Collecting h5py>=2.9.0 Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (2.5 kB) Collecting ml-collections Downloading ml_collections-0.1.1.tar.gz (77 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 77.9/77.9 kB 3.9 MB/s eta 0:00:00 Preparing metadata (setup.py): started Preparing metadata (setup.py): finished with status 'done' Requirement already satisfied: numpy in ./lib/python3.11/site-packages (2.0.0) Collecting opencv-python==4.5.5.62 Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (18 kB) Collecting scipy>=1.3.2 Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (60 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.8/60.8 kB 3.9 MB/s eta 0:00:00 Collecting tensorboard Downloading tensorboard-2.17.0-py3-none-any.whl.metadata (1.6 kB) Collecting tensorboardX>=1.4 Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl.metadata (5.8 kB) Collecting timm Downloading timm-1.0.7-py3-none-any.whl.metadata (47 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 47.5/47.5 kB 11.3 MB/s eta 0:00:00 Collecting absl-py (from ml-collections) Downloading absl_py-2.1.0-py3-none-any.whl.metadata (2.3 kB) Requirement already satisfied: PyYAML in ./lib/python3.11/site-packages (from ml-collections) (6.0.1) Requirement already satisfied: six in ./lib/python3.11/site-packages (from ml-collections) (1.16.0) Collecting contextlib2 (from ml-collections) Downloading contextlib2-21.6.0-py2.py3-none-any.whl.metadata (4.1 kB) Collecting grpcio>=1.48.2 (from tensorboard) Downloading grpcio-1.65.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.3 kB) Collecting markdown>=2.6.8 (from tensorboard) Downloading Markdown-3.6-py3-none-any.whl.metadata (7.0 kB) Collecting protobuf!=4.24.0,<5.0.0,>=3.19.6 (from tensorboard) Downloading protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl.metadata (541 bytes) Requirement already satisfied: setuptools>=41.0.0 in ./lib/python3.11/site-packages (from tensorboard) (71.0.1) Collecting tensorboard-data-server<0.8.0,>=0.7.0 (from tensorboard) Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl.metadata (1.1 kB) Collecting werkzeug>=1.0.1 (from tensorboard) Downloading werkzeug-3.0.3-py3-none-any.whl.metadata (3.7 kB) Collecting packaging (from tensorboardX>=1.4) Downloading packaging-24.1-py3-none-any.whl.metadata (3.2 kB) Requirement already satisfied: torch in ./lib/python3.11/site-packages (from timm) (2.3.1) Requirement already satisfied: torchvision in ./lib/python3.11/site-packages (from timm) (0.18.1) Collecting huggingface_hub (from timm) Downloading huggingface_hub-0.24.0-py3-none-any.whl.metadata (13 kB) Collecting safetensors (from timm) Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl.metadata (3.8 kB) Requirement already satisfied: MarkupSafe>=2.1.1 in ./lib/python3.11/site-packages (from werkzeug>=1.0.1->tensorboard) (2.1.5) Requirement already satisfied: filelock in ./lib/python3.11/site-packages (from huggingface_hub->timm) (3.15.4) Collecting fsspec>=2023.5.0 (from huggingface_hub->timm) Downloading fsspec-2024.6.1-py3-none-any.whl.metadata (11 kB) Requirement already satisfied: requests in ./lib/python3.11/site-packages (from huggingface_hub->timm) (2.32.3) Collecting tqdm>=4.42.1 (from huggingface_hub->timm) Downloading tqdm-4.66.4-py3-none-any.whl.metadata (57 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 57.6/57.6 kB 18.9 MB/s eta 0:00:00 Requirement already satisfied: typing-extensions>=3.7.4.3 in ./lib/python3.11/site-packages (from huggingface_hub->timm) (4.12.2) Requirement already satisfied: sympy in ./lib/python3.11/site-packages (from torch->timm) (1.13.0) Requirement already satisfied: networkx in ./lib/python3.11/site-packages (from torch->timm) (3.3) Requirement already satisfied: jinja2 in ./lib/python3.11/site-packages (from torch->timm) (3.1.4) Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in ./lib/python3.11/site-packages (from torchvision->timm) (9.4.0) Requirement already satisfied: charset-normalizer<4,>=2 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.3.2) Requirement already satisfied: idna<4,>=2.5 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (3.7) Requirement already satisfied: urllib3<3,>=1.21.1 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2.2.2) Requirement already satisfied: certifi>=2017.4.17 in ./lib/python3.11/site-packages (from requests->huggingface_hub->timm) (2024.7.4) Requirement already satisfied: mpmath<1.4,>=1.1.0 in ./lib/python3.11/site-packages (from sympy->torch->timm) (1.3.0) Downloading opencv_python-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (60.4 MB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 60.4/60.4 MB 80.7 MB/s eta 0:00:00 Downloading h5py-3.11.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.4 MB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.4/5.4 MB 195.2 MB/s eta 0:00:00 Downloading scipy-1.14.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (41.1 MB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 41.1/41.1 MB 99.0 MB/s eta 0:00:00 Downloading tensorboard-2.17.0-py3-none-any.whl (5.5 MB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.5/5.5 MB 185.5 MB/s eta 0:00:00 Downloading tensorboardX-2.6.2.2-py2.py3-none-any.whl (101 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 101.7/101.7 kB 32.9 MB/s eta 0:00:00 Downloading timm-1.0.7-py3-none-any.whl (2.3 MB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 2.3/2.3 MB 142.5 MB/s eta 0:00:00 Downloading absl_py-2.1.0-py3-none-any.whl (133 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 133.7/133.7 kB 41.6 MB/s eta 0:00:00 Downloading grpcio-1.65.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.7 MB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.7/5.7 MB 186.4 MB/s eta 0:00:00 Downloading Markdown-3.6-py3-none-any.whl (105 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 105.4/105.4 kB 31.0 MB/s eta 0:00:00 Downloading protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl (294 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 294.6/294.6 kB 76.8 MB/s eta 0:00:00 Downloading tensorboard_data_server-0.7.2-py3-none-manylinux_2_31_x86_64.whl (6.6 MB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 6.6/6.6 MB 181.2 MB/s eta 0:00:00 Downloading werkzeug-3.0.3-py3-none-any.whl (227 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 227.3/227.3 kB 62.6 MB/s eta 0:00:00 Downloading contextlib2-21.6.0-py2.py3-none-any.whl (13 kB) Downloading huggingface_hub-0.24.0-py3-none-any.whl (419 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 419.0/419.0 kB 96.6 MB/s eta 0:00:00 Downloading packaging-24.1-py3-none-any.whl (53 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 54.0/54.0 kB 10.3 MB/s eta 0:00:00 Downloading safetensors-0.4.3-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.2 MB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.2/1.2 MB 143.5 MB/s eta 0:00:00 Downloading fsspec-2024.6.1-py3-none-any.whl (177 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 177.6/177.6 kB 50.9 MB/s eta 0:00:00 Downloading tqdm-4.66.4-py3-none-any.whl (78 kB) ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 78.3/78.3 kB 26.3 MB/s eta 0:00:00 Building wheels for collected packages: ml-collections Building wheel for ml-collections (setup.py): started Building wheel for ml-collections (setup.py): finished with status 'done' Created wheel for ml-collections: filename=ml_collections-0.1.1-py3-none-any.whl size=94508 sha256=310dbc935b23fb0dacb072b058c1f197374aa7241d1279503e2bd3bb0bab5ea9 Stored in directory: /scratch/yuqian_fu/.cache/pip/wheels/28/82/ef/a6971b09a96519d55ce6efef66f0cbcdef2ae9cc1e6b41daf7 Successfully built ml-collections Installing collected packages: werkzeug, tqdm, tensorboard-data-server, scipy, safetensors, protobuf, packaging, opencv-python, markdown, h5py, grpcio, fsspec, contextlib2, absl-py, tensorboardX, tensorboard, ml-collections, huggingface_hub, timm Successfully installed absl-py-2.1.0 contextlib2-21.6.0 fsspec-2024.6.1 grpcio-1.65.1 h5py-3.11.0 huggingface_hub-0.24.0 markdown-3.6 ml-collections-0.1.1 opencv-python-4.5.5.62 packaging-24.1 protobuf-4.25.3 safetensors-0.4.3 scipy-1.14.0 tensorboard-2.17.0 tensorboard-data-server-0.7.2 tensorboardX-2.6.2.2 timm-1.0.7 tqdm-4.66.4 werkzeug-3.0.3 /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_onlyblock2', '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([[-2.5139e-02, -1.7924e-02, 2.9307e-02, ..., -6.4647e-03, -1.8291e-02, -2.4441e-02], [-2.2718e-02, -2.9409e-02, 2.0061e-02, ..., -7.4217e-03, 3.3684e-05, 2.0003e-02], [ 2.5372e-02, 1.9253e-02, -4.3399e-03, ..., -2.4109e-02, -2.7648e-02, -2.6141e-02], ..., [-1.5319e-02, -5.5781e-03, -2.4777e-02, ..., 1.2113e-02, 9.7509e-03, 5.7319e-03], [-2.1644e-02, -3.1244e-02, 2.4834e-03, ..., 4.4214e-03, 4.7801e-03, 2.4475e-02], [ 4.0269e-03, -9.8763e-03, -5.6816e-03, ..., 2.5718e-02, 1.1947e-02, -2.4943e-02]], device='cuda:0'), grad: None Epoch 1, bias, value: tensor([ 0.0088, 0.0188, 0.0148, -0.0022, 0.0142, -0.0066, -0.0144, -0.0159, 0.0256, -0.0067], device='cuda:0'), grad: None 100 0.0001 changing lr ---------------------saving model at epoch 0---------------------------------------------------- epoch 0, time 278.09, cls_loss 1.3028 cls_loss_mapping 1.8310 cls_loss_causal 2.2140 re_mapping 0.1459 re_causal 0.1544 /// teacc 87.55 lr 0.00010000 Epoch 2, weight, value: tensor([[-0.0304, -0.0127, 0.0363, ..., -0.0007, -0.0263, -0.0246], [-0.0255, -0.0372, 0.0133, ..., -0.0162, -0.0066, 0.0178], [ 0.0306, 0.0129, -0.0094, ..., -0.0265, -0.0362, -0.0276], ..., [-0.0109, -0.0030, -0.0264, ..., 0.0158, 0.0134, 0.0089], [-0.0236, -0.0367, -0.0051, ..., -0.0020, 0.0042, 0.0201], [ 0.0010, -0.0127, -0.0044, ..., 0.0254, 0.0161, -0.0236]], device='cuda:0'), grad: tensor([[ 0.0012, 0.0124, -0.0088, ..., -0.0057, 0.0126, 0.0000], [ 0.0092, 0.0006, 0.0007, ..., 0.0030, 0.0177, 0.0000], [-0.0139, -0.0013, 0.0054, ..., 0.0070, 0.0114, 0.0000], ..., [-0.0021, 0.0012, 0.0028, ..., 0.0145, 0.0536, 0.0000], [ 0.0046, 0.0085, -0.0041, ..., -0.0115, -0.0413, 0.0000], [ 0.0007, 0.0008, -0.0082, ..., -0.0449, -0.0875, 0.0000]], device='cuda:0') Epoch 2, bias, value: tensor([ 0.0077, 0.0202, 0.0141, -0.0018, 0.0147, -0.0067, -0.0156, -0.0145, 0.0248, -0.0068], device='cuda:0'), grad: tensor([ 0.0065, 0.0371, -0.0193, 0.0364, 0.0056, -0.0031, -0.0156, 0.0386, -0.0386, -0.0476], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 1---------------------------------------------------- epoch 1, time 277.76, cls_loss 0.4191 cls_loss_mapping 0.7467 cls_loss_causal 1.8580 re_mapping 0.2066 re_causal 0.2696 /// teacc 91.72 lr 0.00010000 Epoch 3, weight, value: tensor([[-0.0329, -0.0113, 0.0411, ..., 0.0012, -0.0298, -0.0248], [-0.0284, -0.0379, 0.0104, ..., -0.0198, -0.0092, 0.0177], [ 0.0338, 0.0105, -0.0126, ..., -0.0277, -0.0388, -0.0277], ..., [-0.0079, -0.0035, -0.0299, ..., 0.0157, 0.0133, 0.0091], [-0.0279, -0.0409, -0.0081, ..., -0.0057, 0.0051, 0.0199], [ 0.0020, -0.0143, -0.0029, ..., 0.0280, 0.0175, -0.0235]], device='cuda:0'), grad: tensor([[ 0.0032, 0.0082, 0.0032, ..., 0.0125, 0.0078, 0.0000], [ 0.0008, 0.0021, 0.0004, ..., 0.0027, 0.0031, 0.0000], [-0.0242, 0.0007, 0.0025, ..., 0.0046, 0.0020, 0.0000], ..., [ 0.0063, -0.0016, 0.0081, ..., -0.0039, -0.0333, 0.0000], [ 0.0065, -0.0195, -0.0086, ..., -0.0059, -0.0210, 0.0000], [-0.0145, -0.0126, -0.0169, ..., -0.0911, -0.0687, 0.0000]], device='cuda:0') Epoch 3, bias, value: tensor([ 0.0077, 0.0202, 0.0136, -0.0020, 0.0149, -0.0061, -0.0164, -0.0147, 0.0249, -0.0064], device='cuda:0'), grad: tensor([ 0.0165, -0.0090, -0.0070, 0.0492, 0.0646, -0.0028, -0.0120, -0.0011, -0.0295, -0.0690], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 2---------------------------------------------------- epoch 2, time 278.39, cls_loss 0.2513 cls_loss_mapping 0.4353 cls_loss_causal 1.7004 re_mapping 0.1454 re_causal 0.2390 /// teacc 94.84 lr 0.00010000 Epoch 4, weight, value: tensor([[-0.0348, -0.0109, 0.0455, ..., 0.0026, -0.0317, -0.0246], [-0.0302, -0.0395, 0.0067, ..., -0.0231, -0.0112, 0.0177], [ 0.0370, 0.0089, -0.0152, ..., -0.0285, -0.0405, -0.0277], ..., [-0.0074, -0.0020, -0.0323, ..., 0.0153, 0.0135, 0.0091], [-0.0308, -0.0436, -0.0095, ..., -0.0078, 0.0055, 0.0199], [ 0.0016, -0.0162, -0.0015, ..., 0.0302, 0.0183, -0.0235]], device='cuda:0'), grad: tensor([[ 1.3237e-03, -5.3749e-03, -1.2703e-02, ..., -7.3166e-03, 2.7561e-03, -2.6379e-03], [ 4.9171e-03, 7.1478e-04, 3.9625e-04, ..., 2.2221e-03, 4.5013e-03, 5.6595e-05], [-3.6255e-02, 3.2949e-04, -3.7861e-04, ..., -4.9829e-04, -3.9101e-03, 2.6420e-05], ..., [ 9.7122e-03, 2.9716e-03, 8.1558e-03, ..., 6.2218e-03, 1.6069e-03, 1.4963e-03], [ 3.3798e-03, 1.9779e-03, -4.4525e-05, ..., 3.2654e-03, 2.7008e-03, 6.1929e-05], [-1.4162e-04, -1.3313e-02, 2.3365e-03, ..., 3.2940e-03, 5.4016e-03, 3.2234e-04]], device='cuda:0') Epoch 4, bias, value: tensor([ 0.0080, 0.0203, 0.0137, -0.0024, 0.0150, -0.0061, -0.0166, -0.0148, 0.0247, -0.0060], device='cuda:0'), grad: tensor([-0.0065, 0.0036, -0.0241, 0.0354, -0.0230, 0.0006, 0.0029, 0.0122, 0.0023, -0.0033], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 3---------------------------------------------------- epoch 3, time 278.35, cls_loss 0.1958 cls_loss_mapping 0.3174 cls_loss_causal 1.4986 re_mapping 0.1158 re_causal 0.2061 /// teacc 95.47 lr 0.00010000 Epoch 5, weight, value: tensor([[-0.0366, -0.0104, 0.0495, ..., 0.0037, -0.0338, -0.0216], [-0.0309, -0.0406, 0.0027, ..., -0.0262, -0.0129, 0.0146], [ 0.0399, 0.0072, -0.0172, ..., -0.0291, -0.0429, -0.0341], ..., [-0.0074, -0.0010, -0.0350, ..., 0.0149, 0.0139, 0.0096], [-0.0323, -0.0443, -0.0099, ..., -0.0092, 0.0056, 0.0160], [ 0.0006, -0.0168, -0.0014, ..., 0.0315, 0.0185, -0.0248]], device='cuda:0'), grad: tensor([[ 5.5962e-03, 6.8486e-05, 1.6289e-03, ..., 5.3787e-04, 1.3046e-03, 4.5466e-04], [ 1.1168e-03, 5.2834e-04, 3.5739e-04, ..., 9.1553e-04, 1.8396e-03, 5.5552e-05], [-3.1204e-02, -6.1646e-03, -7.9727e-03, ..., -3.8795e-03, -3.2654e-03, 2.3782e-04], ..., [ 2.9316e-03, 6.1989e-04, 1.2541e-03, ..., 7.2136e-03, 1.1116e-02, 1.6138e-05], [ 5.4932e-03, 2.3537e-03, 2.2945e-03, ..., 5.1384e-03, 6.1378e-03, 3.2496e-04], [ 6.5470e-04, -3.0403e-03, -1.6451e-04, ..., -3.3173e-02, -5.7648e-02, 2.9624e-05]], device='cuda:0') Epoch 5, bias, value: tensor([ 0.0082, 0.0206, 0.0134, -0.0022, 0.0152, -0.0065, -0.0171, -0.0146, 0.0250, -0.0064], device='cuda:0'), grad: tensor([ 6.0921e-03, 6.5416e-06, -2.9526e-02, 1.8600e-02, 2.3712e-02, 9.3536e-03, -5.4054e-03, 9.9182e-03, 1.0880e-02, -4.3610e-02], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 4---------------------------------------------------- epoch 4, time 277.86, cls_loss 0.1580 cls_loss_mapping 0.2540 cls_loss_causal 1.4358 re_mapping 0.0939 re_causal 0.1804 /// teacc 96.88 lr 0.00010000 Epoch 6, weight, value: tensor([[-0.0371, -0.0106, 0.0526, ..., 0.0051, -0.0359, -0.0241], [-0.0316, -0.0421, 0.0021, ..., -0.0296, -0.0148, 0.0132], [ 0.0419, 0.0069, -0.0193, ..., -0.0297, -0.0445, -0.0408], ..., [-0.0070, -0.0005, -0.0375, ..., 0.0142, 0.0141, 0.0103], [-0.0341, -0.0450, -0.0113, ..., -0.0109, 0.0054, 0.0093], [-0.0003, -0.0177, -0.0017, ..., 0.0328, 0.0191, -0.0273]], device='cuda:0'), grad: tensor([[ 4.4107e-04, 2.5392e-04, 6.0539e-03, ..., 4.4365e-03, 3.5172e-03, 4.9829e-04], [ 1.0284e-02, 6.4075e-05, -1.1568e-03, ..., 1.1606e-03, 1.6823e-03, 1.9205e-04], [-1.8890e-02, 7.5340e-04, 4.4060e-03, ..., 4.4518e-03, 5.4245e-03, 7.7057e-04], ..., [ 1.7633e-03, -4.0680e-05, 1.3533e-03, ..., 4.3726e-04, -2.3251e-03, 3.4243e-05], [ 3.8528e-03, -4.7684e-04, -1.2901e-02, ..., -1.0254e-02, -1.4946e-02, -1.4949e-04], [ 2.5535e-04, 1.4877e-04, 2.3727e-03, ..., 4.8332e-03, 5.4779e-03, 1.2046e-04]], device='cuda:0') Epoch 6, bias, value: tensor([ 0.0083, 0.0203, 0.0133, -0.0024, 0.0153, -0.0067, -0.0168, -0.0141, 0.0249, -0.0064], device='cuda:0'), grad: tensor([ 0.0097, 0.0146, -0.0173, 0.0037, -0.0011, 0.0049, 0.0012, 0.0003, -0.0247, 0.0086], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 5---------------------------------------------------- epoch 5, time 278.31, cls_loss 0.1141 cls_loss_mapping 0.1882 cls_loss_causal 1.3510 re_mapping 0.0799 re_causal 0.1651 /// teacc 97.10 lr 0.00010000 Epoch 7, weight, value: tensor([[-3.8017e-02, -1.0378e-02, 5.4926e-02, ..., 6.2616e-03, -3.7454e-02, -2.8002e-02], [-3.1948e-02, -4.2724e-02, 2.8633e-05, ..., -3.2009e-02, -1.6452e-02, 1.2362e-02], [ 4.4339e-02, 5.8609e-03, -2.1181e-02, ..., -3.0784e-02, -4.6275e-02, -4.8904e-02], ..., [-7.3374e-03, -3.2995e-04, -3.8411e-02, ..., 1.3766e-02, 1.4295e-02, 1.2877e-02], [-3.6592e-02, -4.5429e-02, -1.0921e-02, ..., -1.2124e-02, 5.6063e-03, 5.9953e-03], [-6.4927e-04, -1.8044e-02, -2.7281e-03, ..., 3.3696e-02, 1.9460e-02, -2.9106e-02]], device='cuda:0'), grad: tensor([[ 1.0413e-04, 3.6061e-05, -2.1210e-03, ..., -6.5136e-04, 1.0719e-03, 1.0177e-05], [ 3.7932e-04, 7.9751e-05, -3.3784e-04, ..., 8.4877e-05, -9.9301e-05, -9.6142e-05], [ 3.6430e-04, 1.9574e-04, 1.1282e-03, ..., 6.9332e-04, 1.6088e-03, 9.3162e-05], ..., [ 6.6137e-04, 9.1434e-05, 5.6171e-04, ..., 1.2026e-03, 4.3559e-04, 2.3350e-05], [ 1.5039e-03, 1.6809e-04, 8.0884e-05, ..., 6.8474e-04, 7.5912e-04, 1.5819e-04], [-5.2309e-04, 4.4680e-04, -8.1062e-04, ..., -2.5253e-03, -1.1396e-03, 2.9951e-05]], device='cuda:0') Epoch 7, bias, value: tensor([ 0.0084, 0.0201, 0.0129, -0.0021, 0.0154, -0.0069, -0.0170, -0.0142, 0.0253, -0.0063], device='cuda:0'), grad: tensor([ 0.0002, -0.0031, 0.0030, 0.0030, 0.0009, -0.0077, 0.0006, 0.0011, 0.0027, -0.0007], device='cuda:0') 100 0.0001 changing lr epoch 6, time 261.34, cls_loss 0.1024 cls_loss_mapping 0.1556 cls_loss_causal 1.2240 re_mapping 0.0693 re_causal 0.1462 /// teacc 97.09 lr 0.00010000 Epoch 8, weight, value: tensor([[-3.9099e-02, -9.8635e-03, 5.6857e-02, ..., 7.2413e-03, -3.8952e-02, -2.9268e-02], [-3.2591e-02, -4.2773e-02, 2.7749e-04, ..., -3.3093e-02, -1.7628e-02, 1.0653e-02], [ 4.6330e-02, 4.7390e-03, -2.2939e-02, ..., -3.1849e-02, -4.7772e-02, -5.3329e-02], ..., [-8.2794e-03, -7.3237e-05, -4.0368e-02, ..., 1.3732e-02, 1.5140e-02, 1.3445e-02], [-3.8000e-02, -4.5898e-02, -1.0978e-02, ..., -1.3401e-02, 5.6859e-03, 2.0614e-03], [-1.0782e-03, -1.8746e-02, -2.6144e-03, ..., 3.4746e-02, 1.9653e-02, -3.0435e-02]], device='cuda:0'), grad: tensor([[ 2.1877e-03, 8.6248e-05, -1.1658e-02, ..., -1.8287e-04, 5.7077e-04, -1.3914e-03], [-3.0136e-02, 1.0991e-04, 7.0572e-04, ..., 5.9986e-04, 3.9196e-04, 5.3078e-05], [ 1.3916e-02, -4.7994e-04, 1.6079e-03, ..., 3.2091e-04, 7.0524e-04, 3.0041e-04], ..., [ 1.6842e-03, 9.1076e-05, 5.8031e-04, ..., 7.3385e-04, 7.0274e-05, 3.9458e-05], [ 2.7218e-03, 2.7037e-04, 2.6302e-03, ..., 1.0633e-03, 1.2598e-03, 3.5977e-04], [ 5.4407e-04, 2.8968e-04, 1.4687e-03, ..., 1.1871e-02, 1.3969e-02, 9.3043e-05]], device='cuda:0') Epoch 8, bias, value: tensor([ 0.0085, 0.0201, 0.0130, -0.0020, 0.0154, -0.0069, -0.0171, -0.0142, 0.0253, -0.0065], device='cuda:0'), grad: tensor([-0.0052, -0.0317, 0.0171, 0.0055, -0.0139, 0.0022, 0.0048, 0.0018, 0.0060, 0.0133], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 7---------------------------------------------------- epoch 7, time 278.54, cls_loss 0.1057 cls_loss_mapping 0.1583 cls_loss_causal 1.1747 re_mapping 0.0598 re_causal 0.1285 /// teacc 97.21 lr 0.00010000 Epoch 9, weight, value: tensor([[-0.0406, -0.0095, 0.0590, ..., 0.0081, -0.0404, -0.0299], [-0.0323, -0.0436, 0.0006, ..., -0.0354, -0.0193, 0.0095], [ 0.0482, 0.0039, -0.0247, ..., -0.0325, -0.0494, -0.0596], ..., [-0.0087, 0.0006, -0.0414, ..., 0.0133, 0.0157, 0.0181], [-0.0390, -0.0468, -0.0114, ..., -0.0143, 0.0059, -0.0013], [-0.0018, -0.0192, -0.0028, ..., 0.0356, 0.0198, -0.0341]], device='cuda:0'), grad: tensor([[ 5.8794e-04, 7.5912e-04, -1.3614e-04, ..., -2.4900e-05, 1.1492e-03, 7.3862e-04], [ 1.2827e-03, 1.2815e-04, 5.6744e-04, ..., 1.8561e-04, 5.0449e-04, 2.8953e-05], [-5.6496e-03, 2.6608e-04, -2.0275e-03, ..., 1.8549e-04, 7.8392e-04, 9.2387e-05], ..., [ 1.0729e-03, 2.4045e-04, 2.0206e-04, ..., 1.8377e-03, 3.2005e-03, 3.9227e-06], [ 4.5166e-03, 2.2964e-03, 7.8440e-04, ..., 1.4420e-03, 7.7400e-03, 1.1414e-04], [ 3.5763e-04, 4.3702e-04, 4.1962e-04, ..., 1.7462e-03, 2.3899e-03, 9.5591e-06]], device='cuda:0') Epoch 9, bias, value: tensor([ 0.0085, 0.0200, 0.0130, -0.0020, 0.0157, -0.0071, -0.0174, -0.0139, 0.0256, -0.0067], device='cuda:0'), grad: tensor([ 0.0024, 0.0008, -0.0057, -0.0112, -0.0064, -0.0063, -0.0002, 0.0040, 0.0181, 0.0044], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 8---------------------------------------------------- epoch 8, time 277.99, cls_loss 0.0795 cls_loss_mapping 0.1290 cls_loss_causal 1.1320 re_mapping 0.0545 re_causal 0.1243 /// teacc 97.73 lr 0.00010000 Epoch 10, weight, value: tensor([[-4.1535e-02, -9.4166e-03, 6.0846e-02, ..., 9.1568e-03, -4.1750e-02, -3.0630e-02], [-3.3310e-02, -4.4122e-02, 4.3387e-07, ..., -3.6372e-02, -2.0060e-02, 1.0328e-02], [ 4.9945e-02, 3.2577e-03, -2.6320e-02, ..., -3.3269e-02, -5.0451e-02, -6.3274e-02], ..., [-9.1274e-03, 8.3013e-04, -4.2307e-02, ..., 1.2600e-02, 1.5928e-02, 1.9006e-02], [-4.0328e-02, -4.7789e-02, -1.1606e-02, ..., -1.5415e-02, 5.8589e-03, -3.0546e-03], [-2.0778e-03, -1.9355e-02, -3.0932e-03, ..., 3.6396e-02, 1.9966e-02, -3.5436e-02]], device='cuda:0'), grad: tensor([[ 1.4677e-03, -1.9681e-04, -1.6832e-04, ..., 5.5122e-04, 5.0974e-04, 1.7202e-04], [ 2.7485e-03, 1.1212e-04, 4.4012e-04, ..., 7.2813e-04, 2.2495e-04, 3.8552e-04], [ 3.2482e-03, 2.2757e-04, 1.0004e-03, ..., 4.3416e-04, 1.1978e-03, 1.2827e-04], ..., [ 3.0422e-03, 3.4189e-04, 2.8586e-04, ..., 1.9875e-03, 1.5259e-03, 8.3745e-06], [ 1.1578e-03, 5.9366e-04, -6.7997e-04, ..., 1.7595e-04, 2.3308e-03, 4.9973e-04], [ 3.0828e-04, 4.3440e-04, 2.8706e-04, ..., -4.2000e-03, -3.5553e-03, 2.5362e-05]], device='cuda:0') Epoch 10, bias, value: tensor([ 0.0086, 0.0199, 0.0132, -0.0021, 0.0161, -0.0068, -0.0179, -0.0140, 0.0254, -0.0068], device='cuda:0'), grad: tensor([ 0.0018, 0.0026, 0.0091, -0.0127, -0.0003, -0.0030, -0.0043, 0.0060, 0.0044, -0.0034], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 9---------------------------------------------------- epoch 9, time 278.37, cls_loss 0.0743 cls_loss_mapping 0.1175 cls_loss_causal 1.1044 re_mapping 0.0474 re_causal 0.1133 /// teacc 97.74 lr 0.00010000 Epoch 11, weight, value: tensor([[-0.0427, -0.0095, 0.0623, ..., 0.0102, -0.0433, -0.0320], [-0.0329, -0.0444, 0.0002, ..., -0.0385, -0.0215, 0.0090], [ 0.0513, 0.0026, -0.0286, ..., -0.0345, -0.0516, -0.0673], ..., [-0.0098, 0.0016, -0.0438, ..., 0.0120, 0.0166, 0.0189], [-0.0412, -0.0486, -0.0122, ..., -0.0165, 0.0059, -0.0073], [-0.0024, -0.0201, -0.0031, ..., 0.0375, 0.0201, -0.0384]], device='cuda:0'), grad: tensor([[ 9.6917e-05, 3.4928e-05, -3.9721e-04, ..., -1.0210e-04, 2.5249e-04, 1.1760e-04], [ 2.7204e-04, 1.9777e-04, -1.1319e-04, ..., 8.3268e-05, 3.0732e-04, -9.3639e-05], [-7.8011e-04, 4.4274e-04, 1.1444e-04, ..., 3.2282e-04, 6.3086e-04, 5.2482e-05], ..., [ 5.2595e-04, 8.8215e-04, 4.8727e-05, ..., 7.7629e-04, 5.5742e-04, 1.8343e-05], [ 8.2493e-04, 1.6375e-03, 1.3483e-04, ..., 1.0996e-03, 1.4086e-03, 1.3053e-04], [ 7.4863e-04, 1.7424e-03, -1.9923e-05, ..., 2.5368e-04, 8.8930e-04, 1.8060e-05]], device='cuda:0') Epoch 11, bias, value: tensor([ 0.0086, 0.0199, 0.0129, -0.0022, 0.0162, -0.0071, -0.0175, -0.0136, 0.0255, -0.0070], device='cuda:0'), grad: tensor([ 0.0002, -0.0015, 0.0008, -0.0086, 0.0011, -0.0040, 0.0048, 0.0016, 0.0034, 0.0024], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 10---------------------------------------------------- epoch 10, time 278.08, cls_loss 0.0635 cls_loss_mapping 0.0979 cls_loss_causal 1.0332 re_mapping 0.0441 re_causal 0.1056 /// teacc 98.03 lr 0.00010000 Epoch 12, weight, value: tensor([[-0.0436, -0.0095, 0.0641, ..., 0.0113, -0.0443, -0.0322], [-0.0339, -0.0451, 0.0005, ..., -0.0406, -0.0227, 0.0110], [ 0.0527, 0.0015, -0.0296, ..., -0.0350, -0.0527, -0.0709], ..., [-0.0098, 0.0019, -0.0448, ..., 0.0112, 0.0166, 0.0204], [-0.0422, -0.0491, -0.0125, ..., -0.0172, 0.0063, -0.0073], [-0.0030, -0.0206, -0.0040, ..., 0.0381, 0.0202, -0.0410]], device='cuda:0'), grad: tensor([[ 4.1556e-04, -3.6597e-05, -1.6117e-03, ..., -8.3685e-04, 4.4894e-04, 7.0512e-05], [ 5.7840e-04, 6.2466e-05, 1.1551e-04, ..., 3.1137e-04, 6.4898e-04, 1.5736e-05], [-3.5229e-03, 1.0532e-04, 1.5974e-04, ..., 6.2275e-04, -5.0592e-04, 6.9886e-06], ..., [ 8.4019e-04, 4.0889e-05, 2.3222e-04, ..., 1.8370e-04, -1.1158e-03, 1.8388e-05], [ 1.0710e-03, 1.6105e-04, 2.1911e-04, ..., 2.5868e-04, 2.3985e-04, 8.0109e-05], [ 2.2626e-04, 1.4663e-04, 1.1533e-04, ..., 1.3771e-03, 2.9659e-03, 3.3140e-05]], device='cuda:0') Epoch 12, bias, value: tensor([ 0.0088, 0.0198, 0.0132, -0.0021, 0.0161, -0.0070, -0.0178, -0.0139, 0.0257, -0.0072], device='cuda:0'), grad: tensor([-3.6538e-05, 9.8801e-04, -4.3449e-03, 1.6727e-03, -2.3022e-03, -3.0384e-03, 1.3638e-03, 1.5247e-04, 1.5469e-03, 4.0016e-03], device='cuda:0') 100 0.0001 changing lr epoch 11, time 261.47, cls_loss 0.0624 cls_loss_mapping 0.0958 cls_loss_causal 1.0565 re_mapping 0.0389 re_causal 0.0994 /// teacc 97.79 lr 0.00010000 Epoch 13, weight, value: tensor([[-0.0440, -0.0096, 0.0652, ..., 0.0121, -0.0456, -0.0344], [-0.0341, -0.0454, -0.0007, ..., -0.0415, -0.0237, 0.0070], [ 0.0539, 0.0012, -0.0313, ..., -0.0358, -0.0536, -0.0745], ..., [-0.0098, 0.0021, -0.0458, ..., 0.0105, 0.0171, 0.0210], [-0.0433, -0.0496, -0.0128, ..., -0.0182, 0.0064, -0.0087], [-0.0037, -0.0210, -0.0038, ..., 0.0390, 0.0205, -0.0414]], device='cuda:0'), grad: tensor([[ 4.0030e-04, 8.2195e-05, -1.4377e-04, ..., 3.1376e-04, 6.1464e-04, 1.6844e-04], [ 9.2030e-05, 3.9756e-05, -2.5349e-03, ..., 1.7917e-04, 5.6458e-04, -1.2302e-03], [-1.4591e-03, -8.8692e-05, 2.9683e-04, ..., -2.8778e-06, 3.1781e-04, 1.0830e-04], ..., [ 2.0719e-04, 8.3566e-05, -3.7789e-04, ..., -5.7373e-03, -9.7427e-03, 3.3259e-05], [ 2.9826e-04, 1.6093e-04, 3.8552e-04, ..., 6.3276e-04, 3.8552e-04, 3.4690e-04], [ 2.2733e-04, 1.9467e-04, 4.4298e-04, ..., 2.5444e-03, 6.6986e-03, 4.7654e-05]], device='cuda:0') Epoch 13, bias, value: tensor([ 0.0088, 0.0193, 0.0135, -0.0023, 0.0161, -0.0072, -0.0175, -0.0137, 0.0259, -0.0072], device='cuda:0'), grad: tensor([ 1.3981e-03, -4.5013e-03, -5.0366e-05, 4.0245e-04, 2.1725e-03, -6.3753e-04, 3.2921e-03, -1.2642e-02, 1.6241e-03, 8.9493e-03], device='cuda:0') 100 0.0001 changing lr epoch 12, time 261.50, cls_loss 0.0551 cls_loss_mapping 0.0888 cls_loss_causal 0.9801 re_mapping 0.0372 re_causal 0.0957 /// teacc 98.02 lr 0.00010000 Epoch 14, weight, value: tensor([[-0.0443, -0.0095, 0.0664, ..., 0.0132, -0.0467, -0.0357], [-0.0349, -0.0454, -0.0004, ..., -0.0420, -0.0250, 0.0076], [ 0.0552, 0.0005, -0.0323, ..., -0.0368, -0.0546, -0.0768], ..., [-0.0097, 0.0026, -0.0470, ..., 0.0097, 0.0179, 0.0199], [-0.0442, -0.0501, -0.0128, ..., -0.0195, 0.0066, -0.0099], [-0.0043, -0.0215, -0.0041, ..., 0.0399, 0.0207, -0.0432]], device='cuda:0'), grad: tensor([[ 3.6001e-04, -3.1042e-04, -1.2617e-03, ..., -5.1737e-04, 1.0687e-04, 8.9705e-05], [ 1.8895e-04, 1.1486e-04, -7.5400e-05, ..., -3.4595e-04, 3.3140e-04, -3.0065e-04], [-3.4389e-03, 5.5552e-05, 3.3021e-04, ..., 2.0432e-04, 2.5415e-04, 1.7834e-04], ..., [ 6.9571e-04, -3.4779e-05, 9.2268e-05, ..., 8.6069e-05, -9.0027e-04, 6.7770e-05], [ 7.0763e-04, 1.7381e-04, 1.1265e-04, ..., 1.9109e-04, 3.6502e-04, 1.5438e-04], [ 2.4986e-04, 1.5938e-04, 1.8418e-04, ..., 3.6073e-04, 7.3242e-04, 2.7269e-05]], device='cuda:0') Epoch 14, bias, value: tensor([ 0.0089, 0.0196, 0.0132, -0.0024, 0.0160, -0.0072, -0.0178, -0.0134, 0.0259, -0.0072], device='cuda:0'), grad: tensor([-0.0003, -0.0010, -0.0015, -0.0215, 0.0004, 0.0201, 0.0006, -0.0001, 0.0016, 0.0016], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 13---------------------------------------------------- epoch 13, time 277.69, cls_loss 0.0544 cls_loss_mapping 0.0889 cls_loss_causal 0.9948 re_mapping 0.0349 re_causal 0.0915 /// teacc 98.39 lr 0.00010000 Epoch 15, weight, value: tensor([[-0.0453, -0.0097, 0.0675, ..., 0.0136, -0.0478, -0.0372], [-0.0352, -0.0457, 0.0002, ..., -0.0425, -0.0259, 0.0062], [ 0.0565, -0.0003, -0.0341, ..., -0.0375, -0.0554, -0.0794], ..., [-0.0104, 0.0026, -0.0483, ..., 0.0091, 0.0184, 0.0198], [-0.0454, -0.0504, -0.0125, ..., -0.0198, 0.0068, -0.0111], [-0.0049, -0.0217, -0.0042, ..., 0.0408, 0.0209, -0.0434]], device='cuda:0'), grad: tensor([[-4.3201e-04, -6.2609e-04, -4.6959e-03, ..., -1.4410e-03, 8.6427e-05, 1.4946e-05], [ 1.6689e-05, 5.8338e-06, -2.2709e-04, ..., 1.4818e-04, 3.1734e-04, 3.0939e-06], [ 3.0065e-04, 4.7743e-05, 1.2465e-03, ..., 1.0471e-03, 4.9400e-04, 2.3559e-05], ..., [ 1.0170e-05, 1.5236e-05, 9.4354e-05, ..., 7.2289e-04, 2.7251e-04, 4.2189e-07], [ 1.3280e-04, 1.4603e-04, 5.8794e-04, ..., 1.1663e-03, 1.7929e-03, 1.3083e-05], [ 1.1361e-04, 8.7440e-05, 4.3488e-04, ..., 4.9171e-03, 5.4398e-03, 2.3935e-06]], device='cuda:0') Epoch 15, bias, value: tensor([ 0.0089, 0.0194, 0.0131, -0.0023, 0.0156, -0.0069, -0.0179, -0.0133, 0.0260, -0.0071], device='cuda:0'), grad: tensor([-0.0031, -0.0012, 0.0020, 0.0002, -0.0103, 0.0002, 0.0014, 0.0008, 0.0031, 0.0070], device='cuda:0') 100 0.0001 changing lr epoch 14, time 261.51, cls_loss 0.0468 cls_loss_mapping 0.0780 cls_loss_causal 0.9790 re_mapping 0.0337 re_causal 0.0910 /// teacc 98.19 lr 0.00010000 Epoch 16, weight, value: tensor([[-0.0461, -0.0094, 0.0684, ..., 0.0140, -0.0490, -0.0383], [-0.0356, -0.0462, 0.0001, ..., -0.0432, -0.0264, 0.0071], [ 0.0580, -0.0005, -0.0350, ..., -0.0381, -0.0565, -0.0815], ..., [-0.0113, 0.0020, -0.0492, ..., 0.0085, 0.0186, 0.0188], [-0.0460, -0.0505, -0.0123, ..., -0.0202, 0.0070, -0.0118], [-0.0054, -0.0224, -0.0050, ..., 0.0412, 0.0208, -0.0452]], device='cuda:0'), grad: tensor([[ 3.7551e-05, -1.6391e-04, -1.1444e-03, ..., 3.4362e-05, 2.5535e-04, 3.3528e-05], [ 8.3637e-04, 3.2759e-04, 9.6083e-05, ..., 3.2210e-04, 9.4128e-04, 1.5283e-04], [-1.0300e-03, 6.8998e-04, 2.7156e-04, ..., 5.3406e-04, 1.3151e-03, 4.6015e-05], ..., [ 7.7069e-05, -2.5558e-03, 4.6402e-05, ..., -3.4714e-03, -7.3357e-03, 2.9430e-06], [ 6.3658e-05, 5.4169e-04, 2.6655e-04, ..., 5.7888e-04, 1.2693e-03, 2.2066e-04], [ 1.8597e-05, 1.2863e-04, 5.6267e-05, ..., -6.3181e-04, -4.5633e-04, 1.0006e-05]], device='cuda:0') Epoch 16, bias, value: tensor([ 0.0087, 0.0196, 0.0131, -0.0020, 0.0160, -0.0073, -0.0178, -0.0135, 0.0264, -0.0076], device='cuda:0'), grad: tensor([-2.8801e-04, 2.3460e-03, 2.4471e-03, 7.8773e-04, 7.5912e-03, 1.1139e-03, -5.8899e-03, -1.0933e-02, 2.9011e-03, -6.8843e-05], device='cuda:0') 100 0.0001 changing lr epoch 15, time 260.58, cls_loss 0.0425 cls_loss_mapping 0.0693 cls_loss_causal 0.9267 re_mapping 0.0306 re_causal 0.0850 /// teacc 98.37 lr 0.00010000 Epoch 17, weight, value: tensor([[-0.0468, -0.0093, 0.0696, ..., 0.0147, -0.0498, -0.0391], [-0.0361, -0.0466, -0.0004, ..., -0.0444, -0.0272, 0.0057], [ 0.0595, -0.0010, -0.0364, ..., -0.0387, -0.0573, -0.0835], ..., [-0.0117, 0.0024, -0.0498, ..., 0.0076, 0.0191, 0.0185], [-0.0464, -0.0507, -0.0124, ..., -0.0213, 0.0070, -0.0129], [-0.0054, -0.0225, -0.0054, ..., 0.0421, 0.0210, -0.0461]], device='cuda:0'), grad: tensor([[ 1.8132e-04, 8.6874e-06, -1.4043e-04, ..., -1.3340e-04, 1.1462e-04, 3.3307e-04], [ 2.3556e-03, 1.5102e-05, 2.6703e-03, ..., 2.7955e-05, 2.2447e-04, 5.5313e-03], [ 6.7663e-04, 1.0818e-04, 5.2977e-04, ..., 1.0782e-04, 1.4677e-03, 1.0271e-03], ..., [-6.6090e-04, 1.8001e-05, 2.7850e-05, ..., -5.7846e-05, -1.7281e-03, -1.4454e-05], [ 1.0405e-03, 2.3440e-05, 1.0023e-03, ..., 7.9155e-05, 3.7372e-05, 2.2335e-03], [ 1.5259e-04, 4.8488e-05, 9.5308e-05, ..., 1.6725e-04, 4.5037e-04, 4.7117e-05]], device='cuda:0') Epoch 17, bias, value: tensor([ 0.0088, 0.0195, 0.0131, -0.0025, 0.0162, -0.0071, -0.0177, -0.0133, 0.0262, -0.0075], device='cuda:0'), grad: tensor([ 0.0005, 0.0099, 0.0044, 0.0003, -0.0012, 0.0012, -0.0174, -0.0025, 0.0040, 0.0009], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 16---------------------------------------------------- epoch 16, time 275.63, cls_loss 0.0409 cls_loss_mapping 0.0722 cls_loss_causal 0.9112 re_mapping 0.0299 re_causal 0.0832 /// teacc 98.40 lr 0.00010000 Epoch 18, weight, value: tensor([[-0.0474, -0.0092, 0.0705, ..., 0.0148, -0.0507, -0.0400], [-0.0360, -0.0468, -0.0008, ..., -0.0456, -0.0283, 0.0043], [ 0.0601, -0.0015, -0.0374, ..., -0.0394, -0.0580, -0.0864], ..., [-0.0120, 0.0022, -0.0505, ..., 0.0075, 0.0196, 0.0171], [-0.0471, -0.0508, -0.0122, ..., -0.0226, 0.0067, -0.0150], [-0.0057, -0.0230, -0.0051, ..., 0.0425, 0.0210, -0.0478]], device='cuda:0'), grad: tensor([[ 1.0639e-04, 9.6485e-06, 1.2255e-04, ..., 8.1539e-04, 8.5878e-04, 8.5652e-05], [ 2.6345e-04, 2.2128e-05, 1.2118e-04, ..., 2.3615e-04, 2.5225e-04, 1.2040e-05], [-8.9264e-04, 1.1361e-04, 2.4986e-04, ..., 5.3167e-05, 3.1161e-04, 4.8466e-06], ..., [ 2.2197e-04, 5.8264e-05, 4.6879e-05, ..., -2.3975e-03, -2.0580e-03, -1.8919e-04], [ 1.9109e-04, 5.5462e-05, -2.0885e-03, ..., -3.9268e-04, -1.2131e-03, 6.1095e-05], [ 6.4790e-05, 8.7142e-05, 2.5940e-04, ..., 1.2980e-03, 1.0681e-03, 1.0526e-04]], device='cuda:0') Epoch 18, bias, value: tensor([ 0.0087, 0.0195, 0.0127, -0.0021, 0.0162, -0.0071, -0.0178, -0.0131, 0.0259, -0.0074], device='cuda:0'), grad: tensor([ 1.9703e-03, 7.6008e-04, 2.6393e-04, -6.9320e-05, 1.6761e-04, 2.0561e-03, 2.0754e-04, -4.3411e-03, -4.0703e-03, 3.0537e-03], device='cuda:0') 100 0.0001 changing lr epoch 17, time 254.21, cls_loss 0.0326 cls_loss_mapping 0.0553 cls_loss_causal 0.9035 re_mapping 0.0290 re_causal 0.0834 /// teacc 98.35 lr 0.00010000 Epoch 19, weight, value: tensor([[-0.0481, -0.0092, 0.0715, ..., 0.0155, -0.0514, -0.0405], [-0.0368, -0.0471, -0.0015, ..., -0.0460, -0.0291, 0.0032], [ 0.0612, -0.0023, -0.0380, ..., -0.0399, -0.0588, -0.0876], ..., [-0.0123, 0.0028, -0.0511, ..., 0.0073, 0.0202, 0.0177], [-0.0479, -0.0511, -0.0119, ..., -0.0236, 0.0067, -0.0153], [-0.0063, -0.0237, -0.0058, ..., 0.0428, 0.0211, -0.0485]], device='cuda:0'), grad: tensor([[ 7.7128e-05, 1.0349e-05, -1.1765e-02, ..., 1.4019e-04, -3.6201e-03, -6.1073e-03], [-4.2458e-03, -2.5606e-04, -1.5097e-03, ..., -1.7703e-04, 4.5609e-04, -5.0735e-04], [ 2.4261e-03, 7.5817e-05, 3.3188e-04, ..., -1.1215e-03, -6.1035e-04, 4.6206e-04], ..., [ 4.4084e-04, 2.8789e-05, 4.9448e-04, ..., 2.1057e-03, 1.8988e-03, 5.1618e-05], [ 6.9332e-04, 1.3161e-04, 5.9557e-04, ..., 1.1768e-03, 1.4591e-03, 1.8322e-04], [ 4.9323e-05, 1.9774e-05, 2.2423e-04, ..., -3.1757e-03, -3.5381e-03, 5.5224e-05]], device='cuda:0') Epoch 19, bias, value: tensor([ 0.0087, 0.0191, 0.0128, -0.0024, 0.0164, -0.0067, -0.0176, -0.0128, 0.0259, -0.0076], device='cuda:0'), grad: tensor([-0.0099, -0.0070, -0.0102, 0.0018, 0.0011, 0.0104, 0.0029, 0.0066, 0.0073, -0.0029], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 18---------------------------------------------------- epoch 18, time 268.36, cls_loss 0.0291 cls_loss_mapping 0.0543 cls_loss_causal 0.8886 re_mapping 0.0271 re_causal 0.0799 /// teacc 98.44 lr 0.00010000 Epoch 20, weight, value: tensor([[-0.0486, -0.0093, 0.0723, ..., 0.0157, -0.0521, -0.0409], [-0.0372, -0.0474, -0.0015, ..., -0.0466, -0.0297, 0.0031], [ 0.0622, -0.0027, -0.0391, ..., -0.0402, -0.0594, -0.0889], ..., [-0.0126, 0.0028, -0.0518, ..., 0.0067, 0.0204, 0.0175], [-0.0482, -0.0513, -0.0119, ..., -0.0243, 0.0067, -0.0154], [-0.0067, -0.0239, -0.0060, ..., 0.0434, 0.0214, -0.0489]], device='cuda:0'), grad: tensor([[ 2.2888e-05, 4.6402e-05, -1.9193e-04, ..., -2.4110e-05, 1.3828e-04, 1.7300e-05], [ 1.3840e-04, 2.2447e-04, 1.8045e-05, ..., 1.4758e-04, 3.5596e-04, 1.2904e-05], [-9.3102e-05, 2.3949e-04, 6.0618e-05, ..., 2.5344e-04, 4.5633e-04, 1.8805e-05], ..., [ 8.1122e-05, -8.7440e-05, 2.4021e-05, ..., 8.2169e-03, 6.8359e-03, 2.4512e-06], [-4.9353e-05, 4.9019e-04, 6.0461e-06, ..., 4.5371e-04, 1.0157e-03, -3.0905e-05], [ 7.7188e-05, -5.5647e-04, 4.3422e-05, ..., -9.1248e-03, -8.7433e-03, 7.5251e-06]], device='cuda:0') Epoch 20, bias, value: tensor([ 0.0088, 0.0189, 0.0127, -0.0024, 0.0162, -0.0068, -0.0175, -0.0128, 0.0262, -0.0076], device='cuda:0'), grad: tensor([ 1.9145e-04, 7.4339e-04, 9.8705e-04, -2.6512e-04, 8.2672e-05, -6.5088e-04, 6.3181e-04, 4.5624e-03, 2.0599e-03, -8.3389e-03], device='cuda:0') 100 0.0001 changing lr epoch 19, time 251.02, cls_loss 0.0297 cls_loss_mapping 0.0518 cls_loss_causal 0.8685 re_mapping 0.0260 re_causal 0.0770 /// teacc 98.44 lr 0.00010000 Epoch 21, weight, value: tensor([[-0.0491, -0.0094, 0.0729, ..., 0.0157, -0.0530, -0.0420], [-0.0379, -0.0475, -0.0016, ..., -0.0465, -0.0302, 0.0020], [ 0.0634, -0.0033, -0.0401, ..., -0.0409, -0.0598, -0.0904], ..., [-0.0129, 0.0027, -0.0527, ..., 0.0059, 0.0206, 0.0177], [-0.0487, -0.0518, -0.0120, ..., -0.0246, 0.0069, -0.0158], [-0.0074, -0.0241, -0.0052, ..., 0.0442, 0.0213, -0.0496]], device='cuda:0'), grad: tensor([[-1.9699e-05, -1.9610e-04, -6.4611e-04, ..., 6.0380e-05, 2.1553e-04, 1.2130e-04], [ 3.4720e-05, 2.9191e-05, -1.0081e-05, ..., 8.7798e-05, -1.0364e-05, 2.0400e-05], [ 8.8573e-05, 1.3876e-04, 3.8934e-04, ..., 4.6039e-04, 4.7326e-04, 2.1651e-05], ..., [-2.2840e-04, -1.0371e-04, 1.3542e-04, ..., 2.1243e-04, -3.4118e-04, 1.1936e-05], [-7.8619e-05, -4.9114e-04, 5.8270e-04, ..., 8.7500e-04, 3.6979e-04, 1.5926e-04], [ 2.1219e-05, 6.4790e-05, 1.1349e-03, ..., 1.6603e-03, 7.7486e-04, 6.1095e-05]], device='cuda:0') Epoch 21, bias, value: tensor([ 0.0087, 0.0191, 0.0127, -0.0027, 0.0162, -0.0063, -0.0176, -0.0129, 0.0261, -0.0078], device='cuda:0'), grad: tensor([-1.7479e-05, -2.0962e-03, 1.3027e-03, 2.0733e-03, -4.7531e-03, -1.2169e-03, 1.1215e-03, -3.4022e-04, 1.2465e-03, 2.6817e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 20---------------------------------------------------- epoch 20, time 267.63, cls_loss 0.0279 cls_loss_mapping 0.0494 cls_loss_causal 0.8384 re_mapping 0.0250 re_causal 0.0735 /// teacc 98.66 lr 0.00010000 Epoch 22, weight, value: tensor([[-0.0504, -0.0094, 0.0737, ..., 0.0162, -0.0538, -0.0427], [-0.0383, -0.0478, -0.0019, ..., -0.0470, -0.0308, 0.0013], [ 0.0645, -0.0038, -0.0406, ..., -0.0410, -0.0601, -0.0925], ..., [-0.0134, 0.0030, -0.0529, ..., 0.0054, 0.0207, 0.0179], [-0.0490, -0.0520, -0.0122, ..., -0.0252, 0.0068, -0.0158], [-0.0076, -0.0244, -0.0054, ..., 0.0447, 0.0214, -0.0501]], device='cuda:0'), grad: tensor([[ 6.3181e-05, -1.5765e-05, -1.1873e-03, ..., -4.1080e-04, 1.2338e-04, 4.6343e-05], [ 1.5104e-04, 3.0071e-05, 3.5316e-05, ..., 1.1951e-04, 1.2791e-04, 1.5885e-05], [-9.6464e-04, 3.9697e-05, 1.6725e-04, ..., 1.5330e-04, 1.0496e-04, 1.5020e-05], ..., [ 2.4056e-04, 2.8715e-05, 4.4852e-05, ..., 7.6437e-04, 5.5981e-04, 6.4187e-06], [ 3.4261e-04, 1.8132e-04, 9.6321e-05, ..., 4.8828e-04, 4.9973e-04, 8.8513e-05], [ 1.9491e-05, 2.8789e-05, 1.2362e-04, ..., -4.9438e-03, -3.5496e-03, 1.4216e-05]], device='cuda:0') Epoch 22, bias, value: tensor([ 0.0085, 0.0192, 0.0131, -0.0025, 0.0162, -0.0067, -0.0174, -0.0128, 0.0260, -0.0078], device='cuda:0'), grad: tensor([-5.8794e-04, 4.0531e-04, -5.6219e-04, 5.6177e-05, 3.8719e-03, 1.3483e-04, 4.5609e-04, 1.3571e-03, 1.2903e-03, -6.4201e-03], device='cuda:0') 100 0.0001 changing lr epoch 21, time 251.05, cls_loss 0.0314 cls_loss_mapping 0.0568 cls_loss_causal 0.8744 re_mapping 0.0236 re_causal 0.0714 /// teacc 98.64 lr 0.00010000 Epoch 23, weight, value: tensor([[-0.0510, -0.0091, 0.0747, ..., 0.0167, -0.0544, -0.0440], [-0.0389, -0.0482, -0.0022, ..., -0.0469, -0.0312, -0.0017], [ 0.0659, -0.0042, -0.0417, ..., -0.0413, -0.0607, -0.0938], ..., [-0.0137, 0.0025, -0.0533, ..., 0.0048, 0.0209, 0.0180], [-0.0500, -0.0523, -0.0122, ..., -0.0261, 0.0070, -0.0158], [-0.0081, -0.0250, -0.0060, ..., 0.0455, 0.0216, -0.0511]], device='cuda:0'), grad: tensor([[ 6.2644e-05, 7.7343e-04, 1.2732e-03, ..., 7.0858e-04, 2.8038e-04, 7.4983e-05], [ 1.8522e-05, 5.2750e-05, 7.8201e-05, ..., 8.8334e-05, 9.2328e-05, 2.1830e-05], [-1.0071e-03, -1.3009e-05, 2.1362e-04, ..., -2.1005e-04, 1.4436e-04, 2.3440e-05], ..., [ 8.6010e-05, 2.8896e-04, 4.7016e-04, ..., 3.7456e-04, 1.2696e-04, 3.7085e-06], [ 1.1486e-04, 1.4954e-03, 2.5177e-03, ..., 1.7939e-03, 9.2316e-04, 1.5581e-04], [ 1.0395e-04, 2.4092e-04, 3.8290e-04, ..., 8.0204e-04, 5.6839e-04, 1.8895e-05]], device='cuda:0') Epoch 23, bias, value: tensor([ 0.0085, 0.0187, 0.0135, -0.0027, 0.0161, -0.0068, -0.0172, -0.0127, 0.0261, -0.0079], device='cuda:0'), grad: tensor([ 0.0031, 0.0002, -0.0002, 0.0036, -0.0015, -0.0172, 0.0026, 0.0013, 0.0065, 0.0017], device='cuda:0') 100 0.0001 changing lr epoch 22, time 251.55, cls_loss 0.0260 cls_loss_mapping 0.0504 cls_loss_causal 0.8111 re_mapping 0.0238 re_causal 0.0691 /// teacc 98.43 lr 0.00010000 Epoch 24, weight, value: tensor([[-0.0519, -0.0094, 0.0751, ..., 0.0169, -0.0550, -0.0447], [-0.0393, -0.0486, -0.0016, ..., -0.0475, -0.0319, -0.0009], [ 0.0672, -0.0044, -0.0426, ..., -0.0413, -0.0610, -0.0971], ..., [-0.0142, 0.0023, -0.0539, ..., 0.0045, 0.0212, 0.0175], [-0.0503, -0.0523, -0.0125, ..., -0.0269, 0.0072, -0.0162], [-0.0084, -0.0254, -0.0062, ..., 0.0461, 0.0216, -0.0516]], device='cuda:0'), grad: tensor([[-2.4602e-05, -3.3528e-05, -1.5247e-04, ..., 3.1143e-05, 1.4913e-04, 6.7532e-05], [-4.7088e-06, 2.8491e-05, -7.4565e-05, ..., 1.5393e-05, 1.0884e-04, -7.0274e-05], [ 5.8746e-04, 3.4833e-04, 7.4327e-05, ..., 3.6216e-04, 4.7421e-04, 4.6700e-05], ..., [ 5.9932e-05, 1.6367e-04, 1.5631e-05, ..., 5.4312e-04, 4.8304e-04, -3.7819e-05], [ 2.3949e-04, 4.4608e-04, 1.9681e-04, ..., 3.6788e-04, 1.0042e-03, 3.8600e-04], [ 7.9095e-05, 1.5068e-04, 7.0989e-05, ..., 3.8218e-04, 4.6206e-04, 6.2823e-05]], device='cuda:0') Epoch 24, bias, value: tensor([ 0.0082, 0.0186, 0.0137, -0.0029, 0.0161, -0.0065, -0.0172, -0.0125, 0.0265, -0.0082], device='cuda:0'), grad: tensor([ 0.0004, -0.0003, 0.0028, 0.0023, 0.0020, -0.0147, -0.0008, 0.0028, 0.0037, 0.0018], device='cuda:0') 100 0.0001 changing lr epoch 23, time 251.25, cls_loss 0.0266 cls_loss_mapping 0.0545 cls_loss_causal 0.8280 re_mapping 0.0223 re_causal 0.0669 /// teacc 98.63 lr 0.00010000 Epoch 25, weight, value: tensor([[-0.0528, -0.0094, 0.0760, ..., 0.0177, -0.0555, -0.0452], [-0.0395, -0.0487, -0.0014, ..., -0.0479, -0.0324, -0.0013], [ 0.0678, -0.0054, -0.0434, ..., -0.0419, -0.0620, -0.0968], ..., [-0.0139, 0.0028, -0.0546, ..., 0.0041, 0.0217, 0.0179], [-0.0511, -0.0527, -0.0122, ..., -0.0279, 0.0072, -0.0174], [-0.0090, -0.0259, -0.0070, ..., 0.0465, 0.0214, -0.0537]], device='cuda:0'), grad: tensor([[ 2.0102e-05, 3.2485e-05, 1.1069e-04, ..., 1.1677e-04, 1.0437e-04, 1.2708e-04], [-9.0837e-05, 9.6038e-06, 2.5317e-05, ..., -1.1273e-05, 1.9562e-04, -7.8857e-05], [-6.2609e-04, 1.8954e-05, 3.3945e-05, ..., 7.6890e-05, 1.3208e-04, 2.5377e-05], ..., [ 1.7130e-04, -6.7711e-05, 1.5467e-05, ..., 1.0324e-04, -7.7915e-04, 6.0976e-05], [ 1.4913e-04, 1.2711e-05, 4.5478e-05, ..., 1.0383e-04, 1.9586e-04, 9.5785e-05], [ 1.7717e-05, 1.6034e-05, 2.0996e-05, ..., 3.5793e-05, 1.3840e-04, 2.3097e-05]], device='cuda:0') Epoch 25, bias, value: tensor([ 0.0085, 0.0190, 0.0132, -0.0027, 0.0162, -0.0066, -0.0174, -0.0122, 0.0264, -0.0087], device='cuda:0'), grad: tensor([ 0.0003, -0.0019, 0.0005, 0.0010, -0.0006, 0.0004, -0.0004, -0.0010, 0.0011, 0.0004], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 24---------------------------------------------------- epoch 24, time 268.03, cls_loss 0.0216 cls_loss_mapping 0.0408 cls_loss_causal 0.7950 re_mapping 0.0212 re_causal 0.0650 /// teacc 98.69 lr 0.00010000 Epoch 26, weight, value: tensor([[-0.0534, -0.0094, 0.0763, ..., 0.0180, -0.0562, -0.0464], [-0.0402, -0.0492, -0.0014, ..., -0.0482, -0.0329, -0.0020], [ 0.0689, -0.0055, -0.0442, ..., -0.0423, -0.0627, -0.0964], ..., [-0.0140, 0.0031, -0.0551, ..., 0.0034, 0.0219, 0.0180], [-0.0516, -0.0530, -0.0118, ..., -0.0288, 0.0071, -0.0173], [-0.0093, -0.0262, -0.0074, ..., 0.0472, 0.0217, -0.0542]], device='cuda:0'), grad: tensor([[ 2.8953e-05, 3.8892e-06, -8.5890e-05, ..., 4.0196e-06, 1.7321e-04, 6.9261e-05], [ 4.8846e-05, 3.4153e-05, 1.5001e-03, ..., -3.8910e-03, -5.1618e-05, 3.2663e-05], [-6.4802e-04, 3.1888e-05, 1.3626e-04, ..., 1.9455e-04, 2.6035e-04, 7.0632e-05], ..., [ 1.7536e-04, -3.0017e-04, 6.1333e-05, ..., 4.2992e-03, 2.3901e-04, 1.0863e-05], [ 1.9443e-04, 7.4096e-06, -2.1324e-03, ..., -1.6212e-03, -6.7663e-04, 4.9257e-04], [ 1.4320e-05, 5.2214e-05, 5.0592e-04, ..., 2.9802e-05, 8.5652e-05, 1.2034e-04]], device='cuda:0') Epoch 26, bias, value: tensor([ 0.0081, 0.0190, 0.0134, -0.0030, 0.0166, -0.0066, -0.0176, -0.0121, 0.0262, -0.0085], device='cuda:0'), grad: tensor([ 0.0003, -0.0062, 0.0001, 0.0013, 0.0012, -0.0144, 0.0136, 0.0070, -0.0039, 0.0011], device='cuda:0') 100 0.0001 changing lr epoch 25, time 251.29, cls_loss 0.0267 cls_loss_mapping 0.0460 cls_loss_causal 0.7893 re_mapping 0.0205 re_causal 0.0623 /// teacc 98.51 lr 0.00010000 Epoch 27, weight, value: tensor([[-0.0539, -0.0093, 0.0766, ..., 0.0181, -0.0565, -0.0479], [-0.0409, -0.0495, -0.0017, ..., -0.0483, -0.0335, -0.0027], [ 0.0699, -0.0058, -0.0450, ..., -0.0429, -0.0633, -0.0978], ..., [-0.0143, 0.0032, -0.0557, ..., 0.0028, 0.0219, 0.0170], [-0.0522, -0.0532, -0.0115, ..., -0.0295, 0.0072, -0.0181], [-0.0094, -0.0265, -0.0071, ..., 0.0478, 0.0218, -0.0561]], device='cuda:0'), grad: tensor([[-1.1368e-02, 6.5231e-04, -1.2047e-02, ..., -2.1103e-02, 4.6164e-05, 4.8399e-05], [-4.3392e-05, 1.0587e-05, -1.5676e-04, ..., 4.5419e-05, 2.7704e-04, -4.2170e-05], [ 8.1482e-03, -2.9579e-05, 1.0208e-02, ..., 1.6815e-02, 1.1849e-04, 2.0885e-04], ..., [ 3.8600e-04, 2.8849e-04, 5.5647e-04, ..., 5.4407e-04, -1.3704e-03, -4.7851e-04], [ 2.3317e-04, 4.2558e-05, 3.2902e-04, ..., 3.3951e-04, 6.0844e-04, 3.0828e-04], [ 1.7226e-04, 1.0169e-04, 4.2987e-04, ..., 4.4417e-04, 3.6407e-04, 1.1063e-04]], device='cuda:0') Epoch 27, bias, value: tensor([ 0.0081, 0.0187, 0.0136, -0.0028, 0.0167, -0.0066, -0.0177, -0.0124, 0.0264, -0.0084], device='cuda:0'), grad: tensor([-0.0235, -0.0005, 0.0220, 0.0033, 0.0055, -0.0108, 0.0005, -0.0002, 0.0020, 0.0017], device='cuda:0') 100 0.0001 changing lr epoch 26, time 251.13, cls_loss 0.0219 cls_loss_mapping 0.0425 cls_loss_causal 0.8226 re_mapping 0.0194 re_causal 0.0621 /// teacc 98.67 lr 0.00010000 Epoch 28, weight, value: tensor([[-0.0537, -0.0093, 0.0773, ..., 0.0189, -0.0571, -0.0485], [-0.0413, -0.0497, -0.0020, ..., -0.0482, -0.0342, -0.0040], [ 0.0706, -0.0060, -0.0463, ..., -0.0440, -0.0638, -0.0995], ..., [-0.0146, 0.0035, -0.0562, ..., 0.0026, 0.0225, 0.0164], [-0.0525, -0.0535, -0.0111, ..., -0.0300, 0.0071, -0.0187], [-0.0099, -0.0267, -0.0076, ..., 0.0485, 0.0220, -0.0578]], device='cuda:0'), grad: tensor([[ 1.2153e-04, -3.0935e-05, 1.9730e-02, ..., 1.7975e-02, 6.3777e-05, 5.5194e-05], [ 1.9765e-04, 2.2724e-06, 4.3929e-05, ..., 6.6340e-05, 2.6822e-04, 1.2982e-04], [-2.9430e-03, 5.6960e-06, 1.9455e-04, ..., 1.2219e-04, -1.0425e-04, -3.4571e-05], ..., [ 1.0328e-03, 5.8226e-06, 9.4891e-05, ..., 1.0526e-04, 6.2466e-05, 7.5638e-05], [ 8.3780e-04, 7.6517e-06, 2.1434e-04, ..., 9.1910e-05, -2.0266e-04, -1.3351e-04], [ 1.3185e-04, 6.3665e-06, -2.0584e-02, ..., -1.7471e-02, 5.1641e-04, 4.1097e-05]], device='cuda:0') Epoch 28, bias, value: tensor([ 0.0082, 0.0185, 0.0132, -0.0029, 0.0163, -0.0068, -0.0170, -0.0120, 0.0263, -0.0082], device='cuda:0'), grad: tensor([ 0.0303, 0.0009, -0.0041, 0.0003, -0.0009, 0.0010, -0.0001, 0.0012, 0.0009, -0.0296], device='cuda:0') 100 0.0001 changing lr epoch 27, time 251.08, cls_loss 0.0219 cls_loss_mapping 0.0464 cls_loss_causal 0.8260 re_mapping 0.0200 re_causal 0.0615 /// teacc 98.58 lr 0.00010000 Epoch 29, weight, value: tensor([[-0.0543, -0.0091, 0.0786, ..., 0.0197, -0.0579, -0.0493], [-0.0418, -0.0501, -0.0022, ..., -0.0484, -0.0348, -0.0046], [ 0.0718, -0.0063, -0.0474, ..., -0.0446, -0.0643, -0.1003], ..., [-0.0151, 0.0037, -0.0570, ..., 0.0022, 0.0227, 0.0158], [-0.0530, -0.0539, -0.0111, ..., -0.0308, 0.0071, -0.0175], [-0.0105, -0.0271, -0.0078, ..., 0.0490, 0.0220, -0.0585]], device='cuda:0'), grad: tensor([[ 3.7611e-05, 3.6508e-06, -6.2037e-04, ..., -2.4700e-04, 4.3392e-05, 1.3173e-05], [ 1.6525e-05, 2.7306e-06, 2.0421e-04, ..., 1.2898e-04, 3.3349e-05, -2.3544e-05], [-8.1182e-05, 3.3695e-06, 1.2147e-04, ..., 7.2122e-05, 7.7367e-05, 2.2724e-05], ..., [ 3.9190e-05, 1.3208e-04, 2.9042e-05, ..., 4.8876e-04, 4.5729e-04, 4.0948e-05], [ 1.3877e-06, 1.2748e-05, 9.0659e-05, ..., 3.6240e-04, 2.7585e-04, 3.3200e-05], [-3.1114e-05, -1.9205e-04, 4.1157e-05, ..., -1.2398e-03, -1.2732e-03, -1.0669e-04]], device='cuda:0') Epoch 29, bias, value: tensor([ 0.0086, 0.0185, 0.0134, -0.0026, 0.0163, -0.0065, -0.0175, -0.0121, 0.0263, -0.0086], device='cuda:0'), grad: tensor([-2.9230e-04, -8.6948e-06, 2.2686e-04, 1.3804e-04, 5.0306e-04, 6.7174e-05, 6.2108e-05, 8.0919e-04, 3.5644e-04, -1.8616e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 28---------------------------------------------------- epoch 28, time 268.25, cls_loss 0.0222 cls_loss_mapping 0.0426 cls_loss_causal 0.8053 re_mapping 0.0194 re_causal 0.0594 /// teacc 98.82 lr 0.00010000 Epoch 30, weight, value: tensor([[-0.0549, -0.0093, 0.0790, ..., 0.0201, -0.0588, -0.0506], [-0.0422, -0.0505, -0.0021, ..., -0.0491, -0.0358, -0.0052], [ 0.0732, -0.0075, -0.0484, ..., -0.0451, -0.0649, -0.1013], ..., [-0.0154, 0.0046, -0.0578, ..., 0.0015, 0.0229, 0.0158], [-0.0536, -0.0540, -0.0111, ..., -0.0316, 0.0070, -0.0172], [-0.0112, -0.0275, -0.0082, ..., 0.0493, 0.0221, -0.0586]], device='cuda:0'), grad: tensor([[ 7.0810e-05, 6.9886e-06, -2.8372e-04, ..., -2.1446e-04, 2.3976e-05, 5.8413e-05], [ 2.0969e-04, 2.4676e-05, 3.2097e-05, ..., 8.4400e-05, 5.9515e-05, 2.6441e-04], [-1.4138e-04, 2.7552e-05, 1.1426e-04, ..., 1.3733e-04, 8.0526e-05, 2.0707e-04], ..., [ 1.3304e-04, 2.5295e-06, 2.1338e-05, ..., 5.8800e-05, -4.6670e-05, 2.0042e-05], [ 2.3437e-04, 3.5197e-05, 2.9802e-05, ..., 1.5986e-04, 1.8156e-04, 8.4519e-05], [ 2.2650e-04, 2.7850e-05, 5.0277e-05, ..., 5.6744e-04, 3.5381e-04, 1.4514e-05]], device='cuda:0') Epoch 30, bias, value: tensor([ 0.0082, 0.0183, 0.0135, -0.0027, 0.0166, -0.0063, -0.0176, -0.0120, 0.0263, -0.0087], device='cuda:0'), grad: tensor([-0.0002, 0.0005, 0.0007, -0.0016, -0.0004, 0.0002, -0.0018, 0.0003, 0.0009, 0.0012], device='cuda:0') 100 0.0001 changing lr epoch 29, time 251.21, cls_loss 0.0144 cls_loss_mapping 0.0322 cls_loss_causal 0.8132 re_mapping 0.0186 re_causal 0.0598 /// teacc 98.71 lr 0.00010000 Epoch 31, weight, value: tensor([[-0.0557, -0.0094, 0.0798, ..., 0.0206, -0.0594, -0.0513], [-0.0425, -0.0509, -0.0022, ..., -0.0498, -0.0364, -0.0057], [ 0.0740, -0.0077, -0.0492, ..., -0.0450, -0.0655, -0.1026], ..., [-0.0158, 0.0046, -0.0583, ..., 0.0012, 0.0234, 0.0157], [-0.0542, -0.0542, -0.0114, ..., -0.0323, 0.0069, -0.0170], [-0.0117, -0.0277, -0.0086, ..., 0.0497, 0.0222, -0.0589]], device='cuda:0'), grad: tensor([[-3.3712e-04, -1.1005e-05, -1.3599e-03, ..., -3.1638e-04, -2.7627e-05, 1.4029e-05], [ 5.0247e-05, 1.5363e-05, 1.2517e-05, ..., 5.9783e-05, 5.9694e-05, 1.5218e-06], [ 1.0329e-04, 4.2528e-05, 3.8147e-04, ..., 1.7357e-04, 9.8050e-05, 1.2539e-05], ..., [ 1.4508e-04, 4.2655e-06, 2.5749e-05, ..., 1.0192e-04, 1.6659e-05, 1.1206e-05], [ 2.2209e-04, 2.9385e-05, 4.2987e-04, ..., 2.3830e-04, 1.1617e-04, 1.0896e-07], [ 8.3983e-05, 3.5167e-05, 1.7107e-04, ..., -5.7173e-04, -3.8624e-04, -7.3493e-05]], device='cuda:0') Epoch 31, bias, value: tensor([ 0.0083, 0.0180, 0.0136, -0.0026, 0.0165, -0.0061, -0.0176, -0.0119, 0.0261, -0.0088], device='cuda:0'), grad: tensor([-1.7090e-03, 2.2396e-05, 8.2254e-04, -5.3835e-04, -7.7724e-05, 5.2118e-04, 2.4939e-04, 3.6216e-04, 8.8215e-04, -5.3501e-04], device='cuda:0') 100 0.0001 changing lr epoch 30, time 251.20, cls_loss 0.0200 cls_loss_mapping 0.0382 cls_loss_causal 0.7281 re_mapping 0.0190 re_causal 0.0545 /// teacc 98.61 lr 0.00010000 Epoch 32, weight, value: tensor([[-0.0562, -0.0094, 0.0800, ..., 0.0206, -0.0602, -0.0524], [-0.0432, -0.0512, -0.0021, ..., -0.0502, -0.0370, -0.0057], [ 0.0748, -0.0080, -0.0500, ..., -0.0454, -0.0659, -0.1030], ..., [-0.0160, 0.0047, -0.0588, ..., 0.0008, 0.0236, 0.0156], [-0.0546, -0.0544, -0.0115, ..., -0.0332, 0.0070, -0.0180], [-0.0121, -0.0279, -0.0079, ..., 0.0502, 0.0222, -0.0599]], device='cuda:0'), grad: tensor([[ 1.7738e-04, -1.5557e-05, -3.2139e-04, ..., -2.6369e-04, 3.2043e-04, 1.1295e-04], [ 2.2426e-05, 7.8455e-06, -3.3323e-06, ..., -1.1899e-05, 2.1055e-05, 8.4750e-07], [ 6.1810e-05, 4.0293e-05, 3.6430e-04, ..., 2.1231e-04, 2.6870e-04, 1.2267e-04], ..., [ 5.9277e-05, 3.2336e-05, 7.6890e-05, ..., 1.5700e-04, 5.0552e-06, -1.8477e-05], [ 1.0014e-04, 2.0042e-05, 7.1883e-05, ..., 1.5473e-04, 2.5725e-04, 4.2796e-05], [ 2.4259e-05, -1.5441e-06, 1.0061e-04, ..., -5.0455e-05, -7.3016e-05, 3.3528e-05]], device='cuda:0') Epoch 32, bias, value: tensor([ 0.0077, 0.0176, 0.0139, -0.0022, 0.0166, -0.0062, -0.0177, -0.0118, 0.0262, -0.0086], device='cuda:0'), grad: tensor([ 3.5524e-04, -7.1335e-04, 8.7452e-04, -9.2745e-04, -1.0723e-04, -5.0038e-05, -4.5037e-04, 1.8454e-04, 7.1859e-04, 1.1450e-04], device='cuda:0') 100 0.0001 changing lr epoch 31, time 251.53, cls_loss 0.0179 cls_loss_mapping 0.0347 cls_loss_causal 0.7735 re_mapping 0.0183 re_causal 0.0572 /// teacc 98.78 lr 0.00010000 Epoch 33, weight, value: tensor([[-0.0572, -0.0093, 0.0805, ..., 0.0210, -0.0609, -0.0534], [-0.0435, -0.0514, -0.0019, ..., -0.0505, -0.0375, -0.0065], [ 0.0754, -0.0092, -0.0508, ..., -0.0462, -0.0667, -0.1032], ..., [-0.0158, 0.0049, -0.0596, ..., 0.0002, 0.0238, 0.0158], [-0.0556, -0.0544, -0.0116, ..., -0.0336, 0.0069, -0.0183], [-0.0120, -0.0283, -0.0077, ..., 0.0507, 0.0225, -0.0591]], device='cuda:0'), grad: tensor([[-4.7773e-05, 7.9870e-06, 2.0294e-03, ..., 8.3971e-04, 1.3399e-03, 2.7046e-03], [ 8.0690e-06, 1.9088e-05, 1.3888e-04, ..., 7.0572e-05, 1.1772e-04, 1.3268e-04], [ 3.5107e-05, 2.3022e-05, 1.5795e-04, ..., 1.6403e-04, 7.8261e-05, 5.2005e-05], ..., [ 1.7017e-05, -8.7082e-05, 6.2823e-05, ..., 1.6582e-04, -3.0088e-04, -2.5809e-05], [ 9.2089e-06, 2.0146e-05, 2.1589e-04, ..., 1.1414e-04, 2.3150e-04, 1.6320e-04], [ 4.6849e-05, 7.0632e-05, 2.6512e-04, ..., 1.7548e-04, 1.9252e-04, 1.0437e-04]], device='cuda:0') Epoch 33, bias, value: tensor([ 0.0077, 0.0178, 0.0133, -0.0025, 0.0168, -0.0065, -0.0172, -0.0115, 0.0260, -0.0082], device='cuda:0'), grad: tensor([ 0.0028, 0.0001, 0.0003, 0.0001, 0.0005, -0.0002, -0.0048, -0.0002, 0.0006, 0.0007], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 32---------------------------------------------------- epoch 32, time 267.79, cls_loss 0.0178 cls_loss_mapping 0.0357 cls_loss_causal 0.7641 re_mapping 0.0177 re_causal 0.0552 /// teacc 98.85 lr 0.00010000 Epoch 34, weight, value: tensor([[-0.0579, -0.0093, 0.0811, ..., 0.0215, -0.0614, -0.0549], [-0.0439, -0.0518, -0.0027, ..., -0.0504, -0.0375, -0.0069], [ 0.0763, -0.0094, -0.0515, ..., -0.0465, -0.0672, -0.1044], ..., [-0.0166, 0.0045, -0.0603, ..., -0.0007, 0.0237, 0.0154], [-0.0562, -0.0550, -0.0111, ..., -0.0343, 0.0066, -0.0186], [-0.0123, -0.0281, -0.0078, ..., 0.0513, 0.0227, -0.0592]], device='cuda:0'), grad: tensor([[-9.2760e-06, 7.7812e-07, -1.2932e-03, ..., -7.9679e-04, 1.5065e-05, 2.6450e-05], [ 5.1446e-06, 3.0790e-06, -4.1574e-05, ..., 1.9893e-05, 6.4552e-05, 4.8816e-05], [ 2.6405e-05, 1.3262e-05, 1.4973e-04, ..., 9.7156e-05, 2.4706e-05, 1.7166e-05], ..., [ 1.0513e-05, -1.1362e-06, 4.6402e-05, ..., 4.2170e-05, -1.2648e-04, -4.2021e-05], [ 8.6129e-06, 4.8205e-06, 1.4460e-04, ..., 1.3745e-04, 1.0538e-04, 1.3173e-04], [ 1.3821e-05, 9.9912e-06, 5.7793e-04, ..., 2.6631e-04, -1.3256e-04, -5.8025e-05]], device='cuda:0') Epoch 34, bias, value: tensor([ 0.0073, 0.0177, 0.0133, -0.0024, 0.0168, -0.0058, -0.0171, -0.0120, 0.0257, -0.0079], device='cuda:0'), grad: tensor([-1.1797e-03, -1.1629e-04, 2.6679e-04, 1.3804e-04, -5.6118e-05, 5.6601e-04, -3.9959e-04, -6.6757e-05, 4.7708e-04, 3.7003e-04], device='cuda:0') 100 0.0001 changing lr epoch 33, time 251.15, cls_loss 0.0166 cls_loss_mapping 0.0340 cls_loss_causal 0.7346 re_mapping 0.0169 re_causal 0.0537 /// teacc 98.82 lr 0.00010000 Epoch 35, weight, value: tensor([[-0.0587, -0.0094, 0.0818, ..., 0.0225, -0.0619, -0.0561], [-0.0444, -0.0521, -0.0028, ..., -0.0509, -0.0381, -0.0072], [ 0.0773, -0.0095, -0.0520, ..., -0.0470, -0.0676, -0.1055], ..., [-0.0172, 0.0047, -0.0607, ..., -0.0011, 0.0241, 0.0156], [-0.0568, -0.0552, -0.0108, ..., -0.0346, 0.0066, -0.0190], [-0.0121, -0.0285, -0.0088, ..., 0.0516, 0.0230, -0.0585]], device='cuda:0'), grad: tensor([[ 4.1388e-06, 5.0887e-06, -1.5831e-04, ..., -6.0558e-05, 2.6762e-05, 2.6390e-05], [ 5.3942e-06, 1.3255e-05, -8.8274e-05, ..., -3.3468e-05, 2.5347e-05, 5.6736e-06], [-4.6194e-05, 1.1675e-05, 9.6619e-05, ..., 4.8637e-05, 4.0501e-05, 3.1531e-05], ..., [ 1.3471e-05, 1.6224e-06, 1.6481e-05, ..., 3.0667e-05, -3.0294e-05, 1.2768e-06], [ 1.6361e-05, 1.5914e-05, 7.3195e-05, ..., 7.8261e-05, 6.3837e-05, 3.4183e-05], [ 2.5779e-06, 5.3525e-05, 4.1664e-05, ..., 6.6876e-05, 4.0591e-05, 3.5852e-05]], device='cuda:0') Epoch 35, bias, value: tensor([ 0.0075, 0.0175, 0.0136, -0.0029, 0.0170, -0.0060, -0.0171, -0.0119, 0.0258, -0.0080], device='cuda:0'), grad: tensor([-3.1769e-05, -3.2496e-04, 2.2042e-04, -1.2743e-04, 2.2113e-04, -2.8038e-04, -2.1398e-04, 2.7761e-05, 1.5342e-04, 3.5453e-04], device='cuda:0') 100 0.0001 changing lr epoch 34, time 251.21, cls_loss 0.0146 cls_loss_mapping 0.0335 cls_loss_causal 0.7366 re_mapping 0.0162 re_causal 0.0512 /// teacc 98.58 lr 0.00010000 Epoch 36, weight, value: tensor([[-0.0592, -0.0094, 0.0828, ..., 0.0230, -0.0621, -0.0554], [-0.0442, -0.0524, -0.0027, ..., -0.0509, -0.0381, -0.0075], [ 0.0782, -0.0096, -0.0525, ..., -0.0475, -0.0683, -0.1065], ..., [-0.0176, 0.0044, -0.0611, ..., -0.0018, 0.0242, 0.0156], [-0.0577, -0.0554, -0.0112, ..., -0.0353, 