03, -5.1469e-05, 3.2634e-05, 1.6832e-03, 1.2731e-06, -1.6518e-03, -2.3556e-03, 1.9148e-05, -6.6590e-08, 1.3053e-05], device='cuda:0') 100 0.0001 changing lr epoch 143, time 217.10, cls_loss 0.0029 cls_loss_mapping 0.0051 cls_loss_causal 0.5340 re_mapping 0.0079 re_causal 0.0231 /// teacc 98.89 lr 0.00010000 Epoch 145, weight, value: tensor([[-0.0314, 0.1182, 0.0712, ..., -0.1604, -0.1044, 0.0251], [ 0.1139, -0.0127, -0.0504, ..., 0.0670, 0.0942, -0.1084], [-0.0656, -0.0488, -0.0477, ..., -0.0702, -0.0218, -0.0450], ..., [-0.0472, 0.1352, -0.1552, ..., -0.1155, 0.0057, -0.1245], [ 0.0027, -0.0889, 0.0359, ..., 0.0560, -0.1043, -0.0386], [-0.0554, -0.0827, 0.0539, ..., -0.0737, -0.0527, 0.1154]], device='cuda:0'), grad: tensor([[ 3.4692e-07, -2.7902e-06, -4.6343e-06, ..., 5.4203e-07, 8.5216e-08, -5.3346e-06], [ 2.9877e-06, 2.6003e-06, 3.2391e-06, ..., 3.0864e-06, -1.9977e-07, 4.4517e-07], [ 3.3975e-06, 2.9011e-07, 2.8368e-06, ..., 2.4978e-06, -2.6077e-08, 2.8173e-07], ..., [ 1.6019e-05, -4.6790e-06, 1.4462e-05, ..., 1.2591e-05, 4.2375e-07, 2.6580e-06], [ 1.1779e-05, -1.5646e-06, 5.8532e-05, ..., 2.2382e-05, 1.1228e-05, 8.0943e-05], [ 1.2200e-06, 1.6391e-06, -4.7803e-05, ..., -1.7449e-05, -1.2986e-05, -9.1970e-05]], device='cuda:0') Epoch 145, bias, value: tensor([ 0.0013, 0.0019, 0.0035, 0.0128, 0.0245, 0.0404, -0.0244, 0.0115, -0.0327, -0.0237], device='cuda:0'), grad: tensor([-1.4074e-05, 2.1204e-05, 1.4484e-05, -1.9777e-04, 2.4494e-07, 4.4674e-05, 3.6284e-06, 6.2048e-05, 1.5354e-04, -8.8036e-05], device='cuda:0') 100 0.0001 changing lr epoch 144, time 217.13, cls_loss 0.0030 cls_loss_mapping 0.0042 cls_loss_causal 0.5391 re_mapping 0.0082 re_causal 0.0230 /// teacc 98.87 lr 0.00010000 Epoch 146, weight, value: tensor([[-0.0316, 0.1193, 0.0702, ..., -0.1608, -0.1052, 0.0240], [ 0.1146, -0.0126, -0.0511, ..., 0.0675, 0.0943, -0.1096], [-0.0656, -0.0486, -0.0478, ..., -0.0705, -0.0218, -0.0452], ..., [-0.0479, 0.1353, -0.1557, ..., -0.1160, 0.0056, -0.1250], [ 0.0025, -0.0898, 0.0358, ..., 0.0558, -0.1045, -0.0387], [-0.0559, -0.0829, 0.0543, ..., -0.0740, -0.0520, 0.1160]], device='cuda:0'), grad: tensor([[ 3.4459e-07, -1.8813e-07, -8.9407e-08, ..., 3.8417e-07, -4.7497e-08, -2.2398e-07], [-4.1910e-08, 1.4259e-06, 1.3271e-07, ..., 8.1956e-07, -8.1491e-08, 9.9652e-08], [ 5.6345e-08, 5.4948e-08, 1.2061e-07, ..., 2.0210e-07, -4.0978e-08, 1.8161e-08], ..., [ 3.6787e-08, -2.5071e-06, 5.3318e-07, ..., 3.0231e-06, 1.3271e-07, 3.7346e-07], [ 5.8068e-07, 1.7835e-07, 2.3562e-06, ..., 5.8953e-07, 1.4203e-07, 2.3898e-06], [ 2.2538e-07, 1.3951e-06, -2.0489e-06, ..., 4.2468e-06, 7.1712e-08, -2.9393e-06]], device='cuda:0') Epoch 146, bias, value: tensor([ 0.0005, 0.0019, 0.0036, 0.0128, 0.0245, 0.0409, -0.0246, 0.0110, -0.0329, -0.0233], device='cuda:0'), grad: tensor([ 7.1526e-07, 4.6678e-06, -2.4401e-07, 4.0233e-06, -1.3381e-05, -8.2999e-06, -2.4941e-06, 2.4829e-06, 6.3553e-06, 6.1840e-06], device='cuda:0') 100 0.0001 changing lr epoch 145, time 217.50, cls_loss 0.0024 cls_loss_mapping 0.0041 cls_loss_causal 0.5600 re_mapping 0.0076 re_causal 0.0233 /// teacc 98.90 lr 0.00010000 Epoch 147, weight, value: tensor([[-0.0316, 0.1197, 0.0706, ..., -0.1611, -0.1053, 0.0242], [ 0.1149, -0.0126, -0.0513, ..., 0.0678, 0.0946, -0.1098], [-0.0659, -0.0488, -0.0481, ..., -0.0707, -0.0220, -0.0452], ..., [-0.0483, 0.1358, -0.1558, ..., -0.1164, 0.0054, -0.1254], [ 0.0025, -0.0903, 0.0357, ..., 0.0558, -0.1045, -0.0388], [-0.0562, -0.0833, 0.0543, ..., -0.0743, -0.0521, 0.1162]], device='cuda:0'), grad: tensor([[ 2.1718e-06, 3.2224e-06, -1.9325e-07, ..., 2.2119e-07, 6.5193e-09, 4.6985e-07], [-4.6194e-07, 2.4913e-07, 1.6261e-06, ..., -2.3842e-07, -1.3970e-07, 1.6820e-06], [ 1.1995e-06, 1.8869e-06, 2.2799e-06, ..., 2.7288e-07, 1.5367e-08, 1.7649e-07], ..., [ 2.2687e-06, 3.5036e-06, 6.4932e-06, ..., 4.4284e-07, 7.2177e-08, 1.6149e-06], [ 1.2666e-06, 6.5705e-07, 3.0454e-06, ..., 4.1910e-07, 2.8871e-08, 4.1574e-06], [ 4.1537e-07, 7.1526e-07, -1.3433e-05, ..., 6.0583e-07, 3.4925e-08, -1.8448e-05]], device='cuda:0') Epoch 147, bias, value: tensor([ 0.0007, 0.0020, 0.0033, 0.0132, 0.0246, 0.0408, -0.0243, 0.0108, -0.0330, -0.0233], device='cuda:0'), grad: tensor([ 1.1347e-05, 4.8354e-06, 7.3388e-06, -2.3901e-05, 2.6077e-05, 8.1677e-07, -1.7479e-05, 1.7807e-05, 1.0617e-05, -3.7491e-05], device='cuda:0') 100 0.0001 changing lr epoch 146, time 217.21, cls_loss 0.0027 cls_loss_mapping 0.0049 cls_loss_causal 0.5621 re_mapping 0.0076 re_causal 0.0230 /// teacc 98.92 lr 0.00010000 Epoch 148, weight, value: tensor([[-0.0299, 0.1212, 0.0708, ..., -0.1613, -0.1053, 0.0242], [ 0.1147, -0.0130, -0.0516, ..., 0.0678, 0.0950, -0.1102], [-0.0662, -0.0489, -0.0482, ..., -0.0708, -0.0223, -0.0458], ..., [-0.0486, 0.1365, -0.1561, ..., -0.1176, 0.0052, -0.1261], [ 0.0024, -0.0906, 0.0356, ..., 0.0557, -0.1047, -0.0389], [-0.0564, -0.0843, 0.0546, ..., -0.0746, -0.0523, 0.1167]], device='cuda:0'), grad: tensor([[ 2.2352e-08, -2.1793e-06, -1.1623e-06, ..., 7.1712e-08, -1.2107e-08, -1.4994e-06], [-3.5437e-07, 1.6810e-07, 3.0641e-07, ..., -5.5647e-07, -8.4285e-08, 3.7253e-08], [ 5.8208e-08, 2.3469e-07, 1.3178e-06, ..., 2.1933e-07, 2.2817e-08, 6.5658e-08], ..., [ 8.3819e-08, -4.4098e-07, 2.2277e-06, ..., 1.9884e-07, 1.8626e-08, 7.7300e-07], [-4.4238e-08, 1.6950e-07, 2.9001e-06, ..., -6.1328e-07, 3.7253e-08, 2.3432e-06], [ 4.7032e-08, 1.8133e-06, -1.0710e-06, ..., 1.8626e-07, 2.9337e-08, -2.3693e-06]], device='cuda:0') Epoch 148, bias, value: tensor([ 0.0018, 0.0015, 0.0033, 0.0137, 0.0247, 0.0403, -0.0246, 0.0105, -0.0332, -0.0230], device='cuda:0'), grad: tensor([-4.0531e-06, 1.3933e-06, -1.0729e-05, -2.3812e-05, 5.5693e-06, 6.8322e-06, 7.8380e-06, 7.9796e-06, 8.5235e-06, 5.4203e-07], device='cuda:0') 100 0.0001 changing lr epoch 147, time 216.86, cls_loss 0.0031 cls_loss_mapping 0.0047 cls_loss_causal 0.5443 re_mapping 0.0077 re_causal 0.0223 /// teacc 98.92 lr 0.00010000 Epoch 149, weight, value: tensor([[-0.0300, 0.1235, 0.0722, ..., -0.1619, -0.1054, 0.0248], [ 0.1148, -0.0140, -0.0517, ..., 0.0679, 0.0952, -0.1103], [-0.0664, -0.0493, -0.0485, ..., -0.0709, -0.0225, -0.0460], ..., [-0.0486, 0.1378, -0.1566, ..., -0.1176, 0.0052, -0.1270], [ 0.0028, -0.0920, 0.0355, ..., 0.0558, -0.1047, -0.0391], [-0.0567, -0.0849, 0.0548, ..., -0.0742, -0.0525, 0.1172]], device='cuda:0'), grad: tensor([[ 8.6101e-07, 2.4308e-07, 1.7881e-07, ..., 1.8124e-06, 7.6834e-08, 2.6124e-07], [-1.0945e-05, -7.5484e-07, -2.8759e-06, ..., 7.9691e-05, -7.3016e-06, 2.3097e-07], [ 3.2634e-06, 9.1270e-08, 9.9279e-07, ..., 3.3714e-06, 3.1106e-06, 1.4156e-07], ..., [ 2.9448e-06, 2.6682e-07, 6.2305e-07, ..., 7.7188e-06, 1.0980e-06, 8.4285e-08], [ 1.8124e-06, 3.4180e-07, 6.6264e-07, ..., 9.5665e-06, 4.4750e-07, 5.6531e-07], [ 1.5050e-06, 1.3784e-07, 3.6694e-07, ..., 1.5843e-04, 1.0338e-06, 7.2829e-07]], device='cuda:0') Epoch 149, bias, value: tensor([ 0.0031, 0.0010, 0.0032, 0.0127, 0.0240, 0.0413, -0.0250, 0.0109, -0.0334, -0.0228], device='cuda:0'), grad: tensor([ 4.5896e-06, 1.4436e-04, 6.9775e-06, 2.9244e-06, -4.9686e-04, 6.8620e-06, 2.2929e-06, 1.5572e-05, 1.9208e-05, 2.9469e-04], device='cuda:0') 100 0.0001 changing lr epoch 148, time 217.35, cls_loss 0.0026 cls_loss_mapping 0.0044 cls_loss_causal 0.5725 re_mapping 0.0075 re_causal 0.0234 /// teacc 98.93 lr 0.00010000 Epoch 150, weight, value: tensor([[-0.0309, 0.1238, 0.0723, ..., -0.1636, -0.1076, 0.0248], [ 0.1150, -0.0144, -0.0519, ..., 0.0678, 0.0958, -0.1106], [-0.0665, -0.0492, -0.0487, ..., -0.0711, -0.0227, -0.0460], ..., [-0.0489, 0.1384, -0.1568, ..., -0.1182, 0.0046, -0.1272], [ 0.0026, -0.0927, 0.0354, ..., 0.0556, -0.1051, -0.0392], [-0.0569, -0.0853, 0.0549, ..., -0.0746, -0.0526, 0.1174]], device='cuda:0'), grad: tensor([[ 6.2771e-07, -6.5193e-08, 3.7346e-07, ..., 6.5658e-07, 1.6578e-07, 2.8405e-07], [ 1.7405e-05, 3.6368e-07, 9.1493e-06, ..., 1.0677e-05, 3.5390e-08, 4.6715e-06], [ 1.4501e-06, 2.1234e-07, 7.6741e-07, ..., 9.5274e-07, 2.9802e-08, 3.1292e-07], ..., [ 4.9826e-07, -1.1651e-06, 9.3784e-07, ..., 4.7032e-07, 8.8941e-08, 6.8778e-07], [-3.0786e-05, 1.6904e-07, -7.6890e-06, ..., -2.0549e-05, 1.3709e-06, -2.6692e-06], [ 6.4541e-07, 4.1304e-07, -3.7365e-06, ..., 1.1474e-06, 2.3004e-07, -3.5372e-06]], device='cuda:0') Epoch 150, bias, value: tensor([ 0.0022, 0.0006, 0.0033, 0.0124, 0.0250, 0.0420, -0.0251, 0.0110, -0.0337, -0.0230], device='cuda:0'), grad: tensor([ 1.5134e-06, 2.4244e-05, 3.2000e-06, 6.0163e-06, 7.7020e-07, -9.8050e-06, 7.3798e-06, 6.0210e-07, -2.9281e-05, -4.6194e-06], device='cuda:0') 100 0.0001 changing lr epoch 149, time 217.01, cls_loss 0.0028 cls_loss_mapping 0.0040 cls_loss_causal 0.5482 re_mapping 0.0075 re_causal 0.0220 /// teacc 98.95 lr 0.00010000 Epoch 151, weight, value: tensor([[-0.0311, 0.1242, 0.0724, ..., -0.1644, -0.1081, 0.0248], [ 0.1152, -0.0144, -0.0521, ..., 0.0676, 0.0961, -0.1108], [-0.0668, -0.0489, -0.0486, ..., -0.0712, -0.0228, -0.0461], ..., [-0.0496, 0.1387, -0.1571, ..., -0.1186, 0.0043, -0.1276], [ 0.0028, -0.0935, 0.0353, ..., 0.0558, -0.1053, -0.0394], [-0.0572, -0.0855, 0.0551, ..., -0.0749, -0.0527, 0.1176]], device='cuda:0'), grad: tensor([[ 1.2247e-07, 5.3970e-07, 4.2878e-06, ..., 1.5050e-06, 7.9162e-09, 4.5449e-06], [-4.0419e-06, -2.6077e-07, 1.2107e-07, ..., -3.3807e-06, -9.1735e-08, 1.1269e-07], [ 1.5367e-07, 2.4140e-06, 6.0536e-08, ..., 1.8720e-07, 2.5146e-08, 6.0536e-08], ..., [ 2.7064e-06, -4.1962e-05, 8.0885e-07, ..., 2.5220e-06, 1.9791e-07, 7.8510e-07], [ 3.5577e-07, 2.2259e-06, 2.6554e-05, ..., 8.7768e-06, -1.5600e-07, 2.8074e-05], [ 7.3668e-07, 3.3557e-05, -3.5048e-05, ..., -9.2462e-06, 3.4925e-08, -3.6210e-05]], device='cuda:0') Epoch 151, bias, value: tensor([ 0.0019, 0.0005, 0.0035, 0.0113, 0.0251, 0.0428, -0.0251, 0.0108, -0.0338, -0.0229], device='cuda:0'), grad: tensor([ 1.0140e-05, -7.1898e-06, 3.0100e-06, 6.3293e-06, 4.2506e-06, 6.2771e-07, 5.6904e-07, -6.0588e-05, 5.7846e-05, -1.4879e-05], device='cuda:0') 100 0.0001 changing lr epoch 150, time 217.20, cls_loss 0.0023 cls_loss_mapping 0.0044 cls_loss_causal 0.5272 re_mapping 0.0073 re_causal 0.0221 /// teacc 98.98 lr 0.00010000 Epoch 152, weight, value: tensor([[-0.0317, 0.1247, 0.0726, ..., -0.1650, -0.1085, 0.0249], [ 0.1160, -0.0144, -0.0522, ..., 0.0679, 0.0965, -0.1111], [-0.0671, -0.0489, -0.0486, ..., -0.0713, -0.0230, -0.0461], ..., [-0.0501, 0.1389, -0.1574, ..., -0.1190, 0.0042, -0.1282], [ 0.0010, -0.0964, 0.0350, ..., 0.0544, -0.1054, -0.0398], [-0.0573, -0.0857, 0.0551, ..., -0.0762, -0.0527, 0.1178]], device='cuda:0'), grad: tensor([[ 7.4506e-09, -5.3868e-06, -2.9989e-06, ..., 4.0280e-08, -8.9873e-07, -4.2282e-06], [-2.5798e-07, 2.2724e-07, 9.7789e-09, ..., -3.2317e-07, -1.1642e-08, 4.4703e-08], [ 1.2130e-07, 1.5763e-07, 6.9849e-08, ..., 2.2096e-07, -5.3551e-09, 5.9605e-08], ..., [ 1.1059e-07, -4.2934e-07, 8.1258e-08, ..., 2.3679e-07, 2.1188e-08, 1.4924e-07], [-2.5379e-08, 1.3225e-07, 6.5425e-08, ..., 3.4925e-09, 1.8161e-08, 2.0908e-07], [ 2.5844e-08, 3.8147e-06, 2.1942e-06, ..., 1.7439e-07, 7.2550e-07, 2.8461e-06]], device='cuda:0') Epoch 152, bias, value: tensor([ 0.0019, 0.0006, 0.0034, 0.0114, 0.0262, 0.0435, -0.0250, 0.0108, -0.0346, -0.0235], device='cuda:0'), grad: tensor([-1.3314e-05, 3.7951e-08, 6.7148e-07, 5.3225e-07, 4.8149e-07, 1.2051e-06, 1.0692e-06, 1.4971e-07, 3.1316e-07, 8.8215e-06], device='cuda:0') 100 0.0001 changing lr epoch 151, time 217.66, cls_loss 0.0025 cls_loss_mapping 0.0048 cls_loss_causal 0.5431 re_mapping 0.0076 re_causal 0.0212 /// teacc 98.96 lr 0.00010000 Epoch 153, weight, value: tensor([[-0.0319, 0.1250, 0.0728, ..., -0.1654, -0.1085, 0.0250], [ 0.1162, -0.0146, -0.0524, ..., 0.0678, 0.0966, -0.1114], [-0.0674, -0.0487, -0.0487, ..., -0.0715, -0.0230, -0.0459], ..., [-0.0503, 0.1392, -0.1577, ..., -0.1195, 0.0042, -0.1285], [ 0.0010, -0.0967, 0.0350, ..., 0.0545, -0.1054, -0.0399], [-0.0576, -0.0859, 0.0553, ..., -0.0767, -0.0528, 0.1181]], device='cuda:0'), grad: tensor([[ 2.9267e-07, -1.7406e-06, -7.1852e-07, ..., 5.5972e-07, -1.0813e-06, -1.1697e-06], [ 6.2492e-07, -7.1945e-08, 4.4121e-07, ..., -8.2189e-08, -7.2643e-07, -2.0745e-07], [ 5.5693e-07, 1.2736e-07, 1.2107e-07, ..., 5.2992e-07, 1.6112e-07, 8.8476e-08], ..., [ 1.3104e-06, -4.8243e-07, 2.0466e-07, ..., 1.0617e-06, 4.5565e-07, 2.7148e-07], [-2.2158e-05, 9.8487e-08, -3.3788e-06, ..., -1.1146e-05, -6.5658e-07, 1.5125e-06], [ 1.0235e-06, 1.2415e-06, 4.1467e-07, ..., 9.3970e-07, 9.7509e-07, 6.3330e-07]], device='cuda:0') Epoch 153, bias, value: tensor([ 0.0020, 0.0003, 0.0039, 0.0118, 0.0266, 0.0432, -0.0249, 0.0105, -0.0346, -0.0236], device='cuda:0'), grad: tensor([-5.0776e-06, 4.4773e-07, 1.8235e-06, 2.1845e-05, 1.5693e-07, 6.4671e-06, 9.1502e-07, 2.4382e-06, -3.5048e-05, 6.0722e-06], device='cuda:0') 100 0.0001 changing lr epoch 152, time 217.32, cls_loss 0.0028 cls_loss_mapping 0.0047 cls_loss_causal 0.5532 re_mapping 0.0073 re_causal 0.0214 /// teacc 98.98 lr 0.00010000 Epoch 154, weight, value: tensor([[-0.0322, 0.1253, 0.0722, ..., -0.1662, -0.1087, 0.0242], [ 0.1168, -0.0148, -0.0530, ..., 0.0681, 0.0968, -0.1117], [-0.0675, -0.0488, -0.0487, ..., -0.0715, -0.0228, -0.0461], ..., [-0.0505, 0.1398, -0.1579, ..., -0.1200, 0.0040, -0.1294], [ 0.0008, -0.0973, 0.0350, ..., 0.0542, -0.1055, -0.0401], [-0.0582, -0.0861, 0.0556, ..., -0.0770, -0.0529, 0.1187]], device='cuda:0'), grad: tensor([[ 5.5600e-07, 2.7940e-09, 2.4289e-06, ..., 4.8522e-07, 4.1910e-09, 3.1223e-07], [-6.9663e-06, -1.1008e-06, 2.4196e-06, ..., -7.6666e-06, -5.1223e-09, 1.7360e-06], [ 1.9744e-06, 1.6019e-07, 4.2245e-06, ..., 2.4550e-06, 3.2829e-08, 6.1002e-07], ..., [ 3.1758e-06, -1.6764e-08, 1.3495e-06, ..., 2.4941e-06, 3.6089e-08, 3.7719e-07], [-1.4909e-05, 1.3062e-07, -1.5393e-05, ..., -4.3392e-05, 5.9372e-08, -3.7521e-05], [ 1.2189e-05, 5.9791e-07, 1.6123e-05, ..., 2.9460e-05, 7.6834e-09, 2.3499e-05]], device='cuda:0') Epoch 154, bias, value: tensor([ 0.0013, 0.0002, 0.0044, 0.0115, 0.0267, 0.0434, -0.0245, 0.0103, -0.0351, -0.0234], device='cuda:0'), grad: tensor([ 7.4171e-06, -1.5393e-05, 1.5527e-05, -3.0851e-04, 4.9770e-06, 2.7394e-04, 6.8992e-06, 1.1832e-05, -5.8442e-05, 6.1333e-05], device='cuda:0') 100 0.0001 changing lr epoch 153, time 217.54, cls_loss 0.0031 cls_loss_mapping 0.0066 cls_loss_causal 0.5615 re_mapping 0.0076 re_causal 0.0217 /// teacc 98.96 lr 0.00010000 Epoch 155, weight, value: tensor([[-0.0323, 0.1255, 0.0721, ..., -0.1674, -0.1088, 0.0241], [ 0.1172, -0.0149, -0.0537, ..., 0.0683, 0.0969, -0.1125], [-0.0669, -0.0490, -0.0491, ..., -0.0721, -0.0227, -0.0462], ..., [-0.0511, 0.1403, -0.1584, ..., -0.1206, 0.0037, -0.1304], [ 0.0007, -0.0979, 0.0349, ..., 0.0545, -0.1056, -0.0403], [-0.0589, -0.0864, 0.0560, ..., -0.0779, -0.0531, 0.1193]], device='cuda:0'), grad: tensor([[ 6.1933e-08, -9.9763e-06, -7.6592e-06, ..., -2.4065e-06, 3.7253e-09, -3.1590e-06], [-1.5516e-06, 2.1383e-06, 1.7765e-07, ..., -1.8245e-06, -1.6065e-08, 9.9186e-08], [ 3.3039e-07, 4.9882e-06, 3.8464e-07, ..., 4.8568e-07, -7.0548e-08, 1.6391e-07], ..., [ 6.2305e-07, -2.0042e-05, 2.6799e-07, ..., 8.3307e-07, 1.8626e-08, 2.1770e-07], [ 3.6135e-07, 7.5810e-07, 9.8348e-07, ..., 2.2631e-07, 1.1013e-07, 1.0142e-06], [ 3.4273e-07, 1.0572e-05, 9.1642e-07, ..., 2.1514e-06, 6.0536e-09, -9.5926e-07]], device='cuda:0') Epoch 155, bias, value: tensor([ 0.0012, -0.0004, 0.0049, 0.0105, 0.0272, 0.0433, -0.0230, 0.0104, -0.0357, -0.0234], device='cuda:0'), grad: tensor([-2.9042e-05, 6.6962e-07, 9.6932e-06, 6.1542e-06, 4.3325e-06, 2.5313e-06, 1.4752e-05, -3.7402e-05, 4.2208e-06, 2.4021e-05], device='cuda:0') 100 0.0001 changing lr epoch 154, time 217.20, cls_loss 0.0028 cls_loss_mapping 0.0040 cls_loss_causal 0.5425 re_mapping 0.0075 re_causal 0.0216 /// teacc 98.87 lr 0.00010000 Epoch 156, weight, value: tensor([[-0.0327, 0.1261, 0.0721, ..., -0.1681, -0.1089, 0.0241], [ 0.1211, -0.0151, -0.0516, ..., 0.0707, 0.0975, -0.1127], [-0.0681, -0.0488, -0.0496, ..., -0.0725, -0.0233, -0.0462], ..., [-0.0516, 0.1408, -0.1587, ..., -0.1210, 0.0036, -0.1307], [-0.0009, -0.0983, 0.0343, ..., 0.0537, -0.1057, -0.0404], [-0.0594, -0.0867, 0.0561, ..., -0.0781, -0.0533, 0.1196]], device='cuda:0'), grad: tensor([[-2.8443e-04, -3.0208e-04, -1.5981e-06, ..., 5.3924e-07, 2.8498e-07, -2.7958e-06], [ 1.0121e-04, 1.2082e-04, 8.7777e-08, ..., -1.3448e-05, -1.0140e-05, 6.7987e-08], [ 3.1590e-05, 2.9370e-05, 1.7858e-07, ..., -3.5256e-05, -6.6102e-05, 2.5705e-07], ..., [ 8.0168e-06, 3.1628e-06, 3.1199e-07, ..., 5.5917e-06, 4.5598e-06, 4.2538e-07], [ 3.8054e-06, 3.7868e-06, 4.4610e-07, ..., 1.1679e-06, 4.2585e-07, 9.9745e-07], [ 2.2519e-06, 2.5295e-06, -3.5716e-07, ..., 1.1921e-06, 9.2713e-07, -3.8720e-07]], device='cuda:0') Epoch 156, bias, value: tensor([ 0.0012, 0.0015, 0.0052, 0.0105, 0.0271, 0.0433, -0.0236, 0.0103, -0.0368, -0.0234], device='cuda:0'), grad: tensor([-9.6464e-04, 2.9707e-04, -1.8072e-04, 2.1577e-05, 2.2918e-05, -2.6941e-05, 7.4673e-04, 5.4181e-05, 1.6510e-05, 1.3798e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 155---------------------------------------------------- epoch 155, time 217.97, cls_loss 0.0032 cls_loss_mapping 0.0052 cls_loss_causal 0.5341 re_mapping 0.0076 re_causal 0.0215 /// teacc 99.00 lr 0.00010000 Epoch 157, weight, value: tensor([[-0.0322, 0.1270, 0.0718, ..., -0.1686, -0.1089, 0.0240], [ 0.1222, -0.0141, -0.0519, ..., 0.0726, 0.0976, -0.1131], [-0.0686, -0.0493, -0.0500, ..., -0.0728, -0.0232, -0.0464], ..., [-0.0535, 0.1407, -0.1591, ..., -0.1235, 0.0036, -0.1311], [-0.0012, -0.0989, 0.0344, ..., 0.0534, -0.1058, -0.0406], [-0.0604, -0.0874, 0.0566, ..., -0.0774, -0.0535, 0.1206]], device='cuda:0'), grad: tensor([[ 1.2387e-07, 5.0943e-07, -2.1583e-07, ..., 3.8277e-07, 4.0745e-08, -2.5891e-07], [-4.6613e-07, 8.9593e-07, 1.3388e-07, ..., 6.1840e-07, 2.6077e-08, 6.3796e-08], [ 9.3831e-08, 2.2165e-07, -2.2096e-07, ..., -5.6671e-07, 7.2177e-09, 2.7940e-08], ..., [ 2.0070e-07, 1.6633e-06, 3.8999e-07, ..., 4.6134e-05, -1.0189e-06, -2.4587e-07], [ 1.4552e-07, 1.8859e-07, 5.0664e-07, ..., 2.0433e-06, 8.1491e-09, 1.6391e-07], [ 7.8231e-08, 5.1521e-06, -4.7474e-07, ..., 1.4290e-05, 2.0210e-07, -8.6799e-07]], device='cuda:0') Epoch 157, bias, value: tensor([ 0.0014, 0.0020, 0.0049, 0.0100, 0.0258, 0.0439, -0.0232, 0.0099, -0.0369, -0.0227], device='cuda:0'), grad: tensor([ 1.0589e-06, 3.5968e-06, -1.8729e-06, -2.4438e-06, -1.6296e-04, 1.3679e-05, -3.1237e-06, 1.0687e-04, 5.7369e-06, 3.9458e-05], device='cuda:0') 100 0.0001 changing lr epoch 156, time 217.45, cls_loss 0.0029 cls_loss_mapping 0.0053 cls_loss_causal 0.5657 re_mapping 0.0073 re_causal 0.0213 /// teacc 98.96 lr 0.00010000 Epoch 158, weight, value: tensor([[-0.0323, 0.1284, 0.0717, ..., -0.1687, -0.1090, 0.0235], [ 0.1224, -0.0144, -0.0524, ..., 0.0727, 0.0979, -0.1136], [-0.0698, -0.0507, -0.0505, ..., -0.0729, -0.0233, -0.0470], ..., [-0.0540, 0.1412, -0.1599, ..., -0.1238, 0.0034, -0.1319], [-0.0010, -0.0992, 0.0342, ..., 0.0535, -0.1059, -0.0408], [-0.0611, -0.0882, 0.0574, ..., -0.0777, -0.0537, 0.1216]], device='cuda:0'), grad: tensor([[ 5.9530e-06, 1.5907e-06, 7.7263e-06, ..., 5.0552e-06, 0.0000e+00, 2.5965e-06], [-1.3828e-05, -2.2203e-06, -3.8883e-08, ..., -1.8731e-05, 0.0000e+00, 4.3493e-07], [-5.3830e-07, -1.0468e-05, -1.6242e-05, ..., 9.3058e-06, 0.0000e+00, 1.8673e-07], ..., [ 8.0317e-06, 3.5553e-07, 1.5106e-06, ..., 7.9647e-06, 0.0000e+00, 2.4354e-07], [-7.0445e-06, 7.6815e-06, 1.9325e-08, ..., -1.0669e-05, 0.0000e+00, -5.4799e-06], [ 2.2110e-06, 1.0319e-06, 7.9721e-07, ..., 1.1064e-06, 0.0000e+00, -4.6706e-07]], device='cuda:0') Epoch 158, bias, value: tensor([ 0.0016, 0.0017, 0.0049, 0.0111, 0.0256, 0.0435, -0.0241, 0.0098, -0.0370, -0.0222], device='cuda:0'), grad: tensor([ 2.1622e-05, -2.5481e-05, -1.1981e-04, 4.8168e-06, 5.5553e-07, 1.4216e-05, 8.3260e-07, 1.5453e-05, 8.3029e-05, 4.8578e-06], device='cuda:0') 100 0.0001 changing lr epoch 157, time 217.00, cls_loss 0.0021 cls_loss_mapping 0.0041 cls_loss_causal 0.5195 re_mapping 0.0072 re_causal 0.0210 /// teacc 98.96 lr 0.00010000 Epoch 159, weight, value: tensor([[-0.0325, 0.1293, 0.0725, ..., -0.1693, -0.1089, 0.0241], [ 0.1233, -0.0145, -0.0531, ..., 0.0731, 0.0983, -0.1144], [-0.0716, -0.0511, -0.0503, ..., -0.0736, -0.0237, -0.0473], ..., [-0.0544, 0.1418, -0.1602, ..., -0.1241, 0.0033, -0.1330], [-0.0007, -0.0996, 0.0343, ..., 0.0538, -0.1059, -0.0408], [-0.0616, -0.0885, 0.0573, ..., -0.0779, -0.0539, 0.1217]], device='cuda:0'), grad: tensor([[ 2.9523e-07, -1.2824e-06, -7.4040e-07, ..., 4.3912e-07, 9.3132e-10, -1.7937e-06], [-2.3767e-06, 3.2503e-07, -1.0943e-08, ..., -2.1048e-06, 4.4238e-09, -6.8219e-08], [-3.1311e-06, -3.5129e-06, -2.9560e-06, ..., 6.1002e-07, 6.9849e-10, 2.5379e-07], ..., [ 7.4320e-07, -4.9397e-06, 4.6007e-07, ..., 6.4448e-07, 6.5193e-09, 5.5786e-07], [ 5.0776e-06, 3.8408e-06, 2.7642e-06, ..., 4.0187e-07, 4.0280e-08, 1.5479e-06], [ 5.7369e-07, 4.2245e-06, -6.5472e-07, ..., 9.4622e-07, 2.3283e-09, -3.7253e-07]], device='cuda:0') Epoch 159, bias, value: tensor([ 0.0022, 0.0025, 0.0042, 0.0113, 0.0257, 0.0430, -0.0241, 0.0098, -0.0370, -0.0224], device='cuda:0'), grad: tensor([-3.6284e-06, -2.1346e-06, -6.9797e-05, 8.6427e-06, 2.7828e-06, -6.9737e-06, 1.9986e-06, -4.0494e-06, 6.6698e-05, 6.3963e-06], device='cuda:0') 100 0.0001 changing lr epoch 158, time 217.31, cls_loss 0.0020 cls_loss_mapping 0.0042 cls_loss_causal 0.5469 re_mapping 0.0070 re_causal 0.0215 /// teacc 98.97 lr 0.00010000 Epoch 160, weight, value: tensor([[-0.0331, 0.1297, 0.0724, ..., -0.1705, -0.1089, 