0.0066, -0.0195], [-0.0121, -0.0285, -0.0092, ..., 0.0523, 0.0231, -0.0590]], device='cuda:0'), grad: tensor([[ 1.9819e-05, 1.3066e-06, -1.0663e-04, ..., -1.0431e-06, 3.9160e-05, 1.3523e-05], [ 3.0369e-05, 3.8669e-06, 1.2210e-06, ..., 1.0729e-04, 3.4046e-04, 6.4149e-06], [ 1.6463e-04, 6.2063e-06, 2.6554e-05, ..., 2.1410e-04, 1.0508e-04, 3.7774e-06], ..., [ 2.5466e-05, -4.8190e-05, 4.1761e-06, ..., -6.5207e-05, -5.1022e-04, 4.8615e-07], [ 3.4124e-05, 4.1872e-06, 2.5198e-05, ..., 2.0123e-04, 2.2328e-04, 1.7807e-05], [ 2.3857e-05, 1.8582e-05, 2.2367e-05, ..., 1.5745e-03, 7.2384e-04, 4.0010e-06]], device='cuda:0') Epoch 36, bias, value: tensor([ 0.0079, 0.0183, 0.0135, -0.0029, 0.0168, -0.0062, -0.0170, -0.0120, 0.0253, -0.0080], device='cuda:0'), grad: tensor([ 6.9618e-05, 2.4776e-03, 5.6744e-04, 2.4261e-03, -1.9484e-03, -6.6299e-03, 1.5855e-04, -7.1573e-04, 1.3323e-03, 2.2602e-03], device='cuda:0') 100 0.0001 changing lr epoch 35, time 251.46, cls_loss 0.0120 cls_loss_mapping 0.0265 cls_loss_causal 0.7327 re_mapping 0.0158 re_causal 0.0511 /// teacc 98.77 lr 0.00010000 Epoch 37, weight, value: tensor([[-0.0598, -0.0094, 0.0834, ..., 0.0235, -0.0627, -0.0561], [-0.0446, -0.0528, -0.0027, ..., -0.0513, -0.0389, -0.0079], [ 0.0789, -0.0103, -0.0530, ..., -0.0476, -0.0687, -0.1070], ..., [-0.0180, 0.0050, -0.0615, ..., -0.0024, 0.0247, 0.0157], [-0.0581, -0.0555, -0.0114, ..., -0.0360, 0.0065, -0.0195], [-0.0124, -0.0289, -0.0095, ..., 0.0526, 0.0233, -0.0588]], device='cuda:0'), grad: tensor([[ 5.6356e-05, -6.1178e-04, -6.0616e-03, ..., -3.5686e-03, -1.4372e-03, 2.9281e-05], [ 1.6093e-04, 9.3043e-05, 1.3620e-05, ..., 1.5602e-05, 2.2757e-04, 9.3505e-06], [ 1.8740e-03, 3.9268e-04, 3.3617e-04, ..., 2.0289e-04, 1.2980e-03, 7.9215e-05], ..., [-2.1839e-03, -2.4643e-03, 3.7014e-05, ..., 6.8128e-05, -5.1613e-03, 8.8364e-06], [ 1.9140e-03, 2.1973e-03, 3.0785e-03, ..., 1.8053e-03, 4.8637e-03, 5.8919e-05], [ 5.6952e-05, 5.9694e-05, 4.9925e-04, ..., 4.1771e-04, 1.4782e-04, 1.2644e-05]], device='cuda:0') Epoch 37, bias, value: tensor([ 0.0080, 0.0179, 0.0135, -0.0027, 0.0167, -0.0064, -0.0169, -0.0113, 0.0252, -0.0081], device='cuda:0'), grad: tensor([-7.0190e-03, 4.7088e-04, 4.1542e-03, -3.7575e-03, -7.0989e-05, 1.7929e-03, 2.0905e-03, -1.1101e-02, 1.2581e-02, 8.3923e-04], device='cuda:0') 100 0.0001 changing lr epoch 36, time 251.32, cls_loss 0.0149 cls_loss_mapping 0.0345 cls_loss_causal 0.7418 re_mapping 0.0151 re_causal 0.0499 /// teacc 98.83 lr 0.00010000 Epoch 38, weight, value: tensor([[-0.0604, -0.0095, 0.0840, ..., 0.0238, -0.0634, -0.0567], [-0.0453, -0.0533, -0.0030, ..., -0.0517, -0.0397, -0.0092], [ 0.0803, -0.0107, -0.0537, ..., -0.0480, -0.0690, -0.1069], ..., [-0.0184, 0.0051, -0.0620, ..., -0.0028, 0.0251, 0.0154], [-0.0589, -0.0557, -0.0115, ..., -0.0363, 0.0067, -0.0188], [-0.0130, -0.0293, -0.0101, ..., 0.0528, 0.0234, -0.0596]], device='cuda:0'), grad: tensor([[ 9.1642e-06, -3.8892e-06, -2.4176e-04, ..., -8.9467e-05, 4.1038e-05, 4.8041e-05], [ 1.4342e-05, 2.0582e-06, -6.0126e-06, ..., 3.2876e-06, 4.0740e-05, 2.9162e-05], [-2.1183e-04, 2.7847e-06, 6.0558e-05, ..., 1.4737e-05, 1.4469e-05, 3.6117e-06], ..., [ 1.5244e-05, -2.5108e-06, 3.2812e-05, ..., 1.9118e-05, -3.4630e-05, 1.0766e-05], [ 1.2088e-04, 1.8328e-05, 2.9898e-04, ..., 5.6416e-05, 5.8985e-04, 7.1669e-04], [ 1.6708e-06, 6.3367e-06, 6.1750e-05, ..., -9.3043e-05, -8.3983e-05, 1.6183e-05]], device='cuda:0') Epoch 38, bias, value: tensor([ 0.0078, 0.0176, 0.0140, -0.0027, 0.0167, -0.0065, -0.0166, -0.0112, 0.0254, -0.0086], device='cuda:0'), grad: tensor([-1.1152e-04, -1.2636e-04, -1.3292e-04, 9.9754e-04, 1.8358e-04, -2.0313e-03, -3.7813e-04, 4.3660e-05, 1.5888e-03, -3.2276e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 37---------------------------------------------------- epoch 37, time 269.98, cls_loss 0.0137 cls_loss_mapping 0.0287 cls_loss_causal 0.7411 re_mapping 0.0152 re_causal 0.0483 /// teacc 98.87 lr 0.00010000 Epoch 39, weight, value: tensor([[-0.0610, -0.0096, 0.0846, ..., 0.0242, -0.0642, -0.0577], [-0.0457, -0.0537, -0.0027, ..., -0.0522, -0.0405, -0.0098], [ 0.0810, -0.0113, -0.0547, ..., -0.0485, -0.0697, -0.1078], ..., [-0.0186, 0.0046, -0.0628, ..., -0.0034, 0.0251, 0.0155], [-0.0594, -0.0559, -0.0112, ..., -0.0366, 0.0067, -0.0188], [-0.0130, -0.0295, -0.0105, ..., 0.0531, 0.0235, -0.0598]], device='cuda:0'), grad: tensor([[ 3.2336e-05, 6.4448e-06, -2.2256e-04, ..., -8.0824e-05, 1.9148e-05, 9.4771e-06], [ 7.4327e-05, 8.3968e-06, 6.0834e-06, ..., 9.6783e-06, 4.2617e-05, 1.3746e-06], [-4.6849e-04, -3.2812e-05, 1.8165e-05, ..., 1.8835e-05, 3.4988e-05, 7.6508e-07], ..., [ 1.2624e-04, 1.8150e-05, 4.9584e-06, ..., 7.9721e-06, -1.9717e-04, 1.8356e-06], [ 5.8800e-05, 2.2995e-04, 2.7761e-05, ..., 1.0675e-04, 3.1877e-04, 6.9916e-05], [ 1.8865e-05, 8.3089e-05, 1.2636e-04, ..., 1.3220e-04, 1.4579e-04, 2.5585e-05]], device='cuda:0') Epoch 39, bias, value: tensor([ 0.0078, 0.0173, 0.0140, -0.0027, 0.0172, -0.0068, -0.0163, -0.0112, 0.0252, -0.0084], device='cuda:0'), grad: tensor([-8.5354e-05, 1.9205e-04, -3.5548e-04, 3.3307e-04, -3.1531e-05, -1.1768e-03, 1.2791e-04, -4.6539e-04, 9.3126e-04, 5.3072e-04], device='cuda:0') 100 0.0001 changing lr epoch 38, time 251.41, cls_loss 0.0169 cls_loss_mapping 0.0346 cls_loss_causal 0.7538 re_mapping 0.0148 re_causal 0.0478 /// teacc 98.86 lr 0.00010000 Epoch 40, weight, value: tensor([[-0.0616, -0.0097, 0.0850, ..., 0.0240, -0.0648, -0.0589], [-0.0465, -0.0540, -0.0019, ..., -0.0525, -0.0412, -0.0104], [ 0.0817, -0.0120, -0.0556, ..., -0.0493, -0.0702, -0.1077], ..., [-0.0184, 0.0046, -0.0636, ..., -0.0037, 0.0254, 0.0159], [-0.0599, -0.0564, -0.0114, ..., -0.0372, 0.0065, -0.0188], [-0.0132, -0.0299, -0.0105, ..., 0.0538, 0.0236, -0.0609]], device='cuda:0'), grad: tensor([[ 8.3521e-06, 1.3098e-05, 1.0222e-04, ..., -2.4274e-05, 5.7578e-05, 1.6320e-04], [ 5.4277e-06, 3.9563e-06, 3.7532e-06, ..., 1.3523e-05, 1.9014e-05, 1.1586e-05], [-3.3832e-04, -1.0192e-04, 4.3631e-05, ..., 9.7871e-05, -1.6376e-05, 2.3946e-05], ..., [ 2.6441e-04, 6.3837e-05, 6.8024e-06, ..., 1.4460e-04, 1.0210e-04, 1.4290e-05], [-1.1332e-05, 3.3863e-06, -2.5570e-05, ..., 1.4901e-05, -2.5153e-05, -7.6964e-06], [ 4.6492e-06, 1.1832e-05, 2.3320e-05, ..., 4.5747e-05, 3.7313e-05, 2.2054e-05]], device='cuda:0') Epoch 40, bias, value: tensor([ 0.0073, 0.0165, 0.0142, -0.0024, 0.0171, -0.0068, -0.0161, -0.0110, 0.0249, -0.0080], device='cuda:0'), grad: tensor([ 1.8764e-04, -1.1759e-03, 4.2152e-04, 2.0576e-04, -1.7846e-04, -8.7619e-05, -7.9453e-05, 6.1560e-04, -1.4484e-04, 2.3580e-04], device='cuda:0') 100 0.0001 changing lr epoch 39, time 251.28, cls_loss 0.0109 cls_loss_mapping 0.0264 cls_loss_causal 0.6991 re_mapping 0.0145 re_causal 0.0465 /// teacc 98.81 lr 0.00010000 Epoch 41, weight, value: tensor([[-0.0620, -0.0098, 0.0857, ..., 0.0248, -0.0655, -0.0600], [-0.0465, -0.0542, -0.0017, ..., -0.0531, -0.0415, -0.0107], [ 0.0827, -0.0123, -0.0564, ..., -0.0500, -0.0706, -0.1085], ..., [-0.0189, 0.0047, -0.0640, ..., -0.0040, 0.0257, 0.0151], [-0.0603, -0.0566, -0.0111, ..., -0.0378, 0.0067, -0.0178], [-0.0134, -0.0302, -0.0110, ..., 0.0537, 0.0236, -0.0619]], device='cuda:0'), grad: tensor([[ 1.9133e-05, 2.8126e-07, -5.8115e-07, ..., -2.6952e-06, 3.5942e-05, 3.1114e-05], [ 9.7334e-05, 4.0643e-06, -5.3465e-05, ..., 4.6529e-06, 2.7359e-05, -6.9141e-05], [-5.2333e-05, 5.7146e-06, 2.1413e-05, ..., 5.8562e-06, 1.8492e-05, 1.1161e-05], ..., [ 8.1956e-05, 4.0047e-06, 5.3234e-06, ..., 2.2128e-05, -1.8919e-04, 1.0826e-05], [ 4.4137e-05, 2.4177e-06, 2.7418e-05, ..., 5.3465e-05, 6.3598e-05, 6.3360e-05], [ 3.9697e-05, 8.0690e-06, 1.0103e-05, ..., -2.8539e-04, 1.3120e-05, 2.4021e-05]], device='cuda:0') Epoch 41, bias, value: tensor([ 0.0075, 0.0172, 0.0141, -0.0025, 0.0171, -0.0069, -0.0160, -0.0116, 0.0250, -0.0082], device='cuda:0'), grad: tensor([ 1.7655e-04, -1.9848e-04, 3.1376e-04, 2.3136e-03, 3.3569e-04, -3.2940e-03, -4.5598e-05, -3.9363e-04, 4.9734e-04, 2.9159e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 40---------------------------------------------------- epoch 40, time 269.33, cls_loss 0.0108 cls_loss_mapping 0.0255 cls_loss_causal 0.6745 re_mapping 0.0147 re_causal 0.0457 /// teacc 98.92 lr 0.00010000 Epoch 42, weight, value: tensor([[-0.0624, -0.0099, 0.0865, ..., 0.0251, -0.0660, -0.0602], [-0.0466, -0.0545, -0.0023, ..., -0.0536, -0.0417, -0.0105], [ 0.0835, -0.0132, -0.0570, ..., -0.0507, -0.0710, -0.1095], ..., [-0.0196, 0.0049, -0.0643, ..., -0.0043, 0.0259, 0.0148], [-0.0607, -0.0565, -0.0113, ..., -0.0385, 0.0066, -0.0177], [-0.0139, -0.0305, -0.0110, ..., 0.0545, 0.0238, -0.0629]], device='cuda:0'), grad: tensor([[ 9.8348e-06, 6.1542e-06, 2.9922e-05, ..., 8.3372e-06, 2.0742e-05, 1.6123e-05], [ 1.5378e-04, 1.7062e-05, 1.0356e-05, ..., -2.5362e-05, 5.0753e-05, -6.4909e-05], [-3.1853e-04, 1.8507e-05, 1.0140e-05, ..., 1.0647e-05, 3.4332e-05, 2.2247e-05], ..., [ 5.9545e-05, 3.1620e-05, 1.2713e-06, ..., 8.2374e-05, 1.4871e-05, 2.8998e-05], [ 6.5804e-05, 5.6416e-05, 3.2067e-05, ..., 1.2743e-04, 2.0647e-04, 3.6508e-05], [ 1.5058e-05, 1.2529e-04, 5.0999e-06, ..., -2.0182e-04, -1.5736e-04, 1.8075e-05]], device='cuda:0') Epoch 42, bias, value: tensor([ 0.0075, 0.0174, 0.0142, -0.0026, 0.0168, -0.0067, -0.0159, -0.0116, 0.0248, -0.0081], device='cuda:0'), grad: tensor([ 1.0157e-04, -3.1185e-04, -2.6655e-04, -2.8400e-03, 5.5122e-04, 2.2697e-03, -4.8566e-04, 2.3186e-04, 7.4100e-04, 9.6709e-06], device='cuda:0') 100 0.0001 changing lr epoch 41, time 251.87, cls_loss 0.0130 cls_loss_mapping 0.0275 cls_loss_causal 0.7024 re_mapping 0.0142 re_causal 0.0460 /// teacc 98.74 lr 0.00010000 Epoch 43, weight, value: tensor([[-0.0628, -0.0098, 0.0871, ..., 0.0256, -0.0665, -0.0607], [-0.0469, -0.0549, -0.0023, ..., -0.0553, -0.0423, -0.0110], [ 0.0843, -0.0136, -0.0576, ..., -0.0511, -0.0716, -0.1096], ..., [-0.0198, 0.0049, -0.0646, ..., -0.0048, 0.0263, 0.0143], [-0.0613, -0.0567, -0.0114, ..., -0.0391, 0.0068, -0.0170], [-0.0145, -0.0310, -0.0117, ..., 0.0549, 0.0237, -0.0653]], device='cuda:0'), grad: tensor([[ 8.2478e-06, 6.4597e-06, -4.0233e-05, ..., -1.0341e-05, 3.6508e-05, 3.4600e-05], [ 3.0249e-05, 2.0981e-05, 2.3678e-05, ..., 1.5676e-05, 4.7863e-05, 3.3170e-05], [ 2.3440e-05, 3.5226e-05, 4.0412e-05, ..., 2.3946e-05, 5.2512e-05, 4.3511e-05], ..., [ 3.3557e-05, -8.9943e-05, 4.3213e-06, ..., -4.6939e-05, -4.2963e-04, -3.3557e-05], [ 4.3422e-05, 3.7253e-05, -4.6015e-05, ..., 2.5973e-05, 8.3089e-05, 2.6524e-05], [ 1.5870e-05, 5.6267e-05, 1.3106e-05, ..., 6.8247e-06, 1.7118e-04, 2.6181e-05]], device='cuda:0') Epoch 43, bias, value: tensor([ 0.0077, 0.0169, 0.0146, -0.0028, 0.0172, -0.0063, -0.0157, -0.0118, 0.0248, -0.0084], device='cuda:0'), grad: tensor([ 1.1945e-04, 2.4533e-04, 3.1209e-04, -2.7180e-04, 8.6427e-05, -2.5415e-04, 2.2876e-04, -1.1177e-03, 1.9133e-04, 4.6110e-04], device='cuda:0') 100 0.0001 changing lr epoch 42, time 251.33, cls_loss 0.0105 cls_loss_mapping 0.0265 cls_loss_causal 0.6975 re_mapping 0.0137 re_causal 0.0430 /// teacc 98.68 lr 0.00010000 Epoch 44, weight, value: tensor([[-0.0632, -0.0099, 0.0876, ..., 0.0259, -0.0670, -0.0613], [-0.0470, -0.0554, -0.0017, ..., -0.0554, -0.0428, -0.0114], [ 0.0850, -0.0142, -0.0583, ..., -0.0515, -0.0721, -0.1106], ..., [-0.0201, 0.0053, -0.0648, ..., -0.0055, 0.0268, 0.0145], [-0.0617, -0.0569, -0.0116, ..., -0.0399, 0.0068, -0.0173], [-0.0146, -0.0313, -0.0120, ..., 0.0558, 0.0239, -0.0659]], device='cuda:0'), grad: tensor([[ 3.6627e-05, 4.3400e-06, -1.3813e-05, ..., -8.1956e-06, 1.7911e-05, 1.8194e-05], [ 4.9543e-04, 1.1958e-05, 3.1497e-06, ..., 1.9008e-06, 2.1443e-05, 2.6420e-05], [-1.8253e-03, -1.1194e-04, 7.9796e-06, ..., 6.7018e-06, 1.9237e-05, -1.3113e-04], ..., [ 5.5265e-04, 9.5606e-05, 4.0121e-06, ..., 2.5883e-05, 1.2450e-05, 1.1459e-05], [ 4.3488e-04, 3.1888e-05, -5.5265e-04, ..., 5.9828e-06, -8.3685e-04, -8.5449e-04], [ 1.8492e-05, 8.5160e-06, 1.5169e-05, ..., -2.0325e-05, 6.8128e-05, 4.6074e-05]], device='cuda:0') Epoch 44, bias, value: tensor([ 0.0076, 0.0172, 0.0144, -0.0024, 0.0169, -0.0069, -0.0156, -0.0112, 0.0245, -0.0085], device='cuda:0'), grad: tensor([ 7.5281e-05, 5.9795e-04, -2.7485e-03, -1.3769e-04, 8.1348e-04, 1.8239e-04, 1.4658e-03, 7.5245e-04, -1.2274e-03, 2.2650e-04], device='cuda:0') 100 0.0001 changing lr epoch 43, time 251.54, cls_loss 0.0091 cls_loss_mapping 0.0234 cls_loss_causal 0.6902 re_mapping 0.0132 re_causal 0.0418 /// teacc 98.83 lr 0.00010000 Epoch 45, weight, value: tensor([[-0.0636, -0.0099, 0.0883, ..., 0.0266, -0.0675, -0.0619], [-0.0475, -0.0560, -0.0017, ..., -0.0553, -0.0434, -0.0114], [ 0.0856, -0.0152, -0.0590, ..., -0.0521, -0.0728, -0.1117], ..., [-0.0202, 0.0058, -0.0653, ..., -0.0061, 0.0273, 0.0143], [-0.0621, -0.0567, -0.0114, ..., -0.0403, 0.0066, -0.0175], [-0.0148, -0.0317, -0.0123, ..., 0.0558, 0.0238, -0.0666]], device='cuda:0'), grad: tensor([[ 1.5512e-05, 6.7474e-07, -3.4302e-05, ..., -1.1206e-05, 1.1116e-05, 5.4166e-06], [ 1.4730e-05, 3.1032e-06, -3.2574e-05, ..., 7.9215e-05, 1.0673e-06, -2.7925e-05], [-1.3602e-04, 1.3426e-05, 1.4298e-05, ..., 2.5988e-05, 3.0205e-05, 1.9684e-05], ..., [ 3.9250e-05, 7.5176e-06, 7.9721e-06, ..., 7.1466e-05, -2.0608e-05, 1.1191e-05], [ 4.9770e-05, 7.1526e-06, 7.8827e-06, ..., 2.3961e-05, 1.7837e-05, 1.9614e-06], [ 5.3421e-06, 6.1616e-06, 1.0103e-05, ..., -3.6740e-04, -1.0180e-04, 6.1244e-06]], device='cuda:0') Epoch 45, bias, value: tensor([ 0.0078, 0.0171, 0.0140, -0.0026, 0.0171, -0.0065, -0.0158, -0.0106, 0.0245, -0.0090], device='cuda:0'), grad: tensor([ 2.6360e-05, -2.6441e-04, 2.8133e-05, -3.7681e-06, 1.6284e-04, 9.0122e-05, 1.2589e-04, 1.6522e-04, 1.1390e-04, -4.4394e-04], device='cuda:0') 100 0.0001 changing lr epoch 44, time 251.32, cls_loss 0.0099 cls_loss_mapping 0.0215 cls_loss_causal 0.6998 re_mapping 0.0132 re_causal 0.0423 /// teacc 98.83 lr 0.00010000 Epoch 46, weight, value: tensor([[-0.0642, -0.0102, 0.0885, ..., 0.0267, -0.0680, -0.0626], [-0.0477, -0.0564, -0.0016, ..., -0.0561, -0.0442, -0.0125], [ 0.0868, -0.0157, -0.0597, ..., -0.0525, -0.0729, -0.1109], ..., [-0.0213, 0.0055, -0.0656, ..., -0.0065, 0.0275, 0.0136], [-0.0628, -0.0571, -0.0111, ..., -0.0408, 0.0064, -0.0180], [-0.0147, -0.0315, -0.0125, ..., 0.0562, 0.0239, -0.0665]], device='cuda:0'), grad: tensor([[ 1.3731e-05, 5.3123e-06, -3.1054e-05, ..., -2.5593e-06, 3.5137e-05, 1.9178e-05], [ 4.9400e-04, 6.2108e-05, 7.0632e-06, ..., 3.1769e-05, 1.7369e-04, 1.1548e-05], [-1.5569e-04, 1.1176e-04, 2.0444e-05, ..., 1.0118e-05, 2.0576e-04, 2.4036e-05], ..., [-5.4169e-04, -1.8871e-04, 1.6801e-06, ..., -4.4882e-05, -6.2037e-04, 5.7667e-06], [ 1.3423e-04, 8.1122e-05, 2.0161e-05, ..., 1.3232e-05, 8.1480e-05, 3.8117e-05], [ 3.3200e-05, 2.1651e-05, 1.1072e-05, ..., 2.3901e-04, 1.5724e-04, 1.7613e-05]], device='cuda:0') Epoch 46, bias, value: tensor([ 0.0073, 0.0168, 0.0146, -0.0024, 0.0174, -0.0061, -0.0161, -0.0109, 0.0242, -0.0089], device='cuda:0'), grad: tensor([ 8.6188e-05, 7.6723e-04, 2.3425e-04, 3.7646e-04, -1.9717e-04, -3.9434e-04, -2.1148e-04, -1.5202e-03, 3.5119e-04, 5.0783e-04], device='cuda:0') 100 0.0001 changing lr epoch 45, time 251.63, cls_loss 0.0095 cls_loss_mapping 0.0226 cls_loss_causal 0.7181 re_mapping 0.0129 re_causal 0.0437 /// teacc 98.75 lr 0.00010000 Epoch 47, weight, value: tensor([[-0.0654, -0.0102, 0.0888, ..., 0.0264, -0.0683, -0.0628], [-0.0479, -0.0569, -0.0018, ..., -0.0562, -0.0447, -0.0128], [ 0.0875, -0.0161, -0.0602, ..., -0.0530, -0.0734, -0.1119], ..., [-0.0215, 0.0057, -0.0659, ..., -0.0071, 0.0278, 0.0135], [-0.0632, -0.0574, -0.0112, ..., -0.0412, 0.0066, -0.0180], [-0.0138, -0.0316, -0.0119, ..., 0.0571, 0.0240, -0.0670]], device='cuda:0'), grad: tensor([[ 8.9765e-05, 1.1055e-06, -8.6725e-05, ..., -2.3603e-05, 3.5077e-05, 3.2455e-05], [ 2.8074e-05, 3.5204e-06, 1.4611e-05, ..., 4.6968e-05, 4.2945e-05, 1.3322e-05], [-9.0790e-04, 1.2435e-05, 2.2098e-05, ..., -3.9291e-04, -8.2254e-05, 1.1392e-05], ..., [ 3.0923e-04, 2.1551e-06, 1.2316e-05, ..., 2.2054e-04, 8.3447e-05, 7.2382e-06], [-4.7870e-06, -6.6450e-07, 1.9699e-05, ..., -3.8505e-05, -1.3471e-04, -3.0696e-05], [ 2.6274e-04, -4.3362e-06, 3.4541e-05, ..., -1.7118e-04, -1.1730e-04, 8.8215e-06]], device='cuda:0') Epoch 47, bias, value: tensor([ 0.0066, 0.0172, 0.0145, -0.0027, 0.0174, -0.0063, -0.0162, -0.0111, 0.0244, -0.0082], device='cuda:0'), grad: tensor([ 1.1069e-04, 1.6248e-04, -1.3008e-03, 2.2590e-04, 3.6287e-04, 2.4390e-04, -1.9357e-05, 6.5708e-04, -7.4434e-04, 3.0112e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 46---------------------------------------------------- epoch 46, time 268.10, cls_loss 0.0107 cls_loss_mapping 0.0216 cls_loss_causal 0.6859 re_mapping 0.0127 re_causal 0.0399 /// teacc 99.00 lr 0.00010000 Epoch 48, weight, value: tensor([[-0.0661, -0.0103, 0.0898, ..., 0.0272, -0.0689, -0.0632], [-0.0485, -0.0572, -0.0024, ..., -0.0572, -0.0453, -0.0132], [ 0.0883, -0.0163, -0.0608, ..., -0.0536, -0.0740, -0.1126], ..., [-0.0217, 0.0060, -0.0664, ..., -0.0075, 0.0281, 0.0133], [-0.0637, -0.0576, -0.0108, ..., -0.0414, 0.0068, -0.0182], [-0.0137, -0.0322, -0.0126, ..., 0.0574, 0.0241, -0.0679]], device='cuda:0'), grad: tensor([[ 6.8128e-05, 1.6615e-05, -4.9055e-05, ..., -3.2723e-05, 5.6773e-05, 1.5289e-05], [ 9.5218e-06, 2.4922e-06, 6.4857e-06, ..., 3.9428e-05, -1.3733e-03, 2.2836e-06], [ 6.1095e-06, 9.5740e-06, 5.0068e-05, ..., 4.5836e-05, 4.4346e-05, 1.6857e-06], ..., [ 1.0557e-05, -6.5327e-05, 7.0035e-06, ..., 4.8950e-06, 6.0320e-04, 1.2470e-06], [ 8.2493e-05, 2.2903e-05, 8.0585e-05, ..., 1.6439e-04, 1.6332e-04, 9.3877e-06], [ 2.3633e-05, 5.5343e-05, 7.4804e-05, ..., 8.8024e-04, 1.2026e-03, 7.6517e-06]], device='cuda:0') Epoch 48, bias, value: tensor([ 0.0070, 0.0168, 0.0144, -0.0027, 0.0178, -0.0063, -0.0165, -0.0110, 0.0246, -0.0084], device='cuda:0'), grad: tensor([ 1.5163e-04, -1.6891e-02, 2.1470e-04, 3.6573e-04, -1.7185e-03, -3.7789e-05, 8.5890e-05, 1.3008e-02, 7.9966e-04, 4.0169e-03], device='cuda:0') 100 0.0001 changing lr epoch 47, time 251.62, cls_loss 0.0081 cls_loss_mapping 0.0210 cls_loss_causal 0.6867 re_mapping 0.0129 re_causal 0.0407 /// teacc 98.99 lr 0.00010000 Epoch 49, weight, value: tensor([[-0.0666, -0.0105, 0.0902, ..., 0.0274, -0.0694, -0.0641], [-0.0490, -0.0578, -0.0025, ..., -0.0578, -0.0456, -0.0135], [ 0.0890, -0.0169, -0.0613, ..., -0.0540, -0.0744, -0.1133], ..., [-0.0220, 0.0063, -0.0667, ..., -0.0080, 0.0284, 0.0132], [-0.0641, -0.0575, -0.0107, ..., -0.0414, 0.0068, -0.0183], [-0.0142, -0.0327, -0.0131, ..., 0.0573, 0.0241, -0.0683]], device='cuda:0'), grad: tensor([[ 2.2501e-06, 1.8273e-06, -8.1062e-05, ..., 4.6976e-06, 1.7166e-05, 4.2357e-06], [ 1.0662e-05, 8.3372e-06, 1.8924e-05, ..., 2.9594e-05, 2.7075e-05, 1.4016e-06], [-2.5719e-05, 1.3486e-05, 3.0130e-05, ..., 3.4213e-05, 3.1650e-05, 1.6363e-06], ..., [ 1.0453e-05, 3.1013e-06, 1.8537e-05, ..., 1.8549e-04, 1.2207e-04, 3.6322e-06], [ 3.0726e-05, 1.5780e-05, -7.5281e-05, ..., -2.1998e-06, 3.0071e-05, 1.1042e-05], [ 4.4480e-06, -1.5542e-05, 3.8922e-05, ..., 1.7333e-04, -1.8072e-04, -3.1907e-06]], device='cuda:0') Epoch 49, bias, value: tensor([ 0.0067, 0.0169, 0.0145, -0.0031, 0.0182, -0.0059, -0.0163, -0.0113, 0.0249, -0.0089], device='cuda:0'), grad: tensor([ 4.9770e-05, 8.2791e-05, 2.7466e-04, 4.6939e-05, -4.5609e-04, 1.2457e-04, 3.2872e-05, 5.2834e-04, -7.8678e-04, 1.0151e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 48---------------------------------------------------- epoch 48, time 267.60, cls_loss 0.0091 cls_loss_mapping 0.0235 cls_loss_causal 0.7136 re_mapping 0.0121 re_causal 0.0395 /// teacc 99.03 lr 0.00010000 Epoch 50, weight, value: tensor([[-0.0670, -0.0106, 0.0905, ..., 0.0270, -0.0699, -0.0651], [-0.0493, -0.0583, -0.0027, ..., -0.0584, -0.0461, -0.0139], [ 0.0898, -0.0169, -0.0620, ..., -0.0544, -0.0749, -0.1142], ..., [-0.0223, 0.0064, -0.0671, ..., -0.0088, 0.0284, 0.0130], [-0.0648, -0.0581, -0.0109, ..., -0.0425, 0.0061, -0.0184], [-0.0140, -0.0326, -0.0131, ..., 0.0584, 0.0248, -0.0683]], device='cuda:0'), grad: tensor([[ 2.8580e-05, 3.5856e-07, -1.0389e-04, ..., -2.8685e-05, 5.6624e-05, 1.2785e-05], [ 2.2638e-04, 7.7412e-06, 4.8578e-06, ..., 5.2229e-06, 3.7146e-04, 1.8086e-06], [ 8.0585e-05, 1.5989e-05, 2.4021e-05, ..., 1.7926e-05, 2.1422e-04, 1.9670e-05], ..., [ 1.5545e-03, 9.0897e-06, 6.0946e-06, ..., 9.8869e-06, 2.5749e-03, 5.8999e-07], [ 1.7822e-04, 4.1902e-05, 5.1767e-05, ..., 2.3797e-05, 2.6751e-04, 3.8922e-05], [-2.1286e-03, 1.9237e-05, 2.4229e-05, ..., 1.7822e-05, -3.5934e-03, 3.1646e-06]], device='cuda:0') Epoch 50, bias, value: tensor([ 0.0063, 0.0168, 0.0144, -0.0033, 0.0183, -0.0058, -0.0162, -0.0115, 0.0240, -0.0076], device='cuda:0'), grad: tensor([ 0.0001, 0.0016, 0.0010, 0.0002, 0.0005, 0.0001, -0.0001, 0.0120, 0.0012, -0.0166], device='cuda:0') 100 0.0001 changing lr epoch 49, time 248.13, cls_loss 0.0079 cls_loss_mapping 0.0200 cls_loss_causal 0.6615 re_mapping 0.0126 re_causal 0.0383 /// teacc 98.74 lr 0.00010000 Epoch 51, weight, value: tensor([[-0.0676, -0.0107, 0.0910, ..., 0.0272, -0.0705, -0.0658], [-0.0499, -0.0586, -0.0028, ..., -0.0585, -0.0465, -0.0143], [ 0.0906, -0.0171, -0.0626, ..., -0.0547, -0.0752, -0.1151], ..., [-0.0224, 0.0064, -0.0675, ..., -0.0091, 0.0289, 0.0128], [-0.0653, -0.0585, -0.0105, ..., -0.0431, 0.0059, -0.0187], [-0.0141, -0.0327, -0.0136, ..., 0.0586, 0.0247, -0.0688]], device='cuda:0'), grad: tensor([[ 9.9361e-05, 5.0664e-05, 1.1951e-04, ..., 1.1528e-04, 5.0247e-05, 8.1539e-05], [ 5.2899e-05, 4.8459e-05, 4.0084e-05, ..., 5.4777e-05, 1.0043e-04, 2.6274e-04], [ 1.0943e-04, 1.8728e-04, 4.8190e-05, ..., 2.3639e-04, 1.1247e-04, 4.1664e-05], ..., [ 2.0790e-04, 6.5088e-04, 7.0892e-06, ..., 1.9109e-04, 3.6073e-04, 8.5235e-05], [ 1.2591e-05, 1.3912e-04, -2.6512e-04, ..., 8.6784e-05, -2.4199e-04, -3.2425e-04], [-8.4817e-05, 4.9770e-05, 2.7567e-05, ..., -4.9162e-04, -1.4845e-06, 4.4554e-05]], device='cuda:0') Epoch 51, bias, value: tensor([ 0.0062, 0.0170, 0.0144, -0.0033, 0.0178, -0.0055, -0.0159, -0.0109, 0.0237, -0.0079], device='cuda:0'), grad: tensor([ 0.0007, 0.0006, 0.0009, -0.0027, -0.0008, 0.0049, -0.0037, 0.0017, -0.0012, -0.0004], device='cuda:0') 100 0.0001 changing lr epoch 50, time 251.58, cls_loss 0.0115 cls_loss_mapping 0.0264 cls_loss_causal 0.7045 re_mapping 0.0127 re_causal 0.0390 /// teacc 98.98 lr 0.00010000 Epoch 52, weight, value: tensor([[-0.0688, -0.0108, 0.0920, ..., 0.0278, -0.0709, -0.0663], [-0.0508, -0.0590, -0.0028, ..., -0.0583, -0.0470, -0.0146], [ 0.0915, -0.0172, -0.0631, ..., -0.0552, -0.0756, -0.1152], ..., [-0.0221, 0.0066, -0.0675, ..., -0.0094, 0.0293, 0.0124], [-0.0658, -0.0586, -0.0107, ..., -0.0436, 0.0059, -0.0191], [-0.0139, -0.0333, -0.0142, ..., 0.0594, 0.0248, -0.0695]], device='cuda:0'), grad: tensor([[ 1.2666e-06, -8.8513e-06, -2.4962e-04, ..., -5.6028e-05, 2.4050e-05, -6.0201e-05], [ 1.1504e-05, 6.5193e-06, -1.2526e-07, ..., 2.1875e-05, 1.5289e-05, 2.7884e-06], [-3.2693e-05, 1.0490e-05, 9.1642e-06, ..., 9.2760e-06, 1.0915e-05, 3.7458e-06], ..., [ 1.8254e-05, 2.9251e-05, 4.9472e-06, ..., 3.8624e-05, 3.2485e-05, 5.9567e-06], [ 1.5959e-05, 3.0786e-05, 1.8582e-05, ..., 1.5795e-04, 1.0085e-04, 1.4372e-05], [ 7.8678e-06, -4.5508e-05, 3.1382e-05, ..., -1.4620e-03, -7.4434e-04, 2.1428e-05]], device='cuda:0') Epoch 52, bias, value: tensor([ 0.0062, 0.0167, 0.0143, -0.0037, 0.0178, -0.0051, -0.0158, -0.0108, 0.0236, -0.0077], device='cuda:0'), grad: tensor([-2.8443e-04, -8.2016e-05, 6.1244e-06, 2.8658e-04, 6.2466e-04, 8.9884e-04, 4.9695e-06, 1.5199e-04, 2.9612e-04, -1.9026e-03], device='cuda:0') 100 0.0001 changing lr epoch 51, time 251.59, cls_loss 0.0063 cls_loss_mapping 0.0157 cls_loss_causal 0.6804 re_mapping 0.0122 re_causal 0.0390 /// teacc 98.76 lr 0.00010000 Epoch 53, weight, value: tensor([[-0.0692, -0.0109, 0.0927, ..., 0.0277, -0.0714, -0.0666], [-0.0508, -0.0593, -0.0026, ..., -0.0584, -0.0473, -0.0147], [ 0.0917, -0.0179, -0.0637, ..., -0.0556, -0.0760, -0.1155], ..., [-0.0224, 0.0063, -0.0679, ..., -0.0099, 0.0294, 0.0122], [-0.0661, -0.0587, -0.0107, ..., -0.0441, 0.0061, -0.0189], [-0.0142, -0.0337, -0.0147, ..., 0.0598, 0.0247, -0.0700]], device='cuda:0'), grad: tensor([[ 1.0222e-05, 2.6543e-06, -1.1486e-04, ..., -7.6711e-05, 8.9332e-06, 4.6566e-06], [ 2.5868e-05, 3.5018e-06, 6.6608e-06, ..., 8.5682e-06, 1.3277e-05, 5.4054e-06], [-1.2374e-04, 8.1211e-06, 2.4393e-05, ..., 1.1526e-05, 2.5466e-05, 2.0429e-05], ..., [ 1.7866e-05, -2.3931e-05, 5.8115e-06, ..., 2.6882e-05, -4.0799e-05, 8.4005e-07], [ 7.7412e-06, 4.2357e-06, 1.5348e-05, ..., 2.7925e-05, 4.4495e-05, 2.6062e-05], [ 3.2261e-06, 4.2878e-06, 4.4882e-05, ..., -5.8711e-06, -3.1829e-05, 2.6040e-06]], device='cuda:0') Epoch 53, bias, value: tensor([ 0.0060, 0.0174, 0.0138, -0.0030, 0.0183, -0.0057, -0.0160, -0.0111, 0.0239, -0.0080], device='cuda:0'), grad: tensor([-8.8394e-05, 5.4270e-05, -5.3227e-05, 1.5414e-04, -1.3456e-05, -4.4972e-05, -7.2479e-05, -5.2691e-05, 1.1146e-04, 5.5358e-06], device='cuda:0') 100 0.0001 changing lr epoch 52, time 251.75, cls_loss 0.0083 cls_loss_mapping 0.0192 cls_loss_causal 0.6576 re_mapping 0.0113 re_causal 0.0359 /// teacc 98.90 lr 0.00010000 Epoch 54, weight, value: tensor([[-0.0697, -0.0109, 0.0934, ..., 0.0279, -0.0719, -0.0672], [-0.0506, -0.0598, -0.0030, ..., -0.0587, -0.0481, -0.0170], [ 0.0924, -0.0183, -0.0642, ..., -0.0559, -0.0763, -0.1163], ..., [-0.0228, 0.0061, -0.0683, ..., -0.0102, 0.0295, 0.0121], [-0.0671, -0.0592, -0.0107, ..., -0.0450, 0.0062, -0.0182], [-0.0146, -0.0338, -0.0151, ..., 0.0600, 0.0248, -0.0706]], device='cuda:0'), grad: tensor([[ 3.8117e-05, 4.8578e-06, -2.0072e-05, ..., 2.0955e-06, 2.8074e-05, 6.3106e-06], [ 2.1413e-05, 8.1435e-06, -5.2378e-06, ..., 1.0259e-05, 2.0370e-05, 7.4785e-07], [-3.8892e-05, 8.0466e-06, 2.9206e-06, ..., 2.6211e-05, 1.5467e-05, 1.5600e-06], ..., [ 3.5346e-05, 1.9476e-05, 1.3262e-06, ..., 1.5318e-05, 8.8988e-07, 7.2131e-07], [ 3.1322e-05, 2.3991e-06, 4.5933e-06, ..., 4.2439e-05, 3.7014e-05, 1.2778e-06], [-2.2590e-04, 7.1824e-06, 8.2925e-06, ..., -3.5262e-04, -1.1092e-04, 3.1721e-06]], device='cuda:0') Epoch 54, bias, value: tensor([ 0.0060, 0.0172, 0.0138, -0.0031, 0.0182, -0.0054, -0.0159, -0.0113, 0.0243, -0.0081], device='cuda:0'), grad: tensor([ 1.1927e-04, 6.2990e-04, 2.8700e-05, 5.6362e-04, 5.7936e-05, -1.3266e-03, 6.6996e-05, 7.8261e-05, 3.8600e-04, -6.0225e-04], device='cuda:0') 100 0.0001 changing lr epoch 53, time 251.69, cls_loss 0.0082 cls_loss_mapping 0.0217 cls_loss_causal 0.6682 re_mapping 0.0122 re_causal 0.0373 /// teacc 98.99 lr 0.00010000 Epoch 55, weight, value: tensor([[-0.0708, -0.0110, 0.0945, ..., 0.0278, -0.0726, -0.0676], [-0.0509, -0.0600, -0.0045, ..., -0.0593, -0.0486, -0.0174], [ 0.0936, -0.0184, -0.0649, ..., -0.0566, -0.0768, -0.1168], ..., [-0.0233, 0.0056, -0.0689, ..., -0.0112, 0.0295, 0.0118], [-0.0681, -0.0596, -0.0108, ..., -0.0460, 0.0062, -0.0183], [-0.0144, -0.0341, -0.0154, ..., 0.0613, 0.0249, -0.0714]], device='cuda:0'), grad: tensor([[ 1.7628e-05, 1.3094e-06, -6.1274e-04, ..., -3.0065e-04, -1.4573e-05, -2.3320e-05], [ 1.1265e-04, 6.4187e-06, 1.5229e-05, ..., 8.3864e-05, 4.2289e-05, 2.9523e-06], [-2.4402e-04, 1.1869e-05, 6.3777e-05, ..., 4.8220e-05, 1.3061e-05, 7.3016e-06], ..., [ 5.4598e-05, 7.7635e-06, 1.8239e-05, ..., 5.5885e-04, 3.0279e-04, 1.6298e-06], [ 4.3243e-05, 8.1807e-06, 8.2612e-05, ..., 7.4744e-05, 8.0764e-06, -4.0932e-07], [ 2.0951e-05, 9.2238e-06, 8.8394e-05, ..., -6.7568e-04, -4.3201e-04, 7.8529e-06]], device='cuda:0') Epoch 55, bias, value: tensor([ 0.0061, 0.0168, 0.0143, -0.0032, 0.0179, -0.0052, -0.0158, -0.0117, 0.0239, -0.0076], device='cuda:0'), grad: tensor([-8.6689e-04, 1.8728e-04, -1.2189e-04, 3.2544e-05, 8.7321e-05, 1.2052e-04, 3.5524e-04, 1.0147e-03, 2.0599e-04, -1.0157e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 54---------------------------------------------------- epoch 54, time 268.65, cls_loss 0.0087 cls_loss_mapping 0.0201 cls_loss_causal 0.6687 re_mapping 0.0112 re_causal 0.0357 /// teacc 99.06 lr 0.00010000 Epoch 56, weight, value: tensor([[-0.0713, -0.0113, 0.0956, ..., 0.0283, -0.0730, -0.0681], [-0.0517, -0.0605, -0.0046, ..., -0.0590, -0.0491, -0.0176], [ 0.0944, -0.0201, -0.0656, ..., -0.0570, -0.0773, -0.1186], ..., [-0.0237, 0.0051, -0.0694, ..., -0.0122, 0.0296, 0.0117], [-0.0687, -0.0601, -0.0102, ..., -0.0466, 0.0062, -0.0179], [-0.0151, -0.0341, -0.0160, ..., 0.0611, 0.0254, -0.0717]], device='cuda:0'), grad: tensor([[ 9.0972e-06, 8.2841e-07, -1.7345e-04, ..., -5.5313e-05, 7.4767e-06, -2.9989e-07], [ 6.3717e-05, 1.4538e-06, -2.9534e-05, ..., 2.3544e-05, 1.8731e-05, -5.4464e-06], [-1.5080e-04, 7.8008e-06, 2.2188e-05, ..., -1.6531e-07, 1.4387e-05, 3.1423e-06], ..., [ 3.4153e-05, 2.5406e-06, 4.0419e-06, ..., 1.3404e-05, -5.6714e-05, 5.5227e-07], [ 1.9297e-05, 2.8647e-06, 3.4064e-05, ..., 1.6898e-05, 1.4067e-05, 9.6112e-06], [ 5.7966e-06, 2.9858e-06, 4.9531e-05, ..., 3.1829e-05, 2.7597e-05, 2.7306e-06]], device='cuda:0') Epoch 56, bias, value: tensor([ 0.0064, 0.0167, 0.0142, -0.0028, 0.0184, -0.0049, -0.0164, -0.0118, 0.0238, -0.0079], device='cuda:0'), grad: tensor([-1.8322e-04, 3.4809e-05, -8.6546e-05, 5.6982e-05, -1.2118e-04, 1.9825e-04, -3.8117e-05, -1.0133e-04, 1.0967e-04, 1.3030e-04], device='cuda:0') 100 0.0001 changing lr epoch 55, time 251.08, cls_loss 0.0088 cls_loss_mapping 0.0218 cls_loss_causal 0.6582 re_mapping 0.0113 re_causal 0.0363 /// teacc 98.89 lr 0.00010000 Epoch 57, weight, value: tensor([[-0.0718, -0.0114, 0.0964, ..., 0.0289, -0.0739, -0.0696], [-0.0524, -0.0606, -0.0046, ..., -0.0584, -0.0489, -0.0178], [ 0.0952, -0.0203, -0.0668, ..., -0.0578, -0.0778, -0.1194], ..., [-0.0241, 0.0050, -0.0699, ..., -0.0129, 0.0296, 0.0114], [-0.0690, -0.0602, -0.0097, ..., -0.0479, 0.0063, -0.0175], [-0.0154, -0.0342, -0.0170, ..., 0.0612, 0.0256, -0.0720]], device='cuda:0'), grad: tensor([[ 2.4717e-06, 1.9148e-06, -4.3005e-05, ..., -3.9041e-05, 1.0975e-05, 1.0520e-05], [ 4.7907e-06, 2.0321e-06, -3.1263e-05, ..., -2.2173e-05, 9.9689e-06, -1.8533e-06], [ 6.1631e-05, 4.1366e-05, 1.2346e-05, ..., 1.4640e-05, 1.8477e-05, 2.9933e-06], ..., [ 1.7062e-05, 5.4389e-06, 7.5810e-06, ..., 1.2092e-05, -4.8488e-05, 5.5647e-07], [ 2.8923e-05, 1.4007e-05, 2.1532e-05, ..., 2.1890e-05, 1.5751e-05, 1.2547e-05], [ 1.2055e-05, 7.8082e-06, 2.6703e-05, ..., 1.9267e-05, 1.5795e-05, 2.5947e-06]], device='cuda:0') Epoch 57, bias, value: tensor([ 0.0065, 0.0172, 0.0142, -0.0029, 0.0184, -0.0053, -0.0162, -0.0121, 0.0239, -0.0080], device='cuda:0'), grad: tensor([-5.0180e-06, -1.2338e-04, 1.7774e-04, -8.6725e-05, 8.0615e-06, -5.1588e-05, -1.4782e-05, -1.4865e-04, 1.2398e-04, 1.2016e-04], device='cuda:0') 100 0.0001 changing lr epoch 56, time 251.67, cls_loss 0.0065 cls_loss_mapping 0.0163 cls_loss_causal 0.6549 re_mapping 0.0109 re_causal 0.0354 /// teacc 98.83 lr 0.00010000 Epoch 58, weight, value: tensor([[-0.0724, -0.0114, 0.0967, ..., 0.0289, -0.0746, -0.0704], [-0.0533, -0.0609, -0.0045, ..., -0.0588, -0.0497, -0.0196], [ 0.0963, -0.0206, -0.0674, ..., -0.0583, -0.0784, -0.1174], ..., [-0.0246, 0.0049, -0.0704, ..., -0.0132, 0.0299, 0.0103], [-0.0694, -0.0604, -0.0102, ..., -0.0484, 0.0063, -0.0181], [-0.0156, -0.0345, -0.0173, ..., 0.0617, 0.0257, -0.0730]], device='cuda:0'), grad: tensor([[ 4.0948e-05, 1.1273e-05, -1.1826e-04, ..., -6.0648e-05, 1.6034e-05, 6.3516e-06], [ 7.0274e-05, 2.9013e-05, 3.6955e-06, ..., 1.6242e-05, 3.4928e-05, 7.2680e-06], [-7.1859e-04, 7.4923e-05, 1.9893e-05, ..., 1.4350e-05, 8.9765e-05, 9.6858e-06], ..., [ 5.4538e-05, -2.1303e-04, 5.9977e-06, ..., 6.5826e-06, -3.7456e-04, 8.4490e-06], [ 4.1127e-05, 1.0438e-05, 1.4752e-06, ..., 7.9051e-06, 4.4964e-06, -2.3432e-06], [ 3.2043e-04, 1.2629e-05, 4.5270e-05, ..., 3.0965e-05, 2.1651e-05, 1.1452e-05]], device='cuda:0') Epoch 58, bias, value: tensor([ 0.0061, 0.0165, 0.0147, -0.0029, 0.0181, -0.0053, -0.0154, -0.0118, 0.0237, -0.0079], device='cuda:0'), grad: tensor([-2.9653e-06, 2.3222e-04, -7.7820e-04, 4.4733e-05, 1.4603e-04, 3.2258e-04, 1.2493e-04, -7.6437e-04, 6.1333e-05, 6.1464e-04], device='cuda:0') 100 0.0001 changing lr epoch 57, time 251.38, cls_loss 0.0065 cls_loss_mapping 0.0171 cls_loss_causal 0.6743 re_mapping 0.0111 re_causal 0.0355 /// teacc 98.98 lr 0.00010000 Epoch 59, weight, value: tensor([[-0.0733, -0.0115, 0.0970, ..., 0.0292, -0.0754, -0.0716], [-0.0538, -0.0612, -0.0045, ..., -0.0590, -0.0502, -0.0201], [ 0.0974, -0.0207, -0.0672, ..., -0.0586, -0.0788, -0.1182], ..., [-0.0250, 0.0049, -0.0713, ..., -0.0133, 0.0306, 0.0105], [-0.0702, -0.0605, -0.0096, ..., -0.0492, 0.0065, -0.0174], [-0.0161, -0.0347, -0.0179, ..., 0.0617, 0.0254, -0.0747]], device='cuda:0'), grad: tensor([[ 7.2690e-07, 1.3104e-06, -1.3304e-04, ..., -6.8545e-05, 1.9372e-05, -4.2468e-05], [ 1.0738e-06, 1.9558e-06, -6.6273e-06, ..., 4.8816e-05, 2.9072e-05, -9.5461e-09], [-3.6117e-06, 2.5649e-06, 1.2942e-05, ..., 2.3484e-05, 1.1794e-05, 7.6219e-06], ..., [ 1.5292e-06, 1.0884e-04, 1.4812e-05, ..., 6.1655e-04, 6.2990e-04, 5.7258e-06], [ 2.0657e-06, 5.0701e-06, 2.4468e-05, ..., 6.0588e-05, 4.3631e-05, 1.4566e-05], [ 1.3607e-06, -1.1659e-04, 2.5421e-05, ..., -4.9973e-04, -7.0000e-04, 9.7081e-06]], device='cuda:0') Epoch 59, bias, value: tensor([ 0.0059, 0.0162, 0.0151, -0.0023, 0.0182, -0.0060, -0.0152, -0.0115, 0.0239, -0.0083], device='cuda:0'), grad: tensor([-1.1808e-04, 7.7128e-05, 5.5075e-05, 7.5221e-05, -5.1212e-04, 1.2589e-04, 3.1620e-05, 1.3876e-03, 1.7428e-04, -1.2951e-03], device='cuda:0') 100 0.0001 changing lr epoch 58, time 251.40, cls_loss 0.0060 cls_loss_mapping 0.0167 cls_loss_causal 0.6390 re_mapping 0.0112 re_causal 0.0347 /// teacc 98.89 lr 0.00010000 Epoch 60, weight, value: tensor([[-0.0737, -0.0114, 0.0976, ..., 0.0291, -0.0760, -0.0720], [-0.0541, -0.0616, -0.0043, ..., -0.0593, -0.0507, -0.0206], [ 0.0983, -0.0211, -0.0678, ..., -0.0592, -0.0793, -0.1186], ..., [-0.0255, 0.0052, -0.0717, ..., -0.0138, 0.0310, 0.0093], [-0.0707, -0.0608, -0.0099, ..., -0.0498, 0.0063, -0.0180], [-0.0162, -0.0350, -0.0185, ..., 0.0623, 0.0256, -0.0757]], device='cuda:0'), grad: tensor([[ 1.0490e-05, 4.8615e-07, -4.9496e-04, ..., 1.0127e-04, -3.0100e-05, -4.7874e-04], [ 3.8415e-05, 1.4640e-06, 2.8491e-05, ..., 1.8924e-05, 1.9312e-05, 2.3410e-05], [-1.6642e-04, 2.6580e-06, 5.4598e-05, ..., 2.2531e-05, 3.4481e-05, 3.3408e-05], ..., [ 1.7270e-05, 3.8967e-06, 1.8731e-05, ..., 3.5763e-04, 3.2592e-04, 1.4767e-05], [ 5.3883e-05, 2.8871e-06, -1.3649e-05, ..., 1.3478e-05, -3.3677e-05, 5.7936e-05], [ 4.8988e-06, 3.2764e-06, 1.0657e-04, ..., -2.8491e-05, 2.0072e-05, 7.3016e-05]], device='cuda:0') Epoch 60, bias, value: tensor([ 0.0057, 0.0162, 0.0152, -0.0024, 0.0180, -0.0057, -0.0152, -0.0115, 0.0237, -0.0082], device='cuda:0'), grad: tensor([-9.6416e-04, 1.0800e-04, -4.5657e-05, 6.1750e-05, -7.3767e-04, 5.1498e-04, -4.4763e-05, 9.7609e-04, -1.2040e-05, 1.4555e-04], device='cuda:0') 100 0.0001 changing lr epoch 59, time 251.63, cls_loss 0.0058 cls_loss_mapping 0.0152 cls_loss_causal 0.6424 re_mapping 0.0104 re_causal 0.0341 /// teacc 99.03 lr 0.00010000 Epoch 61, weight, value: tensor([[-0.0741, -0.0115, 0.0985, ..., 0.0297, -0.0767, -0.0721], [-0.0546, -0.0618, -0.0047, ..., -0.0599, -0.0512, -0.0212], [ 0.0986, -0.0234, -0.0687, ..., -0.0598, -0.0802, -0.1184], ..., [-0.0256, 0.0052, -0.0722, ..., -0.0141, 0.0315, 0.0089], [-0.0712, -0.0609, -0.0101, ..., -0.0507, 0.0062, -0.0183], [-0.0164, -0.0349, -0.0187, ..., 0.0629, 0.0258, -0.0763]], device='cuda:0'), grad: tensor([[ 3.0883e-06, 1.7837e-05, 5.5999e-05, ..., 1.4506e-05, 3.1084e-05, 2.7210e-05], [ 4.4443e-06, 2.4792e-06, -5.7481e-06, ..., -9.3132e-07, 5.0291e-06, 3.1833e-06], [-2.6256e-05, 9.1344e-06, 3.2753e-05, ..., 1.2368e-05, 1.1854e-05, 8.4490e-06], ..., [ 4.3288e-06, 1.0142e-06, 7.7784e-06, ..., 2.9001e-06, -1.3625e-06, 6.9384e-07], [ 5.2936e-06, 6.9514e-06, 2.2447e-04, ..., -1.8775e-05, 1.2600e-04, 2.3723e-04], [ 8.1025e-07, 1.6363e-06, 7.1704e-05, ..., 3.0264e-05, 6.7770e-05, 6.1952e-06]], device='cuda:0') Epoch 61, bias, value: tensor([ 0.0061, 0.0161, 0.0147, -0.0020, 0.0176, -0.0059, -0.0150, -0.0111, 0.0232, -0.0078], device='cuda:0'), grad: tensor([ 1.5640e-04, -5.5790e-05, 8.1718e-05, 2.2382e-05, 1.3077e-04, 2.6774e-04, -1.0319e-03, 3.0667e-05, 4.3690e-05, 3.5405e-04], device='cuda:0') 100 0.0001 changing lr epoch 60, time 251.39, cls_loss 0.0055 cls_loss_mapping 0.0131 cls_loss_causal 0.6390 re_mapping 0.0103 re_causal 0.0338 /// teacc 98.92 lr 0.00010000 Epoch 62, weight, value: tensor([[-0.0746, -0.0115, 0.0986, ..., 0.0294, -0.0773, -0.0729], [-0.0550, -0.0620, -0.0047, ..., -0.0603, -0.0516, -0.0219], [ 0.0996, -0.0235, -0.0690, ..., -0.0600, -0.0804, -0.1182], ..., [-0.0262, 0.0051, -0.0726, ..., -0.0143, 0.0317, 0.0085], [-0.0720, -0.0612, -0.0102, ..., -0.0510, 0.0061, -0.0184], [-0.0169, -0.0350, -0.0188, ..., 0.0635, 0.0260, -0.0771]], device='cuda:0'), grad: tensor([[-8.8066e-06, -6.7241e-06, -1.7655e-04, ..., -1.0693e-04, 2.2396e-05, 2.5466e-05], [ 1.7332e-06, 2.5220e-06, 9.3505e-06, ..., 3.2056e-06, 2.8268e-05, 2.8238e-05], [ 5.5246e-06, 5.1484e-06, 4.5151e-05, ..., 3.0011e-05, 2.2486e-05, 2.2978e-05], ..., [ 4.8429e-06, -1.3709e-06, 8.2329e-06, ..., 4.9025e-06, -2.8148e-05, 4.0233e-06], [ 3.1311e-06, 4.1202e-06, 3.5465e-05, ..., 5.1141e-05, 5.2094e-05, 4.1693e-05], [ 9.8720e-06, 9.0450e-06, 8.0228e-05, ..., -1.1415e-03, -5.2547e-04, 2.4006e-05]], device='cuda:0') Epoch 62, bias, value: tensor([ 0.0056, 0.0160, 0.0152, -0.0016, 0.0175, -0.0062, -0.0148, -0.0111, 0.0229, -0.0078], device='cuda:0'), grad: tensor([-1.4198e-04, -8.9929e-06, 1.0914e-04, 1.0557e-03, 1.3771e-03, -1.0147e-03, -2.1696e-04, -1.6227e-05, 1.4615e-04, -1.2884e-03], device='cuda:0') 100 0.0001 changing lr epoch 61, time 251.12, cls_loss 0.0059 cls_loss_mapping 0.0171 cls_loss_causal 0.6245 re_mapping 0.0104 re_causal 0.0321 /// teacc 98.99 lr 0.00010000 Epoch 63, weight, value: tensor([[-0.0747, -0.0115, 0.0992, ..., 0.0294, -0.0778, -0.0737], [-0.0551, -0.0623, -0.0047, ..., -0.0609, -0.0522, -0.0226], [ 0.1002, -0.0237, -0.0700, ..., -0.0606, -0.0808, -0.1190], ..., [-0.0264, 0.0051, -0.0732, ..., -0.0147, 0.0321, 0.0081], [-0.0724, -0.0613, -0.0100, ..., -0.0516, 0.0063, -0.0179], [-0.0172, -0.0352, -0.0193, ..., 0.0637, 0.0259, -0.0785]], device='cuda:0'), grad: tensor([[ 3.0443e-05, 9.0292e-07, -1.8273e-06, ..., -1.2569e-05, 1.0572e-05, 1.7598e-05], [ 6.0424e-06, 2.8126e-07, 2.1216e-06, ..., -2.2678e-07, 3.3733e-06, 2.7344e-06], [-8.5735e-04, 4.4564e-07, 1.6496e-05, ..., -2.2605e-05, -1.8090e-05, 4.3698e-06], ..., [ 8.5473e-05, 1.6857e-07, 2.5928e-06, ..., 3.5837e-06, -3.2242e-06, 5.4017e-07], [ 5.0098e-05, 3.4086e-06, 2.0042e-06, ..., 1.5147e-05, -1.7649e-06, 1.2413e-05], [ 3.4928e-04, 5.9791e-07, 1.3337e-05, ..., -1.6734e-05, 1.1124e-05, 5.3458e-06]], device='cuda:0') Epoch 63, bias, value: tensor([ 0.0055, 0.0157, 0.0150, -0.0016, 0.0179, -0.0063, -0.0149, -0.0110, 0.0233, -0.0079], device='cuda:0'), grad: tensor([ 6.7592e-05, -1.3836e-05, -9.7561e-04, 3.6478e-04, 1.0175e-04, 7.2837e-05, -1.1611e-04, 1.0945e-05, 1.9342e-05, 4.6730e-04], device='cuda:0') 100 0.0001 changing lr epoch 62, time 251.56, cls_loss 0.0058 cls_loss_mapping 0.0144 cls_loss_causal 0.6661 re_mapping 0.0101 re_causal 0.0330 /// teacc 98.99 lr 0.00010000 Epoch 64, weight, value: tensor([[-0.0753, -0.0116, 0.1000, ..., 0.0292, -0.0782, -0.0750], [-0.0552, -0.0625, -0.0056, ..., -0.0616, -0.0525, -0.0231], [ 0.1013, -0.0239, -0.0709, ..., -0.0608, -0.0808, -0.1200], ..., [-0.0275, 0.0051, -0.0737, ..., -0.0157, 0.0321, 0.0079], [-0.0732, -0.0613, -0.0098, ..., -0.0523, 0.0062, -0.0177], [-0.0174, -0.0354, -0.0197, ..., 0.0644, 0.0260, -0.0793]], device='cuda:0'), grad: tensor([[ 3.4878e-07, 5.2154e-08, -2.4810e-05, ..., -1.6659e-05, 4.8801e-06, 4.1351e-06], [ 9.9465e-07, 2.3982e-07, 1.3150e-06, ..., 9.0245e-07, 6.9737e-06, 1.1260e-06], [ 4.6566e-07, 2.7474e-07, 3.2485e-06, ..., 4.5002e-06, 2.7299e-05, 1.4696e-06], ..., [-2.5351e-06, 2.5611e-07, 1.8505e-06, ..., -5.0627e-06, -7.5042e-05, 3.9721e-07], [-5.8971e-06, 2.9337e-07, 3.7514e-06, ..., 3.7365e-06, -2.9638e-05, 3.4962e-06], [ 4.9919e-06, 3.7253e-07, 1.1146e-05, ..., 2.6766e-06, 4.4137e-05, 2.4121e-06]], device='cuda:0') Epoch 64, bias, value: tensor([ 0.0050, 0.0157, 0.0154, -0.0013, 0.0179, -0.0068, -0.0148, -0.0115, 0.0235, -0.0076], device='cuda:0'), grad: tensor([-1.3471e-05, 6.9812e-06, 9.5904e-05, 4.7147e-05, 1.5974e-05, 2.6338e-06, -7.9116e-07, -1.6165e-04, -1.5259e-04, 1.5950e-04], device='cuda:0') 100 0.0001 changing lr epoch 63, time 251.38, cls_loss 0.0058 cls_loss_mapping 0.0172 cls_loss_causal 0.6736 re_mapping 0.0102 re_causal 0.0326 /// teacc 98.89 lr 0.00010000 Epoch 65, weight, value: tensor([[-0.0763, -0.0117, 0.1003, ..., 0.0293, -0.0787, -0.0758], [-0.0559, -0.0627, -0.0058, ..., -0.0615, -0.0525, -0.0236], [ 0.1022, -0.0241, -0.0713, ..., -0.0612, -0.0813, -0.1208], ..., [-0.0278, 0.0048, -0.0740, ..., -0.0162, 0.0323, 0.0077], [-0.0737, -0.0615, -0.0100, ..., -0.0531, 0.0060, -0.0177], [-0.0174, -0.0355, -0.0202, ..., 0.0644, 0.0258, -0.0798]], device='cuda:0'), grad: tensor([[ 2.3282e-04, 1.5413e-07, -5.6314e-04, ..., -9.0075e-04, 2.4974e-05, 2.4587e-05], [ 5.1372e-06, 8.8848e-07, 8.4043e-06, ..., 6.8136e-06, 9.6560e-06, 5.8301e-06], [-4.5919e-04, 1.2098e-06, -2.3377e-04, ..., -1.1635e-04, 7.3351e-06, 5.1707e-06], ..., [ 5.1782e-06, 1.2340e-07, 8.0392e-06, ..., 1.1832e-05, -1.3687e-05, 6.3609e-07], [ 1.3351e-05, 1.1800e-06, 1.6451e-05, ..., 1.9804e-05, -6.0769e-07, -7.3425e-06], [ 8.8960e-06, 4.5337e-06, 6.9714e-04, ..., 8.7309e-04, -6.7987e-06, 5.1893e-06]], device='cuda:0') Epoch 65, bias, value: tensor([ 0.0048, 0.0156, 0.0156, -0.0009, 0.0183, -0.0064, -0.0146, -0.0118, 0.0233, -0.0080], device='cuda:0'), grad: tensor([-1.0233e-03, 1.0200e-05, -5.9509e-04, 2.3174e-04, 4.3333e-05, 6.2943e-05, -7.3969e-05, 2.0921e-05, -5.9903e-06, 1.3285e-03], device='cuda:0') 100 0.0001 changing lr epoch 64, time 251.62, cls_loss 0.0048 cls_loss_mapping 0.0139 cls_loss_causal 0.5864 re_mapping 0.0101 re_causal 0.0305 /// teacc 98.75 lr 0.00010000 Epoch 66, weight, value: tensor([[-0.0775, -0.0117, 0.1003, ..., 0.0298, -0.0792, -0.0778], [-0.0562, -0.0627, -0.0058, ..., -0.0616, -0.0531, -0.0238], [ 0.1028, -0.0244, -0.0710, ..., -0.0616, -0.0818, -0.1217], ..., [-0.0280, 0.0048, -0.0746, ..., -0.0164, 0.0329, 0.0080], [-0.0741, -0.0618, -0.0102, ..., -0.0536, 0.0057, -0.0183], [-0.0178, -0.0358, -0.0209, ..., 0.0646, 0.0258, -0.0809]], device='cuda:0'), grad: tensor([[-3.2922e-07, 2.1942e-06, -1.7971e-05, ..., -2.2113e-05, 7.1190e-06, 5.0217e-06], [ 5.1362e-07, -7.3671e-05, 1.1679e-06, ..., -1.2480e-07, -5.0336e-05, -1.1593e-04], [-1.1690e-05, 1.7449e-05, 3.2127e-05, ..., 1.0088e-05, 2.5526e-05, 3.1531e-05], ..., [ 1.8505e-06, 1.4305e-05, 3.2410e-06, ..., 8.7917e-06, -8.7097e-06, 2.2188e-05], [ 1.1977e-06, 9.1046e-06, -5.3495e-05, ..., 3.8780e-06, -4.1388e-06, 1.9103e-05], [ 6.7055e-07, 1.2126e-06, 2.0817e-05, ..., -1.2122e-05, -3.4869e-06, 2.1998e-06]], device='cuda:0') Epoch 66, bias, value: tensor([ 0.0042, 0.0155, 0.0156, -0.0010, 0.0181, -0.0060, -0.0141, -0.0110, 0.0230, -0.0083], device='cuda:0'), grad: tensor([-9.1642e-06, -5.7030e-04, 1.8191e-04, 1.8215e-04, 2.6152e-05, 7.7128e-05, 9.9391e-06, 7.8917e-05, -1.7527e-06, 2.5317e-05], device='cuda:0') 100 0.0001 changing lr epoch 65, time 251.58, cls_loss 0.0057 cls_loss_mapping 0.0139 cls_loss_causal 0.6449 re_mapping 0.0104 re_causal 0.0311 /// teacc 98.85 lr 0.00010000 Epoch 67, weight, value: tensor([[-0.0781, -0.0118, 0.1013, ..., 0.0305, -0.0792, -0.0778], [-0.0569, -0.0628, -0.0059, ..., -0.0618, -0.0533, -0.0242], [ 0.1039, -0.0247, -0.0716, ..., -0.0618, -0.0823, -0.1228], ..., [-0.0285, 0.0050, -0.0751, ..., -0.0166, 0.0332, 0.0077], [-0.0746, -0.0623, -0.0097, ..., -0.0544, 0.0056, -0.0183], [-0.0180, -0.0359, -0.0216, ..., 0.0647, 0.0259, -0.0816]], device='cuda:0'), grad: tensor([[ 2.8517e-06, 1.5963e-06, -1.9431e-05, ..., -1.0371e-05, 2.6226e-06, 2.2147e-06], [ 4.9734e-04, 9.1083e-07, 3.0883e-06, ..., -6.3069e-06, 1.5758e-06, -2.8200e-06], [-5.7650e-04, 7.5856e-07, 3.7365e-06, ..., 4.7088e-06, 1.1437e-06, -5.1688e-08], ..., [ 4.4346e-05, 3.4422e-06, 1.2685e-06, ..., 8.8811e-06, 3.9451e-06, 2.3376e-06], [ 9.8050e-06, 1.9409e-06, 1.3979e-06, ..., 9.0674e-06, 1.5441e-06, -2.4326e-06], [ 1.3914e-06, 3.5316e-06, 3.9376e-06, ..., -1.0394e-05, -7.2569e-06, 2.0321e-06]], device='cuda:0') Epoch 67, bias, value: tensor([ 0.0047, 0.0153, 0.0162, -0.0018, 0.0181, -0.0061, -0.0142, -0.0113, 0.0234, -0.0083], device='cuda:0'), grad: tensor([ 2.2322e-05, 1.6451e-03, -7.4816e-04, 3.8815e-04, 1.5751e-05, 1.4573e-05, 2.9057e-05, -1.4458e-03, 4.8697e-05, 3.2455e-05], device='cuda:0') 100 0.0001 changing lr epoch 66, time 251.57, cls_loss 0.0065 cls_loss_mapping 0.0158 cls_loss_causal 0.6482 re_mapping 0.0093 re_causal 0.0289 /// teacc 99.01 lr 0.00010000 Epoch 68, weight, value: tensor([[-0.0788, -0.0119, 0.1025, ..., 0.0310, -0.0798, -0.0781], [-0.0573, -0.0630, -0.0059, ..., -0.0638, -0.0540, -0.0246], [ 0.1040, -0.0250, -0.0729, ..., -0.0631, -0.0827, -0.1228], ..., [-0.0273, 0.0048, -0.0755, ..., -0.0168, 0.0335, 0.0073], [-0.0753, -0.0624, -0.0101, ..., -0.0553, 0.0056, -0.0184], [-0.0183, -0.0364, -0.0223, ..., 0.0653, 0.0260, -0.0822]], device='cuda:0'), grad: tensor([[ 2.0042e-06, 1.6615e-06, -1.2740e-05, ..., -6.2808e-06, 6.2361e-06, 3.9600e-06], [ 3.1069e-06, 1.0887e-06, -4.4107e-05, ..., 2.0694e-06, 2.5965e-06, -1.1399e-05], [-2.3544e-05, 3.1237e-06, 5.0925e-06, ..., 2.8938e-05, 1.0118e-05, 1.9111e-06], ..., [ 5.6103e-06, 3.2395e-05, 3.6806e-06, ..., 1.4842e-05, 4.0114e-05, 1.9781e-06], [ 6.5677e-06, 1.6568e-06, 1.2442e-05, ..., 6.8173e-06, -1.7524e-05, -3.2615e-06], [ 1.4668e-06, 3.5819e-06, 8.9481e-06, ..., -5.9530e-06, -8.1658e-06, 2.6003e-06]], device='cuda:0') Epoch 68, bias, value: tensor([ 0.0055, 0.0144, 0.0151, -0.0017, 0.0179, -0.0062, -0.0147, -0.0099, 0.0233, -0.0080], device='cuda:0'), grad: tensor([ 1.5929e-05, -1.1420e-04, 6.0111e-05, -1.7154e-04, -2.4462e-04, 5.8353e-05, 2.3234e-04, 1.8013e-04, -3.0145e-05, 1.3635e-05], device='cuda:0') 100 0.0001 changing lr epoch 67, time 251.62, cls_loss 0.0043 cls_loss_mapping 0.0133 cls_loss_causal 0.6305 re_mapping 0.0098 re_causal 0.0313 /// teacc 98.93 lr 0.00010000 Epoch 69, weight, value: tensor([[-0.0796, -0.0120, 0.1031, ..., 0.0315, -0.0804, -0.0786], [-0.0578, -0.0632, -0.0067, ..., -0.0634, -0.0541, -0.0250], [ 0.1047, -0.0251, -0.0732, ..., -0.0636, -0.0829, -0.1237], ..., [-0.0276, 0.0046, -0.0758, ..., -0.0175, 0.0338, 0.0069], [-0.0755, -0.0624, -0.0103, ..., -0.0560, 0.0055, -0.0185], [-0.0185, -0.0366, -0.0227, ..., 0.0652, 0.0258, -0.0829]], device='cuda:0'), grad: tensor([[-7.6415e-07, 1.1986e-06, 1.3037e-03, ..., -2.2078e-04, 1.8463e-05, 1.2369e-03], [ 9.3281e-06, 3.7625e-06, 1.2875e-05, ..., 4.8466e-06, 5.4948e-06, 7.4692e-06], [-8.6278e-06, 6.9514e-06, 7.5996e-05, ..., 3.0309e-05, 7.7114e-06, 2.3231e-05], ..., [ 1.2524e-05, -6.8434e-06, 1.2152e-05, ..., 2.7299e-05, -4.2588e-05, 5.4017e-06], [ 1.3128e-05, 2.9616e-06, 1.8883e-04, ..., 7.0035e-05, 8.1882e-06, 3.9458e-05], [ 3.7700e-06, 4.2729e-06, 1.3614e-04, ..., -2.7686e-05, -1.0550e-05, 9.6038e-06]], device='cuda:0') Epoch 69, bias, value: tensor([ 0.0056, 0.0145, 0.0153, -0.0021, 0.0183, -0.0057, -0.0146, -0.0099, 0.0232, -0.0087], device='cuda:0'), grad: tensor([ 1.6851e-03, 2.0564e-05, 1.4067e-04, 7.3016e-05, 3.4511e-05, 3.6478e-04, -2.6703e-03, -6.8545e-05, 3.1567e-04, 1.0341e-04], device='cuda:0') 100 0.0001 changing lr epoch 68, time 251.48, cls_loss 0.0044 cls_loss_mapping 0.0112 cls_loss_causal 0.6032 re_mapping 0.0096 re_causal 0.0305 /// teacc 98.98 lr 0.00010000 Epoch 70, weight, value: tensor([[-0.0801, -0.0121, 0.1038, ..., 0.0321, -0.0809, -0.0794], [-0.0578, -0.0632, -0.0069, ..., -0.0637, -0.0543, -0.0250], [ 0.1055, -0.0252, -0.0741, ..., -0.0641, -0.0832, -0.1244], ..., [-0.0279, 0.0043, -0.0761, ..., -0.0178, 0.0340, 0.0065], [-0.0769, -0.0623, -0.0092, ..., -0.0572, 0.0059, -0.0170], [-0.0188, -0.0368, -0.0230, ..., 0.0656, 0.0260, -0.0837]], device='cuda:0'), grad: tensor([[ 2.0154e-06, 8.6613e-08, -5.0306e-05, ..., -2.8849e-05, 1.0744e-05, 1.6272e-05], [ 2.6137e-05, 1.1688e-07, 2.3022e-06, ..., 1.7762e-05, 6.0558e-05, 9.1642e-06], [-1.6585e-05, 3.2829e-07, 3.8259e-06, ..., 1.7419e-05, 1.8179e-05, 1.6605e-06], ..., [-4.5925e-05, 2.7195e-07, 3.0380e-06, ..., 2.0832e-05, -3.4571e-05, 3.4161e-06], [ 1.1005e-05, 5.3551e-08, 2.6256e-05, ..., 2.3976e-05, -7.3731e-05, -3.6985e-05], [ 1.8608e-06, 3.1665e-07, 1.1235e-05, ..., 7.8440e-04, 4.2129e-04, 1.1235e-05]], device='cuda:0') Epoch 70, bias, value: tensor([ 0.0059, 0.0151, 0.0153, -0.0016, 0.0181, -0.0060, -0.0155, -0.0099, 0.0231, -0.0088], device='cuda:0'), grad: tensor([-5.8502e-05, 1.9646e-04, 4.1127e-05, 9.4652e-05, -1.1053e-03, 3.2693e-05, 2.1383e-05, -1.1939e-04, -2.5368e-04, 1.1511e-03], device='cuda:0') 100 0.0001 changing lr epoch 69, time 251.68, cls_loss 0.0046 cls_loss_mapping 0.0135 cls_loss_causal 0.6688 re_mapping 0.0090 re_causal 0.0303 /// teacc 98.93 lr 0.00010000 Epoch 71, weight, value: tensor([[-0.0806, -0.0121, 0.1045, ..., 0.0324, -0.0811, -0.0796], [-0.0586, -0.0634, -0.0067, ..., -0.0637, -0.0550, -0.0256], [ 0.1066, -0.0253, -0.0743, ..., -0.0646, -0.0835, -0.1244], ..., [-0.0277, 0.0042, -0.0764, ..., -0.0185, 0.0341, 0.0061], [-0.0783, -0.0625, -0.0095, ..., -0.0583, 0.0058, -0.0171], [-0.0191, -0.0369, -0.0233, ..., 0.0657, 0.0259, -0.0846]], device='cuda:0'), grad: tensor([[ 5.7258e-06, 2.8983e-06, -4.1723e-06, ..., -2.7064e-06, 1.0252e-05, 3.4049e-06], [ 2.6062e-05, 1.1511e-06, 3.7067e-07, ..., 1.4640e-06, 5.2899e-06, 2.2445e-06], [-1.0031e-04, 1.7080e-06, 1.5572e-06, ..., 2.4904e-06, 4.2506e-06, 1.7397e-06], ..., [ 2.1651e-05, 1.8654e-06, 6.7009e-07, ..., 7.8678e-06, 5.8822e-06, 2.4289e-06], [ 9.4324e-06, -1.4174e-04, -3.1352e-05, ..., 2.0832e-05, -3.7622e-04, -4.6551e-05], [ 2.7474e-06, 2.4792e-06, 2.3991e-06, ..., -6.3777e-05, -4.5866e-05, 3.5372e-06]], device='cuda:0') Epoch 71, bias, value: tensor([ 0.0061, 0.0148, 0.0158, -0.0018, 0.0185, -0.0060, -0.0154, -0.0097, 0.0227, -0.0090], device='cuda:0'), grad: tensor([ 3.3945e-05, -7.3385e-04, -8.2403e-06, 2.2459e-04, 9.0301e-05, 7.5674e-04, 1.1528e-04, 1.7035e-04, -5.9843e-04, -5.1320e-05], device='cuda:0') 100 0.0001 changing lr epoch 70, time 251.68, cls_loss 0.0043 cls_loss_mapping 0.0119 cls_loss_causal 0.6433 re_mapping 0.0093 re_causal 0.0295 /// teacc 98.97 lr 0.00010000 Epoch 72, weight, value: tensor([[-0.0811, -0.0123, 0.1052, ..., 0.0325, -0.0816, -0.0798], [-0.0591, -0.0639, -0.0065, ..., -0.0637, -0.0553, -0.0256], [ 0.1072, -0.0258, -0.0751, ..., -0.0654, -0.0840, -0.1251], ..., [-0.0275, 0.0041, -0.0770, ..., -0.0189, 0.0340, 0.0058], [-0.0795, -0.0627, -0.0098, ..., -0.0593, 0.0057, -0.0172], [-0.0194, -0.0371, -0.0237, ..., 0.0665, 0.0262, -0.0853]], device='cuda:0'), grad: tensor([[ 3.7625e-06, 2.9989e-07, 3.8236e-05, ..., 2.7148e-07, 5.1796e-05, 8.4639e-05], [ 5.7928e-06, 3.6322e-07, 4.2990e-06, ..., 1.7118e-06, 1.4961e-04, 8.8289e-06], [-2.1040e-05, 2.3656e-07, 5.5060e-06, ..., 1.6876e-06, 1.5616e-05, 2.4870e-05], ..., [-2.6114e-06, 2.5285e-07, 2.0824e-06, ..., 2.7135e-05, -1.5771e-04, 2.0731e-06], [ 1.4929e-06, 5.4296e-07, 1.2182e-05, ..., 4.0710e-05, 5.7220e-05, 1.0937e-05], [ 2.9337e-06, 2.1374e-07, 2.7809e-06, ..., -1.1343e-04, -8.8930e-05, 4.9025e-06]], device='cuda:0') Epoch 72, bias, value: tensor([ 0.0060, 0.0150, 0.0157, -0.0011, 0.0180, -0.0062, -0.0151, -0.0098, 0.0221, -0.0086], device='cuda:0'), grad: tensor([ 1.5283e-04, 2.4128e-04, 7.5638e-05, 4.7415e-05, 2.6584e-04, 9.5844e-05, -4.6945e-04, -3.2163e-04, 7.0691e-05, -1.5783e-04], device='cuda:0') 100 0.0001 changing lr epoch 71, time 251.59, cls_loss 0.0040 cls_loss_mapping 0.0118 cls_loss_causal 0.6051 re_mapping 0.0097 re_causal 0.0292 /// teacc 99.05 lr 0.00010000 Epoch 73, weight, value: tensor([[-0.0817, -0.0123, 0.1055, ..., 0.0326, -0.0824, -0.0808], [-0.0599, -0.0641, -0.0064, ..., -0.0639, -0.0559, -0.0263], [ 0.1087, -0.0260, -0.0756, ..., -0.0657, -0.0842, -0.1247], ..., [-0.0281, 0.0040, -0.0772, ..., -0.0199, 0.0341, 0.0055], [-0.0800, -0.0629, -0.0100, ..., -0.0601, 0.0055, -0.0173], [-0.0197, -0.0373, -0.0242, ..., 0.0670, 0.0267, -0.0865]], device='cuda:0'), grad: tensor([[ 1.9297e-06, 9.7789e-07, -1.2470e-06, ..., 7.6694e-07, 5.7258e-06, 1.5730e-06], [ 2.3264e-06, 8.8522e-07, -1.2770e-05, ..., -1.0535e-05, 2.3842e-06, 1.5032e-06], [ 7.6294e-06, 7.3574e-06, 6.5863e-06, ..., 2.7381e-06, 1.9476e-05, 5.9754e-06], ..., [ 8.4192e-06, -1.8049e-06, 1.3523e-06, ..., 7.6056e-05, 3.0324e-05, 1.1744e-06], [ 3.4217e-06, 1.0021e-06, 1.5628e-06, ..., 5.6289e-06, -5.7310e-05, -3.7193e-05], [ 2.7139e-06, 2.4252e-06, 2.1104e-06, ..., -9.6500e-05, -5.5224e-05, 1.3383e-06]], device='cuda:0') Epoch 73, bias, value: tensor([ 0.0057, 0.0149, 0.0164, -0.0017, 0.0177, -0.0060, -0.0148, -0.0101, 0.0219, -0.0082], device='cuda:0'), grad: tensor([ 2.4244e-05, -6.4671e-05, 9.1493e-05, -8.7202e-05, 3.7998e-05, 2.4259e-04, 2.0072e-05, 1.3137e-04, -2.3413e-04, -1.6165e-04], device='cuda:0') 100 0.0001 changing lr epoch 72, time 251.50, cls_loss 0.0039 cls_loss_mapping 0.0123 cls_loss_causal 0.6096 re_mapping 0.0094 re_causal 0.0299 /// teacc 98.95 lr 0.00010000 Epoch 74, weight, value: tensor([[-0.0820, -0.0123, 0.1061, ..., 0.0331, -0.0828, -0.0809], [-0.0600, -0.0642, -0.0063, ..., -0.0638, -0.0562, -0.0275], [ 0.1092, -0.0262, -0.0762, ..., -0.0660, -0.0846, -0.1243], ..., [-0.0283, 0.0038, -0.0775, ..., -0.0201, 0.0344, 0.0053], [-0.0806, -0.0633, -0.0101, ..., -0.0607, 0.0054, -0.0171], [-0.0203, -0.0374, -0.0246, ..., 0.0670, 0.0267, -0.0872]], device='cuda:0'), grad: tensor([[ 5.5693e-06, 3.1758e-07, -3.1404e-06, ..., 1.4615e-04, 3.1322e-05, 6.9812e-06], [ 3.8743e-06, 6.1560e-07, -4.5784e-06, ..., 2.9225e-06, 2.9534e-05, 4.5419e-05], [-9.1255e-05, 1.6177e-06, 1.0587e-05, ..., 1.1265e-05, 1.1660e-05, 1.3903e-05], ..., [ 4.4018e-05, -1.4305e-06, 1.4249e-06, ..., 1.5378e-05, -1.5646e-05, 2.1644e-06], [ 6.2510e-06, -2.5099e-07, 9.8124e-06, ..., 8.9481e-06, -3.2067e-05, -6.0529e-05], [ 6.9439e-06, 9.4296e-07, 4.8615e-06, ..., -2.2268e-04, -2.9534e-05, 3.2149e-06]], device='cuda:0') Epoch 74, bias, value: tensor([ 0.0060, 0.0149, 0.0165, -0.0013, 0.0179, -0.0061, -0.0150, -0.0101, 0.0218, -0.0086], device='cuda:0'), grad: tensor([ 3.0756e-04, 5.3835e-04, -5.8442e-05, 4.7714e-05, 6.4373e-05, 1.6260e-04, -3.1024e-05, 7.3373e-05, -7.4196e-04, -3.6144e-04], device='cuda:0') 100 0.0001 changing lr epoch 73, time 251.60, cls_loss 0.0042 cls_loss_mapping 0.0107 cls_loss_causal 0.5900 re_mapping 0.0092 re_causal 0.0281 /// teacc 98.89 lr 0.00010000 Epoch 75, weight, value: tensor([[-0.0825, -0.0124, 0.1066, ..., 0.0334, -0.0836, -0.0816], [-0.0599, -0.0645, -0.0061, ..., -0.0640, -0.0566, -0.0283], [ 0.1099, -0.0264, -0.0768, ..., -0.0668, -0.0853, -0.1247], ..., [-0.0287, 0.0037, -0.0779, ..., -0.0201, 0.0348, 0.0052], [-0.0815, -0.0635, -0.0102, ..., -0.0612, 0.0052, -0.0170], [-0.0206, -0.0377, -0.0252, ..., 0.0672, 0.0269, -0.0879]], device='cuda:0'), grad: tensor([[ 5.6764e-07, 5.4762e-07, -2.1458e-05, ..., -2.2873e-05, 2.0210e-06, 1.1073e-06], [ 2.4866e-06, 1.9502e-06, 1.0040e-06, ..., 1.4305e-06, 1.7118e-06, -4.6231e-06], [ 2.8927e-06, 4.0196e-06, 3.0212e-06, ..., 3.7123e-06, 3.1069e-06, 1.7891e-06], ..., [ 4.8131e-06, 3.8259e-06, 1.3970e-06, ..., 7.5139e-06, 7.5214e-06, 7.1488e-06], [ 3.5055e-06, 2.1718e-06, 1.6270e-06, ..., 3.1609e-06, 2.7139e-06, 1.4203e-06], [ 6.2101e-06, 5.0738e-06, 9.5144e-06, ..., 2.8666e-06, -4.7218e-07, 2.7716e-06]], device='cuda:0') Epoch 75, bias, value: tensor([ 0.0058, 0.0159, 0.0162, -0.0015, 0.0177, -0.0060, -0.0147, -0.0103, 0.0216, -0.0087], device='cuda:0'), grad: tensor([-3.0547e-05, -8.6367e-05, 2.0415e-05, -5.3763e-05, -9.2462e-06, 2.2098e-05, -1.3318e-06, 9.7871e-05, 1.4901e-05, 2.5988e-05], device='cuda:0') 100 0.0001 changing lr epoch 74, time 251.42, cls_loss 0.0055 cls_loss_mapping 0.0136 cls_loss_causal 0.6052 re_mapping 0.0091 re_causal 0.0282 /// teacc 98.90 lr 0.00010000 Epoch 76, weight, value: tensor([[-0.0832, -0.0124, 0.1074, ..., 0.0337, -0.0845, -0.0823], [-0.0605, -0.0643, -0.0057, ..., -0.0642, -0.0574, -0.0287], [ 0.1105, -0.0267, -0.0774, ..., -0.0673, -0.0861, -0.1251], ..., [-0.0286, 0.0036, -0.0784, ..., -0.0195, 0.0358, 0.0047], [-0.0824, -0.0637, -0.0103, ..., -0.0619, 0.0054, -0.0166], [-0.0210, -0.0378, -0.0259, ..., 0.0673, 0.0266, -0.0897]], device='cuda:0'), grad: tensor([[ 7.8082e-06, 4.6333e-07, -3.7923e-06, ..., 1.0771e-04, 3.3945e-05, 2.0526e-06], [ 3.7067e-06, 1.1818e-06, -2.1327e-06, ..., 1.7602e-06, 3.7905e-06, -2.8824e-07], [-3.7134e-05, 2.7083e-06, 1.1427e-06, ..., 3.4347e-06, 2.4531e-06, 3.8929e-06], ..., [ 1.9625e-05, 2.4475e-06, 4.5588e-07, ..., 9.1866e-06, 7.1675e-06, 2.7753e-06], [ 1.0766e-05, 2.1737e-06, 2.1402e-06, ..., 7.7337e-06, 3.7774e-06, -2.6096e-06], [ 7.6257e-06, 2.9542e-06, 1.5898e-06, ..., -1.3900e-04, -3.7819e-05, 4.2170e-06]], device='cuda:0') Epoch 76, bias, value: tensor([ 0.0061, 0.0151, 0.0159, -0.0016, 0.0175, -0.0059, -0.0146, -0.0088, 0.0219, -0.0095], device='cuda:0'), grad: tensor([ 2.1088e-04, -3.0056e-05, -2.9296e-05, 1.1331e-04, 2.4125e-05, -2.1005e-04, 3.0324e-05, 7.9155e-05, 2.9519e-05, -2.1756e-04], device='cuda:0') 100 0.0001 changing lr epoch 75, time 251.12, cls_loss 0.0055 cls_loss_mapping 0.0139 cls_loss_causal 0.6210 re_mapping 0.0094 re_causal 0.0279 /// teacc 98.99 lr 0.00010000 Epoch 77, weight, value: tensor([[-0.0835, -0.0125, 0.1074, ..., 0.0337, -0.0853, -0.0831], [-0.0616, -0.0645, -0.0060, ..., -0.0647, -0.0581, -0.0293], [ 0.1113, -0.0269, -0.0783, ..., -0.0679, -0.0868, -0.1254], ..., [-0.0287, 0.0033, -0.0788, ..., -0.0199, 0.0362, 0.0041], [-0.0827, -0.0640, -0.0096, ..., -0.0613, 0.0049, -0.0169], [-0.0212, -0.0380, -0.0262, ..., 0.0675, 0.0266, -0.0902]], device='cuda:0'), grad: tensor([[ 2.7306e-06, 5.6950e-07, -8.6129e-06, ..., -4.1611e-06, 4.4815e-06, 1.9614e-06], [ 1.3001e-06, 4.4517e-07, -6.8918e-08, ..., 4.4797e-07, 1.1884e-05, 6.2864e-07], [-2.5686e-06, 8.0001e-07, 1.0114e-06, ..., 3.7672e-07, 7.9945e-06, 2.6431e-06], ..., [ 4.2915e-06, 1.2377e-06, 2.4121e-07, ..., 8.4117e-06, -6.5207e-05, 1.9046e-06], [ 2.5615e-05, 5.5544e-06, 1.6270e-06, ..., 2.6617e-06, 6.5804e-05, 2.2888e-05], [ 7.2867e-06, 2.1756e-06, 1.5255e-06, ..., -1.7732e-05, 1.3947e-05, 6.2883e-06]], device='cuda:0') Epoch 77, bias, value: tensor([ 0.0055, 0.0151, 0.0160, -0.0028, 0.0178, -0.0050, -0.0142, -0.0091, 0.0221, -0.0095], device='cuda:0'), grad: tensor([ 2.0023e-06, 1.4871e-05, 1.9923e-05, 6.4659e-03, 1.7151e-05, -6.5765e-03, 3.7309e-06, -1.5354e-04, 1.6153e-04, 4.3839e-05], device='cuda:0') 100 0.0001 changing lr epoch 76, time 250.82, cls_loss 0.0042 cls_loss_mapping 0.0121 cls_loss_causal 0.6225 re_mapping 0.0087 re_causal 0.0275 /// teacc 99.00 lr 0.00010000 Epoch 78, weight, value: tensor([[-0.0841, -0.0125, 0.1080, ..., 0.0341, -0.0859, -0.0833], [-0.0616, -0.0647, -0.0062, ..., -0.0648, -0.0587, -0.0300], [ 0.1121, -0.0271, -0.0786, ..., -0.0682, -0.0878, -0.1266], ..., [-0.0293, 0.0031, -0.0791, ..., -0.0204, 0.0365, 0.0050], [-0.0834, -0.0643, -0.0098, ..., -0.0622, 0.0049, -0.0167], [-0.0215, -0.0382, -0.0267, ..., 0.0683, 0.0271, -0.0909]], device='cuda:0'), grad: tensor([[ 2.3827e-05, 3.5437e-07, -1.1124e-05, ..., -6.1989e-06, 7.4953e-06, -1.9446e-06], [ 1.0528e-05, 2.1309e-06, 1.3104e-06, ..., 2.6058e-06, 4.4331e-06, -8.2981e-07], [-5.1880e-04, -1.4566e-05, 1.2137e-05, ..., 6.9551e-06, 1.2249e-05, 2.4494e-06], ..., [ 3.7169e-04, 4.0606e-06, 7.2410e-07, ..., 4.8518e-05, 1.6754e-06, 1.0310e-06], [ 8.1360e-05, 1.4968e-05, 8.3521e-06, ..., 5.3458e-06, -5.8934e-06, 3.1609e-06], [ 3.3509e-06, 3.1926e-06, 2.4941e-06, ..., 1.1109e-05, 1.5661e-05, 1.2862e-06]], device='cuda:0') Epoch 78, bias, value: tensor([ 0.0058, 0.0151, 0.0160, -0.0035, 0.0172, -0.0045, -0.0139, -0.0094, 0.0222, -0.0092], device='cuda:0'), grad: tensor([ 3.6836e-05, -7.7114e-06, -6.7711e-04, -1.0088e-05, -1.1081e-04, 5.1051e-05, -3.9995e-05, 5.9748e-04, 9.4712e-05, 6.6042e-05], device='cuda:0') 100 0.0001 changing lr epoch 77, time 251.11, cls_loss 0.0043 cls_loss_mapping 0.0115 cls_loss_causal 0.6136 re_mapping 0.0084 re_causal 0.0269 /// teacc 98.93 lr 0.00010000 Epoch 79, weight, value: tensor([[-0.0852, -0.0127, 0.1081, ..., 0.0342, -0.0865, -0.0840], [-0.0620, -0.0650, -0.0060, ..., -0.0641, -0.0589, -0.0319], [ 0.1132, -0.0272, -0.0789, ..., -0.0687, -0.0883, -0.1270], ..., [-0.0300, 0.0029, -0.0794, ..., -0.0214, 0.0363, 0.0048], [-0.0840, -0.0648, -0.0104, ..., -0.0637, 0.0043, -0.0171], [-0.0214, -0.0383, -0.0270, ..., 0.0687, 0.0271, -0.0919]], device='cuda:0'), grad: tensor([[-7.3649e-06, 6.5193e-08, -1.6022e-04, ..., -6.0529e-05, 2.3078e-06, -2.0638e-05], [ 4.6939e-06, 3.6694e-07, 3.7886e-06, ..., -1.2279e-05, 6.4850e-05, -3.7968e-05], [ 4.1015e-06, 7.5717e-07, 8.0541e-06, ..., 7.8678e-06, 1.9699e-05, 5.4911e-06], ..., [-1.9237e-05, 4.0978e-07, 2.5071e-06, ..., -2.8927e-06, -1.1933e-04, 3.1237e-06], [ 1.3355e-06, 1.3597e-07, 1.1586e-05, ..., 1.4752e-05, -3.6173e-06, 7.0632e-06], [ 7.2680e-06, 6.8545e-07, 2.0668e-05, ..., 1.2062e-05, 1.8030e-05, 8.6203e-06]], device='cuda:0') Epoch 79, bias, value: tensor([ 0.0055, 0.0154, 0.0164, -0.0033, 0.0174, -0.0046, -0.0127, -0.0103, 0.0213, -0.0090], device='cuda:0'), grad: tensor([-1.3852e-04, -2.8834e-05, 1.3077e-04, 1.1426e-04, 4.1872e-05, 8.0705e-05, 8.8990e-05, -5.2929e-04, 9.4771e-05, 1.4532e-04], device='cuda:0') 100 0.0001 changing lr epoch 78, time 251.11, cls_loss 0.0055 cls_loss_mapping 0.0140 cls_loss_causal 0.6175 re_mapping 0.0086 re_causal 0.0274 /// teacc 98.94 lr 0.00010000 Epoch 80, weight, value: tensor([[-0.0859, -0.0128, 0.1087, ..., 0.0344, -0.0869, -0.0843], [-0.0611, -0.0651, -0.0056, ..., -0.0652, -0.0600, -0.0314], [ 0.1135, -0.0274, -0.0793, ..., -0.0682, -0.0888, -0.1280], ..., [-0.0308, 0.0027, -0.0797, ..., -0.0217, 0.0370, 0.0051], [-0.0844, -0.0652, -0.0106, ..., -0.0642, 0.0042, -0.0170], [-0.0219, -0.0385, -0.0274, ..., 0.0699, 0.0272, -0.0930]], device='cuda:0'), grad: tensor([[ 3.8967e-06, 6.5677e-06, -2.4855e-05, ..., -1.0118e-05, 5.3644e-06, 9.0152e-06], [ 3.5353e-06, 5.2340e-06, -3.5793e-05, ..., -2.6718e-05, 4.9658e-06, -2.7686e-05], [ 2.4103e-06, 2.6878e-06, 7.9572e-06, ..., 4.5709e-06, 9.7305e-06, 5.7034e-06], ..., [-1.3737e-08, 2.1700e-06, 4.1462e-06, ..., 6.9216e-06, -5.8189e-06, 1.8151e-06], [ 6.5789e-06, 1.0870e-05, 1.8880e-05, ..., 8.5011e-06, 5.8599e-06, 9.7305e-06], [ 1.1578e-05, 1.8582e-05, 2.4199e-05, ..., 6.5446e-05, 4.0859e-05, 1.2174e-05]], device='cuda:0') Epoch 80, bias, value: tensor([ 0.0055, 0.0158, 0.0160, -0.0023, 0.0166, -0.0052, -0.0130, -0.0099, 0.0212, -0.0086], device='cuda:0'), grad: tensor([ 4.0010e-06, -1.1671e-04, 5.1856e-05, -5.2404e-04, -7.6294e-05, 3.3212e-04, 6.9618e-05, -2.1420e-06, 7.6711e-05, 1.8513e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 79---------------------------------------------------- epoch 79, time 268.73, cls_loss 0.0046 cls_loss_mapping 0.0108 cls_loss_causal 0.6410 re_mapping 0.0084 re_causal 0.0272 /// teacc 99.08 lr 0.00010000 Epoch 81, weight, value: tensor([[-0.0859, -0.0130, 0.1092, ..., 0.0338, -0.0876, -0.0848], [-0.0621, -0.0654, -0.0056, ..., -0.0655, -0.0608, -0.0315], [ 0.1142, -0.0276, -0.0798, ..., -0.0691, -0.0893, -0.1285], ..., [-0.0301, 0.0024, -0.0801, ..., -0.0220, 0.0375, 0.0046], [-0.0850, -0.0653, -0.0103, ..., -0.0649, 0.0043, -0.0167], [-0.0234, -0.0390, -0.0274, ..., 0.0701, 0.0268, -0.0945]], device='cuda:0'), grad: tensor([[ 3.5875e-06, 6.5798e-07, -3.1024e-05, ..., -1.2882e-05, 8.9705e-06, 5.0198e-07], [-1.0651e-04, 1.1092e-06, 4.3842e-07, ..., 3.1382e-05, -1.8513e-04, 2.5192e-07], [ 3.3677e-05, 7.0147e-06, 1.3625e-06, ..., 1.3545e-05, 8.6546e-05, 4.0862e-07], ..., [ 1.4551e-05, 1.3970e-06, 5.2992e-07, ..., 9.1195e-05, 7.6771e-05, 1.9418e-07], [ 8.3089e-05, 2.2307e-05, 3.6061e-06, ..., 9.6142e-05, 1.7011e-04, 1.1669e-06], [ 4.6045e-06, 2.4121e-06, 1.7360e-05, ..., -2.9731e-04, -2.2590e-04, 6.0536e-07]], device='cuda:0') Epoch 81, bias, value: tensor([ 0.0049, 0.0152, 0.0159, -0.0022, 0.0174, -0.0052, -0.0134, -0.0093, 0.0215, -0.0089], device='cuda:0'), grad: tensor([ 3.2149e-06, -1.3905e-03, 5.5218e-04, -5.2243e-05, 1.1981e-04, 5.9068e-05, 1.7059e-04, 3.2783e-04, 9.3603e-04, -7.2622e-04], device='cuda:0') 100 0.0001 changing lr epoch 80, time 251.49, cls_loss 0.0039 cls_loss_mapping 0.0101 cls_loss_causal 0.6096 re_mapping 0.0089 re_causal 0.0266 /// teacc 99.06 lr 0.00010000 Epoch 82, weight, value: tensor([[-0.0862, -0.0131, 0.1097, ..., 0.0341, -0.0883, -0.0851], [-0.0630, -0.0657, -0.0051, ..., -0.0650, -0.0613, -0.0319], [ 0.1153, -0.0280, -0.0804, ..., -0.0693, -0.0898, -0.1288], ..., [-0.0299, 0.0019, -0.0804, ..., -0.0222, 0.0379, 0.0044], [-0.0853, -0.0655, -0.0105, ..., -0.0654, 0.0042, -0.0166], [-0.0250, -0.0394, -0.0280, ..., 0.0703, 0.0269, -0.0953]], device='cuda:0'), grad: tensor([[ 3.8669e-06, 1.5404e-06, 4.8801e-07, ..., 1.2303e-06, 2.9728e-06, 1.9027e-06], [ 1.2323e-05, 4.6454e-06, 1.4063e-07, ..., -3.6787e-06, 4.0904e-06, 3.0752e-06], [ 2.7016e-05, 1.2167e-05, 7.7020e-07, ..., 1.4743e-06, 2.1122e-06, 1.3104e-06], ..., [ 1.4059e-05, 5.1931e-06, 6.1654e-07, ..., 2.6952e-06, 1.6270e-06, 1.1139e-06], [ 5.0403e-06, 2.2706e-06, 5.5581e-06, ..., 1.9819e-06, 1.4231e-05, 1.4223e-05], [ 8.3148e-06, 3.3304e-06, 1.9297e-06, ..., 3.1758e-06, 3.2820e-06, 2.0117e-06]], device='cuda:0') Epoch 82, bias, value: tensor([ 0.0049, 0.0150, 0.0160, -0.0019, 0.0170, -0.0055, -0.0131, -0.0087, 0.0216, -0.0093], device='cuda:0'), grad: tensor([ 1.4551e-05, -9.2089e-06, 6.8069e-05, -1.9276e-04, -7.9453e-05, 1.0294e-04, -4.1090e-06, 4.8220e-05, 1.9923e-05, 3.1710e-05], device='cuda:0') 100 0.0001 changing lr epoch 81, time 251.73, cls_loss 0.0041 cls_loss_mapping 0.0121 cls_loss_causal 0.5936 re_mapping 0.0082 re_causal 0.0254 /// teacc 99.05 lr 0.00010000 Epoch 83, weight, value: tensor([[-0.0868, -0.0132, 0.1101, ..., 0.0344, -0.0892, -0.0860], [-0.0633, -0.0660, -0.0052, ..., -0.0652, -0.0616, -0.0321], [ 0.1159, -0.0282, -0.0810, ..., -0.0696, -0.0903, -0.1297], ..., [-0.0303, 0.0015, -0.0808, ..., -0.0224, 0.0381, 0.0042], [-0.0856, -0.0658, -0.0108, ..., -0.0666, 0.0035, -0.0168], [-0.0251, -0.0395, -0.0281, ..., 0.0705, 0.0270, -0.0959]], device='cuda:0'), grad: tensor([[ 7.4878e-06, 3.3062e-08, -5.2862e-06, ..., -2.3283e-06, 2.6971e-06, 5.7183e-07], [ 4.2021e-05, 8.4750e-08, 4.3167e-07, ..., 4.1984e-06, 2.7269e-05, 3.1590e-06], [-8.7357e-04, 1.0757e-07, -4.7646e-06, ..., 1.2740e-06, 9.4026e-06, -6.7532e-05], ..., [ 3.5971e-05, 1.9558e-07, 1.9409e-06, ..., 7.2047e-06, -1.1218e-04, 7.8231e-07], [ 7.5674e-04, 3.6601e-07, 2.5779e-06, ..., 1.4186e-05, 2.0891e-05, 6.3002e-05], [ 4.7348e-06, 1.9744e-07, 1.8617e-06, ..., -1.0687e-04, 4.1351e-06, 1.5534e-06]], device='cuda:0') Epoch 83, bias, value: tensor([ 0.0049, 0.0152, 0.0159, -0.0010, 0.0167, -0.0053, -0.0134, -0.0087, 0.0209, -0.0093], device='cuda:0'), grad: tensor([ 1.7315e-05, 1.7273e-04, -1.6432e-03, 1.1557e-04, 7.9989e-05, 2.9221e-05, 1.6317e-05, -3.2687e-04, 1.5306e-03, 6.9179e-06], device='cuda:0') 100 0.0001 changing lr epoch 82, time 251.31, cls_loss 0.0047 cls_loss_mapping 0.0118 cls_loss_causal 0.6014 re_mapping 0.0083 re_causal 0.0259 /// teacc 99.00 lr 0.00010000 Epoch 84, weight, value: tensor([[-0.0875, -0.0134, 0.1107, ..., 0.0347, -0.0902, -0.0866], [-0.0637, -0.0662, -0.0057, ..., -0.0653, -0.0620, -0.0321], [ 0.1168, -0.0283, -0.0820, ..., -0.0708, -0.0907, -0.1307], ..., [-0.0308, 0.0015, -0.0810, ..., -0.0231, 0.0382, 0.0039], [-0.0862, -0.0662, -0.0107, ..., -0.0673, 0.0037, -0.0165], [-0.0252, -0.0396, -0.0283, ..., 0.0697, 0.0268, -0.0966]], device='cuda:0'), grad: tensor([[ 5.5544e-06, -4.0699e-07, 1.4102e-04, ..., -2.2054e-05, 1.1569e-04, 1.1331e-04], [-1.4842e-05, 1.5683e-06, 8.4788e-06, ..., 1.1079e-05, 1.4424e-05, 9.6634e-06], [-1.3018e-04, 1.2899e-06, 2.0728e-05, ..., -7.8082e-05, 1.8492e-05, 1.5065e-05], ..., [ 7.9989e-05, 1.7332e-06, 5.5209e-06, ..., 7.9274e-05, 8.9332e-06, 3.0492e-06], [-4.6864e-06, 4.0010e-06, 2.3052e-05, ..., 1.1764e-05, -2.5153e-05, -1.8418e-05], [ 2.9400e-05, 2.6245e-06, 1.2226e-05, ..., 4.8250e-05, 1.1221e-05, 9.2685e-06]], device='cuda:0') Epoch 84, bias, value: tensor([ 0.0047, 0.0150, 0.0160, -0.0013, 0.0186, -0.0051, -0.0139, -0.0087, 0.0213, -0.0106], device='cuda:0'), grad: tensor([ 3.5691e-04, -1.0121e-04, -8.8632e-05, 5.1928e-04, 2.0182e-04, -4.1032e-04, -7.6389e-04, 2.2554e-04, -7.6771e-05, 1.3757e-04], device='cuda:0') 100 0.0001 changing lr epoch 83, time 251.59, cls_loss 0.0043 cls_loss_mapping 0.0096 cls_loss_causal 0.5985 re_mapping 0.0082 re_causal 0.0255 /// teacc 99.06 lr 0.00010000 Epoch 85, weight, value: tensor([[-0.0883, -0.0136, 0.1109, ..., 0.0350, -0.0913, -0.0878], [-0.0644, -0.0666, -0.0055, ..., -0.0646, -0.0624, -0.0327], [ 0.1172, -0.0285, -0.0828, ..., -0.0712, -0.0912, -0.1313], ..., [-0.0304, 0.0011, -0.0815, ..., -0.0237, 0.0384, 0.0039], [-0.0866, -0.0663, -0.0106, ..., -0.0677, 0.0037, -0.0164], [-0.0254, -0.0399, -0.0286, ..., 0.0700, 0.0268, -0.0971]], device='cuda:0'), grad: tensor([[ 1.1418e-06, -1.0030e-06, -4.1515e-05, ..., -2.3887e-05, 1.4352e-06, 2.0601e-06], [ 1.1109e-05, 2.5332e-07, 1.3169e-06, ..., -6.4522e-06, 1.4324e-06, 1.7723e-06], [-5.0336e-05, 3.3807e-07, 8.5756e-06, ..., 5.4017e-06, 1.4687e-06, 1.7649e-06], ..., [ 1.3821e-05, 4.2934e-07, 3.9339e-06, ..., 1.0878e-05, 1.5553e-07, 1.8720e-07], [ 1.4193e-05, 4.4983e-07, 8.3447e-06, ..., 6.2659e-06, 4.0159e-06, 7.4282e-06], [ 1.1194e-06, 5.6624e-07, 6.5118e-06, ..., -8.9779e-06, -5.1782e-06, 7.4180e-07]], device='cuda:0') Epoch 85, bias, value: tensor([ 0.0043, 0.0141, 0.0157, -0.0014, 0.0184, -0.0052, -0.0133, -0.0075, 0.0215, -0.0107], device='cuda:0'), grad: tensor([-4.1664e-05, -5.8025e-05, -1.2958e-04, 1.6332e-05, 8.6725e-06, 7.2382e-06, 8.6725e-06, 9.3997e-05, 7.5996e-05, 1.8299e-05], device='cuda:0') 100 0.0001 changing lr epoch 84, time 251.42, cls_loss 0.0031 cls_loss_mapping 0.0107 cls_loss_causal 0.6105 re_mapping 0.0083 re_causal 0.0258 /// teacc 98.98 lr 0.00010000 Epoch 86, weight, value: tensor([[-0.0886, -0.0137, 0.1111, ..., 0.0352, -0.0919, -0.0890], [-0.0652, -0.0667, -0.0058, ..., -0.0649, -0.0630, -0.0328], [ 0.1183, -0.0289, -0.0827, ..., -0.0715, -0.0916, -0.1321], ..., [-0.0311, 0.0006, -0.0821, ..., -0.0240, 0.0387, 0.0036], [-0.0869, -0.0666, -0.0109, ..., -0.0682, 0.0038, -0.0164], [-0.0256, -0.0400, -0.0289, ..., 0.0703, 0.0268, -0.0981]], device='cuda:0'), grad: tensor([[ 2.9840e-06, 1.0151e-06, -2.2337e-05, ..., -7.1041e-06, 1.5087e-07, -1.9539e-06], [ 1.7779e-06, 6.4913e-07, 1.6689e-06, ..., 7.8231e-06, 2.5213e-05, 8.3297e-06], [ 6.3255e-06, 2.5388e-06, 3.3882e-06, ..., 5.8440e-07, 1.4668e-06, 2.7893e-07], ..., [ 3.4198e-06, 1.2061e-06, 1.1884e-06, ..., -2.5257e-05, -8.4817e-05, -2.0444e-05], [ 4.4052e-07, 1.7509e-07, -1.1817e-05, ..., 9.3924e-07, -4.1686e-06, -7.3574e-06], [ 5.9977e-07, 2.7055e-07, 8.0839e-06, ..., 1.6108e-05, 5.1886e-05, 1.4476e-05]], device='cuda:0') Epoch 86, bias, value: tensor([ 0.0042, 0.0147, 0.0164, -0.0010, 0.0183, -0.0054, -0.0130, -0.0087, 0.0213, -0.0107], device='cuda:0'), grad: tensor([-9.8422e-06, 6.3539e-05, 2.7373e-05, -1.6630e-05, 2.4773e-06, -7.0110e-06, 3.1590e-05, -1.7965e-04, -5.2810e-05, 1.4067e-04], device='cuda:0') 100 0.0001 changing lr epoch 85, time 251.53, cls_loss 0.0043 cls_loss_mapping 0.0127 cls_loss_causal 0.6067 re_mapping 0.0082 re_causal 0.0260 /// teacc 99.00 lr 0.00010000 Epoch 87, weight, value: tensor([[-0.0889, -0.0137, 0.1128, ..., 0.0344, -0.0922, -0.0891], [-0.0654, -0.0669, -0.0079, ..., -0.0656, -0.0638, -0.0340], [ 0.1188, -0.0291, -0.0839, ..., -0.0722, -0.0919, -0.1326], ..., [-0.0315, 0.0005, -0.0830, ..., -0.0242, 0.0392, 0.0038], [-0.0871, -0.0668, -0.0109, ..., -0.0691, 0.0038, -0.0162], [-0.0259, -0.0402, -0.0289, ..., 0.0715, 0.0267, -0.0992]], device='cuda:0'), grad: tensor([[ 1.8766e-07, 3.6787e-08, -4.5188e-06, ..., -2.5816e-06, 5.7034e-06, 1.3746e-06], [ 1.4026e-06, 9.5461e-08, -2.3656e-07, ..., 2.6021e-06, 1.4305e-05, 2.4494e-07], [-1.1642e-08, 6.7754e-07, 9.8124e-06, ..., 2.7623e-06, 1.3456e-05, 1.2657e-06], ..., [ 2.4755e-06, 2.0023e-07, 7.9116e-07, ..., -4.5240e-05, -1.9741e-04, 1.5832e-07], [-1.2312e-06, 2.0536e-07, -1.2897e-05, ..., 2.8256e-06, -3.3975e-05, -1.9576e-06], [ 7.8650e-07, 3.7812e-07, 2.3637e-06, ..., 1.6600e-05, 9.8407e-05, 6.1188e-07]], device='cuda:0') Epoch 87, bias, value: tensor([ 0.0042, 0.0141, 0.0162, -0.0010, 0.0180, -0.0054, -0.0134, -0.0086, 0.0220, -0.0100], device='cuda:0'), grad: tensor([ 8.4639e-06, 2.1487e-05, 8.2374e-05, -7.3127e-06, 1.8787e-04, 6.7234e-05, 1.1079e-05, -4.9162e-04, -1.4377e-04, 2.6393e-04], device='cuda:0') 100 0.0001 changing lr epoch 86, time 251.50, cls_loss 0.0052 cls_loss_mapping 0.0129 cls_loss_causal 0.5978 re_mapping 0.0081 re_causal 0.0245 /// teacc 98.83 lr 0.00010000 Epoch 88, weight, value: tensor([[-9.0728e-02, -1.3838e-02, 1.1302e-01, ..., 3.4492e-02, -9.2871e-02, -8.9492e-02], [-6.4390e-02, -6.7070e-02, -7.1980e-03, ..., -6.6131e-02, -6.4682e-02, -3.3944e-02], [ 1.1879e-01, -2.9280e-02, -8.4015e-02, ..., -7.2789e-02, -9.2746e-02, -1.3326e-01], ..., [-3.1956e-02, 8.8850e-05, -8.3897e-02, ..., -2.3905e-02, 3.9172e-02, 3.4793e-03], [-8.7722e-02, -6.6979e-02, -1.0803e-02, ..., -6.9290e-02, 3.7055e-03, -1.6317e-02], [-2.6299e-02, -4.0429e-02, -2.9215e-02, ..., 7.2029e-02, 2.8190e-02, -9.9485e-02]], device='cuda:0'), grad: tensor([[ 7.9162e-07, 8.0653e-07, 1.1122e-04, ..., 9.9838e-06, 8.4341e-05, 8.7738e-05], [ 2.2780e-06, 2.2575e-06, 7.4804e-06, ..., 3.4496e-06, 1.2428e-05, 7.9945e-06], [ 5.2415e-06, 5.0217e-06, 4.9807e-06, ..., 1.8086e-06, 1.0796e-05, 3.4962e-06], ..., [ 9.0823e-06, 7.9423e-06, 2.4345e-06, ..., 1.0625e-05, -3.1173e-05, 1.8682e-06], [ 3.1684e-06, 3.3192e-06, 8.1509e-06, ..., 1.1235e-05, 8.6054e-06, 1.4259e-06], [ 1.4873e-06, 1.5153e-06, 8.5905e-06, ..., -7.4863e-05, -4.3750e-05, 2.2016e-06]], device='cuda:0') Epoch 88, bias, value: tensor([ 0.0037, 0.0153, 0.0146, -0.0011, 0.0173, -0.0057, -0.0133, -0.0086, 0.0222, -0.0092], device='cuda:0'), grad: tensor([ 1.9717e-04, 2.7746e-05, 5.8681e-05, -5.2124e-05, 1.8680e-04, 8.8394e-05, -3.4356e-04, -8.3089e-05, 2.8908e-05, -1.0902e-04], device='cuda:0') 100 0.0001 changing lr epoch 87, time 251.68, cls_loss 0.0027 cls_loss_mapping 0.0090 cls_loss_causal 0.5971 re_mapping 0.0083 re_causal 0.0255 /// teacc 98.95 lr 0.00010000 Epoch 89, weight, value: tensor([[-9.1141e-02, -1.3863e-02, 1.1343e-01, ..., 3.4667e-02, -9.3379e-02, -8.9823e-02], [-6.4809e-02, -6.7402e-02, -7.2452e-03, ..., -6.6407e-02, -6.5195e-02, -3.4276e-02], [ 1.1990e-01, -2.9514e-02, -8.4112e-02, ..., -7.2821e-02, -9.3190e-02, -1.3319e-01], ..., [-3.2554e-02, -1.0247e-04, -8.4402e-02, ..., -2.4411e-02, 3.9205e-02, 3.0725e-03], [-8.8292e-02, -6.7388e-02, -1.0793e-02, ..., -6.9799e-02, 3.6850e-03, -1.6147e-02], [-2.6505e-02, -4.0651e-02, -2.9509e-02, ..., 7.2278e-02, 2.8510e-02, -9.9802e-02]], device='cuda:0'), grad: tensor([[ 4.0196e-06, 7.7020e-07, 8.2096e-07, ..., 1.5944e-06, 2.8126e-06, 1.2498e-06], [ 1.2154e-06, 3.2363e-07, 2.1188e-07, ..., 7.4282e-06, 9.1270e-06, 2.5835e-06], [-7.8917e-05, -6.1244e-06, 3.8138e-07, ..., 3.5167e-06, 3.9749e-06, 7.9675e-07], ..., [ 6.6571e-06, 6.3190e-07, 3.3528e-08, ..., 5.3011e-06, -4.5449e-06, 1.6550e-06], [ 2.1100e-05, 3.8110e-06, 4.0643e-06, ..., 6.7316e-06, 1.2688e-05, 6.3628e-06], [ 4.1304e-07, 8.8010e-07, 1.6019e-07, ..., 6.1095e-05, 6.4671e-05, 2.1726e-05]], device='cuda:0') Epoch 89, bias, value: tensor([ 0.0038, 0.0151, 0.0153, -0.0002, 0.0172, -0.0066, -0.0134, -0.0090, 0.0225, -0.0090], device='cuda:0'), grad: tensor([ 1.6078e-05, 2.0847e-05, -9.9838e-05, 9.9957e-05, -4.9400e-04, 6.8367e-05, 1.1748e-04, -1.3635e-05, 7.0751e-05, 2.1327e-04], device='cuda:0') 100 0.0001 changing lr epoch 88, time 251.68, cls_loss 0.0032 cls_loss_mapping 0.0096 cls_loss_causal 0.6074 re_mapping 0.0077 re_causal 0.0249 /// teacc 98.98 lr 0.00010000 Epoch 90, weight, value: tensor([[-0.0919, -0.0140, 0.1140, ..., 0.0348, -0.0948, -0.0903], [-0.0650, -0.0676, -0.0078, ..., -0.0666, -0.0658, -0.0345], [ 0.1207, -0.0297, -0.0848, ..., -0.0729, -0.0939, -0.1329], ..., [-0.0328, -0.0002, -0.0850, ..., -0.0248, 0.0397, 0.0028], [-0.0889, -0.0679, -0.0109, ..., -0.0705, 0.0037, -0.0163], [-0.0253, -0.0408, -0.0284, ..., 0.0731, 0.0293, -0.1000]], device='cuda:0'), grad: tensor([[ 1.7174e-06, 4.2189e-07, -2.7448e-05, ..., 1.0371e-05, 6.4000e-06, 3.2391e-06], [ 8.6986e-07, 1.1558e-06, 1.2973e-06, ..., 4.1500e-06, 3.3062e-06, 2.5053e-06], [ 6.0797e-06, 7.2643e-07, 1.2569e-05, ..., 6.5982e-05, 1.4611e-05, 1.1278e-06], ..., [ 2.2184e-06, 6.7577e-06, 3.3900e-06, ..., 1.7077e-05, 1.8194e-05, 4.6864e-06], [ 1.3057e-06, 8.6203e-06, 5.1707e-06, ..., 1.2219e-05, 1.6063e-05, 2.4751e-05], [ 3.7216e-06, 4.7907e-06, 1.4044e-05, ..., 3.1918e-05, 1.1772e-05, 1.0327e-05]], device='cuda:0') Epoch 90, bias, value: tensor([ 0.0036, 0.0149, 0.0154, -0.0002, 0.0170, -0.0081, -0.0136, -0.0088, 0.0226, -0.0079], device='cuda:0'), grad: tensor([ 1.2785e-05, 1.7151e-05, 1.4961e-04, -1.8549e-04, -3.7336e-04, -1.1957e-04, 1.4031e-04, 1.0121e-04, 1.2469e-04, 1.3280e-04], device='cuda:0') 100 0.0001 changing lr epoch 89, time 251.30, cls_loss 0.0032 cls_loss_mapping 0.0088 cls_loss_causal 0.5751 re_mapping 0.0077 re_causal 0.0243 /// teacc 98.95 lr 0.00010000 Epoch 91, weight, value: tensor([[-0.0927, -0.0140, 0.1147, ..., 0.0353, -0.0966, -0.0915], [-0.0653, -0.0679, -0.0072, ..., -0.0662, -0.0665, -0.0347], [ 0.1219, -0.0300, -0.0859, ..., -0.0732, -0.0943, -0.1335], ..., [-0.0336, -0.0006, -0.0857, ..., -0.0253, 0.0394, 0.0025], [-0.0892, -0.0684, -0.0112, ..., -0.0713, 0.0033, -0.0165], [-0.0254, -0.0412, -0.0291, ..., 0.0730, 0.0295, -0.1008]], device='cuda:0'), grad: tensor([[ 9.0338e-08, 3.9116e-08, -5.1409e-06, ..., -5.0925e-06, 1.5516e-06, 1.6792e-06], [ 2.3935e-07, 2.3190e-07, 3.4999e-06, ..., 1.0310e-06, 2.5220e-06, 2.4401e-06], [-7.2271e-07, 4.9733e-07, 5.2191e-06, ..., 1.2424e-06, 3.1199e-06, 3.2093e-06], ..., [ 3.1702e-06, 3.9004e-06, 1.0906e-06, ..., 1.4091e-06, -3.5893e-06, 2.6133e-06], [ 1.0217e-06, 5.6624e-07, 4.7311e-06, ..., 3.4124e-06, 3.1516e-06, 2.5723e-06], [ 4.5355e-07, 8.5402e-07, 1.6838e-06, ..., -4.1723e-07, 3.5632e-06, 4.8801e-07]], device='cuda:0') Epoch 91, bias, value: tensor([ 0.0039, 0.0149, 0.0156, -0.0008, 0.0175, -0.0075, -0.0136, -0.0094, 0.0222, -0.0077], device='cuda:0'), grad: tensor([-3.8594e-06, -6.2212e-06, 1.6525e-05, -1.6794e-05, -1.7677e-06, 8.1211e-06, -3.7432e-05, 6.6981e-06, 1.9699e-05, 1.4953e-05], device='cuda:0') 100 0.0001 changing lr epoch 90, time 251.54, cls_loss 0.0045 cls_loss_mapping 0.0116 cls_loss_causal 0.5618 re_mapping 0.0080 re_causal 0.0238 /// teacc 98.99 lr 0.00010000 Epoch 92, weight, value: tensor([[-0.0931, -0.0141, 0.1153, ..., 0.0355, -0.0971, -0.0918], [-0.0658, -0.0681, -0.0073, ..., -0.0677, -0.0680, -0.0350], [ 0.1230, -0.0301, -0.0866, ..., -0.0735, -0.0948, -0.1341], ..., [-0.0341, -0.0008, -0.0863, ..., -0.0249, 0.0399, 0.0025], [-0.0894, -0.0687, -0.0117, ..., -0.0719, 0.0030, -0.0167], [-0.0262, -0.0415, -0.0295, ..., 0.0723, 0.0289, -0.1013]], device='cuda:0'), grad: tensor([[ 1.7509e-06, 9.3132e-07, 1.8969e-05, ..., -1.9856e-06, 2.4438e-05, 3.3408e-05], [ 1.6382e-06, 1.2498e-06, -2.9914e-06, ..., 4.0047e-06, 2.3283e-06, 1.6056e-06], [-1.5438e-05, 2.4401e-06, 3.4850e-06, ..., 3.0771e-06, 3.2354e-06, 3.0473e-06], ..., [ 3.3136e-06, 3.0193e-06, 7.8138e-07, ..., 7.6070e-06, 2.7250e-06, 9.9279e-07], [ 1.0066e-05, 2.1420e-06, 1.8045e-05, ..., 2.8089e-06, 1.2688e-05, 1.9878e-05], [ 1.7295e-06, 2.9430e-06, 5.5768e-06, ..., -2.5719e-05, -7.5512e-06, 4.2357e-06]], device='cuda:0') Epoch 92, bias, value: tensor([ 0.0040, 0.0139, 0.0162, -0.0013, 0.0189, -0.0073, -0.0132, -0.0089, 0.0223, -0.0089], device='cuda:0'), grad: tensor([ 6.6698e-05, -6.6042e-04, 4.3130e-04, 1.0031e-04, 1.0587e-05, 4.2319e-04, -5.0449e-04, 6.4731e-05, 6.0976e-05, 6.7391e-06], device='cuda:0') 100 0.0001 changing lr epoch 91, time 251.52, cls_loss 0.0025 cls_loss_mapping 0.0074 cls_loss_causal 0.5723 re_mapping 0.0078 re_causal 0.0237 /// teacc 99.00 lr 0.00010000 Epoch 93, weight, value: tensor([[-0.0936, -0.0143, 0.1158, ..., 0.0357, -0.0974, -0.0920], [-0.0660, -0.0684, -0.0073, ..., -0.0677, -0.0682, -0.0351], [ 0.1234, -0.0304, -0.0873, ..., -0.0740, -0.0952, -0.1345], ..., [-0.0341, -0.0011, -0.0871, ..., -0.0253, 0.0401, 0.0025], [-0.0896, -0.0689, -0.0117, ..., -0.0726, 0.0029, -0.0170], [-0.0264, -0.0416, -0.0298, ..., 0.0725, 0.0289, -0.1018]], device='cuda:0'), grad: tensor([[ 2.3749e-07, 1.7695e-08, -6.3241e-05, ..., -6.4790e-05, 6.5938e-07, 6.1654e-07], [ 4.2748e-07, 1.4249e-07, 1.9297e-06, ..., 1.7937e-06, 2.1495e-06, 3.5577e-07], [-4.4145e-06, 2.1886e-07, 4.1649e-06, ..., 3.9861e-06, 8.1211e-07, 7.6648e-07], ..., [ 8.8289e-07, 1.7881e-07, 3.6001e-05, ..., 4.9859e-05, 2.5984e-06, 2.0489e-08], [ 2.7232e-06, 1.7416e-07, 6.7130e-06, ..., 1.3307e-05, 1.8105e-05, 1.5637e-06], [ 1.7416e-07, 1.2014e-07, 7.5586e-06, ..., -8.6054e-06, -2.5257e-05, 1.2759e-07]], device='cuda:0') Epoch 93, bias, value: tensor([ 0.0041, 0.0142, 0.0159, -0.0018, 0.0191, -0.0069, -0.0136, -0.0088, 0.0223, -0.0090], device='cuda:0'), grad: tensor([-1.4186e-04, -4.9509e-06, 7.2345e-06, 1.9092e-06, 2.8759e-06, 1.1005e-05, -1.4286e-06, 1.1510e-04, 6.7472e-05, -5.7667e-05], device='cuda:0') 100 0.0001 changing lr epoch 92, time 251.33, cls_loss 0.0030 cls_loss_mapping 0.0076 cls_loss_causal 0.5763 re_mapping 0.0074 re_causal 0.0234 /// teacc 98.98 lr 0.00010000 Epoch 94, weight, value: tensor([[-0.0938, -0.0144, 0.1161, ..., 0.0358, -0.0980, -0.0925], [-0.0661, -0.0687, -0.0074, ..., -0.0678, -0.0686, -0.0354], [ 0.1239, -0.0308, -0.0878, ..., -0.0744, -0.0955, -0.1348], ..., [-0.0346, -0.0014, -0.0875, ..., -0.0255, 0.0403, 0.0023], [-0.0902, -0.0696, -0.0121, ..., -0.0738, 0.0029, -0.0170], [-0.0265, -0.0417, -0.0300, ..., 0.0730, 0.0293, -0.1035]], device='cuda:0'), grad: tensor([[ 8.3819e-07, 1.6764e-08, 7.5437e-07, ..., -2.4028e-07, 2.1160e-06, 2.3618e-06], [ 6.2492e-07, 1.3970e-08, -1.1269e-06, ..., 4.5355e-07, 2.3320e-06, 6.3144e-07], [-3.4068e-06, 3.9116e-08, 2.9933e-06, ..., 1.3486e-06, 2.9728e-06, 2.5388e-06], ..., [ 1.2908e-06, 2.3283e-08, 3.7067e-07, ..., 5.4836e-06, -3.4645e-06, 2.0303e-07], [ 6.6273e-06, 1.7695e-08, 1.5823e-06, ..., 5.0962e-06, 7.9051e-06, 7.2196e-06], [ 2.3562e-06, 1.7695e-08, 8.2701e-07, ..., -5.2184e-05, -1.6034e-05, 4.8578e-06]], device='cuda:0') Epoch 94, bias, value: tensor([ 0.0040, 0.0141, 0.0158, -0.0014, 0.0188, -0.0074, -0.0127, -0.0089, 0.0221, -0.0089], device='cuda:0'), grad: tensor([ 8.3596e-06, -1.0610e-05, 1.3530e-05, 1.3089e-04, 7.0751e-05, -1.9813e-04, 2.8238e-05, -1.7047e-05, 4.0352e-05, -6.6400e-05], device='cuda:0') 100 0.0001 changing lr epoch 93, time 251.53, cls_loss 0.0032 cls_loss_mapping 0.0092 cls_loss_causal 0.6008 re_mapping 0.0071 re_causal 0.0232 /// teacc 99.04 lr 0.00010000 Epoch 95, weight, value: tensor([[-0.0941, -0.0145, 0.1165, ..., 0.0361, -0.0984, -0.0929], [-0.0663, -0.0689, -0.0073, ..., -0.0681, -0.0689, -0.0355], [ 0.1253, -0.0308, -0.0886, ..., -0.0752, -0.0957, -0.1355], ..., [-0.0353, -0.0016, -0.0878, ..., -0.0266, 0.0400, 0.0025], [-0.0922, -0.0699, -0.0123, ..., -0.0746, 0.0026, -0.0171], [-0.0268, -0.0419, -0.0303, ..., 0.0736, 0.0299, -0.1040]], device='cuda:0'), grad: tensor([[ 1.2383e-05, 3.0547e-07, -5.6982e-05, ..., -2.1994e-05, 5.5693e-07, 7.1526e-07], [ 7.3127e-06, 1.1642e-06, 9.9465e-07, ..., 7.3016e-07, 5.4948e-07, 2.4494e-07], [-5.8353e-05, 4.0755e-06, 3.5111e-06, ..., -1.0943e-06, 1.4408e-06, 5.8208e-07], ..., [ 1.4544e-05, 1.0123e-06, 8.8848e-07, ..., 1.0682e-06, -7.9274e-06, 8.1025e-08], [ 1.1332e-05, 3.3788e-06, 3.3360e-06, ..., 2.2613e-06, 1.3141e-06, 6.2585e-07], [ 4.9733e-06, 1.3383e-06, 1.6063e-05, ..., 2.0470e-06, 2.1886e-06, 3.0734e-07]], device='cuda:0') Epoch 95, bias, value: tensor([ 0.0043, 0.0140, 0.0164, -0.0016, 0.0186, -0.0070, -0.0125, -0.0096, 0.0208, -0.0081], device='cuda:0'), grad: tensor([-6.0230e-05, 3.7372e-05, -1.6287e-05, 3.0249e-05, 2.4617e-05, 3.1114e-05, 1.7032e-05, -1.4436e-04, 3.2365e-05, 4.7922e-05], device='cuda:0') 100 0.0001 changing lr epoch 94, time 251.30, cls_loss 0.0026 cls_loss_mapping 0.0087 cls_loss_causal 0.5810 re_mapping 0.0075 re_causal 0.0234 /// teacc 98.99 lr 0.00010000 Epoch 96, weight, value: tensor([[-0.0946, -0.0146, 0.1170, ..., 0.0364, -0.0987, -0.0931], [-0.0664, -0.0691, -0.0074, ..., -0.0682, -0.0692, -0.0357], [ 0.1265, -0.0306, -0.0893, ..., -0.0756, -0.0959, -0.1362], ..., [-0.0361, -0.0018, -0.0882, ..., -0.0268, 0.0404, 0.0023], [-0.0927, -0.0702, -0.0125, ..., -0.0754, 0.0024, -0.0172], [-0.0270, -0.0421, -0.0305, ..., 0.0739, 0.0297, -0.1042]], device='cuda:0'), grad: tensor([[ 6.2957e-06, 3.0175e-07, -7.2002e-05, ..., -8.7440e-05, 1.9465e-06, 9.5926e-07], [ 4.7274e-06, 2.4959e-07, 2.8200e-06, ..., 1.1094e-05, 4.7050e-06, 1.5739e-07], [-1.3208e-04, 2.7940e-07, 6.0424e-06, ..., 8.6278e-06, 1.0636e-06, 5.3458e-07], ..., [ 1.0198e-04, 5.4296e-07, 3.0287e-06, ..., 2.8521e-05, 7.5847e-06, -1.9278e-07], [ 8.6874e-06, 6.1281e-07, 4.4480e-06, ..., 1.4782e-05, 5.8748e-06, -2.3674e-06], [ 2.8275e-06, 4.6100e-07, 1.6034e-05, ..., -8.6427e-05, -4.7415e-05, 1.6494e-06]], device='cuda:0') Epoch 96, bias, value: tensor([ 0.0043, 0.0142, 0.0166, -0.0020, 0.0184, -0.0070, -0.0124, -0.0097, 0.0206, -0.0081], device='cuda:0'), grad: tensor([-2.2149e-04, 1.8105e-05, -2.0480e-04, 1.2720e-04, 1.8358e-05, 8.4341e-05, 5.0753e-05, 2.3675e-04, 2.1234e-05, -1.2958e-04], device='cuda:0') 100 0.0001 changing lr epoch 95, time 251.01, cls_loss 0.0029 cls_loss_mapping 0.0087 cls_loss_causal 0.5776 re_mapping 0.0072 re_causal 0.0234 /// teacc 99.02 lr 0.00010000 Epoch 97, weight, value: tensor([[-0.0952, -0.0153, 0.1175, ..., 0.0367, -0.0994, -0.0936], [-0.0664, -0.0694, -0.0065, ..., -0.0682, -0.0694, -0.0359], [ 0.1273, -0.0308, -0.0908, ..., -0.0758, -0.0959, -0.1366], ..., [-0.0366, -0.0019, -0.0886, ..., -0.0274, 0.0400, 0.0024], [-0.0930, -0.0694, -0.0123, ..., -0.0756, 0.0023, -0.0173], [-0.0282, -0.0426, -0.0309, ..., 0.0741, 0.0301, -0.1049]], device='cuda:0'), grad: tensor([[ 2.7493e-06, 5.5879e-09, -3.3885e-05, ..., -7.7412e-06, 1.8999e-06, -1.2405e-06], [ 1.1604e-06, 1.4901e-08, -5.6438e-06, ..., -2.3972e-06, 2.8014e-06, 1.7332e-06], [-1.3903e-05, 3.3528e-08, 2.3525e-06, ..., 6.5099e-07, 5.7295e-06, 1.4668e-06], ..., [-3.4459e-08, 3.8184e-08, 1.5832e-06, ..., 1.0412e-06, -1.6406e-05, 9.1270e-08], [ 3.3975e-06, 4.5635e-08, 4.6417e-06, ..., 9.1642e-07, 4.3362e-06, 3.1590e-06], [ 3.4273e-07, 5.0291e-08, 4.7423e-06, ..., -2.1495e-06, 6.3516e-06, 2.1681e-06]], device='cuda:0') Epoch 97, bias, value: tensor([ 0.0040, 0.0150, 0.0162, -0.0016, 0.0182, -0.0072, -0.0122, -0.0099, 0.0211, -0.0082], device='cuda:0'), grad: tensor([-2.1324e-05, -3.4869e-05, 3.9488e-06, 3.1561e-05, 1.5333e-05, -1.1697e-05, -1.5507e-06, -3.5524e-05, 2.0370e-05, 3.3706e-05], device='cuda:0') 100 0.0001 changing lr epoch 96, time 251.52, cls_loss 0.0035 cls_loss_mapping 0.0082 cls_loss_causal 0.5595 re_mapping 0.0074 re_causal 0.0231 /// teacc 98.98 lr 0.00010000 Epoch 98, weight, value: tensor([[-0.0956, -0.0153, 0.1178, ..., 0.0352, -0.1000, -0.0940], [-0.0667, -0.0696, -0.0067, ..., -0.0687, -0.0702, -0.0362], [ 0.1282, -0.0304, -0.0914, ..., -0.0764, -0.0963, -0.1370], ..., [-0.0366, -0.0021, -0.0893, ..., -0.0272, 0.0407, 0.0027], [-0.0936, -0.0698, -0.0125, ..., -0.0766, 0.0019, -0.0175], [-0.0289, -0.0428, -0.0307, ..., 0.0767, 0.0311, -0.1060]], device='cuda:0'), grad: tensor([[ 2.3574e-05, 6.9849e-08, 6.6981e-06, ..., -3.2783e-06, 1.0487e-06, 1.0980e-06], [ 5.3197e-06, 2.3562e-07, -2.4214e-05, ..., -3.2224e-06, 1.2377e-06, -1.1168e-05], [-1.0413e-04, 1.0291e-06, -4.2200e-05, ..., 3.2149e-06, 1.5739e-06, 4.5784e-06], ..., [ 3.4720e-06, 4.7032e-07, 2.1458e-06, ..., 2.9989e-06, 2.2817e-06, 9.7975e-07], [ 4.9233e-05, 1.1884e-06, 3.3587e-05, ..., 6.4746e-06, -2.7716e-06, 1.0766e-06], [ 1.9949e-06, 5.6997e-07, 6.5491e-06, ..., -1.0598e-04, -8.5294e-05, 2.2780e-06]], device='cuda:0') Epoch 98, bias, value: tensor([ 0.0026, 0.0148, 0.0164, -0.0029, 0.0167, -0.0063, -0.0122, -0.0094, 0.0207, -0.0065], device='cuda:0'), grad: tensor([ 3.1978e-05, -1.6654e-04, -8.6188e-05, 4.6305e-06, 2.1744e-04, 4.9621e-05, 2.4945e-05, 2.2665e-05, 1.0264e-04, -2.0123e-04], device='cuda:0') 100 0.0001 changing lr epoch 97, time 251.21, cls_loss 0.0026 cls_loss_mapping 0.0077 cls_loss_causal 0.5573 re_mapping 0.0073 re_causal 0.0234 /// teacc 99.02 lr 0.00010000 Epoch 99, weight, value: tensor([[-0.0957, -0.0154, 0.1185, ..., 0.0354, -0.1004, -0.0942], [-0.0668, -0.0698, -0.0063, ..., -0.0684, -0.0706, -0.0362], [ 0.1290, -0.0306, -0.0918, ..., -0.0765, -0.0968, -0.1374], ..., [-0.0369, -0.0022, -0.0900, ..., -0.0275, 0.0414, 0.0030], [-0.0949, -0.0700, -0.0128, ..., -0.0780, 0.0016, -0.0176], [-0.0293, -0.0430, -0.0311, ..., 0.0763, 0.0304, -0.1071]], device='cuda:0'), grad: tensor([[ 1.1977e-06, 1.5274e-07, -2.6729e-06, ..., 2.7567e-07, 2.5444e-06, 1.0403e-06], [ 4.2357e-06, 1.3970e-08, 8.8066e-06, ..., 1.3169e-06, 8.1658e-06, 6.2883e-06], [-6.2436e-06, 5.6811e-08, 3.8017e-06, ..., 1.5358e-06, 5.4911e-06, 2.8219e-06], ..., [ 4.2003e-07, 1.0245e-07, 1.4063e-07, ..., 4.5039e-06, 1.2293e-07, -1.7881e-07], [ 7.0129e-07, 1.6764e-08, 2.9895e-07, ..., -7.4506e-09, -1.1146e-05, 4.0047e-07], [ 4.9546e-07, 6.2399e-08, 2.3320e-06, ..., 2.3067e-05, 1.7747e-05, 1.9465e-07]], device='cuda:0') Epoch 99, bias, value: tensor([ 0.0029, 0.0150, 0.0167, -0.0023, 0.0177, -0.0067, -0.0123, -0.0092, 0.0197, -0.0074], device='cuda:0'), grad: tensor([ 1.3158e-05, 3.9160e-05, 1.7479e-05, 3.8117e-05, -7.5102e-05, -1.7919e-06, -3.3706e-05, -1.8729e-06, -6.9857e-05, 7.4387e-05], device='cuda:0') 100 0.0001 changing lr epoch 98, time 251.17, cls_loss 0.0026 cls_loss_mapping 0.0090 cls_loss_causal 0.5805 re_mapping 0.0075 re_causal 0.0232 /// teacc 98.98 lr 0.00010000 Epoch 100, weight, value: tensor([[-0.0968, -0.0155, 0.1197, ..., 0.0356, -0.1013, -0.0945], [-0.0671, -0.0699, -0.0074, ..., -0.0685, -0.0712, -0.0370], [ 0.1295, -0.0328, -0.0927, ..., -0.0776, -0.0973, -0.1377], ..., [-0.0375, -0.0025, -0.0905, ..., -0.0278, 0.0416, 0.0029], [-0.0952, -0.0702, -0.0127, ..., -0.0781, 0.0015, -0.0175], [-0.0294, -0.0432, -0.0318, ..., 0.0765, 0.0303, -0.1076]], device='cuda:0'), grad: tensor([[ 1.0366e-06, 9.2201e-08, -1.0766e-05, ..., -5.3644e-06, 2.8741e-06, 7.0781e-07], [ 2.5984e-06, 3.1292e-07, -3.8296e-06, ..., 1.5553e-06, 2.7772e-06, 4.7497e-07], [ 7.6741e-06, 2.1122e-06, 2.6487e-06, ..., 5.5172e-06, 1.8790e-05, 3.2317e-07], ..., [ 4.4703e-06, 1.6587e-06, 1.1651e-06, ..., -2.4021e-05, -1.0967e-04, 1.5553e-07], [ 1.7481e-06, 2.2259e-07, 1.2005e-06, ..., 3.3733e-06, 5.0329e-06, 7.7952e-07], [ 8.6501e-06, 9.7230e-07, 4.2170e-06, ..., 1.3053e-05, 1.6466e-05, 5.9232e-07]], device='cuda:0') Epoch 100, bias, value: tensor([ 0.0030, 0.0150, 0.0163, -0.0022, 0.0178, -0.0067, -0.0121, -0.0093, 0.0202, -0.0076], device='cuda:0'), grad: tensor([-2.4028e-06, -2.6298e-04, 7.7426e-05, -2.9176e-05, 1.1927e-04, -2.4065e-05, 2.3931e-05, -3.8952e-05, 2.0966e-05, 1.1593e-04], device='cuda:0') 100 0.0001 changing lr epoch 99, time 251.16, cls_loss 0.0026 cls_loss_mapping 0.0076 cls_loss_causal 0.5785 re_mapping 0.0072 re_causal 0.0229 /// teacc 99.04 lr 0.00010000 Epoch 101, weight, value: tensor([[-0.0970, -0.0155, 0.1215, ..., 0.0363, -0.1018, -0.0952], [-0.0672, -0.0701, -0.0075, ..., -0.0682, -0.0724, -0.0371], [ 0.1299, -0.0330, -0.0937, ..., -0.0784, -0.0979, -0.1380], ..., [-0.0375, -0.0027, -0.0911, ..., -0.0285, 0.0423, 0.0026], [-0.0954, -0.0703, -0.0127, ..., -0.0780, 0.0016, -0.0174], [-0.0297, -0.0434, -0.0328, ..., 0.0768, 0.0303, -0.1082]], device='cuda:0'), grad: tensor([[ 1.4659e-06, 4.5169e-07, -4.1649e-06, ..., -2.2650e-06, 1.3206e-06, 1.4259e-06], [ 4.8019e-06, 1.6047e-06, 9.3598e-07, ..., -4.4145e-07, 2.9374e-06, 1.2694e-06], [ 7.7426e-05, 2.5362e-05, 1.4631e-06, ..., 2.5015e-06, 1.8254e-05, 1.9316e-06], ..., [-2.1696e-05, 5.6289e-06, 3.8370e-07, ..., 2.0023e-06, -2.5839e-05, 2.3376e-07], [ 5.6252e-06, 1.9595e-06, -7.4469e-06, ..., -1.4119e-06, -1.1846e-05, -1.4089e-05], [ 8.2925e-06, 2.0545e-06, 5.8897e-06, ..., 6.5193e-06, 7.8827e-06, 4.8950e-06]], device='cuda:0') Epoch 101, bias, value: tensor([ 0.0037, 0.0153, 0.0156, -0.0024, 0.0173, -0.0069, -0.0125, -0.0091, 0.0210, -0.0076], device='cuda:0'), grad: tensor([ 3.3826e-06, -1.5318e-05, 1.5402e-04, -1.1295e-04, 5.0366e-06, 2.3305e-05, 1.3031e-05, -6.3837e-05, -6.4850e-05, 5.8204e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 100---------------------------------------------------- epoch 100, time 268.46, cls_loss 0.0034 cls_loss_mapping 0.0093 cls_loss_causal 0.5889 re_mapping 0.0071 re_causal 0.0225 /// teacc 99.09 lr 0.00010000 Epoch 102, weight, value: tensor([[-9.7476e-02, -1.5566e-02, 1.2246e-01, ..., 3.6668e-02, -1.0234e-01, -9.5498e-02], [-6.7772e-02, -7.0457e-02, -7.0759e-03, ..., -6.8958e-02, -7.3398e-02, -3.7248e-02], [ 1.3051e-01, -3.3285e-02, -9.4475e-02, ..., -7.8999e-02, -9.8376e-02, -1.3857e-01], ..., [-3.7536e-02, -3.2772e-03, -9.1633e-02, ..., -2.8687e-02, 4.2397e-02, 1.6970e-03], [-9.5801e-02, -7.0848e-02, -1.4436e-02, ..., -7.8969e-02, 7.6544e-05, -1.9684e-02], [-3.0357e-02, -4.3802e-02, -3.3353e-02, ..., 7.6568e-02, 2.9771e-02, -1.0947e-01]], device='cuda:0'), grad: tensor([[ 1.2340e-06, 1.0710e-06, -1.0312e-05, ..., -4.5374e-06, 6.4727e-07, 6.9849e-07], [ 1.7527e-06, 1.4789e-06, 8.1956e-07, ..., 2.1793e-06, 1.3253e-06, 3.0175e-07], [ 1.2472e-05, 1.1176e-05, 1.6773e-06, ..., 1.2480e-06, 1.2768e-06, 1.3802e-06], ..., [ 6.5714e-06, 5.7369e-06, 4.3586e-07, ..., 1.3169e-06, -7.4413e-07, 6.3889e-07], [ 4.1425e-06, 3.6154e-06, 1.9092e-06, ..., 3.6806e-06, 3.9823e-06, 5.0105e-06], [ 6.3367e-06, 5.6438e-06, 3.8184e-06, ..., 4.4778e-06, 3.0119e-06, 3.0641e-06]], device='cuda:0') Epoch 102, bias, value: tensor([ 0.0041, 0.0149, 0.0154, -0.0013, 0.0185, -0.0077, -0.0116, -0.0088, 0.0199, -0.0086], device='cuda:0'), grad: tensor([-7.2494e-06, 3.6247e-06, 4.8876e-05, -1.1683e-04, -2.9519e-05, -7.8008e-06, 9.5367e-06, 2.1443e-05, 3.5405e-05, 4.2528e-05], device='cuda:0') 100 0.0001 changing lr epoch 101, time 251.57, cls_loss 0.0025 cls_loss_mapping 0.0069 cls_loss_causal 0.5556 re_mapping 0.0068 re_causal 0.0215 /// teacc 98.94 lr 0.00010000 Epoch 103, weight, value: tensor([[-0.0978, -0.0156, 0.1233, ..., 0.0368, -0.1028, -0.0959], [-0.0681, -0.0708, -0.0070, ..., -0.0694, -0.0739, -0.0374], [ 0.1312, -0.0335, -0.0952, ..., -0.0791, -0.0988, -0.1391], ..., [-0.0378, -0.0036, -0.0923, ..., -0.0289, 0.0428, 0.0017], [-0.0962, -0.0712, -0.0150, ..., -0.0799, -0.0005, -0.0200], [-0.0307, -0.0443, -0.0336, ..., 0.0772, 0.0302, -0.1094]], device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, -8.1837e-05, ..., -3.6716e-05, 1.1176e-08, 1.8906e-07], [ 1.7695e-08, 0.0000e+00, 7.4506e-08, ..., -2.1197e-06, 3.1274e-06, 4.2841e-07], [-9.4995e-08, 0.0000e+00, 2.0638e-05, ..., 9.3281e-06, 8.2981e-07, 1.5739e-07], ..., [ 2.5146e-08, 0.0000e+00, 7.8231e-07, ..., 6.7316e-06, -1.0483e-05, 9.4995e-08], [ 2.0489e-08, 0.0000e+00, 1.8507e-05, ..., 8.8662e-06, 1.6019e-07, -2.2072e-06], [ 1.8626e-09, 0.0000e+00, 1.5229e-05, ..., -1.1642e-06, -3.4291e-06, 6.7241e-07]], device='cuda:0') Epoch 103, bias, value: tensor([ 0.0043, 0.0149, 0.0154, -0.0012, 0.0181, -0.0079, -0.0115, -0.0087, 0.0194, -0.0082], device='cuda:0'), grad: tensor([-1.2624e-04, -1.3947e-05, 3.7938e-05, 2.6777e-05, 6.5453e-06, 6.7241e-06, 3.5226e-05, -1.7181e-05, 2.4304e-05, 1.9908e-05], device='cuda:0') 100 0.0001 changing lr epoch 102, time 251.68, cls_loss 0.0024 cls_loss_mapping 0.0074 cls_loss_causal 0.5858 re_mapping 0.0074 re_causal 0.0234 /// teacc 98.94 lr 0.00010000 Epoch 104, weight, value: tensor([[-0.0983, -0.0157, 0.1239, ..., 0.0370, -0.1034, -0.0964], [-0.0683, -0.0711, -0.0072, ..., -0.0697, -0.0744, -0.0379], [ 0.1318, -0.0336, -0.0959, ..., -0.0795, -0.0996, -0.1406], ..., [-0.0383, -0.0041, -0.0928, ..., -0.0290, 0.0432, 0.0018], [-0.0970, -0.0719, -0.0149, ..., -0.0804, -0.0007, -0.0202], [-0.0309, -0.0450, -0.0339, ..., 0.0773, 0.0304, -0.1098]], device='cuda:0'), grad: tensor([[ 6.2399e-08, 2.3022e-06, -4.2096e-07, ..., -1.2852e-07, 1.4063e-06, 7.3388e-07], [ 1.5832e-07, 1.3597e-06, 3.1665e-08, ..., 2.3376e-07, 1.1222e-06, 4.4238e-07], [-1.1688e-06, 3.4366e-07, 2.6077e-08, ..., 1.4063e-07, 3.2410e-07, 1.2480e-07], ..., [ 4.9174e-07, 1.3486e-06, 1.6764e-08, ..., 4.4145e-07, -3.0082e-07, 4.2655e-07], [-4.5355e-07, 5.0277e-05, 1.5739e-07, ..., 3.2187e-06, 3.0965e-05, 1.8850e-05], [ 9.2201e-08, 7.5400e-06, 2.3097e-07, ..., -7.9423e-06, 2.1271e-06, 2.8014e-06]], device='cuda:0') Epoch 104, bias, value: tensor([ 0.0045, 0.0149, 0.0151, -0.0024, 0.0181, -0.0057, -0.0125, -0.0087, 0.0192, -0.0083], device='cuda:0'), grad: tensor([ 5.7369e-06, 4.5523e-06, 1.1725e-06, 6.3419e-04, 4.8578e-06, -7.9107e-04, 3.2689e-06, 1.5665e-06, 1.3006e-04, 5.6773e-06], device='cuda:0') 100 0.0001 changing lr epoch 103, time 251.52, cls_loss 0.0027 cls_loss_mapping 0.0075 cls_loss_causal 0.5576 re_mapping 0.0074 re_causal 0.0224 /// teacc 98.90 lr 0.00010000 Epoch 105, weight, value: tensor([[-0.0991, -0.0159, 0.1242, ..., 0.0371, -0.1042, -0.0972], [-0.0685, -0.0715, -0.0074, ..., -0.0696, -0.0746, -0.0377], [ 0.1324, -0.0338, -0.0969, ..., -0.0798, -0.1005, -0.1417], ..., [-0.0386, -0.0043, -0.0931, ..., -0.0291, 0.0436, 0.0017], [-0.0975, -0.0722, -0.0146, ..., -0.0809, -0.0006, -0.0198], [-0.0310, -0.0454, -0.0342, ..., 0.0774, 0.0302, -0.1104]], device='cuda:0'), grad: tensor([[ 8.8476e-08, -6.9477e-07, -1.3709e-05, ..., -6.1207e-06, 1.9874e-06, 1.6335e-06], [ 1.7136e-07, 1.3132e-07, 2.4904e-06, ..., 8.2329e-07, 4.6864e-06, 5.7779e-06], [-4.7218e-07, 2.9616e-07, 1.4622e-06, ..., 8.7917e-07, 4.4797e-07, 4.5821e-07], ..., [ 4.1164e-07, 3.8370e-07, 8.7172e-07, ..., 1.6717e-06, 5.6718e-07, 2.1048e-07], [ 2.2911e-07, 3.5390e-07, 4.8280e-06, ..., 6.7316e-06, 6.1952e-06, 4.0419e-06], [ 1.1921e-07, 4.4424e-07, 5.0776e-06, ..., -1.0565e-05, -9.9614e-06, -9.6858e-07]], device='cuda:0') Epoch 105, bias, value: tensor([ 0.0043, 0.0141, 0.0164, -0.0027, 0.0184, -0.0061, -0.0123, -0.0087, 0.0198, -0.0087], device='cuda:0'), grad: tensor([-1.4886e-05, 8.8811e-06, 6.6794e-06, 1.7881e-05, 1.6913e-05, 1.2927e-05, -5.3823e-05, 5.7593e-06, 2.4125e-05, -2.4557e-05], device='cuda:0') 100 0.0001 changing lr epoch 104, time 251.78, cls_loss 0.0025 cls_loss_mapping 0.0094 cls_loss_causal 0.5725 re_mapping 0.0071 re_causal 0.0211 /// teacc 98.94 lr 0.00010000 Epoch 106, weight, value: tensor([[-0.0995, -0.0160, 0.1249, ..., 0.0373, -0.1048, -0.0980], [-0.0686, -0.0718, -0.0075, ..., -0.0699, -0.0750, -0.0379], [ 0.1334, -0.0349, -0.0972, ..., -0.0801, -0.1005, -0.1420], ..., [-0.0395, -0.0045, -0.0939, ..., -0.0294, 0.0439, 0.0016], [-0.0977, -0.0740, -0.0147, ..., -0.0820, -0.0010, -0.0202], [-0.0312, -0.0457, -0.0346, ..., 0.0775, 0.0302, -0.1108]], device='cuda:0'), grad: tensor([[ 2.9150e-06, -2.7567e-07, -4.6156e-06, ..., -1.3774e-06, 1.4948e-06, 7.2364e-07], [ 1.5451e-06, 7.8790e-07, 2.7381e-07, ..., 3.8017e-06, 4.2282e-06, 5.3924e-07], [-1.1288e-05, 5.1670e-06, 6.5193e-07, ..., 8.1304e-07, 1.1541e-05, 6.8732e-07], ..., [-3.4552e-06, -2.2516e-05, 1.0151e-07, ..., 2.2259e-06, -5.9545e-05, 3.2783e-07], [ 1.2238e-06, 6.1374e-07, 2.1514e-07, ..., 9.8497e-06, 4.8541e-06, 1.0580e-06], [ 6.3423e-07, 1.2917e-06, 2.2836e-06, ..., 1.9763e-06, 1.5154e-05, 4.7777e-07]], device='cuda:0') Epoch 106, bias, value: tensor([ 0.0045, 0.0141, 0.0163, -0.0029, 0.0187, -0.0055, -0.0125, -0.0085, 0.0190, -0.0089], device='cuda:0'), grad: tensor([ 6.4671e-06, 2.0027e-05, 8.3447e-06, 1.5008e-04, -7.2896e-05, -7.2658e-05, 1.3918e-05, -1.4806e-04, 4.1395e-05, 5.3227e-05], device='cuda:0') 100 0.0001 changing lr epoch 105, time 251.45, cls_loss 0.0024 cls_loss_mapping 0.0075 cls_loss_causal 0.5546 re_mapping 0.0072 re_causal 0.0213 /// teacc 98.99 lr 0.00010000 Epoch 107, weight, value: tensor([[-0.0999, -0.0160, 0.1266, ..., 0.0379, -0.1055, -0.0984], [-0.0689, -0.0722, -0.0081, ..., -0.0700, -0.0757, -0.0384], [ 0.1344, -0.0349, -0.0982, ..., -0.0805, -0.1010, -0.1419], ..., [-0.0400, -0.0046, -0.0945, ..., -0.0301, 0.0439, 0.0018], [-0.0981, -0.0736, -0.0145, ..., -0.0824, -0.0012, -0.0202], [-0.0314, -0.0468, -0.0357, ..., 0.0777, 0.0306, -0.1120]], device='cuda:0'), grad: tensor([[ 3.5502e-06, 9.5833e-07, -5.3644e-06, ..., -2.9542e-06, 3.5036e-06, 1.5181e-06], [ 4.5002e-06, 1.7909e-06, 4.3679e-07, ..., 8.5011e-06, 3.7104e-05, 5.9232e-06], [-2.3973e-04, -4.6939e-06, 1.4044e-06, ..., -3.7193e-05, 1.0327e-05, 1.1940e-06], ..., [ 1.5711e-06, -2.6245e-06, 2.1420e-07, ..., -5.0589e-06, -8.0585e-05, -7.5847e-06], [ 1.1504e-05, -1.0446e-05, 9.0431e-07, ..., 2.8443e-06, 3.2317e-06, -8.3745e-06], [ 2.2316e-04, 1.0222e-05, 2.2780e-06, ..., 6.0380e-05, 2.0653e-05, 3.3807e-06]], device='cuda:0') Epoch 107, bias, value: tensor([ 0.0054, 0.0144, 0.0165, -0.0032, 0.0185, -0.0055, -0.0131, -0.0090, 0.0191, -0.0090], device='cuda:0'), grad: tensor([ 1.5706e-05, 1.7595e-04, -4.6372e-04, 1.8448e-05, -1.7792e-05, -2.2054e-05, 2.1771e-05, -3.0494e-04, -1.5274e-05, 5.9080e-04], device='cuda:0') 100 0.0001 changing lr epoch 106, time 251.77, cls_loss 0.0028 cls_loss_mapping 0.0092 cls_loss_causal 0.5830 re_mapping 0.0072 re_causal 0.0222 /// teacc 99.02 lr 0.00010000 Epoch 108, weight, value: tensor([[-0.1003, -0.0160, 0.1248, ..., 0.0379, -0.1064, -0.1015], [-0.0683, -0.0725, -0.0084, ..., -0.0703, -0.0765, -0.0388], [ 0.1347, -0.0349, -0.0988, ..., -0.0808, -0.1014, -0.1423], ..., [-0.0401, -0.0049, -0.0949, ..., -0.0303, 0.0441, 0.0023], [-0.0989, -0.0740, -0.0122, ..., -0.0827, -0.0007, -0.0186], [-0.0319, -0.0471, -0.0363, ..., 0.0779, 0.0308, -0.1134]], device='cuda:0'), grad: tensor([[ 1.0431e-07, 3.4459e-08, -6.1154e-05, ..., -4.0203e-05, 3.2783e-07, -6.7893e-07], [ 7.1712e-07, 1.1548e-07, 8.0746e-07, ..., -6.0797e-05, 3.4645e-07, 9.9652e-08], [-2.4326e-06, 1.4529e-07, 9.2760e-06, ..., 6.8583e-06, 4.2748e-07, 2.2445e-07], ..., [ 9.4157e-07, 2.3749e-07, 2.9262e-06, ..., 2.0489e-05, -1.0310e-06, -1.6764e-07], [ 2.9616e-07, 1.7323e-07, 8.9109e-06, ..., 7.8306e-06, 1.0785e-06, 6.5286e-07], [ 2.0489e-07, 4.5635e-08, 1.1809e-05, ..., 1.7256e-05, -6.5453e-06, 5.3924e-07]], device='cuda:0') Epoch 108, bias, value: tensor([ 0.0035, 0.0152, 0.0159, -0.0029, 0.0184, -0.0059, -0.0122, -0.0093, 0.0204, -0.0089], device='cuda:0'), grad: tensor([-1.1945e-04, -2.2042e-04, 1.9208e-05, 9.3356e-06, 7.5817e-05, 1.8269e-05, 3.1143e-05, 7.5340e-05, 2.7165e-05, 8.3268e-05], device='cuda:0') 100 0.0001 changing lr epoch 107, time 251.76, cls_loss 0.0026 cls_loss_mapping 0.0076 cls_loss_causal 0.5618 re_mapping 0.0069 re_causal 0.0214 /// teacc 99.01 lr 0.00010000 Epoch 109, weight, value: tensor([[-0.1007, -0.0159, 0.1258, ..., 0.0377, -0.1069, -0.1017], [-0.0695, -0.0728, -0.0085, ..., -0.0708, -0.0771, -0.0390], [ 0.1353, -0.0350, -0.0994, ..., -0.0812, -0.1019, -0.1427], ..., [-0.0391, -0.0051, -0.0956, ..., -0.0308, 0.0442, 0.0023], [-0.0998, -0.0744, -0.0124, ..., -0.0834, -0.0007, -0.0185], [-0.0321, -0.0474, -0.0365, ..., 0.0782, 0.0306, -0.1142]], device='cuda:0'), grad: tensor([[ 2.8610e-06, 1.0394e-06, -3.6713e-06, ..., -2.4028e-07, 1.3513e-06, 1.3784e-07], [ 2.2687e-06, 7.0687e-07, -1.7043e-07, ..., 3.0603e-06, 1.2713e-06, 6.3330e-08], [-7.0870e-05, 1.6037e-06, 2.4121e-07, ..., 9.2108e-07, 9.8068e-07, 6.6124e-08], ..., [ 6.1691e-05, 6.4038e-06, 3.3528e-07, ..., 1.6401e-06, 3.5278e-06, 2.3283e-08], [ 6.1318e-06, 2.6412e-06, 4.1164e-07, ..., 4.8056e-06, 1.2796e-06, -7.4320e-07], [ 6.7167e-06, 5.8189e-06, 2.0918e-06, ..., -1.2779e-04, -3.5703e-05, 4.2189e-07]], device='cuda:0') Epoch 109, bias, value: tensor([ 0.0036, 0.0138, 0.0159, -0.0025, 0.0188, -0.0064, -0.0123, -0.0082, 0.0206, -0.0090], device='cuda:0'), grad: tensor([ 6.3479e-06, 4.9323e-06, -1.2338e-04, -6.8665e-03, 1.4329e-04, 6.8130e-03, 6.3851e-06, 1.4269e-04, 1.7881e-05, -1.4460e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 108---------------------------------------------------- epoch 108, time 268.22, cls_loss 0.0019 cls_loss_mapping 0.0063 cls_loss_causal 0.5519 re_mapping 0.0067 re_causal 0.0209 /// teacc 99.12 lr 0.00010000 Epoch 110, weight, value: tensor([[-0.1011, -0.0160, 0.1267, ..., 0.0383, -0.1076, -0.1019], [-0.0696, -0.0732, -0.0086, ..., -0.0710, -0.0773, -0.0391], [ 0.1356, -0.0361, -0.0998, ..., -0.0815, -0.1027, -0.1435], ..., [-0.0395, -0.0055, -0.0962, ..., -0.0310, 0.0445, 0.0025], [-0.1003, -0.0747, -0.0125, ..., -0.0841, -0.0010, -0.0184], [-0.0324, -0.0480, -0.0381, ..., 0.0779, 0.0305, -0.1151]], device='cuda:0'), grad: tensor([[ 3.6648e-07, 3.1665e-08, -7.2122e-05, ..., -1.5289e-05, 8.0839e-07, 3.3900e-07], [ 1.2061e-06, 3.8184e-08, 1.1986e-06, ..., 2.2557e-06, 1.2023e-06, 1.3784e-07], [-3.7663e-06, 8.8476e-08, 4.2617e-06, ..., 1.5767e-06, 3.0510e-06, 3.9078e-06], ..., [ 1.2387e-06, 5.2061e-07, 6.6962e-07, ..., 4.0270e-06, -2.6450e-06, 1.9930e-07], [ 3.7858e-07, 7.2177e-08, 4.2692e-06, ..., 6.4149e-06, 3.3155e-06, 6.0583e-07], [ 1.2200e-07, 1.5507e-07, 9.6560e-06, ..., -1.6168e-05, -1.0401e-05, 1.0477e-07]], device='cuda:0') Epoch 110, bias, value: tensor([ 0.0041, 0.0139, 0.0158, -0.0013, 0.0191, -0.0068, -0.0123, -0.0083, 0.0205, -0.0096], device='cuda:0'), grad: tensor([-9.4056e-05, -6.0126e-06, 1.1399e-05, 1.0885e-05, 1.5661e-05, 5.0575e-05, 1.1355e-05, -1.3243e-06, 2.0549e-05, -1.9073e-05], device='cuda:0') 100 0.0001 changing lr epoch 109, time 251.23, cls_loss 0.0023 cls_loss_mapping 0.0066 cls_loss_causal 0.5749 re_mapping 0.0065 re_causal 0.0209 /// teacc 99.11 lr 0.00010000 Epoch 111, weight, value: tensor([[-0.1018, -0.0158, 0.1284, ..., 0.0387, -0.1073, -0.1020], [-0.0700, -0.0733, -0.0082, ..., -0.0711, -0.0778, -0.0388], [ 0.1365, -0.0361, -0.1017, ..., -0.0824, -0.1032, -0.1442], ..., [-0.0395, -0.0056, -0.0970, ..., -0.0314, 0.0454, 0.0027], [-0.1012, -0.0751, -0.0125, ..., -0.0848, -0.0013, -0.0185], [-0.0325, -0.0481, -0.0384, ..., 0.0782, 0.0305, -0.1146]], device='cuda:0'), grad: tensor([[ 2.6990e-06, 4.4703e-08, -2.7288e-07, ..., 9.3132e-09, 1.3057e-06, 1.0254e-06], [ 1.1688e-06, 6.9849e-08, 7.6368e-08, ..., 1.0366e-06, 2.9821e-06, 6.3051e-07], [-9.0674e-06, 4.0326e-07, 3.4645e-07, ..., 8.1491e-07, 2.6003e-06, 1.5302e-06], ..., [ 5.6904e-07, 8.7544e-08, 5.3085e-08, ..., -4.0174e-05, -1.4234e-04, -5.2899e-05], [ 3.4962e-06, 1.9651e-07, 4.7199e-06, ..., 1.0906e-06, 3.7365e-06, 3.8929e-06], [ 2.7753e-07, 5.5879e-08, 3.2783e-07, ..., 3.6061e-05, 1.3340e-04, 5.1081e-05]], device='cuda:0') Epoch 111, bias, value: tensor([ 0.0047, 0.0146, 0.0161, -0.0016, 0.0188, -0.0072, -0.0126, -0.0085, 0.0199, -0.0096], device='cuda:0'), grad: tensor([ 9.3579e-06, 3.2596e-06, -6.9775e-06, 1.2323e-05, 2.7698e-06, 1.3426e-05, -1.3486e-05, -4.8065e-04, 1.5339e-06, 4.5800e-04], device='cuda:0') 100 0.0001 changing lr epoch 110, time 251.25, cls_loss 0.0023 cls_loss_mapping 0.0069 cls_loss_causal 0.5397 re_mapping 0.0064 re_causal 0.0198 /// teacc 99.05 lr 0.00010000 Epoch 112, weight, value: tensor([[-0.1025, -0.0167, 0.1285, ..., 0.0388, -0.1080, -0.1023], [-0.0701, -0.0736, -0.0083, ..., -0.0714, -0.0787, -0.0393], [ 0.1373, -0.0363, -0.1022, ..., -0.0831, -0.1036, -0.1448], ..., [-0.0399, -0.0058, -0.0979, ..., -0.0317, 0.0460, 0.0045], [-0.1017, -0.0752, -0.0125, ..., -0.0854, -0.0022, -0.0195], [-0.0324, -0.0483, -0.0388, ..., 0.0788, 0.0307, -0.1151]], device='cuda:0'), grad: tensor([[ 1.3877e-07, 9.7789e-08, -8.8755e-07, ..., 7.3574e-08, 9.3691e-07, 3.2317e-07], [ 1.6764e-07, 4.2468e-07, -7.9442e-07, ..., 3.1497e-06, 4.4517e-06, 6.6869e-07], [ 6.8918e-07, 8.2236e-07, 9.0804e-07, ..., 9.3412e-07, 1.3085e-06, 4.9546e-07], ..., [ 4.1071e-07, 5.4017e-07, 2.1607e-07, ..., -9.6951e-07, -6.6161e-06, 4.7963e-07], [ 3.0920e-07, 3.4459e-07, 7.4506e-07, ..., 5.7556e-06, 6.1747e-07, 1.0915e-06], [ 5.1688e-07, 5.8208e-07, 3.7625e-07, ..., -9.0301e-05, -6.3956e-05, 1.9744e-07]], device='cuda:0') Epoch 112, bias, value: tensor([ 0.0044, 0.0147, 0.0160, -0.0026, 0.0184, -0.0060, -0.0127, -0.0081, 0.0192, -0.0093], device='cuda:0'), grad: tensor([ 2.6412e-06, -3.0734e-06, 2.0415e-05, 6.8638e-07, 1.7166e-04, 1.8388e-05, -7.3016e-06, -3.4034e-05, -4.2133e-06, -1.6510e-04], device='cuda:0') 100 0.0001 changing lr epoch 111, time 251.21, cls_loss 0.0018 cls_loss_mapping 0.0063 cls_loss_causal 0.5464 re_mapping 0.0069 re_causal 0.0214 /// teacc 98.99 lr 0.00010000 Epoch 113, weight, value: tensor([[-0.1030, -0.0167, 0.1289, ..., 0.0391, -0.1090, -0.1027], [-0.0706, -0.0738, -0.0083, ..., -0.0719, -0.0794, -0.0398], [ 0.1382, -0.0365, -0.1027, ..., -0.0834, -0.1041, -0.1453], ..., [-0.0405, -0.0061, -0.0983, ..., -0.0319, 0.0468, 0.0062], [-0.1020, -0.0754, -0.0127, ..., -0.0859, -0.0024, -0.0195], [-0.0327, -0.0486, -0.0392, ..., 0.0786, 0.0301, -0.1166]], device='cuda:0'), grad: tensor([[ 4.6007e-07, 1.1269e-07, 5.5786e-07, ..., 4.8149e-07, 4.5635e-07, 5.1688e-07], [ 8.7731e-07, 3.9767e-07, 2.6356e-07, ..., -1.2234e-05, 1.9297e-06, 9.8255e-07], [-3.0920e-07, 1.2964e-06, 9.0897e-06, ..., 4.3679e-07, 1.2014e-06, 5.3868e-06], ..., [ 1.9111e-06, 8.1398e-07, 9.4995e-08, ..., 1.0490e-05, -7.8008e-06, -7.3481e-07], [ 1.0133e-06, 5.7928e-07, 1.7341e-06, ..., 2.7008e-06, 1.2470e-06, -1.8720e-07], [ 1.0189e-06, 1.1008e-06, 1.9465e-07, ..., -2.5034e-05, -8.1807e-06, -2.9933e-06]], device='cuda:0') Epoch 113, bias, value: tensor([ 0.0045, 0.0146, 0.0162, -0.0027, 0.0186, -0.0059, -0.0125, -0.0079, 0.0190, -0.0098], device='cuda:0'), grad: tensor([ 4.5747e-06, -3.2514e-05, 4.3064e-05, -5.1260e-06, 5.5462e-05, 1.2115e-05, -3.3587e-05, 3.8669e-06, -8.2999e-06, -3.9577e-05], device='cuda:0') 100 0.0001 changing lr epoch 112, time 251.60, cls_loss 0.0023 cls_loss_mapping 0.0083 cls_loss_causal 0.5365 re_mapping 0.0068 re_causal 0.0206 /// teacc 98.90 lr 0.00010000 Epoch 114, weight, value: tensor([[-0.1036, -0.0169, 0.1295, ..., 0.0395, -0.1095, -0.1029], [-0.0711, -0.0743, -0.0083, ..., -0.0740, -0.0798, -0.0402], [ 0.1394, -0.0367, -0.1033, ..., -0.0837, -0.1026, -0.1452], ..., [-0.0407, -0.0065, -0.0989, ..., -0.0324, 0.0463, 0.0057], [-0.1024, -0.0758, -0.0128, ..., -0.0866, -0.0022, -0.0193], [-0.0335, -0.0491, -0.0396, ..., 0.0791, 0.0302, -0.1168]], device='cuda:0'), grad: tensor([[ 1.6559e-06, 4.7218e-07, -1.5469e-06, ..., -1.8906e-07, 1.1669e-06, 6.2957e-07], [ 2.2054e-05, 2.2221e-06, 7.5437e-08, ..., 1.3206e-06, 2.2165e-06, 1.1930e-06], [ 1.6958e-05, 3.8929e-06, 1.6391e-07, ..., 1.5534e-06, 2.5649e-06, 4.6566e-07], ..., [-4.7922e-05, 1.6123e-05, 1.3970e-08, ..., 3.7812e-06, 7.1637e-06, 1.5106e-06], [ 5.4389e-06, 5.3607e-06, -1.3784e-07, ..., 6.1058e-06, 5.6699e-06, 5.5321e-07], [ 8.9407e-07, 8.2329e-07, 1.1148e-06, ..., -7.6517e-06, -6.4075e-06, -5.1446e-06]], device='cuda:0') Epoch 114, bias, value: tensor([ 0.0046, 0.0142, 0.0172, -0.0031, 0.0184, -0.0055, -0.0128, -0.0086, 0.0192, -0.0095], device='cuda:0'), grad: tensor([ 2.2560e-05, 2.7943e-04, 2.4915e-04, 1.2147e-04, -5.3197e-06, 3.0100e-05, 1.0006e-05, -7.4291e-04, 4.7207e-05, -1.1332e-05], device='cuda:0') 100 0.0001 changing lr epoch 113, time 251.30, cls_loss 0.0019 cls_loss_mapping 0.0059 cls_loss_causal 0.5498 re_mapping 0.0065 re_causal 0.0209 /// teacc 98.97 lr 0.00010000 Epoch 115, weight, value: tensor([[-0.1046, -0.0170, 0.1300, ..., 0.0392, -0.1100, -0.1030], [-0.0715, -0.0752, -0.0086, ..., -0.0744, -0.0803, -0.0405], [ 0.1402, -0.0369, -0.1037, ..., -0.0837, -0.1029, -0.1458], ..., [-0.0413, -0.0068, -0.0994, ..., -0.0327, 0.0470, 0.0067], [-0.1030, -0.0763, -0.0129, ..., -0.0870, -0.0026, -0.0193], [-0.0337, -0.0493, -0.0399, ..., 0.0795, 0.0301, -0.1173]], device='cuda:0'), grad: tensor([[ 6.6590e-07, 1.9651e-07, -1.3364e-06, ..., -8.5216e-07, 1.4333e-06, 1.4119e-06], [ 1.3122e-06, 8.4843e-07, 5.1782e-07, ..., 6.6403e-07, 8.3297e-06, 1.0587e-05], [ 5.0925e-06, 3.1181e-06, 4.0606e-07, ..., 9.2573e-07, 6.5379e-06, 8.7097e-06], ..., [ 2.4326e-06, 1.9036e-06, 2.3283e-07, ..., 6.6794e-06, 3.1646e-06, 7.9256e-07], [ 5.4482e-07, 5.6066e-07, 1.1459e-05, ..., 1.1744e-06, 6.0678e-05, 6.7651e-05], [ 6.9942e-07, -1.3011e-06, 3.6880e-07, ..., -1.0431e-05, -7.5847e-06, -6.1467e-08]], device='cuda:0') Epoch 115, bias, value: tensor([ 0.0044, 0.0141, 0.0175, -0.0032, 0.0184, -0.0053, -0.0131, -0.0085, 0.0189, -0.0095], device='cuda:0'), grad: tensor([ 4.2245e-06, 5.1737e-05, 5.8323e-05, -1.7345e-05, 6.2168e-05, -2.4331e-04, -1.3661e-04, 1.9237e-05, 2.1255e-04, -1.1489e-05], device='cuda:0') 100 0.0001 changing lr epoch 114, time 250.87, cls_loss 0.0031 cls_loss_mapping 0.0085 cls_loss_causal 0.5928 re_mapping 0.0061 re_causal 0.0200 /// teacc 98.92 lr 0.00010000 Epoch 116, weight, value: tensor([[-0.1050, -0.0173, 0.1309, ..., 0.0400, -0.1109, -0.1033], [-0.0726, -0.0759, -0.0087, ..., -0.0745, -0.0809, -0.0407], [ 0.1398, -0.0385, -0.1044, ..., -0.0841, -0.1042, -0.1462], ..., [-0.0397, -0.0052, -0.1002, ..., -0.0328, 0.0477, 0.0068], [-0.1037, -0.0767, -0.0131, ..., -0.0876, -0.0028, -0.0195], [-0.0342, -0.0498, -0.0422, ..., 0.0790, 0.0300, -0.1178]], device='cuda:0'), grad: tensor([[ 7.1339e-07, 3.1665e-08, -2.2352e-07, ..., 1.2107e-07, 5.4482e-07, 5.4855e-07], [ 1.2238e-06, 6.9849e-08, 4.0047e-08, ..., 6.2771e-07, 7.2923e-07, 4.1630e-07], [-6.9328e-06, 6.6124e-08, 1.2852e-07, ..., 8.3167e-07, 7.8324e-07, 4.2096e-07], ..., [ 7.1619e-07, 9.2201e-08, 2.9802e-08, ..., 7.6462e-07, -1.2843e-06, 8.7544e-08], [ 1.7555e-06, 5.1223e-08, 5.7742e-08, ..., 6.4448e-07, 2.8107e-06, 3.2876e-06], [ 2.0862e-07, 6.9849e-08, 1.2107e-07, ..., 3.3434e-06, 1.8505e-06, 9.2573e-07]], device='cuda:0') Epoch 116, bias, value: tensor([ 0.0051, 0.0153, 0.0161, -0.0025, 0.0189, -0.0052, -0.0129, -0.0082, 0.0187, -0.0106], device='cuda:0'), grad: tensor([ 3.1032e-06, 2.0601e-06, -5.8301e-06, 7.7114e-06, -2.1875e-05, -2.2560e-05, 1.5423e-05, -1.3513e-06, 1.2144e-05, 1.1064e-05], device='cuda:0') 100 0.0001 changing lr epoch 115, time 251.00, cls_loss 0.0040 cls_loss_mapping 0.0106 cls_loss_causal 0.5624 re_mapping 0.0067 re_causal 0.0198 /// teacc 99.01 lr 0.00010000 Epoch 117, weight, value: tensor([[-0.1059, -0.0174, 0.1311, ..., 0.0397, -0.1117, -0.1037], [-0.0730, -0.0762, -0.0088, ..., -0.0750, -0.0817, -0.0428], [ 0.1407, -0.0385, -0.1048, ..., -0.0855, -0.1047, -0.1461], ..., [-0.0397, -0.0054, -0.1013, ..., -0.0340, 0.0480, 0.0068], [-0.1050, -0.0770, -0.0130, ..., -0.0886, -0.0030, -0.0185], [-0.0343, -0.0498, -0.0426, ..., 0.0815, 0.0308, -0.1181]], device='cuda:0'), grad: tensor([[ 5.4762e-07, -2.5705e-07, -2.9597e-06, ..., 1.4693e-05, 1.5512e-05, 5.8115e-07], [ 9.6764e-07, 5.3737e-07, 9.7137e-07, ..., 4.5858e-06, 2.8983e-06, 1.4901e-07], [ 4.6380e-06, 3.0752e-06, 5.0217e-06, ..., 1.1586e-05, 9.1791e-06, 2.2259e-07], ..., [ 1.2340e-06, 7.5251e-07, 4.0978e-06, ..., 1.6615e-05, 9.9465e-06, 1.0803e-07], [ 4.6100e-07, 3.3434e-07, 4.0680e-05, ..., 1.1390e-04, 9.5129e-05, 1.5832e-08], [ 5.5041e-07, 3.7067e-07, -1.0335e-04, ..., -2.7847e-04, -2.5177e-04, 2.4494e-07]], device='cuda:0') Epoch 117, bias, value: tensor([ 0.0047, 0.0144, 0.0163, -0.0030, 0.0167, -0.0051, -0.0132, -0.0078, 