0.0242], [ 0.1249, -0.0139, -0.0531, ..., 0.0743, 0.0991, -0.1144], [-0.0734, -0.0519, -0.0503, ..., -0.0745, -0.0244, -0.0479], ..., [-0.0556, 0.1417, -0.1604, ..., -0.1254, 0.0032, -0.1340], [-0.0006, -0.1000, 0.0343, ..., 0.0537, -0.1059, -0.0408], [-0.0624, -0.0883, 0.0574, ..., -0.0782, -0.0540, 0.1220]], device='cuda:0'), grad: tensor([[ 1.1874e-06, 2.7427e-07, 6.5845e-07, ..., 1.4799e-06, 0.0000e+00, -1.0454e-07], [ 1.9360e-04, 6.6340e-05, 2.1956e-07, ..., 2.1112e-04, -5.5879e-09, 4.6333e-08], [ 1.0572e-05, 3.6359e-06, -3.5893e-06, ..., 7.4245e-06, 2.3283e-10, 3.4692e-08], ..., [ 2.3302e-06, -7.0501e-07, 2.5891e-07, ..., 3.1032e-06, 2.0955e-09, 1.4226e-07], [-2.0945e-04, -7.1585e-05, 2.7437e-06, ..., -2.2602e-04, 2.3283e-10, 5.8860e-07], [ 4.8010e-07, 9.2015e-07, -2.5565e-07, ..., 1.0375e-06, 1.3970e-09, -9.3551e-07]], device='cuda:0') Epoch 160, bias, value: tensor([ 0.0021, 0.0033, 0.0039, 0.0111, 0.0259, 0.0429, -0.0239, 0.0092, -0.0371, -0.0222], device='cuda:0'), grad: tensor([ 5.3942e-06, 5.2118e-04, -1.1958e-06, -1.4730e-05, 2.7893e-07, 2.4661e-05, 5.9828e-06, 4.0904e-06, -5.4693e-04, 2.3022e-06], device='cuda:0') 100 0.0001 changing lr epoch 159, time 217.56, cls_loss 0.0024 cls_loss_mapping 0.0045 cls_loss_causal 0.5345 re_mapping 0.0070 re_causal 0.0214 /// teacc 98.96 lr 0.00010000 Epoch 161, weight, value: tensor([[-0.0333, 0.1301, 0.0724, ..., -0.1712, -0.1089, 0.0242], [ 0.1257, -0.0134, -0.0530, ..., 0.0745, 0.0997, -0.1145], [-0.0737, -0.0521, -0.0503, ..., -0.0745, -0.0244, -0.0483], ..., [-0.0567, 0.1422, -0.1609, ..., -0.1258, 0.0024, -0.1355], [-0.0006, -0.1004, 0.0341, ..., 0.0537, -0.1059, -0.0410], [-0.0628, -0.0893, 0.0578, ..., -0.0783, -0.0543, 0.1226]], device='cuda:0'), grad: tensor([[ 2.6310e-08, 3.9209e-07, -2.6985e-07, ..., 4.5169e-08, 8.6147e-09, -2.7008e-07], [-2.5844e-08, 3.3751e-06, 3.9348e-08, ..., -3.7486e-08, -1.4668e-08, 6.8452e-08], [ 2.0256e-08, 4.4145e-07, 1.6810e-07, ..., 1.4668e-08, 1.3970e-09, 1.4203e-08], ..., [ 4.6566e-09, -3.6049e-04, 6.2399e-08, ..., 5.6811e-08, 5.5879e-09, -5.6513e-06], [ 6.4727e-08, 6.9197e-07, 4.0629e-07, ..., 4.8196e-08, 5.3551e-09, 2.1956e-07], [ 2.6077e-08, 3.5238e-04, -2.7381e-07, ..., 2.7241e-08, 2.3283e-09, 4.9025e-06]], device='cuda:0') Epoch 161, bias, value: tensor([ 0.0021, 0.0036, 0.0040, 0.0112, 0.0257, 0.0427, -0.0240, 0.0090, -0.0372, -0.0219], device='cuda:0'), grad: tensor([ 4.8336e-07, 6.0834e-06, 8.7451e-07, -7.9442e-07, 4.5449e-06, 2.2147e-06, -2.0396e-07, -6.5851e-04, 2.1234e-06, 6.4373e-04], device='cuda:0') 100 0.0001 changing lr epoch 160, time 217.33, cls_loss 0.0022 cls_loss_mapping 0.0035 cls_loss_causal 0.5651 re_mapping 0.0072 re_causal 0.0216 /// teacc 98.89 lr 0.00010000 Epoch 162, weight, value: tensor([[-0.0337, 0.1302, 0.0723, ..., -0.1720, -0.1090, 0.0241], [ 0.1260, -0.0140, -0.0534, ..., 0.0747, 0.1000, -0.1152], [-0.0740, -0.0526, -0.0502, ..., -0.0749, -0.0245, -0.0485], ..., [-0.0567, 0.1434, -0.1617, ..., -0.1260, 0.0023, -0.1373], [-0.0006, -0.1013, 0.0342, ..., 0.0539, -0.1060, -0.0409], [-0.0632, -0.0895, 0.0579, ..., -0.0785, -0.0544, 0.1229]], device='cuda:0'), grad: tensor([[ 3.4459e-08, 1.1572e-07, -4.4471e-08, ..., 1.1688e-07, 1.2340e-08, -1.1828e-07], [-1.0571e-07, 3.5297e-06, 5.3132e-07, ..., 2.2608e-07, 6.5193e-09, 5.6578e-08], [ 1.8068e-07, 3.0234e-05, 3.6284e-06, ..., 3.1921e-07, -1.1711e-07, 6.8452e-08], ..., [ 2.2235e-07, -9.6917e-05, -1.1854e-05, ..., 4.8522e-07, 2.3050e-08, 5.8906e-08], [-1.6410e-06, 6.2585e-07, -1.1660e-06, ..., -1.9427e-06, 1.0012e-08, -3.9791e-07], [ 2.0163e-07, 5.5730e-05, 6.8992e-06, ..., 2.1905e-06, 9.5461e-09, -2.4005e-07]], device='cuda:0') Epoch 162, bias, value: tensor([ 0.0019, 0.0033, 0.0042, 0.0110, 0.0259, 0.0426, -0.0241, 0.0091, -0.0373, -0.0216], device='cuda:0'), grad: tensor([ 9.9000e-07, 9.6709e-06, 7.0751e-05, 6.2697e-06, 4.5560e-06, 6.6943e-06, 2.4959e-06, -2.4676e-04, -2.2221e-06, 1.4734e-04], device='cuda:0') 100 0.0001 changing lr epoch 161, time 217.28, cls_loss 0.0028 cls_loss_mapping 0.0051 cls_loss_causal 0.5337 re_mapping 0.0075 re_causal 0.0210 /// teacc 98.89 lr 0.00010000 Epoch 163, weight, value: tensor([[-0.0339, 0.1315, 0.0727, ..., -0.1729, -0.1094, 0.0242], [ 0.1261, -0.0142, -0.0539, ..., 0.0745, 0.1000, -0.1159], [-0.0740, -0.0539, -0.0506, ..., -0.0746, -0.0243, -0.0488], ..., [-0.0570, 0.1448, -0.1620, ..., -0.1263, 0.0022, -0.1382], [-0.0010, -0.1011, 0.0342, ..., 0.0536, -0.1060, -0.0412], [-0.0639, -0.0912, 0.0580, ..., -0.0792, -0.0546, 0.1232]], device='cuda:0'), grad: tensor([[ 5.4808e-07, -1.8075e-05, -1.3702e-05, ..., 2.1495e-06, 4.5169e-08, -2.0623e-05], [ 2.4870e-05, 3.2574e-05, 1.8045e-05, ..., 3.2157e-05, 1.1958e-05, 1.5318e-05], [ 4.0643e-06, 5.9698e-07, 1.8142e-06, ..., 3.2276e-05, 2.2873e-06, 8.9873e-07], ..., [-1.2949e-05, -4.3571e-05, 4.3982e-07, ..., -6.3032e-06, -2.1949e-05, 3.2503e-07], [-1.4581e-05, 6.3255e-06, -2.1324e-05, ..., -2.7180e-05, 3.2932e-06, -1.2994e-05], [ 1.5367e-06, 8.7321e-06, 6.3255e-06, ..., 1.5944e-06, 7.0501e-07, 8.8066e-06]], device='cuda:0') Epoch 163, bias, value: tensor([ 0.0024, 0.0029, 0.0040, 0.0095, 0.0261, 0.0434, -0.0233, 0.0094, -0.0372, -0.0221], device='cuda:0'), grad: tensor([-5.2303e-05, 1.4746e-04, 5.8472e-05, 1.0245e-05, -3.8564e-05, 1.4633e-05, -1.2062e-05, -1.2517e-04, -3.3200e-05, 3.0667e-05], device='cuda:0') 100 0.0001 changing lr epoch 162, time 217.42, cls_loss 0.0036 cls_loss_mapping 0.0061 cls_loss_causal 0.5676 re_mapping 0.0075 re_causal 0.0212 /// teacc 98.99 lr 0.00010000 Epoch 164, weight, value: tensor([[-0.0345, 0.1319, 0.0727, ..., -0.1739, -0.1094, 0.0244], [ 0.1257, -0.0162, -0.0561, ..., 0.0750, 0.1004, -0.1164], [-0.0743, -0.0536, -0.0508, ..., -0.0751, -0.0244, -0.0491], ..., [-0.0558, 0.1465, -0.1628, ..., -0.1266, 0.0023, -0.1379], [-0.0011, -0.1018, 0.0342, ..., 0.0533, -0.1063, -0.0412], [-0.0655, -0.0911, 0.0594, ..., -0.0812, -0.0550, 0.1238]], device='cuda:0'), grad: tensor([[-5.8208e-09, -2.9076e-06, -3.8669e-06, ..., 1.8161e-08, -2.3423e-07, -5.2564e-06], [-1.0920e-07, 3.0082e-07, 1.3388e-07, ..., -8.4285e-08, -1.0943e-08, 1.6578e-07], [ 2.1420e-08, 2.0862e-07, 1.0384e-07, ..., 2.1188e-08, -2.3283e-08, 1.2410e-07], ..., [ 4.8662e-08, -8.0885e-07, 3.7253e-07, ..., 8.5682e-08, 3.7253e-08, 6.4075e-07], [-2.1188e-08, 5.8673e-07, 8.4005e-07, ..., -8.1258e-08, 2.3516e-08, 1.1018e-06], [ 3.5623e-08, 1.8142e-06, 1.3411e-06, ..., 1.5227e-07, 1.4156e-07, 1.3849e-06]], device='cuda:0') Epoch 164, bias, value: tensor([ 0.0024, 0.0013, 0.0043, 0.0125, 0.0264, 0.0410, -0.0230, 0.0100, -0.0374, -0.0219], device='cuda:0'), grad: tensor([-1.4581e-05, 1.3094e-06, -1.6186e-06, 4.3400e-07, 7.3388e-07, 8.5495e-07, 2.4308e-06, 8.4797e-07, 3.0920e-06, 6.4671e-06], device='cuda:0') 100 0.0001 changing lr epoch 163, time 217.61, cls_loss 0.0022 cls_loss_mapping 0.0040 cls_loss_causal 0.5235 re_mapping 0.0074 re_causal 0.0205 /// teacc 98.98 lr 0.00010000 Epoch 165, weight, value: tensor([[-0.0354, 0.1313, 0.0730, ..., -0.1746, -0.1095, 0.0242], [ 0.1263, -0.0160, -0.0560, ..., 0.0757, 0.1006, -0.1165], [-0.0745, -0.0540, -0.0508, ..., -0.0752, -0.0245, -0.0493], ..., [-0.0565, 0.1468, -0.1630, ..., -0.1273, 0.0023, -0.1380], [-0.0010, -0.1022, 0.0341, ..., 0.0534, -0.1064, -0.0411], [-0.0666, -0.0917, 0.0595, ..., -0.0814, -0.0556, 0.1243]], device='cuda:0'), grad: tensor([[ 5.6578e-08, -2.2096e-07, -8.0168e-06, ..., 1.2573e-07, 4.8196e-08, -1.0945e-05], [ 2.0070e-07, 1.3039e-08, 1.1828e-07, ..., 1.4985e-06, 8.2422e-07, 1.4622e-07], [ 1.9167e-06, 5.9837e-08, 8.2888e-08, ..., 8.2999e-06, 4.0755e-06, 1.2619e-07], ..., [ 1.6093e-06, -5.1502e-07, 2.1141e-07, ..., 7.0557e-06, 3.2447e-06, 3.3597e-07], [-4.9174e-06, 1.0710e-07, -4.1984e-06, ..., -1.9819e-05, -1.0185e-05, -4.4107e-06], [ 7.7533e-07, 5.0105e-07, 7.0222e-06, ..., 4.0978e-07, 1.6065e-08, 8.1807e-06]], device='cuda:0') Epoch 165, bias, value: tensor([ 0.0019, 0.0016, 0.0044, 0.0124, 0.0261, 0.0409, -0.0221, 0.0101, -0.0375, -0.0220], device='cuda:0'), grad: tensor([-2.1785e-05, 3.5278e-06, 1.5825e-05, 2.7101e-06, 2.2352e-06, 1.7853e-06, 1.3232e-05, 1.3448e-05, -5.5045e-05, 2.4095e-05], device='cuda:0') 100 0.0001 changing lr epoch 164, time 217.33, cls_loss 0.0027 cls_loss_mapping 0.0048 cls_loss_causal 0.5711 re_mapping 0.0067 re_causal 0.0206 /// teacc 99.00 lr 0.00010000 Epoch 166, weight, value: tensor([[-0.0350, 0.1322, 0.0733, ..., -0.1748, -0.1096, 0.0242], [ 0.1268, -0.0159, -0.0561, ..., 0.0760, 0.1005, -0.1168], [-0.0743, -0.0539, -0.0509, ..., -0.0751, -0.0250, -0.0497], ..., [-0.0570, 0.1469, -0.1639, ..., -0.1278, 0.0032, -0.1398], [-0.0010, -0.1024, 0.0341, ..., 0.0538, -0.1067, -0.0412], [-0.0678, -0.0921, 0.0598, ..., -0.0820, -0.0560, 0.1249]], device='cuda:0'), grad: tensor([[ 2.2142e-07, -2.4755e-06, -6.5845e-07, ..., 3.7719e-07, 1.6298e-08, -2.0247e-06], [-3.4878e-07, 8.2655e-07, 2.8592e-07, ..., 4.3749e-07, -1.9791e-07, 4.1095e-07], [ 5.1223e-07, 3.2387e-07, 1.6508e-07, ..., 9.3784e-07, 1.6321e-07, 2.6054e-07], ..., [ 1.8300e-07, -3.4701e-06, 5.5460e-07, ..., 6.8871e-07, 4.5868e-08, 9.3644e-07], [ 1.1325e-06, 3.2131e-07, 3.0193e-06, ..., 1.6233e-06, 1.3062e-07, 3.8296e-06], [ 6.6310e-07, 3.4980e-06, -2.7549e-06, ..., 3.8981e-05, 2.0023e-08, 1.1269e-06]], device='cuda:0') Epoch 166, bias, value: tensor([ 0.0023, 0.0013, 0.0049, 0.0123, 0.0264, 0.0410, -0.0224, 0.0098, -0.0376, -0.0221], device='cuda:0'), grad: tensor([-7.6741e-06, 2.7884e-06, 2.9001e-06, 8.3447e-06, -9.5487e-05, -1.4305e-05, 4.2580e-06, -1.8328e-06, 1.0461e-05, 9.0361e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 165---------------------------------------------------- epoch 165, time 218.57, cls_loss 0.0024 cls_loss_mapping 0.0040 cls_loss_causal 0.5314 re_mapping 0.0070 re_causal 0.0198 /// teacc 99.02 lr 0.00010000 Epoch 167, weight, value: tensor([[-0.0353, 0.1328, 0.0731, ..., -0.1757, -0.1096, 0.0238], [ 0.1273, -0.0159, -0.0557, ..., 0.0767, 0.1006, -0.1167], [-0.0745, -0.0547, -0.0510, ..., -0.0754, -0.0252, -0.0500], ..., [-0.0579, 0.1468, -0.1641, ..., -0.1296, 0.0035, -0.1403], [-0.0008, -0.1005, 0.0339, ..., 0.0545, -0.1069, -0.0415], [-0.0688, -0.0925, 0.0600, ..., -0.0827, -0.0562, 0.1254]], device='cuda:0'), grad: tensor([[ 2.7940e-07, -2.1681e-06, -7.8045e-07, ..., 3.2061e-07, 3.6322e-08, -4.4145e-06], [-1.4305e-06, -1.0384e-07, 4.0978e-08, ..., -1.2247e-06, -3.7812e-07, 1.0221e-07], [ 1.5809e-07, 3.4459e-08, -2.1327e-07, ..., 2.2748e-07, 6.3563e-08, 1.8394e-07], ..., [ 4.4145e-07, -1.1828e-07, 9.6392e-08, ..., 6.5239e-07, 1.1688e-07, 2.0000e-07], [ 7.6368e-08, 1.7090e-07, 3.8277e-07, ..., 2.7684e-07, 2.6543e-08, 4.5076e-07], [ 9.5926e-08, 1.6512e-06, 3.2154e-07, ..., 4.7614e-07, 2.1188e-08, 2.8275e-06]], device='cuda:0') Epoch 167, bias, value: tensor([ 0.0021, 0.0015, 0.0048, 0.0118, 0.0267, 0.0415, -0.0225, 0.0091, -0.0371, -0.0221], device='cuda:0'), grad: tensor([-8.4490e-06, -2.1700e-06, -4.4890e-07, 7.6834e-07, -2.1365e-06, 6.3609e-07, 1.2415e-06, 1.3420e-06, 2.0899e-06, 7.1190e-06], device='cuda:0') 100 0.0001 changing lr epoch 166, time 217.43, cls_loss 0.0025 cls_loss_mapping 0.0035 cls_loss_causal 0.5349 re_mapping 0.0071 re_causal 0.0197 /// teacc 98.97 lr 0.00010000 Epoch 168, weight, value: tensor([[-0.0357, 0.1335, 0.0734, ..., -0.1774, -0.1099, 0.0240], [ 0.1275, -0.0163, -0.0560, ..., 0.0767, 0.1008, -0.1169], [-0.0743, -0.0547, -0.0511, ..., -0.0757, -0.0248, -0.0502], ..., [-0.0581, 0.1475, -0.1645, ..., -0.1299, 0.0035, -0.1406], [-0.0004, -0.1006, 0.0340, ..., 0.0551, -0.1072, -0.0416], [-0.0692, -0.0931, 0.0602, ..., -0.0830, -0.0564, 0.1256]], device='cuda:0'), grad: tensor([[ 5.7835e-07, 2.8173e-08, 1.7788e-06, ..., 8.5123e-07, 3.7719e-08, 2.1718e-06], [-1.6876e-06, 6.5845e-07, 1.1846e-06, ..., -1.6391e-07, -6.8359e-07, 7.0268e-07], [ 6.7279e-06, 5.0478e-06, 8.7395e-06, ..., 8.6725e-06, 6.6869e-07, 1.0617e-07], ..., [ 1.1250e-06, -1.4696e-06, 2.3730e-06, ..., 1.5050e-06, 2.5122e-07, 3.1646e-06], [-6.4671e-05, -3.6627e-05, -6.4552e-05, ..., -1.1295e-04, -6.5519e-07, 9.6262e-06], [ 3.2387e-07, 8.3167e-07, -2.1085e-05, ..., 3.2131e-06, 4.7032e-08, -3.9071e-05]], device='cuda:0') Epoch 168, bias, value: tensor([ 0.0022, 0.0011, 0.0059, 0.0118, 0.0267, 0.0413, -0.0227, 0.0092, -0.0375, -0.0222], device='cuda:0'), grad: tensor([ 6.4224e-06, -1.2377e-06, 4.0740e-05, 3.4094e-05, -5.6997e-06, 2.6965e-04, 5.0426e-05, 9.2760e-06, -3.3832e-04, -6.5327e-05], device='cuda:0') 100 0.0001 changing lr epoch 167, time 217.87, cls_loss 0.0027 cls_loss_mapping 0.0046 cls_loss_causal 0.5459 re_mapping 0.0072 re_causal 0.0200 /// teacc 99.00 lr 0.00010000 Epoch 169, weight, value: tensor([[-0.0361, 0.1339, 0.0732, ..., -0.1791, -0.1100, 0.0239], [ 0.1284, -0.0151, -0.0563, ..., 0.0767, 0.1030, -0.1174], [-0.0745, -0.0530, -0.0512, ..., -0.0760, -0.0248, -0.0508], ..., [-0.0601, 0.1460, -0.1649, ..., -0.1305, 0.0011, -0.1412], [ 0.0012, -0.1000, 0.0348, ..., 0.0574, -0.1074, -0.0409], [-0.0696, -0.0936, 0.0605, ..., -0.0830, -0.0568, 0.1262]], device='cuda:0'), grad: tensor([[ 1.4901e-07, 1.1129e-07, 2.9802e-08, ..., 3.2806e-07, 1.4738e-07, 1.9325e-08], [ 7.2410e-08, 3.3132e-07, 1.4016e-07, ..., 5.5227e-07, 2.4214e-07, 4.3074e-08], [ 1.3155e-07, -1.5842e-06, 1.4692e-07, ..., 1.7439e-07, 2.5379e-08, 1.8394e-08], ..., [ 1.7518e-06, 1.6182e-07, 2.9104e-07, ..., 4.0978e-06, 3.0454e-06, 6.2399e-08], [ 5.4315e-06, 2.4051e-07, 8.6846e-08, ..., 1.2390e-05, 9.3356e-06, -1.3132e-07], [ 3.7905e-07, 9.1689e-07, -1.0338e-07, ..., 5.8003e-06, 4.1211e-08, -2.1118e-07]], device='cuda:0') Epoch 169, bias, value: tensor([ 0.0018, 0.0022, 0.0074, 0.0114, 0.0261, 0.0410, -0.0234, 0.0070, -0.0360, -0.0219], device='cuda:0'), grad: tensor([ 1.3430e-06, 2.9858e-06, -9.8273e-06, 1.7220e-06, -1.0192e-05, -5.8234e-05, 1.2293e-06, 2.1070e-05, 3.8564e-05, 1.1407e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 168---------------------------------------------------- epoch 168, time 217.92, cls_loss 0.0019 cls_loss_mapping 0.0041 cls_loss_causal 0.5386 re_mapping 0.0074 re_causal 0.0211 /// teacc 99.03 lr 0.00010000 Epoch 170, weight, value: tensor([[-0.0363, 0.1344, 0.0733, ..., -0.1801, -0.1100, 0.0241], [ 0.1282, -0.0162, -0.0583, ..., 0.0766, 0.1029, -0.1193], [-0.0741, -0.0530, -0.0503, ..., -0.0759, -0.0244, -0.0513], ..., [-0.0599, 0.1473, -0.1648, ..., -0.1304, 0.0011, -0.1415], [ 0.0013, -0.1003, 0.0346, ..., 0.0573, -0.1075, -0.0411], [-0.0704, -0.0940, 0.0608, ..., -0.0834, -0.0570, 0.1267]], device='cuda:0'), grad: tensor([[ 3.1665e-08, -2.3516e-07, -1.1874e-07, ..., 6.8219e-08, 1.6298e-09, -4.2701e-07], [-6.9104e-07, 8.2655e-08, 5.4017e-08, ..., -6.0722e-07, -3.6554e-08, 1.1199e-07], [ 3.7532e-07, 5.6345e-08, 2.6077e-08, ..., 6.6683e-07, 1.8394e-08, 3.7719e-08], ..., [ 1.3504e-07, -1.3635e-06, 1.1316e-07, ..., 7.5530e-07, 3.7253e-09, 2.3120e-07], [-2.8461e-06, 4.6333e-08, -1.0608e-06, ..., -3.0342e-06, 6.9849e-09, -2.1625e-06], [ 1.9409e-06, 1.0459e-06, 7.1991e-07, ..., 1.7792e-05, 1.3970e-09, 1.5534e-06]], device='cuda:0') Epoch 170, bias, value: tensor([ 0.0019, 0.0008, 0.0082, 0.0113, 0.0261, 0.0409, -0.0228, 0.0075, -0.0362, -0.0218], device='cuda:0'), grad: tensor([-6.5472e-07, -4.5076e-07, 8.0978e-07, 3.4459e-07, -2.8580e-05, 1.6969e-06, 1.3411e-06, 5.7276e-08, -7.1600e-06, 3.2574e-05], device='cuda:0') 100 0.0001 changing lr epoch 169, time 216.63, cls_loss 0.0022 cls_loss_mapping 0.0031 cls_loss_causal 0.5380 re_mapping 0.0070 re_causal 0.0200 /// teacc 99.02 lr 0.00010000 Epoch 171, weight, value: tensor([[-0.0364, 0.1349, 0.0734, ..., -0.1808, -0.1101, 0.0240], [ 0.1282, -0.0165, -0.0589, ..., 0.0765, 0.1026, -0.1202], [-0.0744, -0.0534, -0.0504, ..., -0.0764, -0.0250, -0.0517], ..., [-0.0599, 0.1480, -0.1654, ..., -0.1305, 0.0021, -0.1420], [ 0.0014, -0.1007, 0.0347, ..., 0.0577, -0.1077, -0.0410], [-0.0708, -0.0942, 0.0612, ..., -0.0836, -0.0575, 0.1271]], device='cuda:0'), grad: tensor([[ 1.0375e-06, -3.1060e-07, 1.7742e-07, ..., 2.9299e-06, 3.2596e-09, -7.1712e-08], [-5.7602e-07, 4.0233e-07, 2.9057e-07, ..., -3.9767e-07, -2.3749e-08, 1.7276e-07], [ 4.6706e-07, 2.1467e-07, 1.4659e-06, ..., 1.2135e-06, -5.5879e-08, 1.3784e-07], ..., [ 9.7789e-08, -1.2666e-06, 1.0822e-06, ..., 3.6135e-07, 1.2107e-08, 1.1269e-07], [ 1.0803e-06, 9.4995e-08, 7.8753e-06, ..., 2.7865e-06, 3.2131e-08, 7.7439e-07], [ 1.1595e-07, 5.6298e-07, -2.1011e-05, ..., 4.1490e-07, 4.6566e-09, -1.8448e-05]], device='cuda:0') Epoch 171, bias, value: tensor([ 1.8011e-03, 7.5177e-05, 8.0893e-03, 1.1034e-02, 2.6050e-02, 4.0835e-02, -2.2656e-02, 8.3115e-03, -3.6115e-02, -2.1732e-02], device='cuda:0'), grad: tensor([ 4.4107e-06, 3.5483e-07, 2.5071e-06, -4.0866e-06, 1.0453e-05, 1.2271e-05, -2.3261e-05, 1.2871e-06, 2.2933e-05, -2.6867e-05], device='cuda:0') 100 0.0001 changing lr epoch 170, time 216.84, cls_loss 0.0023 cls_loss_mapping 0.0037 cls_loss_causal 0.5338 re_mapping 0.0066 re_causal 0.0197 /// teacc 98.96 lr 0.00010000 Epoch 172, weight, value: tensor([[-0.0366, 0.1358, 0.0737, ..., -0.1818, -0.1103, 0.0243], [ 0.1303, -0.0137, -0.0591, ..., 0.0796, 0.1025, -0.1207], [-0.0748, -0.0523, -0.0506, ..., -0.0770, -0.0254, -0.0525], ..., [-0.0624, 0.1454, -0.1656, ..., -0.1343, 0.0025, -0.1422], [ 0.0016, -0.1011, 0.0346, ..., 0.0581, -0.1077, -0.0410], [-0.0714, -0.0947, 0.0614, ..., -0.0840, -0.0577, 0.1274]], device='cuda:0'), grad: tensor([[-1.1399e-06, -2.2501e-05, -1.1101e-05, ..., -3.4180e-06, 9.3132e-10, -1.9491e-05], [-4.7684e-07, -4.6566e-09, 4.5169e-08, ..., -4.8894e-07, -1.3970e-09, 6.1002e-08], [ 5.5879e-08, -4.3493e-07, 1.0664e-07, ..., -6.6124e-08, -9.3132e-10, 6.7521e-08], ..., [ 3.6322e-07, 1.2573e-08, 1.3597e-07, ..., 4.9639e-07, 1.3970e-09, 9.8255e-08], [-3.4451e-05, 6.5425e-07, 5.9046e-07, ..., -6.5029e-05, 1.2573e-08, -1.0967e-05], [ 8.1491e-08, 4.2887e-07, 1.1316e-07, ..., 1.4901e-07, 2.3283e-09, 1.6904e-07]], device='cuda:0') Epoch 172, bias, value: tensor([ 0.0020, 0.0017, 0.0081, 0.0113, 0.0262, 0.0402, -0.0218, 0.0068, -0.0361, -0.0219], device='cuda:0'), grad: tensor([-4.7833e-05, -2.7614e-07, -5.1633e-06, 1.5469e-06, 1.2433e-07, 5.3644e-05, 8.9228e-05, 2.4866e-06, -9.4652e-05, 9.4390e-07], device='cuda:0') 100 0.0001 changing lr epoch 171, time 217.17, cls_loss 0.0017 cls_loss_mapping 0.0036 cls_loss_causal 0.5369 re_mapping 0.0068 re_causal 0.0200 /// teacc 98.99 lr 0.00010000 Epoch 173, weight, value: tensor([[-0.0368, 0.1365, 0.0742, ..., -0.1826, -0.1103, 0.0246], [ 0.1301, -0.0141, -0.0595, ..., 0.0796, 0.1028, -0.1212], [-0.0752, -0.0526, -0.0506, ..., -0.0775, -0.0257, -0.0531], ..., [-0.0625, 0.1459, -0.1661, ..., -0.1345, 0.0027, -0.1436], [ 0.0018, -0.1014, 0.0347, ..., 0.0584, -0.1079, -0.0411], [-0.0718, -0.0946, 0.0615, ..., -0.0842, -0.0582, 0.1277]], device='cuda:0'), grad: tensor([[ 1.6112e-07, -1.2685e-06, -1.0468e-06, ..., 1.5134e-07, 1.8161e-08, -1.7826e-06], [-1.1884e-06, 1.8906e-07, -3.6741e-07, ..., -6.1095e-07, -1.2191e-06, 3.3062e-08], [ 8.6101e-07, 2.5053e-07, 3.2596e-07, ..., 7.6927e-07, 8.9267e-07, 3.0734e-08], ..., [ 2.2957e-07, -5.6392e-07, 6.4727e-08, ..., 4.5821e-07, 2.0675e-07, 4.5914e-07], [ 9.4995e-08, 3.4273e-07, 4.5542e-07, ..., 1.7881e-07, 1.9092e-08, 6.7381e-07], [ 1.9418e-07, 1.0412e-06, 5.7695e-07, ..., 3.9907e-07, 2.0768e-07, 8.8522e-07]], device='cuda:0') Epoch 173, bias, value: tensor([ 0.0023, 0.0014, 0.0082, 0.0116, 0.0263, 0.0398, -0.0218, 0.0069, -0.0360, -0.0219], device='cuda:0'), grad: tensor([-3.8035e-06, -1.5264e-06, 2.4624e-06, -4.9025e-06, -2.2016e-06, 3.9451e-06, 2.6589e-07, 5.0524e-07, 1.6503e-06, 3.5893e-06], device='cuda:0') 100 0.0001 changing lr epoch 172, time 217.46, cls_loss 0.0020 cls_loss_mapping 0.0035 cls_loss_causal 0.5576 re_mapping 0.0071 re_causal 0.0206 /// teacc 98.97 lr 0.00010000 Epoch 174, weight, value: tensor([[-0.0373, 0.1370, 0.0753, ..., -0.1841, -0.1095, 0.0251], [ 0.1303, -0.0145, -0.0600, ..., 0.0797, 0.1035, -0.1225], [-0.0758, -0.0531, -0.0507, ..., -0.0779, -0.0266, -0.0537], ..., [-0.0623, 0.1468, -0.1664, ..., -0.1345, 0.0027, -0.1441], [ 0.0017, -0.1020, 0.0348, ..., 0.0584, -0.1080, -0.0411], [-0.0724, -0.0951, 0.0612, ..., -0.0847, -0.0601, 0.1276]], device='cuda:0'), grad: tensor([[ 1.2713e-07, -1.9409e-06, -1.3104e-06, ..., 5.2433e-07, 2.2817e-08, -1.6671e-06], [ 1.3076e-05, 5.1688e-08, 7.4971e-08, ..., 2.9683e-05, 2.9244e-06, 8.7544e-08], [ 6.9337e-07, -1.2480e-07, 2.6822e-07, ..., 2.0731e-06, 9.3132e-08, 2.8312e-07], ..., [ 9.4483e-07, -1.8533e-07, 8.5216e-08, ..., 2.0619e-06, 1.2247e-07, 1.2852e-07], [-1.6540e-05, 2.0256e-07, -9.3132e-08, ..., -3.4750e-05, -3.4645e-06, 4.3726e-07], [ 1.4296e-07, 9.3179e-07, 1.0943e-07, ..., 9.0385e-07, 1.8626e-08, -9.7975e-07]], device='cuda:0') Epoch 174, bias, value: tensor([ 0.0029, 0.0014, 0.0078, 0.0117, 0.0264, 0.0399, -0.0218, 0.0072, -0.0360, -0.0223], device='cuda:0'), grad: tensor([-4.6454e-06, 5.5760e-05, 2.8703e-06, 1.0934e-06, -1.3642e-05, 1.8273e-06, 1.1794e-05, 5.0329e-06, -6.2823e-05, 2.7288e-06], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 173---------------------------------------------------- epoch 173, time 217.45, cls_loss 0.0023 cls_loss_mapping 0.0031 cls_loss_causal 0.5156 re_mapping 0.0068 re_causal 0.0190 /// teacc 99.04 lr 0.00010000 Epoch 175, weight, value: tensor([[-0.0378, 0.1381, 0.0757, ..., -0.1859, -0.1094, 0.0253], [ 0.1305, -0.0146, -0.0618, ..., 0.0795, 0.1039, -0.1266], [-0.0762, -0.0537, -0.0508, ..., -0.0783, -0.0271, -0.0543], ..., [-0.0627, 0.1471, -0.1669, ..., -0.1349, 0.0028, -0.1456], [ 0.0027, -0.1012, 0.0348, ..., 