0.0197, -0.0085], device='cuda:0'), grad: tensor([ 8.5473e-05, 1.7300e-05, 7.5579e-05, 9.2611e-06, 7.3373e-05, 5.6028e-04, 4.8071e-05, 7.9989e-05, 6.0320e-04, -1.5526e-03], device='cuda:0') 100 0.0001 changing lr epoch 116, time 251.16, cls_loss 0.0022 cls_loss_mapping 0.0068 cls_loss_causal 0.5564 re_mapping 0.0063 re_causal 0.0201 /// teacc 98.99 lr 0.00010000 Epoch 118, weight, value: tensor([[-0.1068, -0.0175, 0.1313, ..., 0.0398, -0.1124, -0.1038], [-0.0732, -0.0765, -0.0093, ..., -0.0752, -0.0824, -0.0430], [ 0.1416, -0.0386, -0.1054, ..., -0.0862, -0.1051, -0.1464], ..., [-0.0398, -0.0055, -0.1020, ..., -0.0344, 0.0484, 0.0070], [-0.1053, -0.0772, -0.0131, ..., -0.0892, -0.0034, -0.0186], [-0.0347, -0.0499, -0.0428, ..., 0.0812, 0.0304, -0.1186]], device='cuda:0'), grad: tensor([[ 7.7300e-08, 1.5926e-07, 1.5390e-04, ..., -3.1944e-07, 2.9624e-05, 7.5996e-05], [ 1.8161e-07, 2.9616e-07, -5.1595e-07, ..., 1.4044e-06, 3.2000e-06, 1.2089e-06], [ 8.3167e-07, 2.6356e-07, 3.0696e-06, ..., 2.3842e-07, 2.9951e-06, 2.1961e-06], ..., [-1.5795e-05, 1.3597e-07, 5.4296e-07, ..., -4.8466e-06, -4.6432e-05, -3.6974e-06], [ 5.6811e-08, 1.5814e-06, 1.7658e-06, ..., 2.9057e-07, 2.2314e-06, 3.9786e-06], [ 5.4948e-07, 3.4273e-07, 1.6419e-06, ..., 3.8017e-06, 6.9179e-06, 1.7611e-06]], device='cuda:0') Epoch 118, bias, value: tensor([ 0.0044, 0.0142, 0.0163, -0.0018, 0.0171, -0.0062, -0.0133, -0.0076, 0.0197, -0.0089], device='cuda:0'), grad: tensor([ 3.4785e-04, 1.4938e-06, 1.6168e-05, 7.9274e-05, 1.0997e-05, 1.1332e-05, -3.7193e-04, -1.4794e-04, 1.9863e-05, 3.3379e-05], device='cuda:0') 100 0.0001 changing lr epoch 117, time 251.11, cls_loss 0.0017 cls_loss_mapping 0.0059 cls_loss_causal 0.5639 re_mapping 0.0065 re_causal 0.0206 /// teacc 99.09 lr 0.00010000 Epoch 119, weight, value: tensor([[-0.1073, -0.0176, 0.1315, ..., 0.0400, -0.1130, -0.1043], [-0.0735, -0.0767, -0.0094, ..., -0.0756, -0.0828, -0.0431], [ 0.1429, -0.0387, -0.1062, ..., -0.0865, -0.1055, -0.1463], ..., [-0.0401, -0.0056, -0.1023, ..., -0.0348, 0.0488, 0.0069], [-0.1055, -0.0774, -0.0132, ..., -0.0900, -0.0037, -0.0187], [-0.0371, -0.0500, -0.0430, ..., 0.0811, 0.0304, -0.1193]], device='cuda:0'), grad: tensor([[ 7.7393e-07, 8.3819e-09, -5.5954e-06, ..., -1.3150e-06, 8.9593e-07, 3.2969e-07], [ 4.0159e-06, 3.6322e-08, 2.3991e-06, ..., 7.7765e-07, 4.2766e-06, 3.4831e-07], [-2.0638e-05, 7.7300e-08, 7.9256e-07, ..., 4.5076e-07, 1.3761e-05, 4.4517e-07], ..., [-2.2128e-05, 4.1910e-08, 1.3318e-07, ..., 2.2016e-06, -2.2948e-05, -1.2759e-07], [ 1.7462e-06, 1.4901e-08, 8.2701e-07, ..., 2.8461e-06, -2.4028e-06, -2.3171e-06], [ 5.5507e-07, 3.1665e-08, 3.8743e-07, ..., -3.9078e-06, -1.8422e-06, 3.4459e-07]], device='cuda:0') Epoch 119, bias, value: tensor([ 0.0043, 0.0142, 0.0167, -0.0021, 0.0174, -0.0061, -0.0130, -0.0077, 0.0198, -0.0092], device='cuda:0'), grad: tensor([-5.3346e-06, 2.3633e-05, 2.4825e-05, 5.3883e-05, -5.7928e-07, 1.4603e-05, 5.3234e-06, -9.9719e-05, -1.0028e-05, -6.6347e-06], device='cuda:0') 100 0.0001 changing lr epoch 118, time 251.15, cls_loss 0.0018 cls_loss_mapping 0.0064 cls_loss_causal 0.5562 re_mapping 0.0064 re_causal 0.0202 /// teacc 99.00 lr 0.00010000 Epoch 120, weight, value: tensor([[-0.1078, -0.0179, 0.1319, ..., 0.0402, -0.1134, -0.1044], [-0.0737, -0.0770, -0.0100, ..., -0.0772, -0.0834, -0.0438], [ 0.1433, -0.0389, -0.1067, ..., -0.0873, -0.1057, -0.1468], ..., [-0.0403, -0.0059, -0.1029, ..., -0.0357, 0.0488, 0.0069], [-0.1059, -0.0775, -0.0132, ..., -0.0908, -0.0037, -0.0186], [-0.0371, -0.0502, -0.0432, ..., 0.0811, 0.0304, -0.1197]], device='cuda:0'), grad: tensor([[ 1.1176e-07, 2.0489e-08, -2.1651e-05, ..., -1.3866e-05, 2.8871e-07, 8.4750e-08], [ 4.0699e-07, 6.5193e-08, 1.1828e-07, ..., -2.0023e-07, 3.6508e-07, 6.7055e-08], [-1.0533e-06, 1.0803e-07, 1.0403e-06, ..., 9.1270e-07, 8.6054e-07, 2.0768e-07], ..., [ 2.7660e-07, 7.2643e-08, 2.5425e-07, ..., 6.0536e-07, -1.1865e-06, 4.1910e-08], [ 4.0513e-07, -8.0094e-08, 5.1446e-06, ..., 3.8147e-06, -5.2862e-06, -1.2098e-06], [ 3.0361e-07, 1.7602e-07, 1.2085e-05, ..., 1.1235e-05, 2.1327e-06, 1.5367e-07]], device='cuda:0') Epoch 120, bias, value: tensor([ 0.0044, 0.0143, 0.0166, -0.0024, 0.0179, -0.0060, -0.0134, -0.0081, 0.0200, -0.0093], device='cuda:0'), grad: tensor([-2.8312e-05, -1.0028e-05, 6.5304e-06, 5.1707e-06, -6.9775e-06, 1.8612e-05, 5.8934e-06, -5.9344e-06, -1.5870e-05, 3.0905e-05], device='cuda:0') 100 0.0001 changing lr epoch 119, time 251.45, cls_loss 0.0018 cls_loss_mapping 0.0048 cls_loss_causal 0.5526 re_mapping 0.0061 re_causal 0.0193 /// teacc 99.00 lr 0.00010000 Epoch 121, weight, value: tensor([[-0.1084, -0.0180, 0.1323, ..., 0.0405, -0.1140, -0.1045], [-0.0744, -0.0772, -0.0101, ..., -0.0760, -0.0840, -0.0440], [ 0.1441, -0.0392, -0.1071, ..., -0.0876, -0.1060, -0.1468], ..., [-0.0405, -0.0060, -0.1032, ..., -0.0368, 0.0494, 0.0071], [-0.1063, -0.0777, -0.0137, ..., -0.0924, -0.0043, -0.0189], [-0.0374, -0.0505, -0.0435, ..., 0.0810, 0.0300, -0.1203]], device='cuda:0'), grad: tensor([[ 3.5148e-06, 1.4901e-08, 2.2985e-06, ..., 2.1327e-07, 3.1106e-07, 3.1386e-07], [ 2.9523e-07, 6.0536e-08, 1.9092e-07, ..., -3.1292e-07, 2.8498e-07, 1.7416e-07], [-8.4750e-08, 1.0710e-07, 8.2795e-07, ..., 3.8650e-07, 2.3469e-07, 3.2783e-07], ..., [ 5.1409e-07, 4.8429e-08, 2.9337e-07, ..., 3.8184e-07, -8.1677e-07, 2.0489e-08], [ 1.2303e-06, 6.3330e-08, 1.2033e-06, ..., 3.5763e-07, 4.7870e-07, 4.7777e-07], [ 3.8058e-05, 1.3597e-07, 2.7806e-05, ..., 2.5723e-06, 7.5437e-08, 9.7789e-08]], device='cuda:0') Epoch 121, bias, value: tensor([ 0.0045, 0.0150, 0.0168, -0.0026, 0.0182, -0.0059, -0.0131, -0.0086, 0.0195, -0.0097], device='cuda:0'), grad: tensor([ 1.9655e-05, -1.1094e-05, 6.9812e-06, 1.2726e-05, 3.3826e-06, -2.4390e-04, -2.4438e-06, 2.6543e-06, 7.1079e-06, 2.0456e-04], device='cuda:0') 100 0.0001 changing lr epoch 120, time 250.94, cls_loss 0.0024 cls_loss_mapping 0.0064 cls_loss_causal 0.5702 re_mapping 0.0057 re_causal 0.0194 /// teacc 98.99 lr 0.00010000 Epoch 122, weight, value: tensor([[-0.1088, -0.0180, 0.1327, ..., 0.0405, -0.1144, -0.1046], [-0.0747, -0.0774, -0.0100, ..., -0.0761, -0.0844, -0.0442], [ 0.1444, -0.0393, -0.1077, ..., -0.0882, -0.1063, -0.1473], ..., [-0.0405, -0.0061, -0.1039, ..., -0.0370, 0.0497, 0.0072], [-0.1064, -0.0778, -0.0139, ..., -0.0934, -0.0046, -0.0191], [-0.0377, -0.0507, -0.0437, ..., 0.0809, 0.0301, -0.1207]], device='cuda:0'), grad: tensor([[ 5.2340e-07, 0.0000e+00, -4.9826e-07, ..., 1.8440e-07, 4.0513e-07, 4.9360e-08], [ 8.7917e-07, 1.8626e-09, 2.1420e-08, ..., 6.6217e-07, 8.3260e-07, 1.4901e-08], [-5.4203e-06, 2.7940e-09, 6.8918e-08, ..., 4.6100e-07, 4.8336e-07, 4.0047e-08], ..., [ 9.9838e-07, 1.8626e-09, 3.6322e-08, ..., 8.3633e-07, -5.0105e-07, 8.3819e-09], [ 1.5264e-06, 9.3132e-10, 1.0524e-07, ..., 3.7681e-06, 3.0845e-06, 7.9162e-08], [ 3.8184e-08, 9.3132e-10, 1.7136e-07, ..., 4.3869e-05, 3.6001e-05, 1.9558e-08]], device='cuda:0') Epoch 122, bias, value: tensor([ 0.0045, 0.0147, 0.0162, -0.0027, 0.0187, -0.0058, -0.0132, -0.0080, 0.0191, -0.0102], device='cuda:0'), grad: tensor([ 1.3877e-06, 2.2762e-06, -6.5863e-06, 6.0275e-06, -1.0997e-04, -1.5907e-06, 2.5675e-05, -1.2508e-06, 1.3806e-05, 7.0333e-05], device='cuda:0') 100 0.0001 changing lr epoch 121, time 251.26, cls_loss 0.0017 cls_loss_mapping 0.0041 cls_loss_causal 0.5591 re_mapping 0.0061 re_causal 0.0195 /// teacc 98.90 lr 0.00010000 Epoch 123, weight, value: tensor([[-0.1096, -0.0181, 0.1324, ..., 0.0405, -0.1151, -0.1054], [-0.0749, -0.0775, -0.0110, ..., -0.0762, -0.0848, -0.0446], [ 0.1451, -0.0393, -0.1081, ..., -0.0887, -0.1066, -0.1477], ..., [-0.0409, -0.0062, -0.1048, ..., -0.0374, 0.0501, 0.0073], [-0.1073, -0.0779, -0.0143, ..., -0.0939, -0.0051, -0.0193], [-0.0379, -0.0507, -0.0440, ..., 0.0809, 0.0300, -0.1210]], device='cuda:0'), grad: tensor([[ 1.2526e-06, 3.1665e-08, -4.6305e-06, ..., -3.0342e-06, 6.2771e-07, 1.7043e-07], [ 1.2852e-06, 1.0058e-07, -1.3597e-07, ..., -3.6974e-07, 2.2631e-06, 1.3039e-08], [-1.1288e-05, 2.0489e-07, 6.1281e-07, ..., 4.1816e-07, 8.3074e-06, 3.5390e-08], ..., [ 4.9882e-06, 9.4995e-08, 2.7101e-07, ..., 7.5717e-07, -3.4552e-06, 7.4506e-09], [-4.2841e-06, 7.7020e-07, 3.3528e-07, ..., 4.2655e-07, -1.7568e-05, 7.4506e-09], [ 8.5868e-07, 1.2480e-07, 2.9653e-06, ..., 1.5181e-07, 1.4296e-06, 2.6077e-08]], device='cuda:0') Epoch 123, bias, value: tensor([ 0.0042, 0.0149, 0.0164, -0.0030, 0.0188, -0.0054, -0.0123, -0.0083, 0.0193, -0.0104], device='cuda:0'), grad: tensor([ 4.7758e-06, 1.7419e-05, 4.8518e-05, 4.0174e-05, 2.2247e-05, 5.1022e-05, 1.7017e-05, 4.1127e-06, -2.6751e-04, 6.2168e-05], device='cuda:0') 100 0.0001 changing lr epoch 122, time 251.47, cls_loss 0.0019 cls_loss_mapping 0.0052 cls_loss_causal 0.5382 re_mapping 0.0056 re_causal 0.0188 /// teacc 99.00 lr 0.00010000 Epoch 124, weight, value: tensor([[-0.1113, -0.0179, 0.1341, ..., 0.0411, -0.1137, -0.1048], [-0.0752, -0.0778, -0.0116, ..., -0.0764, -0.0859, -0.0448], [ 0.1457, -0.0395, -0.1092, ..., -0.0897, -0.1072, -0.1480], ..., [-0.0411, -0.0064, -0.1070, ..., -0.0382, 0.0503, 0.0073], [-0.1075, -0.0782, -0.0146, ..., -0.0950, -0.0054, -0.0195], [-0.0374, -0.0510, -0.0445, ..., 0.0809, 0.0299, -0.1212]], device='cuda:0'), grad: tensor([[ 2.2035e-06, 1.7695e-08, -8.8383e-07, ..., -2.8498e-07, 1.5004e-06, 8.9779e-07], [ 1.9334e-06, 4.2841e-08, 1.7788e-07, ..., 1.5674e-06, 8.1025e-07, 1.9372e-07], [-4.0084e-05, 7.6368e-08, 5.7183e-07, ..., 7.7486e-07, 5.2527e-07, 1.8068e-07], ..., [ 2.3797e-05, 1.5087e-07, 2.9616e-07, ..., 2.8070e-06, 1.1064e-06, 5.3085e-08], [ 8.5309e-06, 1.2573e-07, 3.9451e-06, ..., 1.0878e-06, 2.5723e-06, 1.8850e-06], [ 2.8778e-07, 5.0291e-08, 8.0839e-07, ..., 2.7031e-05, 1.2375e-05, 8.4750e-08]], device='cuda:0') Epoch 124, bias, value: tensor([ 0.0054, 0.0149, 0.0161, -0.0033, 0.0189, -0.0058, -0.0121, -0.0084, 0.0196, -0.0105], device='cuda:0'), grad: tensor([ 8.9630e-06, -2.7977e-06, -1.1486e-04, 6.7540e-06, -6.3419e-05, 2.8625e-05, -2.9713e-05, 8.0705e-05, 3.5167e-05, 5.0604e-05], device='cuda:0') 100 0.0001 changing lr epoch 123, time 251.66, cls_loss 0.0016 cls_loss_mapping 0.0058 cls_loss_causal 0.5395 re_mapping 0.0063 re_causal 0.0199 /// teacc 99.02 lr 0.00010000 Epoch 125, weight, value: tensor([[-0.1128, -0.0179, 0.1347, ..., 0.0413, -0.1141, -0.1048], [-0.0755, -0.0780, -0.0118, ..., -0.0766, -0.0864, -0.0449], [ 0.1468, -0.0396, -0.1098, ..., -0.0901, -0.1074, -0.1481], ..., [-0.0407, -0.0065, -0.1078, ..., -0.0391, 0.0503, 0.0072], [-0.1088, -0.0784, -0.0146, ..., -0.0958, -0.0058, -0.0196], [-0.0405, -0.0512, -0.0448, ..., 0.0809, 0.0300, -0.1212]], device='cuda:0'), grad: tensor([[ 6.1467e-08, 7.4506e-09, 1.8440e-07, ..., 7.4506e-08, 3.9116e-07, 3.4552e-07], [ 2.1700e-07, 3.0734e-08, 6.1467e-08, ..., 8.8476e-08, 9.4995e-07, 7.9162e-08], [-6.0815e-07, 9.0338e-08, 7.8231e-08, ..., 9.2201e-08, 2.9989e-07, 8.0094e-08], ..., [ 4.5355e-07, -1.3970e-08, 6.5193e-09, ..., 4.9733e-07, -3.8929e-06, 2.7940e-08], [ 6.9849e-08, 3.5390e-08, 2.8592e-07, ..., 5.9605e-07, 6.2678e-07, 4.0885e-07], [ 4.7497e-08, 4.0047e-08, 5.8673e-08, ..., 1.6481e-05, 9.0674e-06, 1.0431e-07]], device='cuda:0') Epoch 125, bias, value: tensor([ 0.0056, 0.0149, 0.0163, -0.0037, 0.0192, -0.0059, -0.0120, -0.0083, 0.0191, -0.0108], device='cuda:0'), grad: tensor([ 1.6456e-06, 1.3346e-06, 7.9721e-07, 2.2035e-06, -3.4273e-05, -2.6673e-06, -1.7257e-06, -1.5169e-05, 3.6247e-06, 4.4167e-05], device='cuda:0') 100 0.0001 changing lr epoch 124, time 251.28, cls_loss 0.0018 cls_loss_mapping 0.0060 cls_loss_causal 0.5423 re_mapping 0.0061 re_causal 0.0193 /// teacc 98.95 lr 0.00010000 Epoch 126, weight, value: tensor([[-0.1138, -0.0181, 0.1353, ..., 0.0415, -0.1145, -0.1049], [-0.0759, -0.0781, -0.0118, ..., -0.0768, -0.0870, -0.0449], [ 0.1475, -0.0397, -0.1102, ..., -0.0906, -0.1076, -0.1483], ..., [-0.0410, -0.0072, -0.1084, ..., -0.0387, 0.0516, 0.0073], [-0.1091, -0.0785, -0.0147, ..., -0.0967, -0.0060, -0.0197], [-0.0405, -0.0515, -0.0454, ..., 0.0808, 0.0297, -0.1214]], device='cuda:0'), grad: tensor([[ 1.9465e-07, -9.3132e-09, -1.4482e-06, ..., 1.4985e-06, 6.8173e-07, 2.9989e-07], [ 3.3155e-07, 9.1270e-08, 1.5739e-07, ..., 1.2312e-06, 7.9069e-07, 5.4948e-08], [-6.0536e-07, 3.4366e-07, 9.6858e-08, ..., 5.4482e-07, 2.7139e-06, 9.3132e-08], ..., [-5.0757e-07, -4.3306e-07, 2.3842e-07, ..., 8.8811e-06, -2.5108e-06, 2.5146e-08], [ 2.3283e-07, 6.2399e-08, -3.0383e-05, ..., 2.7716e-06, -1.3486e-05, -1.7524e-05], [ 6.5193e-08, 7.1712e-08, 1.1576e-06, ..., -4.3452e-05, -6.2287e-06, 2.6263e-07]], device='cuda:0') Epoch 126, bias, value: tensor([ 0.0057, 0.0148, 0.0165, -0.0036, 0.0193, -0.0061, -0.0120, -0.0080, 0.0195, -0.0114], device='cuda:0'), grad: tensor([ 5.1074e-06, 7.3854e-07, 7.9423e-06, 1.2934e-05, 5.7489e-05, 2.1502e-05, 5.8144e-05, 1.0341e-05, -8.1360e-05, -9.2745e-05], device='cuda:0') 100 0.0001 changing lr epoch 125, time 251.17, cls_loss 0.0020 cls_loss_mapping 0.0061 cls_loss_causal 0.5413 re_mapping 0.0061 re_causal 0.0189 /// teacc 98.90 lr 0.00010000 Epoch 127, weight, value: tensor([[-0.1156, -0.0186, 0.1355, ..., 0.0416, -0.1154, -0.1052], [-0.0761, -0.0787, -0.0122, ..., -0.0772, -0.0876, -0.0452], [ 0.1481, -0.0398, -0.1100, ..., -0.0911, -0.1079, -0.1486], ..., [-0.0412, -0.0077, -0.1091, ..., -0.0391, 0.0517, 0.0069], [-0.1096, -0.0791, -0.0153, ..., -0.0975, -0.0065, -0.0198], [-0.0406, -0.0519, -0.0457, ..., 0.0817, 0.0306, -0.1217]], device='cuda:0'), grad: tensor([[ 4.5355e-07, 3.7253e-09, -2.2929e-06, ..., -9.3412e-07, 2.4382e-06, 2.5760e-06], [ 5.0291e-08, 3.7253e-09, 3.2969e-07, ..., 8.3260e-07, 8.9221e-07, 5.3179e-07], [ 1.1520e-06, 5.5879e-09, 2.5313e-06, ..., 8.2701e-07, 9.8124e-06, 1.0349e-05], ..., [ 1.2852e-07, 7.4506e-09, 5.3551e-07, ..., 3.4589e-06, 1.7993e-06, 3.9674e-07], [ 1.7975e-07, 6.5193e-09, 9.0152e-07, ..., 1.0421e-06, -4.1127e-06, -1.0198e-06], [ 2.2817e-07, 3.7253e-09, 1.0822e-06, ..., 4.6529e-06, 3.6471e-06, 1.8757e-06]], device='cuda:0') Epoch 127, bias, value: tensor([ 0.0055, 0.0147, 0.0166, -0.0035, 0.0186, -0.0056, -0.0120, -0.0080, 0.0190, -0.0107], device='cuda:0'), grad: tensor([ 6.3777e-06, -3.3863e-06, 3.3945e-05, 3.8475e-05, 7.3910e-06, -9.3803e-06, -6.5148e-05, 1.4067e-05, -6.8188e-05, 4.5836e-05], device='cuda:0') 100 0.0001 changing lr epoch 126, time 250.98, cls_loss 0.0020 cls_loss_mapping 0.0053 cls_loss_causal 0.5343 re_mapping 0.0062 re_causal 0.0194 /// teacc 98.94 lr 0.00010000 Epoch 128, weight, value: tensor([[-0.1166, -0.0187, 0.1358, ..., 0.0418, -0.1164, -0.1055], [-0.0763, -0.0789, -0.0117, ..., -0.0776, -0.0903, -0.0453], [ 0.1496, -0.0400, -0.1107, ..., -0.0919, -0.1083, -0.1489], ..., [-0.0420, -0.0080, -0.1098, ..., -0.0396, 0.0530, 0.0069], [-0.1112, -0.0794, -0.0157, ..., -0.0983, -0.0068, -0.0199], [-0.0405, -0.0521, -0.0460, ..., 0.0814, 0.0304, -0.1219]], device='cuda:0'), grad: tensor([[ 6.7428e-07, 7.2177e-07, -2.4475e-06, ..., -1.1632e-06, 2.4773e-07, 4.2841e-08], [ 2.2072e-07, 1.5926e-07, 9.2201e-08, ..., 6.7428e-07, 5.2433e-07, 8.3819e-09], [ 2.1048e-07, 5.9512e-07, 3.6322e-07, ..., 5.0198e-07, 5.8580e-07, 1.4901e-08], ..., [ 3.1572e-07, 2.5425e-07, 1.0151e-07, ..., 2.6338e-06, -1.7229e-07, 3.7253e-09], [ 1.2107e-06, 1.3085e-06, 2.9989e-07, ..., 7.5530e-07, -8.4005e-07, 5.2154e-08], [ 2.7791e-06, 3.0361e-06, 6.7428e-07, ..., -7.1675e-06, -1.6466e-06, 1.6764e-08]], device='cuda:0') Epoch 128, bias, value: tensor([ 0.0054, 0.0143, 0.0177, -0.0041, 0.0192, -0.0059, -0.0114, -0.0083, 0.0195, -0.0112], device='cuda:0'), grad: tensor([ 3.4384e-06, 3.6210e-06, 8.3968e-06, -5.0455e-05, 4.7833e-06, 9.3505e-06, 1.3150e-06, 1.9837e-06, 2.8349e-06, 1.4715e-05], device='cuda:0') 100 0.0001 changing lr epoch 127, time 251.40, cls_loss 0.0019 cls_loss_mapping 0.0059 cls_loss_causal 0.5312 re_mapping 0.0056 re_causal 0.0179 /// teacc 99.04 lr 0.00010000 Epoch 129, weight, value: tensor([[-0.1199, -0.0196, 0.1365, ..., 0.0421, -0.1177, -0.1064], [-0.0772, -0.0796, -0.0123, ..., -0.0767, -0.0907, -0.0454], [ 0.1502, -0.0400, -0.1118, ..., -0.0924, -0.1087, -0.1492], ..., [-0.0416, -0.0081, -0.1099, ..., -0.0412, 0.0534, 0.0068], [-0.1127, -0.0799, -0.0161, ..., -0.1003, -0.0071, -0.0201], [-0.0409, -0.0533, -0.0463, ..., 0.0815, 0.0304, -0.1223]], device='cuda:0'), grad: tensor([[ 5.9605e-08, 1.4156e-07, -3.3528e-06, ..., -5.8115e-07, 4.4610e-07, 9.8720e-08], [ 1.2387e-07, 3.3900e-07, -5.2154e-08, ..., 1.6885e-06, 7.2550e-07, 3.6322e-08], [ 5.4017e-07, 2.1625e-06, 9.1176e-07, ..., 1.1079e-05, 4.4629e-06, 1.5926e-07], ..., [ 6.8359e-07, 2.5071e-06, 2.3376e-07, ..., 1.4238e-05, 5.6289e-06, 4.0047e-08], [ 1.3784e-07, 2.2631e-07, 6.1002e-07, ..., 1.4920e-06, 6.2026e-07, -1.2293e-07], [ 4.2375e-07, 1.7211e-06, 8.3726e-07, ..., 6.7241e-06, 2.6282e-06, 7.3574e-08]], device='cuda:0') Epoch 129, bias, value: tensor([ 0.0048, 0.0141, 0.0176, -0.0033, 0.0192, -0.0056, -0.0115, -0.0080, 0.0188, -0.0112], device='cuda:0'), grad: tensor([-1.8170e-06, 3.2037e-06, 2.4199e-05, -4.0270e-06, -7.5817e-05, 6.7353e-06, 3.0156e-06, 2.8804e-05, 2.1346e-06, 1.3545e-05], device='cuda:0') 100 0.0001 changing lr epoch 128, time 251.61, cls_loss 0.0018 cls_loss_mapping 0.0057 cls_loss_causal 0.5298 re_mapping 0.0057 re_causal 0.0179 /// teacc 99.04 lr 0.00010000 Epoch 130, weight, value: tensor([[-0.1206, -0.0199, 0.1372, ..., 0.0405, -0.1184, -0.1064], [-0.0778, -0.0802, -0.0127, ..., -0.0772, -0.0911, -0.0456], [ 0.1509, -0.0405, -0.1126, ..., -0.0938, -0.1093, -0.1495], ..., [-0.0418, -0.0051, -0.1110, ..., -0.0410, 0.0544, 0.0069], [-0.1132, -0.0809, -0.0165, ..., -0.1008, -0.0075, -0.0203], [-0.0410, -0.0538, -0.0465, ..., 0.0823, 0.0303, -0.1227]], device='cuda:0'), grad: tensor([[ 1.6391e-07, 4.8429e-08, -3.9674e-06, ..., -2.2706e-06, 7.8976e-07, 9.1642e-07], [ 1.4463e-06, 1.0524e-07, -1.4529e-07, ..., 9.9652e-08, 9.0618e-07, 7.9442e-07], [-5.7444e-06, 1.6950e-07, 6.2771e-07, ..., 4.2282e-07, 5.8487e-07, 4.2189e-07], ..., [ 5.1595e-07, 9.4995e-08, 3.2783e-07, ..., 1.5190e-06, -9.6299e-07, 9.5926e-08], [ 4.1313e-06, 3.7905e-07, 7.9535e-07, ..., 1.8440e-07, 1.3830e-06, 1.9409e-06], [ 2.8405e-07, 9.4995e-08, 2.2035e-06, ..., 1.6754e-06, 2.5313e-06, 1.6559e-06]], device='cuda:0') Epoch 130, bias, value: tensor([ 0.0035, 0.0141, 0.0177, -0.0036, 0.0192, -0.0055, -0.0116, -0.0078, 0.0187, -0.0107], device='cuda:0'), grad: tensor([ 4.8131e-06, -5.2862e-06, -5.4538e-06, -1.2666e-07, -2.1178e-06, 3.6918e-06, -1.9863e-05, 6.3032e-06, -1.0580e-05, 2.8566e-05], device='cuda:0') 100 0.0001 changing lr epoch 129, time 251.70, cls_loss 0.0016 cls_loss_mapping 0.0061 cls_loss_causal 0.5686 re_mapping 0.0059 re_causal 0.0190 /// teacc 99.01 lr 0.00010000 Epoch 131, weight, value: tensor([[-0.1212, -0.0201, 0.1377, ..., 0.0408, -0.1194, -0.1067], [-0.0785, -0.0811, -0.0131, ..., -0.0775, -0.0915, -0.0458], [ 0.1523, -0.0405, -0.1137, ..., -0.0944, -0.1095, -0.1497], ..., [-0.0418, -0.0052, -0.1117, ..., -0.0414, 0.0547, 0.0068], [-0.1136, -0.0814, -0.0165, ..., -0.1010, -0.0076, -0.0204], [-0.0426, -0.0545, -0.0469, ..., 0.0823, 0.0302, -0.1231]], device='cuda:0'), grad: tensor([[ 7.2643e-08, 2.0210e-07, 1.2405e-06, ..., 2.6729e-07, 9.9186e-07, 1.9763e-06], [ 3.4785e-07, 1.7229e-08, 1.0012e-07, ..., 6.2445e-07, 4.0606e-07, 1.2573e-07], [-3.3751e-06, 1.3039e-08, 7.0781e-08, ..., 1.0803e-07, 2.0443e-07, 9.5461e-08], ..., [ 1.8068e-07, 8.8476e-09, 9.7789e-09, ..., 3.8892e-06, 2.2724e-06, 1.7229e-08], [ 2.3860e-06, 6.2399e-08, 3.6601e-07, ..., 1.2834e-06, 4.7451e-07, 5.8115e-07], [ 2.1420e-08, 1.3504e-08, 4.7497e-08, ..., -9.5740e-06, -4.1723e-06, 1.2061e-07]], device='cuda:0') Epoch 131, bias, value: tensor([ 0.0037, 0.0142, 0.0179, -0.0036, 0.0192, -0.0055, -0.0121, -0.0078, 0.0190, -0.0110], device='cuda:0'), grad: tensor([ 3.9637e-06, -1.3066e-06, -2.5313e-06, 1.2040e-05, 4.9882e-06, -1.0714e-05, -4.1015e-06, 1.0714e-05, 1.0841e-05, -2.3961e-05], device='cuda:0') 100 0.0001 changing lr epoch 130, time 251.44, cls_loss 0.0019 cls_loss_mapping 0.0065 cls_loss_causal 0.5534 re_mapping 0.0059 re_causal 0.0182 /// teacc 99.02 lr 0.00010000 Epoch 132, weight, value: tensor([[-0.1215, -0.0199, 0.1387, ..., 0.0411, -0.1198, -0.1069], [-0.0788, -0.0818, -0.0130, ..., -0.0778, -0.0921, -0.0459], [ 0.1529, -0.0408, -0.1147, ..., -0.0956, -0.1101, -0.1501], ..., [-0.0420, -0.0053, -0.1126, ..., -0.0417, 0.0553, 0.0067], [-0.1145, -0.0813, -0.0168, ..., -0.1017, -0.0079, -0.0208], [-0.0427, -0.0555, -0.0484, ..., 0.0818, 0.0292, -0.1235]], device='cuda:0'), grad: tensor([[ 1.3039e-08, 9.3132e-09, -5.4389e-07, ..., -6.4261e-08, 5.8394e-07, 1.0245e-07], [ 7.8883e-07, 5.0757e-08, 9.3132e-09, ..., 1.3234e-06, 5.3160e-06, 4.0513e-08], [-9.2434e-07, 4.7497e-08, 8.9873e-08, ..., 4.6520e-07, 1.4743e-06, 8.6147e-08], ..., [ 2.0163e-07, 5.9139e-08, 2.9337e-08, ..., -1.2028e-04, -5.0640e-04, 5.2154e-08], [ 2.8871e-08, -5.6345e-07, 8.0094e-08, ..., -1.0226e-06, 3.0361e-07, -1.2731e-06], [ 1.8161e-08, 3.4692e-07, 1.8068e-07, ..., 1.1629e-04, 4.9448e-04, 8.9360e-07]], device='cuda:0') Epoch 132, bias, value: tensor([ 0.0042, 0.0144, 0.0179, -0.0029, 0.0200, -0.0069, -0.0119, -0.0077, 0.0185, -0.0118], device='cuda:0'), grad: tensor([ 1.7136e-06, 1.5840e-05, 3.7495e-06, 4.1015e-06, 1.1750e-05, 3.3863e-06, 1.5702e-06, -1.3952e-03, -2.0325e-05, 1.3742e-03], device='cuda:0') 100 0.0001 changing lr epoch 131, time 251.64, cls_loss 0.0016 cls_loss_mapping 0.0049 cls_loss_causal 0.5526 re_mapping 0.0061 re_causal 0.0192 /// teacc 98.92 lr 0.00010000 Epoch 133, weight, value: tensor([[-0.1218, -0.0200, 0.1373, ..., 0.0413, -0.1203, -0.1099], [-0.0791, -0.0823, -0.0133, ..., -0.0787, -0.0929, -0.0460], [ 0.1536, -0.0411, -0.1154, ..., -0.0962, -0.1106, -0.1509], ..., [-0.0425, -0.0053, -0.1133, ..., -0.0407, 0.0571, 0.0070], [-0.1148, -0.0815, -0.0171, ..., -0.1028, -0.0084, -0.0210], [-0.0427, -0.0563, -0.0487, ..., 0.0819, 0.0290, -0.1238]], device='cuda:0'), grad: tensor([[ 2.0210e-07, 1.3504e-08, -5.6326e-06, ..., -5.2266e-06, 2.4633e-07, 5.4948e-08], [ 2.3236e-07, 5.7276e-08, 1.9604e-07, ..., 3.5297e-07, 8.7544e-08, 1.0245e-08], [ 2.7614e-07, 7.8697e-08, 5.7137e-07, ..., 1.1222e-06, 1.3411e-07, 7.4040e-08], ..., [ 4.4610e-07, 1.1036e-07, 1.5087e-07, ..., 2.8927e-06, 1.5385e-06, 7.4506e-09], [ 2.8498e-07, 1.9092e-08, 3.2270e-07, ..., 2.5779e-06, 1.0505e-06, 2.3283e-09], [-2.7250e-06, 7.9628e-08, 1.5274e-06, ..., -1.4849e-05, -4.4405e-06, 1.6764e-08]], device='cuda:0') Epoch 133, bias, value: tensor([ 0.0030, 0.0143, 0.0180, -0.0035, 0.0197, -0.0063, -0.0100, -0.0073, 0.0179, -0.0118], device='cuda:0'), grad: tensor([-1.0453e-05, -3.9011e-05, 5.8711e-06, 1.3106e-05, 5.9195e-06, 4.5598e-06, 5.9605e-06, 3.9130e-05, 9.9912e-06, -3.5197e-05], device='cuda:0') 100 0.0001 changing lr epoch 132, time 251.62, cls_loss 0.0016 cls_loss_mapping 0.0054 cls_loss_causal 0.5640 re_mapping 0.0059 re_causal 0.0187 /// teacc 98.95 lr 0.00010000 Epoch 134, weight, value: tensor([[-0.1233, -0.0201, 0.1377, ..., 0.0415, -0.1207, -0.1102], [-0.0794, -0.0828, -0.0133, ..., -0.0792, -0.0938, -0.0461], [ 0.1541, -0.0412, -0.1165, ..., -0.0975, -0.1112, -0.1511], ..., [-0.0427, -0.0066, -0.1141, ..., -0.0412, 0.0568, 0.0069], [-0.1156, -0.0817, -0.0171, ..., -0.1051, -0.0094, -0.0210], [-0.0424, -0.0566, -0.0494, ..., 0.0822, 0.0295, -0.1239]], device='cuda:0'), grad: tensor([[ 6.1467e-08, 1.1642e-08, -9.5461e-07, ..., -7.6927e-07, 4.0280e-07, 3.7998e-07], [ 1.5041e-07, 6.6124e-08, 1.7043e-07, ..., -4.9220e-07, 1.0058e-06, 3.0082e-07], [-3.2876e-07, 6.7055e-08, 4.4098e-07, ..., 1.9046e-07, 5.1828e-07, 3.1246e-07], ..., [ 1.1735e-07, 1.5832e-08, 6.6124e-08, ..., 5.1875e-07, -1.9729e-05, 1.3784e-07], [ 1.9558e-07, 4.7497e-08, -8.0690e-06, ..., 3.7998e-07, 3.7067e-06, -5.0701e-06], [ 8.0094e-08, 5.6811e-08, 1.3430e-06, ..., 9.1735e-08, 8.9705e-06, 1.3970e-06]], device='cuda:0') Epoch 134, bias, value: tensor([ 0.0030, 0.0141, 0.0178, -0.0026, 0.0197, -0.0063, -0.0103, -0.0076, 0.0171, -0.0112], device='cuda:0'), grad: tensor([ 5.5786e-07, -1.5786e-07, 2.7008e-06, 4.0643e-06, 6.8434e-06, 1.9535e-05, 9.7603e-06, -6.4850e-05, -1.6078e-05, 3.7551e-05], device='cuda:0') 100 0.0001 changing lr epoch 133, time 251.57, cls_loss 0.0016 cls_loss_mapping 0.0040 cls_loss_causal 0.5475 re_mapping 0.0056 re_causal 0.0180 /// teacc 99.04 lr 0.00010000 Epoch 135, weight, value: tensor([[-0.1241, -0.0201, 0.1380, ..., 0.0418, -0.1211, -0.1104], [-0.0823, -0.0832, -0.0130, ..., -0.0800, -0.0943, -0.0459], [ 0.1550, -0.0414, -0.1175, ..., -0.0981, -0.1115, -0.1524], ..., [-0.0416, -0.0067, -0.1152, ..., -0.0418, 0.0569, 0.0066], [-0.1159, -0.0820, -0.0167, ..., -0.1060, -0.0096, -0.0204], [-0.0426, -0.0568, -0.0500, ..., 0.0823, 0.0297, -0.1246]], device='cuda:0'), grad: tensor([[ 5.4017e-08, 7.5903e-08, 1.1781e-06, ..., -2.0582e-07, 6.3423e-07, 8.5542e-07], [ 2.9150e-07, 1.1781e-07, 1.1828e-07, ..., 2.2817e-07, 2.1979e-07, 1.2387e-07], [-4.0093e-07, 8.7544e-08, 5.5088e-07, ..., 1.8999e-07, 3.6694e-07, 5.2387e-07], ..., [ 3.3434e-07, 1.7555e-07, 3.0734e-08, ..., 6.1654e-07, 3.7532e-07, 1.8906e-07], [ 4.9220e-07, 9.2201e-08, 3.3341e-07, ..., 9.0711e-07, 7.1060e-07, 1.0291e-07], [ 1.1129e-07, 1.1362e-07, 4.1537e-07, ..., -2.6505e-06, -1.3476e-06, 4.5262e-07]], device='cuda:0') Epoch 135, bias, value: tensor([ 0.0030, 0.0132, 0.0180, -0.0027, 0.0198, -0.0066, -0.0106, -0.0070, 0.0179, -0.0112], device='cuda:0'), grad: tensor([ 4.9211e-06, -2.1607e-06, 4.7050e-06, 1.1593e-05, -5.5879e-09, -2.2963e-05, -3.0398e-06, 5.2936e-06, 3.3192e-06, -1.6503e-06], device='cuda:0') 100 0.0001 changing lr epoch 134, time 251.84, cls_loss 0.0016 cls_loss_mapping 0.0056 cls_loss_causal 0.5659 re_mapping 0.0057 re_causal 0.0189 /// teacc 99.06 lr 0.00010000 Epoch 136, weight, value: tensor([[-0.1246, -0.0201, 0.1386, ..., 0.0421, -0.1215, -0.1104], [-0.0825, -0.0835, -0.0134, ..., -0.0793, -0.0949, -0.0460], [ 0.1555, -0.0417, -0.1192, ..., -0.0990, -0.1120, -0.1529], ..., [-0.0417, -0.0067, -0.1158, ..., -0.0427, 0.0572, 0.0064], [-0.1161, -0.0822, -0.0166, ..., -0.1065, -0.0097, -0.0204], [-0.0427, -0.0570, -0.0504, ..., 0.0825, 0.0299, -0.1248]], device='cuda:0'), grad: tensor([[ 4.2841e-08, 7.8231e-08, -1.0294e-04, ..., -2.2888e-05, 1.1967e-06, 4.1397e-07], [ 7.9628e-08, 7.5903e-08, 5.5600e-07, ..., 8.5216e-07, 1.0356e-06, 1.9465e-07], [ 3.8836e-07, 1.5367e-07, 1.6734e-05, ..., 8.7023e-06, 2.4419e-06, 4.8522e-07], ..., [ 7.9162e-08, -2.9989e-06, 7.3062e-07, ..., -3.2354e-06, -3.1203e-05, 1.6764e-08], [ 4.9826e-08, 2.0023e-08, 3.0566e-06, ..., 9.2667e-07, 6.2026e-07, 1.0878e-06], [ 9.1735e-08, 2.9337e-07, 1.3173e-05, ..., 4.5821e-06, 4.1351e-06, 8.5216e-08]], device='cuda:0') Epoch 136, bias, value: tensor([ 0.0033, 0.0138, 0.0177, -0.0009, 0.0194, -0.0081, -0.0111, -0.0073, 0.0181, -0.0112], device='cuda:0'), grad: tensor([-1.8477e-04, 1.8915e-06, 4.7475e-05, 9.9838e-06, 2.1178e-06, 5.7578e-05, 1.1420e-04, -9.0778e-05, 1.0617e-06, 4.1127e-05], device='cuda:0') 100 0.0001 changing lr epoch 135, time 250.76, cls_loss 0.0019 cls_loss_mapping 0.0057 cls_loss_causal 0.5243 re_mapping 0.0054 re_causal 0.0171 /// teacc 99.00 lr 0.00010000 Epoch 137, weight, value: tensor([[-0.1256, -0.0204, 0.1391, ..., 0.0423, -0.1219, -0.1105], [-0.0828, -0.0840, -0.0133, ..., -0.0798, -0.0960, -0.0462], [ 0.1566, -0.0419, -0.1199, ..., -0.0998, -0.1126, -0.1536], ..., [-0.0420, -0.0072, -0.1167, ..., -0.0438, 0.0577, 0.0064], [-0.1169, -0.0828, -0.0167, ..., -0.1071, -0.0099, -0.0203], [-0.0430, -0.0581, -0.0508, ..., 0.0827, 0.0299, -0.1252]], device='cuda:0'), grad: tensor([[ 3.2596e-09, 4.1910e-09, -2.5988e-05, ..., -5.9940e-06, 1.1036e-07, 7.0315e-08], [ 2.1886e-08, 2.1420e-08, 2.3376e-07, ..., 9.3132e-07, 4.0745e-07, 3.0268e-08], [ 2.0489e-08, 4.3772e-08, 1.4948e-06, ..., 7.2364e-07, 4.0978e-07, 7.8231e-08], ..., [ 3.3528e-08, -9.3598e-08, 2.7008e-07, ..., 3.7439e-07, -2.8946e-06, 1.6298e-08], [ 3.5390e-08, 2.9802e-08, 2.1029e-06, ..., 9.5740e-07, 5.8068e-07, 7.0315e-08], [ 2.1420e-08, 1.8161e-08, 1.7686e-06, ..., 7.4282e-06, 1.3988e-06, 1.3690e-07]], device='cuda:0') Epoch 137, bias, value: tensor([ 0.0034, 0.0138, 0.0179, -0.0030, 0.0196, -0.0061, -0.0111, -0.0074, 0.0180, -0.0112], device='cuda:0'), grad: tensor([-3.9935e-05, 1.7453e-06, 3.9265e-06, 1.3471e-05, -1.5259e-05, 1.0068e-06, 2.4110e-05, -8.2254e-06, 5.2117e-06, 1.3895e-05], device='cuda:0') 100 0.0001 changing lr epoch 136, time 246.95, cls_loss 0.0016 cls_loss_mapping 0.0045 cls_loss_causal 0.5449 re_mapping 0.0058 re_causal 0.0183 /// teacc 99.02 lr 0.00010000 Epoch 138, weight, value: tensor([[-0.1262, -0.0205, 0.1396, ..., 0.0424, -0.1224, -0.1105], [-0.0825, -0.0845, -0.0137, ..., -0.0800, -0.0970, -0.0476], [ 0.1570, -0.0423, -0.1205, ..., -0.1007, -0.1133, -0.1538], ..., [-0.0424, -0.0074, -0.1175, ..., -0.0439, 0.0586, 0.0062], [-0.1186, -0.0832, -0.0178, ..., -0.1104, -0.0110, -0.0208], [-0.0423, -0.0585, -0.0505, ..., 0.0829, 0.0298, -0.1256]], device='cuda:0'), grad: tensor([[ 1.0598e-06, 1.4901e-08, -4.5374e-06, ..., -7.3090e-06, 4.2375e-08, 2.6543e-08], [ 6.3740e-06, 2.9337e-08, 1.2154e-07, ..., 2.0117e-07, 4.8429e-08, 1.1176e-08], [-1.3515e-05, 3.0734e-08, 8.7544e-08, ..., 1.4761e-07, 7.3109e-08, 4.3772e-08], ..., [ 2.9113e-06, 3.9116e-08, 7.2177e-08, ..., 4.9407e-07, 2.2491e-07, 3.7253e-08], [ 1.3504e-06, 7.7393e-07, 1.2107e-07, ..., 2.9337e-07, 9.0431e-07, 8.1584e-07], [ 1.1781e-07, 9.7789e-08, 3.9786e-06, ..., 4.8839e-06, -5.8021e-07, 9.7323e-08]], device='cuda:0') Epoch 138, bias, value: tensor([ 0.0035, 0.0142, 0.0176, -0.0029, 0.0197, -0.0053, -0.0125, -0.0074, 0.0162, -0.0112], device='cuda:0'), grad: tensor([-9.4548e-06, 9.0376e-06, -2.1189e-05, 2.4550e-06, 2.2147e-06, -5.0031e-06, 6.3470e-07, 6.6534e-06, 6.9402e-06, 7.7263e-06], device='cuda:0') 100 0.0001 changing lr epoch 137, time 251.56, cls_loss 0.0009 cls_loss_mapping 0.0036 cls_loss_causal 0.5372 re_mapping 0.0059 re_causal 0.0193 /// teacc 98.98 lr 0.00010000 Epoch 139, weight, value: tensor([[-0.1266, -0.0205, 0.1398, ..., 0.0426, -0.1228, -0.1106], [-0.0828, -0.0849, -0.0139, ..., -0.0801, -0.0976, -0.0481], [ 0.1579, -0.0422, -0.1209, ..., -0.1012, -0.1133, -0.1542], ..., [-0.0428, -0.0075, -0.1179, ..., -0.0441, 0.0586, 0.0060], [-0.1194, -0.0833, -0.0179, ..., -0.1107, -0.0102, -0.0208], [-0.0424, -0.0587, -0.0507, ..., 0.0828, 0.0295, -0.1260]], device='cuda:0'), grad: tensor([[ 6.3255e-06, 1.8626e-09, -6.3330e-08, ..., 5.7742e-08, 8.4750e-08, 1.5832e-08], [ 6.8024e-06, 1.0245e-08, -7.1712e-08, ..., -2.0489e-07, 2.1141e-07, -9.0338e-08], [-7.1466e-05, 1.6764e-08, 1.4901e-08, ..., 8.0094e-08, 1.3132e-07, 2.0489e-08], ..., [ 9.1046e-06, 1.7695e-08, 1.2107e-08, ..., 8.8662e-07, 5.1223e-08, 1.6764e-08], [ 2.4401e-07, 1.1176e-08, 3.1665e-08, ..., 6.7893e-07, 5.8673e-07, 8.2888e-08], [ 1.9819e-06, 9.3132e-09, 2.8871e-08, ..., -4.7088e-06, -3.3043e-06, 3.9116e-08]], device='cuda:0') Epoch 139, bias, value: tensor([ 0.0036, 0.0141, 0.0179, -0.0029, 0.0198, -0.0052, -0.0126, -0.0075, 0.0168, -0.0116], device='cuda:0'), grad: tensor([ 7.5549e-06, 4.9025e-06, -7.9870e-05, 4.4823e-05, 9.1717e-06, 7.1414e-06, 9.6858e-07, 1.1832e-05, 2.9188e-06, -9.3654e-06], device='cuda:0') 100 0.0001 changing lr epoch 138, time 251.90, cls_loss 0.0016 cls_loss_mapping 0.0054 cls_loss_causal 0.5641 re_mapping 0.0057 re_causal 0.0178 /// teacc 99.11 lr 0.00010000 Epoch 140, weight, value: tensor([[-0.1279, -0.0206, 0.1403, ..., 0.0428, -0.1231, -0.1107], [-0.0830, -0.0852, -0.0144, ..., -0.0796, -0.0986, -0.0481], [ 0.1587, -0.0425, -0.1218, ..., -0.1022, -0.1141, -0.1552], ..., [-0.0433, -0.0077, -0.1186, ..., -0.0447, 0.0593, 0.0056], [-0.1199, -0.0829, -0.0178, ..., -0.1112, -0.0106, -0.0207], [-0.0425, -0.0590, -0.0510, ..., 0.0827, 0.0295, -0.1274]], device='cuda:0'), grad: tensor([[ 3.8370e-07, 2.3283e-08, -3.0827e-07, ..., 4.1351e-07, 4.6007e-07, 2.1420e-08], [ 6.9384e-07, 6.5193e-08, -1.0245e-08, ..., 1.0123e-06, 9.9093e-07, 8.3819e-09], [-1.5693e-06, 1.7323e-07, 3.7253e-08, ..., 4.7311e-07, 4.2375e-07, 7.4506e-09], ..., [ 6.4075e-07, 1.1642e-07, 3.7253e-08, ..., 7.1883e-05, 7.1228e-05, 9.3132e-09], [ 5.1502e-07, 5.6811e-08, 4.6566e-08, ..., 7.8380e-06, 7.8976e-06, 2.1979e-07], [ 4.0233e-07, 1.2014e-07, 8.0094e-08, ..., -8.7321e-05, -8.4817e-05, 6.9849e-08]], device='cuda:0') Epoch 140, bias, value: tensor([ 0.0037, 0.0142, 0.0178, -0.0031, 0.0200, -0.0052, -0.0124, -0.0075, 0.0174, -0.0120], device='cuda:0'), grad: tensor([ 2.6934e-06, 1.2508e-06, 7.7579e-07, -4.1798e-06, 1.4447e-05, 6.2361e-06, 1.0375e-06, 2.9683e-04, 4.2498e-05, -3.6168e-04], device='cuda:0') 100 0.0001 changing lr epoch 139, time 251.67, cls_loss 0.0017 cls_loss_mapping 0.0056 cls_loss_causal 0.5770 re_mapping 0.0053 re_causal 0.0176 /// teacc 98.99 lr 0.00010000 Epoch 141, weight, value: tensor([[-0.1289, -0.0206, 0.1413, ..., 0.0433, -0.1237, -0.1108], [-0.0833, -0.0856, -0.0147, ..., -0.0792, -0.0987, -0.0488], [ 0.1593, -0.0427, -0.1237, ..., -0.1035, -0.1147, -0.1565], ..., [-0.0436, -0.0079, -0.1196, ..., -0.0458, 0.0593, 0.0053], [-0.1205, -0.0836, -0.0186, ..., -0.1120, -0.0117, -0.0216], [-0.0425, -0.0595, -0.0515, ..., 0.0830, 0.0300, -0.1282]], device='cuda:0'), grad: tensor([[ 2.0228e-06, 2.7008e-08, 1.6931e-06, ..., 6.5565e-07, 1.1129e-06, 1.2731e-06], [ 9.2238e-06, 9.3132e-08, -5.0478e-07, ..., 1.4445e-06, 7.8883e-07, 6.4168e-07], [-2.5600e-05, 2.8219e-07, 2.0117e-06, ..., 5.3458e-07, 8.7824e-07, 1.0934e-06], ..., [ 1.1986e-06, 3.7067e-07, 1.1642e-07, ..., 1.0971e-06, 4.1910e-08, 1.5087e-07], [ 1.1278e-06, 1.0245e-07, 1.7481e-06, ..., 1.3448e-06, -1.0896e-06, -8.6054e-07], [ 2.7753e-07, 3.8184e-08, 2.3004e-07, ..., -4.1723e-06, 9.8161e-07, 1.3532e-06]], device='cuda:0') Epoch 141, bias, value: tensor([ 0.0041, 0.0148, 0.0176, -0.0027, 0.0196, -0.0051, -0.0128, -0.0081, 0.0169, -0.0117], device='cuda:0'), grad: tensor([ 1.1630e-05, 6.1169e-06, -2.3007e-05, -2.1178e-06, 1.1824e-05, 1.0304e-05, -1.8746e-05, 6.1914e-06, -4.9248e-06, 2.6803e-06], device='cuda:0') 100 0.0001 changing lr epoch 140, time 251.41, cls_loss 0.0022 cls_loss_mapping 0.0060 cls_loss_causal 0.5114 re_mapping 0.0058 re_causal 0.0173 /// teacc 99.02 lr 0.00010000 Epoch 142, weight, value: tensor([[-0.1297, -0.0207, 0.1425, ..., 0.0423, -0.1242, -0.1120], [-0.0834, -0.0862, -0.0143, ..., -0.0797, -0.0991, -0.0483], [ 0.1601, -0.0428, -0.1259, ..., -0.1049, -0.1151, -0.1586], ..., [-0.0438, -0.0076, -0.1217, ..., -0.0465, 0.0596, 0.0051], [-0.1212, -0.0838, -0.0197, ..., -0.1126, -0.0118, -0.0221], [-0.0427, -0.0601, -0.0537, ..., 0.0835, 0.0301, -0.1305]], device='cuda:0'), grad: tensor([[ 6.2305e-07, 1.5460e-07, 1.5832e-08, ..., 5.2154e-07, 1.5832e-07, 3.2596e-08], [ 1.3700e-06, 1.4622e-07, -2.2072e-07, ..., -1.9502e-06, 1.9744e-07, -1.3597e-07], [-6.1207e-06, 3.4925e-07, 8.5682e-08, ..., 4.0699e-07, 3.0547e-07, 9.5926e-08], ..., [ 1.0971e-06, 2.7101e-07, 9.5926e-08, ..., 8.0280e-07, -2.0023e-07, 4.4703e-08], [ 1.4845e-06, 1.4808e-07, 3.5390e-08, ..., 3.0175e-07, -1.1930e-06, -8.6240e-07], [ 1.3076e-06, 3.5297e-07, 1.3132e-07, ..., 2.2594e-06, 7.4226e-07, 6.6124e-08]], device='cuda:0') Epoch 142, bias, value: tensor([ 0.0034, 0.0154, 0.0174, -0.0032, 0.0194, -0.0049, -0.0115, -0.0085, 0.0164, -0.0116], device='cuda:0'), grad: tensor([ 5.8636e-06, -4.1068e-05, -6.7279e-06, -7.0333e-06, -8.8140e-06, 7.9051e-06, 1.3575e-05, 1.3486e-05, 1.1906e-05, 1.0818e-05], device='cuda:0') 100 0.0001 changing lr epoch 141, time 251.53, cls_loss 0.0018 cls_loss_mapping 0.0068 cls_loss_causal 0.5631 re_mapping 0.0057 re_causal 0.0179 /// teacc 99.09 lr 0.00010000 Epoch 143, weight, value: tensor([[-0.1310, -0.0208, 0.1430, ..., 0.0424, -0.1246, -0.1121], [-0.0839, -0.0865, -0.0143, ..., -0.0799, -0.0996, -0.0485], [ 0.1612, -0.0430, -0.1268, ..., -0.1056, -0.1155, -0.1593], ..., [-0.0442, -0.0077, -0.1225, ..., -0.0472, 0.0597, 0.0050], [-0.1231, -0.0840, -0.0178, ..., -0.1147, -0.0121, -0.0214], [-0.0421, -0.0603, -0.0527, ..., 0.0833, 0.0299, -0.1327]], device='cuda:0'), grad: tensor([[ 1.1735e-07, 1.6764e-08, 1.0701e-06, ..., 1.6019e-07, 3.7067e-07, 3.0641e-07], [ 2.7008e-07, 2.2352e-08, -5.5879e-09, ..., 2.1514e-07, 2.3004e-07, 3.9116e-08], [-1.3467e-06, -2.9802e-08, 5.8673e-08, ..., 1.1828e-07, 1.6578e-07, 3.9116e-08], ..., [ 1.8813e-07, 1.5832e-08, 1.2107e-08, ..., 3.3993e-07, -5.4576e-07, -3.7253e-09], [ 1.0990e-07, -4.0047e-08, 3.1013e-07, ..., 1.3113e-06, 4.4983e-07, -2.5146e-08], [ 6.7055e-08, 1.4901e-08, 5.0291e-08, ..., -6.4038e-06, -1.9185e-06, 3.7253e-08]], device='cuda:0') Epoch 143, bias, value: tensor([ 0.0035, 0.0160, 0.0176, -0.0030, 0.0202, -0.0050, -0.0123, -0.0092, 0.0168, -0.0120], device='cuda:0'), grad: tensor([ 4.2357e-06, -3.4533e-06, 2.9281e-06, 6.2548e-06, 8.6278e-06, 2.7809e-06, -1.7807e-06, -1.1392e-05, 3.5912e-06, -1.1846e-05], device='cuda:0') 100 0.0001 changing lr epoch 142, time 251.26, cls_loss 0.0013 cls_loss_mapping 0.0043 cls_loss_causal 0.5588 re_mapping 0.0056 re_causal 0.0181 /// teacc 98.99 lr 0.00010000 Epoch 144, weight, value: tensor([[-0.1316, -0.0208, 0.1434, ..., 0.0427, -0.1250, -0.1121], [-0.0842, -0.0867, -0.0141, ..., -0.0799, -0.0999, -0.0486], [ 0.1622, -0.0432, -0.1275, ..., -0.1063, -0.1162, -0.1597], ..., [-0.0443, -0.0077, -0.1232, ..., -0.0489, 0.0595, 0.0049], [-0.1253, -0.0845, -0.0178, ..., -0.1152, -0.0121, -0.0214], [-0.0421, -0.0604, -0.0533, ..., 0.0834, 0.0301, -0.1328]], device='cuda:0'), grad: tensor([[ 1.2107e-08, 0.0000e+00, -3.9861e-07, ..., -1.7788e-07, 1.4156e-07, 1.3784e-07], [ 2.8405e-07, 3.7253e-09, 5.9605e-08, ..., 4.7777e-07, 2.0210e-07, 9.4064e-08], [ 1.0990e-06, 3.7253e-09, 8.4750e-07, ..., 1.4948e-06, 7.3016e-07, 7.7859e-07], ..., [ 2.0210e-07, 1.3039e-08, 2.3283e-08, ..., 2.1234e-07, -2.0396e-07, 2.9802e-08], [ 1.9558e-08, 1.8626e-09, 1.1083e-07, ..., 8.4750e-08, 3.3304e-06, 6.3516e-06], [ 5.4948e-08, 3.7253e-09, 2.1607e-07, ..., -1.5274e-07, 1.5646e-07, 1.9278e-07]], device='cuda:0') Epoch 144, bias, value: tensor([ 0.0036, 0.0161, 0.0179, -0.0029, 0.0204, -0.0051, -0.0123, -0.0095, 0.0165, -0.0119], device='cuda:0'), grad: tensor([ 1.3132e-07, 1.5507e-06, 7.6257e-06, -7.6462e-07, -3.9265e-06, -2.3156e-05, -3.1684e-06, 3.6601e-07, 2.0042e-05, 1.2973e-06], device='cuda:0') 100 0.0001 changing lr epoch 143, time 251.76, cls_loss 0.0014 cls_loss_mapping 0.0039 cls_loss_causal 0.5304 re_mapping 0.0058 re_causal 0.0177 /// teacc 99.04 lr 0.00010000 Epoch 145, weight, value: tensor([[-0.1321, -0.0208, 0.1441, ..., 0.0430, -0.1255, -0.1121], [-0.0853, -0.0869, -0.0141, ..., -0.0808, -0.1004, -0.0488], [ 0.1646, -0.0433, -0.1286, ..., -0.1070, -0.1170, -0.1610], ..., [-0.0456, -0.0078, -0.1239, ..., -0.0488, 0.0599, 0.0050], [-0.1259, -0.0848, -0.0179, ..., -0.1157, -0.0123, -0.0214], [-0.0423, -0.0606, -0.0540, ..., 0.0835, 0.0300, -0.1332]], device='cuda:0'), grad: tensor([[ 2.0117e-07, 4.5635e-08, -1.0598e-06, ..., 5.8226e-06, 9.4343e-07, 3.9302e-07], [ 1.3886e-06, 2.7567e-07, 8.3819e-08, ..., 2.0470e-06, 1.4165e-06, 4.4610e-07], [ 9.3877e-07, 4.8801e-07, 1.8906e-07, ..., 1.2321e-06, 1.7025e-06, 6.7707e-07], ..., [ 1.9372e-06, 4.8336e-07, 5.2154e-08, ..., 7.1600e-06, 1.8450e-06, 4.6752e-07], [ 2.5891e-07, 1.0896e-07, 6.2399e-08, ..., 3.1162e-06, 3.2987e-06, 2.2911e-06], [ 1.0226e-06, 2.7567e-07, 7.7207e-07, ..., -3.8505e-05, -3.4124e-05, 3.8296e-06]], device='cuda:0') Epoch 145, bias, value: tensor([ 0.0039, 0.0158, 0.0190, -0.0029, 0.0205, -0.0056, -0.0121, -0.0098, 0.0167, -0.0121], device='cuda:0'), grad: tensor([ 1.5080e-05, 1.1854e-05, 1.3702e-05, -2.1666e-05, 3.4899e-05, -2.5451e-05, 1.0818e-05, 1.8910e-05, 1.4521e-05, -7.2479e-05], device='cuda:0') 100 0.0001 changing lr epoch 144, time 251.63, cls_loss 0.0018 cls_loss_mapping 0.0049 cls_loss_causal 0.5219 re_mapping 0.0055 re_causal 0.0168 /// teacc 99.07 lr 0.00010000 Epoch 146, weight, value: tensor([[-0.1333, -0.0209, 0.1452, ..., 0.0434, -0.1266, -0.1123], [-0.0861, -0.0874, -0.0148, ..., -0.0816, -0.1018, -0.0494], [ 0.1655, -0.0435, -0.1311, ..., -0.1093, -0.1192, -0.1637], ..., [-0.0457, -0.0081, -0.1247, ..., -0.0495, 0.0608, 0.0073], [-0.1267, -0.0851, -0.0180, ..., -0.1163, -0.0124, -0.0212], [-0.0425, -0.0609, -0.0542, ..., 0.0837, 0.0304, -0.1348]], device='cuda:0'), grad: tensor([[ 4.7497e-08, 1.5832e-08, 7.0781e-08, ..., 1.1735e-07, 1.4808e-07, 9.0338e-08], [ 2.7381e-07, 9.7789e-08, 2.7008e-08, ..., 4.3306e-07, 7.7672e-07, 6.2399e-08], [ 4.3586e-07, 1.4249e-07, 4.4703e-08, ..., 1.5274e-07, 4.9546e-07, 1.4715e-07], ..., [ 1.8813e-07, 6.1467e-08, 0.0000e+00, ..., 2.9895e-07, -1.0999e-06, 4.3772e-08], [ 2.7940e-07, 4.4703e-08, 4.3772e-08, ..., 3.5297e-07, 3.4831e-07, 3.5390e-07], [ 3.3341e-07, 9.2201e-08, 4.6566e-09, ..., 7.9349e-07, 6.1002e-07, 1.3504e-07]], device='cuda:0') Epoch 146, bias, value: tensor([ 0.0042, 0.0153, 0.0184, -0.0028, 0.0205, -0.0057, -0.0123, -0.0096, 0.0170, -0.0117], device='cuda:0'), grad: tensor([ 1.0654e-06, -6.8657e-06, 5.0291e-06, -7.5586e-06, -6.9626e-06, 2.8163e-06, 1.2759e-06, 9.4064e-08, 5.2899e-06, 5.8115e-06], device='cuda:0') 100 0.0001 changing lr epoch 145, time 251.97, cls_loss 0.0014 cls_loss_mapping 0.0052 cls_loss_causal 0.5501 re_mapping 0.0054 re_causal 0.0177 /// teacc 99.04 lr 0.00010000 Epoch 147, weight, value: tensor([[-0.1346, -0.0210, 0.1459, ..., 0.0436, -0.1281, -0.1126], [-0.0868, -0.0877, -0.0151, ..., -0.0818, -0.1026, -0.0500], [ 0.1667, -0.0438, -0.1319, ..., -0.1098, -0.1203, -0.1646], ..., [-0.0462, -0.0083, -0.1257, ..., -0.0500, 0.0614, 0.0074], [-0.1271, -0.0855, -0.0180, ..., -0.1168, -0.0126, -0.0211], [-0.0427, -0.0614, -0.0544, ..., 0.0839, 0.0304, -0.1353]], device='cuda:0'), grad: tensor([[ 1.0394e-06, 3.7253e-09, -4.8056e-07, ..., -2.3004e-07, 6.7055e-08, 7.3109e-07], [ 2.7753e-07, 2.7940e-09, 5.5879e-08, ..., 4.6566e-08, 2.0768e-07, 1.6764e-08], [-1.5246e-06, 4.6566e-09, 7.9162e-08, ..., 6.1467e-08, 1.7323e-07, 2.3283e-08], ..., [ 6.8080e-07, 1.8626e-09, 3.6322e-08, ..., 8.1025e-08, 5.2154e-08, 2.5146e-08], [ 1.3225e-07, 6.5193e-09, 7.4506e-08, ..., 1.6764e-07, -8.5309e-07, 3.7253e-08], [ 3.3434e-07, 4.6566e-09, 1.4249e-07, ..., -4.4703e-07, -3.6322e-08, 2.3283e-07]], device='cuda:0') Epoch 147, bias, value: tensor([ 0.0043, 0.0152, 0.0186, -0.0029, 0.0201, -0.0054, -0.0121, -0.0097, 0.0170, -0.0116], device='cuda:0'), grad: tensor([ 4.3884e-06, -4.3839e-05, 3.4094e-05, 6.8396e-06, 1.4575e-06, -1.0937e-05, 1.1679e-06, 4.2915e-06, -1.0431e-07, 2.6673e-06], device='cuda:0') 100 0.0001 changing lr epoch 146, time 252.05, cls_loss 0.0011 cls_loss_mapping 0.0037 cls_loss_causal 0.5173 re_mapping 0.0052 re_causal 0.0167 /// teacc 99.02 lr 0.00010000 Epoch 148, weight, value: tensor([[-0.1348, -0.0211, 0.1462, ..., 0.0438, -0.1290, -0.1128], [-0.0870, -0.0880, -0.0159, ..., -0.0823, -0.1031, -0.0504], [ 0.1670, -0.0439, -0.1325, ..., -0.1106, -0.1208, -0.1651], ..., [-0.0463, -0.0083, -0.1263, ..., -0.0502, 0.0619, 0.0074], [-0.1273, -0.0857, -0.0181, ..., -0.1176, -0.0127, -0.0212], [-0.0429, -0.0616, -0.0549, ..., 0.0837, 0.0301, -0.1356]], device='cuda:0'), grad: tensor([[ 3.5577e-07, 6.7987e-08, -5.2992e-07, ..., 8.3819e-08, 3.2689e-07, 1.9558e-07], [ 3.7160e-07, 7.0781e-08, 4.6566e-08, ..., 2.3022e-06, 1.3774e-06, -1.5926e-07], [-2.0079e-06, 1.0058e-07, 4.6566e-08, ..., 8.3819e-07, 1.1567e-06, 2.3562e-07], ..., [ 2.2445e-07, 6.7987e-08, 2.9802e-08, ..., -1.1595e-06, -2.3693e-06, 6.8918e-08], [ 1.4510e-06, 3.3062e-07, 6.8918e-08, ..., 5.1036e-07, 4.2003e-07, 8.5495e-07], [ 1.3877e-07, 6.5193e-08, 2.0303e-07, ..., 3.5197e-05, 1.2003e-05, 2.9709e-07]], device='cuda:0') Epoch 148, bias, value: tensor([ 0.0043, 0.0151, 0.0183, -0.0030, 0.0204, -0.0054, -0.0120, -0.0095, 0.0170, -0.0118], device='cuda:0'), grad: tensor([ 2.3097e-06, 4.2059e-06, 2.1346e-06, -3.9674e-06, -6.5267e-05, -8.2180e-06, 2.2277e-06, -6.7167e-06, 1.1995e-05, 6.1333e-05], device='cuda:0') 100 0.0001 changing lr epoch 147, time 251.69, cls_loss 0.0013 cls_loss_mapping 0.0050 cls_loss_causal 0.5143 re_mapping 0.0053 re_causal 0.0165 /// teacc 98.99 lr 0.00010000 Epoch 149, weight, value: tensor([[-0.1353, -0.0212, 0.1468, ..., 0.0441, -0.1299, -0.1129], [-0.0882, -0.0883, -0.0170, ..., -0.0832, -0.1039, -0.0508], [ 0.1682, -0.0445, -0.1342, ..., -0.1116, -0.1216, -0.1659], ..., [-0.0467, -0.0078, -0.1273, ..., -0.0503, 0.0625, 0.0070], [-0.1277, -0.0858, -0.0182, ..., -0.1173, -0.0123, -0.0212], [-0.0430, -0.0617, -0.0552, ..., 0.0837, 0.0297, -0.1364]], device='cuda:0'), grad: tensor([[ 8.8755e-07, 8.3819e-09, -1.0869e-06, ..., -4.7032e-07, 4.8429e-08, 3.2596e-08], [ 2.1234e-07, 2.4214e-08, 6.7987e-08, ..., 2.9616e-07, 1.3504e-07, 3.9116e-08], [-5.7146e-06, 3.7253e-08, 1.6391e-07, ..., 2.0955e-07, 8.2888e-08, 5.2154e-08], ..., [ 1.1036e-06, 3.2596e-08, 2.1607e-07, ..., 5.0850e-07, 1.6671e-07, 4.9360e-08], [ 3.1572e-07, 7.6368e-08, 1.7416e-07, ..., 2.6263e-07, 1.3225e-07, 1.2480e-07], [ 1.2107e-07, 4.1910e-08, 5.7742e-08, ..., 1.7151e-05, 7.4059e-06, 9.4995e-08]], device='cuda:0') Epoch 149, bias, value: tensor([ 0.0045, 0.0150, 0.0183, -0.0033, 0.0204, -0.0051, -0.0119, -0.0094, 0.0170, -0.0123], device='cuda:0'), grad: tensor([-6.9570e-07, -7.5437e-07, -5.4240e-06, -1.5587e-05, -3.8415e-05, 1.8120e-05, 8.8196e-07, 3.0808e-06, 1.6177e-06, 3.7223e-05], device='cuda:0') 100 0.0001 changing lr epoch 148, time 251.82, cls_loss 0.0015 cls_loss_mapping 0.0047 cls_loss_causal 0.5280 re_mapping 0.0053 re_causal 0.0171 /// teacc 99.04 lr 0.00010000 Epoch 150, weight, value: tensor([[-0.1363, -0.0212, 0.1473, ..., 0.0425, -0.1306, -0.1129], [-0.0906, -0.0885, -0.0174, ..., -0.0834, -0.1067, -0.0510], [ 0.1723, -0.0446, -0.1351, ..., -0.1114, -0.1228, -0.1674], ..., [-0.0492, -0.0078, -0.1317, ..., -0.0513, 0.0647, 0.0084], [-0.1287, -0.0859, -0.0182, ..., -0.1177, -0.0124, -0.0212], [-0.0431, -0.0618, -0.0556, ..., 0.0848, 0.0298, -0.1366]], device='cuda:0'), grad: tensor([[ 6.5938e-07, 6.5193e-09, -1.4789e-06, ..., -1.0701e-06, 1.0990e-07, 8.7544e-08], [ 2.8498e-07, 2.4214e-08, 1.8161e-07, ..., 5.5879e-09, 2.0955e-07, 2.5146e-08], [-1.0446e-05, 8.3819e-08, 1.4054e-06, ..., 9.8720e-08, 3.1479e-07, 2.9802e-08], ..., [ 1.7323e-07, 2.2352e-08, 9.8720e-08, ..., 1.1455e-07, -8.2329e-07, 6.5193e-09], [ 4.3437e-06, 2.0489e-08, -4.2133e-06, ..., 1.2759e-07, 1.2293e-07, 2.5053e-07], [ 6.6124e-08, 1.8626e-08, 3.0752e-06, ..., 3.6601e-07, 1.5367e-07, 9.4995e-08]], device='cuda:0') Epoch 150, bias, value: tensor([ 0.0034, 0.0140, 0.0205, -0.0029, 0.0201, -0.0056, -0.0118, -0.0094, 0.0170, -0.0116], device='cuda:0'), grad: tensor([ 1.3467e-06, 2.8312e-07, -5.0925e-06, 9.1195e-06, 1.3951e-06, -3.4273e-06, 3.0622e-06, -5.9884e-07, -2.4170e-05, 1.8045e-05], device='cuda:0') 100 0.0001 changing lr epoch 149, time 250.73, cls_loss 0.0015 cls_loss_mapping 0.0044 cls_loss_causal 0.5518 re_mapping 0.0053 re_causal 0.0168 /// teacc 99.05 lr 0.00010000 Epoch 151, weight, value: tensor([[-0.1374, -0.0213, 0.1481, ..., 0.0429, -0.1313, -0.1130], [-0.0910, -0.0889, -0.0175, ..., -0.0834, -0.1071, -0.0511], [ 0.1728, -0.0451, -0.1341, ..., -0.1127, -0.1245, -0.1688], ..., [-0.0490, -0.0088, -0.1327, ..., -0.0516, 0.0653, 0.0104], [-0.1297, -0.0862, -0.0183, ..., -0.1182, -0.0127, -0.0213], [-0.0434, -0.0620, -0.0567, ..., 0.0847, 0.0295, -0.1369]], device='cuda:0'), grad: tensor([[ 5.7742e-08, 3.2596e-08, -1.3504e-06, ..., -2.2911e-07, 2.3656e-07, 5.4017e-08], [ 1.8626e-07, 9.4995e-08, -5.5879e-09, ..., 6.3702e-07, 7.9814e-07, 4.3772e-08], [-1.6335e-06, 1.8720e-07, 2.0768e-07, ..., 5.6624e-07, 6.8173e-07, 7.7300e-08], ..., [ 1.0561e-06, 1.1362e-07, 8.2888e-08, ..., 6.5491e-06, 2.9542e-06, 5.1223e-08], [ 7.3295e-07, 1.1735e-07, 2.0768e-07, ..., 1.4743e-06, 1.0356e-06, 8.9407e-08], [ 7.8231e-08, 1.4249e-07, 2.2724e-07, ..., 2.6450e-06, 3.0752e-06, 1.0151e-07]], device='cuda:0') Epoch 151, bias, value: tensor([ 0.0037, 0.0139, 0.0199, -0.0024, 0.0203, -0.0057, -0.0120, -0.0089, 0.0168, -0.0119], device='cuda:0'), grad: tensor([-2.7288e-07, 3.9749e-06, -1.8254e-06, -2.8647e-06, -5.5760e-05, 1.4883e-06, 6.1318e-06, 2.3782e-05, 7.5065e-06, 1.7881e-05], device='cuda:0') 100 0.0001 changing lr epoch 150, time 249.58, cls_loss 0.0017 cls_loss_mapping 0.0052 cls_loss_causal 0.5293 re_mapping 0.0054 re_causal 0.0166 /// teacc 99.00 lr 0.00010000 Epoch 152, weight, value: tensor([[-0.1380, -0.0213, 0.1493, ..., 0.0414, -0.1320, -0.1133], [-0.0913, -0.0893, -0.0195, ..., -0.0842, -0.1080, -0.0515], [ 0.1737, -0.0453, -0.1352, ..., -0.1137, -0.1251, -0.1694], ..., [-0.0496, -0.0088, -0.1336, ..., -0.0515, 0.0665, 0.0114], [-0.1306, -0.0864, -0.0184, ..., -0.1188, -0.0127, -0.0213], [-0.0434, -0.0622, -0.0572, ..., 0.0864, 0.0296, -0.1387]], device='cuda:0'), grad: tensor([[ 3.5483e-07, 1.8626e-09, -5.5581e-06, ..., -5.6028e-06, 4.0196e-06, 2.7716e-06], [ 2.3283e-06, 2.7008e-08, 8.7637e-07, ..., 6.1188e-07, 1.1418e-06, 9.4436e-07], [ 3.3658e-06, 4.1910e-08, 1.4361e-06, ..., 7.7579e-07, 1.7658e-06, 9.7323e-07], ..., [ 3.3043e-06, 5.0291e-08, 5.4296e-07, ..., 5.4054e-06, 2.1141e-06, 5.5041e-07], [ 2.4624e-06, 5.3085e-08, 3.0138e-06, ..., 7.8976e-07, 3.5614e-06, 4.2804e-06], [ 1.9036e-06, 1.0245e-08, 5.3272e-06, ..., -1.1839e-05, -2.7325e-06, 1.0962e-06]], device='cuda:0') Epoch 152, bias, value: tensor([ 0.0027, 0.0137, 0.0199, -0.0022, 0.0194, -0.0061, -0.0117, -0.0085, 0.0167, -0.0108], device='cuda:0'), grad: tensor([-3.7104e-06, 1.3433e-05, 3.1888e-05, -8.8215e-05, 2.1502e-05, 2.2066e-04, -2.4056e-04, 3.1024e-05, 2.7001e-05, -1.3053e-05], device='cuda:0') 100 0.0001 changing lr epoch 151, time 248.50, cls_loss 0.0012 cls_loss_mapping 0.0048 cls_loss_causal 0.5426 re_mapping 0.0050 re_causal 0.0165 /// teacc 99.00 lr 0.00010000 Epoch 153, weight, value: tensor([[-0.1389, -0.0214, 0.1515, ..., 0.0418, -0.1327, -0.1134], [-0.0920, -0.0895, -0.0189, ..., -0.0854, -0.1085, -0.0518], [ 0.1749, -0.0454, -0.1359, ..., -0.1149, -0.1255, -0.1694], ..., [-0.0498, -0.0090, -0.1346, ..., -0.0518, 0.0668, 0.0114], [-0.1314, -0.0865, -0.0187, ..., -0.1202, -0.0127, -0.0213], [-0.0440, -0.0624, -0.0608, ..., 0.0866, 0.0296, -0.1393]], device='cuda:0'), grad: tensor([[ 2.7940e-08, 1.2107e-08, 1.8189e-06, ..., -2.2817e-07, 9.3412e-07, 1.9027e-06], [ 4.8429e-08, 1.4901e-08, 4.2561e-07, ..., 1.1642e-07, 3.9302e-07, 2.2817e-07], [-1.2200e-07, 1.7695e-08, 4.4610e-07, ..., 1.6578e-07, 4.0140e-07, 2.0768e-07], ..., [ 6.4261e-08, 1.5832e-08, 6.3330e-08, ..., 8.1956e-08, 1.0157e-04, 1.9558e-08], [ 1.9558e-08, 6.5193e-09, 4.0326e-07, ..., 9.6858e-08, -1.0389e-04, 1.8254e-07], [ 5.1223e-08, 4.3772e-08, 5.8673e-07, ..., 4.2189e-07, 1.4808e-07, 7.9162e-08]], device='cuda:0') Epoch 153, bias, value: tensor([ 0.0035, 0.0136, 0.0203, -0.0022, 0.0198, -0.0062, -0.0121, -0.0085, 0.0165, -0.0110], device='cuda:0'), grad: tensor([ 7.6219e-06, 2.4810e-06, 2.1327e-06, 4.5896e-06, 8.1807e-06, 3.0808e-06, -2.2426e-05, 3.5906e-04, -3.6645e-04, 1.7621e-06], device='cuda:0') 100 0.0001 changing lr epoch 152, time 250.22, cls_loss 0.0014 cls_loss_mapping 0.0039 cls_loss_causal 0.5152 re_mapping 0.0053 re_causal 0.0163 /// teacc 99.09 lr 0.00010000 Epoch 154, weight, value: tensor([[-0.1395, -0.0215, 0.1519, ..., 0.0420, -0.1344, -0.1138], [-0.0921, -0.0899, -0.0196, ..., -0.0839, -0.1072, -0.0520], [ 0.1754, -0.0456, -0.1367, ..., -0.1162, -0.1264, -0.1697], ..., [-0.0500, -0.0091, -0.1355, ..., -0.0534, 0.0677, 0.0113], [-0.1319, -0.0868, -0.0188, ..., -0.1207, -0.0139, -0.0214], [-0.0442, -0.0633, -0.0610, ..., 0.0867, 0.0294, -0.1398]], device='cuda:0'), grad: tensor([[ 7.1712e-08, 3.6322e-08, -2.0768e-06, ..., -1.7080e-06, 3.7998e-07, 3.0734e-07], [ 2.2538e-07, 1.7323e-07, 1.0803e-07, ..., 6.8918e-08, 7.3295e-07, 1.2293e-07], [-1.8068e-06, 9.1270e-08, 2.0396e-07, ..., 1.2852e-07, 4.1910e-07, 8.9407e-08], ..., [ 1.1465e-06, 2.0955e-07, 1.0524e-07, ..., 1.5181e-07, -2.2668e-06, 1.1269e-07], [ 1.7695e-07, 1.1455e-07, 5.4948e-08, ..., 1.2387e-07, -9.4716e-07, -5.7649e-07], [ 1.0803e-07, 1.8254e-07, 9.7696e-07, ..., 5.9977e-07, 9.4902e-07, 1.3039e-07]], device='cuda:0') Epoch 154, bias, value: tensor([ 0.0033, 0.0152, 0.0198, -0.0023, 0.0198, -0.0065, -0.0118, -0.0089, 0.0157, -0.0111], device='cuda:0'), grad: tensor([-1.1064e-06, 3.7495e-06, -9.2015e-07, -6.9179e-06, 9.7230e-07, 8.0168e-06, 2.6263e-06, -3.6675e-06, -8.8736e-06, 6.0908e-06], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 153---------------------------------------------------- epoch 153, time 268.49, cls_loss 0.0012 cls_loss_mapping 0.0036 cls_loss_causal 0.5272 re_mapping 0.0049 re_causal 0.0159 /// teacc 99.14 lr 0.00010000 Epoch 155, weight, value: tensor([[-0.1400, -0.0222, 0.1516, ..., 0.0422, -0.1356, -0.1146], [-0.0924, -0.0906, -0.0196, ..., -0.0841, -0.1077, -0.0524], [ 0.1759, -0.0458, -0.1373, ..., -0.1168, -0.1271, -0.1702], ..., [-0.0502, -0.0094, -0.1359, ..., -0.0538, 0.0681, 0.0114], [-0.1324, -0.0871, -0.0188, ..., -0.1211, -0.0140, -0.0214], [-0.0444, -0.0636, -0.0618, ..., 0.0870, 0.0298, -0.1403]], device='cuda:0'), grad: tensor([[ 2.0117e-07, 9.3132e-10, 3.6974e-07, ..., 2.9802e-08, 7.5437e-08, 1.1828e-07], [ 7.4040e-07, 2.5146e-08, 2.0992e-06, ..., 2.6450e-07, 6.1374e-07, 5.6997e-07], [-2.0456e-04, 1.6764e-08, 3.7160e-07, ..., 1.0617e-07, 2.8964e-07, 1.0803e-07], ..., [ 1.1241e-06, -6.5193e-08, 5.5879e-09, ..., 8.1584e-07, -5.1409e-07, 5.5879e-09], [ 3.2224e-07, 1.0245e-08, 1.0710e-07, ..., 8.3819e-08, 2.1327e-07, 9.9652e-08], [ 1.9479e-04, 1.6764e-08, 2.1420e-08, ..., -1.0990e-07, 4.1351e-07, 3.3528e-08]], device='cuda:0') Epoch 155, bias, value: tensor([ 0.0030, 0.0153, 0.0196, -0.0024, 0.0194, -0.0067, -0.0114, -0.0090, 0.0160, -0.0108], device='cuda:0'), grad: tensor([ 1.3076e-06, 5.5432e-06, -3.6979e-04, 5.1856e-06, -4.7125e-06, 6.9290e-06, -5.3868e-06, 3.0883e-06, 1.7853e-06, 3.5620e-04], device='cuda:0') 100 0.0001 changing lr epoch 154, time 251.83, cls_loss 0.0014 cls_loss_mapping 0.0043 cls_loss_causal 0.5012 re_mapping 0.0051 re_causal 0.0159 /// teacc 99.06 lr 0.00010000 Epoch 156, weight, value: tensor([[-0.1406, -0.0222, 0.1527, ..., 0.0415, -0.1360, -0.1147], [-0.0930, -0.0914, -0.0198, ..., -0.0845, -0.1087, -0.0531], [ 0.1772, -0.0461, -0.1380, ..., -0.1176, -0.1278, -0.1697], ..., [-0.0504, -0.0094, -0.1366, ..., -0.0549, 0.0685, 0.0116], [-0.1327, -0.0874, -0.0189, ..., -0.1216, -0.0141, -0.0216], [-0.0460, -0.0642, -0.0629, ..., 0.0884, 0.0312, -0.1416]], device='cuda:0'), grad: tensor([[ 5.4855e-07, 1.7881e-07, -1.8273e-06, ..., -1.0794e-06, 1.4622e-07, 7.2643e-08], [ 6.2346e-05, 3.0547e-07, 4.2841e-08, ..., 3.7253e-09, 3.8091e-07, 6.4261e-08], [-6.5625e-05, 1.7695e-07, 1.1269e-07, ..., 5.8673e-08, 1.6391e-07, 6.7055e-08], ..., [ 7.5996e-07, 5.4482e-07, 2.0489e-08, ..., 2.0489e-08, -1.0729e-06, -1.7043e-07], [ 8.1398e-07, 2.8312e-07, 2.0768e-07, ..., 1.0058e-07, 1.9278e-07, 9.8720e-08], [ 5.6531e-07, 1.5348e-06, 1.0701e-06, ..., 6.4448e-07, 1.4985e-06, 2.9057e-07]], device='cuda:0') Epoch 156, bias, value: tensor([ 0.0028, 0.0148, 0.0199, -0.0025, 0.0184, -0.0065, -0.0115, -0.0090, 0.0161, -0.0095], device='cuda:0'), grad: tensor([-4.4610e-07, 1.9944e-04, -2.0671e-04, -9.6112e-06, 1.1437e-06, 5.0105e-07, 8.1677e-07, 5.9605e-08, 3.7607e-06, 1.0811e-05], device='cuda:0') 100 0.0001 changing lr epoch 155, time 251.77, cls_loss 0.0019 cls_loss_mapping 0.0051 cls_loss_causal 0.5568 re_mapping 0.0052 re_causal 0.0171 /// teacc 99.03 lr 0.00010000 Epoch 157, weight, value: tensor([[-0.1415, -0.0223, 0.1543, ..., 0.0415, -0.1367, -0.1148], [-0.0940, -0.0924, -0.0239, ..., -0.0859, -0.1105, -0.0557], [ 0.1788, -0.0479, -0.1394, ..., -0.1186, -0.1297, -0.1704], ..., [-0.0514, -0.0078, -0.1372, ..., -0.0556, 0.0693, 0.0125], [-0.1334, -0.0883, -0.0189, ..., -0.1225, -0.0145, -0.0216], [-0.0462, -0.0634, -0.0635, ..., 0.0894, 0.0320, -0.1413]], device='cuda:0'), grad: tensor([[ 2.4587e-07, 2.9895e-07, -1.7509e-07, ..., 1.6121e-06, 1.1567e-06, 4.7963e-07], [ 6.1747e-07, 6.0722e-07, 7.4506e-08, ..., 5.6960e-06, 2.2389e-06, 5.9512e-07], [ 4.4238e-07, 3.1590e-06, 6.7055e-08, ..., 5.5805e-06, 1.3456e-05, 2.5164e-06], ..., [ 8.7544e-07, -1.5318e-05, 2.5146e-08, ..., 2.8118e-05, -7.4744e-05, -1.2450e-05], [ 1.3877e-07, 1.5739e-07, 1.5460e-07, ..., 2.6077e-06, 1.0189e-06, 3.7625e-07], [ 5.3830e-07, 7.7672e-07, 4.8708e-07, ..., 1.1787e-03, 3.4475e-04, 9.5189e-05]], device='cuda:0') Epoch 157, bias, value: tensor([ 0.0032, 0.0136, 0.0202, -0.0025, 0.0177, -0.0066, -0.0111, -0.0088, 0.0160, -0.0088], device='cuda:0'), grad: tensor([ 6.8583e-06, 1.4544e-05, 4.4048e-05, -7.9572e-06, -2.0237e-03, 1.6928e-05, 4.6715e-06, -1.1408e-04, 6.5267e-06, 2.0504e-03], device='cuda:0') 100 0.0001 changing lr epoch 156, time 251.55, cls_loss 0.0014 cls_loss_mapping 0.0048 cls_loss_causal 0.5464 re_mapping 0.0049 re_causal 0.0161 /// teacc 98.98 lr 0.00010000 Epoch 158, weight, value: tensor([[-0.1425, -0.0223, 0.1562, ..., 0.0418, -0.1373, -0.1149], [-0.0946, -0.0937, -0.0249, ..., -0.0866, -0.1110, -0.0559], [ 0.1798, -0.0481, -0.1416, ..., -0.1202, -0.1302, -0.1698], ..., [-0.0520, -0.0086, -0.1381, ..., -0.0560, 0.0697, 0.0112], [-0.1340, -0.0873, -0.0193, ..., -0.1235, -0.0154, -0.0222], [-0.0465, -0.0656, -0.0643, ..., 0.0889, 0.0314, -0.1424]], device='cuda:0'), grad: tensor([[ 2.5984e-07, 3.9209e-07, -5.5041e-07, ..., -3.6694e-07, 6.5565e-07, 4.9174e-07], [ 4.4610e-07, 1.6298e-07, 1.3225e-07, ..., 9.2201e-08, 2.8312e-07, 2.7753e-07], [-1.7798e-06, 2.2352e-07, 1.8440e-07, ..., 3.4459e-08, 2.2538e-07, 2.8219e-07], ..., [ 3.2783e-07, 8.1025e-08, 4.9360e-08, ..., 1.0245e-07, 3.6322e-08, 4.1910e-08], [ 3.4086e-07, 1.3448e-06, 6.0536e-07, ..., 9.2201e-08, 2.3413e-06, 1.9372e-06], [ 1.7509e-07, 2.0117e-07, 4.5542e-07, ..., -3.0920e-07, 4.5169e-07, 3.5111e-07]], device='cuda:0') Epoch 158, bias, value: tensor([ 0.0039, 