0.0595, -0.1082, -0.0406], [-0.0720, -0.0951, 0.0617, ..., -0.0842, -0.0607, 0.1295]], device='cuda:0'), grad: tensor([[ 2.4028e-06, 1.5367e-08, -8.3819e-09, ..., 9.3356e-06, 0.0000e+00, -1.2573e-08], [-2.1271e-06, -1.2042e-06, 2.7008e-08, ..., -5.1521e-06, 0.0000e+00, 3.3993e-08], [ 2.4196e-06, 1.1735e-06, 1.0477e-07, ..., 6.0759e-06, 0.0000e+00, 9.4529e-08], ..., [ 6.3796e-07, -1.0803e-06, 3.0734e-08, ..., 1.2126e-06, 0.0000e+00, 4.2841e-08], [ 3.6918e-06, 2.2585e-07, -1.8952e-07, ..., 5.4240e-06, 0.0000e+00, -4.5355e-07], [ 2.5099e-07, 4.9500e-07, 1.5367e-08, ..., 9.5274e-07, 0.0000e+00, 1.6764e-07]], device='cuda:0') Epoch 175, bias, value: tensor([ 0.0032, 0.0009, 0.0075, 0.0125, 0.0259, 0.0391, -0.0220, 0.0067, -0.0355, -0.0213], device='cuda:0'), grad: tensor([ 1.6049e-05, -1.3143e-05, 1.5259e-05, 1.8887e-06, 3.3736e-05, 3.8818e-06, -7.4983e-05, 1.1083e-06, 1.3642e-05, 2.5425e-06], device='cuda:0') 100 0.0001 changing lr epoch 174, time 217.55, cls_loss 0.0025 cls_loss_mapping 0.0040 cls_loss_causal 0.5165 re_mapping 0.0064 re_causal 0.0184 /// teacc 98.99 lr 0.00010000 Epoch 176, weight, value: tensor([[-0.0394, 0.1376, 0.0758, ..., -0.1889, -0.1097, 0.0252], [ 0.1317, -0.0149, -0.0622, ..., 0.0806, 0.1043, -0.1272], [-0.0767, -0.0542, -0.0509, ..., -0.0787, -0.0275, -0.0547], ..., [-0.0629, 0.1485, -0.1671, ..., -0.1351, 0.0027, -0.1466], [ 0.0022, -0.1024, 0.0344, ..., 0.0589, -0.1083, -0.0411], [-0.0714, -0.0960, 0.0623, ..., -0.0845, -0.0609, 0.1305]], device='cuda:0'), grad: tensor([[ 7.8697e-08, -1.6578e-07, 7.8082e-06, ..., 4.0494e-06, 0.0000e+00, 8.9109e-06], [-1.6997e-07, 8.0094e-08, 1.1828e-07, ..., -9.0804e-08, -6.0536e-09, 1.3225e-07], [ 3.6787e-08, 8.4750e-08, 7.3109e-08, ..., 8.8010e-08, 9.3132e-10, 7.8231e-08], ..., [ 1.4063e-07, -3.6694e-07, 5.3085e-08, ..., 2.2212e-07, 2.3283e-09, 5.8673e-08], [ 1.1548e-07, 7.4971e-08, 1.4508e-04, ..., 7.0274e-05, 1.3970e-09, 1.6499e-04], [ 4.5169e-08, 3.1106e-07, -1.5461e-04, ..., -7.4089e-05, 9.3132e-10, -1.7560e-04]], device='cuda:0') Epoch 176, bias, value: tensor([ 0.0026, 0.0012, 0.0072, 0.0122, 0.0259, 0.0395, -0.0221, 0.0072, -0.0362, -0.0211], device='cuda:0'), grad: tensor([ 1.8090e-05, 4.9360e-08, 3.4226e-07, 1.6317e-06, 7.2783e-07, -1.5199e-06, 2.9337e-08, 5.1688e-08, 3.3307e-04, -3.5262e-04], device='cuda:0') 100 0.0001 changing lr epoch 175, time 217.39, cls_loss 0.0028 cls_loss_mapping 0.0042 cls_loss_causal 0.5232 re_mapping 0.0068 re_causal 0.0199 /// teacc 99.02 lr 0.00010000 Epoch 177, weight, value: tensor([[-0.0398, 0.1381, 0.0759, ..., -0.1898, -0.1097, 0.0253], [ 0.1319, -0.0149, -0.0636, ..., 0.0805, 0.1034, -0.1280], [-0.0765, -0.0540, -0.0497, ..., -0.0784, -0.0266, -0.0552], ..., [-0.0631, 0.1488, -0.1677, ..., -0.1353, 0.0034, -0.1476], [ 0.0020, -0.1032, 0.0339, ..., 0.0588, -0.1085, -0.0418], [-0.0722, -0.0967, 0.0627, ..., -0.0850, -0.0613, 0.1312]], device='cuda:0'), grad: tensor([[ 1.7369e-07, -1.2619e-07, 2.0117e-07, ..., 1.2992e-07, 1.3970e-09, -8.0094e-08], [-4.5076e-07, -1.6298e-08, 2.4308e-07, ..., -6.5938e-07, -3.9116e-08, -9.7789e-09], [ 2.2305e-07, 4.6100e-08, 2.0349e-07, ..., 1.0943e-07, 6.0536e-09, 1.4435e-08], ..., [ 2.2165e-07, -3.1991e-07, 3.4133e-07, ..., 1.6158e-07, 6.9849e-09, 9.7789e-09], [ 1.2871e-06, 4.8429e-08, 2.1290e-06, ..., 9.4343e-07, 6.9849e-09, 1.7835e-07], [ 1.8813e-07, 2.2491e-07, 1.8021e-07, ..., 2.0256e-07, 7.9162e-09, -1.1642e-08]], device='cuda:0') Epoch 177, bias, value: tensor([ 0.0026, 0.0001, 0.0085, 0.0091, 0.0257, 0.0430, -0.0231, 0.0075, -0.0369, -0.0210], device='cuda:0'), grad: tensor([ 6.1421e-07, -3.7625e-07, 8.0373e-07, -1.0860e-04, 2.1001e-07, 9.8705e-05, -1.3988e-06, 7.6694e-07, 8.0466e-06, 1.2396e-06], device='cuda:0') 100 0.0001 changing lr epoch 176, time 217.16, cls_loss 0.0028 cls_loss_mapping 0.0041 cls_loss_causal 0.5138 re_mapping 0.0065 re_causal 0.0188 /// teacc 98.97 lr 0.00010000 Epoch 178, weight, value: tensor([[-0.0426, 0.1362, 0.0754, ..., -0.1922, -0.1132, 0.0244], [ 0.1329, -0.0151, -0.0627, ..., 0.0818, 0.1041, -0.1274], [-0.0775, -0.0529, -0.0497, ..., -0.0798, -0.0275, -0.0557], ..., [-0.0632, 0.1491, -0.1679, ..., -0.1359, 0.0036, -0.1479], [ 0.0021, -0.1034, 0.0338, ..., 0.0590, -0.1086, -0.0420], [-0.0748, -0.0974, 0.0629, ..., -0.0859, -0.0618, 0.1317]], device='cuda:0'), grad: tensor([[-2.0582e-07, -1.3746e-06, -9.0245e-07, ..., -6.8918e-07, -4.1956e-07, -3.8520e-06], [ 4.7032e-08, 1.9558e-08, 5.2620e-08, ..., 2.6263e-07, 2.5611e-08, 1.1362e-07], [ 2.7474e-08, 3.3062e-08, 3.9581e-08, ..., 1.6978e-06, 1.2573e-08, 4.9360e-08], ..., [ 2.9337e-08, -4.7963e-08, 4.8429e-08, ..., 7.9628e-07, 7.9162e-09, 1.1595e-07], [ 9.7416e-07, 1.9558e-08, 5.4343e-07, ..., -2.0210e-07, 1.0245e-07, 1.7844e-06], [ 7.5437e-08, 4.0047e-08, -6.3330e-08, ..., 1.5991e-06, 2.9802e-08, -2.5565e-07]], device='cuda:0') Epoch 178, bias, value: tensor([ 0.0001, 0.0004, 0.0092, 0.0083, 0.0266, 0.0438, -0.0226, 0.0075, -0.0373, -0.0214], device='cuda:0'), grad: tensor([-2.7508e-05, 8.1724e-07, 9.4026e-06, 1.3094e-06, -4.9807e-06, -4.3996e-06, 2.7567e-05, 1.8477e-06, -7.2196e-06, 3.1553e-06], device='cuda:0') 100 0.0001 changing lr epoch 177, time 217.13, cls_loss 0.0023 cls_loss_mapping 0.0026 cls_loss_causal 0.5396 re_mapping 0.0065 re_causal 0.0193 /// teacc 98.97 lr 0.00010000 Epoch 179, weight, value: tensor([[-0.0429, 0.1379, 0.0743, ..., -0.1924, -0.1131, 0.0227], [ 0.1334, -0.0153, -0.0630, ..., 0.0820, 0.1044, -0.1278], [-0.0780, -0.0508, -0.0496, ..., -0.0802, -0.0273, -0.0550], ..., [-0.0634, 0.1477, -0.1682, ..., -0.1363, 0.0035, -0.1485], [ 0.0020, -0.1040, 0.0336, ..., 0.0589, -0.1088, -0.0423], [-0.0755, -0.0979, 0.0635, ..., -0.0867, -0.0623, 0.1330]], device='cuda:0'), grad: tensor([[ 2.3283e-08, -3.5577e-07, -1.2619e-07, ..., 3.9116e-08, 4.6566e-09, -1.2200e-07], [-8.7079e-08, 1.5367e-08, 3.9581e-08, ..., -1.0058e-07, 5.1223e-09, 8.0094e-08], [ 2.6543e-08, 8.1956e-08, 4.0513e-08, ..., 3.6787e-08, 9.3132e-10, 4.8429e-08], ..., [ 2.1420e-08, -6.5193e-08, 1.1129e-07, ..., 3.1199e-08, 3.2596e-09, 1.9092e-07], [ 5.7742e-08, 2.9802e-08, 7.3202e-07, ..., 1.0384e-07, 2.8871e-08, 1.2880e-06], [ 3.8650e-08, 9.5461e-08, -1.3756e-06, ..., 6.8918e-08, 8.3819e-09, -2.3767e-06]], device='cuda:0') Epoch 179, bias, value: tensor([-0.0005, 0.0003, 0.0110, 0.0087, 0.0272, 0.0434, -0.0230, 0.0063, -0.0377, -0.0211], device='cuda:0'), grad: tensor([-7.9675e-07, -4.6566e-10, 9.7323e-08, 5.4343e-07, 1.5870e-06, -1.1194e-06, 8.9919e-07, 2.8918e-07, 2.0564e-06, -3.5651e-06], device='cuda:0') 100 0.0001 changing lr epoch 178, time 217.81, cls_loss 0.0022 cls_loss_mapping 0.0035 cls_loss_causal 0.4831 re_mapping 0.0066 re_causal 0.0186 /// teacc 98.81 lr 0.00010000 Epoch 180, weight, value: tensor([[-0.0429, 0.1385, 0.0746, ..., -0.1927, -0.1131, 0.0228], [ 0.1352, -0.0147, -0.0633, ..., 0.0841, 0.1045, -0.1281], [-0.0787, -0.0507, -0.0497, ..., -0.0806, -0.0275, -0.0551], ..., [-0.0643, 0.1480, -0.1685, ..., -0.1377, 0.0042, -0.1486], [ 0.0015, -0.1051, 0.0334, ..., 0.0586, -0.1094, -0.0425], [-0.0777, -0.0990, 0.0639, ..., -0.0877, -0.0624, 0.1335]], device='cuda:0'), grad: tensor([[ 1.3225e-07, -1.8813e-07, -6.7521e-08, ..., 1.1083e-07, 2.3283e-09, -4.5169e-08], [ 1.7714e-06, 4.9919e-06, 3.5763e-07, ..., 1.7509e-07, 3.0734e-08, 4.4005e-07], [ 4.4703e-08, 2.3562e-07, -6.0536e-09, ..., 7.5903e-08, 5.1223e-09, 2.3749e-08], ..., [-3.4124e-06, -1.2845e-05, -1.9139e-07, ..., 1.6158e-07, -1.2573e-07, 7.1712e-08], [ 3.5074e-06, 1.1269e-07, 1.8738e-06, ..., 1.5516e-06, 1.0431e-07, 3.2485e-06], [ 2.2221e-06, 6.9663e-06, 4.0513e-07, ..., 6.5938e-07, 1.2200e-07, 3.3667e-07]], device='cuda:0') Epoch 180, bias, value: tensor([-0.0003, 0.0008, 0.0111, 0.0089, 0.0271, 0.0433, -0.0230, 0.0065, -0.0384, -0.0212], device='cuda:0'), grad: tensor([-3.7858e-07, 8.5309e-06, -7.6788e-07, 8.0233e-07, -1.9977e-07, -9.6858e-06, 3.1665e-06, -1.9491e-05, 5.5023e-06, 1.2480e-05], device='cuda:0') 100 0.0001 changing lr epoch 179, time 216.95, cls_loss 0.0021 cls_loss_mapping 0.0029 cls_loss_causal 0.5251 re_mapping 0.0068 re_causal 0.0198 /// teacc 98.94 lr 0.00010000 Epoch 181, weight, value: tensor([[-0.0431, 0.1391, 0.0740, ..., -0.1933, -0.1131, 0.0222], [ 0.1357, -0.0151, -0.0630, ..., 0.0847, 0.1049, -0.1285], [-0.0780, -0.0510, -0.0499, ..., -0.0812, -0.0276, -0.0557], ..., [-0.0645, 0.1488, -0.1689, ..., -0.1380, 0.0044, -0.1490], [ 0.0015, -0.1054, 0.0333, ..., 0.0588, -0.1094, -0.0427], [-0.0782, -0.0994, 0.0643, ..., -0.0882, -0.0631, 0.1343]], device='cuda:0'), grad: tensor([[ 5.1223e-08, -8.8885e-06, -8.8811e-06, ..., -2.0005e-06, 4.3306e-08, -1.0006e-05], [-4.8190e-05, 4.0531e-06, 2.6636e-07, ..., -1.5438e-05, -9.1970e-05, 3.2270e-07], [ 2.4483e-05, 1.2936e-06, 9.9000e-07, ..., 8.2329e-06, 4.6700e-05, 1.0943e-06], ..., [ 2.2918e-05, -9.6709e-06, 1.1269e-07, ..., 7.4431e-06, 4.2677e-05, 1.1455e-07], [-2.5928e-06, 3.3006e-06, -2.5500e-06, ..., -4.0904e-06, 6.7893e-07, -3.5260e-06], [ 2.4661e-06, 4.5337e-06, 4.8727e-06, ..., 3.9861e-06, 9.3179e-07, 6.1542e-06]], device='cuda:0') Epoch 181, bias, value: tensor([-0.0005, 0.0006, 0.0114, 0.0090, 0.0271, 0.0430, -0.0237, 0.0069, -0.0385, -0.0211], device='cuda:0'), grad: tensor([-2.8849e-05, -1.5187e-04, 8.5890e-05, 2.1942e-06, 1.0571e-06, 1.8245e-06, 1.4797e-05, 5.4926e-05, -3.2000e-06, 2.3350e-05], device='cuda:0') 100 0.0001 changing lr epoch 180, time 217.46, cls_loss 0.0017 cls_loss_mapping 0.0041 cls_loss_causal 0.5551 re_mapping 0.0065 re_causal 0.0190 /// teacc 99.01 lr 0.00010000 Epoch 182, weight, value: tensor([[-0.0433, 0.1395, 0.0743, ..., -0.1938, -0.1132, 0.0224], [ 0.1365, -0.0151, -0.0642, ..., 0.0851, 0.1056, -0.1305], [-0.0794, -0.0511, -0.0502, ..., -0.0823, -0.0282, -0.0570], ..., [-0.0646, 0.1493, -0.1692, ..., -0.1382, 0.0041, -0.1481], [ 0.0016, -0.1056, 0.0333, ..., 0.0590, -0.1095, -0.0427], [-0.0793, -0.1004, 0.0651, ..., -0.0881, -0.0632, 0.1355]], device='cuda:0'), grad: tensor([[ 3.0082e-07, 1.8254e-07, 1.2480e-07, ..., 1.0105e-07, 0.0000e+00, 5.5274e-07], [ 4.7730e-07, 2.0117e-07, 3.2037e-07, ..., 4.8941e-07, -1.8626e-09, 1.4529e-07], [ 3.6396e-06, 7.8790e-07, 1.4342e-06, ..., 3.7104e-06, -4.6566e-10, 2.3143e-07], ..., [ 1.1735e-06, 9.4203e-07, 1.5646e-06, ..., 3.4552e-07, 1.3970e-09, 1.1660e-06], [-1.1586e-05, 3.4738e-07, -1.8962e-06, ..., -1.4767e-05, 0.0000e+00, 2.5984e-06], [ 1.7229e-07, 9.6858e-08, -4.2319e-06, ..., 3.0966e-07, 0.0000e+00, -5.9828e-06]], device='cuda:0') Epoch 182, bias, value: tensor([-0.0003, 0.0005, 0.0108, 0.0092, 0.0269, 0.0429, -0.0238, 0.0072, -0.0384, -0.0207], device='cuda:0'), grad: tensor([ 1.8682e-06, 2.3823e-06, 1.3389e-05, 3.4422e-06, 2.5239e-06, 9.9391e-06, 3.0287e-06, 7.7486e-06, -3.4869e-05, -9.4101e-06], device='cuda:0') 100 0.0001 changing lr epoch 181, time 217.56, cls_loss 0.0019 cls_loss_mapping 0.0036 cls_loss_causal 0.5393 re_mapping 0.0061 re_causal 0.0190 /// teacc 99.00 lr 0.00010000 Epoch 183, weight, value: tensor([[-0.0436, 0.1398, 0.0748, ..., -0.1942, -0.1131, 0.0228], [ 0.1375, -0.0147, -0.0647, ..., 0.0856, 0.1060, -0.1313], [-0.0798, -0.0513, -0.0505, ..., -0.0827, -0.0284, -0.0577], ..., [-0.0658, 0.1493, -0.1695, ..., -0.1389, 0.0039, -0.1490], [ 0.0020, -0.1059, 0.0334, ..., 0.0593, -0.1096, -0.0427], [-0.0800, -0.1012, 0.0655, ..., -0.0887, -0.0634, 0.1359]], device='cuda:0'), grad: tensor([[ 1.3411e-07, -7.6881e-07, -8.3959e-07, ..., 8.9873e-08, -7.0315e-08, -1.1865e-06], [-2.7893e-07, 6.7912e-06, -1.0952e-06, ..., -2.5518e-06, -4.1910e-09, -7.5065e-07], [ 3.0315e-07, 5.4669e-07, 9.4064e-08, ..., 1.9465e-07, 1.3970e-09, 5.3085e-08], ..., [-1.6652e-06, -7.7188e-05, -4.4070e-06, ..., 1.6997e-07, 2.3283e-09, 5.4017e-08], [ 7.3807e-07, 1.5106e-06, 5.1735e-07, ..., 1.1362e-07, 9.3132e-10, 1.1269e-07], [ 1.7509e-06, 6.7592e-05, 6.2920e-06, ..., 2.1830e-06, 4.2375e-08, 1.4547e-06]], device='cuda:0') Epoch 183, bias, value: tensor([-2.5532e-05, 7.0022e-04, 1.0523e-02, 9.2398e-03, 2.7203e-02, 4.3542e-02, -2.5544e-02, 6.8184e-03, -3.8187e-02, -2.0872e-02], device='cuda:0'), grad: tensor([-2.3451e-06, 7.9349e-06, 1.4678e-06, -9.8161e-07, 1.4585e-06, 1.2666e-07, -6.2725e-07, -1.2887e-04, 3.7923e-06, 1.1808e-04], device='cuda:0') 100 0.0001 changing lr epoch 182, time 217.45, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.5278 re_mapping 0.0064 re_causal 0.0196 /// teacc 98.98 lr 0.00010000 Epoch 184, weight, value: tensor([[-0.0442, 0.1403, 0.0751, ..., -0.1951, -0.1133, 0.0229], [ 0.1385, -0.0144, -0.0646, ..., 0.0861, 0.1062, -0.1316], [-0.0802, -0.0515, -0.0508, ..., -0.0829, -0.0286, -0.0583], ..., [-0.0660, 0.1504, -0.1687, ..., -0.1392, 0.0038, -0.1489], [ 0.0017, -0.1073, 0.0334, ..., 0.0593, -0.1097, -0.0428], [-0.0821, -0.1035, 0.0653, ..., -0.0893, -0.0635, 0.1358]], device='cuda:0'), grad: tensor([[ 2.3283e-08, -1.3504e-08, 2.9337e-08, ..., 2.1886e-08, 4.6566e-10, 2.7474e-08], [-3.7253e-08, -4.6566e-10, 1.0524e-07, ..., -1.8161e-08, -7.4506e-09, 1.2200e-07], [ 2.3283e-08, 4.6566e-09, 4.4238e-08, ..., 2.0489e-08, 3.2596e-09, 1.2107e-08], ..., [ 4.1910e-08, 3.3528e-08, 2.7474e-07, ..., 5.5414e-08, 7.4506e-09, 2.6077e-07], [ 1.2806e-07, 2.9802e-08, 3.4925e-07, ..., 1.4016e-07, 1.3970e-09, 4.6985e-07], [ 4.9127e-07, -9.4529e-08, -2.2398e-07, ..., 6.8778e-07, 9.3132e-10, 2.3888e-07]], device='cuda:0') Epoch 184, bias, value: tensor([ 9.9534e-05, 9.5707e-04, 1.0190e-02, 9.1858e-03, 2.7522e-02, 4.3569e-02, -2.5344e-02, 7.6505e-03, -3.8411e-02, -2.1817e-02], device='cuda:0'), grad: tensor([ 9.7323e-08, 1.8580e-07, -4.0978e-08, -9.7789e-08, 3.3574e-07, -2.7232e-06, 9.4716e-07, 7.4273e-07, 8.0420e-07, -2.4447e-07], device='cuda:0') 100 0.0001 changing lr epoch 183, time 217.52, cls_loss 0.0029 cls_loss_mapping 0.0046 cls_loss_causal 0.5287 re_mapping 0.0065 re_causal 0.0185 /// teacc 98.92 lr 0.00010000 Epoch 185, weight, value: tensor([[-0.0457, 0.1401, 0.0750, ..., -0.1976, -0.1135, 0.0228], [ 0.1411, -0.0128, -0.0654, ..., 0.0874, 0.1075, -0.1324], [-0.0816, -0.0517, -0.0511, ..., -0.0838, -0.0296, -0.0590], ..., [-0.0698, 0.1474, -0.1693, ..., -0.1424, 0.0036, -0.1497], [ 0.0047, -0.1045, 0.0336, ..., 0.0615, -0.1099, -0.0428], [-0.0841, -0.1046, 0.0658, ..., -0.0903, -0.0637, 0.1376]], device='cuda:0'), grad: tensor([[ 7.0035e-07, -2.3190e-07, -5.2107e-07, ..., 3.8883e-07, 1.1642e-08, -1.0859e-06], [-3.1497e-06, 1.2321e-06, 3.5111e-07, ..., -2.2203e-06, -1.3225e-07, 4.5402e-07], [ 1.8207e-07, 3.5437e-07, 7.2177e-08, ..., 3.5437e-07, -1.8626e-09, 4.9360e-08], ..., [ 9.8720e-08, -1.5780e-05, -2.7865e-06, ..., 2.7847e-07, 5.1223e-09, 6.7055e-08], [ 1.5572e-06, 6.0629e-07, 2.6217e-07, ..., 1.8291e-06, 1.0850e-07, 5.8021e-07], [ 6.2119e-07, 3.7439e-06, 1.7621e-06, ..., 7.7114e-06, 5.1223e-09, 4.4145e-06]], device='cuda:0') Epoch 185, bias, value: tensor([-0.0005, 0.0025, 0.0098, 0.0091, 0.0280, 0.0436, -0.0250, 0.0047, -0.0356, -0.0221], device='cuda:0'), grad: tensor([-6.3516e-07, -6.2026e-07, 1.1632e-06, 2.3231e-05, -2.0981e-05, -7.3433e-05, 7.5936e-05, -3.2812e-05, 4.9658e-06, 2.3171e-05], device='cuda:0') 100 0.0001 changing lr epoch 184, time 217.35, cls_loss 0.0017 cls_loss_mapping 0.0030 cls_loss_causal 0.5288 re_mapping 0.0067 re_causal 0.0195 /// teacc 98.96 lr 0.00010000 Epoch 186, weight, value: tensor([[-0.0462, 0.1401, 0.0750, ..., -0.1985, -0.1136, 0.0228], [ 0.1413, -0.0127, -0.0665, ..., 0.0873, 0.1077, -0.1331], [-0.0820, -0.0516, -0.0504, ..., -0.0837, -0.0295, -0.0599], ..., [-0.0699, 0.1476, -0.1695, ..., -0.1426, 0.0039, -0.1496], [ 0.0049, -0.1045, 0.0338, ..., 0.0622, -0.1101, -0.0427], [-0.0846, -0.1056, 0.0662, ..., -0.0902, -0.0642, 0.1381]], device='cuda:0'), grad: tensor([[ 1.5879e-07, -6.6683e-06, -3.1069e-06, ..., 1.3690e-07, -1.8161e-08, -2.1905e-06], [-5.9307e-06, -4.3446e-07, 3.8184e-08, ..., -9.6411e-06, 4.6566e-10, 1.3970e-08], [ 5.5088e-07, 9.7752e-06, 2.6729e-06, ..., 7.7672e-07, 2.3283e-09, 1.8943e-06], ..., [ 9.9465e-07, -5.4017e-06, 6.2399e-08, ..., 1.4044e-06, 4.6566e-10, 1.9092e-08], [ 3.8408e-06, 1.3430e-06, 2.2305e-07, ..., 6.5416e-06, 2.7940e-09, 3.6228e-07], [ 1.3830e-07, 8.6427e-07, 2.8452e-07, ..., 2.1886e-07, 6.5193e-09, 1.7881e-07]], device='cuda:0') Epoch 186, bias, value: tensor([-0.0006, 0.0023, 0.0104, 0.0090, 0.0277, 0.0436, -0.0251, 0.0047, -0.0355, -0.0219], device='cuda:0'), grad: tensor([-1.7717e-05, -1.1146e-05, 1.1548e-05, -1.1129e-06, 1.1679e-06, 3.8818e-06, -1.2061e-07, -1.7369e-06, 1.2234e-05, 2.9765e-06], device='cuda:0') 100 0.0001 changing lr epoch 185, time 216.97, cls_loss 0.0021 cls_loss_mapping 0.0033 cls_loss_causal 0.5310 re_mapping 0.0066 re_causal 0.0193 /// teacc 98.97 lr 0.00010000 Epoch 187, weight, value: tensor([[-0.0464, 0.1398, 0.0749, ..., -0.1992, -0.1136, 0.0228], [ 0.1415, -0.0131, -0.0667, ..., 0.0873, 0.1082, -0.1332], [-0.0825, -0.0518, -0.0505, ..., -0.0840, -0.0299, -0.0603], ..., [-0.0700, 0.1487, -0.1687, ..., -0.1427, 0.0037, -0.1488], [ 0.0055, -0.1044, 0.0343, ..., 0.0634, -0.1101, -0.0424], [-0.0851, -0.1087, 0.0659, ..., -0.0904, -0.0644, 0.1382]], device='cuda:0'), grad: tensor([[ 1.5507e-07, 7.4040e-08, 4.1910e-09, ..., 2.6077e-07, 4.6566e-10, 1.3039e-08], [-1.6065e-07, 5.6345e-08, 2.8871e-08, ..., -2.1281e-07, 4.8429e-08, 5.0757e-08], [-1.3225e-07, 2.9802e-08, 2.3749e-08, ..., 1.2293e-07, -1.4715e-07, 1.2573e-08], ..., [ 1.5879e-07, -2.8079e-07, 4.4703e-08, ..., 1.0850e-07, 4.7963e-08, 8.0094e-08], [ 2.6952e-06, 1.6969e-06, 2.1234e-07, ..., 4.8093e-06, 2.7940e-09, 4.5029e-07], [ 3.1665e-08, 1.8207e-07, -9.5041e-07, ..., 5.2154e-08, 9.3132e-10, -1.8040e-06]], device='cuda:0') Epoch 187, bias, value: tensor([-0.0009, 0.0021, 0.0102, 0.0088, 0.0278, 0.0438, -0.0266, 0.0058, -0.0349, -0.0229], device='cuda:0'), grad: tensor([ 5.4156e-07, 2.2855e-06, -7.0594e-06, 1.2079e-06, 1.9781e-06, 4.3400e-07, -9.3579e-06, 2.3395e-06, 9.8422e-06, -2.2054e-06], device='cuda:0') 100 0.0001 changing lr epoch 186, time 217.30, cls_loss 0.0022 cls_loss_mapping 0.0029 cls_loss_causal 0.5668 re_mapping 0.0068 re_causal 0.0195 /// teacc 98.91 lr 0.00010000 Epoch 188, weight, value: tensor([[-0.0468, 0.1403, 0.0750, ..., -0.2001, -0.1136, 0.0230], [ 0.1420, -0.0137, -0.0668, ..., 0.0874, 0.1088, -0.1333], [-0.0833, -0.0524, -0.0508, ..., -0.0845, -0.0302, -0.0611], ..., [-0.0699, 0.1494, -0.1696, ..., -0.1429, 0.0034, -0.1495], [ 0.0054, -0.1046, 0.0344, ..., 0.0635, -0.1102, -0.0423], [-0.0864, -0.1094, 0.0662, ..., -0.0908, -0.0645, 0.1383]], device='cuda:0'), grad: tensor([[ 1.8626e-08, -1.5944e-06, -5.1968e-07, ..., 2.3749e-08, 0.0000e+00, -2.4624e-06], [-2.0256e-07, 1.0571e-07, 6.3796e-08, ..., -1.2945e-07, -4.6566e-10, 9.8720e-08], [ 5.2620e-08, 8.0094e-08, 7.2177e-08, ..., 4.2841e-08, 0.0000e+00, 1.3504e-07], ..., [ 1.0943e-07, -3.0641e-07, 1.0477e-07, ..., 1.4296e-07, 0.0000e+00, 2.1374e-07], [ 1.7649e-07, 1.2480e-07, 8.8476e-08, ..., 2.4354e-07, 0.0000e+00, 3.3667e-07], [ 2.5611e-08, 6.3563e-07, 1.3225e-07, ..., 6.1747e-07, 0.0000e+00, 6.5565e-07]], device='cuda:0') Epoch 188, bias, value: tensor([-0.0008, 0.0017, 0.0093, 0.0087, 0.0280, 0.0440, -0.0265, 0.0062, -0.0351, -0.0231], device='cuda:0'), grad: tensor([-7.4133e-06, 2.4633e-07, 1.8161e-07, -6.3796e-08, -1.4175e-06, 1.9390e-06, 1.2014e-06, 5.9465e-07, 1.1530e-06, 3.5670e-06], device='cuda:0') 100 0.0001 changing lr epoch 187, time 217.38, cls_loss 0.0026 cls_loss_mapping 0.0046 cls_loss_causal 0.5645 re_mapping 0.0064 re_causal 0.0190 /// teacc 98.98 lr 0.00010000 Epoch 189, weight, value: tensor([[-0.0474, 0.1403, 0.0749, ..., -0.2009, -0.1136, 0.0230], [ 0.1425, -0.0142, -0.0671, ..., 0.0876, 0.1095, -0.1334], [-0.0854, -0.0525, -0.0516, ..., -0.0856, -0.0311, -0.0616], ..., [-0.0700, 0.1499, -0.1712, ..., -0.1431, 0.0033, -0.1499], [ 0.0055, -0.1046, 0.0346, ..., 0.0634, -0.1103, -0.0425], [-0.0871, -0.1103, 0.0669, ..., -0.0911, -0.0646, 0.1387]], device='cuda:0'), grad: tensor([[ 1.5181e-06, 7.9349e-07, 9.3132e-10, ..., 1.1828e-06, 4.6566e-09, -4.0047e-08], [-6.5193e-08, 1.5786e-07, 2.6543e-08, ..., 3.0734e-08, 1.7229e-08, 1.3039e-08], [ 4.6566e-08, 3.6787e-08, 1.1176e-08, ..., 7.0315e-08, 3.2596e-09, 2.3283e-09], ..., [ 1.7695e-08, -7.7672e-07, 3.6787e-08, ..., 7.5437e-08, -2.8033e-07, 4.2841e-08], [-4.3306e-07, 7.8697e-08, 4.5169e-08, ..., -7.7300e-07, 1.6298e-08, 3.5297e-07], [ 3.9581e-08, 2.4168e-07, -3.6880e-07, ..., 1.9046e-07, 1.1455e-07, -5.0105e-07]], device='cuda:0') Epoch 189, bias, value: tensor([-0.0009, 0.0016, 0.0085, 0.0113, 0.0279, 0.0415, -0.0264, 0.0063, -0.0349, -0.0228], device='cuda:0'), grad: tensor([ 9.6783e-06, 3.3006e-06, -1.6004e-05, 3.7905e-06, 8.7079e-08, 1.9837e-06, -8.1882e-06, 3.2280e-06, 1.5013e-06, 5.8161e-07], device='cuda:0') 100 0.0001 changing lr epoch 188, time 216.92, cls_loss 0.0021 cls_loss_mapping 0.0036 cls_loss_causal 0.5321 re_mapping 0.0062 re_causal 0.0186 /// teacc 98.88 lr 0.00010000 Epoch 190, weight, value: tensor([[-0.0473, 0.1409, 0.0747, ..., -0.2012, -0.1136, 0.0225], [ 0.1429, -0.0145, -0.0672, ..., 0.0878, 0.1098, -0.1336], [-0.0861, -0.0526, -0.0519, ..., -0.0863, -0.0314, -0.0622], ..., [-0.0700, 0.1502, -0.1719, ..., -0.1432, 0.0033, -0.1498], [ 0.0052, -0.1048, 0.0344, ..., 0.0629, -0.1104, -0.0432], [-0.0885, -0.1108, 0.0672, ..., -0.0920, -0.0646, 0.1387]], device='cuda:0'), grad: tensor([[ 1.7323e-07, 2.5611e-08, -1.1176e-08, ..., 2.2724e-07, 1.8626e-09, 6.7055e-08], [-2.5379e-07, 5.0757e-08, 3.8184e-08, ..., -3.4086e-07, -1.5367e-08, 8.7079e-08], [ 3.8603e-07, -1.3970e-09, 3.8650e-08, ..., 5.3737e-07, 1.4435e-08, 1.0012e-07], ..., [ 1.6345e-07, -4.8755e-07, 2.3516e-07, ..., 2.2165e-07, 9.7789e-09, 6.3190e-07], [ 1.4529e-06, 3.0734e-08, 1.8533e-07, ..., 1.7788e-06, 2.0023e-08, 8.3493e-07], [ 2.1327e-07, 7.7300e-08, -9.9279e-07, ..., 3.1525e-07, 4.6566e-09, -2.9616e-06]], device='cuda:0') Epoch 190, bias, value: tensor([-0.0010, 0.0015, 0.0082, 0.0110, 0.0287, 0.0419, -0.0259, 0.0066, -0.0353, -0.0235], device='cuda:0'), grad: tensor([ 8.8708e-07, -2.0163e-07, -3.2922e-07, 4.8848e-07, 5.3495e-06, 1.5870e-05, -2.2128e-05, 1.2135e-06, 2.8741e-06, -4.0084e-06], device='cuda:0') 100 0.0001 changing lr epoch 189, time 216.85, cls_loss 0.0019 cls_loss_mapping 0.0034 cls_loss_causal 0.5367 re_mapping 0.0062 re_causal 0.0185 /// teacc 98.95 lr 0.00010000 Epoch 191, weight, value: tensor([[-0.0476, 0.1414, 0.0753, ..., -0.2019, -0.1136, 0.0228], [ 0.1434, -0.0145, -0.0671, ..., 0.0883, 0.1099, -0.1333], [-0.0866, -0.0527, -0.0519, ..., -0.0867, -0.0313, -0.0625], ..., [-0.0701, 0.1503, -0.1729, ..., -0.1433, 0.0033, -0.1502], [ 0.0050, -0.1049, 0.0342, ..., 0.0627, -0.1105, -0.0435], [-0.0891, -0.1109, 0.0677, ..., -0.0921, -0.0647, 0.1389]], device='cuda:0'), grad: tensor([[ 3.2084e-07, -2.5611e-09, 1.9316e-06, ..., 1.3532e-06, 1.4901e-08, 1.7975e-06], [ 4.7311e-07, 2.9244e-07, 1.0310e-06, ..., 1.0701e-06, 5.5879e-09, 6.1374e-07], [ 1.2284e-06, 1.4538e-06, 9.6038e-06, ..., 5.1223e-06, 1.2876e-07, 5.3123e-06], ..., [ 2.2259e-07, -2.8163e-06, 1.7639e-06, ..., 9.4436e-07, 2.5611e-08, 1.1381e-06], [-5.4464e-06, 6.1700e-08, -1.6186e-06, ..., -8.2105e-06, 1.1409e-07, -9.9000e-07], [ 5.4110e-07, 1.0082e-07, 2.6487e-06, ..., 1.9036e-06, 3.4925e-08, 1.1902e-06]], device='cuda:0') Epoch 191, bias, value: tensor([-0.0003, 0.0014, 0.0091, 0.0112, 0.0282, 0.0415, -0.0258, 0.0067, -0.0355, -0.0234], device='cuda:0'), grad: tensor([ 7.0743e-06, 5.4725e-06, 3.1948e-05, 3.6746e-05, 1.2629e-06, -6.0916e-05, 1.8720e-06, 1.5823e-06, -3.4779e-05, 9.6634e-06], device='cuda:0') 100 0.0001 changing lr epoch 190, time 217.16, cls_loss 0.0019 cls_loss_mapping 0.0035 cls_loss_causal 0.5090 re_mapping 0.0060 re_causal 0.0181 /// teacc 98.90 lr 0.00010000 Epoch 192, weight, value: tensor([[-0.0477, 0.1416, 0.0753, ..., -0.2025, -0.1136, 0.0230], [ 0.1431, -0.0147, -0.0679, ..., 0.0878, 0.1101, -0.1341], [-0.0869, -0.0527, -0.0519, ..., -0.0871, -0.0315, -0.0630], ..., [-0.0701, 0.1507, -0.1734, ..., -0.1435, 0.0030, -0.1505], [ 0.0054, -0.1049, 0.0353, ..., 0.0640, -0.1102, -0.0421], [-0.0913, -0.1113, 0.0679, ..., -0.0924, -0.0649, 0.1390]], device='cuda:0'), grad: tensor([[ 1.3388e-07, 2.4750e-07, -1.4901e-08, ..., 2.1816e-07, 4.6566e-10, -2.0256e-08], [-1.3458e-07, 2.6776e-08, 3.3993e-08, ..., 1.0524e-06, -2.9569e-08, 5.8906e-08], [ 5.3085e-08, 6.0303e-08, 1.0012e-08, ..., 1.8300e-07, 3.0268e-09, 1.8161e-08], ..., [ 2.9197e-07, 4.5169e-08, 2.3516e-07, ..., 1.0636e-06, 8.3819e-09, 1.2037e-07], [-4.6031e-07, 1.9860e-07, -1.3551e-07, ..., -1.4175e-06, 1.2340e-08, -6.7241e-07], [ 4.9639e-07, 8.1956e-08, -3.6485e-07, ..., 3.9749e-06, 1.3970e-09, -1.2782e-07]], device='cuda:0') Epoch 192, bias, value: tensor([-0.0001, 0.0008, 0.0092, 0.0111, 0.0279, 0.0415, -0.0255, 0.0069, -0.0351, -0.0234], device='cuda:0'), grad: tensor([ 5.2806e-07, 2.5257e-06, 2.7660e-07, -9.0990e-07, -1.1027e-05, 4.5002e-06, -4.6864e-06, 3.4459e-06, -1.8831e-06, 7.2271e-06], device='cuda:0') 100 0.0001 changing lr epoch 191, time 217.21, cls_loss 0.0022 cls_loss_mapping 0.0030 cls_loss_causal 0.5340 re_mapping 0.0060 re_causal 0.0182 /// teacc 98.91 lr 0.00010000 Epoch 193, weight, value: tensor([[-0.0484, 0.1415, 0.0751, ..., -0.2035, -0.1138, 0.0228], [ 0.1439, -0.0145, -0.0691, ..., 0.0880, 0.1107, -0.1355], [-0.0881, -0.0531, -0.0523, ..., -0.0887, -0.0322, -0.0641], ..., [-0.0703, 0.1510, -0.1742, ..., -0.1438, 0.0026, -0.1515], [ 0.0054, -0.1050, 0.0355, ..., 0.0637, -0.1094, -0.0422], [-0.0921, -0.1118, 0.0685, ..., -0.0930, -0.0651, 0.1393]], device='cuda:0'), grad: tensor([[ 1.0943e-08, 6.8024e-06, 1.6764e-08, ..., 7.9498e-06, 0.0000e+00, 1.5469e-06], [-6.1747e-07, 1.0729e-06, 1.4901e-08, ..., -3.4925e-08, -3.4925e-09, 9.0571e-08], [ 2.4214e-08, 1.1278e-06, 3.9581e-08, ..., 2.0908e-07, -2.0955e-09, 3.9348e-08], ..., [ 2.9732e-07, -3.6880e-06, 3.4692e-08, ..., 3.8208e-07, 2.3283e-09, 4.0745e-08], [-1.1688e-07, 9.7789e-07, 3.7509e-07, ..., 8.4750e-07, 2.3283e-10, 8.7032e-07], [ 4.1234e-07, 1.3579e-06, -7.5903e-08, ..., 1.6186e-06, 6.9849e-10, 1.5297e-07]], device='cuda:0') Epoch 193, bias, value: tensor([-0.0005, 0.0008, 0.0085, 0.0112, 0.0285, 0.0414, -0.0250, 0.0071, -0.0352, -0.0237], device='cuda:0'), grad: tensor([ 3.1888e-05, 1.6261e-06, 1.6196e-06, 1.4864e-06, 2.4602e-05, 6.2399e-07, -6.8486e-05, -3.6880e-06, 4.5076e-06, 5.8189e-06], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 192---------------------------------------------------- epoch 192, time 217.92, cls_loss 0.0020 cls_loss_mapping 0.0034 cls_loss_causal 0.5147 re_mapping 0.0058 re_causal 0.0172 /// teacc 99.05 lr 0.00010000 Epoch 194, weight, value: tensor([[-0.0493, 0.1398, 0.0756, ..., -0.2064, -0.1150, 0.0232], [ 0.1442, -0.0160, -0.0693, ..., 0.0876, 0.1107, -0.1360], [-0.0884, -0.0534, -0.0525, ..., -0.0890, -0.0323, -0.0652], ..., [-0.0703, 0.1515, -0.1744, ..., -0.1432, 0.0027, -0.1517], [ 0.0053, -0.1048, 0.0353, ..., 0.0625, -0.1095, -0.0425], [-0.0926, -0.1121, 0.0686, ..., -0.0933, -0.0651, 0.1395]], device='cuda:0'), grad: tensor([[ 7.9162e-09, 2.2352e-08, 2.7940e-09, ..., 1.5367e-08, 0.0000e+00, -3.2596e-09], [ 2.6403e-07, 5.5740e-07, 1.9139e-07, ..., 6.7754e-07, -1.3970e-09, 3.3202e-07], [ 1.8626e-08, 9.8161e-07, 2.7008e-08, ..., 3.2131e-08, 0.0000e+00, 6.5193e-09], ..., [ 6.8452e-08, -4.9509e-06, 2.3982e-07, ..., 2.0256e-07, 9.3132e-10, 4.2981e-07], [-4.2468e-07, 3.4878e-07, 1.1129e-07, ..., -8.4797e-07, 0.0000e+00, 1.7555e-07], [ 4.8429e-08, 9.5135e-07, -1.5143e-06, ..., -3.1525e-07, 0.0000e+00, -2.7865e-06]], device='cuda:0') Epoch 194, bias, value: tensor([-0.0016, -0.0007, 0.0085, 0.0111, 0.0286, 0.0416, -0.0238, 0.0076, -0.0353, -0.0239], device='cuda:0'), grad: tensor([ 1.1921e-07, 2.7232e-06, 1.3541e-06, 2.4270e-06, 3.9786e-06, -3.8967e-06, 4.3325e-06, -7.0743e-06, -3.8417e-07, -3.6322e-06], device='cuda:0') 100 0.0001 changing lr epoch 193, time 217.25, cls_loss 0.0022 cls_loss_mapping 0.0040 cls_loss_causal 0.4958 re_mapping 0.0064 re_causal 0.0177 /// teacc 99.01 lr 0.00010000 Epoch 195, weight, value: tensor([[-0.0494, 0.1409, 0.0749, ..., -0.2068, -0.1150, 0.0224], [ 0.1445, -0.0165, -0.0710, ..., 0.0866, 0.1123, -0.1377], [-0.0893, -0.0537, -0.0512, ..., -0.0890, -0.0339, -0.0632], ..., [-0.0703, 0.1521, -0.1756, ..., -0.1432, 0.0025, -0.1526], [ 0.0053, -0.1050, 0.0352, ..., 0.0621, -0.1095, -0.0426], [-0.0929, -0.1125, 0.0702, ..., -0.0925, -0.0653, 0.1409]], device='cuda:0'), grad: tensor([[ 4.1910e-09, -1.8254e-07, -1.3039e-08, ..., 2.5146e-08, 9.3132e-10, -3.1199e-08], [-1.6112e-07, 3.9581e-08, 3.5856e-08, ..., -2.5006e-07, 4.6566e-10, 4.0978e-08], [ 1.4435e-08, 1.0850e-07, 1.2573e-08, ..., 2.9802e-08, -1.2107e-08, 1.2107e-08], ..., [ 2.9802e-08, -1.5274e-07, 1.3830e-07, ..., 9.0338e-08, 2.3283e-09, 1.9558e-07], [ 3.1199e-08, 5.2154e-08, 2.2072e-07, ..., 6.5658e-08, 3.7253e-09, 3.2689e-07], [ 6.1002e-08, 6.5658e-08, -4.8056e-07, ..., 1.2014e-07, 4.6566e-10, -6.9849e-07]], device='cuda:0') Epoch 195, bias, value: tensor([-0.0023, -0.0022, 0.0093, 0.0110, 0.0270, 0.0415, -0.0232, 0.0078, -0.0357, -0.0220], device='cuda:0'), grad: tensor([-1.7788e-07, -2.8033e-07, 3.0268e-08, -6.2305e-07, 8.8010e-08, 7.3574e-07, -6.9849e-08, 3.4878e-07, 8.6939e-07, -9.3831e-07], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 194---------------------------------------------------- epoch 194, time 218.21, cls_loss 0.0019 cls_loss_mapping 0.0037 cls_loss_causal 0.5179 re_mapping 0.0063 re_causal 0.0183 /// teacc 99.09 lr 0.00010000 Epoch 196, weight, value: tensor([[-0.0497, 0.1411, 0.0745, ..., -0.2076, -0.1150, 0.0219], [ 0.1457, -0.0165, -0.0708, ..., 0.0874, 0.1126, -0.1379], [-0.0897, -0.0540, -0.0512, ..., -0.0891, -0.0342, -0.0629], ..., [-0.0705, 0.1523, -0.1762, ..., -0.1436, 0.0025, -0.1536], [ 0.0053, -0.1051, 0.0352, ..., 0.0622, -0.1096, -0.0422], [-0.0940, -0.1126, 0.0705, ..., -0.0929, -0.0654, 0.1411]], device='cuda:0'), grad: tensor([[ 4.4238e-07, 8.7544e-08, -6.5193e-09, ..., 5.6485e-07, 9.3132e-10, 1.2247e-07], [-4.5076e-06, -1.1139e-06, -2.1374e-07, ..., -7.1004e-06, 5.1223e-09, 2.2817e-08], [ 5.0291e-07, 3.2736e-07, 1.3039e-08, ..., 6.5891e-07, -1.3970e-08, 1.6764e-08], ..., [ 1.3821e-06, -3.7719e-07, 1.4668e-07, ..., 2.7362e-06, 1.8626e-09, 1.6764e-08], [ 5.2340e-06, 2.1420e-06, 3.2131e-08, ..., 6.5118e-06, 2.3283e-09, 2.4885e-06], [ 4.8196e-07, 5.1409e-07, 1.1642e-08, ..., 9.7509e-07, 4.6566e-10, -1.8161e-08]], device='cuda:0') Epoch 196, bias, value: tensor([-0.0028, -0.0022, 0.0094, 0.0109, 0.0270, 0.0413, -0.0229, 0.0078, -0.0356, -0.0218], device='cuda:0'), grad: tensor([ 9.1922e-07, -1.4685e-05, 7.7439e-07, 3.7113e-07, 9.2061e-07, 6.7830e-05, -7.5161e-05, 5.1744e-06, 1.1533e-05, 2.5332e-06], device='cuda:0') 100 0.0001 changing lr epoch 195, time 217.14, cls_loss 0.0025 cls_loss_mapping 0.0050 cls_loss_causal 0.5187 re_mapping 0.0066 re_causal 0.0175 /// teacc 98.95 lr 0.00010000 Epoch 197, weight, value: tensor([[-0.0503, 0.1413, 0.0747, ..., -0.2087, -0.1154, 0.0221], [ 0.1467, -0.0167, -0.0708, ..., 0.0881, 0.1134, -0.1381], [-0.0906, -0.0571, -0.0514, ..., -0.0897, -0.0346, -0.0631], ..., [-0.0706, 0.1546, -0.1766, ..., -0.1439, 0.0021, -0.1546], [ 0.0054, -0.1052, 0.0349, ..., 0.0626, -0.1096, -0.0431], [-0.0947, -0.1131, 0.0712, ..., -0.0925, -0.0658, 0.1418]], device='cuda:0'), grad: tensor([[-2.0489e-08, -4.8196e-07, 3.3155e-07, ..., 9.4995e-08, 4.6566e-10, 8.4750e-08], [-1.8850e-06, 1.2852e-07, 9.9186e-08, ..., -2.5053e-06, -5.1223e-09, 3.4366e-07], [ 3.5670e-07, 1.3504e-08, 2.5379e-07, ..., 4.8708e-07, 4.6566e-10, 1.0198e-07], ..., [ 3.5297e-07, -5.9605e-08, 7.0734e-07, ..., 6.0070e-07, 1.8626e-09, 8.6613e-07], [-3.8669e-06, -1.3970e-08, 3.3043e-06, ..., -6.9216e-06, 4.6566e-10, -3.7607e-06], [ 2.9188e-06, 8.7544e-08, 6.5304e-06, ..., 6.6757e-06, 9.3132e-10, 6.2287e-06]], device='cuda:0') Epoch 197, bias, value: tensor([-0.0030, -0.0019, 0.0085, 0.0107, 0.0266, 0.0415, -0.0238, 0.0091, -0.0361, -0.0213], device='cuda:0'), grad: tensor([ 4.3772e-07, -3.9712e-06, 8.0932e-07, -3.4004e-05, -1.6876e-06, 4.3511e-06, 4.2319e-06, 3.2298e-06, -2.6450e-06, 2.9191e-05], device='cuda:0') 100 0.0001 changing lr epoch 196, time 216.86, cls_loss 0.0019 cls_loss_mapping 0.0034 cls_loss_causal 0.5055 re_mapping 0.0065 re_causal 0.0190 /// teacc 99.01 lr 0.00010000 Epoch 198, weight, value: tensor([[-0.0504, 0.1418, 0.0746, ..., -0.2091, -0.1155, 0.0222], [ 0.1474, -0.0169, -0.0709, ..., 0.0883, 0.1140, -0.1383], [-0.0911, -0.0571, -0.0511, ..., -0.0900, -0.0350, -0.0631], ..., [-0.0708, 0.1547, -0.1770, ..., -0.1443, 0.0018, -0.1548], [ 0.0056, -0.1054, 0.0348, ..., 0.0631, -0.1087, -0.0434], [-0.0952, -0.1131, 0.0714, ..., -0.0924, -0.0662, 0.1421]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 8.3353e-07, 1.8869e-06, ..., 1.1232e-06, 3.7253e-08, 2.3544e-06], [-1.6764e-08, 2.6748e-05, 1.3836e-05, ..., 7.9023e-07, 2.7940e-09, 1.7444e-06], [ 1.8626e-09, 1.1735e-05, 9.9167e-06, ..., 4.0904e-06, 1.3039e-08, 7.1302e-06], ..., [ 1.4901e-08, 7.1907e-04, 3.4547e-04, ..., 4.7311e-07, 1.3504e-08, 1.5944e-06], [-9.3132e-09, 9.5218e-06, 5.8152e-06, ..., 7.7672e-07, 2.9337e-08, 2.0936e-06], [ 4.1910e-09, -7.7438e-04, -3.3593e-04, ..., 2.7388e-05, 8.3819e-09, 5.9813e-05]], device='cuda:0') Epoch 198, bias, value: tensor([-0.0029, -0.0018, 0.0086, 0.0107, 0.0262, 0.0413, -0.0241, 0.0089, -0.0358, -0.0209], device='cuda:0'), grad: tensor([ 7.8902e-06, 1.2732e-04, 6.6400e-05, 2.2733e-04, 2.2218e-05, -3.2783e-04, 3.3304e-06, 3.3607e-03, 4.7296e-05, -3.5362e-03], device='cuda:0') 100 0.0001 changing lr epoch 197, time 216.86, cls_loss 0.0019 cls_loss_mapping 0.0040 cls_loss_causal 0.5165 re_mapping 0.0064 re_causal 0.0180 /// teacc 99.02 lr 0.00010000 Epoch 199, weight, value: tensor([[-0.0504, 0.1427, 0.0751, ..., -0.2094, -0.1155, 0.0225], [ 0.1475, -0.0178, -0.0712, ..., 0.0884, 0.1142, -0.1387], [-0.0914, -0.0569, -0.0512, ..., -0.0903, -0.0356, -0.0633], ..., [-0.0708, 0.1545, -0.1789, ..., -0.1444, 0.0026, -0.1567], [ 0.0057, -0.1055, 0.0348, ..., 0.0634, -0.1089, -0.0433], [-0.0957, -0.1113, 0.0709, ..., -0.0933, -0.0664, 0.1416]], device='cuda:0'), grad: tensor([[ 4.9360e-08, -3.1199e-08, -1.3039e-08, ..., 1.5134e-07, 1.8626e-09, -1.9558e-08], [ 1.4435e-08, 1.1036e-07, 8.3819e-09, ..., 4.3772e-08, 4.6566e-10, 2.1420e-08], [ 1.3504e-08, 1.3039e-08, 6.0536e-09, ..., -2.9523e-07, -3.8184e-08, 8.8476e-09], ..., [-1.8626e-09, -2.4401e-07, 1.3970e-08, ..., 5.1688e-08, 3.2596e-09, 3.0268e-08], [ 2.5658e-07, 5.7742e-08, 1.1688e-07, ..., 3.1432e-07, 2.5611e-08, 3.4133e-07], [ 4.4238e-08, 1.2806e-07, -1.1409e-07, ..., 1.3923e-07, 9.3132e-10, -1.7788e-07]], device='cuda:0') Epoch 199, bias, value: tensor([-0.0027, -0.0023, 0.0088, 0.0107, 0.0265, 0.0413, -0.0235, 0.0081, -0.0358, -0.0204], device='cuda:0'), grad: tensor([ 3.7625e-07, 4.3772e-07, -2.8051e-06, 1.0692e-06, 1.6108e-05, -1.7546e-06, -1.5661e-05, 8.2748e-07, 1.3653e-06, 2.3283e-09], device='cuda:0') 100 0.0001 changing lr epoch 198, time 216.85, cls_loss 0.0015 cls_loss_mapping 0.0026 cls_loss_causal 0.4966 re_mapping 0.0064 re_causal 0.0181 /// teacc 98.99 lr 0.00010000 Epoch 200, weight, value: tensor([[-0.0507, 0.1431, 0.0754, ..., -0.2101, -0.1155, 0.0227], [ 0.1477, -0.0180, -0.0714, ..., 0.0886, 0.1144, -0.1389], [-0.0920, -0.0569, -0.0513, ..., -0.0905, -0.0358, -0.0635], ..., [-0.0708, 0.1547, -0.1790, ..., -0.1447, 0.0025, -0.1569], [ 0.0059, -0.1055, 0.0348, ..., 0.0638, -0.1090, -0.0435], [-0.0961, -0.1116, 0.0712, ..., -0.0939, -0.0664, 0.1419]], device='cuda:0'), grad: tensor([[ 2.8405e-08, -1.3039e-08, 1.2107e-07, ..., 1.1502e-07, 0.0000e+00, 2.0862e-07], [-2.1253e-06, 1.1222e-07, 4.4238e-08, ..., -2.5351e-06, 0.0000e+00, 7.6368e-08], [ 1.8086e-06, 6.2166e-07, 2.3376e-07, ..., 2.5742e-06, 0.0000e+00, 3.2736e-07], ..., [ 1.8114e-07, -8.8243e-07, 7.2177e-08, ..., 3.8277e-07, 0.0000e+00, 1.2480e-07], [ 6.6590e-08, 2.0489e-08, 8.7824e-07, ..., 2.0489e-07, 0.0000e+00, 1.6317e-06], [ 4.2841e-08, 1.0058e-07, 3.3900e-06, ..., 2.0750e-06, 0.0000e+00, 6.1169e-06]], device='cuda:0') Epoch 200, bias, value: tensor([-0.0026, -0.0023, 0.0088, 0.0107, 0.0268, 0.0416, -0.0240, 0.0081, -0.0358, -0.0206], device='cuda:0'), grad: tensor([ 5.3924e-07, -4.6045e-06, 4.9360e-06, 8.6203e-06, -4.7497e-08, -2.2501e-05, 6.7009e-07, -1.6531e-07, 2.4941e-06, 1.0096e-05], device='cuda:0') 100 0.0001 changing lr epoch 199, time 216.92, cls_loss 0.0014 cls_loss_mapping 0.0029 cls_loss_causal 0.5240 re_mapping 0.0060 re_causal 0.0182 /// teacc 99.08 lr 0.00010000 Epoch 201, weight, value: tensor([[-0.0514, 0.1431, 0.0759, ..., -0.2107, -0.1163, 0.0230], [ 0.1490, -0.0178, -0.0715, ..., 0.0892, 0.1154, -0.1391], [-0.0925, -0.0568, -0.0514, ..., -0.0908, -0.0360, -0.0637], ..., [-0.0713, 0.1547, -0.1791, ..., -0.1456, 0.0012, -0.1571], [ 0.0058, -0.1056, 0.0346, ..., 0.0637, -0.1092, -0.0439], [-0.0964, -0.1120, 0.0711, ..., -0.0942, -0.0668, 0.1419]], device='cuda:0'), grad: tensor([[ 2.3283e-08, -6.9849e-08, 1.1176e-07, ..., 2.0824e-06, 1.3970e-09, -1.9092e-08], [ 3.8370e-07, 1.6065e-07, 2.2259e-07, ..., 1.0636e-06, 1.2573e-08, 1.7928e-07], [ 4.6566e-08, 7.9628e-08, 5.1036e-07, ..., 3.6508e-07, 5.0757e-08, 1.4110e-07], ..., [ 6.2864e-08, -2.1886e-07, 1.6764e-07, ..., 2.0349e-07, -9.7789e-09, 8.1025e-08], [-1.3284e-05, -4.5635e-07, -7.5214e-06, ..., -3.0175e-05, 3.7253e-09, -6.0573e-06], [ 1.7043e-07, 9.4064e-08, -1.7881e-07, ..., 4.5858e-06, 5.5879e-09, -2.7753e-07]], device='cuda:0') Epoch 201, bias, value: tensor([-0.0025, -0.0017, 0.0088, 0.0109, 0.0269, 0.0416, -0.0242, 0.0077, -0.0359, -0.0209], device='cuda:0'), grad: tensor([ 9.0674e-06, 1.9334e-06, 2.7902e-06, 8.7684e-07, -3.6150e-05, 2.2322e-05, 2.4587e-05, 6.2445e-07, -4.4048e-05, 1.8016e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 200---------------------------------------------------- epoch 200, time 217.82, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.4658 re_mapping 0.0062 re_causal 0.0179 /// teacc 99.11 lr 0.00010000 Epoch 202, weight, value: tensor([[-0.0547, 0.1418, 0.0762, ..., -0.2116, -0.1191, 0.0232], [ 0.1490, -0.0179, -0.0716, ..., 0.0892, 0.1155, -0.1393], [-0.0926, -0.0569, -0.0515, ..., -0.0908, -0.0362, -0.0639], ..., [-0.0713, 0.1549, -0.1794, ..., -0.1456, 0.0012, -0.1576], [ 0.0062, -0.1057, 0.0345, ..., 0.0645, -0.1090, -0.0438], [-0.0970, -0.1122, 0.0713, ..., -0.0945, -0.0670, 0.1421]], device='cuda:0'), grad: tensor([[-6.2818e-07, -1.3560e-06, 8.8476e-09, ..., 2.2911e-07, 9.5926e-08, -4.3120e-07], [ 1.8254e-06, 1.1921e-07, 1.1642e-08, ..., 2.1197e-06, 9.8627e-07, 2.8592e-07], [ 5.3458e-06, 6.3796e-08, 4.0559e-07, ..., 6.3293e-06, 3.3788e-06, 6.5006e-07], ..., [ 1.0924e-06, -1.4435e-08, 4.0978e-08, ..., 1.3830e-06, 6.0257e-07, 1.4156e-07], [-9.9361e-05, 2.1514e-07, -9.9372e-07, ..., -1.1927e-04, -5.4091e-05, -1.0602e-05], [ 2.0629e-07, 3.6787e-08, -8.5682e-08, ..., 1.0431e-06, 9.4529e-08, -1.0245e-07]], device='cuda:0') Epoch 202, bias, value: tensor([-0.0048, -0.0019, 0.0089, 0.0107, 0.0270, 0.0417, -0.0228, 0.0077, -0.0357, -0.0209], device='cuda:0'), grad: tensor([-2.7120e-06, 5.7705e-06, 1.7390e-05, 1.1604e-06, -1.1222e-06, 2.6488e-04, 9.7677e-06, 3.4925e-06, -3.0184e-04, 2.6524e-06], device='cuda:0') 100 0.0001 changing lr epoch 201, time 216.93, cls_loss 0.0016 cls_loss_mapping 0.0042 cls_loss_causal 0.5273 re_mapping 0.0060 re_causal 0.0176 /// teacc 99.01 lr 0.00010000 Epoch 203, weight, value: tensor([[-0.0548, 0.1423, 0.0768, ..., -0.2120, -0.1192, 0.0237], [ 0.1497, -0.0179, -0.0720, ..., 0.0895, 0.1164, -0.1399], [-0.0931, -0.0569, -0.0517, ..., -0.0909, -0.0367, -0.0643], ..., [-0.0716, 0.1550, -0.1802, ..., -0.1462, 0.0004, -0.1580], [ 0.0063, -0.1058, 0.0341, ..., 0.0648, -0.1081, -0.0441], [-0.0974, -0.1123, 0.0719, ..., -0.0943, -0.0685, 0.1425]], device='cuda:0'), grad: tensor([[ 2.0210e-07, 4.3306e-08, 1.3504e-07, ..., 8.5589e-07, 3.7253e-09, -3.1898e-07], [ 3.5763e-07, 2.2519e-06, 5.7276e-08, ..., 1.3178e-07, 1.8915e-06, 2.9337e-08], [-3.2736e-07, 3.4459e-08, -1.2554e-06, ..., 8.4145e-07, -2.0117e-06, -1.8813e-06], ..., [ 2.5611e-08, -9.8050e-06, 3.2783e-07, ..., 2.2352e-08, 1.8347e-07, 6.2399e-08], [ 3.0510e-06, 6.8452e-08, 1.4883e-06, ..., 7.2233e-06, 6.9384e-08, 1.9651e-06], [ 3.2596e-09, 7.5027e-06, 4.9407e-07, ..., 2.2305e-07, 2.1886e-08, 4.8010e-07]], device='cuda:0') Epoch 203, bias, value: tensor([-0.0046, -0.0015, 0.0088, 0.0107, 0.0266, 0.0417, -0.0228, 0.0074, -0.0359, -0.0205], device='cuda:0'), grad: tensor([ 4.2133e-06, 1.5229e-05, -1.3381e-05, -8.2701e-06, 3.0175e-06, 1.0960e-05, -3.9428e-05, -1.3448e-05, 2.8595e-05, 1.2577e-05], device='cuda:0') 100 0.0001 changing lr epoch 202, time 217.26, cls_loss 0.0016 cls_loss_mapping 0.0035 cls_loss_causal 0.4971 re_mapping 0.0063 re_causal 0.0176 /// teacc 98.99 lr 0.00010000 Epoch 204, weight, value: tensor([[-0.0547, 0.1441, 0.0764, ..., -0.2124, -0.1193, 0.0235], [ 0.1505, -0.0176, -0.0725, ..., 0.0902, 0.1165, -0.1407], [-0.0935, -0.0573, -0.0518, ..., -0.0911, -0.0368, -0.0645], ..., [-0.0721, 0.1548, -0.1805, ..., -0.1473, 0.0007, -0.1591], [ 0.0065, -0.1059, 0.0339, ..., 0.0649, -0.1082, -0.0445], [-0.0967, -0.1118, 0.0730, ..., -0.0939, -0.0677, 0.1436]], device='cuda:0'), grad: tensor([[ 1.2442e-06, 1.4482e-07, 1.4473e-06, ..., 1.0580e-06, 4.6566e-10, 1.5730e-06], [-5.9139e-08, 4.3306e-08, 3.0082e-07, ..., 5.4156e-07, -5.5879e-09, 5.5460e-07], [ 2.5146e-08, 2.0489e-08, 3.9116e-08, ..., 3.9581e-08, 9.3132e-10, 7.7300e-08], ..., [ 4.5169e-08, -5.6345e-08, 5.0198e-07, ..., 1.7649e-07, 9.3132e-10, 1.4184e-06], [-4.1723e-06, -8.9360e-07, -3.7923e-06, ..., -2.4792e-06, 1.8626e-09, -3.7216e-06], [ 6.3004e-07, 1.6345e-07, -3.2373e-06, ..., 3.5521e-06, 9.3132e-10, -8.3447e-06]], device='cuda:0') Epoch 204, bias, value: tensor([-0.0036, -0.0017, 0.0089, 0.0107, 0.0261, 0.0415, -0.0228, 0.0058, -0.0361, -0.0182], device='cuda:0'), grad: tensor([ 3.6098e-06, 1.8440e-06, 1.6578e-07, 2.4214e-06, 3.3043e-06, 4.6603e-06, 6.8499e-07, 3.1497e-06, -7.2867e-06, -1.2532e-05], device='cuda:0') 100 0.0001 changing lr epoch 203, time 216.92, cls_loss 0.0012 cls_loss_mapping 0.0029 cls_loss_causal 0.4860 re_mapping 0.0062 re_causal 0.0180 /// teacc 98.89 lr 0.00010000 Epoch 205, weight, value: tensor([[-0.0549, 0.1442, 0.0765, ..., -0.2132, -0.1193, 0.0236], [ 0.1508, -0.0186, -0.0725, ..., 0.0904, 0.1170, -0.1409], [-0.0939, -0.0573, -0.0520, ..., -0.0913, -0.0370, -0.0647], ..., [-0.0721, 0.1556, -0.1807, ..., -0.1474, 0.0014, -0.1597], [ 0.0063, -0.1059, 0.0336, ..., 0.0648, -0.1083, -0.0451], [-0.0963, -0.1121, 0.0742, ..., -0.0940, -0.0681, 0.1448]], device='cuda:0'), grad: tensor([[ 3.2596e-09, -1.0850e-07, -3.8929e-07, ..., 1.7695e-08, 0.0000e+00, -5.2107e-07], [-1.2619e-07, 2.2817e-08, 4.8429e-08, ..., -1.8626e-07, -4.6566e-10, 8.9873e-08], [ 6.5193e-09, 1.8207e-07, 9.5461e-08, ..., -6.0536e-09, 0.0000e+00, 1.0896e-07], ..., [ 7.5437e-08, -9.2993e-07, 1.7369e-07, ..., 2.6729e-07, -4.6566e-10, 2.8918e-07], [ 2.9197e-07, 2.9337e-08, 2.2398e-07, ..., 6.4075e-07, 4.6566e-10, 1.0002e-06], [ 4.5635e-08, 6.7707e-07, -1.5348e-05, ..., -1.2174e-05, 1.3970e-09, -2.9042e-05]], device='cuda:0') Epoch 205, bias, value: tensor([-0.0037, -0.0020, 0.0088, 0.0106, 0.0262, 0.0411, -0.0226, 0.0062, -0.0364, -0.0179], device='cuda:0'), grad: tensor([-1.1437e-06, -1.5600e-07, 3.9069e-07, 5.2992e-07, 5.0455e-05, -5.9232e-07, 6.4820e-07, -5.0850e-07, 9.2434e-07, -5.0515e-05], device='cuda:0') 100 0.0001 changing lr epoch 204, time 216.71, cls_loss 0.0022 cls_loss_mapping 0.0041 cls_loss_causal 0.5371 re_mapping 0.0061 re_causal 0.0173 /// teacc 99.03 lr 0.00010000 Epoch 206, weight, value: tensor([[-0.0568, 0.1415, 0.0751, ..., -0.2169, -0.1197, 0.0221], [ 0.1531, -0.0175, -0.0734, ..., 0.0917, 0.1176, -0.1430], [-0.0943, -0.0574, -0.0521, ..., -0.0915, -0.0372, -0.0649], ..., [-0.0732, 0.1553, -0.1822, ..., -0.1497, 0.0008, -0.1603], [ 0.0062, -0.1060, 0.0335, ..., 0.0647, -0.1085, -0.0454], [-0.0966, -0.1122, 0.0757, ..., -0.0939, -0.0663, 0.1457]], device='cuda:0'), grad: tensor([[ 4.6566e-09, 1.7121e-05, -2.9244e-07, ..., 1.1176e-08, 0.0000e+00, -5.6252e-07], [ 1.9930e-07, 6.8871e-07, 1.0664e-07, ..., 3.4412e-07, 0.0000e+00, 1.3504e-07], [ 6.6590e-08, -4.6611e-05, 4.3772e-08, ..., 4.4703e-08, 0.0000e+00, 6.7055e-08], ..., [ 1.6298e-08, 1.6853e-05, 1.3504e-08, ..., 7.1246e-08, 0.0000e+00, 1.3970e-08], [-7.3249e-07, 7.8678e-06, -2.2724e-07, ..., -1.0170e-06, 0.0000e+00, -2.3609e-07], [ 1.1548e-07, 3.2932e-06, 1.8766e-07, ..., 2.8824e-07, 0.0000e+00, 3.2876e-07]], device='cuda:0') Epoch 206, bias, value: tensor([-0.0059, -0.0015, 0.0087, 0.0109, 0.0261, 0.0411, -0.0210, 0.0055, -0.0365, -0.0175], device='cuda:0'), grad: tensor([ 4.0978e-05, 2.0079e-06, -1.1355e-04, 6.0489e-07, -2.1840e-07, 9.4017e-07, 1.0943e-06, 4.4584e-05, 1.7777e-05, 5.8338e-06], device='cuda:0') 100 0.0001 changing lr epoch 205, time 217.40, cls_loss 0.0020 cls_loss_mapping 0.0037 cls_loss_causal 0.5254 re_mapping 0.0057 re_causal 0.0167 /// teacc 98.96 lr 0.00010000 Epoch 207, weight, value: tensor([[-0.0569, 0.1424, 0.0760, ..., -0.2173, -0.1197, 0.0234], [ 0.1535, -0.0184, -0.0738, ..., 0.0915, 0.1183, -0.1435], [-0.0944, -0.0572, -0.0523, ..., -0.0915, -0.0376, -0.0651], ..., [-0.0734, 0.1566, -0.1825, ..., -0.1499, 0.0003, -0.1614], [ 0.0063, -0.1063, 0.0334, ..., 0.0647, -0.1088, -0.0453], [-0.0976, -0.1146, 0.0757, ..., -0.0946, -0.0668, 0.1456]], device='cuda:0'), grad: tensor([[ 1.0878e-06, -1.2685e-06, -1.6708e-06, ..., 6.6124e-08, -6.0536e-09, -2.8964e-07], [-1.8626e-08, 4.7684e-07, 1.4855e-07, ..., -1.4901e-08, -1.8161e-08, 2.2398e-07], [ 3.2689e-07, -2.6226e-06, 1.6764e-08, ..., 3.0175e-07, -3.1199e-08, 8.8802e-07], ..., [ 2.9104e-07, 2.1756e-06, 4.1118e-07, ..., 3.9767e-07, 1.5367e-08, 4.9500e-07], [-7.4646e-07, -1.1455e-07, -2.4866e-07, ..., -1.2303e-06, 4.6566e-09, 5.9651e-07], [ 1.3551e-07, 1.7649e-06, 3.8138e-07, ..., 5.3644e-07, 1.4435e-08, 1.3188e-06]], device='cuda:0') Epoch 207, bias, value: tensor([-0.0049, -0.0021, 0.0088, 0.0109, 0.0265, 0.0412, -0.0214, 0.0062, -0.0367, -0.0184], device='cuda:0'), grad: tensor([ 2.6934e-06, 1.7062e-06, -9.0078e-06, 8.5915e-07, 3.6741e-07, 4.8690e-06, -1.7226e-05, 9.9242e-06, -1.0990e-06, 6.9141e-06], device='cuda:0') 100 0.0001 changing lr epoch 206, time 217.29, cls_loss 0.0020 cls_loss_mapping 0.0030 cls_loss_causal 0.5063 re_mapping 0.0061 re_causal 0.0171 /// teacc 99.01 lr 0.00010000 Epoch 208, weight, value: tensor([[-0.0574, 0.1429, 0.0760, ..., -0.2178, -0.1200, 0.0234], [ 0.1549, -0.0186, -0.0743, ..., 0.0921, 0.1182, -0.1444], [-0.0947, -0.0574, -0.0521, ..., -0.0918, -0.0378, -0.0653], ..., [-0.0738, 0.1573, -0.1834, ..., -0.1503, 0.0006, -0.1622], [ 0.0063, -0.1065, 0.0333, ..., 0.0647, -0.1092, -0.0456], [-0.1001, -0.1156, 0.0756, ..., -0.0958, -0.0659, 0.1455]], device='cuda:0'), grad: tensor([[ 3.8650e-08, 1.2992e-07, 1.3327e-06, ..., 3.2131e-07, 3.7253e-09, 3.9907e-07], [-6.1747e-07, 2.4475e-06, 2.5285e-07, ..., -8.7498e-07, -3.2596e-09, 1.2899e-07], [ 1.1642e-07, 3.1330e-06, 2.4527e-05, ..., 8.9630e-06, -4.1910e-08, 1.5095e-05], ..., [ 1.9791e-07, -1.9163e-05, 2.1793e-07, ..., 3.5483e-07, -1.1362e-07, 1.2759e-07], [-2.7940e-06, -3.0771e-06, -3.4332e-05, ..., -1.3240e-05, 2.9337e-08, -2.0653e-05], [ 1.0151e-07, 1.3970e-05, 3.8184e-06, ..., 2.3637e-06, 1.2573e-08, 2.4438e-06]], device='cuda:0') Epoch 208, bias, value: tensor([-0.0048, -0.0020, 0.0087, 0.0110, 0.0273, 0.0416, -0.0218, 0.0065, -0.0370, -0.0193], device='cuda:0'), grad: tensor([ 3.3602e-06, 3.0454e-06, 5.1290e-05, 6.6459e-06, -1.4361e-06, 6.8620e-06, 8.1258e-07, -3.0518e-05, -7.2420e-05, 3.2336e-05], device='cuda:0') 100 0.0001 changing lr epoch 207, time 217.06, cls_loss 0.0020 cls_loss_mapping 0.0030 cls_loss_causal 0.5266 re_mapping 0.0059 re_causal 0.0169 /// teacc 98.93 lr 0.00010000 Epoch 209, weight, value: tensor([[-0.0576, 0.1434, 0.0756, ..., -0.2181, -0.1199, 0.0232], [ 0.1549, -0.0185, -0.0760, ..., 0.0916, 0.1186, -0.1468], [-0.0953, -0.0574, -0.0535, ..., -0.0921, -0.0380, -0.0666], ..., [-0.0743, 0.1577, -0.1857, ..., -0.1509, 0.0004, -0.1639], [ 0.0083, -0.1067, 0.0355, ..., 0.0674, -0.1100, -0.0426], [-0.1023, -0.1160, 0.0760, ..., -0.0965, -0.0663, 0.1457]], device='cuda:0'), grad: tensor([[ 4.5635e-08, -1.2051e-06, 5.1688e-08, ..., 8.4750e-08, -7.9488e-07, -1.1213e-06], [-3.5716e-07, 4.5672e-06, 7.6368e-08, ..., 2.5276e-06, 2.4447e-07, 5.3085e-08], [ 4.8894e-08, 1.9651e-07, -5.3877e-07, ..., 1.0058e-07, 2.9802e-08, 5.0757e-08], ..., [-4.5775e-07, -5.0999e-06, 2.2119e-07, ..., 1.0319e-06, -2.2538e-07, 3.0175e-07], [-2.8126e-06, -6.4261e-08, 1.7304e-06, ..., -1.7174e-06, 8.8010e-08, 3.4608e-06], [ 1.7742e-07, 3.6228e-07, -4.9621e-06, ..., 1.1977e-06, 1.5786e-07, -8.9332e-06]], device='cuda:0') Epoch 209, bias, value: tensor([-0.0048, -0.0021, 0.0085, 0.0121, 0.0273, 0.0405, -0.0218, 0.0062, -0.0359, -0.0194], device='cuda:0'), grad: tensor([-3.0678e-06, 1.2301e-05, -6.4746e-06, 4.9267e-07, -3.5930e-06, 6.3442e-06, 3.7961e-06, -3.8147e-06, 1.1176e-05, -1.7270e-05], device='cuda:0') 100 0.0001 changing lr epoch 208, time 217.29, cls_loss 0.0020 cls_loss_mapping 0.0028 cls_loss_causal 0.4938 re_mapping 0.0058 re_causal 0.0168 /// teacc 99.06 lr 0.00010000 Epoch 210, weight, value: tensor([[-0.0577, 0.1443, 0.0762, ..., -0.2187, -0.1199, 0.0242], [ 0.1563, -0.0183, -0.0771, ..., 0.0924, 0.1196, -0.1476], [-0.0955, -0.0576, -0.0540, ..., -0.0924, -0.0381, -0.0672], ..., [-0.0749, 0.1581, -0.1874, ..., -0.1513, -0.0009, -0.1660], [ 0.0083, -0.1069, 0.0357, ..., 0.0675, -0.1101, -0.0424], [-0.1032, -0.1169, 0.0777, ..., -0.0968, -0.0664, 0.1464]], device='cuda:0'), grad: tensor([[ 1.3690e-07, -8.9407e-06, -1.1310e-05, ..., 2.1141e-07, 0.0000e+00, -2.0474e-05], [ 1.7388e-06, 1.2387e-07, 8.2981e-07, ..., 2.6301e-06, 0.0000e+00, 6.1980e-07], [ 1.5292e-06, 2.2398e-07, 2.3097e-07, ..., 2.4550e-06, 0.0000e+00, 7.5949e-07], ..., [ 3.8091e-07, -4.2655e-07, 1.5786e-07, ..., 5.7276e-07, 0.0000e+00, 3.6461e-07], [-9.5740e-06, 1.2321e-06, 4.4564e-07, ..., -1.3642e-05, 0.0000e+00, 1.1502e-07], [ 1.4855e-07, 1.9576e-06, 4.6194e-06, ..., 4.9639e-07, 0.0000e+00, 8.0168e-06]], device='cuda:0') Epoch 210, bias, value: tensor([-0.0043, -0.0017, 0.0084, 0.0121, 0.0268, 0.0392, -0.0199, 0.0059, -0.0360, -0.0190], device='cuda:0'), grad: tensor([-4.0472e-05, 6.4224e-06, 4.7386e-06, 3.1069e-06, -1.6671e-06, 1.8135e-05, 1.8030e-05, 1.2983e-06, -2.7463e-05, 1.7866e-05], device='cuda:0') 100 0.0001 changing lr epoch 209, time 217.16, cls_loss 0.0015 cls_loss_mapping 0.0030 cls_loss_causal 0.5452 re_mapping 0.0058 re_causal 0.0186 /// teacc 99.03 lr 0.00010000 Epoch 211, weight, value: tensor([[-0.0576, 0.1458, 0.0771, ..., -0.2182, -0.1199, 0.0250], [ 0.1566, -0.0200, -0.0774, ..., 0.0927, 0.1203, -0.1478], [-0.0986, -0.0577, -0.0543, ..., -0.0931, -0.0384, -0.0677], ..., [-0.0748, 0.1590, -0.1878, ..., -0.1518, -0.0015, -0.1664], [ 0.0081, -0.1071, 0.0355, ..., 0.0673, -0.1102, -0.0427], [-0.1037, -0.1174, 0.0776, ..., -0.0975, -0.0665, 0.1463]], device='cuda:0'), grad: tensor([[ 3.2596e-09, 5.4482e-08, -9.7789e-09, ..., 1.0710e-08, 0.0000e+00, -1.3504e-08], [-3.9116e-08, 4.9826e-08, 3.2596e-09, ..., -5.8673e-08, 0.0000e+00, 2.7940e-09], [ 5.1223e-09, -4.1090e-06, 1.2573e-08, ..., 3.6787e-08, 0.0000e+00, 1.3970e-09], ..., [ 1.2573e-08, 3.6769e-06, 6.9849e-09, ..., 4.1910e-08, 0.0000e+00, 5.5879e-09], [ 4.2375e-08, 4.2375e-08, 4.6566e-09, ..., 3.0268e-08, 0.0000e+00, 1.9092e-08], [ 9.7789e-09, 4.9360e-08, -6.2305e-07, ..., 1.8021e-07, 0.0000e+00, -6.9616e-07]], device='cuda:0') Epoch 211, bias, value: tensor([-0.0034, -0.0023, 0.0071, 0.0122, 0.0273, 0.0391, -0.0183, 0.0061, -0.0363, -0.0195], device='cuda:0'), grad: tensor([ 1.4529e-07, 3.7253e-08, -1.0625e-05, 4.8801e-07, 9.3644e-07, 5.9139e-08, -1.1316e-07, 9.7528e-06, 1.7043e-07, -8.2469e-07], device='cuda:0') 100 0.0001 changing lr epoch 210, time 217.21, cls_loss 0.0015 cls_loss_mapping 0.0027 cls_loss_causal 0.5085 re_mapping 0.0060 re_causal 0.0176 /// teacc 99.08 lr 0.00010000 Epoch 212, weight, value: tensor([[-0.0579, 0.1453, 0.0758, ..., -0.2194, -0.1199, 0.0245], [ 0.1575, -0.0205, -0.0776, ..., 0.0935, 0.1205, -0.1481], [-0.0988, -0.0578, -0.0545, ..., -0.0933, -0.0384, -0.0679], ..., [-0.0750, 0.1596, -0.1880, ..., -0.1522, -0.0016, -0.1668], [ 0.0081, -0.1072, 0.0358, ..., 0.0674, -0.1104, -0.0426], [-0.1043, -0.1181, 0.0775, ..., -0.0980, -0.0665, 0.1463]], device='cuda:0'), grad: tensor([[ 2.2817e-08, -1.2107e-08, 6.3796e-08, ..., 2.5611e-08, 0.0000e+00, 5.3085e-08], [-2.3935e-07, 2.7809e-06, -8.4285e-08, ..., -2.5379e-07, 0.0000e+00, -2.8871e-08], [ 5.4948e-08, 5.7295e-06, 3.6974e-07, ..., 7.0315e-08, 0.0000e+00, 2.4494e-07], ..., [ 2.8405e-08, -9.2387e-06, 5.4017e-08, ..., 3.5996e-07, 0.0000e+00, 2.5611e-08], [ 2.4447e-07, 2.1607e-07, 2.3982e-07, ..., 2.2817e-07, 0.0000e+00, 2.3609e-07], [ 1.6950e-07, 2.8126e-07, -5.0059e-07, ..., 4.2049e-07, 0.0000e+00, -7.1386e-07]], device='cuda:0') Epoch 212, bias, value: tensor([-0.0040, -0.0019, 0.0070, 0.0121, 0.0277, 0.0393, -0.0184, 0.0063, -0.0363, -0.0200], device='cuda:0'), grad: tensor([ 1.5646e-07, 3.5241e-06, 9.0152e-06, -7.4552e-07, -2.1216e-06, 3.2550e-07, -4.2608e-07, -1.1712e-05, 1.5069e-06, 4.6939e-07], device='cuda:0') 100 0.0001 changing lr epoch 211, time 217.24, cls_loss 0.0014 cls_loss_mapping 0.0020 cls_loss_causal 0.5065 re_mapping 0.0059 re_causal 0.0176 /// teacc 98.98 lr 0.00010000 Epoch 213, weight, value: tensor([[-0.0580, 0.1455, 0.0761, ..., -0.2197, -0.1199, 0.0251], [ 0.1574, -0.0206, -0.0789, ..., 0.0934, 0.1211, -0.1486], [-0.0990, -0.0577, -0.0531, ..., -0.0935, -0.0387, -0.0675], ..., [-0.0751, 0.1602, -0.1878, ..., -0.1526, -0.0021, -0.1673], [ 0.0085, -0.1074, 0.0362, ..., 0.0682, -0.1105, -0.0428], [-0.1046, -0.1188, 0.0775, ..., -0.0990, -0.0665, 0.1465]], device='cuda:0'), grad: tensor([[ 4.6566e-10, -4.6566e-10, 1.1828e-07, ..., 7.9162e-09, 0.0000e+00, -3.2596e-08], [-3.2596e-09, 1.2107e-08, 4.0047e-08, ..., 7.6368e-07, 0.0000e+00, 1.7229e-08], [ 4.1910e-09, 1.4435e-08, 3.9581e-08, ..., 2.5611e-08, 9.3132e-10, 1.2107e-08], ..., [ 1.8626e-09, 4.7823e-07, 9.4436e-07, ..., 1.0012e-07, 0.0000e+00, 8.3866e-07], [-4.6566e-09, 4.1910e-08, 2.3469e-07, ..., 8.0559e-08, 0.0000e+00, 1.4948e-07], [ 1.3970e-09, -7.0967e-07, -8.7079e-07, ..., 1.7602e-07, 0.0000e+00, -2.1365e-06]], device='cuda:0') Epoch 213, bias, value: tensor([-0.0038, -0.0022, 0.0073, 0.0121, 0.0286, 0.0393, -0.0187, 0.0066, -0.0360, -0.0208], device='cuda:0'), grad: tensor([ 1.8347e-07, 1.1437e-06, 1.2526e-07, -2.4084e-06, 1.0226e-06, 2.9989e-07, 1.3597e-07, 3.4738e-06, 7.1805e-07, -4.7013e-06], device='cuda:0') 100 0.0001 changing lr epoch 212, time 217.28, cls_loss 0.0018 cls_loss_mapping 0.0024 cls_loss_causal 0.5026 re_mapping 0.0062 re_causal 0.0173 /// teacc 99.04 lr 0.00010000 Epoch 214, weight, value: tensor([[-0.0580, 0.1459, 0.0757, ..., -0.2200, -0.1200, 0.0252], [ 0.1579, -0.0213, -0.0799, ..., 0.0941, 0.1226, -0.1490], [-0.0992, -0.0579, -0.0532, ..., -0.0936, -0.0392, -0.0678], ..., [-0.0755, 0.1610, -0.1874, ..., -0.1537, -0.0037, -0.1680], [ 0.0087, -0.1075, 0.0365, ..., 0.0684, -0.1106, -0.0430], [-0.1053, -0.1191, 0.0777, ..., -0.0996, -0.0667, 0.1467]], device='cuda:0'), grad: tensor([[ 6.7055e-07, -7.5111e-07, 6.5193e-09, ..., 1.3215e-06, 0.0000e+00, -1.6158e-07], [ 4.4495e-05, 8.0559e-08, 1.7928e-08, ..., 7.9334e-05, 0.0000e+00, 1.1869e-05], [ 7.6368e-07, 3.4925e-08, 3.6089e-08, ..., 1.6834e-07, 0.0000e+00, 2.2794e-07], ..., [ 8.5449e-08, 7.1479e-07, 6.0908e-07, ..., 2.0280e-07, 0.0000e+00, 1.0561e-06], [ 5.9232e-06, 3.5437e-07, 4.7777e-07, ..., 1.2480e-05, 0.0000e+00, 2.3134e-06], [ 1.0733e-07, -3.5670e-07, -6.2957e-07, ..., 7.3574e-07, 0.0000e+00, -8.0839e-07]], device='cuda:0') Epoch 214, bias, value: tensor([-0.0038, -0.0026, 0.0074, 0.0111, 0.0290, 0.0400, -0.0189, 0.0070, -0.0359, -0.0213], device='cuda:0'), grad: tensor([ 1.9744e-06, 1.9920e-04, -1.6198e-05, -1.6508e-07, 2.5928e-06, 5.9009e-05, -2.9635e-04, 3.3285e-06, 4.7743e-05, -7.9628e-07], device='cuda:0') 100 0.0001 changing lr epoch 213, time 217.33, cls_loss 0.0014 cls_loss_mapping 0.0030 cls_loss_causal 0.5260 re_mapping 0.0059 re_causal 0.0177 /// teacc 99.01 lr 0.00010000 Epoch 215, weight, value: tensor([[-0.0581, 0.1459, 0.0754, ..., -0.2209, -0.1200, 0.0248], [ 0.1584, -0.0212, -0.0799, ..., 0.0943, 0.1233, -0.1493], [-0.0993, -0.0580, -0.0533, ..., -0.0938, -0.0393, -0.0683], ..., [-0.0759, 0.1612, -0.1879, ..., -0.1545, -0.0045, -0.1688], [ 0.0088, -0.1076, 0.0360, ..., 0.0684, -0.1107, -0.0440], [-0.1058, -0.1193, 0.0793, ..., -0.0996, -0.0668, 0.1482]], device='cuda:0'), grad: tensor([[ 6.9849e-09, -5.1223e-09, 6.9849e-09, ..., 1.1176e-08, 2.3283e-10, -1.4668e-08], [-3.6787e-08, 6.7148e-07, 2.0256e-08, ..., -4.5402e-08, -1.1642e-09, 2.3283e-09], [ 4.3074e-08, 6.2399e-08, 2.5425e-07, ..., 6.6124e-08, 2.3283e-10, 1.8626e-09], ..., [ 1.6065e-08, -1.0207e-06, 1.5087e-07, ..., 3.2363e-08, 4.6566e-10, 1.0710e-08], [ 1.3271e-08, 6.5891e-08, 1.3900e-07, ..., 3.1432e-08, 0.0000e+00, 4.9127e-08], [ 4.4238e-09, 1.7066e-07, 4.7497e-08, ..., 2.5053e-07, 2.3283e-10, -6.7521e-09]], device='cuda:0') Epoch 215, bias, value: tensor([-0.0040, -0.0022, 0.0074, 0.0111, 0.0289, 0.0403, -0.0194, 0.0067, -0.0363, -0.0206], device='cuda:0'), grad: tensor([ 6.8452e-08, 1.2163e-06, 6.5612e-07, -2.9728e-06, -2.4564e-07, 2.3167e-07, -1.9791e-07, -5.0291e-07, 8.5495e-07, 8.9873e-07], device='cuda:0') 100 0.0001 changing lr epoch 214, time 217.31, cls_loss 0.0015 cls_loss_mapping 0.0022 cls_loss_causal 0.5305 re_mapping 0.0057 re_causal 0.0175 /// teacc 99.09 lr 0.00010000 Epoch 216, weight, value: tensor([[-0.0582, 0.1460, 0.0744, ..., -0.2211, -0.1200, 0.0240], [ 0.1592, -0.0212, -0.0800, ..., 0.0948, 0.1236, -0.1494], [-0.0994, -0.0582, -0.0535, ..., -0.0939, -0.0395, -0.0685], ..., [-0.0760, 0.1616, -0.1886, ..., -0.1549, -0.0049, -0.1697], [ 0.0086, -0.1076, 0.0359, ..., 0.0681, -0.1110, -0.0445], [-0.1067, -0.1198, 0.0800, ..., -0.1000, -0.0668, 0.1489]], device='cuda:0'), grad: tensor([[ 1.1711e-07, -2.0023e-08, 1.6321e-07, ..., 3.9535e-07, 0.0000e+00, 5.3830e-07], [ 3.0082e-07, 1.0594e-07, 1.2293e-07, ..., 1.7392e-07, -9.3132e-10, 2.8149e-07], [ 2.3702e-07, 2.1420e-08, 1.0408e-07, ..., 2.5798e-07, 2.3283e-10, 2.6077e-07], ..., [ 1.6857e-07, -3.9628e-07, 4.7032e-08, ..., 1.5530e-07, 2.3283e-10, 9.8022e-08], [ 4.7982e-05, 2.5146e-08, 1.2510e-05, ..., 3.9101e-05, 2.3283e-10, 3.2097e-05], [ 4.9034e-07, 1.4575e-07, 2.0093e-07, ..., 4.8289e-07, 2.3283e-10, 3.6089e-07]], device='cuda:0') Epoch 216, bias, value: tensor([-0.0045, -0.0019, 0.0073, 0.0113, 0.0291, 0.0403, -0.0196, 0.0065, -0.0367, -0.0205], device='cuda:0'), grad: tensor([ 1.1399e-06, 2.3097e-06, -6.9337e-07, -2.4915e-05, 1.6298e-07, -1.2577e-04, 2.8014e-05, -3.2340e-07, 1.1843e-04, 1.7611e-06], device='cuda:0') 100 0.0001 changing lr epoch 215, time 217.47, cls_loss 0.0018 cls_loss_mapping 0.0033 cls_loss_causal 0.5191 re_mapping 0.0059 re_causal 0.0176 /// teacc 99.05 lr 0.00010000 Epoch 217, weight, value: tensor([[-0.0583, 0.1463, 0.0746, ..., -0.2214, -0.1200, 0.0243], [ 0.1586, -0.0222, -0.0823, ..., 0.0918, 0.1211, -0.1503], [-0.0984, -0.0585, -0.0512, ..., -0.0911, -0.0373, -0.0687], ..., [-0.0761, 0.1625, -0.1887, ..., -0.1550, -0.0050, -0.1701], [ 0.0085, -0.1077, 0.0356, ..., 0.0678, -0.1113, -0.0451], [-0.1072, -0.1202, 0.0799, ..., -0.1027, -0.0665, 0.1482]], device='cuda:0'), grad: tensor([[ 1.0245e-08, -6.4261e-08, 2.9802e-08, ..., 2.2352e-08, 3.8184e-08, -1.7229e-08], [-9.1735e-08, 6.8918e-08, 2.0955e-08, ..., -1.1409e-07, 2.7940e-09, 1.6764e-08], [ 6.6124e-08, 6.1002e-08, 6.7055e-08, ..., 3.4599e-07, -2.5798e-07, 7.5437e-08], ..., [ 3.7719e-08, -2.6822e-07, 1.5832e-07, ..., 9.7323e-08, 6.1467e-08, 2.5006e-07], [-8.4937e-06, 4.5635e-08, 1.0571e-07, ..., -1.7956e-05, 7.5903e-08, -4.2021e-06], [ 1.0710e-08, 9.5926e-08, -3.4971e-07, ..., 1.7416e-07, 2.8871e-08, -6.3144e-07]], device='cuda:0') Epoch 217, bias, value: tensor([-0.0042, -0.0051, 0.0099, 0.0113, 0.0316, 0.0411, -0.0205, 0.0070, -0.0371, -0.0224], device='cuda:0'), grad: tensor([ 1.1129e-07, 1.2573e-07, -1.1213e-06, -2.1979e-07, 7.2177e-08, 8.1137e-06, 2.1562e-05, 6.1048e-07, -2.8655e-05, -5.8347e-07], device='cuda:0') 100 0.0001 changing lr epoch 216, time 217.48, cls_loss 0.0029 cls_loss_mapping 0.0025 cls_loss_causal 0.4983 re_mapping 0.0060 re_causal 0.0169 /// teacc 98.98 lr 0.00010000 Epoch 218, weight, value: tensor([[-0.0588, 0.1465, 0.0747, ..., -0.2231, -0.1200, 0.0246], [ 0.1577, -0.0219, -0.0854, ..., 0.0887, 0.1181, -0.1537], [-0.0963, -0.0584, -0.0481, ..., -0.0881, -0.0345, -0.0692], ..., [-0.0763, 0.1626, -0.1892, ..., -0.1558, -0.0050, -0.1703], [ 0.0077, -0.1079, 0.0354, ..., 0.0672, -0.1118, -0.0454], [-0.1082, -0.1205, 0.0803, ..., -0.1018, -0.0615, 0.1497]], device='cuda:0'), grad: tensor([[ 1.7043e-07, -2.3060e-06, 4.6566e-09, ..., 6.6450e-07, 2.3283e-09, -5.5600e-07], [-1.1846e-06, 4.0419e-07, 1.0710e-08, ..., -1.0356e-06, -1.2107e-06, 2.1001e-07], [ 1.4342e-06, 1.7462e-07, 4.4238e-08, ..., 1.3430e-06, 1.1157e-06, 3.5111e-07], ..., [ 1.7835e-07, 4.6892e-07, 1.8161e-08, ..., 3.5390e-07, 1.7788e-07, 8.8708e-07], [-2.5287e-05, 9.1363e-07, 2.3609e-07, ..., -3.4094e-05, 2.3283e-09, -2.4036e-05], [ 2.9895e-07, 1.0151e-07, 1.6345e-07, ..., 4.0326e-07, 6.5193e-09, 3.0268e-07]], device='cuda:0') Epoch 218, bias, value: tensor([-0.0043, -0.0081, 0.0128, 0.0114, 0.0314, 0.0411, -0.0205, 0.0068, -0.0380, -0.0213], device='cuda:0'), grad: tensor([-3.6750e-06, -3.9674e-07, 2.6599e-06, -3.8333e-06, -7.2177e-08, 1.3970e-05, 6.4492e-05, 2.4028e-06, -7.7188e-05, 1.6298e-06], device='cuda:0') 100 0.0001 changing lr epoch 217, time 217.23, cls_loss 0.0020 cls_loss_mapping 0.0029 cls_loss_causal 0.5147 re_mapping 0.0058 re_causal 0.0169 /// teacc 98.94 lr 0.00010000 Epoch 219, weight, value: tensor([[-0.0590, 0.1468, 0.0749, ..., -0.2261, -0.1228, 0.0230], [ 0.1576, -0.0221, -0.0856, ..., 0.0887, 0.1181, -0.1546], [-0.0965, -0.0586, -0.0481, ..., -0.0882, -0.0345, -0.0699], ..., [-0.0769, 0.1625, -0.1904, ..., -0.1561, -0.0051, -0.1729], [ 0.0086, -0.1072, 0.0357, ..., 0.0679, -0.1119, -0.0451], [-0.1085, -0.1205, 0.0814, ..., -0.1025, -0.0612, 0.1505]], device='cuda:0'), grad: tensor([[ 9.7789e-09, 4.6566e-09, 5.5879e-09, ..., 6.0070e-08, 0.0000e+00, 1.5832e-08], [-5.5740e-07, 6.6590e-08, 5.5879e-09, ..., -1.1241e-06, -2.7940e-09, 7.6368e-08], [ 5.2154e-08, 5.4948e-08, 8.6147e-08, ..., 2.2585e-07, 3.2596e-09, 2.0955e-08], ..., [ 6.9849e-08, -1.4529e-07, 1.5413e-07, ..., 2.3888e-07, 2.3283e-09, 4.7572e-06], [ 1.9092e-08, 1.0431e-07, 1.2573e-08, ..., 2.5239e-07, 2.7940e-09, 2.5611e-07], [ 2.2352e-08, -2.9709e-07, -5.3039e-07, ..., 1.2442e-05, 4.6566e-10, -3.2485e-06]], device='cuda:0') Epoch 219, bias, value: tensor([-0.0065, -0.0082, 0.0127, 0.0114, 0.0327, 0.0414, -0.0192, 0.0056, -0.0366, -0.0214], device='cuda:0'), grad: tensor([ 1.4761e-07, -1.3076e-06, 6.6636e-07, -8.6613e-08, -9.4846e-06, -8.1025e-08, -4.9546e-06, 1.0252e-05, 9.8161e-07, 3.8594e-06], device='cuda:0') 100 0.0001 changing lr epoch 218, time 216.62, cls_loss 0.0016 cls_loss_mapping 0.0022 cls_loss_causal 0.5047 re_mapping 0.0059 re_causal 0.0172 /// teacc 99.02 lr 0.00010000 Epoch 220, weight, value: tensor([[-0.0591, 0.1469, 0.0748, ..., -0.2263, -0.1228, 0.0231], [ 0.1577, -0.0228, -0.0858, ..., 0.0887, 0.1182, -0.1546], [-0.0966, -0.0587, -0.0481, ..., -0.0882, -0.0346, -0.0699], ..., [-0.0770, 0.1628, -0.1917, ..., -0.1564, -0.0051, -0.1730], [ 0.0085, -0.1071, 0.0358, ..., 0.0677, -0.1121, -0.0455], [-0.1087, -0.1204, 0.0823, ..., -0.1030, -0.0612, 0.1506]], device='cuda:0'), grad: tensor([[ 3.7253e-09, -1.8626e-08, 1.7043e-07, ..., 2.1327e-07, 0.0000e+00, 1.0524e-07], [-1.4761e-07, 1.0300e-06, 8.3353e-08, ..., -2.1560e-07, -5.1223e-09, 1.6904e-07], [ 2.5611e-08, 1.8068e-07, 7.5903e-08, ..., 6.6124e-08, 1.3970e-09, 1.1409e-07], ..., [ 3.1199e-08, -5.9716e-06, 2.9113e-06, ..., 1.1502e-07, 2.3283e-09, 4.4852e-06], [ 6.7521e-08, 3.5251e-07, 1.6810e-07, ..., -9.2089e-06, 0.0000e+00, 1.9632e-06], [ 1.8626e-08, 3.9823e-06, -7.2345e-06, ..., 4.2289e-05, 9.3132e-10, -1.5348e-05]], device='cuda:0') Epoch 220, bias, value: tensor([-0.0065, -0.0082, 0.0127, 0.0113, 0.0330, 0.0414, -0.0189, 0.0054, -0.0365, -0.0213], device='cuda:0'), grad: tensor([ 3.3621e-07, 2.1905e-06, 7.5437e-07, -2.1281e-07, -7.6354e-05, 8.5607e-06, 8.5123e-07, -6.0536e-06, -2.1532e-06, 7.1824e-05], device='cuda:0') 100 0.0001 changing lr epoch 219, time 216.92, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.5224 re_mapping 0.0055 re_causal 0.0169 /// teacc 99.06 lr 0.00010000 