0.0130, 0.0198, -0.0022, 0.0185, -0.0080, -0.0101, -0.0087, 0.0160, -0.0094], device='cuda:0'), grad: tensor([ 2.1551e-06, 1.9744e-06, -1.2610e-06, 4.6007e-06, 2.0005e-06, -2.5332e-05, 5.8487e-07, 1.2433e-06, 1.0364e-05, 3.6377e-06], device='cuda:0') 100 0.0001 changing lr epoch 157, time 251.73, cls_loss 0.0028 cls_loss_mapping 0.0061 cls_loss_causal 0.5370 re_mapping 0.0051 re_causal 0.0165 /// teacc 99.09 lr 0.00010000 Epoch 159, weight, value: tensor([[-0.1433, -0.0225, 0.1566, ..., 0.0420, -0.1379, -0.1151], [-0.0946, -0.0946, -0.0250, ..., -0.0876, -0.1119, -0.0561], [ 0.1801, -0.0482, -0.1424, ..., -0.1212, -0.1308, -0.1703], ..., [-0.0521, -0.0086, -0.1390, ..., -0.0563, 0.0704, 0.0113], [-0.1346, -0.0875, -0.0194, ..., -0.1238, -0.0156, -0.0224], [-0.0467, -0.0664, -0.0648, ..., 0.0881, 0.0305, -0.1430]], device='cuda:0'), grad: tensor([[ 4.9826e-08, -1.6950e-07, -9.5088e-07, ..., -2.9104e-07, 8.8941e-08, 1.4761e-07], [ 8.7079e-08, 2.7474e-08, 3.9116e-08, ..., 9.4995e-08, 5.5879e-08, -2.1476e-06], [-4.5681e-07, 4.2841e-08, 1.7323e-07, ..., 6.9849e-08, 6.5193e-08, 9.2201e-08], ..., [ 2.1234e-07, 3.3528e-08, 1.1316e-07, ..., 2.9337e-07, 7.0781e-08, 9.2853e-07], [ 6.0070e-08, 1.2573e-08, 2.0443e-07, ..., 4.8894e-08, 2.5611e-08, 8.6147e-08], [ 7.5717e-07, 6.9803e-07, 1.5507e-07, ..., -5.5879e-07, -2.8685e-07, 1.4249e-07]], device='cuda:0') Epoch 159, bias, value: tensor([ 0.0039, 0.0119, 0.0191, -0.0021, 0.0196, -0.0085, -0.0097, -0.0074, 0.0160, -0.0104], device='cuda:0'), grad: tensor([-6.9663e-07, -2.8655e-05, 8.8196e-07, -2.3302e-06, 4.9407e-07, 1.4514e-05, -1.8729e-06, 1.3649e-05, 1.0505e-06, 2.9169e-06], device='cuda:0') 100 0.0001 changing lr epoch 158, time 251.51, cls_loss 0.0012 cls_loss_mapping 0.0036 cls_loss_causal 0.5242 re_mapping 0.0052 re_causal 0.0166 /// teacc 98.95 lr 0.00010000 Epoch 160, weight, value: tensor([[-0.1442, -0.0230, 0.1567, ..., 0.0420, -0.1384, -0.1153], [-0.0949, -0.0953, -0.0248, ..., -0.0879, -0.1124, -0.0560], [ 0.1818, -0.0484, -0.1429, ..., -0.1216, -0.1311, -0.1713], ..., [-0.0525, -0.0087, -0.1403, ..., -0.0568, 0.0707, 0.0112], [-0.1372, -0.0855, -0.0196, ..., -0.1242, -0.0150, -0.0221], [-0.0482, -0.0695, -0.0652, ..., 0.0880, 0.0297, -0.1458]], device='cuda:0'), grad: tensor([[ 2.6403e-07, 4.6566e-10, -3.5577e-07, ..., 6.4354e-07, 2.8266e-07, 9.2201e-08], [ 1.9558e-08, 1.2107e-08, 8.8476e-09, ..., 1.1846e-06, 6.2073e-07, 3.2131e-08], [ 3.6741e-07, 1.8626e-09, 8.5682e-08, ..., 1.9539e-06, 6.5379e-07, 4.2841e-08], ..., [ 1.1688e-07, 2.3982e-07, 2.0489e-08, ..., 2.6897e-06, 2.0042e-06, 1.0896e-07], [ 4.1910e-08, 9.3132e-10, 5.9139e-08, ..., 5.5740e-07, 3.1013e-07, 1.5460e-07], [-2.1011e-06, 4.1910e-09, 2.4820e-07, ..., -1.2200e-06, 1.0384e-06, 8.9407e-08]], device='cuda:0') Epoch 160, bias, value: tensor([ 0.0038, 0.0117, 0.0195, -0.0025, 0.0198, -0.0078, -0.0096, -0.0071, 0.0163, -0.0114], device='cuda:0'), grad: tensor([ 2.8461e-06, 4.1872e-06, 9.0525e-06, -9.9018e-06, -2.8268e-05, -4.5588e-07, 6.6981e-06, 1.9476e-05, -6.5984e-07, -2.9560e-06], device='cuda:0') 100 0.0001 changing lr epoch 159, time 251.70, cls_loss 0.0010 cls_loss_mapping 0.0034 cls_loss_causal 0.5220 re_mapping 0.0053 re_causal 0.0180 /// teacc 99.04 lr 0.00010000 Epoch 161, weight, value: tensor([[-0.1446, -0.0232, 0.1571, ..., 0.0421, -0.1388, -0.1154], [-0.0953, -0.0955, -0.0247, ..., -0.0895, -0.1127, -0.0561], [ 0.1822, -0.0485, -0.1434, ..., -0.1226, -0.1315, -0.1724], ..., [-0.0525, -0.0089, -0.1413, ..., -0.0571, 0.0715, 0.0115], [-0.1374, -0.0856, -0.0196, ..., -0.1244, -0.0151, -0.0221], [-0.0489, -0.0696, -0.0659, ..., 0.0878, 0.0294, -0.1462]], device='cuda:0'), grad: tensor([[ 1.0245e-08, 2.7940e-09, -7.2177e-08, ..., 6.9384e-08, 1.6391e-07, 2.6543e-08], [ 3.5856e-08, 9.3132e-09, 2.3283e-09, ..., 3.1199e-08, 3.4086e-07, 4.4703e-08], [ 9.3132e-08, 1.8161e-08, 1.2573e-08, ..., 1.9558e-08, 3.8650e-08, 2.0489e-08], ..., [ 4.1444e-08, 1.2107e-08, 4.1910e-09, ..., -2.3097e-07, -7.2364e-07, -8.7544e-08], [ 1.0710e-08, 2.7940e-09, 1.2107e-08, ..., 8.8476e-09, -4.5635e-08, -1.3644e-07], [ 1.6624e-07, 6.1467e-08, 4.7032e-08, ..., 1.1129e-07, 1.6205e-07, 1.1548e-07]], device='cuda:0') Epoch 161, bias, value: tensor([ 0.0039, 0.0116, 0.0192, -0.0028, 0.0210, -0.0074, -0.0097, -0.0071, 0.0163, -0.0120], device='cuda:0'), grad: tensor([ 1.3337e-06, 8.1956e-07, 6.6450e-07, -9.7696e-07, -2.4680e-08, 7.9768e-07, 1.3364e-07, -3.6284e-06, -2.7530e-06, 3.6340e-06], device='cuda:0') 100 0.0001 changing lr epoch 160, time 251.77, cls_loss 0.0012 cls_loss_mapping 0.0033 cls_loss_causal 0.5139 re_mapping 0.0049 re_causal 0.0161 /// teacc 98.96 lr 0.00010000 Epoch 162, weight, value: tensor([[-0.1451, -0.0234, 0.1577, ..., 0.0423, -0.1399, -0.1156], [-0.0955, -0.0958, -0.0248, ..., -0.0898, -0.1132, -0.0560], [ 0.1829, -0.0486, -0.1442, ..., -0.1237, -0.1319, -0.1729], ..., [-0.0531, -0.0090, -0.1419, ..., -0.0573, 0.0721, 0.0110], [-0.1377, -0.0858, -0.0196, ..., -0.1248, -0.0150, -0.0221], [-0.0492, -0.0698, -0.0664, ..., 0.0877, 0.0292, -0.1464]], device='cuda:0'), grad: tensor([[ 2.4447e-07, 7.8697e-08, -1.5832e-08, ..., 5.3551e-08, 1.7881e-07, 1.4016e-07], [ 6.3749e-07, 7.7300e-08, 2.2817e-08, ..., 3.7253e-07, 2.7381e-07, 3.8184e-08], [-1.3262e-05, 4.2235e-07, 2.7940e-08, ..., 2.3050e-07, 3.4133e-07, 2.7940e-08], ..., [ 1.3299e-05, 1.7136e-07, 4.0513e-08, ..., 5.1968e-07, -3.1013e-07, 2.7008e-08], [ 1.7928e-07, 2.6450e-07, 2.0722e-07, ..., 4.5449e-07, -8.9873e-08, -1.0803e-07], [ 1.2806e-07, 9.4995e-08, 1.6950e-07, ..., 3.0752e-06, 9.8627e-07, 1.6997e-07]], device='cuda:0') Epoch 162, bias, value: tensor([ 0.0039, 0.0114, 0.0192, -0.0025, 0.0214, -0.0075, -0.0100, -0.0070, 0.0164, -0.0123], device='cuda:0'), grad: tensor([ 1.5171e-06, 1.9465e-06, -1.9729e-05, 1.5870e-06, -1.2122e-05, -7.1190e-06, 2.1905e-06, 2.4691e-05, -1.7462e-06, 8.7172e-06], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 161---------------------------------------------------- epoch 161, time 268.22, cls_loss 0.0013 cls_loss_mapping 0.0040 cls_loss_causal 0.5487 re_mapping 0.0050 re_causal 0.0160 /// teacc 99.19 lr 0.00010000 Epoch 163, weight, value: tensor([[-0.1457, -0.0235, 0.1583, ..., 0.0426, -0.1406, -0.1158], [-0.0958, -0.0962, -0.0252, ..., -0.0902, -0.1135, -0.0561], [ 0.1843, -0.0488, -0.1451, ..., -0.1246, -0.1325, -0.1740], ..., [-0.0536, -0.0091, -0.1438, ..., -0.0580, 0.0723, 0.0119], [-0.1381, -0.0860, -0.0197, ..., -0.1253, -0.0151, -0.0221], [-0.0514, -0.0699, -0.0675, ..., 0.0872, 0.0291, -0.1465]], device='cuda:0'), grad: tensor([[ 1.6298e-08, 7.4506e-09, -1.3504e-06, ..., -5.8534e-07, 1.2387e-07, 1.1502e-07], [ 3.2596e-08, 1.3970e-08, 4.6100e-08, ..., -2.4680e-08, 2.9337e-07, -1.0710e-08], [ 1.0245e-08, 2.9802e-08, 1.1967e-07, ..., 6.3796e-08, 1.1316e-07, 6.5193e-08], ..., [ 7.9628e-08, 2.7008e-08, 4.9826e-08, ..., 1.6158e-07, -1.1362e-06, 2.7008e-08], [ 7.8231e-08, 7.2177e-08, 2.8173e-07, ..., 2.1420e-07, 3.0827e-07, 3.7951e-07], [ 3.0734e-08, 2.8405e-08, 9.7882e-07, ..., -1.3970e-09, 4.7684e-07, 3.4459e-08]], device='cuda:0') Epoch 163, bias, value: tensor([ 0.0041, 0.0115, 0.0192, -0.0036, 0.0219, -0.0064, -0.0100, -0.0070, 0.0165, -0.0131], device='cuda:0'), grad: tensor([-1.4044e-06, -4.7032e-07, 9.2620e-07, -2.2743e-06, 6.0489e-07, 2.6207e-06, -3.6042e-06, -3.1181e-06, 3.1590e-06, 3.5521e-06], device='cuda:0') 100 0.0001 changing lr epoch 162, time 251.81, cls_loss 0.0010 cls_loss_mapping 0.0040 cls_loss_causal 0.5267 re_mapping 0.0051 re_causal 0.0164 /// teacc 99.15 lr 0.00010000 Epoch 164, weight, value: tensor([[-0.1466, -0.0237, 0.1583, ..., 0.0426, -0.1411, -0.1162], [-0.0962, -0.0966, -0.0247, ..., -0.0898, -0.1137, -0.0558], [ 0.1848, -0.0487, -0.1456, ..., -0.1250, -0.1328, -0.1744], ..., [-0.0539, -0.0094, -0.1450, ..., -0.0583, 0.0726, 0.0120], [-0.1384, -0.0861, -0.0197, ..., -0.1261, -0.0152, -0.0221], [-0.0515, -0.0700, -0.0680, ..., 0.0873, 0.0292, -0.1468]], device='cuda:0'), grad: tensor([[ 9.3132e-08, 1.3504e-08, -1.0291e-07, ..., -8.3353e-08, 1.5926e-07, 7.8697e-08], [ 2.0443e-07, 6.2399e-08, -4.2375e-08, ..., 3.4459e-07, 1.3039e-07, 1.7695e-08], [-7.0082e-07, 8.3819e-08, 1.8673e-07, ..., 3.4785e-07, 1.4482e-07, 2.6543e-08], ..., [ 2.7474e-07, 6.6124e-08, 1.4482e-07, ..., 3.5623e-07, -3.4226e-07, 1.5832e-08], [ 4.9593e-07, 2.9802e-08, 1.5739e-07, ..., 7.9442e-07, 2.0722e-07, 1.3737e-07], [-3.6452e-06, 2.3749e-08, 6.3330e-08, ..., -1.4976e-05, -8.8941e-08, 2.1746e-07]], device='cuda:0') Epoch 164, bias, value: tensor([ 0.0039, 0.0117, 0.0192, -0.0034, 0.0218, -0.0069, -0.0096, -0.0072, 0.0164, -0.0132], device='cuda:0'), grad: tensor([ 7.6136e-07, 1.7118e-06, 3.0696e-06, 8.0233e-07, 1.6838e-05, 1.9884e-04, 9.6112e-07, 1.6000e-06, 7.1563e-06, -2.3139e-04], device='cuda:0') 100 0.0001 changing lr epoch 163, time 251.36, cls_loss 0.0011 cls_loss_mapping 0.0039 cls_loss_causal 0.5292 re_mapping 0.0052 re_causal 0.0166 /// teacc 99.06 lr 0.00010000 Epoch 165, weight, value: tensor([[-0.1473, -0.0240, 0.1588, ..., 0.0428, -0.1420, -0.1163], [-0.0963, -0.0970, -0.0247, ..., -0.0899, -0.1141, -0.0558], [ 0.1853, -0.0488, -0.1462, ..., -0.1256, -0.1332, -0.1747], ..., [-0.0542, -0.0096, -0.1463, ..., -0.0588, 0.0729, 0.0120], [-0.1386, -0.0863, -0.0210, ..., -0.1265, -0.0160, -0.0224], [-0.0517, -0.0700, -0.0692, ..., 0.0871, 0.0291, -0.1468]], device='cuda:0'), grad: tensor([[ 1.5879e-07, 5.4948e-08, 3.5949e-06, ..., -1.3039e-08, 9.9372e-07, 6.2538e-07], [ 9.8255e-08, 2.9802e-08, 3.7439e-07, ..., 4.0513e-08, 2.2305e-07, 2.1746e-07], [ 1.8347e-07, 1.6764e-07, 6.9756e-07, ..., 3.9581e-08, 2.1374e-07, 1.8813e-07], ..., [ 6.7521e-07, 2.4447e-07, 9.3132e-08, ..., 7.4040e-08, -4.9779e-07, -2.0955e-08], [ 1.5274e-07, 1.9092e-08, 5.9716e-06, ..., 8.4750e-08, 1.6624e-06, 4.4703e-07], [ 3.8650e-08, 1.3039e-08, 1.1595e-07, ..., -2.5192e-07, -6.9849e-09, 3.4459e-08]], device='cuda:0') Epoch 165, bias, value: tensor([ 0.0040, 0.0113, 0.0189, -0.0034, 0.0221, -0.0071, -0.0084, -0.0067, 0.0153, -0.0135], device='cuda:0'), grad: tensor([ 7.4431e-06, 3.2812e-05, 3.2969e-06, 7.3481e-07, 3.2950e-06, -6.2168e-05, -1.9833e-05, 3.2242e-06, 2.4870e-05, 6.1877e-06], device='cuda:0') 100 0.0001 changing lr epoch 164, time 251.27, cls_loss 0.0015 cls_loss_mapping 0.0047 cls_loss_causal 0.5421 re_mapping 0.0049 re_causal 0.0156 /// teacc 99.15 lr 0.00010000 Epoch 166, weight, value: tensor([[-0.1495, -0.0241, 0.1606, ..., 0.0410, -0.1430, -0.1165], [-0.0963, -0.0976, -0.0249, ..., -0.0896, -0.1145, -0.0560], [ 0.1865, -0.0476, -0.1472, ..., -0.1263, -0.1337, -0.1749], ..., [-0.0545, -0.0100, -0.1480, ..., -0.0595, 0.0736, 0.0120], [-0.1390, -0.0866, -0.0212, ..., -0.1272, -0.0163, -0.0226], [-0.0522, -0.0702, -0.0722, ..., 0.0886, 0.0287, -0.1468]], device='cuda:0'), grad: tensor([[ 4.2841e-08, 2.7940e-09, -9.4250e-07, ..., -5.2806e-07, 4.2375e-08, 1.2573e-08], [ 1.1129e-07, 2.7940e-09, -7.4040e-08, ..., 1.0850e-07, 4.2841e-08, 4.1910e-09], [-7.7719e-07, 1.3970e-09, 2.9802e-08, ..., 8.0559e-08, 4.2375e-08, 5.1223e-09], ..., [ 1.2573e-07, 5.5879e-09, 7.3109e-08, ..., 2.0536e-07, -4.0978e-08, 9.3132e-10], [ 8.6613e-08, 2.7940e-09, 6.2864e-08, ..., 6.6590e-08, -7.2177e-08, -4.6566e-10], [ 2.6077e-08, 9.3132e-10, 6.7614e-07, ..., -1.5460e-07, -1.2107e-08, 3.2596e-09]], device='cuda:0') Epoch 166, bias, value: tensor([ 0.0024, 0.0114, 0.0195, -0.0024, 0.0223, -0.0082, -0.0085, -0.0067, 0.0151, -0.0127], device='cuda:0'), grad: tensor([-1.0580e-06, -1.0263e-06, -5.5879e-07, 2.7381e-07, -5.9372e-07, 1.1995e-06, 6.5332e-07, 1.3625e-06, -1.8626e-06, 1.5954e-06], device='cuda:0') 100 0.0001 changing lr epoch 165, time 251.69, cls_loss 0.0011 cls_loss_mapping 0.0037 cls_loss_causal 0.5375 re_mapping 0.0051 re_causal 0.0159 /// teacc 99.19 lr 0.00010000 Epoch 167, weight, value: tensor([[-0.1491, -0.0236, 0.1619, ..., 0.0412, -0.1435, -0.1166], [-0.0977, -0.0980, -0.0256, ..., -0.0898, -0.1148, -0.0561], [ 0.1876, -0.0477, -0.1483, ..., -0.1276, -0.1346, -0.1757], ..., [-0.0546, -0.0102, -0.1492, ..., -0.0597, 0.0744, 0.0125], [-0.1394, -0.0867, -0.0213, ..., -0.1277, -0.0164, -0.0226], [-0.0528, -0.0705, -0.0729, ..., 0.0885, 0.0285, -0.1469]], device='cuda:0'), grad: tensor([[ 2.5658e-07, 1.0710e-08, -5.4482e-08, ..., -4.8429e-08, 2.6543e-08, 2.2352e-08], [ 8.2888e-08, 1.1642e-08, -8.1025e-08, ..., -3.6322e-08, 7.3574e-08, 2.7940e-09], [-9.2480e-07, 3.5390e-08, 1.8161e-08, ..., 9.3132e-09, 5.7742e-08, 4.1910e-09], ..., [ 6.0443e-07, 2.1420e-08, 3.3993e-08, ..., 2.7474e-08, -3.7486e-07, 0.0000e+00], [ 7.4971e-08, 1.8161e-08, 2.3283e-08, ..., 9.3132e-09, 2.4680e-08, 9.7789e-09], [ 1.3690e-07, 8.1491e-08, 5.5879e-08, ..., 4.2561e-07, 2.5285e-07, 1.8626e-09]], device='cuda:0') Epoch 167, bias, value: tensor([ 0.0027, 0.0116, 0.0189, -0.0026, 0.0224, -0.0082, -0.0088, -0.0067, 0.0152, -0.0131], device='cuda:0'), grad: tensor([ 7.0268e-07, -2.3041e-06, -1.1921e-06, -1.2778e-06, -7.4040e-08, 1.7369e-07, 3.9535e-07, 7.7439e-07, 6.1700e-07, 2.1830e-06], device='cuda:0') 100 0.0001 changing lr epoch 166, time 251.88, cls_loss 0.0009 cls_loss_mapping 0.0032 cls_loss_causal 0.5277 re_mapping 0.0048 re_causal 0.0158 /// teacc 99.06 lr 0.00010000 Epoch 168, weight, value: tensor([[-0.1493, -0.0237, 0.1623, ..., 0.0412, -0.1441, -0.1166], [-0.0981, -0.0985, -0.0256, ..., -0.0899, -0.1151, -0.0562], [ 0.1881, -0.0481, -0.1488, ..., -0.1280, -0.1349, -0.1757], ..., [-0.0547, -0.0104, -0.1499, ..., -0.0608, 0.0744, 0.0125], [-0.1403, -0.0868, -0.0213, ..., -0.1282, -0.0164, -0.0226], [-0.0531, -0.0706, -0.0731, ..., 0.0884, 0.0286, -0.1469]], device='cuda:0'), grad: tensor([[ 1.4901e-08, 9.7789e-09, 2.7940e-09, ..., 7.7300e-08, 6.1002e-08, 1.3970e-08], [ 5.7742e-08, 5.1223e-09, 1.3970e-09, ..., 3.7951e-07, 4.6846e-07, 1.0245e-08], [-1.0990e-07, 9.3132e-09, 1.8626e-09, ..., 7.7765e-08, 1.2992e-07, 2.1886e-08], ..., [ 4.3772e-08, 3.3528e-08, 0.0000e+00, ..., 3.2969e-07, -2.0768e-06, 1.5832e-08], [ 6.4261e-08, 8.8476e-08, 9.3132e-10, ..., 9.2108e-07, 2.7521e-07, -1.1269e-07], [-9.4529e-08, 1.6764e-08, 0.0000e+00, ..., -2.9728e-06, 9.1223e-07, 1.0245e-08]], device='cuda:0') Epoch 168, bias, value: tensor([ 0.0027, 0.0118, 0.0185, -0.0028, 0.0227, -0.0078, -0.0091, -0.0068, 0.0151, -0.0132], device='cuda:0'), grad: tensor([ 4.9965e-07, 2.4512e-06, 9.0152e-07, 2.8349e-06, 7.9535e-07, -2.9840e-06, 1.0682e-06, -4.3996e-06, 1.2927e-06, -2.4531e-06], device='cuda:0') 100 0.0001 changing lr epoch 167, time 251.70, cls_loss 0.0011 cls_loss_mapping 0.0049 cls_loss_causal 0.5312 re_mapping 0.0051 re_causal 0.0160 /// teacc 99.07 lr 0.00010000 Epoch 169, weight, value: tensor([[-0.1505, -0.0241, 0.1627, ..., 0.0413, -0.1449, -0.1167], [-0.0986, -0.0991, -0.0256, ..., -0.0910, -0.1161, -0.0563], [ 0.1887, -0.0483, -0.1492, ..., -0.1285, -0.1355, -0.1761], ..., [-0.0551, -0.0109, -0.1516, ..., -0.0620, 0.0752, 0.0128], [-0.1407, -0.0873, -0.0220, ..., -0.1291, -0.0174, -0.0231], [-0.0532, -0.0708, -0.0735, ..., 0.0886, 0.0287, -0.1469]], device='cuda:0'), grad: tensor([[ 1.1642e-08, 0.0000e+00, 3.2987e-06, ..., 7.9945e-06, 9.2341e-07, 1.5832e-08], [ 1.5367e-08, 0.0000e+00, -5.4482e-08, ..., 3.7067e-07, 8.8941e-08, 7.9162e-09], [-3.2596e-07, 1.3970e-09, 3.8836e-07, ..., 2.0815e-07, 2.3190e-07, 2.7940e-09], ..., [ 2.4354e-07, 9.3132e-10, 1.2107e-08, ..., 3.8231e-07, -3.3975e-06, 1.8626e-09], [ 1.8161e-08, 0.0000e+00, 4.7535e-06, ..., 5.4482e-07, 1.2927e-06, 1.8626e-09], [ 4.6566e-09, 0.0000e+00, 9.8255e-08, ..., 9.0227e-06, 1.3737e-07, 1.8626e-09]], device='cuda:0') Epoch 169, bias, value: tensor([ 0.0027, 0.0118, 0.0184, -0.0022, 0.0228, -0.0079, -0.0087, -0.0068, 0.0146, -0.0132], device='cuda:0'), grad: tensor([ 2.2605e-05, -2.5344e-04, 3.1665e-06, 6.5923e-05, -3.9369e-05, 2.5276e-06, -1.6034e-05, 1.8311e-04, 1.1489e-05, 1.9699e-05], device='cuda:0') 100 0.0001 changing lr epoch 168, time 251.83, cls_loss 0.0011 cls_loss_mapping 0.0046 cls_loss_causal 0.5317 re_mapping 0.0053 re_causal 0.0170 /// teacc 99.05 lr 0.00010000 Epoch 170, weight, value: tensor([[-0.1508, -0.0246, 0.1659, ..., 0.0417, -0.1454, -0.1166], [-0.1012, -0.0999, -0.0289, ..., -0.0925, -0.1169, -0.0577], [ 0.1907, -0.0494, -0.1501, ..., -0.1292, -0.1357, -0.1764], ..., [-0.0554, -0.0112, -0.1523, ..., -0.0634, 0.0752, 0.0128], [-0.1408, -0.0884, -0.0223, ..., -0.1303, -0.0177, -0.0232], [-0.0534, -0.0711, -0.0747, ..., 0.0888, 0.0289, -0.1470]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 5.5879e-09, -5.7742e-08, ..., 1.3048e-06, 9.3132e-08, 8.3819e-09], [ 9.3132e-09, 1.0245e-08, 2.4214e-08, ..., 6.7987e-07, 1.5181e-07, 6.5193e-09], [ 4.8429e-08, 5.2154e-08, 2.7008e-08, ..., 6.0722e-07, 1.2852e-07, 7.4506e-09], ..., [ 3.7253e-09, 5.5879e-09, 4.2841e-08, ..., 1.4324e-06, 3.1665e-07, 1.3039e-08], [ 1.4249e-07, 1.5646e-07, 2.7008e-08, ..., 9.8348e-07, 2.1141e-07, 5.5879e-08], [ 1.8626e-09, 1.8626e-09, 1.1642e-07, ..., 5.8115e-07, 4.3120e-07, 2.0489e-08]], device='cuda:0') Epoch 170, bias, value: tensor([ 0.0038, 0.0115, 0.0191, -0.0013, 0.0225, -0.0080, -0.0088, -0.0071, 0.0143, -0.0130], device='cuda:0'), grad: tensor([ 2.8946e-06, 2.0154e-06, 3.0305e-06, -8.0373e-07, -2.6599e-05, -7.1339e-07, 1.1951e-05, 1.4752e-06, 3.7514e-06, 3.0193e-06], device='cuda:0') 100 0.0001 changing lr epoch 169, time 251.58, cls_loss 0.0011 cls_loss_mapping 0.0031 cls_loss_causal 0.4971 re_mapping 0.0049 re_causal 0.0150 /// teacc 99.12 lr 0.00010000 Epoch 171, weight, value: tensor([[-0.1513, -0.0251, 0.1662, ..., 0.0418, -0.1470, -0.1174], [-0.1014, -0.1004, -0.0289, ..., -0.0927, -0.1174, -0.0579], [ 0.1911, -0.0497, -0.1515, ..., -0.1301, -0.1360, -0.1768], ..., [-0.0555, -0.0113, -0.1542, ..., -0.0639, 0.0756, 0.0127], [-0.1408, -0.0886, -0.0224, ..., -0.1308, -0.0179, -0.0233], [-0.0537, -0.0713, -0.0751, ..., 0.0889, 0.0289, -0.1471]], device='cuda:0'), grad: tensor([[ 2.4494e-07, 8.1956e-08, -2.4308e-07, ..., -1.5832e-08, 3.2596e-08, 3.5390e-08], [ 1.7192e-06, 3.6322e-08, 3.0734e-08, ..., 6.7893e-07, 9.9652e-08, 2.7008e-08], [-4.0643e-06, 6.1467e-08, 4.0978e-08, ..., 8.3819e-08, 6.0536e-08, 2.0489e-08], ..., [ 6.5286e-07, 7.4506e-08, 3.5390e-08, ..., 7.2271e-07, -2.6077e-08, 5.6811e-08], [ 8.2608e-07, 8.1956e-08, 7.8231e-08, ..., 3.3155e-07, 4.6566e-09, 4.4703e-08], [ 1.7509e-07, 1.4622e-07, 1.2480e-07, ..., -3.6042e-07, -2.0955e-07, 1.6764e-08]], device='cuda:0') Epoch 171, bias, value: tensor([ 0.0038, 0.0115, 0.0185, -0.0013, 0.0222, -0.0082, -0.0084, -0.0071, 0.0145, -0.0130], device='cuda:0'), grad: tensor([ 5.4296e-07, 5.5134e-06, -6.0983e-06, -1.1176e-07, -5.4576e-06, -3.5204e-07, 1.2554e-06, 2.6636e-06, 1.4277e-06, 6.2492e-07], device='cuda:0') 100 0.0001 changing lr epoch 170, time 251.56, cls_loss 0.0012 cls_loss_mapping 0.0041 cls_loss_causal 0.5399 re_mapping 0.0051 re_causal 0.0157 /// teacc 98.97 lr 0.00010000 Epoch 172, weight, value: tensor([[-0.1519, -0.0260, 0.1664, ..., 0.0418, -0.1485, -0.1179], [-0.1017, -0.1012, -0.0289, ..., -0.0920, -0.1179, -0.0579], [ 0.1937, -0.0498, -0.1523, ..., -0.1292, -0.1346, -0.1772], ..., [-0.0559, -0.0111, -0.1557, ..., -0.0648, 0.0760, 0.0129], [-0.1417, -0.0889, -0.0225, ..., -0.1320, -0.0181, -0.0234], [-0.0536, -0.0715, -0.0755, ..., 0.0892, 0.0292, -0.1472]], device='cuda:0'), grad: tensor([[ 6.9849e-08, 0.0000e+00, 3.8184e-08, ..., 3.2596e-08, 1.6205e-07, 9.6858e-08], [ 4.6566e-09, 2.7940e-09, 1.8626e-09, ..., 1.4808e-07, 2.5984e-07, 3.0734e-08], [ 1.7695e-08, 1.8626e-09, 1.0245e-08, ..., 7.7300e-08, 1.1083e-07, 2.0489e-08], ..., [ 3.1665e-08, 2.7940e-09, 5.5879e-09, ..., 3.5577e-07, -4.3958e-07, 1.3970e-08], [ 2.1420e-08, 1.2107e-08, 2.7940e-08, ..., 1.7881e-07, 3.6787e-07, 1.7323e-07], [ 1.3784e-07, 9.3132e-10, 5.5879e-09, ..., 1.9744e-07, 1.4156e-07, 7.4506e-09]], device='cuda:0') Epoch 172, bias, value: tensor([ 0.0038, 0.0116, 0.0197, -0.0017, 0.0215, -0.0082, -0.0088, -0.0071, 0.0144, -0.0127], device='cuda:0'), grad: tensor([ 1.3709e-06, -4.4018e-05, 5.9828e-06, 2.5760e-06, 2.0787e-06, -4.9360e-06, 2.1793e-06, -4.6194e-07, 3.3557e-05, 1.6587e-06], device='cuda:0') 100 0.0001 changing lr epoch 171, time 251.58, cls_loss 0.0012 cls_loss_mapping 0.0044 cls_loss_causal 0.5582 re_mapping 0.0047 re_causal 0.0149 /// teacc 99.03 lr 0.00010000 Epoch 173, weight, value: tensor([[-0.1508, -0.0264, 0.1672, ..., 0.0420, -0.1494, -0.1181], [-0.1021, -0.1020, -0.0289, ..., -0.0922, -0.1183, -0.0576], [ 0.1945, -0.0499, -0.1550, ..., -0.1308, -0.1343, -0.1777], ..., [-0.0555, -0.0109, -0.1563, ..., -0.0651, 0.0765, 0.0129], [-0.1427, -0.0894, -0.0226, ..., -0.1327, -0.0183, -0.0236], [-0.0540, -0.0719, -0.0762, ..., 0.0891, 0.0294, -0.1473]], device='cuda:0'), grad: tensor([[ 6.0536e-08, 3.8184e-08, 1.4473e-06, ..., -7.0781e-08, 6.5286e-07, 9.7416e-07], [ 3.8184e-08, 1.5832e-08, 6.2957e-07, ..., 2.3097e-06, 3.7439e-06, 4.6939e-07], [-4.5635e-08, 1.2107e-08, 1.1362e-07, ..., 9.4995e-08, 1.3039e-07, 7.3574e-08], ..., [ 7.7300e-08, 1.3970e-08, 4.4703e-08, ..., -3.4198e-06, -5.3495e-06, 3.1665e-08], [ 3.6322e-08, 2.6915e-07, 4.1611e-06, ..., 1.2200e-07, 2.4345e-06, 3.7309e-06], [-5.4017e-08, 9.6858e-08, 2.7567e-07, ..., 3.6322e-07, 1.4594e-06, 1.7043e-07]], device='cuda:0') Epoch 173, bias, value: tensor([ 0.0040, 0.0117, 0.0198, -0.0020, 0.0216, -0.0082, -0.0089, -0.0070, 0.0142, -0.0130], device='cuda:0'), grad: tensor([ 4.0568e-06, 2.3931e-05, 1.3141e-06, 4.8056e-06, 2.4512e-06, -9.1344e-06, -1.5855e-05, -3.8207e-05, 1.5154e-05, 1.1533e-05], device='cuda:0') 100 0.0001 changing lr epoch 172, time 251.50, cls_loss 0.0011 cls_loss_mapping 0.0030 cls_loss_causal 0.5185 re_mapping 0.0046 re_causal 0.0148 /// teacc 99.12 lr 0.00010000 Epoch 174, weight, value: tensor([[-0.1512, -0.0269, 0.1677, ..., 0.0421, -0.1503, -0.1183], [-0.1026, -0.1034, -0.0292, ..., -0.0927, -0.1186, -0.0580], [ 0.1952, -0.0501, -0.1564, ..., -0.1316, -0.1346, -0.1771], ..., [-0.0559, -0.0118, -0.1569, ..., -0.0661, 0.0765, 0.0130], [-0.1431, -0.0898, -0.0226, ..., -0.1335, -0.0185, -0.0237], [-0.0542, -0.0719, -0.0765, ..., 0.0889, 0.0295, -0.1473]], device='cuda:0'), grad: tensor([[ 1.9558e-08, 5.5879e-09, -2.8908e-05, ..., -1.4998e-05, 7.8231e-08, 6.5193e-09], [ 4.8429e-08, 1.0245e-08, 8.1584e-07, ..., 5.1130e-07, 9.3970e-07, -4.3772e-08], [ 5.9232e-07, 1.4901e-07, 3.1032e-06, ..., 1.6456e-06, 5.7649e-07, 5.5879e-09], ..., [ 1.1176e-08, 2.6077e-08, 1.9893e-06, ..., 1.0980e-06, -4.1164e-06, 8.3819e-09], [ 2.3283e-08, 5.5879e-09, 1.2815e-06, ..., 6.8825e-07, 2.8871e-08, 2.2352e-08], [ 1.0245e-08, 2.7940e-09, 1.3120e-05, ..., 6.7465e-06, 4.4797e-07, 5.5879e-09]], device='cuda:0') Epoch 174, bias, value: tensor([ 0.0041, 0.0117, 0.0200, -0.0018, 0.0221, -0.0077, -0.0093, -0.0071, 0.0141, -0.0133], device='cuda:0'), grad: tensor([-6.2287e-05, 4.7013e-06, 1.5736e-05, 1.4320e-05, 1.6717e-06, 5.2936e-06, 1.0833e-05, -2.0608e-05, -3.5092e-06, 3.3915e-05], device='cuda:0') 100 0.0001 changing lr epoch 173, time 251.06, cls_loss 0.0010 cls_loss_mapping 0.0034 cls_loss_causal 0.5372 re_mapping 0.0047 re_causal 0.0153 /// teacc 99.05 lr 0.00010000 Epoch 175, weight, value: tensor([[-0.1518, -0.0274, 0.1685, ..., 0.0422, -0.1518, -0.1185], [-0.1028, -0.1044, -0.0298, ..., -0.0936, -0.1190, -0.0579], [ 0.1954, -0.0503, -0.1572, ..., -0.1323, -0.1349, -0.1777], ..., [-0.0561, -0.0124, -0.1583, ..., -0.0671, 0.0765, 0.0131], [-0.1434, -0.0902, -0.0228, ..., -0.1342, -0.0187, -0.0238], [-0.0526, -0.0719, -0.0770, ..., 0.0891, 0.0293, -0.1474]], device='cuda:0'), grad: tensor([[ 2.4214e-08, -1.4063e-07, -1.5525e-06, ..., -3.4366e-07, 8.1025e-08, 8.6613e-08], [ 3.1665e-08, 1.2107e-08, -2.6077e-08, ..., 1.3039e-08, 1.0245e-07, 5.9605e-08], [-3.0734e-08, 4.0978e-08, 1.6205e-07, ..., 5.1223e-08, 5.9605e-08, 6.7055e-08], ..., [ 7.1712e-08, 3.6322e-08, 9.3132e-08, ..., -1.2107e-08, -2.3469e-07, 1.8626e-08], [ 3.8184e-08, 7.0781e-08, 5.6438e-07, ..., 1.3318e-07, 8.3819e-09, -1.8720e-07], [ 4.4703e-08, 3.4459e-08, 2.9709e-07, ..., 3.7253e-08, 1.5367e-07, 6.4261e-08]], device='cuda:0') Epoch 175, bias, value: tensor([ 0.0043, 0.0118, 0.0197, -0.0021, 0.0223, -0.0075, -0.0095, -0.0073, 0.0140, -0.0130], device='cuda:0'), grad: tensor([-3.0361e-06, -1.3784e-07, 7.1805e-07, -3.7253e-08, 4.1537e-07, -1.9670e-06, 3.3733e-06, -3.5949e-07, -4.7404e-07, 1.4668e-06], device='cuda:0') 100 0.0001 changing lr epoch 174, time 251.63, cls_loss 0.0010 cls_loss_mapping 0.0034 cls_loss_causal 0.5300 re_mapping 0.0050 re_causal 0.0156 /// teacc 99.08 lr 0.00010000 Epoch 176, weight, value: tensor([[-0.1524, -0.0291, 0.1687, ..., 0.0422, -0.1525, -0.1186], [-0.1021, -0.1056, -0.0298, ..., -0.0935, -0.1194, -0.0579], [ 0.1951, -0.0514, -0.1578, ..., -0.1328, -0.1358, -0.1794], ..., [-0.0561, -0.0123, -0.1599, ..., -0.0679, 0.0770, 0.0139], [-0.1437, -0.0904, -0.0228, ..., -0.1351, -0.0189, -0.0238], [-0.0527, -0.0725, -0.0777, ..., 0.0893, 0.0296, -0.1476]], device='cuda:0'), grad: tensor([[ 1.1176e-08, 1.3318e-07, -9.4995e-08, ..., -8.1025e-08, 1.9465e-07, 1.3039e-08], [ 5.1223e-08, 2.3283e-08, -1.7229e-07, ..., 2.9802e-08, 8.6613e-08, 1.1176e-08], [ 2.2352e-08, 3.6322e-08, 1.0896e-07, ..., 3.9116e-08, 2.6077e-08, 1.2107e-08], ..., [ 4.9360e-08, 3.6322e-08, 7.0781e-08, ..., 1.9837e-07, -8.2888e-08, 6.5193e-09], [ 4.7497e-08, 4.7777e-07, 3.5949e-07, ..., 1.6205e-07, 6.2771e-07, 3.4459e-08], [ 2.4214e-08, 7.0781e-08, 1.6298e-07, ..., -9.1642e-07, -1.5646e-07, 4.6566e-09]], device='cuda:0') Epoch 176, bias, value: tensor([ 0.0042, 0.0121, 0.0178, -0.0022, 0.0220, -0.0073, -0.0093, -0.0074, 0.0146, -0.0130], device='cuda:0'), grad: tensor([ 8.7265e-07, -2.0582e-06, 1.0561e-06, -5.9195e-06, 9.1828e-07, 1.5749e-06, 1.1716e-06, 1.1465e-06, 3.2783e-06, -2.0824e-06], device='cuda:0') 100 0.0001 changing lr epoch 175, time 251.68, cls_loss 0.0010 cls_loss_mapping 0.0032 cls_loss_causal 0.5041 re_mapping 0.0047 re_causal 0.0146 /// teacc 99.06 lr 0.00010000 Epoch 177, weight, value: tensor([[-0.1534, -0.0292, 0.1691, ..., 0.0423, -0.1534, -0.1188], [-0.1035, -0.1065, -0.0298, ..., -0.0937, -0.1197, -0.0581], [ 0.1956, -0.0518, -0.1590, ..., -0.1337, -0.1361, -0.1783], ..., [-0.0551, -0.0124, -0.1608, ..., -0.0682, 0.0773, 0.0127], [-0.1443, -0.0906, -0.0227, ..., -0.1360, -0.0190, -0.0238], [-0.0531, -0.0728, -0.0782, ..., 0.0894, 0.0297, -0.1478]], device='cuda:0'), grad: tensor([[ 9.6858e-08, 4.2841e-08, -3.7532e-07, ..., -2.1514e-07, 1.4156e-07, 6.0536e-08], [ 4.1723e-07, 1.7695e-08, 6.7987e-08, ..., 2.5146e-08, 1.1362e-07, 2.2352e-08], [-3.0547e-07, 3.5390e-08, 8.9407e-08, ..., 1.9372e-07, 2.0955e-07, 1.4901e-08], ..., [-1.7509e-06, 2.9802e-08, 3.1665e-08, ..., 3.6322e-08, -1.0449e-06, 9.3132e-09], [ 1.6019e-07, 6.5193e-08, 3.1758e-07, ..., 1.2200e-07, 2.0955e-07, 1.3318e-07], [ 1.0310e-06, 3.4459e-08, 2.1514e-07, ..., -6.1002e-07, 6.4634e-07, 2.7940e-08]], device='cuda:0') Epoch 177, bias, value: tensor([ 0.0041, 0.0119, 0.0176, -0.0024, 0.0217, -0.0070, -0.0091, -0.0071, 0.0149, -0.0130], device='cuda:0'), grad: tensor([ 1.2666e-07, -2.9225e-06, 1.7220e-06, 9.3039e-07, 1.0645e-06, 1.5274e-07, -1.0720e-06, -5.8860e-06, 1.6261e-06, 4.2319e-06], device='cuda:0') 100 0.0001 changing lr epoch 176, time 251.80, cls_loss 0.0014 cls_loss_mapping 0.0032 cls_loss_causal 0.5021 re_mapping 0.0047 re_causal 0.0141 /// teacc 99.10 lr 0.00010000 Epoch 178, weight, value: tensor([[-0.1550, -0.0293, 0.1699, ..., 0.0408, -0.1543, -0.1190], [-0.1025, -0.1074, -0.0298, ..., -0.0943, -0.1201, -0.0581], [ 0.1942, -0.0544, -0.1600, ..., -0.1360, -0.1366, -0.1787], ..., [-0.0564, -0.0129, -0.1622, ..., -0.0704, 0.0777, 0.0126], [-0.1451, -0.0908, -0.0228, ..., -0.1373, -0.0194, -0.0241], [-0.0511, -0.0730, -0.0798, ..., 0.0913, 0.0299, -0.1478]], device='cuda:0'), grad: tensor([[ 8.7544e-08, 0.0000e+00, -4.2096e-07, ..., -2.4121e-07, 4.0047e-08, 3.3528e-08], [ 1.0151e-07, 9.3132e-10, 4.9360e-08, ..., 7.6368e-08, 2.0303e-07, 4.8429e-08], [-6.8825e-07, 1.8626e-09, 9.2201e-08, ..., 2.7008e-08, 6.7987e-08, -2.0489e-08], ..., [ 3.8650e-07, 1.8626e-09, 1.4901e-08, ..., 1.0990e-07, -4.7870e-07, 2.7940e-08], [ 4.0047e-08, 0.0000e+00, -3.2224e-07, ..., -6.6124e-08, -6.7055e-08, -4.1910e-08], [ 4.2841e-08, 9.3132e-10, 4.3400e-07, ..., -1.9558e-07, 4.2841e-08, 4.0978e-08]], device='cuda:0') Epoch 178, bias, value: tensor([ 0.0025, 0.0124, 0.0147, -0.0015, 0.0212, -0.0069, -0.0089, -0.0076, 0.0144, -0.0106], device='cuda:0'), grad: tensor([-3.4459e-08, 1.5469e-06, 1.8068e-07, 1.6298e-06, 7.3854e-07, 4.6846e-07, 1.6019e-07, -9.3412e-07, -7.8008e-06, 4.0345e-06], device='cuda:0') 100 0.0001 changing lr epoch 177, time 251.13, cls_loss 0.0011 cls_loss_mapping 0.0031 cls_loss_causal 0.4937 re_mapping 0.0046 re_causal 0.0138 /// teacc 99.03 lr 0.00010000 Epoch 179, weight, value: tensor([[-0.1559, -0.0300, 0.1704, ..., 0.0408, -0.1556, -0.1193], [-0.1029, -0.1084, -0.0299, ..., -0.0943, -0.1206, -0.0585], [ 0.1947, -0.0546, -0.1609, ..., -0.1365, -0.1368, -0.1787], ..., [-0.0561, -0.0131, -0.1635, ..., -0.0705, 0.0784, 0.0129], [-0.1460, -0.0910, -0.0229, ..., -0.1379, -0.0196, -0.0243], [-0.0518, -0.0736, -0.0807, ..., 0.0914, 0.0298, -0.1479]], device='cuda:0'), grad: tensor([[ 5.1130e-07, 9.3132e-10, -3.6322e-08, ..., 6.5658e-07, 9.8161e-07, 3.6322e-08], [-2.3860e-06, 6.5193e-09, 6.1467e-08, ..., 8.6799e-07, 4.5858e-06, 1.3970e-08], [ 3.5353e-06, 3.7253e-09, 3.5390e-08, ..., 3.5204e-07, 6.2492e-07, 6.5193e-09], ..., [-4.8578e-06, 4.6566e-09, 1.7695e-08, ..., -4.1500e-06, -1.0155e-05, 1.8626e-09], [ 7.7300e-08, 1.8626e-09, 4.3958e-07, ..., 2.1234e-07, 1.9558e-07, 3.4459e-08], [ 7.8417e-07, 6.5193e-09, 1.0151e-07, ..., 2.1279e-05, 4.9658e-06, 3.7253e-08]], device='cuda:0') Epoch 179, bias, value: tensor([ 0.0024, 0.0124, 0.0144, -0.0020, 0.0216, -0.0063, -0.0092, -0.0074, 0.0142, -0.0107], device='cuda:0'), grad: tensor([ 5.8226e-06, -4.6372e-05, 5.3138e-05, 1.1802e-05, -2.6196e-05, 3.5856e-06, -4.3120e-07, -4.8131e-05, 6.0629e-07, 4.6104e-05], device='cuda:0') 100 0.0001 changing lr epoch 178, time 251.65, cls_loss 0.0010 cls_loss_mapping 0.0033 cls_loss_causal 0.5307 re_mapping 0.0046 re_causal 0.0149 /// teacc 98.91 lr 0.00010000 Epoch 180, weight, value: tensor([[-0.1558, -0.0300, 0.1716, ..., 0.0408, -0.1566, -0.1194], [-0.1026, -0.1097, -0.0299, ..., -0.0922, -0.1212, -0.0586], [ 0.1950, -0.0545, -0.1618, ..., -0.1376, -0.1370, -0.1788], ..., [-0.0563, -0.0134, -0.1648, ..., -0.0729, 0.0787, 0.0129], [-0.1462, -0.0908, -0.0230, ..., -0.1393, -0.0197, -0.0244], [-0.0527, -0.0749, -0.0826, ..., 0.0914, 0.0285, -0.1496]], device='cuda:0'), grad: tensor([[ 7.3109e-07, 1.1176e-08, -3.4459e-08, ..., 1.2573e-07, 2.5705e-07, 3.5390e-08], [ 1.3253e-06, 1.2200e-07, 8.3819e-09, ..., 1.1176e-07, 5.7463e-07, 2.3469e-07], [-4.6752e-07, 2.7101e-07, 1.3970e-08, ..., 4.0978e-08, 8.0094e-08, 1.6764e-08], ..., [ 1.0226e-06, 3.2596e-08, 1.8626e-09, ..., 9.1270e-08, -1.2415e-06, -5.4017e-07], [ 4.1071e-07, 9.3132e-09, 1.3039e-08, ..., 1.5926e-07, 7.3574e-08, -1.1642e-07], [ 1.1744e-06, 1.7695e-08, 5.1223e-08, ..., -3.1106e-07, 5.1875e-07, 3.0734e-08]], device='cuda:0') Epoch 180, bias, value: tensor([ 0.0026, 0.0131, 0.0138, -0.0023, 0.0215, -0.0050, -0.0093, -0.0079, 0.0146, -0.0118], device='cuda:0'), grad: tensor([ 2.3246e-06, 4.7013e-06, 1.0524e-07, -3.4831e-07, 1.2442e-06, -8.6650e-06, 9.7137e-07, -2.7437e-06, -1.5143e-06, 3.8780e-06], device='cuda:0') 100 0.0001 changing lr epoch 179, time 251.14, cls_loss 0.0011 cls_loss_mapping 0.0039 cls_loss_causal 0.5655 re_mapping 0.0046 re_causal 0.0148 /// teacc 99.15 lr 0.00010000 Epoch 181, weight, value: tensor([[-0.1568, -0.0301, 0.1720, ..., 0.0409, -0.1582, -0.1196], [-0.1038, -0.1116, -0.0300, ..., -0.0925, -0.1222, -0.0588], [ 0.1958, -0.0546, -0.1622, ..., -0.1381, -0.1372, -0.1794], ..., [-0.0560, -0.0136, -0.1653, ..., -0.0730, 0.0797, 0.0131], [-0.1465, -0.0911, -0.0230, ..., -0.1405, -0.0198, -0.0244], [-0.0530, -0.0751, -0.0834, ..., 0.0917, 0.0285, -0.1498]], device='cuda:0'), grad: tensor([[ 1.5832e-08, 4.6566e-09, -3.1851e-07, ..., -1.1828e-07, 3.2596e-08, 1.4901e-08], [ 4.5635e-08, 9.3132e-09, -2.7940e-09, ..., 6.2399e-08, 6.2399e-08, 5.5879e-09], [-2.1886e-07, 3.7253e-09, 2.7008e-08, ..., 1.4901e-08, 1.7695e-08, 5.5879e-09], ..., [ 1.2200e-07, 1.5832e-08, 2.1420e-08, ..., 9.4064e-08, -6.1467e-08, 1.8626e-09], [ 2.9802e-08, 2.7008e-08, 1.0338e-07, ..., 2.5239e-07, 1.7695e-07, 3.7253e-08], [ 1.4901e-08, 4.3772e-08, 1.0245e-07, ..., -4.7497e-07, 7.6368e-08, 8.3819e-09]], device='cuda:0') Epoch 181, bias, value: tensor([ 0.0027, 0.0128, 0.0140, -0.0022, 0.0212, -0.0054, -0.0093, -0.0076, 0.0151, -0.0116], device='cuda:0'), grad: tensor([-2.9150e-07, -4.0978e-07, -9.9652e-08, 1.6093e-06, 6.7987e-07, -3.6955e-06, 4.6566e-08, 6.2212e-07, 1.8114e-06, -2.9616e-07], device='cuda:0') 100 0.0001 changing lr epoch 180, time 251.42, cls_loss 0.0013 cls_loss_mapping 0.0043 cls_loss_causal 0.5228 re_mapping 0.0048 re_causal 0.0151 /// teacc 99.02 lr 0.00010000 Epoch 182, weight, value: tensor([[-0.1572, -0.0301, 0.1729, ..., 0.0409, -0.1599, -0.1198], [-0.1046, -0.1130, -0.0301, ..., -0.0932, -0.1231, -0.0589], [ 0.1963, -0.0548, -0.1627, ..., -0.1385, -0.1375, -0.1794], ..., [-0.0560, -0.0137, -0.1663, ..., -0.0745, 0.0791, 0.0131], [-0.1471, -0.0915, -0.0232, ..., -0.1415, -0.0200, -0.0244], [-0.0532, -0.0754, -0.0839, ..., 0.0911, 0.0277, -0.1499]], device='cuda:0'), grad: tensor([[ 2.3376e-07, 4.6566e-09, -6.7987e-07, ..., -2.0582e-07, 1.0245e-07, 6.5193e-09], [ 4.8280e-06, 1.0245e-08, -2.2445e-07, ..., -1.1828e-07, 6.9384e-07, 2.7940e-09], [-1.5795e-05, 1.1176e-08, 3.5390e-08, ..., 5.3085e-08, 4.4145e-07, 3.7253e-09], ..., [ 7.5623e-06, 1.4901e-08, 1.7509e-07, ..., 2.3656e-07, -1.5292e-06, 0.0000e+00], [ 2.0713e-06, 0.0000e+00, 1.0338e-07, ..., 3.7067e-07, 2.6450e-07, 3.7253e-09], [ 1.2945e-07, 1.7695e-08, 2.9616e-07, ..., 6.8825e-07, 2.7474e-07, 9.3132e-10]], device='cuda:0') Epoch 182, bias, value: tensor([ 0.0028, 0.0119, 0.0142, -0.0022, 0.0220, -0.0058, -0.0090, -0.0070, 0.0150, -0.0122], device='cuda:0'), grad: tensor([ 3.9209e-07, 9.6858e-06, -2.6733e-05, 2.3358e-06, -1.8151e-06, 2.1234e-07, 1.3784e-06, 3.5670e-06, 6.9663e-06, 4.0308e-06], device='cuda:0') 100 0.0001 changing lr epoch 181, time 251.77, cls_loss 0.0012 cls_loss_mapping 0.0035 cls_loss_causal 0.5478 re_mapping 0.0044 re_causal 0.0144 /// teacc 98.92 lr 0.00010000 Epoch 183, weight, value: tensor([[-0.1581, -0.0315, 0.1762, ..., 0.0410, -0.1574, -0.1208], [-0.1049, -0.1150, -0.0301, ..., -0.0935, -0.1239, -0.0590], [ 0.1969, -0.0550, -0.1638, ..., -0.1391, -0.1378, -0.1795], ..., [-0.0562, -0.0143, -0.1676, ..., -0.0750, 0.0798, 0.0131], [-0.1478, -0.0920, -0.0233, ..., -0.1425, -0.0202, -0.0244], [-0.0543, -0.0780, -0.0845, ..., 0.0913, 0.0279, -0.1499]], device='cuda:0'), grad: tensor([[ 4.9360e-08, -3.0734e-08, -1.0682e-06, ..., -3.5111e-07, 9.5926e-08, 3.7253e-09], [ 1.4901e-07, 1.4901e-08, 3.0734e-08, ..., 2.1420e-07, 5.6252e-07, 9.3132e-10], [-4.7870e-07, 9.3132e-09, 8.2888e-08, ..., 1.7881e-07, 8.8476e-08, 9.3132e-10], ..., [ 1.3411e-07, 1.6764e-08, 9.3132e-08, ..., 3.6880e-07, -5.4389e-07, 0.0000e+00], [ 4.8429e-08, 1.3970e-08, 8.2888e-08, ..., 4.0233e-07, 2.0862e-07, 1.8626e-09], [ 3.1665e-08, 2.1420e-08, 2.5798e-07, ..., -2.1867e-06, -1.1278e-06, 9.3132e-10]], device='cuda:0') Epoch 183, bias, value: tensor([ 0.0044, 0.0118, 0.0141, -0.0015, 0.0218, -0.0056, -0.0111, -0.0069, 0.0150, -0.0123], device='cuda:0'), grad: tensor([-9.7975e-07, 2.6859e-06, 5.5227e-07, 8.2795e-07, 1.5339e-06, 1.5507e-06, 7.8231e-07, -7.9256e-07, -1.6000e-06, -4.5784e-06], device='cuda:0') 100 0.0001 changing lr epoch 182, time 251.90, cls_loss 0.0011 cls_loss_mapping 0.0040 cls_loss_causal 0.4919 re_mapping 0.0045 re_causal 0.0146 /// teacc 98.92 lr 0.00010000 Epoch 184, weight, value: tensor([[-0.1594, -0.0317, 0.1763, ..., 0.0411, -0.1577, -0.1211], [-0.1052, -0.1159, -0.0303, ..., -0.0936, -0.1247, -0.0591], [ 0.1973, -0.0552, -0.1647, ..., -0.1394, -0.1382, -0.1794], ..., [-0.0559, -0.0142, -0.1687, ..., -0.0750, 0.0804, 0.0131], [-0.1482, -0.0923, -0.0233, ..., -0.1435, -0.0204, -0.0244], [-0.0547, -0.0782, -0.0848, ..., 0.0909, 0.0273, -0.1500]], device='cuda:0'), grad: tensor([[ 5.5879e-09, 9.3132e-10, -1.5004e-06, ..., -3.6415e-07, 3.1665e-08, 2.7940e-09], [ 4.3772e-08, 1.1176e-08, 3.5390e-08, ..., 3.6135e-07, 7.3388e-07, 3.0734e-08], [ 2.4214e-08, 1.3970e-08, 1.6764e-08, ..., 1.4901e-08, 1.7695e-08, 1.2107e-08], ..., [ 1.1921e-07, 2.2352e-08, 5.2154e-08, ..., -7.4506e-09, -7.4506e-07, 9.3132e-10], [ 4.1910e-08, 1.3970e-08, 8.0094e-08, ..., 1.9651e-07, 2.0117e-07, 1.7043e-07], [ 1.3039e-08, 4.6566e-09, 6.0722e-07, ..., -1.2424e-06, -5.0198e-07, 1.1176e-08]], device='cuda:0') Epoch 184, bias, value: tensor([ 0.0043, 0.0118, 0.0139, -0.0016, 0.0224, -0.0060, -0.0104, -0.0068, 0.0151, -0.0130], device='cuda:0'), grad: tensor([-2.0713e-06, 2.0750e-06, 2.6636e-07, 1.5302e-06, 1.4128e-06, -3.0883e-06, 7.2643e-07, -1.5274e-06, 2.2184e-06, -1.5413e-06], device='cuda:0') 100 0.0001 changing lr epoch 183, time 251.71, cls_loss 0.0008 cls_loss_mapping 0.0028 cls_loss_causal 0.4967 re_mapping 0.0047 re_causal 0.0149 /// teacc 99.01 lr 0.00010000 Epoch 185, weight, value: tensor([[-0.1597, -0.0315, 0.1764, ..., 0.0411, -0.1578, -0.1212], [-0.1054, -0.1164, -0.0302, ..., -0.0937, -0.1253, -0.0598], [ 0.1977, -0.0554, -0.1650, ..., -0.1398, -0.1386, -0.1796], ..., [-0.0562, -0.0144, -0.1702, ..., -0.0753, 0.0808, 0.0131], [-0.1491, -0.0925, -0.0234, ..., -0.1457, -0.0209, -0.0246], [-0.0545, -0.0784, -0.0856, ..., 0.0910, 0.0273, -0.1501]], device='cuda:0'), grad: tensor([[ 5.5879e-08, 9.3132e-10, -3.2783e-07, ..., -2.8126e-07, 1.3039e-07, 1.6112e-07], [ 1.0058e-07, 9.3132e-10, 5.0291e-08, ..., -9.3132e-10, 3.4571e-06, 1.2200e-07], [-1.0207e-06, 9.3132e-10, 6.8918e-08, ..., 3.7253e-08, 3.6322e-08, 1.3039e-08], ..., [ 2.6915e-07, 9.3132e-10, 6.1467e-08, ..., 1.3039e-07, -3.8072e-06, 1.6764e-08], [ 1.6298e-07, 1.8626e-09, 1.2480e-07, ..., 4.4703e-08, 1.4622e-07, 1.6391e-07], [ 2.3283e-08, 9.3132e-10, 2.9523e-07, ..., 1.7602e-07, 1.0617e-07, 4.9360e-08]], device='cuda:0') Epoch 185, bias, value: tensor([ 0.0044, 0.0119, 0.0139, -0.0017, 0.0224, -0.0056, -0.0107, -0.0068, 0.0147, -0.0130], device='cuda:0'), grad: tensor([-2.4587e-07, 2.1935e-05, -1.3700e-06, 1.7313e-06, -2.5332e-07, -1.0319e-06, -2.6450e-07, -2.3291e-05, 1.4212e-06, 1.3355e-06], device='cuda:0') 100 0.0001 changing lr epoch 184, time 252.01, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.5379 re_mapping 0.0044 re_causal 0.0147 /// teacc 99.09 lr 0.00010000 Epoch 186, weight, value: tensor([[-0.1602, -0.0316, 0.1765, ..., 0.0412, -0.1580, -0.1215], [-0.1060, -0.1175, -0.0300, ..., -0.0933, -0.1265, -0.0601], [ 0.1983, -0.0555, -0.1657, ..., -0.1400, -0.1392, -0.1800], ..., [-0.0563, -0.0144, -0.1720, ..., -0.0760, 0.0818, 0.0135], [-0.1497, -0.0925, -0.0235, ..., -0.1468, -0.0210, -0.0245], [-0.0547, -0.0789, -0.0859, ..., 0.0910, 0.0273, -0.1502]], device='cuda:0'), grad: tensor([[ 1.6671e-07, 5.5879e-09, 2.3376e-06, ..., -5.5134e-07, 8.4843e-07, 2.1420e-06], [ 1.1502e-06, 6.5193e-09, 2.4959e-07, ..., -7.0781e-08, 3.2876e-07, 1.6950e-07], [-3.1181e-06, 5.5879e-09, 8.4937e-07, ..., 1.7229e-07, 2.8126e-07, 4.0140e-07], ..., [ 1.2871e-06, 1.7695e-08, 1.4622e-07, ..., 2.4680e-07, -9.2667e-07, 6.5193e-09], [-3.2131e-07, 9.3132e-09, 4.7777e-07, ..., 3.5949e-07, 1.5553e-07, 1.3318e-07], [ 1.5926e-07, 6.5193e-09, 4.2375e-07, ..., -4.6659e-07, -5.3085e-08, 3.4459e-08]], device='cuda:0') Epoch 186, bias, value: tensor([ 0.0043, 0.0118, 0.0138, -0.0017, 0.0224, -0.0056, -0.0108, -0.0066, 0.0151, -0.0132], device='cuda:0'), grad: tensor([ 4.2617e-06, 1.0990e-07, -3.5502e-06, 1.0151e-06, 2.5909e-06, 4.9472e-06, -1.1563e-05, 1.6112e-06, -1.7509e-06, 2.2799e-06], device='cuda:0') 100 0.0001 changing lr epoch 185, time 251.81, cls_loss 0.0009 cls_loss_mapping 0.0036 cls_loss_causal 0.5405 re_mapping 0.0047 re_causal 0.0151 /// teacc 98.97 lr 0.00010000 Epoch 187, weight, value: tensor([[-0.1607, -0.0316, 0.1768, ..., 0.0413, -0.1581, -0.1217], [-0.1067, -0.1185, -0.0302, ..., -0.0934, -0.1271, -0.0602], [ 0.1990, -0.0558, -0.1665, ..., -0.1404, -0.1398, -0.1802], ..., [-0.0565, -0.0140, -0.1737, ..., -0.0761, 0.0828, 0.0139], [-0.1500, -0.0928, -0.0232, ..., -0.1475, -0.0211, -0.0245], [-0.0548, -0.0791, -0.0873, ..., 0.0912, 0.0280, -0.1503]], device='cuda:0'), grad: tensor([[ 4.1910e-08, 1.4901e-08, -2.5146e-08, ..., -9.3132e-09, 2.1420e-08, 2.7940e-09], [ 1.2573e-07, 4.2841e-08, -3.7253e-09, ..., 6.5193e-09, 7.7300e-08, 9.3132e-10], [-3.3528e-06, 5.3085e-08, 7.4506e-09, ..., 2.7940e-09, 3.8184e-08, 1.8626e-09], ..., [ 1.9185e-07, 4.0047e-08, 2.7940e-09, ..., 8.3819e-09, -3.3621e-07, 9.3132e-10], [ 3.2131e-07, 2.2352e-08, 6.5193e-09, ..., 8.3819e-09, 2.1420e-08, 1.8626e-09], [ 2.6077e-08, 2.3283e-08, 2.1420e-08, ..., -1.7788e-07, 8.1956e-08, 9.3132e-10]], device='cuda:0') Epoch 187, bias, value: tensor([ 0.0045, 0.0117, 0.0140, -0.0021, 0.0219, -0.0055, -0.0109, -0.0065, 0.0158, -0.0132], device='cuda:0'), grad: tensor([ 1.9185e-07, 2.3190e-07, -3.7830e-06, -2.5406e-06, 3.2689e-06, 1.9372e-06, 8.1025e-08, -4.2934e-07, 7.4320e-07, 2.9057e-07], device='cuda:0') 100 0.0001 changing lr epoch 186, time 251.77, cls_loss 0.0015 cls_loss_mapping 0.0038 cls_loss_causal 0.5297 re_mapping 0.0045 re_causal 0.0137 /// teacc 99.03 lr 0.00010000 Epoch 188, weight, value: tensor([[-0.1615, -0.0336, 0.1765, ..., 0.0415, -0.1587, -0.1230], [-0.1071, -0.1211, -0.0300, ..., -0.0935, -0.1285, -0.0603], [ 0.1997, -0.0559, -0.1683, ..., -0.1414, -0.1402, -0.1805], ..., [-0.0569, -0.0155, -0.1764, ..., -0.0770, 0.0825, 0.0140], [-0.1511, -0.0942, -0.0236, ..., -0.1509, -0.0220, -0.0246], [-0.0549, -0.0791, -0.0890, ..., 0.0916, 0.0288, -0.1504]], device='cuda:0'), grad: tensor([[ 7.1712e-08, 2.6077e-08, -3.7532e-07, ..., 3.4180e-07, 1.3877e-07, 2.7940e-09], [ 1.7975e-07, 7.3574e-08, -1.1176e-08, ..., 4.1630e-07, 5.0291e-07, 2.1420e-08], [ 1.6671e-07, 2.1141e-07, 1.7695e-08, ..., 4.4610e-07, 1.6391e-07, 2.7940e-09], ..., [ 3.0920e-07, 1.2759e-07, 3.1665e-08, ..., 9.9652e-07, -7.7393e-07, -3.2596e-08], [ 1.1083e-07, 3.2596e-08, 5.5879e-08, ..., 9.2015e-07, 4.8243e-07, 9.3132e-10], [ 5.8673e-08, 2.7008e-08, 1.3318e-07, ..., -1.3439e-06, -9.3691e-07, 6.5193e-09]], device='cuda:0') Epoch 188, bias, value: tensor([ 0.0043, 0.0117, 0.0141, -0.0026, 0.0217, -0.0043, -0.0103, -0.0067, 0.0152, -0.0130], device='cuda:0'), grad: tensor([ 1.2424e-06, 9.8627e-07, 2.4289e-06, -1.7760e-06, -7.0632e-06, 1.3961e-06, 1.3188e-06, -5.9605e-08, 3.6731e-06, -2.1793e-06], device='cuda:0') 100 0.0001 changing lr epoch 187, time 251.51, cls_loss 0.0012 cls_loss_mapping 0.0031 cls_loss_causal 0.5084 re_mapping 0.0045 re_causal 0.0143 /// teacc 99.12 lr 0.00010000 Epoch 189, weight, value: tensor([[-0.1622, -0.0337, 0.1767, ..., 0.0395, -0.1588, -0.1233], [-0.1073, -0.1224, -0.0300, ..., -0.0936, -0.1299, -0.0605], [ 0.2003, -0.0561, -0.1697, ..., -0.1421, -0.1409, -0.1806], ..., [-0.0574, -0.0167, -0.1780, ..., -0.0774, 0.0828, 0.0142], [-0.1522, -0.0950, -0.0236, ..., -0.1514, -0.0222, -0.0247], [-0.0550, -0.0792, -0.0897, ..., 0.0933, 0.0287, -0.1504]], device='cuda:0'), grad: tensor([[ 2.3656e-07, 1.4901e-08, -5.6811e-08, ..., -1.6950e-07, 6.0536e-08, 9.4995e-08], [-1.2787e-06, 8.0094e-08, 1.8626e-07, ..., 3.3528e-08, 1.2945e-07, 4.2841e-08], [-1.9521e-06, -1.0245e-08, 4.6566e-08, ..., 7.4506e-09, 1.5926e-07, 1.4901e-08], ..., [ 1.8552e-06, 2.2724e-07, 2.1420e-08, ..., 5.1223e-08, -4.6939e-07, 9.3132e-10], [ 5.7928e-07, 6.2399e-08, 2.1979e-07, ..., 3.7253e-08, 5.8673e-08, 6.5193e-08], [ 7.3574e-08, 9.4995e-08, 1.9837e-07, ..., -1.6298e-07, 3.3528e-08, 4.6566e-09]], device='cuda:0') Epoch 189, bias, value: tensor([ 0.0024, 0.0117, 0.0142, -0.0033, 0.0216, -0.0031, -0.0101, -0.0067, 0.0149, -0.0117], device='cuda:0'), grad: tensor([ 7.3668e-07, -2.0072e-05, 1.7695e-06, -4.2655e-06, 2.6915e-06, 3.1181e-06, -2.2352e-06, 1.5438e-05, 1.9539e-06, 8.6892e-07], device='cuda:0') 100 0.0001 changing lr epoch 188, time 251.42, cls_loss 0.0009 cls_loss_mapping 0.0026 cls_loss_causal 0.5049 re_mapping 0.0046 re_causal 0.0147 /// teacc 98.95 lr 0.00010000 Epoch 190, weight, value: tensor([[-0.1626, -0.0337, 0.1769, ..., 0.0395, -0.1591, -0.1236], [-0.1075, -0.1234, -0.0302, ..., -0.0937, -0.1305, -0.0606], [ 0.2025, -0.0564, -0.1712, ..., -0.1429, -0.1414, -0.1809], ..., [-0.0582, -0.0168, -0.1802, ..., -0.0770, 0.0847, 0.0142], [-0.1557, -0.0954, -0.0241, ..., -0.1517, -0.0223, -0.0248], [-0.0554, -0.0793, -0.0904, ..., 0.0932, 0.0278, -0.1505]], device='cuda:0'), grad: tensor([[ 1.5367e-08, 6.5193e-09, 1.5832e-07, ..., 1.4435e-08, 7.7765e-08, 7.2643e-08], [ 4.7963e-08, 1.5832e-08, -6.9849e-09, ..., 2.6077e-08, 3.1665e-08, -2.5388e-06], [-7.4506e-08, 1.2573e-08, 2.7940e-08, ..., 2.1420e-08, 2.2352e-08, 5.8673e-08], ..., [ 8.4750e-08, 4.3772e-08, 1.0710e-08, ..., 4.6659e-07, 2.8219e-07, 4.1258e-07], [ 3.0734e-08, 5.6345e-08, 1.7276e-07, ..., 4.1910e-08, 2.1514e-07, 1.0161e-06], [ 7.4506e-09, 4.1910e-09, 2.6543e-08, ..., 9.8255e-08, 5.8673e-08, 4.0513e-08]], device='cuda:0') Epoch 190, bias, value: tensor([ 0.0024, 0.0119, 0.0161, -0.0032, 0.0215, -0.0033, -0.0096, -0.0067, 0.0130, -0.0121], device='cuda:0'), grad: tensor([ 1.3644e-06, -1.3769e-04, 2.8387e-06, -4.6566e-08, 2.6658e-05, 1.5870e-05, 1.1757e-05, 2.3857e-05, 5.3197e-05, 2.2110e-06], device='cuda:0') 100 0.0001 changing lr epoch 189, time 251.64, cls_loss 0.0008 cls_loss_mapping 0.0029 cls_loss_causal 0.4971 re_mapping 0.0048 re_causal 0.0148 /// teacc 98.99 lr 0.00010000 Epoch 191, weight, value: tensor([[-0.1628, -0.0338, 0.1767, ..., 0.0393, -0.1594, -0.1247], [-0.1079, -0.1241, -0.0303, ..., -0.0940, -0.1311, -0.0602], [ 0.2029, -0.0565, -0.1717, ..., -0.1433, -0.1419, -0.1810], ..., [-0.0582, -0.0169, -0.1813, ..., -0.0775, 0.0850, 0.0141], [-0.1558, -0.0960, -0.0243, ..., -0.1520, -0.0226, -0.0248], [-0.0555, -0.0794, -0.0908, ..., 0.0937, 0.0281, -0.1505]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 3.7253e-09, -4.8429e-07, ..., -1.3318e-07, 1.2573e-07, 2.5425e-07], [ 2.6077e-08, 3.7253e-09, 3.5018e-07, ..., 2.1700e-07, 5.3365e-07, 6.5845e-07], [-7.5437e-08, 9.3132e-10, 4.0326e-07, ..., 1.9558e-08, 3.5204e-07, 9.1735e-07], ..., [ 3.7253e-08, 5.5879e-09, 4.0978e-08, ..., 7.9907e-07, -2.2724e-07, -4.6566e-09], [ 3.7253e-09, 3.7253e-09, 3.1367e-06, ..., 1.3970e-08, 1.9819e-06, 7.0632e-06], [ 9.3132e-10, 4.6566e-09, 2.0582e-07, ..., -1.0030e-06, -5.1968e-07, 2.7008e-08]], device='cuda:0') Epoch 191, bias, value: tensor([ 0.0019, 0.0120, 0.0161, -0.0032, 0.0213, -0.0032, -0.0093, -0.0068, 0.0128, -0.0117], device='cuda:0'), grad: tensor([ 2.8871e-07, 1.6876e-06, 3.7439e-06, 2.4233e-06, -1.1921e-07, 5.8524e-06, -3.7879e-05, 1.0943e-06, 2.5809e-05, -2.8983e-06], device='cuda:0') 100 0.0001 changing lr epoch 190, time 251.57, cls_loss 0.0012 cls_loss_mapping 0.0035 cls_loss_causal 0.5013 re_mapping 0.0045 re_causal 0.0137 /// teacc 98.85 lr 0.00010000 Epoch 192, weight, value: tensor([[-0.1632, -0.0338, 0.1765, ..., 0.0393, -0.1597, -0.1261], [-0.1086, -0.1252, -0.0303, ..., -0.0943, -0.1320, -0.0592], [ 0.2029, -0.0594, -0.1734, ..., -0.1449, -0.1428, -0.1810], ..., [-0.0580, -0.0167, -0.1832, ..., -0.0779, 0.0856, 0.0135], [-0.1559, -0.0962, -0.0245, ..., -0.1525, -0.0227, -0.0251], [-0.0554, -0.0794, -0.0914, ..., 0.0911, 0.0250, -0.1506]], device='cuda:0'), grad: tensor([[ 1.5832e-08, 9.3132e-10, -1.6131e-06, ..., -5.2433e-07, 4.9360e-08, 6.7055e-08], [ 6.3330e-08, 6.5193e-09, 1.0245e-07, ..., 1.1176e-07, 6.3330e-08, 2.8871e-08], [-1.7416e-07, 5.5879e-09, 2.4028e-07, ..., 8.8476e-08, 2.8871e-08, 6.3330e-08], ..., [ 2.5146e-08, 7.4506e-09, 6.7987e-08, ..., 6.0257e-07, 2.8964e-07, 3.7253e-09], [ 3.0734e-08, 2.7940e-09, 1.3132e-07, ..., 1.1921e-07, 5.3085e-08, 3.2596e-08], [ 4.6566e-09, 9.3132e-10, 4.4703e-07, ..., -1.9576e-06, -1.0831e-06, 1.8626e-09]], device='cuda:0') Epoch 192, bias, value: tensor([ 0.0018, 0.0121, 0.0156, -0.0030, 0.0244, -0.0035, -0.0091, -0.0067, 0.0127, -0.0144], device='cuda:0'), grad: tensor([-2.5332e-06, -6.6962e-07, 6.6217e-07, 6.3423e-07, 4.8652e-06, 8.7731e-07, -1.6391e-07, 2.6189e-06, -1.0859e-06, -5.2303e-06], device='cuda:0') 100 0.0001 changing lr epoch 191, time 251.72, cls_loss 0.0010 cls_loss_mapping 0.0041 cls_loss_causal 0.5120 re_mapping 0.0048 re_causal 0.0145 /// teacc 98.99 lr 0.00010000 Epoch 193, weight, value: tensor([[-0.1637, -0.0339, 0.1749, ..., 0.0393, -0.1612, -0.1305], [-0.1088, -0.1260, -0.0304, ..., -0.0951, -0.1328, -0.0598], [ 0.2031, -0.0595, -0.1752, ..., -0.1458, -0.1437, -0.1824], ..., [-0.0581, -0.0168, -0.1841, ..., -0.0791, 0.0856, 0.0134], [-0.1561, -0.0960, -0.0247, ..., -0.1529, -0.0228, -0.0252], [-0.0555, -0.0797, -0.0916, ..., 0.0916, 0.0251, -0.1507]], device='cuda:0'), grad: tensor([[ 2.7195e-07, 2.6356e-07, -6.6236e-06, ..., -1.9316e-06, 5.4948e-08, 7.2643e-08], [ 1.3970e-07, 1.6019e-07, 2.5332e-07, ..., 2.8871e-08, 6.9849e-08, 8.1025e-08], [ 1.8671e-05, 4.1753e-05, 1.1707e-06, ..., 2.3190e-07, 2.9802e-08, 2.1420e-08], ..., [ 3.2783e-07, 7.2084e-07, 1.2107e-07, ..., 6.5193e-08, -5.4948e-08, 2.7940e-09], [-1.2349e-06, -9.7323e-07, 2.1216e-06, ..., 1.1334e-06, 7.3574e-08, 1.0058e-07], [ 3.1013e-07, 2.7288e-07, 1.8850e-06, ..., 1.3690e-07, -7.7300e-08, 3.7253e-09]], device='cuda:0') Epoch 193, bias, value: tensor([ 0.0007, 0.0121, 0.0152, -0.0041, 0.0239, -0.0027, -0.0074, -0.0069, 0.0130, -0.0138], device='cuda:0'), grad: tensor([-1.0349e-05, 2.1625e-06, 8.4281e-05, -8.1658e-05, 8.5868e-07, -2.3738e-05, 3.3915e-05, 1.9670e-06, -1.6168e-05, 8.6501e-06], device='cuda:0') 100 0.0001 changing lr epoch 192, time 251.27, cls_loss 0.0010 cls_loss_mapping 0.0035 cls_loss_causal 0.5266 re_mapping 0.0044 re_causal 0.0136 /// teacc 99.08 lr 0.00010000 Epoch 194, weight, value: tensor([[-0.1641, -0.0340, 0.1752, ..., 0.0392, -0.1614, -0.1306], [-0.1092, -0.1273, -0.0309, ..., -0.0954, -0.1337, -0.0600], [ 0.2032, -0.0603, -0.1773, ..., -0.1462, -0.1450, -0.1837], ..., [-0.0575, -0.0173, -0.1858, ..., -0.0794, 0.0860, 0.0135], [-0.1562, -0.0961, -0.0250, ..., -0.1533, -0.0228, -0.0251], [-0.0567, -0.0799, -0.0922, ..., 0.0917, 0.0253, -0.1508]], device='cuda:0'), grad: tensor([[-3.5390e-08, 9.3132e-10, 5.6066e-06, ..., -4.8894e-07, 1.6019e-07, 1.6158e-06], [ 2.5146e-08, 4.6566e-09, 9.4995e-08, ..., 5.0291e-08, 2.8871e-08, 3.9116e-08], [ 5.5879e-09, 3.7253e-09, 1.6857e-07, ..., 1.1548e-07, 1.3970e-08, 1.0245e-08], ..., [ 2.5146e-08, 5.5879e-09, 3.6322e-08, ..., 4.3772e-08, -1.4901e-08, 1.8626e-09], [ 3.5390e-08, 4.6566e-09, 3.1944e-07, ..., 5.2154e-08, 8.4750e-08, 9.8720e-08], [ 1.1269e-07, 1.3970e-08, 4.1444e-07, ..., 4.2934e-07, 1.5832e-08, 1.6764e-08]], device='cuda:0') Epoch 194, bias, value: tensor([ 0.0007, 0.0121, 0.0146, -0.0038, 0.0237, -0.0030, -0.0074, -0.0068, 0.0133, -0.0137], device='cuda:0'), grad: tensor([ 8.8587e-06, -3.5353e-06, 8.8476e-07, -2.5127e-06, -1.4128e-06, 1.9185e-06, -9.0599e-06, 2.2314e-06, 4.2934e-07, 2.1532e-06], device='cuda:0') 100 0.0001 changing lr epoch 193, time 251.80, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.4930 re_mapping 0.0043 re_causal 0.0140 /// teacc 99.11 lr 0.00010000 Epoch 195, weight, value: tensor([[-0.1646, -0.0338, 0.1756, ..., 0.0393, -0.1616, -0.1307], [-0.1103, -0.1285, -0.0308, ..., -0.0956, -0.1338, -0.0596], [ 0.2037, -0.0604, -0.1780, ..., -0.1466, -0.1453, -0.1841], ..., [-0.0568, -0.0174, -0.1891, ..., -0.0802, 0.0857, 0.0135], [-0.1562, -0.0963, -0.0250, ..., -0.1538, -0.0229, -0.0250], [-0.0572, -0.0799, -0.0930, ..., 0.0918, 0.0254, -0.1509]], device='cuda:0'), grad: tensor([[ 7.4506e-09, 1.8626e-09, -6.3702e-07, ..., -2.2445e-07, 5.6811e-08, 6.5193e-09], [ 5.2154e-08, 1.2107e-08, 2.6077e-08, ..., 1.3039e-07, 3.1292e-07, 2.7940e-08], [-4.6566e-09, 7.4506e-09, 2.1420e-08, ..., 8.9407e-08, 1.8813e-07, 1.6764e-08], ..., [-6.1002e-07, 1.3039e-08, 2.1420e-08, ..., -1.5022e-06, -3.6657e-06, -2.9709e-07], [ 3.9116e-08, 1.4529e-07, 3.9116e-08, ..., 2.7176e-06, 1.5963e-06, 6.5193e-09], [ 4.3493e-07, -1.3877e-07, 2.2445e-07, ..., -2.1793e-06, 5.9046e-07, 2.0023e-07]], device='cuda:0') Epoch 195, bias, value: tensor([ 0.0009, 0.0123, 0.0145, -0.0046, 0.0237, -0.0024, -0.0077, -0.0070, 0.0136, -0.0137], device='cuda:0'), grad: tensor([-6.6403e-07, 1.1064e-06, 6.2864e-07, -2.6915e-07, 2.7455e-06, 1.0561e-06, 3.9954e-07, -1.0505e-05, 6.8471e-06, -1.3895e-06], device='cuda:0') 100 0.0001 changing lr epoch 194, time 251.58, cls_loss 0.0010 cls_loss_mapping 0.0024 cls_loss_causal 0.4918 re_mapping 0.0044 re_causal 0.0139 /// teacc 99.14 lr 0.00010000 Epoch 196, weight, value: tensor([[-0.1657, -0.0338, 0.1761, ..., 0.0394, -0.1617, -0.1306], [-0.1110, -0.1331, -0.0310, ..., -0.0955, -0.1344, -0.0598], [ 0.2045, -0.0605, -0.1789, ..., -0.1472, -0.1455, -0.1842], ..., [-0.0572, -0.0170, -0.1906, ..., -0.0808, 0.0857, 0.0137], [-0.1565, -0.0987, -0.0234, ..., -0.1521, -0.0232, -0.0249], [-0.0577, -0.0806, -0.0962, ..., 0.0918, 0.0256, -0.1510]], device='cuda:0'), grad: tensor([[ 5.3085e-08, 1.6764e-08, 8.8476e-08, ..., 1.1642e-06, 2.8498e-07, 9.7789e-08], [ 2.9895e-07, 5.7742e-08, 3.7253e-09, ..., 5.8115e-07, 1.0338e-07, 1.3877e-07], [-1.4659e-06, -1.1362e-07, 1.5832e-08, ..., 7.5903e-07, 1.5739e-07, 1.2573e-07], ..., [ 1.0198e-06, 1.1083e-07, 2.9802e-08, ..., 9.0897e-07, 1.9558e-08, 2.7474e-07], [ 8.0094e-08, 3.5390e-08, 3.0734e-08, ..., 3.2596e-06, 8.4564e-07, 1.3411e-07], [ 1.3970e-08, 1.4901e-08, 2.7940e-08, ..., -1.1005e-05, -2.4512e-06, 4.2375e-07]], device='cuda:0') Epoch 196, bias, value: tensor([ 0.0011, 0.0123, 0.0147, -0.0046, 0.0236, -0.0022, -0.0083, -0.0071, 0.0153, -0.0139], device='cuda:0'), grad: tensor([ 2.8051e-06, 1.2983e-06, -1.4808e-07, -1.2573e-07, 5.1223e-07, 9.6187e-06, 3.8128e-06, 4.4517e-06, -3.0436e-06, -1.9237e-05], device='cuda:0') 100 0.0001 changing lr epoch 195, time 251.51, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.5240 re_mapping 0.0044 re_causal 0.0142 /// teacc 99.07 lr 0.00010000 Epoch 197, weight, value: tensor([[-0.1677, -0.0335, 0.1768, ..., 0.0398, -0.1618, -0.1307], [-0.1111, -0.1352, -0.0313, ..., -0.0960, -0.1353, -0.0600], [ 0.2051, -0.0604, -0.1810, ..., -0.1488, -0.1458, -0.1845], ..., [-0.0582, -0.0180, -0.1926, ..., -0.0813, 0.0858, 0.0139], [-0.1565, -0.0995, -0.0233, ..., -0.1525, -0.0237, -0.0253], [-0.0580, -0.0815, -0.0988, ..., 0.0918, 0.0258, -0.1510]], device='cuda:0'), grad: tensor([[ 3.2503e-07, 0.0000e+00, 6.5565e-07, ..., 2.4680e-07, 2.6263e-07, 1.6019e-07], [-1.9856e-06, 0.0000e+00, -2.4047e-06, ..., 1.5274e-07, 1.2014e-07, 1.9558e-08], [ 8.4098e-07, 9.3132e-10, 1.0505e-06, ..., 9.2201e-08, 3.9116e-08, 2.0489e-08], ..., [ 2.7101e-07, 9.3132e-10, 2.5798e-07, ..., 5.4482e-07, 1.5460e-07, -1.3039e-08], [ 5.2806e-07, 9.3132e-10, 7.2550e-07, ..., 2.8778e-07, 1.3597e-07, 4.8429e-08], [-8.7544e-08, 0.0000e+00, 1.1362e-07, ..., -2.7586e-06, -1.0859e-06, 4.0978e-08]], device='cuda:0') Epoch 197, bias, value: tensor([ 0.0015, 0.0123, 0.0145, -0.0041, 0.0236, -0.0036, -0.0072, -0.0072, 0.0156, -0.0141], device='cuda:0'), grad: tensor([ 8.1733e-06, -4.5955e-05, 1.9863e-05, 1.5618e-06, 3.2187e-06, 9.5740e-07, -1.2238e-06, 6.2510e-06, 1.2979e-05, -5.8822e-06], device='cuda:0') 100 0.0001 changing lr epoch 196, time 251.48, cls_loss 0.0008 cls_loss_mapping 0.0032 cls_loss_causal 0.5345 re_mapping 0.0045 re_causal 0.0148 /// teacc 99.15 lr 0.00010000 Epoch 198, weight, value: tensor([[-0.1688, -0.0340, 0.1767, ..., 0.0397, -0.1623, -0.1307], [-0.1112, -0.1368, -0.0312, ..., -0.0962, -0.1362, -0.0602], [ 0.2054, -0.0604, -0.1820, ..., -0.1494, -0.1461, -0.1847], ..., [-0.0585, -0.0194, -0.1950, ..., -0.0819, 0.0857, 0.0141], [-0.1567, -0.1001, -0.0233, ..., -0.1527, -0.0237, -0.0253], [-0.0591, -0.0819, -0.0992, ..., 0.0918, 0.0258, -0.1511]], device='cuda:0'), grad: tensor([[ 2.4214e-08, 0.0000e+00, -9.0338e-08, ..., 1.1176e-08, 4.2841e-08, 8.3819e-09], [ 9.5926e-08, 1.8626e-09, 2.7940e-09, ..., 1.5087e-07, 9.5926e-08, 2.7940e-09], [-9.8348e-07, 1.8626e-09, 9.3132e-09, ..., 1.0990e-07, 5.7742e-08, 1.8626e-09], ..., [ 4.2375e-07, 1.8626e-09, 9.3132e-10, ..., 1.8990e-06, 8.9686e-07, 0.0000e+00], [ 1.9930e-07, 0.0000e+00, 1.7695e-08, ..., 3.5577e-07, 1.8533e-07, 6.5193e-09], [ 8.3819e-09, 0.0000e+00, 3.9116e-08, ..., 2.5481e-06, 1.2824e-06, 1.8626e-09]], device='cuda:0') Epoch 198, bias, value: tensor([ 0.0012, 0.0124, 0.0144, -0.0038, 0.0238, -0.0034, -0.0073, -0.0073, 0.0159, -0.0142], device='cuda:0'), grad: tensor([ 1.3504e-07, 1.2107e-07, -1.3011e-06, 1.7695e-07, -1.8358e-05, 7.4133e-07, 9.5181e-07, 7.3016e-06, 1.6447e-06, 8.5756e-06], device='cuda:0') 100 0.0001 changing lr epoch 197, time 251.62, cls_loss 0.0006 cls_loss_mapping 0.0029 cls_loss_causal 0.5086 re_mapping 0.0045 re_causal 0.0146 /// teacc 99.06 lr 0.00010000 Epoch 199, weight, value: tensor([[-0.1691, -0.0340, 0.1769, ..., 0.0397, -0.1624, -0.1308], [-0.1113, -0.1382, -0.0312, ..., -0.0963, -0.1372, -0.0607], [ 0.2056, -0.0606, -0.1826, ..., -0.1506, -0.1466, -0.1851], ..., [-0.0585, -0.0197, -0.1965, ..., -0.0823, 0.0861, 0.0157], [-0.1568, -0.1023, -0.0234, ..., -0.1529, -0.0240, -0.0254], [-0.0591, -0.0820, -0.0995, ..., 0.0918, 0.0258, -0.1512]], device='cuda:0'), grad: tensor([[ 1.0617e-07, 9.3132e-10, 5.3085e-08, ..., -1.4063e-07, 2.7940e-09, 9.3132e-10], [ 2.5891e-07, 5.5879e-09, -3.0845e-06, ..., 4.8429e-08, 4.5635e-08, 5.5879e-09], [-3.8855e-06, 9.3132e-10, 8.9128e-07, ..., 6.5193e-09, 9.3132e-10, 0.0000e+00], ..., [ 2.5108e-06, -5.5879e-09, 1.8198e-06, ..., 2.7008e-08, -1.0896e-07, -1.6764e-08], [ 1.6671e-07, 4.2841e-08, 2.7008e-08, ..., 2.4214e-08, 3.0734e-08, 1.7695e-08], [ 7.4506e-09, 1.8626e-09, 1.1828e-07, ..., 2.7940e-09, -4.0047e-08, 