Epoch 221, weight, value: tensor([[-0.0591, 0.1474, 0.0750, ..., -0.2264, -0.1228, 0.0233], [ 0.1587, -0.0205, -0.0858, ..., 0.0888, 0.1182, -0.1548], [-0.0967, -0.0590, -0.0481, ..., -0.0882, -0.0346, -0.0701], ..., [-0.0782, 0.1624, -0.1918, ..., -0.1591, -0.0052, -0.1732], [ 0.0084, -0.1072, 0.0360, ..., 0.0678, -0.1128, -0.0459], [-0.1096, -0.1214, 0.0827, ..., -0.1037, -0.0612, 0.1508]], device='cuda:0'), grad: tensor([[ 8.3819e-09, -7.8231e-08, -8.8476e-08, ..., 1.0245e-08, 9.3132e-10, -2.5565e-07], [-2.9383e-07, 3.3528e-08, 3.7253e-09, ..., -4.1910e-07, -2.2817e-08, 2.7940e-09], [ 1.3318e-07, 4.8429e-08, 2.3283e-09, ..., 2.2212e-07, -6.0536e-09, 7.4506e-09], ..., [ 6.7987e-08, -3.6322e-07, 1.0245e-08, ..., 1.0710e-07, 1.3970e-08, 1.2107e-08], [-2.8405e-08, 2.0023e-08, 4.1910e-09, ..., -4.1910e-08, 5.5879e-09, 1.6298e-08], [ 2.5146e-08, 3.0687e-07, -4.0978e-08, ..., 1.7090e-07, 2.3283e-09, 8.6613e-08]], device='cuda:0') Epoch 221, bias, value: tensor([-0.0064, -0.0081, 0.0127, 0.0111, 0.0343, 0.0417, -0.0200, 0.0050, -0.0366, -0.0218], device='cuda:0'), grad: tensor([-4.2375e-07, -7.7160e-07, 1.0477e-07, 3.2596e-08, 0.0000e+00, 1.5227e-07, 1.8161e-07, -1.5041e-07, 7.8697e-08, 8.1956e-07], device='cuda:0') 100 0.0001 changing lr epoch 220, time 216.54, cls_loss 0.0014 cls_loss_mapping 0.0027 cls_loss_causal 0.4930 re_mapping 0.0055 re_causal 0.0160 /// teacc 99.03 lr 0.00010000 Epoch 222, weight, value: tensor([[-0.0592, 0.1480, 0.0733, ..., -0.2265, -0.1228, 0.0216], [ 0.1596, -0.0203, -0.0857, ..., 0.0889, 0.1184, -0.1548], [-0.0971, -0.0593, -0.0481, ..., -0.0882, -0.0347, -0.0704], ..., [-0.0786, 0.1628, -0.1920, ..., -0.1598, -0.0058, -0.1736], [ 0.0084, -0.1073, 0.0359, ..., 0.0678, -0.1129, -0.0460], [-0.1101, -0.1218, 0.0842, ..., -0.1038, -0.0612, 0.1519]], device='cuda:0'), grad: tensor([[ 9.8720e-08, -8.2422e-08, 4.1910e-09, ..., 1.2666e-07, 0.0000e+00, -1.4901e-08], [-1.0282e-06, 3.5856e-08, 8.3819e-09, ..., -1.4063e-06, 0.0000e+00, 1.1176e-08], [ 2.6729e-07, 2.3283e-08, 1.2573e-08, ..., 3.6322e-07, 4.6566e-10, 1.9092e-08], ..., [ 7.4692e-07, -1.4622e-07, 3.2596e-08, ..., 1.0179e-06, 4.6566e-10, 4.4703e-08], [ 7.0920e-07, 3.3993e-08, 1.0524e-07, ..., 8.4937e-07, 9.3132e-10, 2.3004e-07], [ 7.9628e-08, 5.6345e-08, -2.9942e-07, ..., 1.0943e-07, 0.0000e+00, -4.8755e-07]], device='cuda:0') Epoch 222, bias, value: tensor([-0.0073, -0.0079, 0.0126, 0.0108, 0.0341, 0.0421, -0.0204, 0.0051, -0.0367, -0.0211], device='cuda:0'), grad: tensor([ 1.1921e-07, -3.4031e-06, 1.0990e-06, -2.9337e-08, 2.9542e-06, 1.5656e-06, -6.7167e-06, 2.4159e-06, 2.4382e-06, -4.2794e-07], device='cuda:0') 100 0.0001 changing lr epoch 221, time 217.14, cls_loss 0.0017 cls_loss_mapping 0.0026 cls_loss_causal 0.5020 re_mapping 0.0057 re_causal 0.0165 /// teacc 99.03 lr 0.00010000 Epoch 223, weight, value: tensor([[-0.0594, 0.1482, 0.0713, ..., -0.2269, -0.1228, 0.0198], [ 0.1602, -0.0204, -0.0858, ..., 0.0890, 0.1184, -0.1549], [-0.0973, -0.0594, -0.0481, ..., -0.0883, -0.0347, -0.0706], ..., [-0.0787, 0.1630, -0.1935, ..., -0.1600, -0.0063, -0.1739], [ 0.0082, -0.1074, 0.0353, ..., 0.0676, -0.1130, -0.0465], [-0.1105, -0.1219, 0.0856, ..., -0.1047, -0.0612, 0.1529]], device='cuda:0'), grad: tensor([[ 9.3132e-09, -2.5425e-07, -2.9244e-07, ..., 2.7008e-08, 0.0000e+00, -5.3644e-07], [-2.1420e-08, 7.4506e-08, 1.3970e-08, ..., -2.5146e-08, -1.8626e-09, 2.4214e-08], [ 1.8626e-07, 4.5728e-07, 1.2387e-07, ..., 2.9057e-07, 9.3132e-10, 1.3132e-07], ..., [ 2.3283e-08, -8.2608e-07, 5.3085e-08, ..., 5.6811e-08, 9.3132e-10, 9.2201e-08], [-1.7043e-07, 6.2399e-08, 1.4910e-06, ..., -2.1141e-07, 0.0000e+00, 3.4273e-06], [ 7.3574e-08, 3.9488e-07, -1.4240e-06, ..., 6.2492e-07, 0.0000e+00, -3.1367e-06]], device='cuda:0') Epoch 223, bias, value: tensor([-0.0088, -0.0079, 0.0125, 0.0109, 0.0351, 0.0422, -0.0203, 0.0046, -0.0370, -0.0205], device='cuda:0'), grad: tensor([-2.6040e-06, 1.4249e-07, 1.6605e-06, 1.0179e-06, -4.1164e-07, -1.1949e-06, -8.3819e-09, -9.9372e-07, 6.6608e-06, -4.2729e-06], device='cuda:0') 100 0.0001 changing lr epoch 222, time 217.06, cls_loss 0.0022 cls_loss_mapping 0.0034 cls_loss_causal 0.5202 re_mapping 0.0056 re_causal 0.0163 /// teacc 98.91 lr 0.00010000 Epoch 224, weight, value: tensor([[-0.0595, 0.1483, 0.0718, ..., -0.2272, -0.1228, 0.0199], [ 0.1603, -0.0203, -0.0863, ..., 0.0890, 0.1184, -0.1556], [-0.0977, -0.0599, -0.0480, ..., -0.0883, -0.0347, -0.0711], ..., [-0.0791, 0.1639, -0.1938, ..., -0.1605, -0.0064, -0.1743], [ 0.0091, -0.1077, 0.0362, ..., 0.0685, -0.1132, -0.0462], [-0.1124, -0.1228, 0.0852, ..., -0.1069, -0.0612, 0.1519]], device='cuda:0'), grad: tensor([[ 1.7695e-08, -8.3819e-09, 1.6578e-07, ..., 2.4214e-08, 2.7940e-09, 6.5193e-09], [-8.4750e-08, 6.2399e-08, 7.8231e-08, ..., -1.4342e-07, -3.3528e-08, 9.3132e-09], [ 1.0803e-07, 6.0536e-08, 3.5763e-07, ..., 1.8906e-07, 5.5879e-09, 3.7253e-08], ..., [ 4.0047e-08, -5.3458e-07, 4.4424e-07, ..., 7.2643e-08, -1.3970e-08, 3.1665e-08], [-5.4352e-06, 1.5832e-08, -1.2763e-05, ..., -1.0870e-05, 1.0245e-08, -6.8173e-06], [ 1.8533e-07, 3.6415e-07, 1.5302e-06, ..., 3.6601e-07, 1.9558e-08, 2.4866e-07]], device='cuda:0') Epoch 224, bias, value: tensor([-0.0086, -0.0080, 0.0127, 0.0107, 0.0376, 0.0420, -0.0203, 0.0048, -0.0368, -0.0227], device='cuda:0'), grad: tensor([ 3.6787e-07, -5.8673e-08, 5.6811e-07, -1.5125e-06, 1.5832e-07, 2.3916e-05, -3.8743e-06, 4.0233e-07, -2.3559e-05, 3.6154e-06], device='cuda:0') 100 0.0001 changing lr epoch 223, time 216.75, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.4844 re_mapping 0.0060 re_causal 0.0168 /// teacc 98.87 lr 0.00010000 Epoch 225, weight, value: tensor([[-0.0596, 0.1493, 0.0725, ..., -0.2275, -0.1228, 0.0206], [ 0.1616, -0.0204, -0.0864, ..., 0.0891, 0.1186, -0.1557], [-0.0982, -0.0603, -0.0481, ..., -0.0884, -0.0348, -0.0714], ..., [-0.0789, 0.1686, -0.1908, ..., -0.1608, -0.0063, -0.1748], [ 0.0091, -0.1078, 0.0362, ..., 0.0686, -0.1134, -0.0465], [-0.1167, -0.1290, 0.0826, ..., -0.1074, -0.0613, 0.1521]], device='cuda:0'), grad: tensor([[ 1.8626e-08, -1.2387e-07, -4.0047e-08, ..., 1.3970e-08, 1.3970e-08, -7.4506e-08], [ 3.6322e-08, 8.8476e-08, 3.3528e-08, ..., 5.5879e-09, 5.2154e-08, 4.6566e-08], [ 2.6077e-08, 2.2631e-07, 2.9802e-08, ..., 1.9558e-08, 1.2387e-07, 2.4214e-08], ..., [ 8.3819e-09, -1.8105e-06, 7.8231e-08, ..., 9.3132e-09, -8.2981e-07, 1.1455e-07], [ 1.4249e-07, 8.1025e-08, 1.5926e-07, ..., 8.0094e-08, 1.0803e-07, 1.7695e-07], [ 3.3528e-08, 7.5437e-08, -1.7602e-07, ..., 3.2596e-08, 2.7008e-08, -2.6450e-07]], device='cuda:0') Epoch 225, bias, value: tensor([-0.0080, -0.0079, 0.0126, 0.0106, 0.0375, 0.0421, -0.0204, 0.0089, -0.0370, -0.0266], device='cuda:0'), grad: tensor([-2.7567e-07, 4.3213e-07, 2.2724e-07, 5.4650e-06, 1.9670e-06, -5.6587e-06, 1.7416e-07, -2.9486e-06, 7.4785e-07, -1.5087e-07], device='cuda:0') 100 0.0001 changing lr epoch 224, time 217.68, cls_loss 0.0022 cls_loss_mapping 0.0030 cls_loss_causal 0.5144 re_mapping 0.0060 re_causal 0.0165 /// teacc 98.99 lr 0.00010000 Epoch 226, weight, value: tensor([[-0.0597, 0.1498, 0.0721, ..., -0.2278, -0.1228, 0.0192], [ 0.1633, -0.0186, -0.0864, ..., 0.0891, 0.1186, -0.1558], [-0.0984, -0.0599, -0.0481, ..., -0.0884, -0.0349, -0.0736], ..., [-0.0811, 0.1683, -0.1910, ..., -0.1621, -0.0060, -0.1767], [ 0.0096, -0.1080, 0.0367, ..., 0.0696, -0.1142, -0.0464], [-0.1175, -0.1290, 0.0829, ..., -0.1081, -0.0615, 0.1538]], device='cuda:0'), grad: tensor([[ 3.1665e-08, -1.4529e-07, -1.8906e-07, ..., 4.1910e-08, 6.5193e-09, -1.5646e-07], [-1.1828e-07, 6.7987e-08, 7.4506e-09, ..., -1.9185e-07, -1.8626e-08, 2.6077e-08], [ 6.4261e-08, 1.5553e-07, 1.0803e-07, ..., 1.0431e-07, 1.2107e-08, 1.0803e-07], ..., [ 1.0058e-07, -2.0955e-07, 1.6764e-08, ..., 1.4342e-07, 6.2399e-08, 2.3097e-07], [ 5.3644e-07, 1.8626e-08, 5.3830e-07, ..., 7.3109e-07, 1.1548e-07, 1.7788e-06], [ 9.2201e-08, 7.6368e-08, -1.5832e-08, ..., 1.4901e-07, 2.8871e-08, -1.3039e-08]], device='cuda:0') Epoch 226, bias, value: tensor([-0.0090, -0.0075, 0.0126, 0.0106, 0.0370, 0.0419, -0.0201, 0.0083, -0.0369, -0.0261], device='cuda:0'), grad: tensor([-2.6636e-06, -1.8254e-07, 1.9558e-06, 1.1483e-06, 7.0781e-08, -3.1590e-05, 2.7820e-05, 5.9605e-08, 2.8759e-06, 4.7684e-07], device='cuda:0') 100 0.0001 changing lr epoch 225, time 218.10, cls_loss 0.0019 cls_loss_mapping 0.0031 cls_loss_causal 0.4940 re_mapping 0.0056 re_causal 0.0161 /// teacc 98.94 lr 0.00010000 Epoch 227, weight, value: tensor([[-0.0598, 0.1510, 0.0723, ..., -0.2282, -0.1228, 0.0195], [ 0.1639, -0.0189, -0.0865, ..., 0.0892, 0.1187, -0.1568], [-0.0989, -0.0605, -0.0482, ..., -0.0884, -0.0350, -0.0738], ..., [-0.0826, 0.1684, -0.1911, ..., -0.1624, -0.0058, -0.1786], [ 0.0108, -0.1070, 0.0359, ..., 0.0689, -0.1149, -0.0475], [-0.1177, -0.1291, 0.0834, ..., -0.1082, -0.0610, 0.1550]], device='cuda:0'), grad: tensor([[ 5.8766e-07, -1.5832e-08, 5.4017e-08, ..., 4.4052e-07, 1.8626e-08, 1.0058e-07], [-5.1260e-06, 1.6764e-08, 2.0489e-08, ..., -4.6566e-06, -2.6450e-07, 7.6368e-08], [ 1.2713e-06, 5.3085e-08, 1.7416e-07, ..., 1.2089e-06, 1.7043e-07, 5.4017e-08], ..., [ 1.1120e-06, -5.0291e-08, 1.9744e-07, ..., 7.7300e-07, 7.0781e-08, 1.5274e-06], [ 9.8813e-07, 4.7497e-08, -4.5076e-07, ..., -6.6869e-07, 2.8871e-08, -2.0489e-08], [-1.4435e-07, 4.0978e-08, 2.9616e-07, ..., 1.1800e-06, 3.1665e-08, -3.4831e-06]], device='cuda:0') Epoch 227, bias, value: tensor([-0.0086, -0.0077, 0.0126, 0.0112, 0.0366, 0.0416, -0.0200, 0.0076, -0.0349, -0.0259], device='cuda:0'), grad: tensor([ 1.7267e-06, -1.5303e-05, 4.9621e-06, -3.6377e-06, 4.6417e-06, 3.3434e-07, 2.3767e-06, 5.4948e-06, 1.8617e-06, -2.5108e-06], device='cuda:0') 100 0.0001 changing lr epoch 226, time 217.40, cls_loss 0.0018 cls_loss_mapping 0.0026 cls_loss_causal 0.4829 re_mapping 0.0056 re_causal 0.0158 /// teacc 98.94 lr 0.00010000 Epoch 228, weight, value: tensor([[-0.0601, 0.1515, 0.0722, ..., -0.2285, -0.1228, 0.0194], [ 0.1644, -0.0199, -0.0862, ..., 0.0894, 0.1189, -0.1562], [-0.0998, -0.0607, -0.0484, ..., -0.0885, -0.0352, -0.0763], ..., [-0.0824, 0.1687, -0.1912, ..., -0.1631, -0.0053, -0.1791], [ 0.0109, -0.1073, 0.0362, ..., 0.0693, -0.1169, -0.0475], [-0.1181, -0.1291, 0.0837, ..., -0.1082, -0.0613, 0.1560]], device='cuda:0'), grad: tensor([[ 2.0489e-08, -9.8720e-08, 1.4808e-07, ..., 3.0734e-08, 2.1420e-08, 3.9395e-07], [-2.1651e-05, 2.0582e-07, 1.9558e-08, ..., -3.2306e-05, -2.4185e-05, 3.7253e-08], [ 1.9923e-05, -3.1386e-07, -1.3784e-07, ..., 2.9758e-05, 2.2277e-05, 1.8999e-07], ..., [ 1.0384e-06, -6.7614e-07, 6.0536e-08, ..., 1.5581e-06, 1.1111e-06, 9.4064e-08], [ 1.4715e-07, 1.0710e-07, 8.4564e-07, ..., 3.4086e-07, 9.1270e-08, 1.4752e-06], [ 5.7742e-08, 7.0687e-07, -1.2834e-06, ..., 9.4995e-08, 8.6613e-08, -2.5295e-06]], device='cuda:0') Epoch 228, bias, value: tensor([-0.0087, -0.0078, 0.0125, 0.0112, 0.0361, 0.0412, -0.0198, 0.0079, -0.0350, -0.0258], device='cuda:0'), grad: tensor([ 6.1467e-07, -9.4473e-05, 7.0930e-05, 1.4389e-06, 1.0312e-05, 5.4669e-07, 2.8275e-06, 5.6662e-06, 5.8040e-06, -3.8035e-06], device='cuda:0') 100 0.0001 changing lr epoch 227, time 217.62, cls_loss 0.0017 cls_loss_mapping 0.0030 cls_loss_causal 0.5094 re_mapping 0.0058 re_causal 0.0163 /// teacc 98.94 lr 0.00010000 Epoch 229, weight, value: tensor([[-0.0599, 0.1530, 0.0729, ..., -0.2287, -0.1229, 0.0197], [ 0.1647, -0.0201, -0.0863, ..., 0.0895, 0.1189, -0.1564], [-0.1001, -0.0611, -0.0484, ..., -0.0886, -0.0353, -0.0764], ..., [-0.0824, 0.1688, -0.1913, ..., -0.1643, -0.0053, -0.1806], [ 0.0107, -0.1076, 0.0363, ..., 0.0692, -0.1176, -0.0479], [-0.1183, -0.1291, 0.0836, ..., -0.1086, -0.0614, 0.1568]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 8.3819e-09, 1.0245e-08, ..., 6.5193e-09, 5.5879e-09, 9.3132e-10], [-2.0768e-07, 6.8247e-05, 9.3132e-10, ..., -4.0140e-07, 5.2243e-05, 0.0000e+00], [ 1.7416e-07, 1.8999e-07, 1.9558e-08, ..., 3.8277e-07, 3.4925e-07, 1.8626e-09], ..., [ 8.3819e-09, -7.4744e-05, 9.3132e-09, ..., 3.1665e-08, -5.7399e-05, 2.7940e-09], [ 1.1176e-08, 5.4985e-06, 7.4506e-09, ..., 1.8626e-08, 4.2245e-06, 5.5879e-09], [ 4.6566e-09, 1.9185e-07, -4.6566e-09, ..., 5.3085e-08, 1.3132e-07, -1.9558e-08]], device='cuda:0') Epoch 229, bias, value: tensor([-0.0079, -0.0077, 0.0125, 0.0108, 0.0356, 0.0427, -0.0207, 0.0079, -0.0353, -0.0256], device='cuda:0'), grad: tensor([ 3.7812e-07, 3.7384e-04, -8.3148e-06, 2.3358e-06, 1.8254e-07, 9.9745e-07, 1.0896e-06, -4.0436e-04, 3.1590e-05, 1.6131e-06], device='cuda:0') 100 0.0001 changing lr epoch 228, time 217.74, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.5074 re_mapping 0.0057 re_causal 0.0163 /// teacc 99.01 lr 0.00010000 Epoch 230, weight, value: tensor([[-0.0601, 0.1533, 0.0731, ..., -0.2288, -0.1229, 0.0199], [ 0.1652, -0.0205, -0.0864, ..., 0.0896, 0.1189, -0.1566], [-0.1004, -0.0609, -0.0485, ..., -0.0887, -0.0353, -0.0772], ..., [-0.0827, 0.1692, -0.1913, ..., -0.1636, -0.0047, -0.1814], [ 0.0109, -0.1091, 0.0364, ..., 0.0690, -0.1188, -0.0477], [-0.1190, -0.1291, 0.0841, ..., -0.1089, -0.0614, 0.1574]], device='cuda:0'), grad: tensor([[ 1.4529e-07, -1.5339e-06, -3.9712e-06, ..., 1.4901e-07, -3.4180e-06, -5.3197e-06], [-2.4214e-06, -2.6077e-08, 9.4995e-08, ..., -2.4997e-06, -2.3562e-07, 8.1956e-08], [ 1.0887e-06, 2.9057e-07, 3.7178e-06, ..., 1.1204e-06, 4.3958e-07, 2.4904e-06], ..., [ 3.5018e-07, -4.5821e-07, 2.0582e-07, ..., 3.6880e-07, 7.6368e-08, 1.5926e-07], [ 5.4669e-07, 1.1176e-07, 2.2091e-06, ..., 5.6904e-07, 1.3132e-07, 1.4277e-06], [ 1.3225e-07, 5.0850e-07, -9.3803e-06, ..., 2.2724e-07, 2.5146e-07, -5.4054e-06]], device='cuda:0') Epoch 230, bias, value: tensor([-0.0078, -0.0078, 0.0125, 0.0107, 0.0350, 0.0427, -0.0208, 0.0082, -0.0358, -0.0255], device='cuda:0'), grad: tensor([-2.2903e-05, -6.9961e-06, 1.1533e-05, 1.0030e-06, 3.4496e-06, 1.8671e-05, 8.9593e-07, 7.5530e-07, 5.8673e-06, -1.2323e-05], device='cuda:0') 100 0.0001 changing lr epoch 229, time 216.99, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.4928 re_mapping 0.0053 re_causal 0.0158 /// teacc 99.11 lr 0.00010000 Epoch 231, weight, value: tensor([[-0.0604, 0.1536, 0.0737, ..., -0.2292, -0.1228, 0.0202], [ 0.1655, -0.0204, -0.0866, ..., 0.0896, 0.1190, -0.1571], [-0.1009, -0.0612, -0.0486, ..., -0.0888, -0.0354, -0.0775], ..., [-0.0830, 0.1693, -0.1914, ..., -0.1642, -0.0045, -0.1816], [ 0.0114, -0.1092, 0.0369, ..., 0.0697, -0.1201, -0.0476], [-0.1190, -0.1292, 0.0848, ..., -0.1086, -0.0615, 0.1580]], device='cuda:0'), grad: tensor([[ 7.1712e-08, 1.2051e-06, -3.1758e-07, ..., -7.2364e-07, 4.2841e-07, -1.3057e-06], [-7.1377e-06, -5.2974e-06, 1.3970e-08, ..., -5.1335e-06, 2.5146e-08, 9.3132e-09], [ 4.0140e-07, 5.6252e-07, 9.0338e-08, ..., 3.1944e-07, 1.4156e-07, 1.0245e-08], ..., [ 6.2585e-06, -6.7428e-07, 3.2037e-07, ..., 4.1947e-06, -1.7844e-06, 8.1025e-08], [ 1.5646e-07, 4.2468e-07, 1.2480e-07, ..., 3.1572e-07, 1.2852e-07, 1.6671e-07], [ 6.4727e-07, 2.6971e-06, 2.7940e-08, ..., 8.0466e-07, 7.4040e-07, 8.8476e-08]], device='cuda:0') Epoch 231, bias, value: tensor([-0.0075, -0.0078, 0.0124, 0.0110, 0.0340, 0.0424, -0.0207, 0.0081, -0.0358, -0.0251], device='cuda:0'), grad: tensor([-7.7300e-07, -3.8117e-05, 3.0696e-06, -2.9877e-06, 6.8638e-07, 1.4370e-06, 2.0042e-06, 2.4036e-05, 2.1011e-06, 8.5756e-06], device='cuda:0') 100 0.0001 changing lr epoch 230, time 216.99, cls_loss 0.0018 cls_loss_mapping 0.0031 cls_loss_causal 0.4933 re_mapping 0.0056 re_causal 0.0155 /// teacc 99.10 lr 0.00010000 Epoch 232, weight, value: tensor([[-0.0609, 0.1548, 0.0742, ..., -0.2297, -0.1228, 0.0207], [ 0.1665, -0.0212, -0.0869, ..., 0.0899, 0.1194, -0.1582], [-0.1017, -0.0613, -0.0486, ..., -0.0889, -0.0356, -0.0777], ..., [-0.0832, 0.1696, -0.1914, ..., -0.1657, -0.0053, -0.1818], [ 0.0113, -0.1096, 0.0368, ..., 0.0697, -0.1215, -0.0479], [-0.1195, -0.1292, 0.0851, ..., -0.1087, -0.0619, 0.1585]], device='cuda:0'), grad: tensor([[ 6.5193e-09, -6.4261e-08, -1.0617e-07, ..., -3.1665e-08, -2.7940e-08, -2.0396e-07], [-1.7509e-07, -3.0734e-08, -6.5193e-09, ..., -3.7160e-07, -4.5635e-08, 2.7940e-09], [ 1.6764e-08, 1.4901e-08, 1.3039e-08, ..., -6.2399e-08, 1.2107e-08, 1.9558e-08], ..., [ 8.1025e-08, -9.3132e-10, 2.7940e-08, ..., 1.7881e-07, 2.5146e-08, 1.8626e-08], [ 6.2399e-08, 3.7253e-08, 8.4750e-08, ..., 2.4308e-07, 2.1420e-08, 1.3784e-07], [ 9.3132e-09, 2.5146e-08, -1.0338e-07, ..., 2.8871e-08, 7.4506e-09, -1.7136e-07]], device='cuda:0') Epoch 232, bias, value: tensor([-0.0070, -0.0078, 0.0123, 0.0111, 0.0337, 0.0423, -0.0210, 0.0084, -0.0361, -0.0250], device='cuda:0'), grad: tensor([-5.2899e-07, -6.1933e-07, -6.8825e-07, -1.3225e-07, 2.4121e-07, 8.7544e-08, 4.0978e-08, 4.5169e-07, 1.2917e-06, -1.2480e-07], device='cuda:0') 100 0.0001 changing lr epoch 231, time 217.25, cls_loss 0.0016 cls_loss_mapping 0.0024 cls_loss_causal 0.5177 re_mapping 0.0054 re_causal 0.0159 /// teacc 99.02 lr 0.00010000 Epoch 233, weight, value: tensor([[-0.0613, 0.1559, 0.0746, ..., -0.2300, -0.1229, 0.0210], [ 0.1667, -0.0226, -0.0870, ..., 0.0900, 0.1196, -0.1583], [-0.1020, -0.0615, -0.0487, ..., -0.0891, -0.0358, -0.0781], ..., [-0.0833, 0.1700, -0.1916, ..., -0.1658, -0.0053, -0.1826], [ 0.0118, -0.1097, 0.0372, ..., 0.0704, -0.1221, -0.0479], [-0.1199, -0.1292, 0.0854, ..., -0.1092, -0.0621, 0.1587]], device='cuda:0'), grad: tensor([[ 5.4948e-08, -7.2550e-07, -2.1048e-07, ..., 1.4622e-07, 9.3132e-10, -3.2410e-07], [ 6.2287e-06, 3.7160e-07, 3.2783e-06, ..., 1.0237e-05, -2.0489e-08, 3.6098e-06], [ 1.1642e-07, 1.2107e-07, 1.7788e-07, ..., 3.3621e-07, 2.0582e-07, 7.1712e-08], ..., [-1.5832e-08, -3.5856e-07, 2.6077e-08, ..., 3.9674e-07, 2.0489e-08, 1.9558e-08], [-8.1658e-06, 1.5087e-07, -4.1835e-06, ..., -1.3486e-05, 1.8626e-08, -4.6119e-06], [ 1.7881e-07, 1.9744e-07, 1.3597e-07, ..., 1.3644e-06, 5.5879e-09, 1.6112e-07]], device='cuda:0') Epoch 233, bias, value: tensor([-0.0068, -0.0076, 0.0120, 0.0120, 0.0336, 0.0413, -0.0211, 0.0086, -0.0360, -0.0250], device='cuda:0'), grad: tensor([-1.7602e-06, 1.8716e-05, 2.5742e-06, -1.2228e-06, -4.3437e-06, 2.0899e-06, 1.9670e-06, 7.9814e-07, -2.3156e-05, 4.3213e-06], device='cuda:0') 100 0.0001 changing lr epoch 232, time 216.79, cls_loss 0.0014 cls_loss_mapping 0.0023 cls_loss_causal 0.4906 re_mapping 0.0054 re_causal 0.0156 /// teacc 98.94 lr 0.00010000 Epoch 234, weight, value: tensor([[-0.0614, 0.1566, 0.0749, ..., -0.2302, -0.1229, 0.0211], [ 0.1669, -0.0231, -0.0870, ..., 0.0901, 0.1197, -0.1584], [-0.1024, -0.0616, -0.0487, ..., -0.0892, -0.0358, -0.0783], ..., [-0.0834, 0.1704, -0.1918, ..., -0.1661, -0.0051, -0.1853], [ 0.0117, -0.1104, 0.0374, ..., 0.0706, -0.1226, -0.0480], [-0.1202, -0.1294, 0.0855, ..., -0.1096, -0.0622, 0.1595]], device='cuda:0'), grad: tensor([[ 1.3039e-08, 7.4506e-09, 0.0000e+00, ..., 3.9116e-08, 6.5193e-09, -4.6566e-09], [-2.7105e-05, 2.4214e-08, 9.3132e-10, ..., -1.4082e-05, -4.2945e-05, 7.4506e-09], [ 2.4542e-05, 3.4459e-08, 9.3132e-10, ..., 1.4089e-05, 3.8922e-05, 9.3132e-10], ..., [ 2.2389e-06, -2.2892e-06, 1.7695e-08, ..., 1.8952e-06, 3.5372e-06, 1.7323e-07], [ 2.7940e-08, 2.8871e-08, 7.4506e-09, ..., 3.9823e-06, 1.9558e-08, 2.8871e-08], [ 1.6298e-07, 2.1867e-06, -2.3283e-08, ..., 1.0319e-05, 2.3749e-07, -2.4959e-07]], device='cuda:0') Epoch 234, bias, value: tensor([-0.0067, -0.0077, 0.0119, 0.0134, 0.0334, 0.0401, -0.0212, 0.0087, -0.0362, -0.0249], device='cuda:0'), grad: tensor([ 1.0245e-07, -9.7692e-05, 9.0778e-05, 1.5181e-07, -3.3855e-05, -3.1665e-08, 8.9332e-06, 6.0797e-06, 6.2250e-06, 1.9163e-05], device='cuda:0') 100 0.0001 changing lr epoch 233, time 216.78, cls_loss 0.0016 cls_loss_mapping 0.0036 cls_loss_causal 0.5011 re_mapping 0.0060 re_causal 0.0166 /// teacc 99.01 lr 0.00010000 Epoch 235, weight, value: tensor([[-0.0624, 0.1587, 0.0760, ..., -0.2308, -0.1231, 0.0225], [ 0.1686, -0.0221, -0.0871, ..., 0.0907, 0.1199, -0.1581], [-0.1028, -0.0606, -0.0486, ..., -0.0892, -0.0360, -0.0786], ..., [-0.0837, 0.1700, -0.1918, ..., -0.1683, -0.0064, -0.1862], [ 0.0116, -0.1107, 0.0373, ..., 0.0703, -0.1228, -0.0481], [-0.1208, -0.1294, 0.0854, ..., -0.1105, -0.0623, 0.1595]], device='cuda:0'), grad: tensor([[ 3.0734e-08, 1.9558e-08, 4.6566e-09, ..., 4.4703e-08, 9.3132e-10, 3.7253e-09], [-3.2131e-07, -2.0303e-07, 1.0245e-08, ..., -5.6531e-07, -7.8231e-08, 5.5879e-09], [ 8.5682e-08, 6.5193e-08, 1.6764e-08, ..., 4.2748e-07, 1.8626e-09, 2.7940e-09], ..., [ 1.8626e-07, 7.3574e-08, 5.4948e-08, ..., 3.7346e-07, 4.6566e-08, 9.4064e-08], [ 1.3448e-06, 6.2399e-08, 2.6785e-06, ..., 5.7928e-07, 1.4901e-08, 1.4417e-06], [ 1.0058e-07, 3.4459e-08, 5.3085e-08, ..., 1.2852e-07, 9.3132e-09, -3.3062e-07]], device='cuda:0') Epoch 235, bias, value: tensor([-0.0058, -0.0070, 0.0124, 0.0109, 0.0335, 0.0418, -0.0229, 0.0079, -0.0364, -0.0249], device='cuda:0'), grad: tensor([ 4.4424e-07, -1.0617e-06, -3.5483e-07, -8.2478e-06, 1.5832e-08, 1.9111e-06, -4.3586e-07, 1.0589e-06, 6.4895e-06, 1.7695e-07], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 234---------------------------------------------------- epoch 234, time 217.92, cls_loss 0.0014 cls_loss_mapping 0.0021 cls_loss_causal 0.5009 re_mapping 0.0057 re_causal 0.0163 /// teacc 99.12 lr 0.00010000 Epoch 236, weight, value: tensor([[-0.0626, 0.1589, 0.0775, ..., -0.2311, -0.1231, 0.0236], [ 0.1688, -0.0226, -0.0872, ..., 0.0908, 0.1199, -0.1585], [-0.1030, -0.0606, -0.0487, ..., -0.0893, -0.0360, -0.0789], ..., [-0.0838, 0.1702, -0.1919, ..., -0.1689, -0.0058, -0.1874], [ 0.0128, -0.1109, 0.0372, ..., 0.0724, -0.1220, -0.0490], [-0.1213, -0.1294, 0.0852, ..., -0.1107, -0.0624, 0.1595]], device='cuda:0'), grad: tensor([[ 6.3330e-08, 2.3283e-08, 1.3607e-06, ..., 6.4261e-08, 9.3132e-10, 1.9558e-08], [ 1.2107e-08, 6.5193e-09, 3.3528e-08, ..., 1.4901e-08, 1.8626e-09, 3.7253e-09], [ 1.3411e-07, 2.1420e-08, -1.8729e-06, ..., 1.4715e-07, 8.3819e-09, 2.7008e-08], ..., [ 2.2352e-08, -1.2107e-08, 8.2888e-08, ..., 2.0489e-08, 3.7253e-09, 2.3283e-08], [-2.8778e-07, -3.6322e-08, 2.0675e-07, ..., -3.3993e-07, 6.8918e-08, -3.5390e-08], [ 1.4901e-08, 1.4901e-08, 2.7008e-08, ..., 2.4214e-08, 1.8626e-09, -4.8429e-08]], device='cuda:0') Epoch 236, bias, value: tensor([-0.0050, -0.0070, 0.0124, 0.0106, 0.0333, 0.0418, -0.0229, 0.0079, -0.0358, -0.0250], device='cuda:0'), grad: tensor([ 1.4380e-05, 1.9372e-07, -2.1130e-05, -9.9242e-06, 4.9360e-08, 1.2182e-05, -7.5437e-07, 3.9581e-07, 4.2729e-06, 2.8964e-07], device='cuda:0') 100 0.0001 changing lr epoch 235, time 216.99, cls_loss 0.0014 cls_loss_mapping 0.0025 cls_loss_causal 0.4868 re_mapping 0.0056 re_causal 0.0156 /// teacc 99.05 lr 0.00010000 Epoch 237, weight, value: tensor([[-0.0630, 0.1592, 0.0779, ..., -0.2314, -0.1232, 0.0238], [ 0.1694, -0.0224, -0.0872, ..., 0.0910, 0.1201, -0.1586], [-0.1033, -0.0607, -0.0487, ..., -0.0894, -0.0361, -0.0790], ..., [-0.0840, 0.1703, -0.1919, ..., -0.1693, -0.0063, -0.1880], [ 0.0124, -0.1113, 0.0370, ..., 0.0718, -0.1231, -0.0493], [-0.1215, -0.1294, 0.0852, ..., -0.1111, -0.0626, 0.1597]], device='cuda:0'), grad: tensor([[ 1.1176e-08, 3.0603e-06, 1.8626e-09, ..., 9.3132e-09, 0.0000e+00, -3.2783e-07], [ 7.4506e-09, 1.4558e-05, 1.8626e-09, ..., 1.4901e-08, 0.0000e+00, 1.0245e-08], [ 2.7008e-08, 4.9546e-07, 1.7695e-08, ..., 7.0781e-08, -1.8626e-09, 2.4214e-08], ..., [ 3.7253e-09, -2.3916e-05, 1.3039e-08, ..., 1.7695e-08, 9.3132e-10, 7.9162e-08], [-1.3039e-08, 4.6939e-07, -1.9558e-08, ..., -1.2852e-07, 0.0000e+00, 6.4261e-08], [ 6.5193e-09, 3.0212e-06, -8.9407e-08, ..., 2.0396e-07, 0.0000e+00, -1.8906e-07]], device='cuda:0') Epoch 237, bias, value: tensor([-0.0049, -0.0068, 0.0123, 0.0078, 0.0332, 0.0446, -0.0222, 0.0078, -0.0363, -0.0251], device='cuda:0'), grad: tensor([ 5.5060e-06, 2.6748e-05, 1.0496e-06, 4.7591e-07, 2.2296e-06, 1.3039e-06, 1.8254e-07, -4.3690e-05, 6.1374e-07, 5.6289e-06], device='cuda:0') 100 0.0001 changing lr epoch 236, time 216.96, cls_loss 0.0018 cls_loss_mapping 0.0022 cls_loss_causal 0.5370 re_mapping 0.0053 re_causal 0.0160 /// teacc 99.10 lr 0.00010000 Epoch 238, weight, value: tensor([[-0.0636, 0.1589, 0.0777, ..., -0.2321, -0.1232, 0.0237], [ 0.1702, -0.0216, -0.0871, ..., 0.0911, 0.1202, -0.1585], [-0.1039, -0.0619, -0.0487, ..., -0.0894, -0.0361, -0.0789], ..., [-0.0842, 0.1704, -0.1922, ..., -0.1698, -0.0063, -0.1889], [ 0.0116, -0.1124, 0.0360, ..., 0.0711, -0.1235, -0.0499], [-0.1222, -0.1294, 0.0854, ..., -0.1140, -0.0629, 0.1597]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 8.5682e-08, -7.4506e-09, ..., 7.4506e-09, 0.0000e+00, -1.8999e-07], [ 7.4506e-09, 4.2561e-07, 1.8626e-09, ..., 8.3819e-09, 0.0000e+00, 2.0489e-08], [ 2.3283e-08, 6.1691e-06, 5.5879e-09, ..., 5.7742e-08, 0.0000e+00, 2.7940e-08], ..., [ 6.2399e-08, -7.3537e-06, 1.1362e-07, ..., 1.5367e-07, 0.0000e+00, 8.8196e-07], [-1.3225e-07, 9.4064e-08, 3.7253e-09, ..., -3.1851e-07, 0.0000e+00, 2.5146e-08], [ 8.3819e-09, -2.3283e-06, -1.0859e-06, ..., 4.8429e-08, 0.0000e+00, -8.5086e-06]], device='cuda:0') Epoch 238, bias, value: tensor([-0.0051, -0.0066, 0.0121, 0.0079, 0.0356, 0.0448, -0.0215, 0.0077, -0.0374, -0.0261], device='cuda:0'), grad: tensor([-1.9744e-07, 7.7300e-07, 9.5963e-06, 1.8440e-07, 1.7568e-05, 4.2841e-07, 2.9895e-07, -9.3803e-06, -4.4703e-07, -1.8805e-05], device='cuda:0') 100 0.0001 changing lr epoch 237, time 216.70, cls_loss 0.0016 cls_loss_mapping 0.0026 cls_loss_causal 0.5184 re_mapping 0.0056 re_causal 0.0160 /// teacc 99.12 lr 0.00010000 Epoch 239, weight, value: tensor([[-0.0643, 0.1589, 0.0776, ..., -0.2333, -0.1232, 0.0236], [ 0.1708, -0.0209, -0.0872, ..., 0.0913, 0.1201, -0.1587], [-0.1042, -0.0627, -0.0487, ..., -0.0895, -0.0361, -0.0796], ..., [-0.0848, 0.1705, -0.1923, ..., -0.1715, -0.0059, -0.1902], [ 0.0117, -0.1128, 0.0356, ..., 0.0712, -0.1237, -0.0509], [-0.1227, -0.1294, 0.0857, ..., -0.1140, -0.0630, 0.1607]], device='cuda:0'), grad: tensor([[ 9.3132e-09, -2.5146e-08, 9.3132e-10, ..., 1.5832e-08, 0.0000e+00, -1.0245e-08], [-5.2154e-08, 2.1420e-08, 7.4506e-09, ..., -1.7323e-07, 0.0000e+00, 1.3970e-08], [ 4.2841e-08, 7.4506e-09, 4.6566e-09, ..., 1.3877e-07, -1.8626e-09, 5.5879e-09], ..., [ 7.4506e-09, -5.4017e-08, 2.7940e-09, ..., 2.0489e-08, 0.0000e+00, 1.1176e-08], [ 2.4214e-08, 1.3039e-08, 2.1420e-08, ..., 3.6322e-08, 0.0000e+00, 6.1467e-08], [ 8.3819e-09, 2.2352e-08, 3.7253e-09, ..., 2.5146e-08, 0.0000e+00, -5.9605e-08]], device='cuda:0') Epoch 239, bias, value: tensor([-0.0054, -0.0065, 0.0120, 0.0080, 0.0354, 0.0450, -0.0224, 0.0077, -0.0377, -0.0259], device='cuda:0'), grad: tensor([-2.6077e-08, -2.4214e-07, 1.9744e-07, 6.7521e-07, 9.4064e-08, -7.4599e-07, -3.5390e-08, -2.6077e-08, 1.4715e-07, -3.9116e-08], device='cuda:0') 100 0.0001 changing lr epoch 238, time 216.69, cls_loss 0.0015 cls_loss_mapping 0.0023 cls_loss_causal 0.5046 re_mapping 0.0055 re_causal 0.0156 /// teacc 99.02 lr 0.00010000 Epoch 240, weight, value: tensor([[-0.0647, 0.1596, 0.0778, ..., -0.2336, -0.1233, 0.0238], [ 0.1718, -0.0219, -0.0870, ..., 0.0916, 0.1203, -0.1580], [-0.1047, -0.0601, -0.0489, ..., -0.0896, -0.0362, -0.0808], ..., [-0.0847, 0.1703, -0.1923, ..., -0.1716, -0.0060, -0.1912], [ 0.0124, -0.1130, 0.0353, ..., 0.0720, -0.1231, -0.0512], [-0.1233, -0.1297, 0.0859, ..., -0.1143, -0.0630, 0.1613]], device='cuda:0'), grad: tensor([[ 2.5239e-07, 1.4901e-08, 3.0827e-07, ..., 4.0513e-07, 0.0000e+00, 3.7253e-09], [ 8.9034e-07, 1.2107e-07, 1.3039e-07, ..., 5.4482e-07, -9.3132e-10, 1.8626e-09], [ 2.2799e-06, 5.9977e-07, 2.5295e-06, ..., 4.1015e-06, 7.4506e-09, 2.7940e-09], ..., [ 4.0978e-08, -8.7358e-07, 2.2072e-07, ..., 1.0896e-07, 0.0000e+00, 6.4261e-08], [-3.2093e-06, -1.3039e-08, -3.7812e-06, ..., -1.0379e-05, -9.3132e-09, 1.5367e-07], [ 1.0896e-07, 3.6322e-08, 7.2177e-07, ..., 8.2050e-07, 0.0000e+00, -2.4214e-08]], device='cuda:0') Epoch 240, bias, value: tensor([-0.0053, -0.0064, 0.0126, 0.0087, 0.0354, 0.0441, -0.0228, 0.0074, -0.0375, -0.0260], device='cuda:0'), grad: tensor([ 1.9073e-06, 2.8536e-06, 2.1964e-05, -9.5546e-05, 8.3260e-07, 1.0413e-04, -5.9698e-07, -2.2538e-07, -3.9160e-05, 3.8128e-06], device='cuda:0') 100 0.0001 changing lr epoch 239, time 217.40, cls_loss 0.0011 cls_loss_mapping 0.0017 cls_loss_causal 0.4959 re_mapping 0.0057 re_causal 0.0162 /// teacc 99.02 lr 0.00010000 Epoch 241, weight, value: tensor([[-0.0651, 0.1597, 0.0778, ..., -0.2341, -0.1233, 0.0238], [ 0.1719, -0.0220, -0.0871, ..., 0.0917, 0.1203, -0.1582], [-0.1049, -0.0602, -0.0490, ..., -0.0896, -0.0363, -0.0807], ..., [-0.0847, 0.1704, -0.1925, ..., -0.1717, -0.0059, -0.1920], [ 0.0129, -0.1134, 0.0349, ..., 0.0724, -0.1234, -0.0514], [-0.1235, -0.1297, 0.0862, ..., -0.1143, -0.0631, 0.1618]], device='cuda:0'), grad: tensor([[ 2.7940e-09, -2.5425e-07, -1.3784e-07, ..., 6.5193e-09, 1.3970e-08, -3.1479e-07], [ 1.0245e-08, 1.6764e-08, 1.9558e-07, ..., 2.8592e-07, 4.5635e-08, 1.6298e-07], [ 9.3132e-09, -4.0047e-08, 8.6892e-07, ..., 2.7008e-08, 4.5411e-06, 1.7695e-08], ..., [ 4.6566e-09, 4.2841e-08, 2.6915e-07, ..., 1.7695e-08, 1.3364e-06, 2.2352e-08], [-2.5332e-07, 4.8429e-08, -1.5125e-06, ..., -2.8834e-06, 5.4017e-08, -1.2238e-06], [ 1.7975e-07, 8.1956e-08, 1.0198e-06, ..., 1.9260e-06, 2.6077e-08, 7.3481e-07]], device='cuda:0') Epoch 241, bias, value: tensor([-0.0053, -0.0064, 0.0126, 0.0087, 0.0353, 0.0441, -0.0230, 0.0073, -0.0376, -0.0258], device='cuda:0'), grad: tensor([-7.7579e-07, 7.7393e-07, 3.1441e-05, -4.3392e-05, 8.2795e-07, 1.6317e-06, 3.4273e-07, 9.7007e-06, -3.5129e-06, 2.9802e-06], device='cuda:0') 100 0.0001 changing lr epoch 240, time 217.24, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4897 re_mapping 0.0053 re_causal 0.0151 /// teacc 98.99 lr 0.00010000 Epoch 242, weight, value: tensor([[-0.0657, 0.1594, 0.0776, ..., -0.2349, -0.1233, 0.0236], [ 0.1723, -0.0221, -0.0871, ..., 0.0918, 0.1203, -0.1582], [-0.1053, -0.0605, -0.0491, ..., -0.0897, -0.0363, -0.0813], ..., [-0.0849, 0.1707, -0.1929, ..., -0.1723, -0.0059, -0.1926], [ 0.0131, -0.1134, 0.0345, ..., 0.0726, -0.1241, -0.0516], [-0.1239, -0.1299, 0.0868, ..., -0.1144, -0.0631, 0.1622]], device='cuda:0'), grad: tensor([[ 2.5518e-07, 7.3574e-08, 3.6508e-07, ..., 4.4424e-07, 0.0000e+00, 7.5903e-07], [ 3.8296e-06, 1.1930e-06, 1.0803e-07, ..., 4.9248e-06, -2.7940e-09, 1.2042e-06], [ 7.4506e-08, 2.7940e-08, 1.2852e-07, ..., 1.6764e-07, 9.3132e-10, 2.3842e-07], ..., [ 4.1910e-08, -4.3772e-08, 3.5856e-07, ..., 2.1048e-07, 9.3132e-10, 6.3609e-07], [ 4.0308e-06, 3.6880e-07, 6.0678e-05, ..., 2.1815e-05, 0.0000e+00, 1.1122e-04], [ 7.4506e-07, 1.4901e-08, -5.8800e-05, ..., -2.2098e-05, 9.3132e-10, -1.2612e-04]], device='cuda:0') Epoch 242, bias, value: tensor([-0.0056, -0.0064, 0.0125, 0.0088, 0.0352, 0.0440, -0.0227, 0.0073, -0.0378, -0.0256], device='cuda:0'), grad: tensor([ 1.7304e-06, 7.8306e-06, 3.2410e-07, -5.0783e-05, 1.5432e-06, 6.8903e-05, -3.3736e-05, 1.2880e-06, 1.9443e-04, -1.9145e-04], device='cuda:0') 100 0.0001 changing lr epoch 241, time 217.09, cls_loss 0.0013 cls_loss_mapping 0.0025 cls_loss_causal 0.5291 re_mapping 0.0056 re_causal 0.0165 /// teacc 99.10 lr 0.00010000 Epoch 243, weight, value: tensor([[-0.0661, 0.1593, 0.0775, ..., -0.2354, -0.1233, 0.0235], [ 0.1727, -0.0222, -0.0872, ..., 0.0918, 0.1204, -0.1583], [-0.1056, -0.0606, -0.0490, ..., -0.0897, -0.0363, -0.0805], ..., [-0.0850, 0.1708, -0.1935, ..., -0.1727, -0.0059, -0.1934], [ 0.0138, -0.1132, 0.0336, ..., 0.0735, -0.1245, -0.0519], [-0.1245, -0.1299, 0.0882, ..., -0.1141, -0.0632, 0.1641]], device='cuda:0'), grad: tensor([[ 5.6811e-08, -3.8045e-07, -1.8161e-08, ..., 4.3772e-08, 0.0000e+00, -2.3982e-07], [-2.0396e-07, 4.0513e-08, -1.8161e-08, ..., -2.4680e-07, 0.0000e+00, -1.7229e-08], [ 6.3796e-08, -1.9744e-07, 2.3283e-08, ..., 7.1246e-08, -4.6566e-10, 2.6077e-08], ..., [ 2.4680e-08, 2.1188e-07, 3.1199e-08, ..., 7.8697e-08, 0.0000e+00, 4.3306e-08], [ 1.4482e-07, 2.3516e-07, 3.3528e-08, ..., 1.8580e-07, 0.0000e+00, 1.6252e-07], [ 2.8871e-08, 1.2433e-07, -7.5763e-07, ..., 1.2154e-07, 0.0000e+00, -1.2517e-06]], device='cuda:0') Epoch 243, bias, value: tensor([-0.0058, -0.0064, 0.0125, 0.0088, 0.0340, 0.0439, -0.0229, 0.0072, -0.0378, -0.0250], device='cuda:0'), grad: tensor([-7.0827e-07, -3.8417e-07, -4.4964e-06, -2.2491e-07, 2.7008e-06, 3.0966e-07, -3.9069e-07, 4.3176e-06, 1.4119e-06, -2.5369e-06], device='cuda:0') 100 0.0001 changing lr epoch 242, time 217.17, cls_loss 0.0011 cls_loss_mapping 0.0019 cls_loss_causal 0.5190 re_mapping 0.0057 re_causal 0.0163 /// teacc 98.96 lr 0.00010000 Epoch 244, weight, value: tensor([[-0.0669, 0.1590, 0.0774, ..., -0.2364, -0.1234, 0.0234], [ 0.1736, -0.0220, -0.0872, ..., 0.0918, 0.1209, -0.1584], [-0.1061, -0.0610, -0.0491, ..., -0.0898, -0.0365, -0.0807], ..., [-0.0858, 0.1710, -0.1939, ..., -0.1729, -0.0081, -0.1947], [ 0.0139, -0.1134, 0.0335, ..., 0.0737, -0.1246, -0.0518], [-0.1248, -0.1299, 0.0887, ..., -0.1139, -0.0633, 0.1650]], device='cuda:0'), grad: tensor([[ 1.3784e-06, 2.1234e-06, -1.3597e-07, ..., 2.6841e-06, 4.6566e-10, 1.5190e-06], [-1.9558e-08, 2.3283e-08, 7.9162e-09, ..., -1.3504e-08, -4.1910e-09, 2.4214e-08], [ 5.7276e-08, 7.5903e-08, 8.3819e-09, ..., 9.9652e-08, -4.1910e-09, 6.0536e-08], ..., [ 1.5832e-08, 9.3132e-10, 1.2247e-07, ..., 1.7229e-08, 5.1223e-09, 1.6298e-07], [ 1.2340e-07, 1.9837e-07, 1.4715e-07, ..., 1.9092e-07, 1.8626e-09, 3.5390e-07], [ 5.4482e-08, 7.2643e-08, -9.7323e-08, ..., 1.5227e-07, 1.8626e-09, -7.8697e-08]], device='cuda:0') Epoch 244, bias, value: tensor([-0.0061, -0.0059, 0.0124, 0.0090, 0.0333, 0.0439, -0.0227, 0.0066, -0.0379, -0.0246], device='cuda:0'), grad: tensor([ 8.8811e-06, 3.8184e-08, 2.3702e-07, 1.2107e-08, 8.3679e-07, 8.3353e-08, -1.1668e-05, 4.9546e-07, 9.8906e-07, 8.8010e-08], device='cuda:0') 100 0.0001 changing lr epoch 243, time 216.89, cls_loss 0.0012 cls_loss_mapping 0.0028 cls_loss_causal 0.5106 re_mapping 0.0057 re_causal 0.0160 /// teacc 98.96 lr 0.00010000 Epoch 245, weight, value: tensor([[-0.0674, 0.1593, 0.0774, ..., -0.2367, -0.1235, 0.0235], [ 0.1739, -0.0224, -0.0873, ..., 0.0919, 0.1210, -0.1588], [-0.1066, -0.0611, -0.0492, ..., -0.0898, -0.0365, -0.0808], ..., [-0.0858, 0.1712, -0.1940, ..., -0.1732, -0.0082, -0.1949], [ 0.0138, -0.1136, 0.0330, ..., 0.0737, -0.1251, -0.0522], [-0.1250, -0.1300, 0.0890, ..., -0.1145, -0.0635, 0.1654]], device='cuda:0'), grad: tensor([[ 6.5193e-09, -6.1467e-08, -1.8626e-09, ..., 5.5879e-09, 0.0000e+00, -1.3039e-08], [ 5.5507e-07, 6.5193e-08, 1.0245e-08, ..., 4.7311e-07, -9.3132e-10, 4.9360e-08], [ 2.5146e-08, 1.6764e-08, 6.5193e-09, ..., 2.3283e-08, -9.3132e-10, 6.5193e-09], ..., [ 4.2841e-08, -1.5087e-07, 3.9116e-08, ..., 3.8184e-08, 0.0000e+00, 6.0536e-08], [-8.2608e-07, 2.3283e-08, 9.5926e-08, ..., -7.0874e-07, 0.0000e+00, 1.2573e-07], [ 4.8429e-08, 7.5437e-08, -2.3656e-07, ..., 4.3772e-08, 9.3132e-10, -4.9546e-07]], device='cuda:0') Epoch 245, bias, value: tensor([-0.0060, -0.0059, 0.0123, 0.0089, 0.0337, 0.0440, -0.0227, 0.0067, -0.0381, -0.0248], device='cuda:0'), grad: tensor([-2.5798e-07, 1.2284e-06, 7.7300e-08, 9.3784e-07, 5.8487e-07, -1.0412e-06, 2.8871e-07, 6.9849e-08, -1.1967e-06, -6.9197e-07], device='cuda:0') 100 0.0001 changing lr epoch 244, time 217.16, cls_loss 0.0013 cls_loss_mapping 0.0023 cls_loss_causal 0.4824 re_mapping 0.0056 re_causal 0.0158 /// teacc 99.09 lr 0.00010000 Epoch 246, weight, value: tensor([[-0.0677, 0.1596, 0.0773, ..., -0.2370, -0.1236, 0.0234], [ 0.1748, -0.0225, -0.0872, ..., 0.0921, 0.1211, -0.1585], [-0.1075, -0.0613, -0.0493, ..., -0.0899, -0.0367, -0.0810], ..., [-0.0860, 0.1713, -0.1942, ..., -0.1737, -0.0082, -0.1957], [ 0.0133, -0.1140, 0.0323, ..., 0.0730, -0.1254, -0.0527], [-0.1256, -0.1300, 0.0893, ..., -0.1173, -0.0637, 0.1649]], device='cuda:0'), grad: tensor([[ 4.0047e-08, 1.3690e-07, 6.5193e-09, ..., 4.3772e-08, 8.0094e-08, -9.3132e-10], [-8.2888e-08, 1.9092e-07, 6.5193e-09, ..., 4.6566e-08, 6.6124e-08, 7.4506e-09], [ 6.7987e-08, 6.4261e-08, 8.3819e-09, ..., 8.1956e-08, 3.4459e-08, 3.7253e-09], ..., [ 1.4901e-08, -1.1828e-06, 1.2107e-08, ..., 1.0058e-07, -4.2748e-07, 1.3039e-08], [ 8.6799e-07, 1.4156e-07, 2.9802e-07, ..., 6.9570e-07, 4.9919e-07, 3.9861e-07], [ 6.5193e-09, 4.8801e-07, -2.2352e-08, ..., 1.8934e-06, 1.8161e-07, -1.0896e-07]], device='cuda:0') Epoch 246, bias, value: tensor([-0.0061, -0.0057, 0.0122, 0.0088, 0.0363, 0.0442, -0.0223, 0.0066, -0.0386, -0.0265], device='cuda:0'), grad: tensor([ 3.1572e-07, 5.0012e-07, -8.0094e-07, 3.5297e-07, -3.9749e-06, -1.5236e-06, -1.7695e-08, -1.7723e-06, 2.6673e-06, 4.2692e-06], device='cuda:0') 100 0.0001 changing lr epoch 245, time 216.47, cls_loss 0.0013 cls_loss_mapping 0.0029 cls_loss_causal 0.4911 re_mapping 0.0055 re_causal 0.0152 /// teacc 98.95 lr 0.00010000 Epoch 247, weight, value: tensor([[-0.0678, 0.1599, 0.0774, ..., -0.2373, -0.1236, 0.0235], [ 0.1773, -0.0232, -0.0872, ..., 0.0930, 0.1240, -0.1587], [-0.1113, -0.0621, -0.0493, ..., -0.0909, -0.0393, -0.0812], ..., [-0.0863, 0.1719, -0.1942, ..., -0.1741, -0.0099, -0.1960], [ 0.0132, -0.1142, 0.0322, ..., 0.0731, -0.1253, -0.0530], [-0.1256, -0.1300, 0.0902, ..., -0.1164, -0.0640, 0.1675]], device='cuda:0'), grad: tensor([[ 1.0524e-07, -1.1176e-08, -6.5193e-09, ..., 1.0151e-07, 1.7695e-08, -8.3819e-09], [-1.3616e-06, 5.5879e-09, 9.5926e-08, ..., -1.1949e-06, -3.0361e-07, 9.0338e-08], [ 1.2079e-06, 2.7940e-09, 2.7940e-09, ..., 1.4268e-06, 1.0896e-07, 1.0245e-08], ..., [ 1.8161e-07, -1.7695e-08, 3.2596e-08, ..., 2.0396e-07, 2.7008e-08, 7.7300e-08], [-2.5295e-06, 2.7940e-09, -7.1339e-07, ..., -3.0678e-06, 7.1712e-08, -6.8359e-07], [ 2.3376e-07, 9.3132e-09, -4.8243e-07, ..., 3.8929e-07, 4.0047e-08, -1.1781e-06]], device='cuda:0') Epoch 247, bias, value: tensor([-0.0061, -0.0049, 0.0111, 0.0087, 0.0351, 0.0440, -0.0220, 0.0069, -0.0388, -0.0255], device='cuda:0'), grad: tensor([ 2.3469e-07, -3.0808e-06, 4.4666e-06, 3.9209e-07, 1.5646e-06, 3.8147e-06, 1.8030e-06, 6.8638e-07, -8.4490e-06, -1.4678e-06], device='cuda:0') 100 0.0001 changing lr epoch 246, time 216.55, cls_loss 0.0013 cls_loss_mapping 0.0027 cls_loss_causal 0.5105 re_mapping 0.0057 re_causal 0.0164 /// teacc 99.03 lr 0.00010000 Epoch 248, weight, value: tensor([[-0.0687, 0.1592, 0.0765, ..., -0.2384, -0.1237, 0.0228], [ 0.1798, -0.0208, -0.0873, ..., 0.0934, 0.1248, -0.1589], [-0.1136, -0.0644, -0.0494, ..., -0.0911, -0.0398, -0.0814], ..., [-0.0874, 0.1719, -0.1943, ..., -0.1753, -0.0114, -0.1963], [ 0.0136, -0.1143, 0.0325, ..., 0.0736, -0.1259, -0.0528], [-0.1263, -0.1301, 0.0907, ..., -0.1165, -0.0645, 0.1683]], device='cuda:0'), grad: tensor([[ 2.4214e-08, 1.0245e-08, -5.5879e-09, ..., 3.5390e-08, 1.0245e-08, -7.4506e-09], [-2.0489e-08, 3.7253e-09, 2.7940e-09, ..., -5.9605e-08, 0.0000e+00, 2.7940e-09], [ 1.4901e-08, 4.6566e-09, 3.2596e-08, ..., 3.5390e-08, 0.0000e+00, 3.7253e-09], ..., [ 5.5879e-09, -7.4506e-09, 4.6566e-09, ..., 1.3039e-08, 0.0000e+00, 1.8626e-09], [ 2.2352e-08, 3.7253e-09, 2.4214e-08, ..., -4.6566e-09, 0.0000e+00, 3.3528e-08], [ 2.2352e-08, 5.5879e-09, 1.8626e-09, ..., 5.8673e-08, 0.0000e+00, 2.7940e-09]], device='cuda:0') Epoch 248, bias, value: tensor([-0.0068, -0.0038, 0.0103, 0.0093, 0.0348, 0.0434, -0.0213, 0.0066, -0.0386, -0.0254], device='cuda:0'), grad: tensor([ 4.1910e-08, -1.2573e-07, 2.0489e-07, 8.9128e-07, 7.4506e-09, -1.2610e-06, 9.3132e-10, 5.8673e-08, 6.3330e-08, 1.1083e-07], device='cuda:0') 100 0.0001 changing lr epoch 247, time 216.59, cls_loss 0.0008 cls_loss_mapping 0.0020 cls_loss_causal 0.5057 re_mapping 0.0056 re_causal 0.0167 /// teacc 99.09 lr 0.00010000 Epoch 249, weight, value: tensor([[-0.0694, 0.1594, 0.0763, ..., -0.2392, -0.1237, 0.0228], [ 0.1792, -0.0226, -0.0872, ..., 0.0935, 0.1249, -0.1589], [-0.1139, -0.0645, -0.0494, ..., -0.0912, -0.0398, -0.0819], ..., [-0.0858, 0.1729, -0.1944, ..., -0.1755, -0.0114, -0.1966], [ 0.0135, -0.1145, 0.0324, ..., 0.0736, -0.1261, -0.0531], [-0.1267, -0.1301, 0.0909, ..., -0.1165, -0.0647, 0.1686]], device='cuda:0'), grad: tensor([[ 3.2037e-07, -1.8626e-09, 2.9802e-07, ..., 3.8929e-07, 9.3132e-10, 2.0303e-07], [-1.1176e-08, 1.8440e-07, 1.6764e-08, ..., 5.5879e-09, -1.6764e-08, 1.2107e-08], [ 4.7497e-08, 3.1386e-07, 2.8871e-08, ..., 6.1467e-08, 3.7253e-09, 2.1420e-08], ..., [ 3.4459e-08, -8.3540e-07, 2.3283e-08, ..., 1.7136e-07, -5.5879e-09, 2.0489e-08], [-1.5711e-06, 2.5518e-07, -1.3327e-06, ..., -1.9018e-06, 5.5879e-09, -9.0990e-07], [ 4.4890e-07, 9.9652e-08, 2.1327e-07, ..., 6.2957e-07, 9.3132e-09, 9.0338e-08]], device='cuda:0') Epoch 249, bias, value: tensor([-0.0069, -0.0045, 0.0102, 0.0093, 0.0347, 0.0433, -0.0204, 0.0075, -0.0388, -0.0254], device='cuda:0'), grad: tensor([ 1.7770e-06, 3.2131e-07, 7.0687e-07, 2.3656e-06, -2.4028e-07, 1.5516e-06, 1.4901e-07, -1.0431e-06, -7.9423e-06, 2.3562e-06], device='cuda:0') 100 0.0001 changing lr epoch 248, time 217.40, cls_loss 0.0017 cls_loss_mapping 0.0021 cls_loss_causal 0.5201 re_mapping 0.0053 re_causal 0.0151 /// teacc 99.06 lr 0.00010000 Epoch 250, weight, value: tensor([[-0.0697, 0.1601, 0.0763, ..., -0.2399, -0.1237, 0.0228], [ 0.1794, -0.0229, -0.0874, ..., 0.0935, 0.1249, -0.1595], [-0.1140, -0.0646, -0.0496, ..., -0.0912, -0.0399, -0.0822], ..., [-0.0860, 0.1729, -0.1944, ..., -0.1767, -0.0114, -0.1977], [ 0.0135, -0.1148, 0.0322, ..., 0.0739, -0.1264, -0.0537], [-0.1261, -0.1301, 0.0920, ..., -0.1172, -0.0648, 0.1703]], device='cuda:0'), grad: tensor([[ 1.8626e-08, 5.5879e-09, 1.3039e-08, ..., 1.7695e-08, 2.7940e-09, 2.7940e-09], [-3.3587e-05, -3.0145e-05, 1.3039e-08, ..., -4.3005e-05, -3.0734e-08, 1.8626e-09], [ 3.7253e-08, 3.3528e-08, 7.5437e-08, ..., 4.2841e-08, 1.1176e-08, 1.8626e-09], ..., [ 2.8029e-05, 2.5049e-05, 7.2643e-08, ..., 3.5971e-05, 8.3819e-09, 2.9802e-08], [ 2.4680e-07, 2.4401e-07, 1.2107e-08, ..., 3.0547e-07, 9.3132e-09, -1.8626e-09], [ 3.1739e-06, 2.9132e-06, -1.6671e-07, ..., 4.0941e-06, 4.6566e-09, -3.5577e-07]], device='cuda:0') Epoch 250, bias, value: tensor([-0.0068, -0.0046, 0.0102, 0.0093, 0.0349, 0.0433, -0.0211, 0.0073, -0.0388, -0.0251], device='cuda:0'), grad: tensor([ 3.1386e-07, -1.0931e-04, -1.4342e-06, -2.2501e-06, 7.2122e-06, 8.1118e-07, 9.8068e-07, 9.2387e-05, 1.2508e-06, 1.0066e-05], device='cuda:0') 100 0.0001 changing lr epoch 249, time 217.34, cls_loss 0.0014 cls_loss_mapping 0.0022 cls_loss_causal 0.5086 re_mapping 0.0052 re_causal 0.0152 /// teacc 99.01 lr 0.00010000 Epoch 251, weight, value: tensor([[-0.0706, 0.1596, 0.0761, ..., -0.2408, -0.1242, 0.0227], [ 0.1802, -0.0233, -0.0875, ..., 0.0936, 0.1255, -0.1597], [-0.1142, -0.0647, -0.0497, ..., -0.0913, -0.0400, -0.0826], ..., [-0.0868, 0.1736, -0.1945, ..., -0.1772, -0.0136, -0.1979], [ 0.0135, -0.1154, 0.0325, ..., 0.0742, -0.1267, -0.0540], [-0.1269, -0.1303, 0.0923, ..., -0.1172, -0.0654, 0.1708]], device='cuda:0'), grad: tensor([[ 8.0373e-07, 2.5090e-06, 1.1055e-06, ..., 2.3954e-06, 9.3132e-10, 7.5996e-07], [ 5.5879e-09, 9.3132e-09, 5.8673e-08, ..., 1.5832e-08, -1.6764e-08, 2.6077e-08], [ 3.0734e-08, 7.4506e-09, -6.2399e-07, ..., 8.6613e-08, 3.7253e-09, 8.3819e-09], ..., [ 2.5146e-08, -9.3132e-09, 2.4214e-08, ..., 5.0291e-08, 5.5879e-09, 1.6764e-08], [-2.2613e-06, -3.2783e-07, -1.9372e-06, ..., -2.0452e-06, 