1.8626e-09]], device='cuda:0') Epoch 199, bias, value: tensor([ 0.0013, 0.0126, 0.0141, -0.0039, 0.0238, -0.0033, -0.0074, -0.0074, 0.0157, -0.0143], device='cuda:0'), grad: tensor([ 9.2946e-07, -1.3046e-05, -2.9467e-06, 1.2536e-06, 2.2911e-07, -5.2154e-07, 5.4203e-07, 1.2249e-05, 1.1371e-06, 1.5274e-07], device='cuda:0') 100 0.0001 changing lr epoch 198, time 251.80, cls_loss 0.0007 cls_loss_mapping 0.0028 cls_loss_causal 0.4991 re_mapping 0.0044 re_causal 0.0142 /// teacc 99.13 lr 0.00010000 Epoch 200, weight, value: tensor([[-0.1694, -0.0341, 0.1771, ..., 0.0398, -0.1625, -0.1308], [-0.1116, -0.1401, -0.0319, ..., -0.0962, -0.1382, -0.0617], [ 0.2061, -0.0608, -0.1831, ..., -0.1510, -0.1469, -0.1852], ..., [-0.0587, -0.0196, -0.1975, ..., -0.0826, 0.0865, 0.0159], [-0.1571, -0.1038, -0.0234, ..., -0.1532, -0.0243, -0.0254], [-0.0592, -0.0824, -0.0997, ..., 0.0919, 0.0259, -0.1513]], device='cuda:0'), grad: tensor([[ 2.3823e-06, 7.4506e-09, 4.3772e-08, ..., 9.3132e-09, 3.8184e-08, 1.3970e-08], [ 3.5111e-07, 4.5635e-08, 2.8871e-08, ..., 5.0291e-08, 8.4657e-07, 1.3039e-08], [-1.0416e-05, 1.1828e-07, 9.3132e-09, ..., 1.8626e-09, 2.5146e-08, 2.7940e-09], ..., [ 6.0908e-07, 6.5193e-08, 0.0000e+00, ..., -5.1223e-08, -1.1018e-06, 0.0000e+00], [ 5.1372e-06, 1.9558e-08, 2.3283e-08, ..., 4.9360e-08, 3.8184e-08, 8.3819e-09], [ 1.9185e-07, 1.8626e-08, 1.8626e-09, ..., -1.8533e-07, 1.1642e-07, 9.3132e-10]], device='cuda:0') Epoch 200, bias, value: tensor([ 0.0014, 0.0126, 0.0141, -0.0037, 0.0237, -0.0035, -0.0073, -0.0075, 0.0157, -0.0142], device='cuda:0'), grad: tensor([ 6.5416e-06, 5.7071e-06, -2.7671e-05, 3.5539e-06, 3.7998e-07, 1.5814e-06, -1.1921e-07, -4.4927e-06, 1.3724e-05, 7.7859e-07], device='cuda:0') 100 0.0001 changing lr epoch 199, time 251.56, cls_loss 0.0010 cls_loss_mapping 0.0041 cls_loss_causal 0.4995 re_mapping 0.0044 re_causal 0.0136 /// teacc 98.96 lr 0.00010000 Epoch 201, weight, value: tensor([[-0.1709, -0.0344, 0.1774, ..., 0.0399, -0.1627, -0.1310], [-0.1133, -0.1425, -0.0333, ..., -0.0966, -0.1395, -0.0635], [ 0.2073, -0.0610, -0.1823, ..., -0.1532, -0.1473, -0.1856], ..., [-0.0593, -0.0203, -0.1985, ..., -0.0836, 0.0872, 0.0159], [-0.1575, -0.1036, -0.0233, ..., -0.1533, -0.0240, -0.0254], [-0.0591, -0.0830, -0.1003, ..., 0.0921, 0.0262, -0.1513]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.8626e-09, -3.6322e-08, ..., 1.5832e-08, 2.7940e-08, 7.4506e-09], [ 3.3528e-08, 7.4506e-09, -0.0000e+00, ..., 4.7497e-07, 1.9744e-07, 2.7940e-09], [-5.0291e-08, 4.6566e-09, 5.5879e-09, ..., 4.3772e-08, 2.7008e-08, 2.7940e-09], ..., [ 3.4459e-08, 2.3283e-08, 4.6566e-09, ..., 1.5181e-07, 3.8184e-08, 2.7940e-09], [ 4.6566e-09, 5.5879e-09, 1.6764e-08, ..., 4.1910e-08, 4.8429e-08, 2.1420e-08], [ 4.6566e-09, 5.5879e-09, 3.1665e-08, ..., 4.5970e-06, 1.9027e-06, 6.5193e-09]], device='cuda:0') Epoch 201, bias, value: tensor([ 0.0014, 0.0122, 0.0145, -0.0030, 0.0235, -0.0042, -0.0071, -0.0073, 0.0163, -0.0141], device='cuda:0'), grad: tensor([ 8.7544e-08, 1.1278e-06, 9.4064e-08, 2.0638e-06, -1.1101e-05, -2.4512e-06, 5.7369e-07, 4.5262e-07, -1.6671e-07, 9.3356e-06], device='cuda:0') 100 0.0001 changing lr epoch 200, time 251.42, cls_loss 0.0009 cls_loss_mapping 0.0031 cls_loss_causal 0.5043 re_mapping 0.0040 re_causal 0.0128 /// teacc 99.06 lr 0.00010000 Epoch 202, weight, value: tensor([[-0.1716, -0.0345, 0.1778, ..., 0.0401, -0.1628, -0.1311], [-0.1135, -0.1436, -0.0336, ..., -0.0971, -0.1422, -0.0644], [ 0.2076, -0.0611, -0.1844, ..., -0.1562, -0.1483, -0.1863], ..., [-0.0596, -0.0207, -0.1997, ..., -0.0849, 0.0878, 0.0161], [-0.1577, -0.1036, -0.0235, ..., -0.1538, -0.0241, -0.0255], [-0.0594, -0.0832, -0.1011, ..., 0.0924, 0.0269, -0.1515]], device='cuda:0'), grad: tensor([[ 1.8626e-08, 1.3039e-08, -2.7381e-07, ..., -1.3411e-07, 4.3772e-08, 6.4261e-08], [ 1.9558e-07, 3.3528e-08, 4.6566e-08, ..., 5.8673e-08, 1.8626e-07, 4.3772e-08], [ 1.7975e-07, 7.4506e-08, 4.2841e-08, ..., 1.3039e-08, 8.8476e-08, 2.9802e-08], ..., [-2.4773e-07, 7.3574e-08, 1.0245e-08, ..., 1.0245e-07, -1.4994e-07, -1.5832e-08], [ 1.6019e-07, 1.6019e-07, 6.5193e-08, ..., 5.0291e-08, 5.1223e-08, 3.3528e-08], [ 1.9558e-08, 1.2107e-08, 1.5181e-07, ..., -3.1479e-07, -3.2037e-07, 3.7253e-09]], device='cuda:0') Epoch 202, bias, value: tensor([ 0.0017, 0.0127, 0.0138, -0.0031, 0.0233, -0.0040, -0.0076, -0.0078, 0.0162, -0.0138], device='cuda:0'), grad: tensor([ 7.4506e-08, 2.4512e-06, 2.5257e-06, 1.6745e-06, 1.0664e-06, 1.9651e-06, 4.3679e-07, -2.3320e-06, -9.6858e-06, 1.8021e-06], device='cuda:0') 100 0.0001 changing lr epoch 201, time 251.90, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.4989 re_mapping 0.0043 re_causal 0.0139 /// teacc 99.04 lr 0.00010000 Epoch 203, weight, value: tensor([[-0.1747, -0.0346, 0.1780, ..., 0.0402, -0.1629, -0.1312], [-0.1141, -0.1453, -0.0337, ..., -0.0975, -0.1440, -0.0646], [ 0.2088, -0.0611, -0.1849, ..., -0.1584, -0.1497, -0.1853], ..., [-0.0606, -0.0209, -0.2003, ..., -0.0862, 0.0884, 0.0160], [-0.1580, -0.1039, -0.0237, ..., -0.1543, -0.0247, -0.0255], [-0.0596, -0.0834, -0.1015, ..., 0.0925, 0.0269, -0.1515]], device='cuda:0'), grad: tensor([[ 2.0489e-08, 2.7940e-09, -2.0768e-07, ..., -2.0768e-07, 4.5635e-08, 2.4214e-08], [ 1.9837e-07, 1.7695e-08, 2.3283e-08, ..., 6.5193e-08, 6.6124e-08, 1.8626e-09], [-1.1176e-08, 1.3970e-08, 1.5832e-08, ..., 5.3085e-08, 2.1420e-08, 1.8626e-09], ..., [ 2.5146e-07, 9.3132e-08, 5.0291e-08, ..., 7.1712e-08, -3.0734e-08, 0.0000e+00], [ 2.5798e-07, 2.1420e-08, 7.3574e-08, ..., 3.3528e-08, 3.5018e-07, 1.3970e-08], [ 7.1712e-08, 6.5193e-09, 9.4064e-08, ..., 7.8231e-08, 8.3819e-08, 9.3132e-10]], device='cuda:0') Epoch 203, bias, value: tensor([ 0.0017, 0.0123, 0.0140, -0.0032, 0.0235, -0.0041, -0.0074, -0.0077, 0.0161, -0.0138], device='cuda:0'), grad: tensor([-3.2783e-07, -1.4849e-05, 6.0536e-07, -6.8247e-06, 8.5589e-07, 6.3702e-07, 8.7824e-07, 4.1649e-06, 1.2673e-05, 2.1346e-06], device='cuda:0') 100 0.0001 changing lr epoch 202, time 251.59, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.4875 re_mapping 0.0043 re_causal 0.0129 /// teacc 99.01 lr 0.00010000 Epoch 204, weight, value: tensor([[-0.1753, -0.0346, 0.1781, ..., 0.0388, -0.1633, -0.1313], [-0.1144, -0.1459, -0.0337, ..., -0.0982, -0.1454, -0.0649], [ 0.2114, -0.0612, -0.1856, ..., -0.1587, -0.1493, -0.1856], ..., [-0.0637, -0.0211, -0.2012, ..., -0.0865, 0.0885, 0.0162], [-0.1582, -0.1041, -0.0239, ..., -0.1547, -0.0252, -0.0257], [-0.0602, -0.0835, -0.1018, ..., 0.0931, 0.0269, -0.1516]], device='cuda:0'), grad: tensor([[ 5.5879e-09, 9.3132e-10, -4.5076e-07, ..., -3.2689e-07, 2.4214e-08, 1.2107e-08], [ 2.4214e-08, 2.7940e-09, -2.2724e-07, ..., 1.6205e-07, 5.9605e-08, -3.4459e-08], [-2.5332e-07, 2.7940e-09, 5.2154e-08, ..., 1.0990e-07, 3.6322e-08, 7.4506e-09], ..., [ 3.3528e-08, 3.7253e-09, 2.4214e-08, ..., 1.0617e-07, -1.4901e-08, 2.7940e-09], [ 2.0582e-07, 4.6566e-09, 1.1548e-07, ..., 1.1176e-07, 1.2480e-07, 4.3772e-08], [ 1.1176e-08, 3.7253e-09, 4.1071e-07, ..., 2.0489e-06, 9.0990e-07, 2.7940e-09]], device='cuda:0') Epoch 204, bias, value: tensor([ 0.0005, 0.0123, 0.0155, -0.0025, 0.0236, -0.0046, -0.0070, -0.0084, 0.0160, -0.0135], device='cuda:0'), grad: tensor([-5.4669e-07, -6.5845e-07, 1.2014e-07, -3.5241e-06, -5.2005e-06, 3.4235e-06, 6.8918e-08, 3.0641e-07, 1.0803e-06, 4.9099e-06], device='cuda:0') 100 0.0001 changing lr epoch 203, time 251.37, cls_loss 0.0011 cls_loss_mapping 0.0034 cls_loss_causal 0.5562 re_mapping 0.0041 re_causal 0.0136 /// teacc 98.95 lr 0.00010000 Epoch 205, weight, value: tensor([[-0.1769, -0.0347, 0.1790, ..., 0.0386, -0.1634, -0.1313], [-0.1147, -0.1481, -0.0336, ..., -0.0985, -0.1473, -0.0651], [ 0.2118, -0.0613, -0.1864, ..., -0.1591, -0.1501, -0.1859], ..., [-0.0638, -0.0216, -0.2030, ..., -0.0870, 0.0892, 0.0167], [-0.1585, -0.1044, -0.0239, ..., -0.1547, -0.0254, -0.0259], [-0.0597, -0.0844, -0.1044, ..., 0.0948, 0.0287, -0.1518]], device='cuda:0'), grad: tensor([[ 3.9116e-08, 9.3132e-10, -1.1697e-06, ..., -3.7439e-07, -1.3597e-07, 1.1176e-08], [ 7.6368e-08, 9.3132e-10, 1.9558e-08, ..., 2.7940e-08, 2.1420e-08, 2.7940e-09], [-4.0047e-07, 0.0000e+00, 2.3283e-08, ..., 8.5682e-08, 8.3819e-09, 1.5832e-08], ..., [ 1.4901e-07, 9.3132e-10, 9.3132e-09, ..., 5.2154e-08, 1.1176e-08, -9.3132e-10], [ 4.4703e-08, 6.5193e-09, 3.1665e-08, ..., 1.1083e-07, -3.0734e-08, 8.3819e-09], [ 1.3039e-08, 1.8626e-09, 1.4529e-07, ..., -8.0094e-08, 1.8626e-08, 3.7253e-09]], device='cuda:0') Epoch 205, bias, value: tensor([ 0.0004, 0.0125, 0.0152, -0.0021, 0.0223, -0.0048, -0.0075, -0.0083, 0.0160, -0.0121], device='cuda:0'), grad: tensor([-1.4370e-06, -1.0179e-06, -4.4052e-07, 4.1816e-07, -3.1386e-07, -3.0454e-07, 1.4985e-06, 1.3430e-06, 1.7509e-07, 7.7300e-08], device='cuda:0') 100 0.0001 changing lr epoch 204, time 251.30, cls_loss 0.0013 cls_loss_mapping 0.0034 cls_loss_causal 0.5219 re_mapping 0.0042 re_causal 0.0135 /// teacc 99.04 lr 0.00010000 Epoch 206, weight, value: tensor([[-0.1776, -0.0348, 0.1794, ..., 0.0387, -0.1637, -0.1314], [-0.1155, -0.1505, -0.0338, ..., -0.0982, -0.1510, -0.0658], [ 0.2121, -0.0616, -0.1873, ..., -0.1593, -0.1509, -0.1861], ..., [-0.0635, -0.0216, -0.2039, ..., -0.0873, 0.0925, 0.0182], [-0.1586, -0.1046, -0.0241, ..., -0.1554, -0.0258, -0.0259], [-0.0614, -0.0852, -0.1052, ..., 0.0939, 0.0270, -0.1519]], device='cuda:0'), grad: tensor([[ 4.9360e-08, 0.0000e+00, 2.8778e-07, ..., -9.3132e-10, 3.6508e-07, 6.7987e-08], [ 9.4064e-08, 9.3132e-10, 4.6566e-08, ..., 1.2107e-08, 9.7789e-08, 1.4901e-08], [-1.2545e-06, 0.0000e+00, 1.5181e-07, ..., 3.7253e-09, 1.6671e-07, 2.8871e-08], ..., [ 9.3970e-07, 9.3132e-10, 2.0489e-08, ..., 1.4901e-08, -2.7940e-09, 3.7253e-09], [ 3.0734e-08, 0.0000e+00, 2.5332e-07, ..., 5.6811e-08, 3.7346e-07, 7.7300e-08], [ 4.6566e-09, 0.0000e+00, 1.7695e-07, ..., -3.7253e-07, 1.1083e-07, 3.5390e-08]], device='cuda:0') Epoch 206, bias, value: tensor([ 0.0005, 0.0137, 0.0146, -0.0020, 0.0233, -0.0018, -0.0104, -0.0083, 0.0163, -0.0139], device='cuda:0'), grad: tensor([ 1.2182e-06, -8.4750e-08, -1.4249e-06, 2.5332e-07, 1.2452e-06, 8.4788e-06, -1.1019e-05, 1.6587e-06, -2.2445e-07, -1.6019e-07], device='cuda:0') 100 0.0001 changing lr epoch 205, time 251.46, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.4995 re_mapping 0.0042 re_causal 0.0135 /// teacc 99.09 lr 0.00010000 Epoch 207, weight, value: tensor([[-0.1785, -0.0349, 0.1795, ..., 0.0387, -0.1640, -0.1315], [-0.1162, -0.1516, -0.0339, ..., -0.0989, -0.1534, -0.0659], [ 0.2105, -0.0648, -0.1877, ..., -0.1595, -0.1533, -0.1860], ..., [-0.0633, -0.0203, -0.2045, ..., -0.0882, 0.0935, 0.0183], [-0.1589, -0.1049, -0.0243, ..., -0.1562, -0.0264, -0.0260], [-0.0615, -0.0853, -0.1054, ..., 0.0940, 0.0271, -0.1519]], device='cuda:0'), grad: tensor([[2.7940e-08, 1.0245e-08, 5.4948e-08, ..., 9.3132e-10, 4.2841e-08, 5.4017e-08], [5.8673e-08, 3.2596e-08, 1.0338e-07, ..., 7.4506e-09, 6.8918e-08, 1.1642e-07], [2.5146e-08, 6.1467e-08, 6.1467e-08, ..., 8.3819e-09, 5.1223e-08, 8.3819e-08], ..., [3.0082e-07, 2.2352e-07, 3.7253e-09, ..., 2.7940e-09, 3.4459e-08, 1.9185e-07], [3.0734e-08, 1.6764e-08, 1.1548e-07, ..., 1.1083e-07, 9.1270e-08, 1.0896e-07], [9.3132e-09, 5.5879e-09, 1.7695e-08, ..., 2.0210e-07, 6.7987e-08, 1.4901e-08]], device='cuda:0') Epoch 207, bias, value: tensor([ 0.0005, 0.0135, 0.0125, -0.0003, 0.0232, -0.0016, -0.0106, -0.0081, 0.0163, -0.0137], device='cuda:0'), grad: tensor([ 2.6822e-07, 4.8336e-07, 4.1630e-07, -2.5816e-06, -1.6671e-07, 6.1654e-07, -1.7826e-06, 1.3085e-06, 6.8173e-07, 7.6182e-07], device='cuda:0') 100 0.0001 changing lr epoch 206, time 251.34, cls_loss 0.0008 cls_loss_mapping 0.0025 cls_loss_causal 0.4754 re_mapping 0.0040 re_causal 0.0126 /// teacc 99.11 lr 0.00010000 Epoch 208, weight, value: tensor([[-0.1787, -0.0349, 0.1802, ..., 0.0381, -0.1639, -0.1315], [-0.1165, -0.1526, -0.0352, ..., -0.0997, -0.1539, -0.0660], [ 0.2108, -0.0648, -0.1884, ..., -0.1597, -0.1537, -0.1858], ..., [-0.0633, -0.0205, -0.2058, ..., -0.0886, 0.0936, 0.0183], [-0.1595, -0.1051, -0.0244, ..., -0.1565, -0.0268, -0.0261], [-0.0617, -0.0854, -0.1060, ..., 0.0942, 0.0270, -0.1520]], device='cuda:0'), grad: tensor([[ 1.0245e-08, -2.7940e-09, -1.2673e-05, ..., -6.0052e-06, 2.0489e-08, 6.5193e-09], [ 2.3749e-07, 3.7253e-09, 9.2201e-08, ..., 5.2154e-08, 8.5682e-08, 2.7940e-09], [-3.7998e-07, 2.7940e-09, 7.2271e-07, ..., 3.5297e-07, 2.0489e-08, 2.7940e-09], ..., [ 5.9605e-08, 5.5879e-09, 2.3469e-07, ..., 1.2573e-07, -9.3132e-08, 0.0000e+00], [ 5.4017e-08, 6.5193e-09, 4.4443e-06, ..., 2.0750e-06, 3.7625e-07, 1.1083e-07], [ 1.3039e-08, 9.3132e-09, 5.3830e-06, ..., 2.5127e-06, -3.2596e-08, 9.3132e-10]], device='cuda:0') Epoch 208, bias, value: tensor([ 0.0002, 0.0136, 0.0125, -0.0003, 0.0234, -0.0016, -0.0106, -0.0082, 0.0161, -0.0136], device='cuda:0'), grad: tensor([-2.9102e-05, 1.2713e-06, 9.1083e-07, 9.6578e-07, 2.1141e-07, 3.4831e-07, 1.9558e-06, 8.8476e-08, 1.0937e-05, 1.2390e-05], device='cuda:0') 100 0.0001 changing lr epoch 207, time 251.34, cls_loss 0.0008 cls_loss_mapping 0.0031 cls_loss_causal 0.5189 re_mapping 0.0040 re_causal 0.0132 /// teacc 99.06 lr 0.00010000 Epoch 209, weight, value: tensor([[-0.1793, -0.0349, 0.1810, ..., 0.0383, -0.1640, -0.1316], [-0.1182, -0.1536, -0.0362, ..., -0.1003, -0.1550, -0.0662], [ 0.2117, -0.0648, -0.1894, ..., -0.1600, -0.1541, -0.1861], ..., [-0.0636, -0.0206, -0.2071, ..., -0.0891, 0.0939, 0.0195], [-0.1599, -0.1053, -0.0246, ..., -0.1571, -0.0273, -0.0262], [-0.0616, -0.0855, -0.1065, ..., 0.0942, 0.0270, -0.1520]], device='cuda:0'), grad: tensor([[ 6.5193e-09, 3.7253e-09, -2.3376e-07, ..., -1.9651e-07, 3.4831e-07, 2.0862e-07], [ 5.8673e-08, 3.9116e-08, 1.8906e-07, ..., 7.8697e-07, 9.3970e-07, 1.1269e-07], [ 1.2107e-08, 2.1420e-08, 6.7055e-08, ..., 6.5193e-09, 1.0151e-07, 3.6322e-08], ..., [ 1.1362e-07, 7.5437e-08, 5.5879e-09, ..., 2.8498e-07, -8.1956e-07, 5.7742e-08], [ 6.5193e-08, 4.4703e-08, 2.5947e-06, ..., 3.3528e-08, 1.8133e-06, 1.2955e-06], [ 1.6764e-08, 1.3039e-08, 2.2259e-07, ..., -1.5767e-06, -4.1537e-07, 2.5146e-08]], device='cuda:0') Epoch 209, bias, value: tensor([ 0.0006, 0.0135, 0.0129, -0.0002, 0.0235, -0.0018, -0.0106, -0.0083, 0.0158, -0.0136], device='cuda:0'), grad: tensor([ 2.9895e-07, 5.6550e-06, 6.0070e-07, -3.4496e-06, 1.2210e-06, 2.0817e-05, -2.4840e-05, -3.1963e-06, 6.5081e-06, -3.6545e-06], device='cuda:0') 100 0.0001 changing lr epoch 208, time 251.28, cls_loss 0.0010 cls_loss_mapping 0.0035 cls_loss_causal 0.4957 re_mapping 0.0042 re_causal 0.0126 /// teacc 99.09 lr 0.00010000 Epoch 210, weight, value: tensor([[-0.1800, -0.0349, 0.1818, ..., 0.0385, -0.1639, -0.1315], [-0.1201, -0.1547, -0.0366, ..., -0.1016, -0.1578, -0.0665], [ 0.2129, -0.0649, -0.1903, ..., -0.1602, -0.1546, -0.1863], ..., [-0.0639, -0.0207, -0.2082, ..., -0.0898, 0.0951, 0.0202], [-0.1603, -0.1054, -0.0248, ..., -0.1584, -0.0279, -0.0263], [-0.0622, -0.0856, -0.1073, ..., 0.0950, 0.0282, -0.1521]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -9.6858e-08, ..., -5.3085e-08, 2.7940e-08, 2.0489e-08], [ 1.2107e-08, 1.8626e-09, 3.7253e-09, ..., 5.3085e-08, 1.2489e-06, 1.9446e-06], [-1.1548e-07, 9.3132e-10, 2.6077e-08, ..., 3.9116e-08, 1.2107e-08, 5.5879e-09], ..., [ 1.3970e-08, 2.7940e-09, 9.3132e-10, ..., 7.0781e-08, 6.4261e-08, 5.6811e-08], [ 9.6858e-08, 9.3132e-10, 2.9802e-08, ..., 9.0338e-08, 1.7351e-06, 2.6822e-06], [ 1.8626e-09, 9.3132e-10, 5.4948e-08, ..., -5.2620e-07, -2.8685e-07, 7.4506e-09]], device='cuda:0') Epoch 210, bias, value: tensor([ 0.0012, 0.0129, 0.0140, 0.0007, 0.0223, -0.0020, -0.0106, -0.0081, 0.0154, -0.0126], device='cuda:0'), grad: tensor([-6.2399e-08, 1.0230e-05, -2.2352e-08, 2.6636e-07, 3.4180e-07, -4.0084e-05, 1.5572e-05, 4.6846e-07, 1.4298e-05, -1.0300e-06], device='cuda:0') 100 0.0001 changing lr epoch 209, time 249.11, cls_loss 0.0009 cls_loss_mapping 0.0024 cls_loss_causal 0.5007 re_mapping 0.0040 re_causal 0.0129 /// teacc 99.06 lr 0.00010000 Epoch 211, weight, value: tensor([[-0.1804, -0.0349, 0.1821, ..., 0.0386, -0.1642, -0.1316], [-0.1202, -0.1552, -0.0384, ..., -0.1021, -0.1606, -0.0690], [ 0.2121, -0.0649, -0.1926, ..., -0.1605, -0.1587, -0.1884], ..., [-0.0640, -0.0208, -0.2090, ..., -0.0901, 0.0954, 0.0200], [-0.1608, -0.1056, -0.0252, ..., -0.1591, -0.0298, -0.0266], [-0.0623, -0.0857, -0.1079, ..., 0.0950, 0.0282, -0.1522]], device='cuda:0'), grad: tensor([[ 8.3726e-07, 3.6322e-08, 2.1104e-06, ..., -3.7253e-08, 1.5954e-06, 1.7695e-08], [ 1.2945e-07, 1.5553e-07, 4.5635e-08, ..., 1.6764e-08, 2.9244e-07, 4.1910e-08], [-2.1644e-06, 3.6322e-08, -3.5111e-07, ..., 4.6566e-09, 9.3132e-08, 1.3039e-08], ..., [ 2.0210e-07, 1.5087e-07, 3.5390e-08, ..., 3.0734e-08, 3.1665e-08, 3.7253e-08], [ 5.3179e-07, 9.4064e-08, 4.2003e-07, ..., 1.2107e-08, 3.4180e-07, 5.0291e-08], [ 4.6566e-08, 4.9360e-08, 5.6811e-08, ..., 1.2759e-07, 1.7788e-07, 1.3970e-08]], device='cuda:0') Epoch 211, bias, value: tensor([ 0.0013, 0.0124, 0.0124, 0.0006, 0.0223, -0.0016, -0.0106, -0.0081, 0.0149, -0.0125], device='cuda:0'), grad: tensor([ 7.5810e-06, 2.1458e-06, -3.5651e-06, -1.0140e-05, 8.8476e-08, 4.7162e-06, -7.1861e-06, 2.1644e-06, 2.1905e-06, 1.9744e-06], device='cuda:0') 100 0.0001 changing lr epoch 210, time 246.70, cls_loss 0.0010 cls_loss_mapping 0.0028 cls_loss_causal 0.4842 re_mapping 0.0042 re_causal 0.0129 /// teacc 99.19 lr 0.00010000 Epoch 212, weight, value: tensor([[-0.1810, -0.0350, 0.1823, ..., 0.0386, -0.1648, -0.1318], [-0.1226, -0.1559, -0.0385, ..., -0.1020, -0.1616, -0.0691], [ 0.2137, -0.0649, -0.1937, ..., -0.1611, -0.1598, -0.1892], ..., [-0.0652, -0.0210, -0.2102, ..., -0.0906, 0.0957, 0.0202], [-0.1613, -0.1057, -0.0247, ..., -0.1597, -0.0277, -0.0268], [-0.0623, -0.0858, -0.1082, ..., 0.0949, 0.0281, -0.1523]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.8626e-09, 3.7253e-09, ..., 2.7940e-09, 2.7940e-08, 3.7253e-09], [ 2.7940e-09, 3.7253e-09, -1.8626e-09, ..., 5.5879e-09, 6.3330e-08, 4.6566e-09], [-0.0000e+00, 1.8626e-09, 1.8626e-09, ..., 1.8626e-09, 1.4901e-08, 9.3132e-10], ..., [ 3.7253e-09, 5.5879e-09, 1.8626e-09, ..., 9.3132e-09, -5.0943e-07, 6.5193e-09], [ 2.7940e-09, 8.3819e-09, -6.5193e-09, ..., 3.1665e-08, 4.5635e-08, 1.3970e-08], [ 1.8626e-09, 2.7940e-09, 2.7940e-09, ..., -6.2399e-08, 3.7998e-07, 8.3819e-09]], device='cuda:0') Epoch 212, bias, value: tensor([ 0.0012, 0.0121, 0.0140, 0.0006, 0.0224, -0.0018, -0.0106, -0.0086, 0.0165, -0.0127], device='cuda:0'), grad: tensor([ 1.7136e-07, -2.9467e-06, 1.9483e-06, 3.7402e-06, 9.0338e-08, -4.7535e-06, 4.1164e-07, -6.7055e-07, 4.0606e-07, 1.5888e-06], device='cuda:0') 100 0.0001 changing lr epoch 211, time 246.81, cls_loss 0.0007 cls_loss_mapping 0.0027 cls_loss_causal 0.4968 re_mapping 0.0044 re_causal 0.0138 /// teacc 99.08 lr 0.00010000 Epoch 213, weight, value: tensor([[-0.1816, -0.0350, 0.1825, ..., 0.0387, -0.1652, -0.1320], [-0.1231, -0.1565, -0.0385, ..., -0.1022, -0.1629, -0.0693], [ 0.2143, -0.0649, -0.1946, ..., -0.1613, -0.1606, -0.1896], ..., [-0.0655, -0.0213, -0.2109, ..., -0.0906, 0.0962, 0.0206], [-0.1620, -0.1063, -0.0250, ..., -0.1602, -0.0283, -0.0270], [-0.0625, -0.0859, -0.1086, ..., 0.0949, 0.0281, -0.1524]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.2538e-07, ..., -1.9185e-07, 5.0291e-08, 5.5879e-09], [ 1.8626e-09, 1.8626e-09, 6.5193e-09, ..., -2.5146e-08, 2.1886e-07, 9.3132e-10], [-2.7940e-09, 0.0000e+00, 1.4901e-08, ..., 9.3132e-09, 2.9802e-08, 1.8626e-09], ..., [ 3.7253e-09, 9.3132e-10, 1.8626e-09, ..., 3.5390e-08, -3.9861e-07, 0.0000e+00], [ 1.8626e-09, 9.3132e-10, 2.5146e-08, ..., 5.8673e-08, 1.3225e-07, 3.7253e-09], [ 9.3132e-09, 8.3819e-09, 2.2631e-07, ..., 1.6019e-07, 8.1956e-08, 9.3132e-10]], device='cuda:0') Epoch 213, bias, value: tensor([ 0.0013, 0.0117, 0.0142, 0.0007, 0.0224, -0.0019, -0.0105, -0.0078, 0.0161, -0.0130], device='cuda:0'), grad: tensor([-1.6671e-07, 3.7625e-07, 2.0582e-07, 5.8115e-07, 1.2387e-07, -2.8685e-07, -8.0839e-07, -1.0254e-06, 1.8347e-07, 8.2608e-07], device='cuda:0') 100 0.0001 changing lr epoch 212, time 246.11, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.5036 re_mapping 0.0041 re_causal 0.0132 /// teacc 99.06 lr 0.00010000 Epoch 214, weight, value: tensor([[-0.1825, -0.0351, 0.1829, ..., 0.0390, -0.1657, -0.1321], [-0.1236, -0.1575, -0.0386, ..., -0.1025, -0.1640, -0.0694], [ 0.2148, -0.0649, -0.1953, ..., -0.1616, -0.1619, -0.1896], ..., [-0.0659, -0.0219, -0.2121, ..., -0.0933, 0.0958, 0.0207], [-0.1628, -0.1067, -0.0249, ..., -0.1605, -0.0283, -0.0272], [-0.0630, -0.0860, -0.1096, ..., 0.0950, 0.0283, -0.1524]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.8405e-07, ..., -2.4401e-07, 0.0000e+00, 0.0000e+00], [ 3.7253e-09, 9.3132e-10, 3.7253e-09, ..., 5.5879e-09, 5.5879e-09, 0.0000e+00], [-2.2352e-08, 0.0000e+00, 3.7253e-09, ..., 3.7253e-09, 1.8626e-09, 0.0000e+00], ..., [ 1.8626e-08, 2.7940e-09, 4.6566e-09, ..., 6.5193e-09, -1.0245e-08, 9.3132e-10], [ 1.8626e-09, 1.8626e-09, 5.5879e-09, ..., 4.6566e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 1.8626e-09, 2.0862e-07, ..., 2.0396e-07, 8.3819e-09, 0.0000e+00]], device='cuda:0') Epoch 214, bias, value: tensor([ 0.0014, 0.0114, 0.0141, 0.0006, 0.0225, -0.0018, -0.0105, -0.0078, 0.0162, -0.0130], device='cuda:0'), grad: tensor([-4.6194e-07, -9.6858e-07, 4.1164e-07, 5.4017e-08, 7.4506e-09, 5.8673e-08, 2.9244e-07, 5.5134e-07, -5.4855e-07, 5.9325e-07], device='cuda:0') 100 0.0001 changing lr epoch 213, time 246.69, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.5134 re_mapping 0.0040 re_causal 0.0129 /// teacc 99.06 lr 0.00010000 Epoch 215, weight, value: tensor([[-0.1828, -0.0352, 0.1831, ..., 0.0388, -0.1660, -0.1322], [-0.1240, -0.1582, -0.0386, ..., -0.1024, -0.1651, -0.0695], [ 0.2149, -0.0649, -0.1959, ..., -0.1619, -0.1629, -0.1902], ..., [-0.0655, -0.0224, -0.2127, ..., -0.0937, 0.0964, 0.0212], [-0.1631, -0.1071, -0.0252, ..., -0.1611, -0.0286, -0.0273], [-0.0633, -0.0862, -0.1099, ..., 0.0951, 0.0282, -0.1524]], device='cuda:0'), grad: tensor([[ 2.5146e-08, 5.6811e-08, 2.7940e-08, ..., 1.2107e-08, 9.4064e-08, 9.3132e-10], [ 1.3039e-08, 8.3819e-09, 5.5879e-09, ..., 4.9360e-08, 3.1665e-08, 9.3132e-10], [-1.2852e-07, 2.2352e-08, 4.6566e-09, ..., 2.1420e-08, 2.3283e-08, 0.0000e+00], ..., [ 1.8626e-08, 3.3807e-07, 6.5193e-09, ..., 1.6261e-06, 9.1270e-07, -9.3132e-10], [ 3.7253e-08, 5.3085e-08, 8.1956e-08, ..., 5.5879e-08, 8.7544e-08, 9.3132e-10], [ 8.3819e-09, 1.3039e-08, 1.5832e-08, ..., -1.9241e-06, -8.9314e-07, 0.0000e+00]], device='cuda:0') Epoch 215, bias, value: tensor([ 1.2109e-03, 1.2355e-02, 1.3674e-02, 7.1428e-05, 2.2415e-02, -1.7808e-03, -1.0427e-02, -7.6227e-03, 1.4705e-02, -1.3002e-02], device='cuda:0'), grad: tensor([ 7.2457e-07, -1.7975e-07, 4.0047e-08, 1.3448e-06, 1.2899e-06, -7.4916e-06, -2.8033e-07, 7.7114e-06, 8.3912e-07, -4.0159e-06], device='cuda:0') 100 0.0001 changing lr epoch 214, time 246.85, cls_loss 0.0010 cls_loss_mapping 0.0031 cls_loss_causal 0.5196 re_mapping 0.0039 re_causal 0.0127 /// teacc 99.10 lr 0.00010000 Epoch 216, weight, value: tensor([[-0.1834, -0.0356, 0.1834, ..., 0.0390, -0.1666, -0.1323], [-0.1243, -0.1593, -0.0387, ..., -0.1040, -0.1703, -0.0700], [ 0.2158, -0.0649, -0.1966, ..., -0.1617, -0.1634, -0.1904], ..., [-0.0662, -0.0228, -0.2142, ..., -0.0948, 0.1004, 0.0213], [-0.1633, -0.1054, -0.0252, ..., -0.1615, -0.0307, -0.0274], [-0.0649, -0.0866, -0.1106, ..., 0.0952, 0.0274, -0.1525]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 3.7253e-09, -1.0151e-07, ..., -1.9558e-08, 5.5134e-07, 6.6310e-07], [ 1.9558e-08, 9.5926e-08, 3.7253e-09, ..., 2.1048e-07, 2.3022e-05, 2.9147e-05], [-2.2352e-08, 2.7940e-08, 1.5832e-08, ..., 1.2107e-08, 8.2981e-07, 9.7323e-07], ..., [-4.0978e-08, -2.3935e-07, 1.8626e-09, ..., 6.9570e-07, -4.6566e-09, 2.4587e-07], [ 1.8626e-09, 5.5879e-09, 1.5832e-08, ..., 1.3225e-07, 4.0159e-06, 5.0403e-06], [ 9.3132e-10, 6.5193e-09, 6.7987e-08, ..., -1.1940e-06, -5.8766e-07, 1.4156e-07]], device='cuda:0') Epoch 216, bias, value: tensor([ 0.0013, 0.0111, 0.0139, -0.0003, 0.0223, -0.0018, -0.0105, -0.0050, 0.0136, -0.0137], device='cuda:0'), grad: tensor([ 2.7865e-06, 1.2350e-04, 4.3362e-06, 1.3150e-06, 9.3952e-06, 7.8380e-05, -2.3949e-04, 1.7565e-06, 2.1636e-05, -3.5968e-06], device='cuda:0') 100 0.0001 changing lr epoch 215, time 246.47, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4891 re_mapping 0.0039 re_causal 0.0131 /// teacc 98.99 lr 0.00010000 Epoch 217, weight, value: tensor([[-0.1838, -0.0357, 0.1833, ..., 0.0390, -0.1684, -0.1340], [-0.1248, -0.1609, -0.0387, ..., -0.1047, -0.1717, -0.0715], [ 0.2162, -0.0649, -0.1971, ..., -0.1619, -0.1637, -0.1910], ..., [-0.0661, -0.0231, -0.2153, ..., -0.0953, 0.1006, 0.0223], [-0.1637, -0.1053, -0.0254, ..., -0.1621, -0.0309, -0.0276], [-0.0651, -0.0868, -0.1110, ..., 0.0953, 0.0275, -0.1527]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.8626e-09, 1.2107e-08, ..., 6.5193e-09, 5.6811e-08, 4.3772e-08], [ 5.5879e-09, 8.3819e-09, -1.2964e-06, ..., -1.2489e-06, 3.3528e-08, 2.9802e-08], [ 1.1176e-08, 1.9558e-08, 3.7253e-08, ..., 2.0489e-08, 9.3132e-09, 8.3819e-09], ..., [ 1.1176e-08, 1.6764e-08, 7.1712e-08, ..., 1.7136e-07, 2.3283e-08, 9.3132e-10], [ 1.2107e-08, 2.5146e-08, 5.8580e-07, ..., 5.6345e-07, 9.9652e-08, 6.1467e-08], [ 3.7253e-09, 8.3819e-09, 6.1654e-07, ..., -2.9616e-07, -2.0396e-07, 4.6566e-09]], device='cuda:0') Epoch 217, bias, value: tensor([ 0.0009, 0.0107, 0.0138, -0.0007, 0.0224, -0.0018, -0.0103, -0.0047, 0.0136, -0.0135], device='cuda:0'), grad: tensor([ 4.8615e-07, -8.8811e-06, 3.2876e-07, -4.1351e-07, 3.2447e-06, 8.9500e-07, -7.1526e-07, 1.0552e-06, 3.0883e-06, 9.1363e-07], device='cuda:0') 100 0.0001 changing lr epoch 216, time 246.63, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.5218 re_mapping 0.0041 re_causal 0.0137 /// teacc 99.06 lr 0.00010000 Epoch 218, weight, value: tensor([[-0.1837, -0.0357, 0.1844, ..., 0.0395, -0.1687, -0.1342], [-0.1258, -0.1617, -0.0387, ..., -0.1049, -0.1723, -0.0718], [ 0.2163, -0.0650, -0.1981, ..., -0.1624, -0.1657, -0.1913], ..., [-0.0664, -0.0233, -0.2165, ..., -0.0957, 0.1009, 0.0227], [-0.1646, -0.1055, -0.0257, ..., -0.1627, -0.0309, -0.0277], [-0.0655, -0.0870, -0.1136, ..., 0.0951, 0.0275, -0.1529]], device='cuda:0'), grad: tensor([[ 1.3318e-07, 1.1828e-07, -3.4366e-07, ..., -2.6450e-07, 5.5879e-09, 2.7940e-09], [ 1.0245e-08, 2.7940e-09, 1.2107e-08, ..., 6.5193e-09, 5.1223e-08, 1.8626e-09], [-3.7253e-09, 1.2107e-08, 4.0047e-08, ..., 1.6764e-08, 7.4506e-09, 1.8626e-09], ..., [ 3.3528e-08, 9.3132e-09, 2.0489e-08, ..., 1.8626e-08, -2.6729e-07, 9.3132e-10], [-6.5193e-09, 3.4459e-08, 1.0431e-07, ..., 3.1665e-08, 3.2596e-08, 4.6566e-09], [ 3.4459e-08, 3.0734e-08, 3.7439e-07, ..., 1.5832e-07, 4.6566e-09, 2.7940e-09]], device='cuda:0') Epoch 218, bias, value: tensor([ 0.0016, 0.0103, 0.0135, -0.0008, 0.0224, -0.0017, -0.0103, -0.0045, 0.0136, -0.0138], device='cuda:0'), grad: tensor([ 1.9558e-07, -2.6673e-06, 1.2899e-06, -3.7905e-07, 1.6950e-07, 2.8405e-07, 1.6671e-07, 1.8463e-05, -1.8626e-05, 1.0701e-06], device='cuda:0') 100 0.0001 changing lr epoch 217, time 246.42, cls_loss 0.0008 cls_loss_mapping 0.0023 cls_loss_causal 0.4871 re_mapping 0.0041 re_causal 0.0130 /// teacc 99.10 lr 0.00010000 Epoch 219, weight, value: tensor([[-0.1842, -0.0358, 0.1847, ..., 0.0396, -0.1689, -0.1344], [-0.1253, -0.1620, -0.0387, ..., -0.1066, -0.1740, -0.0719], [ 0.2164, -0.0650, -0.2002, ..., -0.1626, -0.1662, -0.1939], ..., [-0.0668, -0.0234, -0.2178, ..., -0.0976, 0.1007, 0.0230], [-0.1649, -0.1056, -0.0260, ..., -0.1632, -0.0311, -0.0280], [-0.0656, -0.0871, -0.1139, ..., 0.0949, 0.0274, -0.1530]], device='cuda:0'), grad: tensor([[ 3.3341e-07, 0.0000e+00, -1.1269e-07, ..., -6.7987e-08, 3.9116e-08, 1.2107e-08], [ 6.3982e-07, 1.8626e-09, 2.7940e-09, ..., -1.3132e-07, 8.9034e-07, 3.7253e-09], [-6.2473e-06, 3.7253e-09, 5.5879e-09, ..., 2.7940e-09, 6.1467e-08, 2.7940e-09], ..., [ 2.3190e-06, 1.8626e-09, 5.5879e-09, ..., 3.5390e-08, -2.0210e-06, -5.5879e-09], [ 4.2468e-07, 0.0000e+00, 2.5146e-08, ..., 1.2107e-08, 2.1420e-08, 1.0245e-08], [ 4.9360e-08, 9.3132e-10, 6.3330e-08, ..., 1.3970e-08, 8.4378e-07, 3.7253e-09]], device='cuda:0') Epoch 219, bias, value: tensor([ 0.0016, 0.0110, 0.0123, -0.0011, 0.0228, -0.0017, -0.0102, -0.0054, 0.0136, -0.0136], device='cuda:0'), grad: tensor([ 5.6624e-07, 2.4624e-06, -1.0826e-05, 1.7341e-06, 2.6114e-06, 6.7800e-07, 9.7044e-07, -2.3656e-06, 8.6240e-07, 3.2838e-06], device='cuda:0') 100 0.0001 changing lr epoch 218, time 246.11, cls_loss 0.0009 cls_loss_mapping 0.0025 cls_loss_causal 0.5103 re_mapping 0.0039 re_causal 0.0128 /// teacc 99.17 lr 0.00010000 Epoch 220, weight, value: tensor([[-0.1847, -0.0358, 0.1850, ..., 0.0397, -0.1689, -0.1344], [-0.1250, -0.1627, -0.0387, ..., -0.1064, -0.1754, -0.0722], [ 0.2167, -0.0650, -0.2010, ..., -0.1629, -0.1667, -0.1937], ..., [-0.0671, -0.0238, -0.2202, ..., -0.0983, 0.1011, 0.0230], [-0.1659, -0.1057, -0.0261, ..., -0.1639, -0.0312, -0.0283], [-0.0661, -0.0872, -0.1141, ..., 0.0951, 0.0276, -0.1531]], device='cuda:0'), grad: tensor([[ 3.0734e-08, 8.3819e-09, -1.2573e-07, ..., -6.7055e-08, 1.3039e-08, 0.0000e+00], [ 8.2888e-08, 2.3283e-08, 2.4214e-08, ..., 2.5146e-08, 7.6368e-08, 9.3132e-10], [ 4.6566e-08, 1.9558e-08, 1.1269e-07, ..., 2.8871e-08, 7.9162e-08, 0.0000e+00], ..., [ 1.3039e-07, 3.6322e-08, 6.5193e-09, ..., -8.3819e-09, -2.6636e-07, -1.8626e-09], [ 1.2945e-07, 3.6322e-08, -9.4902e-07, ..., 1.6019e-07, -6.1840e-07, 9.3132e-10], [ 1.0710e-07, 2.3283e-08, 9.7789e-08, ..., -2.4121e-06, -7.1991e-07, 9.3132e-10]], device='cuda:0') Epoch 220, bias, value: tensor([ 0.0018, 0.0114, 0.0110, -0.0022, 0.0226, -0.0016, -0.0102, -0.0049, 0.0135, -0.0136], device='cuda:0'), grad: tensor([ 3.3528e-08, 8.4005e-07, 1.4417e-06, -5.9940e-06, 4.3213e-06, 4.4890e-06, 7.0557e-06, -4.4983e-07, -8.9332e-06, -2.8461e-06], device='cuda:0') 100 0.0001 changing lr epoch 219, time 246.11, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4976 re_mapping 0.0040 re_causal 0.0131 /// teacc 99.08 lr 0.00010000 Epoch 221, weight, value: tensor([[-0.1857, -0.0359, 0.1854, ..., 0.0398, -0.1692, -0.1345], [-0.1251, -0.1631, -0.0387, ..., -0.1065, -0.1758, -0.0723], [ 0.2172, -0.0650, -0.2019, ..., -0.1633, -0.1669, -0.1943], ..., [-0.0675, -0.0240, -0.2217, ..., -0.0990, 0.1010, 0.0230], [-0.1665, -0.1058, -0.0264, ..., -0.1644, -0.0314, -0.0285], [-0.0665, -0.0874, -0.1146, ..., 0.0953, 0.0278, -0.1533]], device='cuda:0'), grad: tensor([[ 4.4703e-08, 0.0000e+00, -1.8450e-06, ..., -7.9535e-07, 2.7567e-07, 3.7253e-09], [ 1.0990e-07, 0.0000e+00, 2.6077e-08, ..., 2.6077e-08, 5.7183e-07, 9.3132e-10], [-8.4098e-07, 9.3132e-10, 4.0047e-08, ..., 1.3970e-08, 4.1910e-08, 1.8626e-09], ..., [ 1.0803e-07, 9.3132e-10, 1.5832e-08, ..., 4.2841e-08, -5.1595e-07, 9.3132e-10], [ 2.0210e-07, 7.4506e-09, 5.3085e-08, ..., 3.0361e-07, -5.5879e-07, 1.1176e-08], [ 1.8626e-08, 0.0000e+00, 7.5903e-07, ..., -1.2107e-07, -1.5367e-07, 0.0000e+00]], device='cuda:0') Epoch 221, bias, value: tensor([ 0.0019, 0.0115, 0.0112, -0.0028, 0.0224, -0.0017, -0.0098, -0.0052, 0.0132, -0.0134], device='cuda:0'), grad: tensor([-9.1270e-08, 9.0823e-06, -1.0375e-06, -1.4128e-06, 5.6718e-07, 9.7603e-07, 2.1867e-06, -7.9721e-06, -2.8461e-06, 5.2806e-07], device='cuda:0') 100 0.0001 changing lr epoch 220, time 246.32, cls_loss 0.0010 cls_loss_mapping 0.0027 cls_loss_causal 0.5060 re_mapping 0.0038 re_causal 0.0124 /// teacc 99.07 lr 0.00010000 Epoch 222, weight, value: tensor([[-0.1887, -0.0361, 0.1851, ..., 0.0398, -0.1704, -0.1350], [-0.1257, -0.1634, -0.0389, ..., -0.1073, -0.1790, -0.0726], [ 0.2179, -0.0650, -0.2041, ..., -0.1636, -0.1677, -0.1960], ..., [-0.0682, -0.0243, -0.2234, ..., -0.1005, 0.1012, 0.0235], [-0.1668, -0.1059, -0.0264, ..., -0.1648, -0.0310, -0.0284], [-0.0667, -0.0874, -0.1150, ..., 0.0954, 0.0280, -0.1535]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 5.5917e-06, ..., -4.0047e-08, 3.7700e-06, 3.8631e-06], [ 0.0000e+00, 0.0000e+00, 7.8231e-08, ..., 5.5879e-08, 8.0094e-08, 5.0291e-08], [ 0.0000e+00, 0.0000e+00, 5.3085e-08, ..., 1.2107e-08, 3.7253e-08, 3.0734e-08], ..., [ 0.0000e+00, 0.0000e+00, 2.3283e-08, ..., 3.8184e-08, 8.3819e-09, 1.1176e-08], [ 9.3132e-10, 0.0000e+00, 4.1164e-07, ..., 1.4249e-07, 3.4552e-07, 2.9616e-07], [ 9.3132e-10, 0.0000e+00, 8.8476e-08, ..., -6.9942e-07, -2.2445e-07, 4.8429e-08]], device='cuda:0') Epoch 222, bias, value: tensor([ 0.0008, 0.0108, 0.0114, -0.0020, 0.0224, -0.0020, -0.0096, -0.0049, 0.0138, -0.0133], device='cuda:0'), grad: tensor([ 1.2219e-05, 3.2131e-07, 1.6857e-07, 7.0781e-08, 1.2880e-06, 2.5965e-06, -1.6123e-05, 1.2480e-07, 1.0459e-06, -1.7136e-06], device='cuda:0') 100 0.0001 changing lr epoch 221, time 246.43, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.5171 re_mapping 0.0041 re_causal 0.0131 /// teacc 99.11 lr 0.00010000 Epoch 223, weight, value: tensor([[-0.1893, -0.0361, 0.1853, ..., 0.0399, -0.1709, -0.1352], [-0.1263, -0.1639, -0.0390, ..., -0.1081, -0.1806, -0.0729], [ 0.2185, -0.0650, -0.2046, ..., -0.1638, -0.1681, -0.1962], ..., [-0.0686, -0.0247, -0.2253, ..., -0.1003, 0.1022, 0.0240], [-0.1670, -0.1060, -0.0267, ..., -0.1654, -0.0308, -0.0286], [-0.0669, -0.0877, -0.1153, ..., 0.0948, 0.0275, -0.1536]], device='cuda:0'), grad: tensor([[ 2.7940e-08, 2.7940e-09, -5.5693e-07, ..., 7.4506e-09, 1.8347e-07, 2.0489e-08], [ 8.1956e-08, 1.0245e-08, 1.8626e-08, ..., 2.7381e-07, 2.5332e-07, 1.2107e-08], [-4.0419e-07, 2.7008e-08, 2.0489e-08, ..., 6.8918e-08, 5.6811e-08, 4.6566e-09], ..., [ 1.1921e-07, 3.0734e-08, 1.2107e-08, ..., 1.4435e-07, -4.2561e-07, 9.3132e-10], [ 1.8068e-07, 2.7008e-08, 9.4995e-08, ..., 3.9563e-06, 2.9430e-06, 7.4506e-09], [ 1.1176e-08, 3.7253e-09, 1.0617e-07, ..., -7.2010e-06, -5.4836e-06, 1.8626e-09]], device='cuda:0') Epoch 223, bias, value: tensor([ 0.0009, 0.0108, 0.0116, -0.0016, 0.0231, -0.0022, -0.0095, -0.0050, 0.0142, -0.0140], device='cuda:0'), grad: tensor([-1.3318e-07, 1.2908e-06, -3.1199e-07, 2.0061e-06, 6.8173e-06, 1.6289e-06, 5.5879e-07, -7.4320e-07, 1.2159e-05, -2.3291e-05], device='cuda:0') 100 0.0001 changing lr epoch 222, time 246.31, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.4871 re_mapping 0.0039 re_causal 0.0126 /// teacc 99.09 lr 0.00010000 Epoch 224, weight, value: tensor([[-0.1896, -0.0361, 0.1860, ..., 0.0402, -0.1712, -0.1355], [-0.1268, -0.1642, -0.0391, ..., -0.1088, -0.1817, -0.0736], [ 0.2200, -0.0650, -0.2048, ..., -0.1641, -0.1681, -0.1947], ..., [-0.0704, -0.0248, -0.2277, ..., -0.1008, 0.1023, 0.0223], [-0.1676, -0.1060, -0.0270, ..., -0.1666, -0.0312, -0.0284], [-0.0679, -0.0877, -0.1157, ..., 0.0949, 0.0279, -0.1538]], device='cuda:0'), grad: tensor([[ 2.7940e-08, 0.0000e+00, -2.3283e-07, ..., -1.2573e-07, 2.7940e-08, 1.0245e-08], [ 2.3432e-06, 9.3132e-10, 1.1176e-08, ..., 1.5832e-08, 1.0747e-06, 1.8626e-09], [ 1.3905e-06, 0.0000e+00, 1.0245e-08, ..., 4.0978e-08, 6.3237e-07, 3.7253e-09], ..., [-4.4629e-06, 9.3132e-10, 5.5879e-09, ..., 3.0547e-07, -1.7332e-06, 1.8626e-09], [ 2.8871e-08, 0.0000e+00, 2.8871e-08, ..., 6.5193e-08, 4.9360e-08, 1.3039e-08], [ 2.7008e-08, 9.3132e-10, 1.4808e-07, ..., -9.6951e-07, -1.6941e-06, 2.7940e-09]], device='cuda:0') Epoch 224, bias, value: tensor([ 0.0013, 0.0110, 0.0123, -0.0014, 0.0230, -0.0022, -0.0095, -0.0057, 0.0143, -0.0139], device='cuda:0'), grad: tensor([-7.3574e-08, 8.2925e-06, 5.7444e-06, 2.5220e-06, 1.9893e-06, -1.2014e-07, -5.1316e-07, -1.6287e-05, 6.3051e-07, -2.2314e-06], device='cuda:0') 100 0.0001 changing lr epoch 223, time 246.41, cls_loss 0.0007 cls_loss_mapping 0.0025 cls_loss_causal 0.5024 re_mapping 0.0040 re_causal 0.0134 /// teacc 99.07 lr 0.00010000 Epoch 225, weight, value: tensor([[-0.1899, -0.0359, 0.1883, ..., 0.0409, -0.1715, -0.1356], [-0.1260, -0.1645, -0.0391, ..., -0.1090, -0.1825, -0.0737], [ 0.2202, -0.0650, -0.2057, ..., -0.1643, -0.1684, -0.1947], ..., [-0.0710, -0.0249, -0.2298, ..., -0.1013, 0.1025, 0.0223], [-0.1679, -0.1062, -0.0274, ..., -0.1674, -0.0313, -0.0285], [-0.0680, -0.0878, -0.1174, ..., 0.0943, 0.0272, -0.1539]], device='cuda:0'), grad: tensor([[ 1.3039e-08, 0.0000e+00, -1.6857e-06, ..., -9.6858e-07, 4.0047e-08, 1.4901e-08], [ 4.2096e-07, 0.0000e+00, 1.3039e-08, ..., 1.0431e-07, 6.8266e-07, 3.8743e-07], [ 2.5425e-07, 0.0000e+00, 9.7789e-08, ..., 5.8673e-08, 6.3516e-07, 3.6880e-07], ..., [-1.2955e-06, 9.3132e-10, 2.7940e-08, ..., 2.4214e-08, -2.2817e-06, -1.2992e-06], [ 5.5879e-08, 0.0000e+00, 7.3574e-08, ..., 4.0047e-08, 1.1176e-08, 1.3970e-08], [ 2.7940e-08, 9.3132e-10, 1.3895e-06, ..., 8.6613e-07, 6.1467e-08, 2.9802e-08]], device='cuda:0') Epoch 225, bias, value: tensor([ 0.0027, 0.0116, 0.0115, -0.0007, 0.0237, -0.0025, -0.0097, -0.0058, 0.0144, -0.0146], device='cuda:0'), grad: tensor([-3.2745e-06, 3.8072e-06, 2.7269e-06, 5.0291e-07, 3.0510e-06, -7.0687e-07, 9.7230e-07, -9.2834e-06, -1.3318e-06, 3.5204e-06], device='cuda:0') 100 0.0001 changing lr epoch 224, time 246.20, cls_loss 0.0008 cls_loss_mapping 0.0019 cls_loss_causal 0.4964 re_mapping 0.0037 re_causal 0.0121 /// teacc 99.02 lr 0.00010000 Epoch 226, weight, value: tensor([[-0.1915, -0.0360, 0.1888, ..., 0.0410, -0.1720, -0.1358], [-0.1262, -0.1653, -0.0391, ..., -0.1098, -0.1835, -0.0738], [ 0.2207, -0.0650, -0.2066, ..., -0.1647, -0.1688, -0.1949], ..., [-0.0712, -0.0252, -0.2344, ..., -0.1018, 0.1027, 0.0230], [-0.1683, -0.1063, -0.0277, ..., -0.1683, -0.0315, -0.0287], [-0.0683, -0.0881, -0.1181, ..., 0.0944, 0.0274, -0.1541]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, -3.1665e-08, ..., -3.7253e-08, 3.2596e-08, 2.1420e-08], [ 4.6566e-09, 9.3132e-10, -4.0326e-07, ..., 1.5832e-08, 5.1223e-08, 2.7940e-09], [ 3.3528e-08, 1.8626e-09, 1.4901e-08, ..., 3.0454e-07, 1.1176e-08, 5.7742e-08], ..., [ 8.3819e-09, 3.7253e-09, 4.7497e-08, ..., 1.6764e-08, -9.9652e-08, 1.8626e-09], [ 3.7253e-09, 9.3132e-10, 1.5646e-07, ..., 1.3039e-08, 6.2399e-08, 4.0978e-08], [ 2.7940e-09, 0.0000e+00, 3.4459e-08, ..., -1.8533e-07, 2.0489e-08, 4.6566e-09]], device='cuda:0') Epoch 226, bias, value: tensor([ 0.0029, 0.0120, 0.0114, -0.0007, 0.0235, -0.0025, -0.0097, -0.0062, 0.0144, -0.0145], device='cuda:0'), grad: tensor([ 6.4261e-08, -1.4324e-06, 5.9325e-07, 1.9465e-07, -4.7311e-07, -2.4214e-08, 4.2375e-07, -1.3690e-07, 7.7579e-07, 9.3132e-10], device='cuda:0') 100 0.0001 changing lr epoch 225, time 246.10, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.5311 re_mapping 0.0038 re_causal 0.0126 /// teacc 99.02 lr 0.00010000 Epoch 227, weight, value: tensor([[-0.1918, -0.0361, 0.1890, ..., 0.0411, -0.1724, -0.1359], [-0.1271, -0.1660, -0.0394, ..., -0.1100, -0.1842, -0.0740], [ 0.2214, -0.0650, -0.2074, ..., -0.1652, -0.1701, -0.1951], ..., [-0.0714, -0.0254, -0.2353, ..., -0.1021, 0.1032, 0.0233], [-0.1687, -0.1063, -0.0280, ..., -0.1697, -0.0317, -0.0290], [-0.0683, -0.0882, -0.1183, ..., 0.0944, 0.0274, -0.1542]], device='cuda:0'), grad: tensor([[ 4.6566e-09, 1.8626e-09, -1.1176e-08, ..., -1.1176e-08, 9.6858e-08, 1.3970e-08], [ 2.0489e-08, 9.3132e-10, 2.7940e-09, ..., 5.5879e-09, 2.7381e-07, 6.5193e-09], [-1.5926e-07, 9.3132e-10, 1.8626e-09, ..., 1.8626e-09, 1.1828e-07, 8.3819e-09], ..., [ 1.4901e-08, 9.3132e-10, 9.3132e-10, ..., 8.3819e-09, -1.8962e-06, -1.0058e-07], [ 1.0245e-08, 4.9360e-08, 1.9558e-08, ..., 4.7497e-08, 4.3493e-07, 1.4901e-08], [ 2.7940e-09, 1.8626e-09, 1.3970e-08, ..., -8.6613e-08, 9.4250e-07, 5.9605e-08]], device='cuda:0') Epoch 227, bias, value: tensor([ 0.0029, 0.0121, 0.0113, -0.0007, 0.0236, -0.0025, -0.0097, -0.0061, 0.0145, -0.0146], device='cuda:0'), grad: tensor([ 3.9488e-07, 9.8068e-07, 2.2259e-07, 1.3141e-06, 2.7195e-07, -3.7290e-06, 1.9185e-07, -6.8918e-06, 3.7812e-06, 3.4608e-06], device='cuda:0') 100 0.0001 changing lr epoch 226, time 246.74, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.4831 re_mapping 0.0037 re_causal 0.0116 /// teacc 99.16 lr 0.00010000 Epoch 228, weight, value: tensor([[-0.1922, -0.0361, 0.1898, ..., 0.0415, -0.1731, -0.1361], [-0.1272, -0.1668, -0.0401, ..., -0.1109, -0.1859, -0.0743], [ 0.2218, -0.0651, -0.2095, ..., -0.1660, -0.1706, -0.1955], ..., [-0.0714, -0.0257, -0.2372, ..., -0.1026, 0.1036, 0.0236], [-0.1691, -0.1072, -0.0260, ..., -0.1714, -0.0296, -0.0270], [-0.0686, -0.0884, -0.1187, ..., 0.0943, 0.0273, -0.1544]], device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.8626e-09, -1.5553e-07, ..., -2.9802e-08, 7.4506e-08, 2.6077e-08], [ 2.5146e-08, 2.7940e-09, 3.5390e-08, ..., 3.1665e-08, 2.5798e-07, 1.0245e-08], [-1.3690e-07, 1.8626e-09, 2.1420e-08, ..., 4.6566e-09, 1.6764e-08, 2.7940e-09], ..., [ 6.7055e-08, 1.1176e-08, 3.2596e-08, ..., 1.8626e-08, -2.3916e-06, -6.5193e-09], [ 2.0489e-08, 3.1665e-08, 1.7695e-07, ..., 9.6858e-08, 1.5926e-07, 1.3039e-08], [ 2.7940e-09, 9.3132e-09, 2.4214e-08, ..., -7.8976e-07, 1.8580e-06, 1.0245e-08]], device='cuda:0') Epoch 228, bias, value: tensor([ 0.0035, 0.0131, 0.0105, -0.0013, 0.0237, -0.0025, -0.0102, -0.0067, 0.0165, -0.0147], device='cuda:0'), grad: tensor([-6.7987e-08, 4.3120e-07, -1.3411e-07, 3.9712e-06, 1.4435e-06, -5.0068e-06, 5.1223e-08, -7.2829e-06, 1.1101e-06, 5.4352e-06], device='cuda:0') 100 0.0001 changing lr epoch 227, time 246.46, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.4733 re_mapping 0.0038 re_causal 0.0121 /// teacc 99.06 lr 0.00010000 Epoch 229, weight, value: tensor([[-0.1950, -0.0365, 0.1899, ..., 0.0416, -0.1745, -0.1368], [-0.1276, -0.1700, -0.0401, ..., -0.1116, -0.1882, -0.0746], [ 0.2225, -0.0651, -0.2101, ..., -0.1664, -0.1710, -0.1957], ..., [-0.0716, -0.0261, -0.2385, ..., -0.1033, 0.1042, 0.0241], [-0.1694, -0.1085, -0.0264, ..., -0.1729, -0.0301, -0.0275], [-0.0693, -0.0887, -0.1190, ..., 0.0938, 0.0270, -0.1546]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 8.0094e-08, -1.1921e-07, ..., 2.2352e-08, 1.4529e-07, 7.5437e-08], [ 1.8626e-09, 2.1420e-07, 1.2107e-08, ..., 9.8720e-08, 3.5856e-07, 6.7987e-08], [ 1.8626e-09, 1.2759e-07, 7.4506e-09, ..., 3.6322e-07, 5.7835e-07, 1.9558e-08], ..., [ 4.6566e-09, 1.8515e-06, 3.7253e-09, ..., 1.4063e-07, -1.2759e-06, -4.3958e-07], [ 1.8626e-09, 1.1735e-07, 1.4994e-07, ..., 8.0094e-08, -4.8522e-07, -2.7008e-08], [ 0.0000e+00, 6.8918e-08, 1.6484e-07, ..., 1.8803e-06, 3.8370e-07, 8.3819e-09]], device='cuda:0') Epoch 229, bias, value: tensor([ 0.0032, 0.0121, 0.0107, -0.0016, 0.0245, -0.0024, -0.0101, -0.0059, 0.0163, -0.0154], device='cuda:0'), grad: tensor([ 8.7731e-07, 3.3733e-06, 2.3499e-05, -1.3128e-05, -6.8136e-06, 6.1579e-06, 8.7172e-07, 5.0217e-06, -2.8208e-05, 8.2850e-06], device='cuda:0') 100 0.0001 changing lr epoch 228, time 246.71, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.5037 re_mapping 0.0040 re_causal 0.0131 /// teacc 99.09 lr 0.00010000 Epoch 230, weight, value: tensor([[-0.1958, -0.0367, 0.1937, ..., 0.0431, -0.1759, -0.1386], [-0.1277, -0.1716, -0.0406, ..., -0.1097, -0.1888, -0.0748], [ 0.2245, -0.0650, -0.2106, ..., -0.1666, -0.1714, -0.1960], ..., [-0.0719, -0.0266, -0.2399, ..., -0.1059, 0.1043, 0.0246], [-0.1701, -0.1090, -0.0268, ..., -0.1740, -0.0305, -0.0280], [-0.0726, -0.0889, -0.1254, ..., 0.0921, 0.0270, -0.1550]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -1.2945e-07, ..., -6.0536e-08, 1.2107e-08, 1.8626e-09], [ 7.4506e-09, 6.5193e-09, -6.5193e-09, ..., 0.0000e+00, 3.5390e-08, 0.0000e+00], [ 5.9605e-08, 5.4948e-08, 2.3283e-08, ..., 5.5879e-09, 8.3819e-09, 9.3132e-10], ..., [ 3.9116e-08, 2.8871e-08, 9.3132e-09, ..., 1.3970e-08, -4.1910e-08, 0.0000e+00], [ 4.6566e-09, 2.7940e-09, 6.5193e-08, ..., 2.7940e-08, 3.1665e-08, 1.8626e-09], [ 9.3132e-10, 1.8626e-09, 7.2643e-08, ..., -6.9849e-08, -3.8184e-08, 0.0000e+00]], device='cuda:0') Epoch 230, bias, value: tensor([ 0.0049, 0.0129, 0.0115, -0.0018, 0.0244, -0.0025, -0.0099, -0.0065, 0.0160, -0.0176], device='cuda:0'), grad: tensor([-1.0803e-07, -1.2200e-06, 2.6450e-07, -2.4680e-07, 2.2724e-07, 1.2387e-07, -9.6858e-08, 7.7859e-07, 3.8464e-07, -9.5926e-08], device='cuda:0') 100 0.0001 changing lr epoch 229, time 246.65, cls_loss 0.0007 cls_loss_mapping 0.0030 cls_loss_causal 0.5095 re_mapping 0.0037 re_causal 0.0124 /// teacc 99.10 lr 0.00010000 Epoch 231, weight, value: tensor([[-0.1961, -0.0377, 0.1940, ..., 0.0434, -0.1768, -0.1388], [-0.1279, -0.1735, -0.0410, ..., -0.1110, -0.1899, -0.0749], [ 0.2250, -0.0650, -0.2111, ..., -0.1670, -0.1718, -0.1962], ..., [-0.0722, -0.0282, -0.2413, ..., -0.1064, 0.1046, 0.0246], [-0.1705, -0.1093, -0.0272, ..., -0.1751, -0.0308, -0.0281], [-0.0730, -0.0894, -0.1257, ..., 0.0914, 0.0268, -0.1551]], device='cuda:0'), grad: tensor([[ 9.3132e-08, 6.5193e-09, -3.2596e-08, ..., 2.1420e-08, 3.3993e-08, 2.7940e-09], [ 1.5739e-07, 2.5146e-08, 2.7940e-09, ..., 4.5169e-08, 3.6787e-08, 4.6566e-10], [-1.2163e-06, -3.1665e-08, 2.3283e-09, ..., 6.5193e-09, 2.0489e-08, 0.0000e+00], ..., [ 1.9465e-07, 2.0955e-08, 2.3283e-09, ..., 1.7835e-07, 3.6322e-08, 9.3132e-10], [ 6.2678e-07, 3.6787e-08, 7.4506e-09, ..., 5.4482e-08, -3.2596e-09, 9.7789e-09], [ 1.1176e-08, 3.7253e-09, 1.2107e-08, ..., -1.1986e-06, -4.4098e-07, 4.6566e-10]], device='cuda:0') Epoch 231, bias, value: tensor([ 0.0051, 0.0127, 0.0116, -0.0017, 0.0262, -0.0026, -0.0097, -0.0066, 0.0160, -0.0187], device='cuda:0'), grad: tensor([ 3.3854e-07, 1.9185e-07, -2.1700e-06, 4.5635e-08, 2.2650e-06, -9.0804e-08, 4.0932e-07, 1.1735e-06, 5.9651e-07, -2.7753e-06], device='cuda:0') 100 0.0001 changing lr epoch 230, time 246.84, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.5102 re_mapping 0.0038 re_causal 0.0129 /// teacc 99.01 lr 0.00010000 Epoch 232, weight, value: tensor([[-0.1973, -0.0366, 0.1943, ..., 0.0434, -0.1772, -0.1390], [-0.1289, -0.1757, -0.0411, ..., -0.1116, -0.1912, -0.0750], [ 0.2262, -0.0651, -0.2117, ..., -0.1667, -0.1720, -0.1964], ..., [-0.0730, -0.0292, -0.2441, ..., -0.1073, 0.1047, 0.0249], [-0.1717, -0.1097, -0.0275, ..., -0.1766, -0.0311, -0.0282], [-0.0730, -0.0902, -0.1258, ..., 0.0913, 0.0269, -0.1551]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 2.7940e-09, -4.0326e-07, ..., -4.3772e-08, 1.5832e-08, 1.0245e-08], [ 9.2201e-08, 5.1223e-08, 1.9558e-08, ..., -8.3819e-09, 1.2107e-08, 9.3132e-10], [ 9.3132e-10, 2.7940e-09, 6.5193e-09, ..., 9.3132e-10, 1.8626e-09, 9.3132e-10], ..., [ 1.7695e-08, 1.3039e-08, 7.4506e-09, ..., 2.7940e-09, -2.6077e-08, 9.3132e-10], [ 3.7253e-09, 6.5193e-09, 2.8871e-08, ..., 2.7940e-09, 1.3039e-08, 3.7253e-09], [ 7.4506e-09, 1.1176e-08, 2.9802e-08, ..., 6.5193e-09, 3.7253e-08, 1.8626e-09]], device='cuda:0') Epoch 232, bias, value: tensor([ 0.0052, 0.0119, 0.0124, -0.0017, 0.0265, -0.0026, -0.0098, -0.0063, 0.0158, -0.0187], device='cuda:0'), grad: tensor([-5.3830e-07, 2.6915e-07, 5.8673e-08, -6.5565e-07, 8.1025e-08, -2.3004e-07, 3.7160e-07, 2.2259e-07, 1.3411e-07, 2.9244e-07], device='cuda:0') 100 0.0001 changing lr epoch 231, time 246.39, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.5026 re_mapping 0.0037 re_causal 0.0121 /// teacc 99.17 lr 0.00010000 Epoch 233, weight, value: tensor([[-0.1977, -0.0362, 0.1946, ..., 0.0432, -0.1775, -0.1391], [-0.1296, -0.1789, -0.0414, ..., -0.1092, -0.1918, -0.0751], [ 0.2268, -0.0651, -0.2128, ..., -0.1673, -0.1730, -0.1966], ..., [-0.0732, -0.0261, -0.2464, ..., -0.1050, 0.1079, 0.0248], [-0.1722, -0.1096, -0.0275, ..., -0.1779, -0.0313, -0.0282], [-0.0731, -0.0936, -0.1259, ..., 0.0911, 0.0243, -0.1553]], device='cuda:0'), grad: tensor([[ 1.4901e-07, 7.5437e-08, -5.0291e-08, ..., -1.7695e-08, 3.7253e-09, 1.8626e-09], [ 3.5949e-07, 1.7602e-07, 7.4506e-09, ..., 6.5193e-09, 1.1176e-08, 9.3132e-10], [ 1.8440e-07, 1.5274e-07, 2.7940e-09, ..., 1.8626e-09, 9.3132e-09, 0.0000e+00], ..., [ 5.5321e-07, 2.5425e-07, 2.7940e-09, ..., 1.3039e-08, -1.4901e-08, 0.0000e+00], [ 1.5087e-07, 1.0338e-07, 9.3132e-09, ..., 2.7008e-08, 6.5193e-09, 2.7940e-09], [ 7.8231e-08, 4.9360e-08, 2.3283e-08, ..., -1.4994e-07, -2.2352e-08, 0.0000e+00]], device='cuda:0') Epoch 233, bias, value: tensor([ 0.0051, 0.0113, 0.0124, -0.0020, 0.0261, -0.0026, -0.0098, -0.0035, 0.0157, -0.0200], device='cuda:0'), grad: tensor([ 6.8452e-07, 1.3895e-06, 1.3281e-06, -9.9316e-06, 2.6356e-07, 2.2538e-06, 1.3225e-07, 2.7530e-06, 9.8068e-07, 1.3504e-07], device='cuda:0') 100 0.0001 changing lr epoch 232, time 246.54, cls_loss 0.0010 cls_loss_mapping 0.0026 cls_loss_causal 0.4874 re_mapping 0.0039 re_causal 0.0122 /// teacc 99.00 lr 0.00010000 Epoch 234, weight, value: tensor([[-0.1990, -0.0363, 0.1948, ..., 0.0427, -0.1778, -0.1394], [-0.1298, -0.1819, -0.0405, ..., -0.1070, -0.1924, -0.0754], [ 0.2285, -0.0654, -0.2133, ..., -0.1677, -0.1739, -0.1964], ..., [-0.0755, -0.0257, -0.2490, ..., -0.1052, 0.1080, 0.0244], [-0.1731, -0.1115, -0.0279, ..., -0.1843, -0.0323, -0.0289], [-0.0731, -0.0936, -0.1262, ..., 0.0931, 0.0263, -0.1555]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-08, -1.0990e-07, ..., -7.8231e-08, 1.8626e-08, 7.8231e-08], [ 4.6566e-09, 9.3132e-09, 1.4901e-08, ..., 2.1420e-08, 1.0338e-07, 1.3039e-06], [-2.7940e-09, 4.1910e-08, 8.3819e-09, ..., 1.0245e-08, 7.5437e-08, 1.0412e-06], ..., [ 1.1176e-08, 3.3528e-08, 2.8871e-08, ..., 2.3004e-07, 2.7940e-08, 3.2596e-08], [ 0.0000e+00, 3.9116e-08, 1.6764e-08, ..., 3.1665e-08, -3.3528e-08, 9.3132e-10], [ 0.0000e+00, 1.5832e-08, 3.8184e-08, ..., -1.2107e-08, -5.5879e-08, 4.5635e-08]], device='cuda:0') Epoch 234, bias, value: tensor([ 0.0043, 0.0126, 0.0132, -0.0028, 0.0235, -0.0025, -0.0096, -0.0036, 0.0139, -0.0178], device='cuda:0'), grad: tensor([ 3.0454e-07, 6.6683e-06, 5.5879e-06, 4.7963e-07, 1.1306e-06, 4.5300e-06, -1.9461e-05, 8.4005e-07, -5.1782e-07, 4.1910e-07], device='cuda:0') 100 0.0001 changing lr epoch 233, time 246.77, cls_loss 0.0008 cls_loss_mapping 0.0028 cls_loss_causal 0.5076 re_mapping 0.0042 re_causal 0.0127 /// teacc 99.06 lr 0.00010000 Epoch 235, weight, value: tensor([[-0.2002, -0.0367, 0.1949, ..., 0.0427, -0.1780, -0.1397], [-0.1309, -0.1861, -0.0406, ..., -0.1070, -0.1928, -0.0759], [ 0.2322, -0.0646, -0.2140, ..., -0.1681, -0.1742, -0.1963], ..., [-0.0764, -0.0259, -0.2501, ..., -0.1058, 0.1078, 0.0243], [-0.1790, -0.1158, -0.0281, ..., -0.1847, -0.0325, -0.0291], [-0.0732, -0.0936, -0.1263, ..., 0.0931, 0.0264, -0.1558]], device='cuda:0'), grad: tensor([[ 1.1269e-07, 4.6566e-09, -7.5437e-08, ..., -6.7055e-08, 7.4506e-09, 6.3330e-08], [ 1.1981e-05, 2.7940e-09, 6.0536e-08, ..., 2.0489e-08, 3.7253e-09, 1.0245e-08], [-1.5497e-05, 6.5193e-09, 2.6077e-08, ..., 1.1176e-08, 1.5832e-08, 1.1176e-08], ..., [ 2.1514e-07, 7.4506e-09, 4.6566e-09, ..., 1.8626e-09, 2.7940e-09, 5.5879e-09], [ 8.8569e-07, -2.4959e-07, 1.0990e-07, ..., 1.8626e-08, -8.1025e-07, -5.0478e-07], [ 9.0338e-08, 1.9558e-08, 2.5146e-08, ..., 1.2107e-08, 1.1176e-08, 1.2107e-08]], device='cuda:0') Epoch 235, bias, value: tensor([ 0.0043, 0.0128, 0.0163, -0.0039, 0.0238, -0.0025, -0.0096, -0.0042, 0.0120, -0.0178], device='cuda:0'), grad: tensor([ 3.9022e-07, 3.9846e-05, -5.1260e-05, 1.3663e-06, 4.1351e-06, 3.2503e-06, 1.0729e-06, 7.5903e-07, -1.1921e-07, 5.5041e-07], device='cuda:0') 100 0.0001 changing lr epoch 234, time 246.64, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.5255 re_mapping 0.0036 re_causal 0.0124 /// teacc 99.11 lr 0.00010000 Epoch 236, weight, value: tensor([[-0.2005, -0.0369, 0.1949, ..., 0.0427, -0.1785, -0.1402], [-0.1317, -0.1911, -0.0409, ..., -0.1065, -0.1932, -0.0765], [ 0.2320, -0.0653, -0.2149, ..., -0.1686, -0.1749, -0.1977], ..., [-0.0764, -0.0260, -0.2514, ..., -0.1069, 0.1077, 0.0270], [-0.1799, -0.1161, -0.0283, ..., -0.1849, -0.0326, -0.0294], [-0.0732, -0.0937, -0.1263, ..., 0.0929, 0.0264, -0.1559]], device='cuda:0'), grad: tensor([[ 4.1910e-08, 1.6764e-08, 1.6764e-08, ..., -4.8429e-08, 9.3132e-09, 2.7940e-08], [ 1.0524e-07, 7.1712e-08, 8.3819e-09, ..., -6.3330e-08, 1.6764e-08, 1.8626e-09], [ 2.6450e-07, 1.3132e-07, 2.1048e-07, ..., 1.5832e-08, 1.4901e-08, 4.1910e-08], ..., [ 6.8918e-08, 2.3283e-08, 4.6566e-09, ..., 1.2200e-07, -6.5193e-09, 0.0000e+00], [ 6.8918e-08, 5.1223e-08, 5.6718e-07, ..., 2.5146e-08, 4.2841e-08, 1.2759e-07], [ 5.5879e-09, 2.7940e-09, 6.0536e-08, ..., -2.0489e-07, -5.7742e-08, 3.7253e-09]], device='cuda:0') Epoch 236, bias, value: tensor([ 0.0042, 0.0131, 0.0156, -0.0026, 0.0240, -0.0026, -0.0095, -0.0047, 0.0117, -0.0178], device='cuda:0'), grad: tensor([ 3.1199e-07, -1.2293e-07, 1.9129e-06, -2.6673e-06, 3.7067e-07, 2.0470e-06, -4.2059e-06, 6.1188e-07, 1.9893e-06, -2.3935e-07], device='cuda:0') 100 0.0001 changing lr epoch 235, time 246.40, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4789 re_mapping 0.0035 re_causal 0.0114 /// teacc 99.12 lr 0.00010000 Epoch 237, weight, value: tensor([[-0.2015, -0.0371, 0.1951, ..., 0.0427, -0.1786, -0.1404], [-0.1343, -0.1943, -0.0413, ..., -0.1066, -0.1944, -0.0766], [ 0.2337, -0.0654, -0.2153, ..., -0.1692, -0.1751, -0.1981], ..., [-0.0773, -0.0260, -0.2536, ..., -0.1086, 0.1065, 0.0270], [-0.1802, -0.1166, -0.0286, ..., -0.1852, -0.0327, -0.0294], [-0.0733, -0.0937, -0.1264, ..., 0.0932, 0.0275, -0.1560]], device='cuda:0'), grad: tensor([[ 3.7253e-09, -2.6077e-08, -6.4448e-07, ..., -3.2317e-07, 1.3039e-08, 1.8626e-09], [ 1.6764e-08, 9.3132e-09, 3.7253e-09, ..., 9.3132e-10, 2.1420e-08, 0.0000e+00], [-1.9558e-08, 3.1665e-08, 2.9895e-07, ..., 1.8533e-07, 3.7253e-09, 0.0000e+00], ..., [ 1.2107e-08, 2.7008e-08, 8.3819e-09, ..., 9.3132e-08, -1.4901e-08, 0.0000e+00], [ 6.5193e-09, 1.9092e-07, 1.7509e-07, ..., 1.6578e-07, 2.0489e-08, 9.3132e-10], [ 5.7742e-07, 1.4268e-06, 7.2643e-08, ..., -3.4925e-07, -1.2480e-07, 9.3132e-10]], device='cuda:0') Epoch 237, bias, value: tensor([ 0.0042, 0.0125, 0.0159, -0.0026, 0.0240, -0.0026, -0.0094, -0.0053, 0.0116, -0.0172], device='cuda:0'), grad: tensor([-1.1688e-06, 6.7987e-08, 6.5751e-07, -5.1111e-06, 1.7229e-07, -1.4538e-06, 2.0023e-07, 1.9278e-07, 2.4214e-06, 3.9861e-06], device='cuda:0') 100 0.0001 changing lr epoch 236, time 246.57, cls_loss 0.0009 cls_loss_mapping 0.0018 cls_loss_causal 0.4685 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.07 lr 0.00010000 Epoch 238, weight, value: tensor([[-0.2019, -0.0378, 0.1951, ..., 0.0427, -0.1790, -0.1406], [-0.1348, -0.1963, -0.0415, ..., -0.1041, -0.1950, -0.0769], [ 0.2340, -0.0655, -0.2175, ..., -0.1695, -0.1756, -0.1989], ..., [-0.0775, -0.0261, -0.2547, ..., -0.1100, 0.1065, 0.0281], [-0.1802, -0.1161, -0.0288, ..., -0.1854, -0.0327, -0.0292], [-0.0734, -0.0937, -0.1264, ..., 0.0931, 0.0276, -0.1562]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 6.5193e-09, 1.7695e-08, ..., -5.5879e-09, 1.4901e-08, 1.9558e-08], [ 1.8626e-09, 4.6566e-09, 6.5193e-09, ..., 1.6764e-08, 2.4214e-08, 3.7253e-09], [ 1.8626e-09, 3.7253e-09, 1.0245e-08, ..., 1.3039e-08, 6.5193e-09, 4.6566e-09], ..., [ 1.8626e-09, 3.7253e-09, 9.3132e-10, ..., 5.5879e-09, -6.8918e-08, 9.3132e-10], [ 1.8626e-09, -1.8626e-09, 2.0489e-08, ..., 4.6566e-09, 8.3819e-09, 4.6566e-09], [ 0.0000e+00, 2.7940e-09, 7.4506e-09, ..., 1.4063e-07, 9.3132e-09, 1.8626e-09]], device='cuda:0') Epoch 238, bias, value: tensor([ 0.0041, 0.0152, 0.0157, -0.0026, 0.0239, -0.0029, -0.0093, -0.0071, 0.0122, -0.0175], device='cuda:0'), grad: tensor([ 3.2037e-07, 6.7335e-07, 4.9639e-07, 1.2331e-06, -1.0803e-07, -2.1886e-07, 1.8626e-09, -7.7114e-07, -2.1011e-06, 4.7497e-07], device='cuda:0') 100 0.0001 changing lr epoch 237, time 246.57, cls_loss 0.0009 cls_loss_mapping 0.0035 cls_loss_causal 0.5220 re_mapping 0.0035 re_causal 0.0116 /// teacc 99.07 lr 0.00010000 Epoch 239, weight, value: tensor([[-0.2028, -0.0391, 0.1952, ..., 0.0427, -0.1794, -0.1409], [-0.1353, -0.2009, -0.0416, ..., -0.1034, -0.1958, -0.0772], [ 0.2344, -0.0658, -0.2199, ..., -0.1702, -0.1762, -0.2017], ..., [-0.0779, -0.0259, -0.2558, ..., -0.1105, 0.1066, 0.0284], [-0.1805, -0.1176, -0.0306, ..., -0.1859, -0.0329, -0.0307], [-0.0735, -0.0933, -0.1265, ..., 0.0931, 0.0276, -0.1571]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.3039e-08, 6.5193e-09, ..., 2.0489e-08, 1.3970e-08, 9.3132e-09], [ 9.3132e-10, 1.2107e-08, -2.4214e-08, ..., 6.5193e-08, 1.6764e-08, 6.5193e-09], [ 0.0000e+00, 1.0245e-08, 9.3132e-10, ..., 6.5193e-09, 6.5193e-09, 5.5879e-09], ..., [ 2.7940e-09, 2.5146e-08, 7.4506e-09, ..., 1.3039e-07, 3.9116e-08, 1.3970e-08], [ 0.0000e+00, 8.6799e-07, 1.8626e-09, ..., 7.2643e-08, 5.1688e-07, 5.1502e-07], [ 0.0000e+00, 5.0291e-08, 2.7940e-09, ..., -2.5239e-07, -1.2852e-07, 2.5146e-08]], device='cuda:0') Epoch 239, bias, value: tensor([ 0.0041, 0.0157, 0.0154, -0.0029, 0.0239, -0.0029, -0.0089, -0.0073, 0.0115, -0.0177], device='cuda:0'), grad: tensor([ 3.0547e-07, 2.2072e-07, 1.7323e-07, 7.6517e-06, 3.3341e-07, -1.8403e-05, 1.5870e-06, 8.5961e-07, 7.9200e-06, -6.3796e-07], device='cuda:0') 100 0.0001 changing lr epoch 238, time 246.68, cls_loss 0.0006 cls_loss_mapping 0.0029 cls_loss_causal 0.4674 re_mapping 0.0037 re_causal 0.0121 /// teacc 99.09 lr 0.00010000 Epoch 240, weight, value: tensor([[-0.2030, -0.0393, 0.1953, ..., 0.0427, -0.1798, -0.1411], [-0.1357, -0.2031, -0.0417, ..., -0.1034, -0.1966, -0.0774], [ 0.2347, -0.0658, -0.2207, ..., -0.1708, -0.1765, -0.2019], ..., [-0.0780, -0.0260, -0.2564, ..., -0.1106, 0.1066, 0.0285], [-0.1806, -0.1180, -0.0309, ..., -0.1862, -0.0328, -0.0309], [-0.0736, -0.0933, -0.1266, ..., 0.0941, 0.0282, -0.1575]], device='cuda:0'), grad: tensor([[ 1.8626e-08, 9.3132e-10, -1.5926e-07, ..., -1.4249e-07, 1.7695e-08, 7.6368e-08], [ 1.3504e-07, 1.8626e-09, 1.0245e-08, ..., 1.9558e-08, 3.7253e-09, 2.6077e-08], [-1.8522e-05, 0.0000e+00, 3.1665e-08, ..., 2.5146e-08, 2.7940e-09, 9.3132e-09], ..., [ 9.6858e-08, 3.7253e-09, 1.0245e-08, ..., 1.4435e-07, 1.8626e-09, 1.8626e-09], [ 1.8239e-05, 1.4901e-08, 3.8184e-08, ..., 2.8871e-08, -1.8626e-09, -1.5460e-07], [ 2.7940e-09, 1.0151e-07, 1.7788e-07, ..., -4.4890e-07, 5.0291e-08, 9.3132e-10]], device='cuda:0') Epoch 240, bias, value: tensor([ 0.0040, 0.0156, 0.0153, -0.0024, 0.0227, -0.0030, -0.0088, -0.0073, 0.0116, -0.0168], device='cuda:0'), grad: tensor([-3.4459e-08, 4.9360e-07, -3.2693e-05, 7.0687e-07, 6.6590e-07, -1.4100e-06, 1.7323e-07, 4.5449e-07, 3.1143e-05, 5.5041e-07], device='cuda:0') 100 0.0001 changing lr epoch 239, time 248.58, cls_loss 0.0006 cls_loss_mapping 0.0039 cls_loss_causal 0.4939 re_mapping 0.0038 re_causal 0.0125 /// teacc 99.13 lr 0.00010000 Epoch 241, weight, value: tensor([[-0.2037, -0.0394, 0.1952, ..., 0.0427, -0.1801, -0.1413], [-0.1362, -0.2046, -0.0417, ..., -0.1035, -0.1972, -0.0778], [ 0.2352, -0.0658, -0.2221, ..., -0.1715, -0.1767, -0.2021], ..., [-0.0782, -0.0260, -0.2578, ..., -0.1108, 0.1066, 0.0285], [-0.1804, -0.1183, -0.0309, ..., -0.1859, -0.0329, -0.0313], [-0.0739, -0.0934, -0.1267, ..., 0.0939, 0.0280, -0.1576]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 1.8626e-09, -1.8161e-07, ..., -2.7008e-08, 1.8626e-09, 1.8626e-09], [ 4.6566e-09, 5.5879e-09, 1.1176e-08, ..., 3.7253e-08, 1.1176e-08, 0.0000e+00], [ 2.7940e-09, 2.7940e-09, 1.8626e-09, ..., 1.3039e-08, 9.3132e-10, 0.0000e+00], ..., [ 1.4901e-08, 1.3039e-08, 2.7940e-09, ..., 1.0338e-07, 1.4901e-08, 0.0000e+00], [ 5.5879e-09, 5.5879e-09, 1.8626e-09, ..., 1.3504e-07, 3.7253e-09, 0.0000e+00], [ 1.5832e-08, 3.2596e-08, 1.5832e-08, ..., -8.3353e-07, -1.3970e-07, 0.0000e+00]], device='cuda:0') Epoch 241, bias, value: tensor([ 0.0040, 0.0156, 0.0153, -0.0026, 0.0229, -0.0031, -0.0085, -0.0073, 0.0117, -0.0170], device='cuda:0'), grad: tensor([-1.6484e-07, 1.0245e-07, 5.2154e-08, -9.2108e-07, -1.1884e-06, 3.2112e-06, 3.7625e-07, 2.8592e-07, 2.9802e-07, -2.0489e-06], device='cuda:0') 100 0.0001 changing lr epoch 240, time 248.44, cls_loss 0.0007 cls_loss_mapping 0.0021 cls_loss_causal 0.5154 re_mapping 0.0037 re_causal 0.0120 /// teacc 99.09 lr 0.00010000 Epoch 242, weight, value: tensor([[-0.2054, -0.0397, 0.1953, ..., 0.0427, -0.1804, -0.1415], [-0.1366, -0.2064, -0.0418, ..., -0.1036, -0.1982, -0.0810], [ 0.2364, -0.0658, -0.2225, ..., -0.1718, -0.1770, -0.2024], ..., [-0.0802, -0.0261, -0.2589, ..., -0.1109, 0.1067, 0.0285], [-0.1806, -0.1189, -0.0310, ..., -0.1861, -0.0329, -0.0317], [-0.0741, -0.0935, -0.1267, ..., 0.0939, 0.0280, -0.1577]], device='cuda:0'), grad: tensor([[ 1.3039e-08, 9.3132e-09, -2.9802e-07, ..., -1.7043e-07, 5.5879e-09, 6.5193e-09], [ 3.4273e-07, 6.6124e-08, 2.4214e-08, ..., 5.5879e-09, 1.6764e-08, 2.7940e-09], [-4.0699e-07, -5.4948e-08, 1.5832e-08, ..., 6.5193e-09, 1.2107e-08, 9.3132e-10], ..., [ 4.5635e-08, 1.7695e-08, 5.5879e-09, ..., 9.3132e-10, -4.6566e-08, 1.8626e-09], [ 1.5832e-08, -1.6764e-08, -4.3772e-08, ..., -1.3039e-08, 3.7253e-09, 9.3132e-09], [ 6.5193e-09, 1.6764e-08, 2.7567e-07, ..., 1.4808e-07, 9.3132e-10, 6.5193e-09]], device='cuda:0') Epoch 242, bias, value: tensor([ 0.0039, 0.0148, 0.0161, -0.0020, 0.0231, -0.0035, -0.0073, -0.0076, 0.0116, -0.0171], device='cuda:0'), grad: tensor([-2.9802e-07, -3.5651e-06, 8.1956e-07, -6.3609e-07, 1.3132e-07, -1.8254e-07, 7.5344e-07, 2.2314e-06, -6.0536e-08, 8.2143e-07], device='cuda:0') 100 0.0001 changing lr epoch 241, time 248.46, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4886 re_mapping 0.0036 re_causal 0.0114 /// teacc 99.17 lr 0.00010000 Epoch 243, weight, value: tensor([[-0.2055, -0.0398, 0.1956, ..., 0.0428, -0.1807, -0.1418], [-0.1370, -0.2076, -0.0418, ..., -0.1036, -0.1992, -0.0811], [ 0.2369, -0.0659, -0.2233, ..., -0.1721, -0.1772, -0.2028], ..., [-0.0805, -0.0261, -0.2600, ..., -0.1110, 0.1068, 0.0285], [-0.1807, -0.1192, -0.0313, ..., -0.1864, -0.0330, -0.0322], [-0.0743, -0.0935, -0.1268, ..., 0.0939, 0.0281, -0.1580]], device='cuda:0'), grad: tensor([[ 9.3132e-09, 0.0000e+00, -1.6298e-07, ..., -7.5437e-08, 9.3132e-10, 1.8626e-09], [ 1.3877e-07, 1.8626e-09, 2.5146e-08, ..., 4.6473e-07, 2.7940e-09, 0.0000e+00], [-2.5984e-07, 1.8626e-09, 7.4506e-09, ..., 4.3772e-08, 9.3132e-10, 9.3132e-10], ..., [ 5.5879e-08, 9.3132e-09, 6.5193e-09, ..., 8.3819e-09, 1.8626e-09, 0.0000e+00], [ 1.8626e-08, -0.0000e+00, 8.3819e-09, ..., 2.0489e-08, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 2.7940e-09, 5.4948e-08, ..., 1.1642e-07, -1.3970e-08, 0.0000e+00]], device='cuda:0') Epoch 243, bias, value: tensor([ 0.0040, 0.0147, 0.0161, 0.0009, 0.0230, -0.0052, -0.0071, -0.0076, 0.0115, -0.0171], device='cuda:0'), grad: tensor([-1.3132e-07, 1.4119e-06, 2.0675e-07, 1.8440e-07, -9.4157e-07, 1.9278e-07, 2.2817e-07, -1.1632e-06, -7.6927e-07, 7.7952e-07], device='cuda:0') 100 0.0001 changing lr epoch 242, time 248.54, cls_loss 0.0006 cls_loss_mapping 0.0024 cls_loss_causal 0.4717 re_mapping 0.0037 re_causal 0.0118 /// teacc 99.03 lr 0.00010000 Epoch 244, weight, value: tensor([[-0.2059, -0.0399, 0.1957, ..., 0.0428, -0.1810, -0.1424], [-0.1377, -0.2087, -0.0418, ..., -0.1039, -0.2007, -0.0811], [ 0.2375, -0.0659, -0.2246, ..., -0.1728, -0.1777, -0.2031], ..., [-0.0809, -0.0262, -0.2618, ..., -0.1113, 0.1068, 0.0284], [-0.1808, -0.1192, -0.0314, ..., -0.1866, -0.0331, -0.0323], [-0.0744, -0.0935, -0.1269, ..., 0.0939, 0.0282, -0.1581]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.5146e-08, 1.8533e-07, ..., 0.0000e+00, 5.5879e-09, 3.2596e-08], [ 1.8626e-09, 1.7695e-08, 1.2200e-07, ..., 7.4506e-09, 8.3819e-09, 3.3975e-06], [-9.3132e-10, 1.5832e-08, 1.1642e-07, ..., 1.8626e-09, 6.5193e-09, 3.0380e-06], ..., [ 9.3132e-10, 9.3132e-10, 3.7253e-09, ..., 2.7940e-09, -1.1176e-08, 1.5832e-08], [ 3.7253e-09, 3.7253e-09, 2.3283e-08, ..., 9.3132e-10, 1.8626e-09, 1.2759e-07], [ 9.3132e-10, 1.8626e-09, 1.0245e-08, ..., 6.3330e-08, 1.8626e-09, 6.5193e-09]], device='cuda:0') Epoch 244, bias, value: tensor([ 0.0040, 0.0146, 0.0162, 0.0007, 0.0231, -0.0052, -0.0070, -0.0077, 0.0117, -0.0170], device='cuda:0'), grad: tensor([ 1.5963e-06, 1.8939e-05, 1.6943e-05, 4.0699e-07, 3.4347e-06, 3.6582e-06, -4.6164e-05, 5.5879e-08, 8.5961e-07, 2.9150e-07], device='cuda:0') 100 0.0001 changing lr epoch 243, time 248.46, cls_loss 0.0007 cls_loss_mapping 0.0025 cls_loss_causal 0.5240 re_mapping 0.0036 re_causal 0.0123 /// teacc 99.11 lr 0.00010000 Epoch 245, weight, value: tensor([[-0.2061, -0.0401, 0.1958, ..., 0.0425, -0.1812, -0.1424], [-0.1383, -0.2094, -0.0416, ..., -0.1045, -0.2017, -0.0813], [ 0.2380, -0.0659, -0.2254, ..., -0.1731, -0.1780, -0.2039], ..., [-0.0809, -0.0262, -0.2626, ..., -0.1110, 0.1069, 0.0285], [-0.1809, -0.1193, -0.0315, ..., -0.1869, -0.0331, -0.0323], [-0.0745, -0.0935, -0.1270, ..., 0.0941, 0.0281, -0.1580]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 9.3132e-10, -9.8720e-08, ..., -5.1223e-08, 9.3132e-10, 9.3132e-10], [ 1.8626e-09, 2.7940e-09, 8.3819e-09, ..., 4.6566e-09, 1.3970e-08, 9.3132e-10], [-7.4506e-09, 9.3132e-10, 7.4506e-09, ..., 5.5879e-09, 9.3132e-10, 0.0000e+00], ..., [ 6.5193e-09, 3.0734e-08, 1.1176e-08, ..., 1.0245e-08, -2.3283e-08, 0.0000e+00], [ 0.0000e+00, 9.3132e-10, 6.5193e-09, ..., 4.6566e-09, 9.3132e-10, 9.3132e-10], [ 0.0000e+00, 2.1420e-08, 3.9116e-08, ..., 6.1467e-08, 4.6566e-09, 0.0000e+00]], device='cuda:0') Epoch 245, bias, value: tensor([ 0.0037, 0.0137, 0.0160, 0.0017, 0.0233, -0.0059, -0.0068, -0.0069, 0.0118, -0.0169], device='cuda:0'), grad: tensor([-1.6484e-07, 1.1176e-07, 1.4901e-08, -3.5018e-07, -6.7987e-08, 8.4750e-08, 1.3970e-08, 6.9849e-08, 2.7008e-08, 2.5798e-07], device='cuda:0') 100 0.0001 changing lr epoch 244, time 248.12, cls_loss 0.0005 cls_loss_mapping 0.0019 cls_loss_causal 0.4723 re_mapping 0.0039 re_causal 0.0125 /// teacc 99.10 lr 0.00010000 Epoch 246, weight, value: tensor([[-0.2048, -0.0403, 0.1973, ..., 0.0434, -0.1818, -0.1444], [-0.1386, -0.2098, -0.0416, ..., -0.1045, -0.2021, -0.0813], [ 0.2385, -0.0659, -0.2269, ..., -0.1739, -0.1781, -0.2042], ..., [-0.0813, -0.0262, -0.2633, ..., -0.1110, 0.1070, 0.0285], [-0.1811, -0.1193, -0.0309, ..., -0.1866, -0.0332, -0.0325], [-0.0751, -0.0936, -0.1290, ..., 0.0934, 0.0281, -0.1587]], device='cuda:0'), grad: tensor([[ 9.3132e-09, 5.5879e-09, -2.1420e-08, ..., 9.3132e-10, 3.7253e-09, 9.3132e-10], [ 1.8161e-07, 3.1665e-08, -9.3132e-10, ..., 1.6764e-08, 2.2352e-08, 9.3132e-10], [-3.0547e-07, 2.5146e-08, 2.7940e-09, ..., 8.3819e-09, 4.6566e-09, 1.8626e-09], ..., [ 5.5879e-08, 8.4750e-08, 9.3132e-10, ..., -3.7253e-09, -6.3330e-08, 9.3132e-10], [ 2.2352e-08, 1.9558e-08, 1.8626e-09, ..., 9.3132e-10, 9.3132e-10, 9.3132e-10], [ 2.7940e-09, 3.7253e-09, 7.4506e-09, ..., 9.3132e-10, 2.1420e-08, 1.8626e-09]], device='cuda:0') Epoch 246, bias, value: tensor([ 0.0046, 0.0137, 0.0162, 0.0016, 0.0233, -0.0059, -0.0064, -0.0070, 0.0121, -0.0177], device='cuda:0'), grad: tensor([ 7.7300e-08, 7.1898e-07, -5.8953e-07, -1.1735e-06, -1.6764e-08, 3.7905e-07, 2.5332e-07, -1.4901e-07, 1.7416e-07, 3.3900e-07], device='cuda:0') 100 0.0001 changing lr epoch 245, time 248.64, cls_loss 0.0006 cls_loss_mapping 0.0025 cls_loss_causal 0.5059 re_mapping 0.0038 re_causal 0.0128 /// teacc 99.08 lr 0.00010000 Epoch 247, weight, value: tensor([[-0.2046, -0.0407, 0.1982, ..., 0.0437, -0.1824, -0.1449], [-0.1388, -0.2106, -0.0417, ..., -0.1047, -0.2035, -0.0814], [ 0.2387, -0.0659, -0.2285, ..., -0.1747, -0.1785, -0.2045], ..., [-0.0815, -0.0263, -0.2646, ..., -0.1110, 0.1071, 0.0285], [-0.1812, -0.1197, -0.0305, ..., -0.1868, -0.0333, -0.0328], [-0.0753, -0.0936, -0.1300, ..., 0.0932, 0.0281, -0.1590]], device='cuda:0'), grad: tensor([[ 1.0245e-08, 0.0000e+00, -1.2917e-06, ..., -5.0198e-07, 1.8626e-09, -9.3132e-10], [ 1.8626e-09, 9.3132e-10, 6.4261e-08, ..., 5.9605e-08, 5.6811e-08, 1.2107e-08], [-5.1223e-08, 0.0000e+00, 1.0990e-07, ..., 8.5682e-08, 8.3819e-09, 1.3039e-08], ..., [ 7.4506e-09, 9.3132e-10, 1.1735e-07, ..., 9.1270e-08, -7.7300e-08, 4.6566e-09], [ 1.8626e-08, 9.3132e-10, 1.3411e-07, ..., 1.1828e-07, 4.4703e-08, 6.0536e-08], [ 0.0000e+00, 9.3132e-10, 3.0175e-07, ..., 1.0058e-07, 1.8626e-08, 1.4901e-08]], device='cuda:0') Epoch 247, bias, value: tensor([ 0.0050, 0.0136, 0.0160, 0.0013, 0.0233, -0.0057, -0.0063, -0.0069, 0.0124, -0.0181], device='cuda:0'), grad: tensor([-2.4028e-06, 1.5153e-06, 3.0547e-07, 3.3807e-07, -9.4343e-07, -7.0687e-07, 1.8161e-06, -1.5236e-06, 5.1595e-07, 1.0785e-06], device='cuda:0') 100 0.0001 changing lr epoch 246, time 248.41, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4806 re_mapping 0.0036 re_causal 0.0115 /// teacc 99.14 lr 0.00010000 Epoch 248, weight, value: tensor([[-0.2046, -0.0409, 0.1985, ..., 0.0438, -0.1826, -0.1450], [-0.1390, -0.2114, -0.0420, ..., -0.1049, -0.2042, -0.0816], [ 0.2405, -0.0659, -0.2295, ..., -0.1756, -0.1789, -0.2047], ..., [-0.0818, -0.0263, -0.2677, ..., -0.1111, 0.1071, 0.0284], [-0.1838, -0.1199, -0.0304, ..., -0.1873, -0.0337, -0.0334], [-0.0754, -0.0936, -0.1301, ..., 0.0931, 0.0282, -0.1593]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, -8.3521e-06, ..., -5.0776e-06, -9.3132e-10, 1.8626e-09], [ 1.8626e-09, 9.3132e-10, 2.8964e-07, ..., 1.9558e-07, 2.6077e-08, 9.3132e-10], [-1.3970e-08, 9.3132e-10, 1.5646e-07, ..., 9.8720e-08, 1.8626e-09, 0.0000e+00], ..., [ 3.7253e-09, 9.3132e-10, 3.1106e-07, ..., 2.1234e-07, -4.8429e-08, 0.0000e+00], [ 8.3819e-09, 3.7253e-09, 1.9465e-07, ..., 1.0896e-07, 1.8626e-09, 1.8626e-09], [ 0.0000e+00, 0.0000e+00, 3.3416e-06, ..., 2.2296e-06, 8.3819e-09, 9.3132e-10]], device='cuda:0') Epoch 248, bias, value: tensor([ 0.0051, 0.0131, 0.0170, 0.0010, 0.0235, -0.0054, -0.0064, -0.0065, 0.0114, -0.0182], device='cuda:0'), grad: tensor([-2.3499e-05, 1.0105e-06, 4.5449e-07, 1.3355e-06, 1.0710e-07, 4.9695e-06, 4.3400e-06, 6.7614e-07, 5.8021e-07, 1.0036e-05], device='cuda:0') 100 0.0001 changing lr epoch 247, time 248.47, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.4942 re_mapping 0.0036 re_causal 0.0115 /// teacc 99.12 lr 0.00010000 Epoch 249, weight, value: tensor([[-0.2047, -0.0411, 0.1987, ..., 0.0438, -0.1834, -0.1454], [-0.1394, -0.2120, -0.0424, ..., -0.1051, -0.2049, -0.0816], [ 0.2408, -0.0660, -0.2311, ..., -0.1762, -0.1794, -0.2051], ..., [-0.0817, -0.0263, -0.2717, ..., -0.1113, 0.1072, 0.0284], [-0.1839, -0.1204, -0.0308, ..., -0.1882, -0.0340, -0.0338], [-0.0755, -0.0934, -0.1302, ..., 0.0933, 0.0282, -0.1597]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 1.8626e-09, 9.4995e-08, ..., -4.0978e-08, 5.6811e-08, 9.0338e-08], [ 1.8626e-09, 5.5879e-09, 1.6764e-08, ..., 1.7695e-08, 7.8231e-08, 8.3819e-09], [-2.7940e-09, 4.6566e-09, 1.0245e-08, ..., 6.5193e-09, 1.8626e-08, 2.7940e-09], ..., [ 3.7253e-09, 1.4901e-08, 9.3132e-09, ..., 0.0000e+00, -3.5856e-07, 0.0000e+00], [ 2.7940e-09, 1.1176e-08, 4.5635e-08, ..., 2.1420e-08, 2.9802e-08, 2.4214e-08], [ 9.3132e-10, 2.7940e-09, 3.6322e-08, ..., -0.0000e+00, 2.3749e-07, 9.3132e-10]], device='cuda:0') Epoch 249, bias, value: tensor([ 0.0050, 0.0132, 0.0168, 0.0005, 0.0238, -0.0055, -0.0060, -0.0067, 0.0113, -0.0182], device='cuda:0'), grad: tensor([ 3.0734e-07, -1.2837e-05, 1.8906e-07, -2.6263e-07, 2.7008e-08, 3.8370e-07, -6.6310e-07, 1.1638e-05, 2.6356e-07, 9.7137e-07], device='cuda:0') 100 0.0001 changing lr epoch 248, time 247.42, cls_loss 0.0007 cls_loss_mapping 0.0027 cls_loss_causal 0.4820 re_mapping 0.0037 re_causal 0.0115 /// teacc 99.13 lr 0.00010000 Epoch 250, weight, value: tensor([[-0.2051, -0.0415, 0.1979, ..., 0.0438, -0.1857, -0.1486], [-0.1402, -0.2131, -0.0436, ..., -0.1053, -0.2060, -0.0817], [ 0.2412, -0.0661, -0.2313, ..., -0.1765, -0.1805, -0.2054], ..., [-0.0818, -0.0263, -0.2733, ..., -0.1114, 0.1074, 0.0283], [-0.1839, -0.1202, -0.0312, ..., -0.1899, -0.0352, -0.0350], [-0.0755, -0.0934, -0.1303, ..., 0.0934, 0.0284, -0.1601]], device='cuda:0'), grad: tensor([[ 8.3819e-09, 9.3132e-10, -3.1386e-07, ..., -1.0245e-08, 9.3132e-09, -3.4459e-08], [ 1.2621e-05, 6.5193e-09, 4.6566e-09, ..., 3.9116e-08, 1.7695e-07, 9.3132e-10], [ 3.5390e-08, -1.3039e-08, 3.7253e-09, ..., 5.5879e-09, 5.5879e-09, 9.3132e-10], ..., [-1.2726e-05, 8.3819e-09, 1.8626e-09, ..., 2.7940e-08, -2.0768e-07, 0.0000e+00], [ 3.7253e-09, 1.2107e-08, 8.3819e-09, ..., 1.7043e-07, 4.0978e-08, 1.8626e-09], [ 4.6566e-09, -3.0734e-08, 1.5832e-08, ..., -8.9314e-07, -1.5087e-07, 1.8626e-09]], device='cuda:0') Epoch 250, bias, value: tensor([ 0.0045, 0.0130, 0.0169, 0.0005, 0.0237, -0.0057, -0.0049, -0.0067, 0.0103, -0.0179], device='cuda:0'), grad: tensor([-2.7288e-07, 1.0926e-04, 1.5842e-06, 7.6462e-07, 1.5842e-06, 2.0023e-07, 3.1479e-07, -1.1122e-04, 4.7777e-07, -2.5723e-06], device='cuda:0') 100 0.0001 changing lr epoch 249, time 246.71, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4779 re_mapping 0.0039 re_causal 0.0120 /// teacc 99.13 lr 0.00010000 Epoch 251, weight, value: tensor([[-0.2052, -0.0417, 0.1980, ..., 0.0437, -0.1860, -0.1486], [-0.1411, -0.2135, -0.0443, ..., -0.1056, -0.2080, -0.0818], [ 0.2415, -0.0661, -0.2317, ..., -0.1769, -0.1811, -0.2055], ..., [-0.0816, -0.0263, -0.2742, ..., -0.1117, 0.1075, 0.0280], [-0.1840, -0.1201, -0.0313, ..., -0.1903, -0.0351, -0.0351], [-0.0752, -0.0935, -0.1303, ..., 0.0936, 0.0285, -0.1604]], device='cuda:0'), grad: tensor([[ 1.2107e-08, 1.6764e-08, -6.5193e-09, ..., 1.3039e-08, 5.5879e-09, 9.3132e-10], [-2.5146e-08, 4.6566e-09, 0.0000e+00, ..., -4.4703e-08, 4.9360e-08, 0.0000e+00], [ 2.7008e-08, 2.7008e-08, 9.3132e-10, ..., 2.7940e-08, 1.3970e-08, 9.3132e-10], ..., [ 2.1420e-08, 1.4901e-08, 0.0000e+00, ..., 1.4808e-07, -1.3411e-07, 0.0000e+00], [ 1.2107e-08, 1.5832e-08, 9.3132e-10, ..., 5.5879e-09, 1.8626e-09, 0.0000e+00], [ 1.6764e-08, 1.8626e-08, 2.7940e-09, ..., 2.4214e-08, 5.0291e-08, 9.3132e-10]], device='cuda:0') Epoch 251, bias, value: tensor([ 0.0044, 0.0125, 0.0168, 0.0003, 0.0236, -0.0053, -0.0052, -0.0065, 0.0104, -0.0177], device='cuda:0'), grad: tensor([ 2.8871e-07, -3.2596e-07, 6.2957e-07, 2.8498e-06, -4.7870e-07, -4.3698e-06, 4.8708e-07, 6.7987e-08, 2.2817e-07, 6.3609e-07], device='cuda:0') 100 0.0001 changing lr epoch 250, time 246.83, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4668 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.17 lr 0.00010000 Epoch 252, weight, value: tensor([[-0.2055, -0.0420, 0.1980, ..., 0.0437, -0.1864, -0.1487], [-0.1435, -0.2144, -0.0440, ..., -0.1054, -0.2086, -0.0818], [ 0.2427, -0.0661, -0.2320, ..., -0.1774, -0.1815, -0.2057], ..., [-0.0819, -0.0263, -0.2752, ..., -0.1118, 0.1076, 0.0280], [-0.1841, -0.1200, -0.0308, ..., -0.1910, -0.0350, -0.0352], [-0.0752, -0.0935, -0.1303, ..., 0.0935, 0.0285, -0.1605]], device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, -5.3551e-08, ..., -8.3819e-09, -4.6566e-10, 3.7253e-09], [ 7.4971e-08, 4.6566e-10, 2.3283e-09, ..., 1.1688e-07, 4.1444e-08, 9.3132e-10], [-1.9511e-07, 1.8626e-09, 1.8626e-09, ..., 6.4261e-08, 4.4703e-08, 0.0000e+00], ..., [ 9.1270e-08, 4.6566e-10, 1.8626e-09, ..., 6.1933e-08, -1.0012e-07, 0.0000e+00], [ 8.8476e-09, -6.9849e-09, 2.0489e-08, ..., 3.6787e-08, 1.0710e-08, 1.8626e-09], [ 1.3970e-09, 4.6566e-10, 1.7229e-08, ..., 5.6028e-06, 1.6252e-07, 0.0000e+00]], device='cuda:0') Epoch 252, bias, value: tensor([ 0.0044, 0.0126, 0.0174, 0.0001, 0.0238, -0.0054, -0.0053, -0.0065, 0.0107, -0.0179], device='cuda:0'), grad: tensor([ 6.1933e-08, 8.6799e-07, 3.2829e-07, 1.9791e-07, -1.5363e-05, 9.5926e-08, 2.8778e-07, -1.0263e-06, -1.6531e-07, 1.4678e-05], device='cuda:0') 100 0.0001 changing lr epoch 251, time 245.88, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4886 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.15 lr 0.00010000 Epoch 253, weight, value: tensor([[-0.2060, -0.0427, 0.1980, ..., 0.0438, -0.1881, -0.1497], [-0.1443, -0.2166, -0.0440, ..., -0.1054, -0.2093, -0.0818], [ 0.2434, -0.0667, -0.2327, ..., -0.1780, -0.1826, -0.2062], ..., [-0.0823, -0.0267, -0.2766, ..., -0.1120, 0.1077, 0.0281], [-0.1843, -0.1207, -0.0311, ..., -0.1916, -0.0352, -0.0355], [-0.0750, -0.0935, -0.1307, ..., 0.0935, 0.0285, -0.1607]], device='cuda:0'), grad: tensor([[ 6.5193e-09, 2.7940e-09, -1.9465e-07, ..., -1.7788e-07, 3.5390e-08, 4.0978e-08], [ 7.5903e-07, 1.8626e-09, 2.1420e-08, ..., 7.1712e-08, 1.7695e-08, 2.7940e-09], [-8.6613e-07, 9.3132e-10, 9.9652e-08, ..., 7.7300e-08, 5.5879e-09, 1.8626e-09], ..., [ 5.0291e-08, 1.5832e-08, 1.7695e-08, ..., 1.3039e-07, -2.3283e-08, 0.0000e+00], [ 6.5193e-09, 9.3132e-10, 4.6566e-09, ..., 5.9605e-08, 6.5193e-09, 3.7253e-09], [ 1.8626e-09, -3.6322e-08, 3.6322e-08, ..., -3.2503e-07, -5.5879e-09, 0.0000e+00]], device='cuda:0') Epoch 253, bias, value: tensor([ 0.0043, 0.0125, 0.0174, 0.0007, 0.0238, -0.0056, -0.0047, -0.0066, 0.0104, -0.0179], device='cuda:0'), grad: tensor([-2.7567e-07, 2.3879e-06, -2.1495e-06, 2.6636e-07, 1.2852e-07, 5.1968e-07, 2.5425e-07, 3.0547e-07, -1.1120e-06, -3.2689e-07], device='cuda:0') 100 0.0001 changing lr epoch 252, time 246.14, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.5114 re_mapping 0.0035 re_causal 0.0119 /// teacc 99.11 lr 0.00010000 Epoch 254, weight, value: tensor([[-0.2062, -0.0433, 0.1981, ..., 0.0437, -0.1884, -0.1497], [-0.1446, -0.2176, -0.0442, ..., -0.1055, -0.2114, -0.0819], [ 0.2437, -0.0667, -0.2336, ..., -0.1788, -0.1834, -0.2065], ..., [-0.0824, -0.0268, -0.2776, ..., -0.1122, 0.1080, 0.0281], [-0.1844, -0.1217, -0.0313, ..., -0.1919, -0.0352, -0.0356], [-0.0750, -0.0935, -0.1307, ..., 0.0937, 0.0285, -0.1608]], device='cuda:0'), grad: tensor([[-1.9558e-08, 0.0000e+00, -9.9093e-07, ..., -4.0419e-07, 9.3132e-09, 4.6566e-09], [ 2.7940e-09, 9.3132e-10, 1.3970e-08, ..., 1.3132e-07, 2.3283e-08, 1.8626e-09], [-6.5193e-09, 1.8626e-09, 8.9407e-08, ..., 1.6112e-07, 7.4506e-09, 1.8626e-09], ..., [ 1.6764e-08, 3.7253e-09, 1.5832e-08, ..., 4.4145e-07, 1.0338e-07, 0.0000e+00], [ 4.6566e-09, 9.3132e-10, 2.7008e-08, ..., 6.3330e-08, 0.0000e+00, 9.3132e-10], [ 3.7253e-09, -4.6566e-09, 1.3690e-07, ..., -1.3132e-07, -1.5646e-07, 0.0000e+00]], device='cuda:0') Epoch 254, bias, value: tensor([ 0.0042, 0.0124, 0.0172, 0.0006, 0.0237, -0.0056, -0.0045, -0.0063, 0.0102, -0.0179], device='cuda:0'), grad: tensor([-1.8338e-06, -4.8429e-08, 6.5379e-07, 1.9278e-07, -4.0531e-06, 1.4715e-07, 3.3211e-06, 1.3215e-06, -1.0151e-07, 3.9767e-07], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 253---------------------------------------------------- epoch 253, time 262.94, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4621 re_mapping 0.0037 re_causal 0.0115 /// teacc 99.20 lr 0.00010000 Epoch 255, weight, value: tensor([[-0.2065, -0.0437, 0.1983, ..., 0.0437, -0.1886, -0.1497], [-0.1452, -0.2182, -0.0443, ..., -0.1056, -0.2120, -0.0819], [ 0.2443, -0.0668, -0.2344, ..., -0.1793, -0.1836, -0.2067], ..., [-0.0828, -0.0268, -0.2798, ..., -0.1124, 0.1080, 0.0281], [-0.1846, -0.1219, -0.0315, ..., -0.1921, -0.0349, -0.0356], [-0.0753, -0.0939, -0.1308, ..., 0.0937, 0.0286, -0.1608]], device='cuda:0'), grad: tensor([[-2.6077e-08, 0.0000e+00, -8.0559e-07, ..., -4.1910e-07, 2.7940e-09, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 1.6764e-08, 2.6356e-07, 0.0000e+00], [ 1.3970e-08, 0.0000e+00, 6.7055e-08, ..., 4.6566e-08, 2.7940e-09, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 1.3039e-08, -3.1386e-07, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 6.5193e-08, ..., 7.9162e-08, 1.3970e-08, 9.3132e-10], [ 6.5193e-09, -0.0000e+00, 5.0943e-07, ..., 1.3318e-07, 1.3039e-08, 0.0000e+00]], device='cuda:0') Epoch 255, bias, value: tensor([ 0.0043, 0.0124, 0.0176, 0.0012, 0.0237, -0.0059, -0.0047, -0.0066, 0.0105, -0.0179], device='cuda:0'), grad: tensor([-1.5106e-06, -1.9837e-07, 6.9011e-07, 2.2538e-07, 2.1141e-07, 5.5879e-08, 1.7136e-07, -5.5321e-07, 3.2037e-07, 6.1560e-07], device='cuda:0') 100 0.0001 changing lr epoch 254, time 246.02, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4952 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.20 lr 0.00010000 Epoch 256, weight, value: tensor([[-0.2063, -0.0438, 0.1987, ..., 0.0438, -0.1891, -0.1499], [-0.1457, -0.2192, -0.0450, ..., -0.1056, -0.2142, -0.0819], [ 0.2448, -0.0670, -0.2356, ..., -0.1800, -0.1854, -0.2068], ..., [-0.0831, -0.0267, -0.2809, ..., -0.1126, 0.1083, 0.0280], [-0.1848, -0.1221, -0.0316, ..., -0.1925, -0.0351, -0.0357], [-0.0753, -0.0939, -0.1310, ..., 0.0939, 0.0287, -0.1609]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 1.8626e-09, -7.5437e-08, ..., -4.2841e-08, 8.3819e-09, 4.6566e-09], [ 7.4506e-09, 3.7253e-09, 5.5879e-09, ..., 8.3819e-09, 1.8626e-09, 9.3132e-10], [-6.5193e-09, 6.5193e-09, 5.5879e-09, ..., 8.3819e-09, 9.3132e-10, 0.0000e+00], ..., [ 1.3039e-08, 9.3132e-09, 0.0000e+00, ..., 3.7253e-09, 9.3132e-10, 0.0000e+00], [ 1.3970e-08, 5.5879e-09, 1.1176e-08, ..., 9.3132e-09, 2.7940e-09, 1.8626e-09], [ 6.5193e-09, 5.5879e-09, 1.4901e-08, ..., 1.6671e-07, 1.4901e-08, 0.0000e+00]], device='cuda:0') Epoch 256, bias, value: tensor([ 0.0045, 0.0124, 0.0175, 0.0010, 0.0235, -0.0058, -0.0048, -0.0066, 0.0108, -0.0178], device='cuda:0'), grad: tensor([-9.0338e-08, 1.8626e-09, 2.3283e-08, -2.0489e-07, -4.3213e-07, 4.0047e-08, 1.2852e-07, 8.1025e-08, 7.3574e-08, 3.8557e-07], device='cuda:0') 100 0.0001 changing lr epoch 255, time 246.14, cls_loss 0.0006 cls_loss_mapping 0.0020 cls_loss_causal 0.4706 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.17 lr 0.00010000 Epoch 257, weight, value: tensor([[-0.2070, -0.0443, 0.1987, ..., 0.0438, -0.1902, -0.1500], [-0.1461, -0.2201, -0.0452, ..., -0.1056, -0.2163, -0.0819], [ 0.2452, -0.0674, -0.2361, ..., -0.1804, -0.1861, -0.2070], ..., [-0.0835, -0.0267, -0.2814, ..., -0.1129, 0.1085, 0.0280], [-0.1849, -0.1224, -0.0318, ..., -0.1928, -0.0354, -0.0359], [-0.0756, -0.0939, -0.1310, ..., 0.0939, 0.0286, -0.1609]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -2.4214e-08, ..., -1.3970e-08, 4.6566e-09, 0.0000e+00], [ 5.5879e-09, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, 7.3574e-08, 0.0000e+00], [-1.9372e-07, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, -1.8626e-09, 0.0000e+00], ..., [ 1.7043e-07, 0.0000e+00, 1.8626e-09, ..., 1.0245e-08, -3.7346e-07, 0.0000e+00], [ 3.7253e-09, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, 9.3132e-10, 0.0000e+00], [ 9.3132e-10, 9.3132e-10, 1.2107e-08, ..., 1.9558e-07, 2.7940e-07, 0.0000e+00]], device='cuda:0') Epoch 257, bias, value: tensor([ 0.0043, 0.0127, 0.0143, 0.0011, 0.0235, -0.0057, -0.0048, -0.0055, 0.0107, -0.0179], device='cuda:0'), grad: tensor([ 1.0468e-06, -1.2606e-05, -3.0454e-07, 3.9395e-07, -5.5321e-07, 2.9989e-07, 3.1292e-07, 1.2498e-06, 8.4564e-06, 1.6829e-06], device='cuda:0') 100 0.0001 changing lr epoch 256, time 246.01, cls_loss 0.0009 cls_loss_mapping 0.0023 cls_loss_causal 0.4805 re_mapping 0.0034 re_causal 0.0111 /// teacc 99.13 lr 0.00010000 Epoch 258, weight, value: tensor([[-0.2074, -0.0445, 0.1983, ..., 0.0438, -0.1934, -0.1513], [-0.1466, -0.2206, -0.0454, ..., -0.1057, -0.2175, -0.0820], [ 0.2456, -0.0674, -0.2373, ..., -0.1812, -0.1867, -0.2071], ..., [-0.0835, -0.0267, -0.2830, ..., -0.1132, 0.1086, 0.0280], [-0.1850, -0.1225, -0.0320, ..., -0.1930, -0.0358, -0.0362], [-0.0743, -0.0939, -0.1310, ..., 0.0950, 0.0294, -0.1612]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 9.3132e-10, 1.8626e-09, 9.3132e-10], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 1.2107e-08, 9.3132e-10, 0.0000e+00], [-6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, 9.3132e-10, 0.0000e+00], [ 4.6566e-09, 0.0000e+00, 1.8626e-09, ..., 1.9558e-08, 7.4506e-09, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -1.7695e-08, -1.8626e-09, 0.0000e+00]], device='cuda:0') Epoch 258, bias, value: tensor([ 0.0038, 0.0131, 0.0128, 0.0011, 0.0222, -0.0056, -0.0041, -0.0054, 0.0105, -0.0167], device='cuda:0'), grad: tensor([ 5.5879e-09, 1.8626e-09, -2.7940e-09, 3.1665e-08, -9.0338e-08, -1.4715e-07, 1.1362e-07, 2.3283e-08, 8.7544e-08, -1.1176e-08], device='cuda:0') 100 0.0001 changing lr epoch 257, time 246.21, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.4792 re_mapping 0.0035 re_causal 0.0111 /// teacc 99.17 lr 0.00010000 Epoch 259, weight, value: tensor([[-0.2065, -0.0447, 0.1982, ..., 0.0440, -0.1945, -0.1523], [-0.1471, -0.2212, -0.0482, ..., -0.1068, -0.2185, -0.0821], [ 0.2461, -0.0675, -0.2379, ..., -0.1817, -0.1868, -0.2072], ..., [-0.0843, -0.0268, -0.2843, ..., -0.1134, 0.1082, 0.0280], [-0.1850, -0.1227, -0.0321, ..., -0.1935, -0.0360, -0.0363], [-0.0736, -0.0939, -0.1311, ..., 0.0956, 0.0299, -0.1613]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 5.5879e-09, ..., 9.3132e-09, 3.7253e-09, 9.3132e-10], [ 8.3819e-09, 9.3132e-10, 1.8626e-09, ..., 2.7940e-08, 6.5193e-09, 9.3132e-10], [-3.1665e-08, 1.8626e-09, 9.3132e-10, ..., 2.7940e-09, 9.3132e-10, 0.0000e+00], ..., [ 2.0489e-08, 1.8626e-09, 0.0000e+00, ..., 1.1176e-08, -1.1176e-08, 0.0000e+00], [ 3.7253e-09, 9.3132e-10, 2.7940e-09, ..., 7.4506e-09, 1.8626e-09, 9.3132e-10], [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., -9.6858e-08, 3.7253e-09, 1.8626e-09]], device='cuda:0') Epoch 259, bias, value: tensor([ 0.0039, 0.0105, 0.0125, 0.0041, 0.0215, -0.0072, -0.0039, -0.0029, 0.0106, -0.0160], device='cuda:0'), grad: tensor([ 5.1223e-08, 1.0291e-06, 3.2317e-07, 3.4273e-07, 2.3283e-08, 4.1723e-07, 1.7975e-07, 4.5635e-08, -2.5444e-06, 1.2945e-07], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 258---------------------------------------------------- epoch 258, time 262.90, cls_loss 0.0009 cls_loss_mapping 0.0026 cls_loss_causal 0.4868 re_mapping 0.0034 re_causal 0.0109 /// teacc 99.22 lr 0.00010000 Epoch 260, weight, value: tensor([[-0.2084, -0.0447, 0.1982, ..., 0.0440, -0.1949, -0.1526], [-0.1472, -0.2222, -0.0481, ..., -0.1070, -0.2221, -0.0840], [ 0.2471, -0.0675, -0.2385, ..., -0.1827, -0.1875, -0.2079], ..., [-0.0860, -0.0270, -0.2857, ..., -0.1139, 0.1082, 0.0278], [-0.1852, -0.1227, -0.0324, ..., -0.1942, -0.0344, -0.0366], [-0.0737, -0.0939, -0.1311, ..., 0.0955, 0.0299, -0.1615]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, -6.5193e-09, ..., -9.3132e-10, 9.3132e-10, 0.0000e+00], [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, 4.6566e-09, 0.0000e+00], [-5.1223e-08, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, 1.8626e-09, 0.0000e+00], ..., [ 7.4506e-09, 9.3132e-10, 0.0000e+00, ..., 9.5926e-08, 3.8184e-08, 9.3132e-10], [ 5.5879e-09, 1.8626e-09, 1.8626e-09, ..., 5.5879e-09, 6.5193e-09, 3.7253e-09], [ 3.7253e-09, 9.3132e-10, 2.7940e-09, ..., 2.7008e-08, 4.6566e-09, 9.3132e-10]], device='cuda:0') Epoch 260, bias, value: tensor([ 0.0037, 0.0106, 0.0120, 0.0037, 0.0218, -0.0101, -0.0008, -0.0030, 0.0122, -0.0161], device='cuda:0'), grad: tensor([ 4.6566e-09, 4.0047e-08, -7.0781e-08, 6.7987e-08, -3.4738e-07, -9.0990e-07, 7.6089e-07, 2.6450e-07, 7.8231e-08, 1.1735e-07], device='cuda:0') 100 0.0001 changing lr epoch 259, time 246.11, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4982 re_mapping 0.0035 re_causal 0.0116 /// teacc 99.11 lr 0.00010000 Epoch 261, weight, value: tensor([[-0.2110, -0.0449, 0.1983, ..., 0.0430, -0.1957, -0.1533], [-0.1477, -0.2234, -0.0488, ..., -0.1072, -0.2239, -0.0840], [ 0.2478, -0.0676, -0.2388, ..., -0.1829, -0.1882, -0.2081], ..., [-0.0865, -0.0270, -0.2863, ..., -0.1142, 0.1084, 0.0282], [-0.1855, -0.1231, -0.0327, ..., -0.1946, -0.0344, -0.0369], [-0.0729, -0.0939, -0.1311, ..., 0.0962, 0.0301, -0.1630]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -3.7253e-09, ..., -1.8626e-09, 2.7940e-09, 9.3132e-10], [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.7008e-08, 9.3132e-10], [-3.5390e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 9.3132e-10], ..., [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, -5.3085e-08, 0.0000e+00], [-9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 0.0000e+00, -4.6566e-09], [ 9.3132e-10, 9.3132e-10, 1.8626e-09, ..., -9.3132e-10, 6.5193e-09, 1.8626e-09]], device='cuda:0') Epoch 261, bias, value: tensor([ 0.0029, 0.0107, 0.0122, 0.0036, 0.0219, -0.0101, -0.0007, -0.0032, 0.0121, -0.0156], device='cuda:0'), grad: tensor([ 1.9558e-08, -1.9714e-05, -4.6566e-08, 3.7514e-06, -1.4901e-08, -1.2852e-07, -2.7008e-08, 1.5840e-05, -3.5390e-08, 3.4273e-07], device='cuda:0') 100 0.0001 changing lr epoch 260, time 246.80, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4963 re_mapping 0.0035 re_causal 0.0116 /// teacc 99.11 lr 0.00010000 Epoch 262, weight, value: tensor([[-0.2112, -0.0450, 0.1978, ..., 0.0430, -0.1968, -0.1549], [-0.1489, -0.2237, -0.0489, ..., -0.1073, -0.2247, -0.0842], [ 0.2488, -0.0676, -0.2393, ..., -0.1832, -0.1885, -0.2082], ..., [-0.0867, -0.0270, -0.2868, ..., -0.1145, 0.1085, 0.0285], [-0.1861, -0.1236, -0.0327, ..., -0.1949, -0.0351, -0.0382], [-0.0729, -0.0939, -0.1312, ..., 0.0962, 0.0301, -0.1668]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -5.6811e-08, ..., -2.7940e-08, 2.7940e-09, 1.8626e-09], [ 7.4506e-09, 2.7940e-09, 5.5879e-09, ..., 2.7940e-09, 5.5879e-09, 1.8626e-09], [-2.8871e-08, 1.8626e-09, 8.3819e-09, ..., 2.7940e-09, 2.7940e-09, 1.8626e-09], ..., [ 2.3283e-08, 1.3039e-08, 9.3132e-10, ..., 3.2596e-08, 2.0489e-08, 9.3132e-10], [ 9.3132e-09, 0.0000e+00, 2.0489e-08, ..., 4.6566e-09, 8.3819e-09, 9.3132e-09], [ 9.3132e-10, 1.8626e-09, 3.5390e-08, ..., 1.8161e-07, 4.6566e-09, 9.3132e-10]], device='cuda:0') Epoch 262, bias, value: tensor([ 0.0025, 0.0107, 0.0125, 0.0035, 0.0220, -0.0093, -0.0013, -0.0031, 0.0117, -0.0157], device='cuda:0'), grad: tensor([-5.7742e-08, -4.0047e-07, -6.5193e-09, -5.5879e-08, -3.3621e-07, -2.3562e-07, 1.1455e-07, 5.0850e-07, 1.0058e-07, 3.5018e-07], device='cuda:0') 100 0.0001 changing lr epoch 261, time 246.08, cls_loss 0.0008 cls_loss_mapping 0.0024 cls_loss_causal 0.4695 re_mapping 0.0034 re_causal 0.0110 /// teacc 99.14 lr 0.00010000 Epoch 263, weight, value: tensor([[-0.2111, -0.0451, 0.1980, ..., 0.0431, -0.1970, -0.1551], [-0.1494, -0.2240, -0.0501, ..., -0.1075, -0.2253, -0.0848], [ 0.2495, -0.0677, -0.2421, ..., -0.1846, -0.1888, -0.2096], ..., [-0.0869, -0.0270, -0.2885, ..., -0.1155, 0.1085, 0.0263], [-0.1866, -0.1238, -0.0332, ..., -0.1955, -0.0358, -0.0400], [-0.0732, -0.0939, -0.1313, ..., 0.0934, 0.0288, -0.1706]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -8.8476e-08, ..., -5.8673e-08, 9.3132e-09, 6.5193e-09], [ 3.7253e-09, 0.0000e+00, 1.7695e-08, ..., 2.7008e-08, 7.4506e-09, 4.6566e-09], [-3.0082e-07, -9.3132e-10, 1.3970e-08, ..., 4.6566e-09, 5.5879e-09, 4.6566e-09], ..., [ 2.8685e-07, 9.3132e-10, 1.8626e-09, ..., 1.0245e-08, 9.3132e-10, 0.0000e+00], [ 4.6566e-09, 0.0000e+00, 2.2352e-08, ..., 4.7497e-08, 5.5879e-09, 4.6566e-09], [ 1.8626e-09, 0.0000e+00, 2.8871e-08, ..., 3.3434e-07, -1.8626e-09, 9.3132e-10]], device='cuda:0') Epoch 263, bias, value: tensor([ 0.0025, 0.0106, 0.0125, 0.0032, 0.0250, -0.0087, -0.0016, -0.0031, 0.0109, -0.0187], device='cuda:0'), grad: tensor([-7.0781e-08, 1.0524e-07, -3.1479e-07, 3.8370e-06, 7.6741e-07, -1.0796e-05, 8.1304e-07, 4.6752e-07, 9.5181e-07, 4.2506e-06], device='cuda:0') 100 0.0001 changing lr epoch 262, time 245.97, cls_loss 0.0007 cls_loss_mapping 0.0024 cls_loss_causal 0.4872 re_mapping 0.0035 re_causal 0.0113 /// teacc 99.19 lr 0.00010000 Epoch 264, weight, value: tensor([[-0.2098, -0.0452, 0.1983, ..., 0.0432, -0.1971, -0.1551], [-0.1510, -0.2244, -0.0508, ..., -0.1080, -0.2260, -0.0850], [ 0.2500, -0.0677, -0.2438, ..., -0.1860, -0.1890, -0.2099], ..., [-0.0871, -0.0271, -0.2899, ..., -0.1188, 0.1056, 0.0261], [-0.1867, -0.1247, -0.0333, ..., -0.1961, -0.0360, -0.0402], [-0.0732, -0.0940, -0.1314, ..., 0.0944, 0.0322, -0.1708]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.0245e-07, ..., -7.3574e-08, 9.3132e-10, 9.3132e-10], [ 1.8626e-09, 1.8626e-09, 4.6566e-09, ..., -4.6566e-09, 9.3132e-10, 2.7940e-09], [ 1.8626e-09, 9.3132e-10, 4.6566e-09, ..., 6.5193e-09, 9.3132e-10, 1.8626e-09], ..., [ 4.6566e-09, 2.7940e-09, 6.5193e-09, ..., 1.1176e-08, 0.0000e+00, 9.3132e-10], [ 9.3132e-10, 0.0000e+00, 2.5146e-08, ..., 1.3039e-08, 4.6566e-09, -1.4901e-08], [ 2.7940e-09, 3.7253e-09, 4.7497e-08, ..., 6.8918e-08, 9.3132e-10, 9.3132e-10]], device='cuda:0') Epoch 264, bias, value: tensor([ 0.0027, 0.0105, 0.0124, 0.0019, 0.0247, -0.0082, -0.0017, -0.0039, 0.0107, -0.0174], device='cuda:0'), grad: tensor([-2.2631e-07, -9.3132e-09, 1.2293e-07, 9.3132e-10, -2.1048e-07, 8.7544e-08, 2.2445e-07, 1.1269e-07, -3.8557e-07, 2.8312e-07], device='cuda:0') 100 0.0001 changing lr epoch 263, time 246.17, cls_loss 0.0006 cls_loss_mapping 0.0021 cls_loss_causal 0.4788 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.04 lr 0.00010000 Epoch 265, weight, value: tensor([[-0.2091, -0.0452, 0.1989, ..., 0.0434, -0.1979, -0.1561], [-0.1512, -0.2249, -0.0516, ..., -0.1080, -0.2265, -0.0850], [ 0.2503, -0.0677, -0.2451, ..., -0.1872, -0.1893, -0.2099], ..., [-0.0877, -0.0271, -0.2909, ..., -0.1202, 0.1052, 0.0262], [-0.1868, -0.1249, -0.0334, ..., -0.1965, -0.0361, -0.0405], [-0.0730, -0.0940, -0.1315, ..., 0.0946, 0.0326, -0.1708]], device='cuda:0'), grad: tensor([[-9.3132e-10, 0.0000e+00, -8.9128e-07, ..., -6.2957e-07, 9.3132e-10, -1.3039e-08], [ 3.9116e-08, 9.3132e-10, 2.4214e-08, ..., 5.5879e-08, 3.7253e-08, 6.5193e-09], [-3.7439e-07, 9.3132e-10, 2.4214e-08, ..., 3.8184e-08, 9.3132e-10, -4.6566e-08], ..., [ 3.2037e-07, 1.8626e-09, 9.3132e-09, ..., -1.1176e-08, -5.6811e-08, 4.0047e-08], [ 4.6566e-09, 9.3132e-10, 4.0047e-08, ..., 3.1665e-08, 9.3132e-10, 9.3132e-10], [ 1.8626e-09, 9.3132e-10, 4.8056e-07, ..., 3.6694e-07, 8.3819e-09, 1.8626e-09]], device='cuda:0') Epoch 265, bias, value: tensor([ 0.0029, 0.0105, 0.0123, 0.0017, 0.0247, -0.0081, -0.0017, -0.0042, 0.0105, -0.0171], device='cuda:0'), grad: tensor([-1.6913e-06, 4.6194e-07, -3.9395e-07, 1.4063e-07, -1.5367e-07, 1.2293e-07, 4.1071e-07, -2.4680e-07, 1.0803e-07, 1.2526e-06], device='cuda:0') 100 0.0001 changing lr epoch 264, time 246.22, cls_loss 0.0008 cls_loss_mapping 0.0022 cls_loss_causal 0.5080 re_mapping 0.0036 re_causal 0.0113 /// teacc 99.15 lr 0.00010000 Epoch 266, weight, value: tensor([[-0.2089, -0.0453, 0.1997, ..., 0.0437, -0.1979, -0.1562], [-0.1520, -0.2256, -0.0528, ..., -0.1088, -0.2280, -0.0850], [ 0.2515, -0.0677, -0.2462, ..., -0.1886, -0.1899, -0.2096], ..., [-0.0891, -0.0272, -0.2932, ..., -0.1203, 0.1053, 0.0257], [-0.1870, -0.1250, -0.0338, ..., -0.1988, -0.0362, -0.0406], [-0.0729, -0.0940, -0.1316, ..., 0.0964, 0.0326, -0.1708]], device='cuda:0'), grad: tensor([[-9.3132e-10, 9.3132e-10, -2.7008e-08, ..., -3.4459e-08, 2.6077e-08, 1.3970e-08], [ 8.3819e-09, 9.3132e-10, 1.2107e-08, ..., 8.3819e-09, 3.7253e-09, 9.3132e-10], [-5.3085e-08, 1.8626e-09, 6.5193e-09, ..., 3.7253e-09, 1.8626e-09, 1.8626e-09], ..., [ 3.3528e-08, 3.7253e-09, 9.3132e-10, ..., 6.5193e-09, -2.7940e-09, 0.0000e+00], [ 4.6566e-09, 9.3132e-10, 3.4459e-08, ..., 1.8626e-08, 6.5193e-09, 3.7253e-09], [ 1.0245e-08, 3.7253e-09, 3.9116e-08, ..., 7.2643e-08, 4.6566e-09, 2.7940e-09]], device='cuda:0') Epoch 266, bias, value: tensor([ 0.0034, 0.0104, 0.0125, 0.0015, 0.0229, -0.0081, -0.0019, -0.0041, 0.0100, -0.0153], device='cuda:0'), grad: tensor([-2.7008e-08, -1.4994e-07, -5.1223e-08, -2.9709e-07, -8.5682e-08, 1.9614e-06, -1.9446e-06, 2.3749e-07, 1.3225e-07, 2.3469e-07], device='cuda:0') 100 0.0001 changing lr epoch 265, time 245.67, cls_loss 0.0006 cls_loss_mapping 0.0023 cls_loss_causal 0.4885 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.15 lr 0.00010000 Epoch 267, weight, value: tensor([[-0.2089, -0.0457, 0.1999, ..., 0.0437, -0.1981, -0.1564], [-0.1527, -0.2262, -0.0530, ..., -0.1089, -0.2290, -0.0851], [ 0.2521, -0.0678, -0.2471, ..., -0.1887, -0.1906, -0.2100], ..., [-0.0894, -0.0272, -0.2948, ..., -0.1203, 0.1054, 0.0260], [-0.1872, -0.1252, -0.0340, ..., -0.1995, -0.0364, -0.0408], [-0.0732, -0.0941, -0.1317, ..., 0.0964, 0.0325, -0.1708]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 9.3132e-10, -9.1270e-08, ..., -6.8918e-08, 3.7253e-09, 9.3132e-10], [-8.1025e-08, 8.3819e-09, 3.7253e-09, ..., 1.8626e-09, 3.6322e-08, 0.0000e+00], [ 1.1921e-07, 5.5879e-09, 3.7253e-09, ..., 2.7940e-09, 1.8626e-09, 0.0000e+00], ..., [ 2.1420e-08, 8.3819e-09, 3.7253e-09, ..., 3.7253e-09, -5.9605e-08, 0.0000e+00], [ 1.0245e-08, 2.7940e-09, 9.3132e-09, ..., 6.5193e-09, 2.7940e-09, 9.3132e-10], [ 1.2107e-08, 3.7253e-09, 4.0047e-08, ..., 2.1420e-08, 9.3132e-09, 0.0000e+00]], device='cuda:0') Epoch 267, bias, value: tensor([ 0.0035, 0.0105, 0.0124, 0.0015, 0.0230, -0.0081, -0.0019, -0.0041, 0.0093, -0.0155], device='cuda:0'), grad: tensor([-1.5367e-07, -2.5090e-06, 2.7195e-06, -3.1386e-07, 2.4214e-08, 1.9092e-07, -4.8429e-08, -1.2200e-07, 6.5193e-08, 1.6484e-07], device='cuda:0') 100 0.0001 changing lr epoch 266, time 245.76, cls_loss 0.0006 cls_loss_mapping 0.0031 cls_loss_causal 0.4658 re_mapping 0.0035 re_causal 0.0111 /// teacc 99.16 lr 0.00010000 Epoch 268, weight, value: tensor([[-0.2095, -0.0458, 0.2004, ..., 0.0437, -0.1980, -0.1564], [-0.1536, -0.2273, -0.0531, ..., -0.1090, -0.2307, -0.0854], [ 0.2516, -0.0693, -0.2486, ..., -0.1900, -0.1914, -0.2105], ..., [-0.0891, -0.0274, -0.2956, ..., -0.1203, 0.1055, 0.0258], [-0.1874, -0.1252, -0.0342, ..., -0.1999, -0.0365, -0.0419], [-0.0732, -0.0941, -0.1320, ..., 0.0963, 0.0325, -0.1708]], device='cuda:0'), grad: tensor([[ 5.5879e-09, 2.7940e-09, 3.7253e-09, ..., 0.0000e+00, 9.3132e-10, 9.3132e-10], [ 1.8626e-08, 4.6566e-09, 9.3132e-10, ..., 1.8626e-09, 9.3132e-10, 9.3132e-10], [-2.7008e-08, 8.3819e-09, 1.8626e-09, ..., 0.0000e+00, 9.3132e-10, 9.3132e-10], ..., [ 2.9802e-08, 1.7695e-08, 0.0000e+00, ..., 3.7253e-09, -9.3132e-10, 9.3132e-10], [ 7.4506e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, 4.6566e-09], [ 1.0245e-08, 9.3132e-09, 0.0000e+00, ..., 5.5879e-09, 1.8626e-09, 1.8626e-09]], device='cuda:0') Epoch 268, bias, value: tensor([ 0.0036, 0.0105, 0.0112, 0.0020, 0.0230, -0.0079, -0.0020, -0.0041, 0.0093, -0.0155], device='cuda:0'), grad: tensor([ 6.1467e-08, 6.4261e-08, -3.7253e-09, 2.6338e-06, 1.3039e-08, -4.0345e-06, 2.0582e-07, 3.1292e-07, 4.5728e-07, 2.8405e-07], device='cuda:0') 100 0.0001 changing lr epoch 267, time 246.50, cls_loss 0.0005 cls_loss_mapping 0.0030 cls_loss_causal 0.4801 re_mapping 0.0038 re_causal 0.0118 /// teacc 99.09 lr 0.00010000 Epoch 269, weight, value: tensor([[-0.2098, -0.0460, 0.2005, ..., 0.0435, -0.1981, -0.1564], [-0.1541, -0.2283, -0.0531, ..., -0.1095, -0.2317, -0.0854], [ 0.2523, -0.0694, -0.2490, ..., -0.1907, -0.1916, -0.2107], ..., [-0.0896, -0.0274, -0.2961, ..., -0.1204, 0.1056, 0.0258], [-0.1878, -0.1254, -0.0341, ..., -0.2000, -0.0365, -0.0419], [-0.0734, -0.0941, -0.1321, ..., 0.0964, 0.0325, -0.1708]], device='cuda:0'), grad: tensor([[ 1.8161e-07, 0.0000e+00, 1.5832e-08, ..., 9.3132e-09, 2.7940e-09, 3.7253e-09], [ 7.4506e-09, 0.0000e+00, 1.2107e-08, ..., 1.8626e-09, 1.3039e-08, 9.3132e-10], [-4.9639e-07, -0.0000e+00, 1.2107e-08, ..., -3.4459e-08, 3.7253e-09, 0.0000e+00], ..., [ 7.7300e-08, 1.8626e-09, 1.8626e-09, ..., 1.3039e-08, -3.0734e-08, 0.0000e+00], [ 8.3819e-09, 0.0000e+00, -8.6613e-08, ..., 1.1176e-08, 6.5193e-09, 1.8626e-09], [ 5.5879e-09, 0.0000e+00, 6.5193e-09, ..., -3.5390e-08, -3.7253e-09, 0.0000e+00]], device='cuda:0') Epoch 269, bias, value: tensor([ 0.0034, 0.0105, 0.0113, 0.0020, 0.0231, -0.0072, -0.0027, -0.0041, 0.0093, -0.0155], device='cuda:0'), grad: tensor([ 7.0315e-07, 5.4948e-08, -6.9570e-07, 4.9826e-07, 5.1223e-08, -1.5367e-07, 8.0466e-07, 2.5518e-07, -1.6205e-06, 1.0058e-07], device='cuda:0') 100 0.0001 changing lr epoch 268, time 246.37, cls_loss 0.0005 cls_loss_mapping 0.0018 cls_loss_causal 0.5013 re_mapping 0.0037 re_causal 0.0117 /// teacc 99.18 lr 0.00010000 Epoch 270, weight, value: tensor([[-0.2095, -0.0461, 0.2006, ..., 0.0436, -0.1984, -0.1565], [-0.1544, -0.2289, -0.0535, ..., -0.1099, -0.2334, -0.0859], [ 0.2533, -0.0694, -0.2501, ..., -0.1913, -0.1923, -0.2113], ..., [-0.0909, -0.0275, -0.2977, ..., -0.1205, 0.1057, 0.0256], [-0.1879, -0.1254, -0.0340, ..., -0.2001, -0.0364, -0.0420], [-0.0739, -0.0942, -0.1322, ..., 0.0964, 0.0324, -0.1708]], device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, -2.7940e-08, ..., 1.8626e-09, 2.7940e-09, 9.3132e-10], [ 5.0012e-07, 0.0000e+00, 9.3132e-10, ..., 3.2969e-07, 1.1176e-08, 0.0000e+00], [ 2.4866e-07, 0.0000e+00, -0.0000e+00, ..., 4.6566e-09, 9.3132e-10, 0.0000e+00], ..., [-7.7020e-07, 0.0000e+00, 9.3132e-10, ..., 2.7008e-08, -1.9558e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.4901e-08, 2.7940e-09, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 5.5879e-09, ..., -3.8929e-07, -2.4214e-08, 0.0000e+00]], device='cuda:0') Epoch 270, bias, value: tensor([ 0.0035, 0.0105, 0.0113, 0.0018, 0.0230, -0.0071, -0.0027, -0.0041, 0.0097, -0.0155], device='cuda:0'), grad: tensor([ 7.5437e-08, 8.7395e-06, 4.4517e-06, 1.4529e-07, -5.3272e-07, 7.0781e-08, 8.6613e-08, -1.2293e-05, -1.1176e-08, -7.1153e-07], device='cuda:0') 100 0.0001 changing lr epoch 269, time 246.71, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4739 re_mapping 0.0040 re_causal 0.0117 /// teacc 99.10 lr 0.00010000 Epoch 271, weight, value: tensor([[-0.2096, -0.0463, 0.2010, ..., 0.0435, -0.1987, -0.1566], [-0.1552, -0.2300, -0.0542, ..., -0.1107, -0.2342, -0.0859], [ 0.2543, -0.0694, -0.2512, ..., -0.1927, -0.1929, -0.2115], ..., [-0.0917, -0.0275, -0.3025, ..., -0.1206, 0.1057, 0.0252], [-0.1881, -0.1256, -0.0342, ..., -0.2005, -0.0365, -0.0421], [-0.0741, -0.0944, -0.1323, ..., 0.0965, 0.0325, -0.1708]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.9116e-08, ..., -2.4214e-08, 4.6566e-09, -9.3132e-10], [ 2.7940e-09, 9.3132e-10, 1.8626e-09, ..., 6.5193e-09, 9.3132e-09, 0.0000e+00], [-1.1176e-08, 9.3132e-10, 1.8626e-09, ..., 9.3132e-10, 2.7940e-09, 0.0000e+00], ..., [ 1.2107e-08, 9.3132e-10, 1.8626e-09, ..., 3.6322e-08, -6.5193e-09, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 1.0524e-07, 4.5635e-08, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 2.1420e-08, ..., -1.1083e-07, -6.1467e-08, 9.3132e-10]], device='cuda:0') Epoch 271, bias, value: tensor([ 0.0035, 0.0105, 0.0114, 0.0017, 0.0231, -0.0071, -0.0027, -0.0042, 0.0103, -0.0155], device='cuda:0'), grad: tensor([ 2.9895e-07, 3.0734e-08, 2.1420e-08, 6.7055e-08, -7.7300e-08, 4.1910e-08, 4.7497e-08, 1.1176e-07, -2.6636e-07, -2.9523e-07], device='cuda:0') 100 0.0001 changing lr epoch 270, time 246.30, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.5153 re_mapping 0.0036 re_causal 0.0116 /// teacc 99.09 lr 0.00010000 Epoch 272, weight, value: tensor([[-0.2098, -0.0464, 0.2020, ..., 0.0435, -0.1991, -0.1562], [-0.1557, -0.2313, -0.0543, ..., -0.1109, -0.2354, -0.0860], [ 0.2546, -0.0696, -0.2529, ..., -0.1945, -0.1955, -0.2134], ..., [-0.0915, -0.0275, -0.3053, ..., -0.1210, 0.1056, 0.0274], [-0.1882, -0.1261, -0.0343, ..., -0.2012, -0.0369, -0.0422], [-0.0741, -0.0944, -0.1324, ..., 0.0965, 0.0327, -0.1708]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, 1.5702e-06, ..., -7.4506e-08, 7.7114e-07, 5.5879e-07], [-6.8359e-07, 0.0000e+00, 3.2596e-08, ..., 3.0734e-08, 2.4214e-08, 2.7940e-09], [ 5.6066e-07, 0.0000e+00, 1.3970e-08, ..., 1.4901e-08, 1.3039e-08, 2.7940e-09], ..., [ 8.1025e-08, 0.0000e+00, 8.3819e-09, ..., 1.4901e-07, 2.1420e-08, 0.0000e+00], [ 4.6566e-09, 0.0000e+00, 3.2596e-08, ..., 2.6077e-08, 2.3283e-08, 8.3819e-09], [ 1.8626e-09, 0.0000e+00, 4.4703e-08, ..., 4.6566e-09, 4.6566e-08, 2.7940e-09]], device='cuda:0') Epoch 272, bias, value: tensor([ 0.0039, 0.0104, 0.0110, 0.0015, 0.0232, -0.0070, -0.0030, -0.0042, 0.0104, -0.0154], device='cuda:0'), grad: tensor([ 3.0026e-06, -7.1749e-06, 6.1691e-06, 1.8906e-07, -7.0129e-07, -5.3365e-07, -2.3339e-06, 1.0747e-06, 2.3283e-07, 3.5390e-08], device='cuda:0') 100 0.0001 changing lr epoch 271, time 245.98, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4690 re_mapping 0.0035 re_causal 0.0111 /// teacc 98.99 lr 0.00010000 Epoch 273, weight, value: tensor([[-0.2102, -0.0469, 0.2023, ..., 0.0436, -0.1996, -0.1562], [-0.1558, -0.2327, -0.0544, ..., -0.1110, -0.2367, -0.0861], [ 0.2554, -0.0696, -0.2531, ..., -0.1956, -0.1963, -0.2138], ..., [-0.0918, -0.0275, -0.3080, ..., -0.1211, 0.1059, 0.0279], [-0.1886, -0.1264, -0.0348, ..., -0.2014, -0.0371, -0.0425], [-0.0742, -0.0944, -0.1326, ..., 0.0965, 0.0325, -0.1708]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, -1.3178e-07, ..., -8.6613e-08, 1.6298e-08, 5.1223e-09], [ 9.3132e-10, 9.3132e-10, 2.3283e-09, ..., 6.8918e-08, 3.6322e-08, 1.6298e-08], [ 0.0000e+00, 4.6566e-10, 4.1910e-09, ..., 2.7474e-08, 1.3039e-08, 6.0536e-09], ..., [ 1.3970e-09, 1.8626e-09, 6.0536e-09, ..., -1.1977e-06, -5.4482e-07, -2.9616e-07], [ 4.6566e-10, 4.6566e-10, 6.5193e-09, ..., 7.4506e-09, 5.1223e-09, 1.3970e-09], [ 0.0000e+00, 0.0000e+00, 9.0804e-08, ..., 1.4715e-07, 3.7253e-08, 2.2352e-08]], device='cuda:0') Epoch 273, bias, value: tensor([ 0.0040, 0.0104, 0.0109, 0.0014, 0.0232, -0.0069, -0.0031, -0.0042, 0.0102, -0.0155], device='cuda:0'), grad: tensor([-4.5169e-08, 6.1980e-07, 2.9476e-07, 7.5549e-06, 1.1176e-07, -1.6317e-06, 1.0133e-06, -8.7991e-06, 9.9186e-08, 7.8278e-07], device='cuda:0') 100 0.0001 changing lr epoch 272, time 246.49, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.5005 re_mapping 0.0037 re_causal 0.0121 /// teacc 99.10 lr 0.00010000 Epoch 274, weight, value: tensor([[-0.2105, -0.0471, 0.2029, ..., 0.0437, -0.2003, -0.1565], [-0.1566, -0.2333, -0.0555, ..., -0.1116, -0.2383, -0.0861], [ 0.2576, -0.0697, -0.2539, ..., -0.1967, -0.1969, -0.2139], ..., [-0.0943, -0.0275, -0.3104, ..., -0.1209, 0.1069, 0.0291], [-0.1886, -0.1265, -0.0344, ..., -0.2006, -0.0373, -0.0425], [-0.0744, -0.0945, -0.1329, ..., 0.0956, 0.0306, -0.1709]], device='cuda:0'), grad: tensor([[ 4.6100e-08, 0.0000e+00, -5.4948e-08, ..., -3.6322e-08, 8.2422e-08, 3.7253e-08], [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., -1.3970e-09, 9.3132e-09, 4.6566e-10], [ 7.4506e-09, 0.0000e+00, 2.3749e-08, ..., 2.2352e-08, 1.9558e-08, 9.3132e-09], ..., [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 1.6298e-08, -1.8161e-08, 4.6566e-10], [ 8.3819e-09, 0.0000e+00, 4.5635e-08, ..., 3.8184e-08, 1.7229e-08, 7.9162e-09], [ 9.3132e-10, 0.0000e+00, 8.7544e-08, ..., 5.1223e-08, -1.3970e-09, 1.8626e-09]], device='cuda:0') Epoch 274, bias, value: tensor([ 0.0043, 0.0104, 0.0118, 0.0005, 0.0240, -0.0069, -0.0032, -0.0041, 0.0109, -0.0164], device='cuda:0'), grad: tensor([ 8.1956e-08, -3.8650e-08, 8.4285e-08, 4.9826e-08, 8.4285e-08, 1.7090e-07, -7.0361e-07, 2.0955e-08, 1.2573e-07, 1.4435e-07], device='cuda:0') 100 0.0001 changing lr epoch 273, time 246.91, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4930 re_mapping 0.0037 re_causal 0.0115 /// teacc 99.17 lr 0.00010000 Epoch 275, weight, value: tensor([[-0.2108, -0.0478, 0.2039, ..., 0.0441, -0.2010, -0.1566], [-0.1573, -0.2349, -0.0557, ..., -0.1119, -0.2397, -0.0862], [ 0.2598, -0.0697, -0.2562, ..., -0.1975, -0.1980, -0.2146], ..., [-0.0968, -0.0276, -0.3139, ..., -0.1209, 0.1070, 0.0290], [-0.1891, -0.1272, -0.0351, ..., -0.2017, -0.0378, -0.0426], [-0.0750, -0.0948, -0.1339, ..., 0.0955, 0.0306, -0.1709]], device='cuda:0'), grad: tensor([[ 2.3283e-09, 1.3970e-09, -6.0024e-07, ..., -4.4471e-07, 7.9162e-09, 3.7253e-09], [ 4.1910e-09, 1.8626e-09, 1.3504e-08, ..., -5.1223e-09, 9.3132e-10, 0.0000e+00], [-8.3819e-09, 1.3970e-09, 3.8184e-08, ..., 2.5611e-08, 1.3970e-09, 4.6566e-10], ..., [ 9.3132e-09, 3.2596e-09, 2.7940e-08, ..., 2.4680e-08, -4.6566e-10, 0.0000e+00], [ 4.1910e-09, 2.7940e-09, 2.1420e-08, ..., 3.1199e-08, 8.3819e-09, 4.6566e-10], [ 2.3283e-09, 1.8626e-09, 2.4494e-07, ..., 1.3271e-07, -1.4435e-08, 4.6566e-10]], device='cuda:0') Epoch 275, bias, value: tensor([ 0.0049, 0.0105, 0.0128, 0.0004, 0.0240, -0.0069, -0.0031, -0.0042, 0.0105, -0.0165], device='cuda:0'), grad: tensor([-1.2582e-06, -1.7416e-07, 7.4971e-08, -6.4261e-08, 9.1735e-08, 3.4925e-08, 4.9220e-07, 1.7183e-07, 1.1036e-07, 5.2294e-07], device='cuda:0') 100 0.0001 changing lr epoch 274, time 246.46, cls_loss 0.0008 cls_loss_mapping 0.0025 cls_loss_causal 0.4909 re_mapping 0.0035 re_causal 0.0113 /// teacc 99.21 lr 0.00010000 Epoch 276, weight, value: tensor([[-0.2111, -0.0484, 0.2043, ..., 0.0437, -0.2014, -0.1567], [-0.1572, -0.2373, -0.0558, ..., -0.1120, -0.2415, -0.0862], [ 0.2601, -0.0699, -0.2574, ..., -0.1985, -0.1994, -0.2153], ..., [-0.0975, -0.0278, -0.3162, ..., -0.1209, 0.1072, 0.0289], [-0.1893, -0.1294, -0.0358, ..., -0.2023, -0.0386, -0.0428], [-0.0756, -0.0950, -0.1341, ..., 0.0956, 0.0304, -0.1709]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.6124e-07, ..., -1.7183e-07, 1.1176e-08, 1.8626e-09], [ 4.6566e-10, 0.0000e+00, 1.2107e-08, ..., 1.9558e-08, 2.3283e-08, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 5.1223e-09, 1.0245e-08, 4.6566e-10], ..., [ 4.6566e-10, 0.0000e+00, 4.6566e-09, ..., 7.4506e-09, 1.6484e-07, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 1.5832e-08, ..., 3.4925e-08, -2.6403e-07, 2.7940e-09], [ 0.0000e+00, 0.0000e+00, 1.0198e-07, ..., 3.7532e-07, 6.9384e-08, 4.6566e-10]], device='cuda:0') Epoch 276, bias, value: tensor([ 0.0047, 0.0105, 0.0110, -0.0010, 0.0241, -0.0049, -0.0032, -0.0042, 0.0115, -0.0166], device='cuda:0'), grad: tensor([-4.4843e-07, -6.4913e-07, 8.5216e-08, 1.5879e-07, -6.5658e-07, 7.6368e-08, 1.6624e-07, 1.9185e-06, -1.6671e-06, 1.0226e-06], device='cuda:0') 100 0.0001 changing lr epoch 275, time 246.66, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4996 re_mapping 0.0035 re_causal 0.0116 /// teacc 99.15 lr 0.00010000 Epoch 277, weight, value: tensor([[-0.2114, -0.0492, 0.2045, ..., 0.0437, -0.2017, -0.1567], [-0.1576, -0.2384, -0.0559, ..., -0.1122, -0.2429, -0.0863], [ 0.2624, -0.0700, -0.2578, ..., -0.1988, -0.1998, -0.2156], ..., [-0.1005, -0.0279, -0.3170, ..., -0.1209, 0.1074, 0.0292], [-0.1898, -0.1296, -0.0361, ..., -0.2042, -0.0391, -0.0432], [-0.0755, -0.0950, -0.1342, ..., 0.0958, 0.0306, -0.1709]], device='cuda:0'), grad: tensor([[ 2.3749e-08, 3.7253e-09, -5.0291e-08, ..., 1.9791e-07, 1.2107e-08, 2.0489e-08], [ 2.5611e-08, 3.7253e-09, 1.8626e-09, ..., 2.0489e-08, 1.0710e-08, 5.1223e-09], [-1.4435e-07, -9.3132e-10, 5.5879e-09, ..., 3.0268e-08, 1.8626e-09, 1.2107e-08], ..., [ 5.7276e-08, 6.0536e-09, 9.3132e-10, ..., 4.6566e-09, -1.6298e-08, 0.0000e+00], [ 1.7695e-08, 4.6566e-10, 5.1223e-09, ..., 4.0513e-08, 6.5193e-09, -2.7940e-09], [ 6.9849e-09, 2.7940e-09, 2.5146e-08, ..., -2.5053e-07, -2.6077e-08, 1.4901e-08]], device='cuda:0') Epoch 277, bias, value: tensor([ 0.0047, 0.0108, 0.0120, -0.0010, 0.0239, -0.0048, -0.0032, -0.0045, 0.0109, -0.0165], device='cuda:0'), grad: tensor([ 4.2887e-07, 1.8300e-07, -1.9092e-07, -9.0804e-08, -5.8711e-06, 7.8231e-08, 5.8524e-06, 3.9581e-08, 9.7789e-09, -4.2608e-07], device='cuda:0') 100 0.0001 changing lr epoch 276, time 246.26, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4922 re_mapping 0.0034 re_causal 0.0108 /// teacc 99.00 lr 0.00010000 Epoch 278, weight, value: tensor([[-0.2120, -0.0493, 0.2047, ..., 0.0437, -0.2020, -0.1568], [-0.1586, -0.2391, -0.0559, ..., -0.1125, -0.2452, -0.0865], [ 0.2641, -0.0701, -0.2567, ..., -0.1997, -0.2003, -0.2139], ..., [-0.1012, -0.0279, -0.3178, ..., -0.1208, 0.1077, 0.0294], [-0.1905, -0.1297, -0.0365, ..., -0.2051, -0.0397, -0.0423], [-0.0765, -0.0951, -0.1343, ..., 0.0957, 0.0304, -0.1709]], device='cuda:0'), grad: tensor([[ 3.9209e-07, 1.8626e-09, 1.5739e-07, ..., 3.5297e-07, 1.5832e-08, 3.0361e-07], [ 9.3132e-08, 9.3132e-10, 1.1176e-08, ..., 9.2201e-08, 5.7742e-08, 1.3970e-08], [ 6.1933e-07, 1.8626e-09, 3.1386e-07, ..., 7.2550e-07, 4.9360e-08, 5.7463e-07], ..., [-2.7008e-08, 0.0000e+00, 1.8626e-09, ..., 1.6205e-07, -9.2201e-08, 1.8626e-09], [ 9.1270e-08, 9.3132e-10, 3.7253e-08, ..., 7.4506e-08, 9.3132e-10, 6.1467e-08], [ 3.4459e-08, 0.0000e+00, 2.1420e-08, ..., -3.8929e-07, -2.7008e-08, 1.4901e-08]], device='cuda:0') Epoch 278, bias, value: tensor([ 0.0046, 0.0110, 0.0128, -0.0011, 0.0240, -0.0047, -0.0034, -0.0047, 0.0114, -0.0166], device='cuda:0'), grad: tensor([ 1.9781e-06, -7.9796e-06, 3.7029e-06, 7.7765e-07, 2.5928e-06, 1.0401e-05, -1.2808e-05, 9.2909e-06, -9.6336e-06, 1.6065e-06], device='cuda:0') 100 0.0001 changing lr epoch 277, time 246.28, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.4906 re_mapping 0.0038 re_causal 0.0115 /// teacc 99.15 lr 0.00010000 Epoch 279, weight, value: tensor([[-0.2123, -0.0496, 0.2049, ..., 0.0434, -0.2026, -0.1570], [-0.1625, -0.2397, -0.0561, ..., -0.1127, -0.2473, -0.0867], [ 0.2676, -0.0702, -0.2579, ..., -0.1984, -0.2010, -0.2154], ..., [-0.1019, -0.0281, -0.3198, ..., -0.1209, 0.1079, 0.0297], [-0.1908, -0.1299, -0.0373, ..., -0.2058, -0.0401, -0.0428], [-0.0775, -0.0951, -0.1346, ..., 0.0958, 0.0305, -0.1709]], device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, -2.1514e-07, ..., -1.2666e-07, 9.3132e-10, -9.3132e-10], [ 8.1956e-08, 0.0000e+00, 2.2352e-08, ..., 2.0489e-08, 2.7940e-09, 0.0000e+00], [-1.1269e-06, 0.0000e+00, 9.3132e-09, ..., 8.3819e-09, 1.8626e-09, 0.0000e+00], ..., [ 9.7882e-07, 2.7940e-09, 1.0245e-08, ..., -9.3132e-10, -1.0245e-08, 0.0000e+00], [ 1.6764e-08, 0.0000e+00, 2.5146e-08, ..., 2.0489e-08, -9.3132e-10, 0.0000e+00], [ 9.3132e-09, 0.0000e+00, 8.5682e-08, ..., 2.8871e-08, 1.8626e-09, 0.0000e+00]], device='cuda:0') Epoch 279, bias, value: tensor([ 0.0045, 0.0102, 0.0150, -0.0012, 0.0240, -0.0048, -0.0032, -0.0041, 0.0115, -0.0166], device='cuda:0'), grad: tensor([-3.8557e-07, 2.8871e-07, -2.2408e-06, 6.4261e-08, 5.6811e-08, 3.7253e-08, 1.0338e-07, 1.7826e-06, 7.0781e-08, 2.0955e-07], device='cuda:0') 100 0.0001 changing lr epoch 278, time 246.77, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.4949 re_mapping 0.0034 re_causal 0.0109 /// teacc 99.13 lr 0.00010000 Epoch 280, weight, value: tensor([[-0.2130, -0.0497, 0.2049, ..., 0.0435, -0.2039, -0.1578], [-0.1630, -0.2404, -0.0561, ..., -0.1147, -0.2488, -0.0868], [ 0.2681, -0.0703, -0.2596, ..., -0.1989, -0.2021, -0.2171], ..., [-0.1022, -0.0282, -0.3211, ..., -0.1211, 0.1080, 0.0291], [-0.1910, -0.1302, -0.0376, ..., -0.2066, -0.0416, -0.0449], [-0.0792, -0.0953, -0.1347, ..., 0.0958, 0.0303, -0.1710]], device='cuda:0'), grad: tensor([[ 3.1665e-08, 9.3132e-10, -1.4789e-06, ..., 1.2293e-07, -2.3283e-08, 8.3819e-09], [ 2.7940e-09, 9.3132e-10, 1.3039e-08, ..., 1.3504e-07, 4.6566e-09, 7.4506e-09], [ 2.9802e-08, 2.7940e-09, 4.6566e-09, ..., 5.1223e-08, 1.0245e-08, 2.7940e-09], ..., [-1.3039e-08, 9.3132e-10, 6.5193e-09, ..., 3.7625e-07, 0.0000e+00, 1.6764e-08], [ 9.3132e-09, 1.8626e-09, 2.6077e-08, ..., 4.0978e-08, 5.5879e-09, 1.8626e-09], [ 2.5146e-08, 1.8626e-09, 5.1223e-08, ..., 1.6853e-05, 1.2107e-07, 9.3319e-07]], device='cuda:0') Epoch 280, bias, value: tensor([ 0.0043, 0.0099, 0.0149, -0.0011, 0.0241, -0.0048, -0.0030, -0.0040, 0.0103, -0.0166], device='cuda:0'), grad: tensor([-2.4140e-06, -3.2596e-08, 4.1910e-07, -2.7940e-07, -4.2707e-05, 2.7567e-07, 2.7530e-06, 8.7358e-07, 2.9150e-07, 4.0889e-05], device='cuda:0') 100 0.0001 changing lr epoch 279, time 246.68, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4762 re_mapping 0.0035 re_causal 0.0107 /// teacc 99.21 lr 0.00010000 Epoch 281, weight, value: tensor([[-0.2134, -0.0499, 0.2057, ..., 0.0437, -0.2039, -0.1578], [-0.1631, -0.2408, -0.0563, ..., -0.1151, -0.2504, -0.0870], [ 0.2683, -0.0703, -0.2615, ..., -0.2003, -0.2032, -0.2177], ..., [-0.1023, -0.0282, -0.3226, ..., -0.1210, 0.1085, 0.0293], [-0.1906, -0.1303, -0.0373, ..., -0.2067, -0.0423, -0.0457], [-0.0796, -0.0953, -0.1353, ..., 0.0956, 0.0300, -0.1712]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., 1.3039e-08, 2.7940e-09, -1.8626e-09], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, 1.9558e-08, 0.0000e+00], [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 2.3283e-08, 0.0000e+00], ..., [ 2.7940e-09, 0.0000e+00, 1.8626e-09, ..., 7.4506e-09, -7.4506e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.8871e-08, 5.5879e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., -2.0675e-07, 9.3132e-10, 1.8626e-09]], device='cuda:0') Epoch 281, bias, value: tensor([ 0.0047, 0.0099, 0.0146, -0.0011, 0.0243, -0.0047, -0.0030, -0.0039, 