2.7940e-09, -1.2731e-06], [ 4.0419e-07, 6.9849e-08, 3.5204e-07, ..., 3.7719e-07, 2.7940e-09, 1.7602e-07]], device='cuda:0') Epoch 251, bias, value: tensor([-0.0074, -0.0045, 0.0101, 0.0094, 0.0347, 0.0433, -0.0210, 0.0075, -0.0388, -0.0251], device='cuda:0'), grad: tensor([ 1.1213e-05, 2.5798e-07, -3.9265e-06, 3.4906e-06, -2.0582e-07, 1.7053e-06, -7.2569e-06, 2.1979e-07, -6.9179e-06, 1.4063e-06], device='cuda:0') 100 0.0001 changing lr epoch 250, time 217.23, cls_loss 0.0014 cls_loss_mapping 0.0026 cls_loss_causal 0.5175 re_mapping 0.0055 re_causal 0.0159 /// teacc 99.04 lr 0.00010000 Epoch 252, weight, value: tensor([[-0.0711, 0.1599, 0.0760, ..., -0.2414, -0.1243, 0.0226], [ 0.1804, -0.0236, -0.0880, ..., 0.0937, 0.1261, -0.1599], [-0.1153, -0.0648, -0.0499, ..., -0.0915, -0.0406, -0.0829], ..., [-0.0868, 0.1745, -0.1945, ..., -0.1763, -0.0139, -0.1984], [ 0.0133, -0.1177, 0.0326, ..., 0.0736, -0.1284, -0.0539], [-0.1281, -0.1306, 0.0930, ..., -0.1171, -0.0658, 0.1720]], device='cuda:0'), grad: tensor([[ 9.3132e-09, 1.0245e-08, 1.3970e-08, ..., 1.0245e-08, 0.0000e+00, -5.5879e-09], [ 5.7369e-07, 4.9174e-06, 1.4808e-07, ..., 1.0747e-06, 0.0000e+00, 1.5832e-07], [ 1.5832e-08, 4.1910e-08, 5.5879e-09, ..., 9.9652e-08, 0.0000e+00, 1.7695e-08], ..., [-1.0403e-06, -9.9316e-06, 1.0245e-08, ..., 1.5553e-07, 0.0000e+00, 1.3970e-08], [-2.6077e-08, 6.2399e-08, 1.9558e-08, ..., -1.4957e-06, 0.0000e+00, 2.3283e-08], [ 4.9081e-07, 4.6901e-06, -4.8243e-07, ..., 1.3970e-08, 0.0000e+00, -5.3924e-07]], device='cuda:0') Epoch 252, bias, value: tensor([-0.0075, -0.0044, 0.0097, 0.0094, 0.0341, 0.0433, -0.0203, 0.0081, -0.0400, -0.0247], device='cuda:0'), grad: tensor([ 7.4506e-08, 9.9093e-06, 2.3283e-07, 1.0431e-07, 5.0478e-07, 8.7731e-07, -2.5891e-07, -1.4395e-05, -3.0696e-06, 6.0126e-06], device='cuda:0') 100 0.0001 changing lr epoch 251, time 216.99, cls_loss 0.0012 cls_loss_mapping 0.0021 cls_loss_causal 0.5002 re_mapping 0.0052 re_causal 0.0150 /// teacc 99.05 lr 0.00010000 Epoch 253, weight, value: tensor([[-0.0714, 0.1604, 0.0753, ..., -0.2418, -0.1245, 0.0219], [ 0.1822, -0.0239, -0.0882, ..., 0.0939, 0.1279, -0.1600], [-0.1173, -0.0648, -0.0497, ..., -0.0918, -0.0423, -0.0836], ..., [-0.0870, 0.1750, -0.1946, ..., -0.1763, -0.0145, -0.1987], [ 0.0135, -0.1180, 0.0328, ..., 0.0741, -0.1294, -0.0537], [-0.1284, -0.1309, 0.0936, ..., -0.1173, -0.0660, 0.1727]], device='cuda:0'), grad: tensor([[ 9.3132e-09, -5.0478e-07, -4.3772e-08, ..., 7.4506e-09, -9.3132e-09, -9.1735e-07], [-2.0489e-08, 1.2666e-07, 2.0489e-08, ..., -2.8871e-08, -3.7253e-08, 1.1269e-07], [ 4.9360e-08, 3.3528e-08, 7.4506e-09, ..., 6.3330e-08, -5.3085e-08, 3.7253e-08], ..., [ 3.4180e-07, -1.1176e-07, 1.6764e-07, ..., 2.7940e-08, 8.1956e-08, 9.8068e-07], [-2.8871e-08, 1.5832e-08, -3.9116e-08, ..., -7.1712e-08, 3.7253e-09, 1.9558e-08], [-5.1316e-07, 9.4995e-08, -2.5332e-07, ..., 6.6124e-08, 5.5879e-09, -1.5320e-06]], device='cuda:0') Epoch 253, bias, value: tensor([-0.0079, -0.0036, 0.0089, 0.0095, 0.0341, 0.0432, -0.0201, 0.0082, -0.0401, -0.0248], device='cuda:0'), grad: tensor([-3.8147e-06, 4.9267e-07, -8.6240e-07, 1.9558e-07, 2.0638e-06, 1.7416e-07, 2.2985e-06, 4.0196e-06, -7.1712e-08, -4.4890e-06], device='cuda:0') 100 0.0001 changing lr epoch 252, time 218.08, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4895 re_mapping 0.0052 re_causal 0.0150 /// teacc 98.93 lr 0.00010000 Epoch 254, weight, value: tensor([[-0.0719, 0.1627, 0.0754, ..., -0.2424, -0.1246, 0.0220], [ 0.1830, -0.0238, -0.0887, ..., 0.0942, 0.1286, -0.1603], [-0.1178, -0.0649, -0.0492, ..., -0.0920, -0.0426, -0.0836], ..., [-0.0878, 0.1750, -0.1948, ..., -0.1770, -0.0162, -0.1994], [ 0.0132, -0.1183, 0.0328, ..., 0.0741, -0.1303, -0.0542], [-0.1294, -0.1310, 0.0940, ..., -0.1174, -0.0664, 0.1734]], device='cuda:0'), grad: tensor([[ 4.6566e-09, -1.9595e-06, 5.5879e-09, ..., 4.6566e-09, -9.3132e-10, 2.7940e-09], [ 7.5437e-08, 3.6322e-07, 6.9849e-08, ..., 6.9849e-08, -4.6566e-09, 1.1176e-08], [ 9.2480e-07, 8.5682e-08, 7.4785e-07, ..., 8.6334e-07, 3.7253e-09, 9.3132e-10], ..., [ 7.9162e-08, -1.5087e-07, 1.0431e-07, ..., 8.5682e-08, 9.3132e-10, 6.5193e-08], [-2.1905e-06, 1.0617e-07, -1.7239e-06, ..., -2.0452e-06, 0.0000e+00, 6.1467e-08], [ 2.5146e-08, 2.6077e-08, -5.8580e-07, ..., 3.5390e-08, 9.3132e-10, -1.0859e-06]], device='cuda:0') Epoch 254, bias, value: tensor([-0.0069, -0.0034, 0.0088, 0.0096, 0.0340, 0.0431, -0.0205, 0.0080, -0.0403, -0.0247], device='cuda:0'), grad: tensor([-7.9572e-06, 1.7174e-06, 8.0541e-06, -1.9088e-05, 2.4997e-06, 3.5074e-06, 6.5081e-06, 1.5050e-05, -7.6964e-06, -2.6077e-06], device='cuda:0') 100 0.0001 changing lr epoch 253, time 216.90, cls_loss 0.0014 cls_loss_mapping 0.0023 cls_loss_causal 0.4777 re_mapping 0.0056 re_causal 0.0154 /// teacc 99.07 lr 0.00010000 Epoch 255, weight, value: tensor([[-0.0724, 0.1644, 0.0751, ..., -0.2433, -0.1254, 0.0215], [ 0.1830, -0.0240, -0.0901, ..., 0.0941, 0.1281, -0.1604], [-0.1176, -0.0647, -0.0479, ..., -0.0919, -0.0420, -0.0847], ..., [-0.0878, 0.1752, -0.1950, ..., -0.1774, -0.0160, -0.2012], [ 0.0126, -0.1188, 0.0323, ..., 0.0736, -0.1327, -0.0550], [-0.1299, -0.1312, 0.0949, ..., -0.1175, -0.0668, 0.1748]], device='cuda:0'), grad: tensor([[ 0.0000e+00, -9.0338e-08, 4.0978e-08, ..., 3.3528e-08, 1.8626e-09, 2.8871e-08], [-2.3190e-07, 2.4624e-06, 3.7253e-09, ..., -2.9150e-07, 6.8080e-07, 4.6566e-09], [ 4.8429e-08, 5.0850e-07, 9.3132e-09, ..., 6.7055e-08, 1.3504e-07, 6.5193e-09], ..., [ 8.3819e-08, -5.7593e-06, 5.5879e-09, ..., 8.1956e-08, -1.6745e-06, 4.6566e-09], [ 6.3330e-08, 2.5593e-06, 1.9558e-08, ..., 1.0151e-07, 7.7393e-07, 7.5437e-08], [ 4.2841e-08, 2.7847e-07, 1.0245e-08, ..., 1.5553e-07, 6.6124e-08, -2.6077e-08]], device='cuda:0') Epoch 255, bias, value: tensor([-0.0067, -0.0039, 0.0094, 0.0104, 0.0338, 0.0422, -0.0203, 0.0077, -0.0409, -0.0243], device='cuda:0'), grad: tensor([-2.2352e-08, 3.6582e-06, 1.0226e-06, 4.1015e-06, -2.3842e-07, -4.1462e-06, -8.1025e-08, -9.8348e-06, 4.7460e-06, 7.8231e-07], device='cuda:0') 100 0.0001 changing lr epoch 254, time 217.28, cls_loss 0.0017 cls_loss_mapping 0.0022 cls_loss_causal 0.5136 re_mapping 0.0055 re_causal 0.0148 /// teacc 99.02 lr 0.00010000 Epoch 256, weight, value: tensor([[-0.0733, 0.1656, 0.0749, ..., -0.2445, -0.1253, 0.0219], [ 0.1831, -0.0240, -0.0915, ..., 0.0941, 0.1281, -0.1610], [-0.1181, -0.0650, -0.0467, ..., -0.0920, -0.0419, -0.0852], ..., [-0.0884, 0.1754, -0.1955, ..., -0.1782, -0.0163, -0.2016], [ 0.0133, -0.1191, 0.0330, ..., 0.0744, -0.1331, -0.0552], [-0.1304, -0.1312, 0.0955, ..., -0.1177, -0.0673, 0.1751]], device='cuda:0'), grad: tensor([[ 3.1665e-08, -1.9222e-06, -1.9278e-07, ..., 3.1665e-08, 9.3132e-10, -1.7807e-06], [-2.5146e-08, 5.4017e-08, 5.5879e-09, ..., -5.5879e-09, -9.3132e-10, 3.9116e-08], [ 3.5390e-08, 2.0489e-08, 5.6811e-08, ..., -8.3819e-09, -3.3528e-08, 1.3970e-08], ..., [ 1.3039e-08, 1.8626e-08, 8.3819e-09, ..., 2.0862e-07, 3.7253e-09, 2.5146e-08], [ 1.3970e-07, 4.8429e-08, 8.0094e-08, ..., 1.7881e-07, 1.8626e-09, 1.8626e-08], [ 1.1176e-08, 1.6950e-06, 1.6950e-07, ..., 9.3877e-07, 1.8626e-09, 1.5628e-06]], device='cuda:0') Epoch 256, bias, value: tensor([-0.0064, -0.0041, 0.0095, 0.0101, 0.0338, 0.0421, -0.0186, 0.0076, -0.0407, -0.0242], device='cuda:0'), grad: tensor([-6.1169e-06, 3.1572e-07, -4.4797e-07, -2.8498e-07, -1.8813e-06, 4.9267e-07, -3.4273e-07, 4.8615e-07, 7.1153e-07, 7.0743e-06], device='cuda:0') 100 0.0001 changing lr epoch 255, time 217.54, cls_loss 0.0013 cls_loss_mapping 0.0021 cls_loss_causal 0.4901 re_mapping 0.0054 re_causal 0.0154 /// teacc 98.98 lr 0.00010000 Epoch 257, weight, value: tensor([[-0.0739, 0.1671, 0.0753, ..., -0.2449, -0.1253, 0.0229], [ 0.1816, -0.0265, -0.0920, ..., 0.0942, 0.1281, -0.1614], [-0.1183, -0.0651, -0.0464, ..., -0.0920, -0.0419, -0.0856], ..., [-0.0865, 0.1774, -0.1955, ..., -0.1795, -0.0159, -0.2020], [ 0.0138, -0.1195, 0.0334, ..., 0.0751, -0.1336, -0.0551], [-0.1315, -0.1314, 0.0956, ..., -0.1188, -0.0675, 0.1750]], device='cuda:0'), grad: tensor([[ 4.6566e-09, -1.0822e-06, -3.0734e-08, ..., -2.7660e-07, 0.0000e+00, -1.0226e-06], [ 8.3819e-09, 7.4506e-09, 3.0734e-08, ..., 2.7940e-09, 0.0000e+00, 5.4948e-08], [ 1.8626e-09, 5.5879e-09, 4.6566e-09, ..., 1.8626e-09, 0.0000e+00, 5.5879e-09], ..., [ 1.9558e-08, 1.4901e-08, 1.1269e-07, ..., 8.3819e-09, 0.0000e+00, 2.5425e-07], [-4.4703e-08, 2.0489e-08, 1.7695e-08, ..., -6.3330e-08, 0.0000e+00, 3.6322e-08], [-4.6566e-08, 1.7975e-07, -2.2445e-07, ..., 2.2352e-08, 0.0000e+00, -3.7812e-07]], device='cuda:0') Epoch 257, bias, value: tensor([-0.0058, -0.0054, 0.0097, 0.0100, 0.0344, 0.0420, -0.0185, 0.0091, -0.0407, -0.0248], device='cuda:0'), grad: tensor([-3.2298e-06, 1.6205e-07, 8.3819e-09, -5.5879e-09, 4.7684e-07, 2.4345e-06, 3.3528e-07, 7.4599e-07, 4.2841e-08, -9.5740e-07], device='cuda:0') 100 0.0001 changing lr epoch 256, time 217.53, cls_loss 0.0012 cls_loss_mapping 0.0032 cls_loss_causal 0.5167 re_mapping 0.0052 re_causal 0.0154 /// teacc 99.05 lr 0.00010000 Epoch 258, weight, value: tensor([[-0.0743, 0.1677, 0.0751, ..., -0.2451, -0.1254, 0.0229], [ 0.1813, -0.0272, -0.0922, ..., 0.0942, 0.1282, -0.1617], [-0.1184, -0.0648, -0.0463, ..., -0.0921, -0.0421, -0.0857], ..., [-0.0879, 0.1766, -0.1957, ..., -0.1813, -0.0160, -0.2026], [ 0.0170, -0.1166, 0.0330, ..., 0.0768, -0.1348, -0.0552], [-0.1318, -0.1315, 0.0960, ..., -0.1203, -0.0676, 0.1753]], device='cuda:0'), grad: tensor([[ 4.6566e-08, -1.8347e-07, 6.8918e-08, ..., 1.0245e-07, 0.0000e+00, -8.7544e-08], [-1.1176e-08, 1.3970e-08, 2.9802e-08, ..., -1.5832e-08, -1.8626e-09, 3.0734e-08], [ 4.2841e-08, 5.4948e-08, 5.3085e-08, ..., 8.7544e-08, -5.5879e-09, 4.2841e-08], ..., [ 2.1420e-08, -1.3318e-07, 8.2888e-08, ..., 5.2154e-08, 1.8626e-09, 1.0524e-07], [-1.2293e-07, 1.3039e-07, -4.3493e-07, ..., -1.0934e-06, 2.7940e-09, -2.7753e-07], [ 1.4901e-07, 7.8231e-08, -1.8720e-07, ..., 3.3248e-07, 9.3132e-10, -5.5321e-07]], device='cuda:0') Epoch 258, bias, value: tensor([-0.0057, -0.0057, 0.0098, 0.0104, 0.0353, 0.0418, -0.0188, 0.0083, -0.0382, -0.0254], device='cuda:0'), grad: tensor([-4.2189e-07, 1.5181e-07, 2.4308e-07, 1.7099e-06, 1.5600e-06, 3.1851e-07, -1.0431e-07, 3.7625e-07, -3.4049e-06, -4.4145e-07], device='cuda:0') 100 0.0001 changing lr epoch 257, time 216.89, cls_loss 0.0012 cls_loss_mapping 0.0018 cls_loss_causal 0.4978 re_mapping 0.0053 re_causal 0.0149 /// teacc 99.02 lr 0.00010000 Epoch 259, weight, value: tensor([[-0.0755, 0.1685, 0.0754, ..., -0.2461, -0.1254, 0.0233], [ 0.1816, -0.0272, -0.0925, ..., 0.0944, 0.1284, -0.1631], [-0.1189, -0.0648, -0.0463, ..., -0.0924, -0.0422, -0.0862], ..., [-0.0881, 0.1764, -0.1970, ..., -0.1822, -0.0164, -0.2056], [ 0.0170, -0.1167, 0.0329, ..., 0.0770, -0.1352, -0.0553], [-0.1325, -0.1312, 0.0978, ..., -0.1197, -0.0676, 0.1780]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.2107e-08, 0.0000e+00, ..., 3.7253e-09, 0.0000e+00, 1.8626e-09], [-8.3260e-07, -5.2713e-07, 0.0000e+00, ..., -7.2643e-07, 0.0000e+00, 2.0489e-08], [ 5.5879e-09, -1.3551e-06, 0.0000e+00, ..., 1.5739e-07, 0.0000e+00, 2.3283e-08], ..., [ 6.7241e-07, 9.1176e-07, 2.7940e-09, ..., 1.4035e-06, 0.0000e+00, 9.6858e-08], [ 1.6764e-08, 4.7497e-08, 1.8626e-09, ..., 2.7008e-08, 0.0000e+00, 7.4506e-09], [ 1.3597e-07, 8.4098e-07, -2.8871e-08, ..., 5.9485e-05, 0.0000e+00, 9.3505e-06]], device='cuda:0') Epoch 259, bias, value: tensor([-0.0056, -0.0056, 0.0095, 0.0104, 0.0344, 0.0418, -0.0177, 0.0078, -0.0383, -0.0244], device='cuda:0'), grad: tensor([ 3.9116e-08, -1.2731e-06, -2.4661e-06, 1.2573e-07, -1.6522e-04, -9.0338e-08, 2.0303e-07, 3.5502e-06, 1.6205e-07, 1.6499e-04], device='cuda:0') 100 0.0001 changing lr epoch 258, time 217.23, cls_loss 0.0013 cls_loss_mapping 0.0032 cls_loss_causal 0.4974 re_mapping 0.0052 re_causal 0.0153 /// teacc 98.99 lr 0.00010000 Epoch 260, weight, value: tensor([[-0.0758, 0.1687, 0.0754, ..., -0.2467, -0.1254, 0.0234], [ 0.1819, -0.0270, -0.0926, ..., 0.0946, 0.1286, -0.1633], [-0.1191, -0.0672, -0.0463, ..., -0.0925, -0.0423, -0.0864], ..., [-0.0882, 0.1776, -0.1972, ..., -0.1831, -0.0166, -0.2058], [ 0.0166, -0.1169, 0.0319, ..., 0.0762, -0.1377, -0.0562], [-0.1323, -0.1313, 0.0984, ..., -0.1219, -0.0678, 0.1784]], device='cuda:0'), grad: tensor([[ 8.1956e-08, -6.7055e-08, 1.6764e-08, ..., 1.8068e-07, 0.0000e+00, 2.1420e-08], [ 2.8592e-07, 5.9605e-08, 1.7136e-07, ..., 3.9767e-07, 0.0000e+00, 1.7695e-07], [ 4.7497e-08, 6.7055e-08, 3.5390e-08, ..., 6.9849e-08, 0.0000e+00, 2.1420e-08], ..., [ 1.9651e-07, -4.7497e-08, 8.1956e-08, ..., 3.0920e-07, 0.0000e+00, 8.0094e-08], [-3.1423e-06, -3.7160e-07, -1.0645e-06, ..., -4.8727e-06, 0.0000e+00, -1.6764e-06], [ 7.6555e-07, 2.4308e-07, 3.8277e-07, ..., 1.1856e-06, 0.0000e+00, 3.5297e-07]], device='cuda:0') Epoch 260, bias, value: tensor([-0.0056, -0.0055, 0.0088, 0.0105, 0.0364, 0.0419, -0.0178, 0.0084, -0.0391, -0.0260], device='cuda:0'), grad: tensor([ 2.8126e-07, 1.0878e-06, 3.0454e-07, 7.1898e-07, 4.8336e-07, 3.6880e-06, 1.2200e-06, 5.9139e-07, -1.1377e-05, 2.9951e-06], device='cuda:0') 100 0.0001 changing lr epoch 259, time 217.25, cls_loss 0.0012 cls_loss_mapping 0.0023 cls_loss_causal 0.5104 re_mapping 0.0053 re_causal 0.0153 /// teacc 98.92 lr 0.00010000 Epoch 261, weight, value: tensor([[-0.0761, 0.1691, 0.0754, ..., -0.2471, -0.1255, 0.0234], [ 0.1821, -0.0271, -0.0927, ..., 0.0948, 0.1290, -0.1633], [-0.1194, -0.0670, -0.0463, ..., -0.0927, -0.0425, -0.0871], ..., [-0.0884, 0.1776, -0.1974, ..., -0.1837, -0.0171, -0.2060], [ 0.0168, -0.1170, 0.0319, ..., 0.0769, -0.1382, -0.0562], [-0.1331, -0.1313, 0.0987, ..., -0.1221, -0.0680, 0.1788]], device='cuda:0'), grad: tensor([[ 1.3970e-08, -3.7253e-09, 1.5832e-08, ..., 2.5146e-08, 0.0000e+00, 1.3039e-08], [ 1.8626e-09, 4.2003e-07, 1.3039e-08, ..., 2.2352e-08, 2.9802e-08, 1.3970e-08], [ 5.5879e-09, 3.7253e-09, 6.5193e-09, ..., 1.2573e-07, 0.0000e+00, 5.5879e-09], ..., [ 4.6566e-09, -4.6846e-07, 5.1223e-08, ..., 4.6566e-08, -3.1665e-08, 7.3574e-08], [-1.9372e-07, 1.8626e-09, -3.2876e-07, ..., -6.2585e-07, 0.0000e+00, -3.0641e-07], [ 1.1362e-07, 4.0047e-08, 2.3283e-08, ..., 2.4308e-07, 9.3132e-10, -7.3574e-08]], device='cuda:0') Epoch 261, bias, value: tensor([-0.0055, -0.0053, 0.0091, 0.0108, 0.0363, 0.0417, -0.0177, 0.0079, -0.0391, -0.0260], device='cuda:0'), grad: tensor([ 8.5682e-08, 9.6392e-07, -2.7847e-07, 1.3039e-07, 4.5169e-07, 2.9895e-07, 2.5425e-07, -3.0734e-07, -1.7826e-06, 2.0117e-07], device='cuda:0') 100 0.0001 changing lr epoch 260, time 219.29, cls_loss 0.0019 cls_loss_mapping 0.0031 cls_loss_causal 0.4680 re_mapping 0.0054 re_causal 0.0142 /// teacc 98.94 lr 0.00010000 Epoch 262, weight, value: tensor([[-0.0774, 0.1694, 0.0748, ..., -0.2480, -0.1255, 0.0231], [ 0.1822, -0.0272, -0.0928, ..., 0.0947, 0.1295, -0.1635], [-0.1200, -0.0670, -0.0462, ..., -0.0929, -0.0429, -0.0875], ..., [-0.0885, 0.1785, -0.1978, ..., -0.1846, -0.0177, -0.2063], [ 0.0177, -0.1173, 0.0320, ..., 0.0791, -0.1383, -0.0558], [-0.1343, -0.1314, 0.0994, ..., -0.1246, -0.0685, 0.1797]], device='cuda:0'), grad: tensor([[ 1.8161e-07, -6.5193e-09, 2.9895e-07, ..., 1.4156e-07, 1.3039e-08, 3.7998e-07], [-3.5949e-07, 6.7055e-08, 4.7497e-08, ..., -3.8184e-07, -2.7288e-07, 9.6858e-08], [ 1.3225e-07, 1.9558e-08, 3.4459e-08, ..., 1.7788e-07, 3.3528e-08, 4.4703e-08], ..., [ 1.2945e-07, -3.2037e-07, 1.1092e-06, ..., 2.4773e-07, 6.1467e-08, 1.5972e-06], [-8.1956e-08, 3.7253e-08, 3.3528e-08, ..., -9.6206e-07, 3.5390e-08, 3.8743e-07], [ 2.2352e-07, 2.0489e-07, -4.5821e-06, ..., 1.1520e-06, 7.4506e-08, -7.2792e-06]], device='cuda:0') Epoch 262, bias, value: tensor([-0.0059, -0.0054, 0.0093, 0.0091, 0.0381, 0.0413, -0.0171, 0.0094, -0.0388, -0.0276], device='cuda:0'), grad: tensor([ 1.4091e-06, -3.8557e-07, 2.0210e-07, 7.3574e-07, 1.1601e-05, 3.0994e-06, -1.0114e-06, 4.9546e-06, -1.2415e-06, -1.9401e-05], device='cuda:0') 100 0.0001 changing lr epoch 261, time 222.27, cls_loss 0.0013 cls_loss_mapping 0.0019 cls_loss_causal 0.4975 re_mapping 0.0057 re_causal 0.0160 /// teacc 99.08 lr 0.00010000 Epoch 263, weight, value: tensor([[-0.0780, 0.1700, 0.0747, ..., -0.2486, -0.1260, 0.0227], [ 0.1822, -0.0276, -0.0929, ..., 0.0948, 0.1296, -0.1637], [-0.1204, -0.0669, -0.0463, ..., -0.0931, -0.0431, -0.0877], ..., [-0.0884, 0.1788, -0.1981, ..., -0.1855, -0.0174, -0.2068], [ 0.0178, -0.1174, 0.0321, ..., 0.0795, -0.1389, -0.0558], [-0.1345, -0.1314, 0.1011, ..., -0.1246, -0.0689, 0.1812]], device='cuda:0'), grad: tensor([[ 1.8626e-09, -2.7940e-09, 1.0245e-08, ..., 2.7940e-09, 0.0000e+00, -1.8626e-09], [-1.6764e-08, 6.7055e-08, 9.3132e-09, ..., -2.4214e-08, -7.4506e-09, 9.3132e-10], [ 8.3819e-09, 6.6124e-08, 1.3039e-08, ..., 1.7695e-08, 1.0245e-08, 0.0000e+00], ..., [ 1.8626e-08, -3.9302e-07, 3.9116e-08, ..., 5.0291e-08, -3.3528e-08, 1.8626e-09], [-2.0023e-07, 2.2259e-07, 1.1083e-07, ..., -4.7870e-07, 2.8871e-08, -9.3132e-09], [ 1.3970e-08, 1.5832e-08, 7.4506e-09, ..., 1.0803e-07, 1.8626e-09, -3.6322e-08]], device='cuda:0') Epoch 263, bias, value: tensor([-0.0060, -0.0057, 0.0093, 0.0092, 0.0377, 0.0412, -0.0164, 0.0093, -0.0387, -0.0272], device='cuda:0'), grad: tensor([ 3.1665e-08, 9.7789e-08, 5.7649e-07, -1.3277e-05, -3.1944e-07, 1.1794e-05, 3.3062e-07, 3.9954e-07, 7.6368e-08, 2.7381e-07], device='cuda:0') 100 0.0001 changing lr epoch 262, time 221.87, cls_loss 0.0016 cls_loss_mapping 0.0027 cls_loss_causal 0.4670 re_mapping 0.0052 re_causal 0.0147 /// teacc 99.03 lr 0.00010000 Epoch 264, weight, value: tensor([[-0.0803, 0.1704, 0.0744, ..., -0.2508, -0.1275, 0.0224], [ 0.1814, -0.0280, -0.0929, ..., 0.0946, 0.1291, -0.1638], [-0.1206, -0.0670, -0.0463, ..., -0.0931, -0.0432, -0.0879], ..., [-0.0881, 0.1790, -0.1994, ..., -0.1863, -0.0167, -0.2085], [ 0.0178, -0.1175, 0.0305, ..., 0.0797, -0.1394, -0.0570], [-0.1354, -0.1314, 0.1014, ..., -0.1250, -0.0689, 0.1815]], device='cuda:0'), grad: tensor([[ 3.5390e-08, -1.2675e-06, -3.7346e-07, ..., 4.5635e-08, 5.5879e-09, -4.7777e-07], [-7.1712e-08, 8.2888e-08, 9.3132e-08, ..., -1.7695e-07, -8.1025e-08, 1.1735e-07], [ 4.2841e-08, 2.5332e-07, 1.4808e-07, ..., 6.9849e-08, 5.2154e-08, 1.0151e-07], ..., [ 1.6950e-07, -9.4064e-08, 1.5832e-07, ..., 3.0268e-07, 3.9116e-08, 1.5832e-07], [ 1.1921e-07, 3.0827e-07, 7.4506e-09, ..., -1.2107e-08, 9.8720e-08, 3.3714e-07], [ 1.8906e-07, 2.8405e-07, 1.9837e-07, ..., 2.9150e-07, 9.7789e-08, 2.9057e-07]], device='cuda:0') Epoch 264, bias, value: tensor([-0.0068, -0.0063, 0.0093, 0.0097, 0.0377, 0.0410, -0.0142, 0.0092, -0.0390, -0.0271], device='cuda:0'), grad: tensor([-6.4075e-06, 4.9360e-08, 1.5637e-06, -8.1025e-08, 1.0524e-07, -1.6484e-06, 2.6133e-06, 7.4878e-07, 1.1744e-06, 1.8906e-06], device='cuda:0') 100 0.0001 changing lr epoch 263, time 220.78, cls_loss 0.0016 cls_loss_mapping 0.0029 cls_loss_causal 0.5002 re_mapping 0.0053 re_causal 0.0152 /// teacc 98.99 lr 0.00010000 Epoch 265, weight, value: tensor([[-0.0805, 0.1710, 0.0750, ..., -0.2517, -0.1275, 0.0224], [ 0.1825, -0.0274, -0.0931, ..., 0.0954, 0.1295, -0.1638], [-0.1208, -0.0676, -0.0461, ..., -0.0933, -0.0432, -0.0884], ..., [-0.0893, 0.1791, -0.1998, ..., -0.1907, -0.0184, -0.2090], [ 0.0170, -0.1178, 0.0293, ..., 0.0769, -0.1399, -0.0603], [-0.1369, -0.1317, 0.1019, ..., -0.1255, -0.0698, 0.1819]], device='cuda:0'), grad: tensor([[ 4.2375e-08, -6.6124e-08, 2.3283e-09, ..., 4.8894e-08, 0.0000e+00, -2.7940e-09], [ 3.7719e-08, 2.2305e-07, 5.0757e-08, ..., 4.7963e-08, 0.0000e+00, 4.5169e-08], [ 1.3039e-08, -6.7521e-08, 1.8161e-08, ..., 2.0955e-08, 0.0000e+00, 8.3819e-09], ..., [ 5.0757e-08, -3.6368e-07, 7.5158e-07, ..., 8.0094e-08, 0.0000e+00, 7.2317e-07], [-3.6508e-07, 1.5087e-07, 2.0489e-08, ..., -4.0093e-07, 0.0000e+00, 5.8673e-08], [ 1.0524e-07, -2.6077e-08, -2.0154e-06, ..., 1.1176e-07, 0.0000e+00, -2.1234e-06]], device='cuda:0') Epoch 265, bias, value: tensor([-0.0065, -0.0055, 0.0093, 0.0108, 0.0380, 0.0404, -0.0140, 0.0086, -0.0407, -0.0274], device='cuda:0'), grad: tensor([ 2.2585e-07, 6.9151e-07, -3.8138e-07, 1.1418e-06, 3.9749e-06, -3.2689e-07, 3.9209e-07, 2.9579e-06, -1.7658e-06, -6.9141e-06], device='cuda:0') 100 0.0001 changing lr epoch 264, time 222.38, cls_loss 0.0013 cls_loss_mapping 0.0020 cls_loss_causal 0.5126 re_mapping 0.0051 re_causal 0.0147 /// teacc 99.01 lr 0.00010000 Epoch 266, weight, value: tensor([[-0.0806, 0.1719, 0.0751, ..., -0.2522, -0.1275, 0.0225], [ 0.1825, -0.0276, -0.0933, ..., 0.0954, 0.1295, -0.1641], [-0.1209, -0.0676, -0.0466, ..., -0.0933, -0.0433, -0.0886], ..., [-0.0893, 0.1794, -0.2009, ..., -0.1911, -0.0185, -0.2091], [ 0.0181, -0.1178, 0.0294, ..., 0.0789, -0.1400, -0.0590], [-0.1371, -0.1319, 0.1022, ..., -0.1256, -0.0698, 0.1823]], device='cuda:0'), grad: tensor([[ 1.0338e-07, 2.0489e-08, 1.3504e-08, ..., 7.5437e-08, 1.3970e-09, 4.6566e-09], [-4.9472e-06, -4.6100e-07, -3.8650e-08, ..., -7.7337e-06, -3.8892e-06, 1.0710e-08], [ 3.3751e-06, 2.1886e-08, 1.2107e-08, ..., 6.2287e-06, 3.5241e-06, 7.4506e-09], ..., [ 5.1735e-07, 2.1420e-08, 2.3749e-08, ..., 6.6636e-07, 2.8824e-07, 1.3039e-08], [ 6.1886e-07, 1.4482e-07, 1.9651e-07, ..., 6.9151e-07, 2.3283e-09, 2.1001e-07], [ 4.7917e-07, 1.8766e-07, 1.0803e-07, ..., 4.1211e-07, 5.5879e-09, 9.6858e-08]], device='cuda:0') Epoch 266, bias, value: tensor([-0.0061, -0.0057, 0.0092, 0.0117, 0.0379, 0.0402, -0.0154, 0.0082, -0.0397, -0.0274], device='cuda:0'), grad: tensor([ 2.0396e-07, -1.8522e-05, 1.4029e-05, 5.4855e-07, 1.6065e-07, -8.8513e-06, 7.7784e-06, 1.6447e-06, 1.7984e-06, 1.2144e-06], device='cuda:0') 100 0.0001 changing lr