0.0105, -0.0169], device='cuda:0'), grad: tensor([ 2.6077e-08, 3.8184e-08, 1.5367e-07, 6.7987e-08, 2.8871e-07, -9.3132e-09, 2.6077e-08, -3.5670e-07, 9.3132e-08, -3.2131e-07], device='cuda:0') 100 0.0001 changing lr epoch 280, time 246.63, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4326 re_mapping 0.0036 re_causal 0.0110 /// teacc 99.16 lr 0.00010000 Epoch 282, weight, value: tensor([[-0.2137, -0.0507, 0.2059, ..., 0.0430, -0.2040, -0.1577], [-0.1631, -0.2411, -0.0561, ..., -0.1152, -0.2516, -0.0872], [ 0.2685, -0.0704, -0.2627, ..., -0.2013, -0.2041, -0.2181], ..., [-0.1023, -0.0282, -0.3242, ..., -0.1209, 0.1089, 0.0298], [-0.1910, -0.1306, -0.0377, ..., -0.2073, -0.0426, -0.0459], [-0.0798, -0.0954, -0.1354, ..., 0.0957, 0.0297, -0.1712]], device='cuda:0'), grad: tensor([[ 9.3132e-10, -4.6566e-09, -5.4482e-07, ..., -2.4214e-07, 2.7940e-09, -2.0489e-08], [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 7.4506e-09, 9.3132e-09, 9.3132e-10], [-7.4506e-09, 9.3132e-10, 2.1420e-08, ..., 1.4901e-08, 8.3819e-09, -2.7940e-09], ..., [ 1.8626e-09, 9.3132e-10, 1.4901e-08, ..., 1.1176e-08, -3.9116e-08, 9.3132e-10], [ 3.7253e-09, 2.7940e-09, 7.0781e-08, ..., 3.2596e-08, 1.8626e-09, 3.7253e-09], [ 9.3132e-10, 9.3132e-10, 2.1793e-07, ..., 2.0582e-07, 3.2596e-08, 1.2107e-08]], device='cuda:0') Epoch 282, bias, value: tensor([ 0.0042, 0.0099, 0.0140, -0.0012, 0.0243, -0.0047, -0.0031, -0.0037, 0.0102, -0.0169], device='cuda:0'), grad: tensor([-1.1008e-06, 1.1269e-07, 4.8056e-07, 1.5274e-07, -2.9802e-07, 2.7008e-07, 3.0920e-07, -1.0021e-06, 1.9465e-07, 8.8103e-07], device='cuda:0') 100 0.0001 changing lr epoch 281, time 246.12, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4959 re_mapping 0.0033 re_causal 0.0106 /// teacc 99.18 lr 0.00010000 Epoch 283, weight, value: tensor([[-0.2149, -0.0508, 0.2060, ..., 0.0430, -0.2043, -0.1578], [-0.1624, -0.2416, -0.0561, ..., -0.1154, -0.2520, -0.0873], [ 0.2686, -0.0705, -0.2632, ..., -0.2009, -0.2052, -0.2186], ..., [-0.1027, -0.0283, -0.3250, ..., -0.1210, 0.1089, 0.0301], [-0.1915, -0.1308, -0.0381, ..., -0.2076, -0.0426, -0.0462], [-0.0800, -0.0954, -0.1355, ..., 0.0957, 0.0297, -0.1712]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.0990e-07, ..., -1.6764e-08, 1.8626e-09, 9.3132e-10], [ 5.5879e-09, 9.3132e-10, 2.7940e-09, ..., 7.0781e-08, 6.5193e-09, 0.0000e+00], [-5.0291e-08, 2.7940e-09, 7.4506e-09, ..., 2.3283e-08, 2.7940e-09, 0.0000e+00], ..., [ 5.9605e-08, 2.7940e-09, 4.6566e-09, ..., 3.1665e-08, -6.5193e-09, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 1.2107e-08, ..., -2.5146e-08, 1.8626e-09, 1.8626e-09], [ 1.8626e-09, 9.3132e-10, 2.4214e-08, ..., 5.0012e-07, 1.4901e-08, 9.3132e-10]], device='cuda:0') Epoch 283, bias, value: tensor([ 0.0041, 0.0101, 0.0131, -0.0014, 0.0243, -0.0046, -0.0030, -0.0037, 0.0102, -0.0169], device='cuda:0'), grad: tensor([-3.8184e-08, -6.2399e-08, -9.3132e-10, 6.0536e-08, -1.8794e-06, -5.0291e-08, 2.0489e-07, 3.7160e-07, -3.0082e-07, 1.6931e-06], device='cuda:0') 100 0.0001 changing lr epoch 282, time 246.46, cls_loss 0.0005 cls_loss_mapping 0.0017 cls_loss_causal 0.5171 re_mapping 0.0033 re_causal 0.0112 /// teacc 99.11 lr 0.00010000 Epoch 284, weight, value: tensor([[-0.2162, -0.0510, 0.2060, ..., 0.0431, -0.2049, -0.1584], [-0.1625, -0.2427, -0.0562, ..., -0.1156, -0.2526, -0.0875], [ 0.2696, -0.0706, -0.2637, ..., -0.2004, -0.2059, -0.2177], ..., [-0.1034, -0.0285, -0.3260, ..., -0.1210, 0.1090, 0.0285], [-0.1917, -0.1315, -0.0385, ..., -0.2082, -0.0430, -0.0471], [-0.0802, -0.0952, -0.1356, ..., 0.0957, 0.0297, -0.1713]], device='cuda:0'), grad: tensor([[ 1.0245e-08, 2.7940e-09, -6.0722e-07, ..., -1.8533e-07, 3.7253e-09, -8.1956e-08], [ 8.9407e-08, 2.9802e-08, 2.7940e-09, ..., 1.4901e-08, 1.8626e-08, 1.8626e-09], [-5.5879e-09, 1.3970e-08, 1.8626e-09, ..., 3.7253e-09, 9.3132e-10, -1.6764e-08], ..., [ 1.0896e-07, 3.6322e-08, 4.6566e-09, ..., 5.3085e-08, -2.1420e-08, 2.7940e-09], [ 2.9802e-08, 1.2107e-08, 9.3132e-09, ..., 1.6112e-07, 2.6077e-08, 1.0245e-08], [ 2.7008e-08, 6.5193e-09, 6.5193e-09, ..., -2.3283e-08, -1.3039e-08, 1.8626e-09]], device='cuda:0') Epoch 284, bias, value: tensor([ 0.0040, 0.0101, 0.0130, -0.0015, 0.0244, -0.0046, -0.0030, -0.0038, 0.0100, -0.0170], device='cuda:0'), grad: tensor([-9.9186e-07, 3.6974e-07, 1.3597e-07, -1.3346e-06, -5.4482e-07, -1.8533e-07, 1.2293e-06, 5.6159e-07, 5.8487e-07, 1.6391e-07], device='cuda:0') 100 0.0001 changing lr epoch 283, time 246.54, cls_loss 0.0007 cls_loss_mapping 0.0021 cls_loss_causal 0.5079 re_mapping 0.0032 re_causal 0.0107 /// teacc 99.12 lr 0.00010000 Epoch 285, weight, value: tensor([[-0.2165, -0.0512, 0.2066, ..., 0.0432, -0.2052, -0.1585], [-0.1626, -0.2434, -0.0563, ..., -0.1166, -0.2542, -0.0877], [ 0.2710, -0.0704, -0.2645, ..., -0.2010, -0.2066, -0.2177], ..., [-0.1052, -0.0287, -0.3276, ..., -0.1216, 0.1090, 0.0286], [-0.1919, -0.1315, -0.0389, ..., -0.2115, -0.0430, -0.0475], [-0.0803, -0.0952, -0.1359, ..., 0.0960, 0.0298, -0.1713]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 2.7940e-09, -2.7008e-08, ..., -1.0245e-08, 1.0245e-08, 4.6566e-09], [ 4.6566e-09, 4.6566e-09, 5.5879e-09, ..., 4.3772e-08, 1.9558e-08, 2.7940e-09], [ 5.5879e-09, 9.3132e-09, 2.7940e-09, ..., 1.4901e-08, 5.5879e-09, 9.3132e-10], ..., [ 1.7695e-08, 1.0245e-08, 1.8626e-09, ..., 3.0454e-07, 1.4901e-07, 0.0000e+00], [ 3.7253e-09, 3.7253e-09, 9.3132e-09, ..., 8.3819e-09, 1.2107e-08, 5.5879e-09], [ 1.0245e-08, 1.1176e-08, 1.1176e-08, ..., 1.8533e-07, -2.0489e-08, 0.0000e+00]], device='cuda:0') Epoch 285, bias, value: tensor([ 0.0042, 0.0101, 0.0137, -0.0011, 0.0245, -0.0049, -0.0030, -0.0040, 0.0085, -0.0167], device='cuda:0'), grad: tensor([ 1.4901e-08, 1.0431e-07, 1.2014e-07, -4.3027e-07, -2.4233e-06, 4.2655e-07, 2.4773e-07, 1.1660e-06, -1.9744e-07, 9.7789e-07], device='cuda:0') 100 0.0001 changing lr epoch 284, time 246.56, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4787 re_mapping 0.0035 re_causal 0.0108 /// teacc 99.15 lr 0.00010000 Epoch 286, weight, value: tensor([[-0.2167, -0.0515, 0.2072, ..., 0.0414, -0.2055, -0.1586], [-0.1627, -0.2445, -0.0565, ..., -0.1169, -0.2558, -0.0882], [ 0.2713, -0.0705, -0.2654, ..., -0.2018, -0.2074, -0.2179], ..., [-0.1054, -0.0288, -0.3295, ..., -0.1216, 0.1092, 0.0292], [-0.1921, -0.1318, -0.0397, ..., -0.2120, -0.0432, -0.0480], [-0.0805, -0.0954, -0.1362, ..., 0.0964, 0.0297, -0.1713]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., -1.8626e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], ..., [-0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, -2.7940e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -2.9802e-08, -3.7253e-09, 0.0000e+00]], device='cuda:0') Epoch 286, bias, value: tensor([ 0.0028, 0.0102, 0.0135, -0.0011, 0.0245, -0.0050, -0.0030, -0.0040, 0.0083, -0.0165], device='cuda:0'), grad: tensor([ 0.0000e+00, -8.1956e-08, 1.0245e-08, 4.6566e-09, 5.2154e-08, -2.7940e-09, 6.5193e-09, 4.6566e-08, 1.1176e-08, -4.9360e-08], device='cuda:0') 100 0.0001 changing lr epoch 285, time 246.64, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.5198 re_mapping 0.0034 re_causal 0.0109 /// teacc 99.12 lr 0.00010000 Epoch 287, weight, value: tensor([[-0.2167, -0.0518, 0.2075, ..., 0.0414, -0.2058, -0.1587], [-0.1629, -0.2454, -0.0565, ..., -0.1170, -0.2567, -0.0883], [ 0.2714, -0.0705, -0.2666, ..., -0.2026, -0.2086, -0.2187], ..., [-0.1054, -0.0294, -0.3340, ..., -0.1219, 0.1092, 0.0294], [-0.1921, -0.1320, -0.0402, ..., -0.2123, -0.0434, -0.0481], [-0.0806, -0.0955, -0.1364, ..., 0.0965, 0.0298, -0.1713]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.4214e-08, ..., -1.2107e-08, 2.7940e-09, 2.7940e-09], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.5832e-08, 9.3132e-10, 3.7253e-09], [ 9.3132e-10, 9.3132e-10, 2.7940e-09, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 2.7940e-08, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 1.5832e-08, 5.5879e-09, 2.7940e-09], [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., -9.8720e-08, -6.5193e-09, 2.7940e-09]], device='cuda:0') Epoch 287, bias, value: tensor([ 0.0028, 0.0102, 0.0129, -0.0010, 0.0246, -0.0050, -0.0030, -0.0041, 0.0086, -0.0164], device='cuda:0'), grad: tensor([-3.1665e-08, 1.9558e-08, 1.8626e-08, 2.0489e-08, 7.4506e-08, -9.4995e-08, 1.1455e-07, 1.0617e-07, -4.8429e-08, -1.7323e-07], device='cuda:0') 100 0.0001 changing lr epoch 286, time 246.47, cls_loss 0.0005 cls_loss_mapping 0.0024 cls_loss_causal 0.4721 re_mapping 0.0037 re_causal 0.0111 /// teacc 99.03 lr 0.00010000 Epoch 288, weight, value: tensor([[-0.2168, -0.0549, 0.2073, ..., 0.0411, -0.2061, -0.1588], [-0.1634, -0.2461, -0.0567, ..., -0.1174, -0.2589, -0.0901], [ 0.2720, -0.0706, -0.2673, ..., -0.2034, -0.2092, -0.2191], ..., [-0.1057, -0.0295, -0.3352, ..., -0.1217, 0.1099, 0.0321], [-0.1928, -0.1323, -0.0407, ..., -0.2126, -0.0440, -0.0491], [-0.0809, -0.0955, -0.1366, ..., 0.0964, 0.0293, -0.1716]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 9.3132e-10, -1.8533e-07, ..., -2.5425e-07, 1.8626e-09, 1.9558e-08], [ 5.5879e-09, 9.3132e-10, 1.8626e-09, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], [-2.8126e-07, 1.8626e-09, 9.3132e-10, ..., 5.5879e-09, 0.0000e+00, 0.0000e+00], ..., [ 2.6450e-07, 1.8626e-09, 0.0000e+00, ..., 2.7940e-09, -2.7940e-09, 1.8626e-09], [ 6.5193e-09, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, 0.0000e+00, 4.8429e-08], [ 2.7940e-09, 0.0000e+00, 1.9651e-07, ..., 2.6356e-07, 1.8626e-09, 1.8626e-08]], device='cuda:0') Epoch 288, bias, value: tensor([ 0.0019, 0.0102, 0.0129, -0.0010, 0.0247, -0.0046, -0.0029, -0.0039, 0.0080, -0.0167], device='cuda:0'), grad: tensor([-3.7532e-07, -2.7008e-08, -4.7125e-07, 2.9802e-08, -7.6368e-08, -7.8883e-07, 1.9837e-07, 5.1130e-07, 3.6601e-07, 6.1747e-07], device='cuda:0') 100 0.0001 changing lr epoch 287, time 246.68, cls_loss 0.0006 cls_loss_mapping 0.0023 cls_loss_causal 0.4984 re_mapping 0.0035 re_causal 0.0112 /// teacc 99.11 lr 0.00010000 Epoch 289, weight, value: tensor([[-0.2169, -0.0549, 0.2103, ..., 0.0439, -0.2062, -0.1585], [-0.1641, -0.2465, -0.0592, ..., -0.1195, -0.2594, -0.0906], [ 0.2733, -0.0707, -0.2680, ..., -0.2033, -0.2093, -0.2197], ..., [-0.1069, -0.0296, -0.3360, ..., -0.1219, 0.1100, 0.0320], [-0.1929, -0.1324, -0.0407, ..., -0.2126, -0.0439, -0.0491], [-0.0815, -0.0956, -0.1390, ..., 0.0952, 0.0293, -0.1716]], device='cuda:0'), grad: tensor([[ 7.9162e-08, 1.8626e-09, -6.5193e-09, ..., 6.5193e-09, 9.3132e-10, -1.8626e-09], [ 3.0734e-08, 1.2107e-08, 4.6566e-09, ..., 5.5879e-09, 1.8626e-09, 3.7253e-09], [-4.6473e-07, 3.3528e-08, 0.0000e+00, ..., -7.8231e-08, 9.3132e-10, 1.8626e-09], ..., [ 5.3085e-08, 2.3283e-08, 9.3132e-10, ..., 5.5879e-09, -4.6566e-09, 9.3132e-09], [ 3.4459e-08, 1.3970e-08, 9.3132e-10, ..., 1.8626e-09, 0.0000e+00, 2.7940e-09], [ 1.3970e-08, 1.8626e-09, 1.8626e-09, ..., 1.7695e-08, 1.8626e-09, 9.3132e-10]], device='cuda:0') Epoch 289, bias, value: tensor([ 0.0049, 0.0098, 0.0135, -0.0013, 0.0247, -0.0043, -0.0032, -0.0039, 0.0083, -0.0175], device='cuda:0'), grad: tensor([ 1.8254e-07, 1.2014e-07, -8.8569e-07, -1.9185e-07, -5.6811e-08, 2.5798e-07, 1.9372e-07, 1.7229e-07, 9.9652e-08, 1.0431e-07], device='cuda:0') 100 0.0001 changing lr epoch 288, time 246.98, cls_loss 0.0005 cls_loss_mapping 0.0018 cls_loss_causal 0.4949 re_mapping 0.0034 re_causal 0.0113 /// teacc 99.12 lr 0.00010000 Epoch 290, weight, value: tensor([[-0.2172, -0.0551, 0.2105, ..., 0.0440, -0.2064, -0.1587], [-0.1642, -0.2471, -0.0592, ..., -0.1196, -0.2600, -0.0907], [ 0.2736, -0.0708, -0.2692, ..., -0.2032, -0.2097, -0.2199], ..., [-0.1071, -0.0297, -0.3368, ..., -0.1220, 0.1101, 0.0320], [-0.1931, -0.1326, -0.0421, ..., -0.2127, -0.0439, -0.0490], [-0.0819, -0.0956, -0.1392, ..., 0.0951, 0.0292, -0.1717]], device='cuda:0'), grad: tensor([[ 1.7695e-08, 1.8626e-09, 9.3132e-09, ..., -0.0000e+00, 2.9802e-08, 1.2107e-08], [ 2.4214e-08, 1.8626e-09, 6.5193e-09, ..., 2.7008e-08, 4.2003e-07, 5.5879e-09], [-6.0350e-07, -5.7742e-08, 1.8626e-09, ..., 4.4703e-08, 2.2352e-08, -3.2596e-08], ..., [ 3.7253e-08, 3.7253e-09, 9.3132e-10, ..., 1.8626e-09, -6.7707e-07, 1.8626e-09], [ 3.9488e-07, 4.0978e-08, 6.5193e-09, ..., 6.5193e-09, 1.0245e-08, 2.7008e-08], [ 1.8626e-09, 0.0000e+00, 3.7253e-09, ..., 4.8429e-08, 1.4435e-07, 9.3132e-10]], device='cuda:0') Epoch 290, bias, value: tensor([ 0.0049, 0.0099, 0.0126, -0.0013, 0.0248, -0.0043, -0.0032, -0.0038, 0.0089, -0.0176], device='cuda:0'), grad: tensor([ 2.2165e-07, 2.6785e-06, -1.9819e-06, 3.5949e-07, -1.3039e-07, 2.3376e-07, -9.9652e-08, -3.9339e-06, 1.6056e-06, 1.0431e-06], device='cuda:0') 100 0.0001 changing lr epoch 289, time 246.77, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4844 re_mapping 0.0032 re_causal 0.0108 /// teacc 99.12 lr 0.00010000 Epoch 291, weight, value: tensor([[-0.2181, -0.0551, 0.2107, ..., 0.0440, -0.2068, -0.1590], [-0.1641, -0.2476, -0.0601, ..., -0.1200, -0.2613, -0.0910], [ 0.2740, -0.0708, -0.2705, ..., -0.2038, -0.2103, -0.2202], ..., [-0.1077, -0.0298, -0.3373, ..., -0.1221, 0.1102, 0.0320], [-0.1933, -0.1328, -0.0423, ..., -0.2127, -0.0439, -0.0491], [-0.0822, -0.0957, -0.1392, ..., 0.0950, 0.0292, -0.1717]], device='cuda:0'), grad: tensor([[ 5.5879e-09, 9.3132e-10, -8.6725e-06, ..., -5.7369e-06, 2.7940e-09, 2.7940e-09], [ 1.3970e-08, 1.8626e-09, 3.2857e-06, ..., 2.1849e-06, 5.5879e-08, 9.3132e-10], [-1.6298e-07, 9.3132e-10, 4.3772e-08, ..., 8.2888e-08, 6.5193e-09, 0.0000e+00], ..., [ 9.6858e-08, 5.5879e-09, 9.3132e-09, ..., 4.7497e-08, -6.5193e-08, -1.8626e-09], [ 5.5879e-09, 9.3132e-10, 7.4506e-09, ..., 8.3819e-09, 9.3132e-10, 1.8626e-09], [ 1.8626e-09, 9.3132e-10, 4.3772e-08, ..., 1.1548e-07, 1.2107e-08, -9.3132e-10]], device='cuda:0') Epoch 291, bias, value: tensor([ 0.0049, 0.0099, 0.0123, -0.0014, 0.0249, -0.0043, -0.0032, -0.0038, 0.0092, -0.0177], device='cuda:0'), grad: tensor([-2.1994e-05, 7.0147e-06, -1.7695e-08, 3.1665e-08, -1.6689e-06, 1.2163e-06, 1.3471e-05, 1.4072e-06, 1.7788e-07, 3.8743e-07], device='cuda:0') 100 0.0001 changing lr epoch 290, time 246.49, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4912 re_mapping 0.0033 re_causal 0.0108 /// teacc 99.09 lr 0.00010000 Epoch 292, weight, value: tensor([[-0.2179, -0.0554, 0.2113, ..., 0.0441, -0.2073, -0.1593], [-0.1647, -0.2482, -0.0604, ..., -0.1203, -0.2621, -0.0907], [ 0.2749, -0.0710, -0.2725, ..., -0.2048, -0.2110, -0.2210], ..., [-0.1084, -0.0300, -0.3386, ..., -0.1221, 0.1104, 0.0319], [-0.1935, -0.1330, -0.0427, ..., -0.2130, -0.0438, -0.0493], [-0.0829, -0.0957, -0.1396, ..., 0.0950, 0.0291, -0.1717]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 1.3411e-07, 1.4901e-08, 9.3132e-10], [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 1.4808e-07, 1.5553e-07, 0.0000e+00], [-5.5879e-09, 9.3132e-10, 0.0000e+00, ..., 9.7789e-08, 1.2107e-08, 0.0000e+00], ..., [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., -1.9595e-06, -3.5074e-06, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.9558e-08, 5.5879e-09, 0.0000e+00], [ 1.8626e-09, 1.8626e-09, 9.3132e-10, ..., 1.1018e-06, 3.2466e-06, 0.0000e+00]], device='cuda:0') Epoch 292, bias, value: tensor([ 0.0050, 0.0099, 0.0125, -0.0014, 0.0250, -0.0042, -0.0033, -0.0039, 0.0093, -0.0178], device='cuda:0'), grad: tensor([ 4.3027e-07, 1.0319e-06, 3.0175e-07, 3.2224e-07, 6.3051e-07, 4.5262e-07, 8.4750e-08, -1.9625e-05, 9.5926e-08, 1.6257e-05], device='cuda:0') 100 0.0001 changing lr epoch 291, time 246.99, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.4834 re_mapping 0.0032 re_causal 0.0103 /// teacc 99.10 lr 0.00010000 Epoch 293, weight, value: tensor([[-0.2187, -0.0559, 0.2119, ..., 0.0442, -0.2077, -0.1594], [-0.1649, -0.2492, -0.0607, ..., -0.1205, -0.2641, -0.0908], [ 0.2752, -0.0711, -0.2740, ..., -0.2063, -0.2119, -0.2211], ..., [-0.1088, -0.0302, -0.3403, ..., -0.1239, 0.1089, 0.0319], [-0.1942, -0.1333, -0.0431, ..., -0.2137, -0.0441, -0.0495], [-0.0832, -0.0958, -0.1396, ..., 0.0953, 0.0306, -0.1717]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1921e-07, ..., -5.4948e-08, 1.3970e-08, -4.6566e-09], [ 1.8626e-09, 0.0000e+00, 2.7940e-09, ..., -0.0000e+00, 3.7253e-09, 9.3132e-10], [-2.7940e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, 1.8626e-09, 9.3132e-10], ..., [ 2.7940e-09, 9.3132e-10, 9.3132e-10, ..., 1.8626e-09, -4.6566e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, -1.8626e-09, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., -1.8626e-08, -3.7253e-09, 9.3132e-10]], device='cuda:0') Epoch 293, bias, value: tensor([ 0.0050, 0.0102, 0.0121, -0.0014, 0.0251, -0.0040, -0.0035, -0.0047, 0.0087, -0.0174], device='cuda:0'), grad: tensor([-1.4994e-07, 4.6566e-09, 1.0245e-08, -4.0047e-08, 4.7497e-08, 4.4331e-07, -2.6729e-07, 7.4506e-09, -4.5635e-08, -4.6566e-09], device='cuda:0') 100 0.0001 changing lr epoch 292, time 246.59, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4715 re_mapping 0.0032 re_causal 0.0104 /// teacc 99.12 lr 0.00010000 Epoch 294, weight, value: tensor([[-0.2185, -0.0561, 0.2122, ..., 0.0442, -0.2079, -0.1595], [-0.1650, -0.2502, -0.0609, ..., -0.1207, -0.2653, -0.0909], [ 0.2758, -0.0711, -0.2749, ..., -0.2072, -0.2126, -0.2212], ..., [-0.1097, -0.0303, -0.3409, ..., -0.1227, 0.1100, 0.0320], [-0.1946, -0.1337, -0.0435, ..., -0.2138, -0.0444, -0.0498], [-0.0838, -0.0958, -0.1398, ..., 0.0949, 0.0296, -0.1717]], device='cuda:0'), grad: tensor([[-3.5390e-08, 0.0000e+00, -1.1828e-07, ..., -7.0781e-08, 9.3132e-10, 9.3132e-10], [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 1.2107e-08, 7.4506e-09, 9.3132e-10], [ 3.7253e-09, 0.0000e+00, 1.2107e-08, ..., 7.4506e-09, 9.3132e-10, 0.0000e+00], ..., [ 3.7253e-09, 0.0000e+00, 1.1176e-08, ..., 3.5390e-08, -1.3039e-08, 0.0000e+00], [ 2.7940e-09, 0.0000e+00, 1.2107e-08, ..., 4.6566e-09, 2.7940e-09, 2.7940e-09], [ 1.4901e-08, 0.0000e+00, 5.1223e-08, ..., -5.0291e-08, 3.7253e-09, 0.0000e+00]], device='cuda:0') Epoch 294, bias, value: tensor([ 0.0050, 0.0103, 0.0120, -0.0014, 0.0252, -0.0039, -0.0034, -0.0042, 0.0088, -0.0181], device='cuda:0'), grad: tensor([-3.2596e-07, -5.5879e-09, 9.4064e-08, 6.9477e-07, 1.4622e-07, -7.0781e-07, 1.4435e-07, 1.5087e-07, -2.5705e-07, 6.1467e-08], device='cuda:0') 100 0.0001 changing lr epoch 293, time 246.84, cls_loss 0.0005 cls_loss_mapping 0.0020 cls_loss_causal 0.5207 re_mapping 0.0032 re_causal 0.0107 /// teacc 99.14 lr 0.00010000 Epoch 295, weight, value: tensor([[-0.2187, -0.0561, 0.2102, ..., 0.0443, -0.2101, -0.1606], [-0.1652, -0.2507, -0.0610, ..., -0.1209, -0.2678, -0.0910], [ 0.2760, -0.0713, -0.2765, ..., -0.2079, -0.2136, -0.2212], ..., [-0.1096, -0.0304, -0.3428, ..., -0.1230, 0.1101, 0.0320], [-0.1953, -0.1343, -0.0442, ..., -0.2143, -0.0445, -0.0501], [-0.0839, -0.0955, -0.1399, ..., 0.0949, 0.0295, -0.1717]], device='cuda:0'), grad: tensor([[ 2.1420e-08, 9.3132e-10, -1.0803e-07, ..., -2.7940e-08, 0.0000e+00, 9.3132e-10], [ 6.8918e-08, 1.8626e-09, 5.5879e-09, ..., -1.8626e-09, 2.7940e-09, 1.8626e-09], [-6.7893e-07, 1.6764e-08, 1.3970e-08, ..., 3.7253e-09, 0.0000e+00, 3.7253e-09], ..., [ 4.3679e-07, 1.4901e-08, 6.5193e-09, ..., 2.7940e-09, -6.5193e-09, 3.7253e-09], [ 1.5832e-07, 2.7940e-09, 1.6764e-08, ..., 3.7253e-09, 1.8626e-09, 1.8626e-09], [ 2.7940e-08, 3.7253e-09, 3.4459e-08, ..., 1.9558e-08, 3.7253e-09, 0.0000e+00]], device='cuda:0') Epoch 295, bias, value: tensor([ 0.0037, 0.0102, 0.0114, -0.0015, 0.0253, -0.0038, -0.0023, -0.0041, 0.0084, -0.0181], device='cuda:0'), grad: tensor([-6.9849e-08, -8.1584e-07, -1.8450e-06, -2.8592e-07, 2.4214e-08, -4.7497e-08, 1.2852e-07, 1.3448e-06, 1.2908e-06, 2.7288e-07], device='cuda:0') 100 0.0001 changing lr epoch 294, time 246.55, cls_loss 0.0007 cls_loss_mapping 0.0022 cls_loss_causal 0.4760 re_mapping 0.0033 re_causal 0.0101 /// teacc 99.13 lr 0.00010000 Epoch 296, weight, value: tensor([[-0.2190, -0.0562, 0.2100, ..., 0.0442, -0.2109, -0.1610], [-0.1654, -0.2515, -0.0612, ..., -0.1239, -0.2690, -0.0912], [ 0.2767, -0.0714, -0.2772, ..., -0.2087, -0.2142, -0.2213], ..., [-0.1100, -0.0306, -0.3438, ..., -0.1219, 0.1123, 0.0319], [-0.1961, -0.1346, -0.0461, ..., -0.2157, -0.0451, -0.0511], [-0.0842, -0.0956, -0.1396, ..., 0.0953, 0.0275, -0.1717]], device='cuda:0'), grad: tensor([[ 1.3039e-08, 0.0000e+00, 0.0000e+00, ..., 3.6322e-08, 4.6566e-09, 0.0000e+00], [ 1.3411e-07, 0.0000e+00, 0.0000e+00, ..., 5.8673e-08, 5.6811e-08, 0.0000e+00], [ 6.5006e-07, 0.0000e+00, 0.0000e+00, ..., 3.9674e-07, 3.9488e-07, 0.0000e+00], ..., [-1.0543e-06, 0.0000e+00, 0.0000e+00, ..., -5.3924e-07, -5.7463e-07, 0.0000e+00], [ 1.3039e-08, 0.0000e+00, 0.0000e+00, ..., 4.4703e-08, 4.6566e-09, 0.0000e+00], [ 6.6124e-08, 0.0000e+00, 9.3132e-10, ..., -1.4901e-07, 2.8871e-08, 0.0000e+00]], device='cuda:0') Epoch 296, bias, value: tensor([ 0.0033, 0.0112, 0.0116, -0.0014, 0.0251, -0.0027, -0.0035, -0.0043, 0.0074, -0.0186], device='cuda:0'), grad: tensor([ 1.3504e-07, 4.5076e-07, 4.2357e-06, 7.3854e-07, 2.9523e-07, -6.8918e-08, 1.0524e-07, -6.0797e-06, 1.4249e-07, 3.5390e-08], device='cuda:0') 100 0.0001 changing lr epoch 295, time 246.73, cls_loss 0.0006 cls_loss_mapping 0.0028 cls_loss_causal 0.4764 re_mapping 0.0033 re_causal 0.0102 /// teacc 99.20 lr 0.00010000 Epoch 297, weight, value: tensor([[-0.2187, -0.0562, 0.2105, ..., 0.0439, -0.2111, -0.1612], [-0.1656, -0.2520, -0.0612, ..., -0.1238, -0.2697, -0.0914], [ 0.2770, -0.0715, -0.2783, ..., -0.2103, -0.2152, -0.2217], ..., [-0.1101, -0.0308, -0.3448, ..., -0.1222, 0.1123, 0.0317], [-0.1963, -0.1348, -0.0465, ..., -0.2153, -0.0433, -0.0512], [-0.0849, -0.0956, -0.1403, ..., 0.0956, 0.0277, -0.1718]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.0896e-07, ..., -2.5146e-08, 2.7940e-09, 7.4506e-09], [ 1.0245e-08, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, 3.7253e-09, 4.6566e-09], [-4.0978e-08, -1.8626e-09, 2.7940e-09, ..., 9.3132e-10, 9.3132e-10, 9.3132e-10], ..., [ 8.3819e-09, 0.0000e+00, 4.6566e-09, ..., 2.9802e-08, 3.7253e-09, 4.6566e-09], [ 1.3970e-08, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, 9.4064e-08, 1.5926e-07], [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., -8.8476e-08, 9.3132e-10, 1.1176e-08]], device='cuda:0') Epoch 297, bias, value: tensor([ 0.0031, 0.0116, 0.0113, -0.0015, 0.0251, -0.0026, -0.0039, -0.0046, 0.0086, -0.0185], device='cuda:0'), grad: tensor([-1.7788e-07, -5.8673e-08, -7.9162e-08, 4.6566e-08, 1.7416e-07, -2.2277e-06, 1.7285e-06, 1.9185e-07, 5.8580e-07, -1.7788e-07], device='cuda:0') 100 0.0001 changing lr epoch 296, time 246.59, cls_loss 0.0005 cls_loss_mapping 0.0016 cls_loss_causal 0.4677 re_mapping 0.0032 re_causal 0.0104 /// teacc 99.17 lr 0.00010000 Epoch 298, weight, value: tensor([[-0.2182, -0.0562, 0.2118, ..., 0.0444, -0.2112, -0.1612], [-0.1664, -0.2524, -0.0641, ..., -0.1252, -0.2705, -0.0915], [ 0.2779, -0.0715, -0.2797, ..., -0.2123, -0.2158, -0.2223], ..., [-0.1105, -0.0309, -0.3465, ..., -0.1222, 0.1123, 0.0317], [-0.1965, -0.1350, -0.0456, ..., -0.2154, -0.0427, -0.0503], [-0.0854, -0.0957, -0.1406, ..., 0.0956, 0.0277, -0.1718]], device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, 2.3283e-08, ..., 3.2596e-09, 1.6764e-08, 9.3132e-09], [ 2.0955e-08, 0.0000e+00, 1.7695e-08, ..., -1.8626e-09, 2.3749e-08, 8.8476e-09], [-3.9581e-08, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, 4.1910e-09, 1.3970e-09], ..., [ 6.9849e-09, 4.6566e-10, 0.0000e+00, ..., 2.7474e-08, -3.6322e-08, 0.0000e+00], [ 5.1223e-09, 4.6566e-10, 2.0489e-08, ..., 2.7940e-09, 1.6764e-08, 1.0245e-08], [ 1.3970e-09, 0.0000e+00, 1.3970e-09, ..., -1.4435e-08, 1.3970e-08, 0.0000e+00]], device='cuda:0') Epoch 298, bias, value: tensor([ 0.0041, 0.0113, 0.0118, -0.0015, 0.0251, -0.0026, -0.0045, -0.0046, 0.0102, -0.0185], device='cuda:0'), grad: tensor([ 1.3551e-07, -1.7928e-07, -2.9337e-08, -3.4925e-08, 1.1642e-08, 1.0338e-07, -3.6228e-07, 5.9139e-08, 1.5879e-07, 1.3504e-07], device='cuda:0') 100 0.0001 changing lr epoch 297, time 246.73, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4799 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.12 lr 0.00010000 Epoch 299, weight, value: tensor([[-0.2188, -0.0564, 0.2119, ..., 0.0445, -0.2115, -0.1614], [-0.1668, -0.2528, -0.0641, ..., -0.1254, -0.2714, -0.0916], [ 0.2787, -0.0716, -0.2812, ..., -0.2136, -0.2168, -0.2231], ..., [-0.1108, -0.0309, -0.3472, ..., -0.1223, 0.1125, 0.0317], [-0.1969, -0.1352, -0.0465, ..., -0.2155, -0.0428, -0.0511], [-0.0868, -0.0957, -0.1410, ..., 0.0956, 0.0277, -0.1718]], device='cuda:0'), grad: tensor([[ 1.2107e-08, 2.7940e-09, 1.5367e-08, ..., 8.3819e-09, 1.7229e-08, 4.6566e-09], [ 9.3132e-09, 4.6566e-10, 7.9162e-09, ..., 1.3039e-08, 5.2620e-08, 2.3283e-09], [-4.8056e-07, 4.6566e-10, 4.1910e-09, ..., 4.1910e-09, 1.7695e-08, 1.3970e-09], ..., [ 4.3027e-07, 9.3132e-10, 0.0000e+00, ..., 1.1642e-08, -1.1176e-07, 0.0000e+00], [ 3.3062e-08, 3.2596e-09, 8.3353e-08, ..., 1.0710e-08, 1.2107e-08, 2.5611e-08], [ 8.3819e-09, 6.5193e-09, 5.1223e-09, ..., -9.2201e-08, 1.4901e-08, 1.3970e-09]], device='cuda:0') Epoch 299, bias, value: tensor([ 0.0041, 0.0112, 0.0116, -0.0016, 0.0252, -0.0028, -0.0043, -0.0045, 0.0103, -0.0185], device='cuda:0'), grad: tensor([ 1.9372e-07, 3.6741e-07, -1.0096e-06, 4.0280e-07, 1.2759e-07, -5.1036e-07, -1.4808e-07, 4.3353e-07, -1.4389e-07, 3.1106e-07], device='cuda:0') 100 0.0001 changing lr epoch 298, time 246.94, cls_loss 0.0005 cls_loss_mapping 0.0021 cls_loss_causal 0.4658 re_mapping 0.0033 re_causal 0.0106 /// teacc 99.06 lr 0.00010000 Epoch 300, weight, value: tensor([[-0.2191, -0.0565, 0.2120, ..., 0.0445, -0.2119, -0.1616], [-0.1672, -0.2538, -0.0645, ..., -0.1254, -0.2730, -0.0928], [ 0.2798, -0.0717, -0.2819, ..., -0.2151, -0.2173, -0.2232], ..., [-0.1118, -0.0311, -0.3478, ..., -0.1225, 0.1125, 0.0316], [-0.1973, -0.1358, -0.0465, ..., -0.2156, -0.0428, -0.0515], [-0.0876, -0.0958, -0.1411, ..., 0.0957, 0.0278, -0.1718]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 7.9768e-07, ..., 4.7497e-08, 1.5832e-07, 1.4715e-07], [ 3.2596e-09, 0.0000e+00, 4.1910e-09, ..., 7.4971e-08, 3.5390e-08, 9.3132e-10], [-9.7789e-09, 0.0000e+00, 2.7940e-09, ..., 2.3283e-09, 1.8626e-09, 4.6566e-10], ..., [ 3.2596e-09, 0.0000e+00, 0.0000e+00, ..., -5.6531e-07, -4.8475e-07, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 4.6566e-09, ..., 5.1223e-09, 6.5193e-09, 9.3132e-10], [ 0.0000e+00, 4.6566e-10, 3.2596e-09, ..., 6.1328e-07, 4.4284e-07, 4.6566e-10]], device='cuda:0') Epoch 300, bias, value: tensor([ 0.0041, 0.0112, 0.0120, -0.0015, 0.0251, -0.0023, -0.0047, -0.0047, 0.0107, -0.0185], device='cuda:0'), grad: tensor([ 1.4342e-06, 4.4797e-07, 1.1176e-08, 8.9407e-08, -3.1525e-07, -1.3830e-07, -1.4417e-06, -3.1702e-06, 1.1595e-07, 2.9784e-06], device='cuda:0') 100 0.0001 changing lr epoch 299, time 246.64, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4549 re_mapping 0.0034 re_causal 0.0103 /// teacc 99.13 lr 0.00010000 Epoch 301, weight, value: tensor([[-0.2194, -0.0565, 0.2120, ..., 0.0442, -0.2123, -0.1619], [-0.1684, -0.2543, -0.0647, ..., -0.1257, -0.2740, -0.0929], [ 0.2814, -0.0717, -0.2826, ..., -0.2158, -0.2182, -0.2233], ..., [-0.1121, -0.0314, -0.3482, ..., -0.1226, 0.1126, 0.0316], [-0.1980, -0.1359, -0.0468, ..., -0.2159, -0.0432, -0.0521], [-0.0887, -0.0958, -0.1412, ..., 0.0959, 0.0278, -0.1718]], device='cuda:0'), grad: tensor([[ 3.5390e-08, 0.0000e+00, 8.3819e-09, ..., 3.2596e-09, 4.6566e-09, 1.6764e-08], [ 6.7987e-08, 0.0000e+00, 2.7940e-09, ..., 1.3504e-08, 1.2899e-07, 4.1910e-09], [ 2.1886e-08, 0.0000e+00, 2.0023e-08, ..., 8.8476e-09, 6.6124e-08, 3.8184e-08], ..., [-1.2433e-07, 0.0000e+00, 0.0000e+00, ..., 1.1176e-08, -2.5565e-07, 0.0000e+00], [ 6.5193e-09, 0.0000e+00, 6.0536e-09, ..., 4.1910e-09, 1.4435e-08, 4.1910e-09], [ 3.8184e-08, 0.0000e+00, 3.2596e-09, ..., 1.9372e-07, 2.1886e-08, 1.8626e-09]], device='cuda:0') Epoch 301, bias, value: tensor([ 0.0037, 0.0110, 0.0118, -0.0016, 0.0253, -0.0024, -0.0047, -0.0045, 0.0107, -0.0184], device='cuda:0'), grad: tensor([ 1.9977e-07, 1.0431e-06, 5.6392e-07, 1.5218e-06, -3.2037e-07, -1.3150e-06, -7.3807e-07, -1.8440e-06, -9.3132e-09, 9.1456e-07], device='cuda:0') 100 0.0001 changing lr epoch 300, time 246.40, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.4682 re_mapping 0.0034 re_causal 0.0102 /// teacc 99.13 lr 0.00010000 Epoch 302, weight, value: tensor([[-0.2203, -0.0566, 0.2121, ..., 0.0442, -0.2127, -0.1621], [-0.1686, -0.2548, -0.0647, ..., -0.1258, -0.2752, -0.0930], [ 0.2820, -0.0718, -0.2835, ..., -0.2169, -0.2194, -0.2217], ..., [-0.1126, -0.0316, -0.3489, ..., -0.1228, 0.1126, 0.0314], [-0.1983, -0.1361, -0.0475, ..., -0.2160, -0.0436, -0.0526], [-0.0895, -0.0961, -0.1414, ..., 0.0959, 0.0278, -0.1718]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.8626e-09, ..., 4.6566e-09, 1.2107e-08, 4.1910e-09], [ 4.1910e-09, 0.0000e+00, 3.7253e-09, ..., 2.5611e-08, 4.1910e-09, 2.3283e-09], [-1.5041e-07, 0.0000e+00, 2.7940e-09, ..., 2.3749e-08, 1.3970e-09, 1.8626e-09], ..., [ 1.3737e-07, 4.6566e-10, 0.0000e+00, ..., 5.2154e-08, 8.3819e-09, 9.3132e-10], [ 9.3132e-09, 0.0000e+00, 8.3819e-09, ..., 3.2131e-08, 9.3132e-09, 2.3283e-09], [ 9.3132e-10, 0.0000e+00, 6.9849e-09, ..., -3.3528e-08, -3.7253e-09, 1.3970e-09]], device='cuda:0') Epoch 302, bias, value: tensor([ 0.0036, 0.0114, 0.0105, -0.0015, 0.0254, -0.0031, -0.0037, -0.0047, 0.0108, -0.0185], device='cuda:0'), grad: tensor([ 7.9628e-08, 8.5682e-08, -1.4761e-07, 2.9802e-08, -4.8382e-07, 2.6356e-07, -2.1607e-07, 4.0233e-07, 2.9802e-08, -2.3749e-08], device='cuda:0') 100 0.0001 changing lr epoch 301, time 246.42, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4819 re_mapping 0.0034 re_causal 0.0106 /// teacc 99.07 lr 0.00010000 Epoch 303, weight, value: tensor([[-0.2203, -0.0567, 0.2123, ..., 0.0443, -0.2131, -0.1622], [-0.1689, -0.2558, -0.0648, ..., -0.1257, -0.2757, -0.0931], [ 0.2824, -0.0720, -0.2839, ..., -0.2183, -0.2198, -0.2216], ..., [-0.1129, -0.0317, -0.3490, ..., -0.1229, 0.1127, 0.0315], [-0.1988, -0.1365, -0.0479, ..., -0.2161, -0.0440, -0.0528], [-0.0900, -0.0960, -0.1415, ..., 0.0958, 0.0278, -0.1718]], device='cuda:0'), grad: tensor([[ 4.1444e-08, 1.3970e-09, -2.1607e-07, ..., -1.5041e-07, 4.6566e-09, 1.3970e-09], [ 2.4680e-08, 1.8626e-09, 3.7253e-09, ..., 2.7940e-09, 2.8452e-07, 4.6566e-10], [-2.7614e-07, -2.7940e-09, 1.3970e-09, ..., 1.3970e-09, 9.3132e-10, 4.6566e-10], ..., [ 2.6543e-08, 4.6566e-09, 2.3283e-09, ..., 4.1910e-09, -3.8883e-07, 4.6566e-10], [ 1.9697e-07, 8.3819e-09, 3.2596e-09, ..., 1.3970e-09, 2.3283e-09, 9.3132e-10], [ 2.7940e-09, 4.6566e-10, 1.8114e-07, ..., 1.2061e-07, 7.9628e-08, 0.0000e+00]], device='cuda:0') Epoch 303, bias, value: tensor([ 0.0036, 0.0129, 0.0104, -0.0013, 0.0255, -0.0033, -0.0035, -0.0061, 0.0106, -0.0186], device='cuda:0'), grad: tensor([-1.2200e-07, 5.9158e-06, 8.4285e-08, 8.1491e-08, 1.0291e-07, 5.8673e-08, 6.7055e-08, -1.0394e-05, 5.4250e-07, 3.6657e-06], device='cuda:0') 100 0.0001 changing lr epoch 302, time 246.55, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4950 re_mapping 0.0037 re_causal 0.0114 /// teacc 99.06 lr 0.00010000 Epoch 304, weight, value: tensor([[-0.2208, -0.0567, 0.2123, ..., 0.0443, -0.2134, -0.1624], [-0.1695, -0.2563, -0.0649, ..., -0.1258, -0.2780, -0.0935], [ 0.2836, -0.0720, -0.2852, ..., -0.2194, -0.2209, -0.2217], ..., [-0.1131, -0.0317, -0.3493, ..., -0.1229, 0.1129, 0.0318], [-0.2006, -0.1368, -0.0485, ..., -0.2163, -0.0444, -0.0532], [-0.0903, -0.0961, -0.1415, ..., 0.0957, 0.0277, -0.1719]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, 6.5193e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3039e-08, -5.5879e-09, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.0675e-07, 7.8231e-08, 0.0000e+00]], device='cuda:0') Epoch 304, bias, value: tensor([ 0.0036, 0.0128, 0.0109, 0.0004, 0.0256, -0.0047, -0.0035, -0.0059, 0.0101, -0.0188], device='cuda:0'), grad: tensor([ 2.7940e-09, 4.0978e-08, 5.5879e-09, 5.5879e-09, -4.6380e-07, 1.8626e-09, 8.3819e-09, -7.4506e-09, 1.4901e-08, 3.9767e-07], device='cuda:0') 100 0.0001 changing lr epoch 303, time 246.36, cls_loss 0.0005 cls_loss_mapping 0.0017 cls_loss_causal 0.4523 re_mapping 0.0035 re_causal 0.0106 /// teacc 99.17 lr 0.00010000 Epoch 305, weight, value: tensor([[-0.2201, -0.0567, 0.2125, ..., 0.0444, -0.2137, -0.1625], [-0.1704, -0.2566, -0.0649, ..., -0.1258, -0.2789, -0.0936], [ 0.2844, -0.0721, -0.2865, ..., -0.2201, -0.2216, -0.2217], ..., [-0.1133, -0.0317, -0.3495, ..., -0.1236, 0.1125, 0.0318], [-0.2008, -0.1369, -0.0489, ..., -0.2164, -0.0446, -0.0537], [-0.0909, -0.0961, -0.1418, ..., 0.0959, 0.0281, -0.1719]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., 2.4214e-08, 6.5193e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.4904e-06, 5.8301e-07, 0.0000e+00], [-2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 7.4506e-09, 1.8626e-09, 0.0000e+00], ..., [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-06, 6.4727e-07, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, -1.8626e-09, ..., 1.3039e-08, 2.7940e-09, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -6.0089e-06, -1.4007e-06, 9.3132e-10]], device='cuda:0') Epoch 305, bias, value: tensor([ 0.0036, 0.0105, 0.0112, 0.0006, 0.0256, -0.0046, -0.0037, -0.0039, 0.0097, -0.0186], device='cuda:0'), grad: tensor([ 1.3225e-07, 1.0341e-05, 1.0338e-07, 7.7300e-08, 2.8554e-06, 9.3132e-09, 5.2154e-08, 1.2375e-05, -3.9116e-08, -2.5883e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 304---------------------------------------------------- epoch 304, time 262.99, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4683 re_mapping 0.0033 re_causal 0.0102 /// teacc 99.24 lr 0.00010000 Epoch 306, weight, value: tensor([[-0.2217, -0.0567, 0.2125, ..., 0.0443, -0.2139, -0.1626], [-0.1712, -0.2571, -0.0650, ..., -0.1261, -0.2808, -0.0938], [ 0.2887, -0.0724, -0.2866, ..., -0.2204, -0.2230, -0.2217], ..., [-0.1176, -0.0319, -0.3498, ..., -0.1237, 0.1126, 0.0317], [-0.2020, -0.1372, -0.0493, ..., -0.2167, -0.0449, -0.0539], [-0.0913, -0.0962, -0.1419, ..., 0.0959, 0.0281, -0.1719]], device='cuda:0'), grad: tensor([[ 9.8720e-08, 0.0000e+00, 4.0047e-08, ..., -9.3132e-10, 1.7695e-08, 4.6566e-09], [ 2.7940e-08, 0.0000e+00, 9.3132e-10, ..., -0.0000e+00, 9.3132e-10, 9.3132e-10], [-1.5814e-06, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, 1.8626e-09, 0.0000e+00], ..., [ 1.2387e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 2.5146e-08, 0.0000e+00, -8.3819e-09, ..., 0.0000e+00, -1.0245e-08, 2.7940e-09], [ 1.9558e-08, 0.0000e+00, 8.3819e-09, ..., 1.2107e-08, 7.4506e-09, 9.3132e-10]], device='cuda:0') Epoch 306, bias, value: tensor([ 0.0034, 0.0106, 0.0138, 0.0006, 0.0258, -0.0047, -0.0036, -0.0058, 0.0095, -0.0187], device='cuda:0'), grad: tensor([ 3.7160e-07, 6.9849e-08, -3.4608e-06, 2.3153e-06, 1.9558e-08, 2.5611e-07, -1.9465e-07, 5.4203e-07, -1.0524e-07, 1.6857e-07], device='cuda:0') 100 0.0001 changing lr epoch 305, time 246.41, cls_loss 0.0007 cls_loss_mapping 0.0012 cls_loss_causal 0.4697 re_mapping 0.0033 re_causal 0.0101 /// teacc 99.16 lr 0.00010000 Epoch 307, weight, value: tensor([[-0.2226, -0.0567, 0.2129, ..., 0.0443, -0.2142, -0.1626], [-0.1723, -0.2581, -0.0650, ..., -0.1263, -0.2842, -0.0939], [ 0.2891, -0.0739, -0.2875, ..., -0.2208, -0.2250, -0.2210], ..., [-0.1178, -0.0320, -0.3513, ..., -0.1238, 0.1130, 0.0317], [-0.2038, -0.1381, -0.0498, ..., -0.2171, -0.0454, -0.0544], [-0.0918, -0.0960, -0.1421, ..., 0.0959, 0.0279, -0.1719]], device='cuda:0'), grad: tensor([[ 1.4901e-08, 0.0000e+00, -1.4901e-08, ..., -4.6566e-09, 1.0245e-08, 9.3132e-10], [ 9.2201e-08, 1.8626e-09, 9.3132e-10, ..., 0.0000e+00, 1.0710e-07, 0.0000e+00], [-1.5274e-07, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 1.5832e-08, 0.0000e+00], ..., [ 2.5146e-08, 2.7940e-09, 0.0000e+00, ..., 3.7253e-09, -2.0117e-07, 0.0000e+00], [ 1.3970e-08, 9.3132e-10, 5.5879e-09, ..., 1.8626e-09, 1.4901e-08, 4.6566e-09], [ 5.5879e-09, 0.0000e+00, 5.5879e-09, ..., -7.4506e-09, 4.1910e-08, 0.0000e+00]], device='cuda:0') Epoch 307, bias, value: tensor([ 0.0035, 0.0107, 0.0140, 0.0006, 0.0260, -0.0047, -0.0037, -0.0059, 0.0089, -0.0188], device='cuda:0'), grad: tensor([ 7.3574e-08, 1.0878e-06, -2.4028e-07, 2.6077e-08, 8.7544e-08, -1.2480e-07, 1.0058e-07, -1.5385e-06, 1.5087e-07, 3.8557e-07], device='cuda:0') 100 0.0001 changing lr epoch 306, time 246.43, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4284 re_mapping 0.0033 re_causal 0.0099 /// teacc 99.12 lr 0.00010000 Epoch 308, weight, value: tensor([[-0.2229, -0.0568, 0.2132, ..., 0.0443, -0.2146, -0.1628], [-0.1729, -0.2584, -0.0651, ..., -0.1268, -0.2895, -0.0941], [ 0.2892, -0.0741, -0.2879, ..., -0.2211, -0.2271, -0.2191], ..., [-0.1178, -0.0322, -0.3521, ..., -0.1248, 0.1128, 0.0317], [-0.2041, -0.1383, -0.0502, ..., -0.2173, -0.0461, -0.0549], [-0.0926, -0.0961, -0.1425, ..., 0.0961, 0.0283, -0.1719]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 1.1176e-08, 3.7253e-09, ..., 3.7998e-07, 4.6566e-09, 2.7940e-09], [ 2.3283e-08, 3.2596e-08, 9.3132e-10, ..., 4.4703e-08, 2.7008e-08, 1.8626e-09], [ 7.0781e-08, 9.6764e-07, 0.0000e+00, ..., 2.7008e-08, 2.2352e-08, 9.3132e-10], ..., [ 4.0047e-08, 1.0990e-07, 0.0000e+00, ..., 5.7742e-08, -4.8429e-08, 9.3132e-10], [ 2.7940e-09, 4.1910e-08, 1.8626e-09, ..., 2.2817e-07, 1.4249e-07, 8.1025e-08], [ 1.8626e-09, -3.4645e-07, 0.0000e+00, ..., -3.7253e-06, -1.2573e-07, 3.7253e-09]], device='cuda:0') Epoch 308, bias, value: tensor([ 0.0037, 0.0106, 0.0140, 0.0003, 0.0261, -0.0045, -0.0036, -0.0060, 0.0087, -0.0184], device='cuda:0'), grad: tensor([ 1.3718e-06, 3.8836e-07, 4.4070e-06, -2.1178e-06, 5.1558e-06, -1.1576e-06, 6.5938e-07, 5.2247e-07, 1.3616e-06, -1.0580e-05], device='cuda:0') 100 0.0001 changing lr epoch 307, time 246.12, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4796 re_mapping 0.0033 re_causal 0.0099 /// teacc 98.97 lr 0.00010000 Epoch 309, weight, value: tensor([[-0.2233, -0.0568, 0.2135, ..., 0.0440, -0.2149, -0.1629], [-0.1732, -0.2590, -0.0651, ..., -0.1270, -0.2943, -0.0941], [ 0.2893, -0.0748, -0.2885, ..., -0.2215, -0.2300, -0.2189], ..., [-0.1179, -0.0323, -0.3528, ..., -0.1251, 0.1132, 0.0318], [-0.2044, -0.1385, -0.0507, ..., -0.2182, -0.0466, -0.0553], [-0.0927, -0.0959, -0.1428, ..., 0.0958, 0.0283, -0.1720]], device='cuda:0'), grad: tensor([[-2.6077e-08, 0.0000e+00, 9.9652e-08, ..., 7.4785e-07, 1.0245e-08, 1.8626e-09], [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 3.2596e-08, 1.0245e-08, 9.3132e-10], [ 1.8626e-09, 0.0000e+00, 1.2107e-08, ..., 4.6566e-09, 9.3132e-10, 9.3132e-10], ..., [ 2.7940e-09, 0.0000e+00, 8.3819e-09, ..., 5.4948e-08, 1.3039e-08, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 2.3283e-08, 4.6566e-09, 9.3132e-10], [ 1.0245e-08, 0.0000e+00, -1.4063e-07, ..., -2.3916e-06, -2.4401e-07, 0.0000e+00]], device='cuda:0') Epoch 309, bias, value: tensor([ 0.0034, 0.0105, 0.0140, 0.0001, 0.0269, -0.0045, -0.0036, -0.0060, 0.0079, -0.0188], device='cuda:0'), grad: tensor([ 1.1083e-06, 8.1956e-08, 4.2841e-08, 3.9116e-08, 3.5688e-06, 2.1048e-07, -1.5181e-07, 2.3097e-07, 8.1025e-08, -5.2229e-06], device='cuda:0') 100 0.0001 changing lr epoch 308, time 246.61, cls_loss 0.0007 cls_loss_mapping 0.0020 cls_loss_causal 0.4903 re_mapping 0.0032 re_causal 0.0102 /// teacc 99.15 lr 0.00010000 Epoch 310, weight, value: tensor([[-0.2226, -0.0568, 0.2150, ..., 0.0443, -0.2152, -0.1628], [-0.1739, -0.2590, -0.0662, ..., -0.1272, -0.2954, -0.0943], [ 0.2893, -0.0748, -0.2900, ..., -0.2223, -0.2316, -0.2184], ..., [-0.1179, -0.0323, -0.3545, ..., -0.1252, 0.1136, 0.0318], [-0.2045, -0.1386, -0.0516, ..., -0.2185, -0.0472, -0.0565], [-0.0933, -0.0960, -0.1439, ..., 0.0970, 0.0283, -0.1720]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-08, ..., -1.7695e-08, 2.7940e-09, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 2.0489e-08, ..., 9.3132e-09, 2.2352e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 8.3819e-09, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 2.7940e-09, -5.9605e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 1.8626e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., -9.3132e-09, 1.3039e-08, 0.0000e+00]], device='cuda:0') Epoch 310, bias, value: tensor([ 4.3602e-03, 1.0489e-02, 1.3952e-02, -5.2807e-05, 2.5495e-02, -3.7253e-03, -4.8555e-03, -5.8842e-03, 7.9361e-03, -1.7865e-02], device='cuda:0'), grad: tensor([-2.5146e-08, -4.4983e-07, 8.4750e-08, 8.0094e-08, 4.1910e-08, -2.7940e-08, 3.8184e-08, -3.7253e-09, 3.8184e-08, 2.4680e-07], device='cuda:0') 100 0.0001 changing lr epoch 309, time 246.43, cls_loss 0.0005 cls_loss_mapping 0.0018 cls_loss_causal 0.4725 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.18 lr 0.00010000 Epoch 311, weight, value: tensor([[-0.2226, -0.0568, 0.2153, ..., 0.0443, -0.2157, -0.1628], [-0.1744, -0.2591, -0.0663, ..., -0.1273, -0.2971, -0.0943], [ 0.2893, -0.0748, -0.2907, ..., -0.2233, -0.2331, -0.2186], ..., [-0.1178, -0.0323, -0.3556, ..., -0.1252, 0.1143, 0.0317], [-0.2050, -0.1387, -0.0527, ..., -0.2187, -0.0480, -0.0567], [-0.0937, -0.0959, -0.1443, ..., 0.0970, 0.0279, -0.1720]], device='cuda:0'), grad: tensor([[-9.3132e-10, 0.0000e+00, -4.4703e-08, ..., -1.5832e-08, 9.3132e-10, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 5.5879e-09, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 1.8626e-09, 0.0000e+00], ..., [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, -2.0489e-08, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, -0.0000e+00, -0.0000e+00], [ 9.3132e-10, 0.0000e+00, 1.2107e-08, ..., -5.5879e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 311, bias, value: tensor([ 0.0044, 0.0104, 0.0139, -0.0002, 0.0256, -0.0037, -0.0050, -0.0057, 0.0077, -0.0180], device='cuda:0'), grad: tensor([-8.6613e-08, -2.6077e-08, 1.9558e-08, -1.2107e-08, 7.0781e-08, 3.9116e-08, 2.4214e-08, -8.6613e-08, 9.3132e-10, 5.0291e-08], device='cuda:0') 100 0.0001 changing lr epoch 310, time 246.43, cls_loss 0.0007 cls_loss_mapping 0.0017 cls_loss_causal 0.4976 re_mapping 0.0030 re_causal 0.0098 /// teacc 99.14 lr 0.00010000 Epoch 312, weight, value: tensor([[-0.2237, -0.0569, 0.2153, ..., 0.0444, -0.2171, -0.1647], [-0.1747, -0.2596, -0.0664, ..., -0.1274, -0.2982, -0.0944], [ 0.2894, -0.0748, -0.2931, ..., -0.2238, -0.2347, -0.2186], ..., [-0.1179, -0.0327, -0.3583, ..., -0.1253, 0.1144, 0.0317], [-0.2067, -0.1394, -0.0537, ..., -0.2189, -0.0492, -0.0573], [-0.0960, -0.0973, -0.1453, ..., 0.0980, 0.0292, -0.1720]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 0.0000e+00, 1.8626e-09, 9.3132e-10], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 3.0734e-08, 9.3132e-10], [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 3.7253e-09, 0.0000e+00], ..., [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, -3.6322e-08, 0.0000e+00], [-7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 9.3132e-10, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -6.5193e-09, -9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 312, bias, value: tensor([ 0.0043, 0.0104, 0.0139, -0.0002, 0.0246, -0.0034, -0.0047, -0.0057, 0.0062, -0.0173], device='cuda:0'), grad: tensor([ 1.0245e-08, 2.0210e-07, 1.0058e-07, 2.1420e-08, 9.3132e-09, -2.2631e-07, 2.0303e-07, -2.1979e-07, -9.4064e-08, -2.7940e-09], device='cuda:0') 100 0.0001 changing lr epoch 311, time 246.64, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.5101 re_mapping 0.0033 re_causal 0.0109 /// teacc 99.20 lr 0.00010000 Epoch 313, weight, value: tensor([[-0.2242, -0.0570, 0.2158, ..., 0.0444, -0.2173, -0.1647], [-0.1754, -0.2601, -0.0671, ..., -0.1277, -0.2990, -0.0946], [ 0.2894, -0.0749, -0.2937, ..., -0.2242, -0.2367, -0.2187], ..., [-0.1178, -0.0327, -0.3588, ..., -0.1254, 0.1147, 0.0319], [-0.2076, -0.1406, -0.0522, ..., -0.2193, -0.0498, -0.0574], [-0.0970, -0.0982, -0.1455, ..., 0.0982, 0.0292, -0.1720]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.1130e-07, ..., -1.2107e-07, 5.5879e-09, 3.7253e-09], [ 3.7253e-09, 0.0000e+00, 1.7695e-08, ..., 2.2352e-08, 1.4901e-08, 9.3132e-10], [-2.7940e-08, 0.0000e+00, 2.7008e-08, ..., 9.3132e-09, 1.8626e-09, 0.0000e+00], ..., [ 1.3970e-08, 0.0000e+00, 2.4214e-08, ..., 1.0151e-07, 3.3528e-08, 2.7940e-09], [ 1.8626e-09, 0.0000e+00, 2.1420e-08, ..., 1.2107e-08, 7.4506e-09, 1.8626e-09], [ 9.3132e-10, 0.0000e+00, 1.8720e-07, ..., 4.2841e-08, -3.0734e-08, -6.5193e-09]], device='cuda:0') Epoch 313, bias, value: tensor([ 4.3289e-03, 1.0270e-02, 1.3852e-02, -6.8141e-05, 2.4522e-02, -3.5653e-03, -4.8375e-03, -5.5576e-03, 6.8824e-03, -1.7254e-02], device='cuda:0'), grad: tensor([-1.0822e-06, 2.4680e-07, 4.0978e-08, 1.2014e-07, -3.4552e-07, 2.8498e-07, 1.5367e-07, 2.0489e-07, 1.2014e-07, 2.6543e-07], device='cuda:0') 100 0.0001 changing lr epoch 312, time 246.08, cls_loss 0.0009 cls_loss_mapping 0.0029 cls_loss_causal 0.4798 re_mapping 0.0034 re_causal 0.0102 /// teacc 99.12 lr 0.00010000 Epoch 314, weight, value: tensor([[-0.2233, -0.0570, 0.2186, ..., 0.0447, -0.2176, -0.1650], [-0.1758, -0.2605, -0.0672, ..., -0.1279, -0.2995, -0.0950], [ 0.2892, -0.0749, -0.2973, ..., -0.2276, -0.2376, -0.2189], ..., [-0.1178, -0.0328, -0.3606, ..., -0.1277, 0.1141, 0.0319], [-0.2088, -0.1408, -0.0505, ..., -0.2197, -0.0490, -0.0578], [-0.0986, -0.0988, -0.1462, ..., 0.0979, 0.0288, -0.1721]], device='cuda:0'), grad: tensor([[ 2.2352e-08, 1.8626e-09, -2.3656e-07, ..., -2.1420e-08, 3.3528e-08, 1.8626e-08], [ 1.9558e-08, 1.8626e-09, 2.6077e-08, ..., 7.4506e-09, 3.7253e-09, 3.7253e-09], [ 5.4948e-08, 3.7253e-09, 2.4214e-08, ..., 2.9802e-08, 1.0245e-08, 3.7253e-09], ..., [ 5.8673e-08, 4.6566e-09, 2.6077e-08, ..., 1.3970e-08, 3.7253e-09, 0.0000e+00], [ 1.6764e-08, 9.3132e-10, -2.7940e-09, ..., 7.4506e-09, -5.5879e-09, 6.5193e-09], [-2.1420e-08, 3.7253e-09, 1.5087e-07, ..., 3.3528e-08, -3.0734e-08, 1.8626e-09]], device='cuda:0') Epoch 314, bias, value: tensor([ 0.0066, 0.0085, 0.0135, 0.0002, 0.0258, -0.0038, -0.0065, -0.0041, 0.0056, -0.0178], device='cuda:0'), grad: tensor([-3.9954e-07, 1.7881e-07, 3.4925e-07, -1.0934e-06, -3.1106e-07, 1.7500e-06, -1.2508e-06, 3.3341e-07, -5.1502e-07, 9.3319e-07], device='cuda:0') 100 0.0001 changing lr epoch 313, time 246.25, cls_loss 0.0007 cls_loss_mapping 0.0019 cls_loss_causal 0.4960 re_mapping 0.0032 re_causal 0.0102 /// teacc 99.18 lr 0.00010000 Epoch 315, weight, value: tensor([[-0.2234, -0.0571, 0.2187, ..., 0.0447, -0.2184, -0.1652], [-0.1760, -0.2606, -0.0673, ..., -0.1280, -0.3014, -0.0962], [ 0.2893, -0.0750, -0.2976, ..., -0.2278, -0.2412, -0.2193], ..., [-0.1178, -0.0329, -0.3625, ..., -0.1278, 0.1155, 0.0324], [-0.2094, -0.1409, -0.0497, ..., -0.2200, -0.0494, -0.0581], [-0.0989, -0.0988, -0.1466, ..., 0.0975, 0.0276, -0.1722]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -2.0489e-08, ..., -7.4506e-09, 9.3132e-10, 0.0000e+00], [ 2.3283e-08, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 5.5879e-09, 0.0000e+00], [-5.2154e-08, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 2.7940e-09, 0.0000e+00], ..., [ 8.3819e-09, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, -1.1176e-08, 0.0000e+00], [ 1.6764e-08, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 1.0245e-08, 0.0000e+00, 8.3819e-09, ..., 3.6322e-08, -9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 315, bias, value: tensor([ 0.0065, 0.0103, 0.0134, 0.0002, 0.0264, -0.0038, -0.0064, -0.0051, 0.0060, -0.0196], device='cuda:0'), grad: tensor([-3.7253e-09, 1.1735e-07, -1.6484e-07, 2.4401e-06, -6.6124e-08, -3.1441e-06, 1.4715e-07, -1.8626e-09, 1.0338e-07, 5.6438e-07], device='cuda:0') 100 0.0001 changing lr epoch 314, time 246.15, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4761 re_mapping 0.0035 re_causal 0.0105 /// teacc 99.10 lr 0.00010000 Epoch 316, weight, value: tensor([[-0.2236, -0.0571, 0.2191, ..., 0.0448, -0.2192, -0.1653], [-0.1765, -0.2607, -0.0673, ..., -0.1282, -0.3036, -0.0973], [ 0.2893, -0.0750, -0.2978, ..., -0.2279, -0.2430, -0.2193], ..., [-0.1179, -0.0329, -0.3637, ..., -0.1280, 0.1164, 0.0328], [-0.2099, -0.1409, -0.0501, ..., -0.2202, -0.0497, -0.0585], [-0.0994, -0.0989, -0.1469, ..., 0.0975, 0.0269, -0.1722]], device='cuda:0'), grad: tensor([[ 4.6566e-08, 0.0000e+00, -1.1176e-08, ..., 2.2072e-07, 1.3039e-07, 0.0000e+00], [ 2.1420e-08, 0.0000e+00, 4.6566e-09, ..., 6.1467e-08, 2.9802e-08, 0.0000e+00], [-1.8626e-08, 0.0000e+00, 9.3132e-10, ..., 3.2596e-08, 1.8626e-08, 0.0000e+00], ..., [ 4.0978e-08, 0.0000e+00, 9.3132e-10, ..., 1.9744e-07, 8.5682e-08, 0.0000e+00], [ 1.7695e-08, 0.0000e+00, -8.3819e-09, ..., 4.1910e-08, -4.1910e-08, 0.0000e+00], [-2.5239e-07, 0.0000e+00, 8.3819e-09, ..., -1.4165e-06, -8.2888e-07, 0.0000e+00]], device='cuda:0') Epoch 316, bias, value: tensor([ 0.0066, 0.0103, 0.0134, 0.0009, 0.0265, -0.0042, -0.0066, -0.0050, 0.0060, -0.0199], device='cuda:0'), grad: tensor([ 1.0729e-06, 2.9057e-07, 1.2666e-07, 6.1747e-07, 1.4426e-06, 1.6140e-06, 5.8860e-07, 9.2480e-07, -2.1607e-07, -6.4895e-06], device='cuda:0') 100 0.0001 changing lr epoch 315, time 245.92, cls_loss 0.0006 cls_loss_mapping 0.0015 cls_loss_causal 0.5122 re_mapping 0.0031 re_causal 0.0102 /// teacc 99.15 lr 0.00010000 Epoch 317, weight, value: tensor([[-0.2238, -0.0571, 0.2204, ..., 0.0449, -0.2216, -0.1672], [-0.1769, -0.2607, -0.0678, ..., -0.1286, -0.3052, -0.0975], [ 0.2894, -0.0750, -0.2979, ..., -0.2279, -0.2438, -0.2194], ..., [-0.1179, -0.0329, -0.3655, ..., -0.1282, 0.1168, 0.0328], [-0.2103, -0.1409, -0.0514, ..., -0.2205, -0.0499, -0.0588], [-0.0996, -0.0989, -0.1472, ..., 0.0976, 0.0267, -0.1722]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, -9.3132e-10, ..., 3.7253e-09, 0.0000e+00, 0.0000e+00], [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 2.2352e-08, -1.8626e-09, 0.0000e+00]], device='cuda:0') Epoch 317, bias, value: tensor([ 0.0072, 0.0107, 0.0134, 0.0008, 0.0265, -0.0042, -0.0070, -0.0054, 0.0067, -0.0199], device='cuda:0'), grad: tensor([ 1.9558e-08, -2.5891e-07, 2.6077e-08, 3.9116e-08, -8.6613e-08, -9.7789e-08, 6.0536e-08, 1.6671e-07, 3.1665e-08, 1.0151e-07], device='cuda:0') 100 0.0001 changing lr epoch 316, time 246.46, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4892 re_mapping 0.0034 re_causal 0.0106 /// teacc 99.11 lr 0.00010000 Epoch 318, weight, value: tensor([[-0.2238, -0.0571, 0.2204, ..., 0.0450, -0.2223, -0.1676], [-0.1773, -0.2608, -0.0670, ..., -0.1287, -0.3062, -0.0976], [ 0.2895, -0.0750, -0.2982, ..., -0.2281, -0.2456, -0.2194], ..., [-0.1180, -0.0329, -0.3667, ..., -0.1283, 0.1170, 0.0328], [-0.2105, -0.1409, -0.0526, ..., -0.2211, -0.0501, -0.0589], [-0.1003, -0.0989, -0.1474, ..., 0.0977, 0.0267, -0.1722]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -9.3132e-10, ..., 7.5437e-08, 1.7695e-08, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 2.4214e-08, 4.6566e-08, 0.0000e+00], [-1.8626e-09, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, 1.2107e-08, 0.0000e+00], ..., [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.0245e-08, -1.8068e-07, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, -2.7940e-09, ..., 4.6566e-09, -3.7253e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -1.5395e-06, 9.4064e-08, 0.0000e+00]], device='cuda:0') Epoch 318, bias, value: tensor([ 0.0072, 0.0108, 0.0135, 0.0006, 0.0265, -0.0041, -0.0070, -0.0054, 0.0059, -0.0199], device='cuda:0'), grad: tensor([ 3.1479e-07, 2.2352e-08, 7.8231e-08, 2.5015e-06, 1.7099e-06, 1.0896e-07, -1.3039e-08, -5.4762e-07, -1.3039e-08, -4.1649e-06], device='cuda:0') 100 0.0001 changing lr epoch 317, time 246.08, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4694 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.10 lr 0.00010000 Epoch 319, weight, value: tensor([[-0.2240, -0.0571, 0.2205, ..., 0.0450, -0.2227, -0.1676], [-0.1777, -0.2612, -0.0673, ..., -0.1288, -0.3071, -0.0977], [ 0.2897, -0.0750, -0.2982, ..., -0.2282, -0.2466, -0.2194], ..., [-0.1180, -0.0330, -0.3678, ..., -0.1286, 0.1170, 0.0329], [-0.2110, -0.1410, -0.0514, ..., -0.2214, -0.0516, -0.0589], [-0.1005, -0.0989, -0.1484, ..., 0.0977, 0.0267, -0.1722]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, 5.5879e-09, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, -2.0489e-08, ..., 9.3132e-09, 5.7742e-08, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 0.0000e+00, 2.7940e-09, 0.0000e+00], ..., [ 1.8626e-09, 0.0000e+00, 1.8626e-08, ..., -1.7695e-08, -1.4435e-07, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, -5.1223e-08, ..., 1.8626e-09, -3.5390e-08, 0.0000e+00], [ 2.7940e-09, 9.3132e-10, 2.3283e-08, ..., -5.5879e-09, 6.3330e-08, 0.0000e+00]], device='cuda:0') Epoch 319, bias, value: tensor([ 0.0071, 0.0108, 0.0135, -0.0033, 0.0265, -0.0003, -0.0071, -0.0054, 0.0066, -0.0200], device='cuda:0'), grad: tensor([ 7.9162e-08, -3.9767e-07, 1.1642e-07, 1.1083e-07, 5.5879e-08, 2.2538e-07, 1.0896e-07, -3.8370e-07, -8.5309e-07, 9.2480e-07], device='cuda:0') 100 0.0001 changing lr epoch 318, time 246.61, cls_loss 0.0006 cls_loss_mapping 0.0018 cls_loss_causal 0.4738 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.12 lr 0.00010000 Epoch 320, weight, value: tensor([[-0.2241, -0.0571, 0.2210, ..., 0.0453, -0.2237, -0.1678], [-0.1778, -0.2613, -0.0694, ..., -0.1299, -0.3097, -0.0981], [ 0.2898, -0.0750, -0.2985, ..., -0.2284, -0.2498, -0.2195], ..., [-0.1180, -0.0330, -0.3693, ..., -0.1286, 0.1174, 0.0333], [-0.2123, -0.1410, -0.0511, ..., -0.2215, -0.0522, -0.0590], [-0.1006, -0.0989, -0.1489, ..., 0.0977, 0.0265, -0.1723]], device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, -2.7940e-09, ..., -1.8626e-09, 9.3132e-10, 0.0000e+00], [ 5.2154e-08, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 4.6566e-09, 0.0000e+00], [-4.5262e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.6857e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -7.4506e-09, 0.0000e+00], [ 1.2107e-08, 0.0000e+00, 5.5879e-09, ..., 0.0000e+00, -6.5193e-09, 2.7940e-09], [ 1.2666e-07, 0.0000e+00, 1.8626e-09, ..., 2.3283e-08, 5.5879e-09, 0.0000e+00]], device='cuda:0') Epoch 320, bias, value: tensor([ 0.0074, 0.0107, 0.0135, -0.0033, 0.0265, -0.0005, -0.0062, -0.0054, 0.0072, -0.0203], device='cuda:0'), grad: tensor([ 1.0245e-08, 9.6858e-08, -9.3225e-07, 1.8533e-07, -2.9802e-08, 9.0338e-08, 8.6613e-08, 4.0140e-07, -1.7043e-07, 2.6822e-07], device='cuda:0') 100 0.0001 changing lr epoch 319, time 246.46, cls_loss 0.0004 cls_loss_mapping 0.0009 cls_loss_causal 0.4717 re_mapping 0.0033 re_causal 0.0109 /// teacc 99.11 lr 0.00010000 Epoch 321, weight, value: tensor([[-0.2241, -0.0571, 0.2210, ..., 0.0453, -0.2244, -0.1678], [-0.1781, -0.2613, -0.0693, ..., -0.1299, -0.3106, -0.0981], [ 0.2898, -0.0751, -0.2985, ..., -0.2285, -0.2507, -0.2195], ..., [-0.1181, -0.0330, -0.3707, ..., -0.1288, 0.1177, 0.0335], [-0.2125, -0.1410, -0.0513, ..., -0.2216, -0.0520, -0.0591], [-0.1009, -0.0989, -0.1490, ..., 0.0977, 0.0264, -0.1723]], device='cuda:0'), grad: tensor([[ 3.7253e-09, 0.0000e+00, -3.5390e-08, ..., -8.3819e-09, 9.3132e-10, 0.0000e+00], [ 7.8231e-08, 0.0000e+00, 0.0000e+00, ..., 1.9558e-08, 1.5832e-08, 0.0000e+00], [-1.6484e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 5.3085e-08, 0.0000e+00, 1.8626e-09, ..., 1.5274e-07, 1.1921e-07, 0.0000e+00], [ 2.0489e-08, 0.0000e+00, 7.4506e-09, ..., 3.7253e-09, -0.0000e+00, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 1.5832e-08, ..., -1.8533e-07, -1.4808e-07, 0.0000e+00]], device='cuda:0') Epoch 321, bias, value: tensor([ 0.0072, 0.0109, 0.0135, -0.0034, 0.0266, -0.0005, -0.0061, -0.0054, 0.0072, -0.0203], device='cuda:0'), grad: tensor([-6.6124e-08, 1.9185e-07, -3.1758e-07, -5.6811e-08, 9.5926e-08, 3.7253e-08, 4.4703e-08, 9.3784e-07, 6.5193e-08, -9.1270e-07], device='cuda:0') 100 0.0001 changing lr epoch 320, time 246.30, cls_loss 0.0006 cls_loss_mapping 0.0014 cls_loss_causal 0.4567 re_mapping 0.0033 re_causal 0.0103 /// teacc 99.03 lr 0.00010000 Epoch 322, weight, value: tensor([[-0.2243, -0.0571, 0.2212, ..., 0.0452, -0.2250, -0.1679], [-0.1788, -0.2614, -0.0693, ..., -0.1302, -0.3120, -0.0982], [ 0.2900, -0.0752, -0.2987, ..., -0.2286, -0.2520, -0.2195], ..., [-0.1181, -0.0331, -0.3716, ..., -0.1291, 0.1177, 0.0335], [-0.2131, -0.1411, -0.0516, ..., -0.2220, -0.0522, -0.0592], [-0.1012, -0.0989, -0.1492, ..., 0.0974, 0.0261, -0.1723]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, 1.8626e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, 1.9558e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 1.8626e-09, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.0978e-08, -4.6566e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, 2.7940e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.2331e-06, -4.6659e-07, 0.0000e+00]], device='cuda:0') Epoch 322, bias, value: tensor([ 0.0070, 0.0113, 0.0135, -0.0034, 0.0269, -0.0005, -0.0060, -0.0059, 0.0071, -0.0205], device='cuda:0'), grad: tensor([ 1.5832e-08, 7.9162e-08, 2.5146e-08, 4.5635e-08, 2.7418e-06, -1.7602e-07, 6.8918e-08, -9.3132e-09, 2.7008e-08, -2.8238e-06], device='cuda:0') 100 0.0001 changing lr epoch 321, time 246.05, cls_loss 0.0004 cls_loss_mapping 0.0012 cls_loss_causal 0.4564 re_mapping 0.0033 re_causal 0.0104 /// teacc 99.10 lr 0.00010000 Epoch 323, weight, value: tensor([[-0.2243, -0.0572, 0.2211, ..., 0.0440, -0.2260, -0.1679], [-0.1790, -0.2615, -0.0693, ..., -0.1302, -0.3127, -0.0983], [ 0.2900, -0.0752, -0.2988, ..., -0.2288, -0.2528, -0.2195], ..., [-0.1182, -0.0331, -0.3725, ..., -0.1292, 0.1178, 0.0336], [-0.2132, -0.1412, -0.0525, ..., -0.2231, -0.0523, -0.0593], [-0.1015, -0.0991, -0.1487, ..., 0.0977, 0.0257, -0.1723]], device='cuda:0'), grad: tensor([[ 1.6764e-08, 0.0000e+00, 1.2061e-06, ..., 3.3528e-08, 2.9057e-07, 0.0000e+00], [ 8.8476e-08, 0.0000e+00, 2.7008e-08, ..., 1.8626e-09, 1.5832e-08, 0.0000e+00], [-3.2317e-06, 0.0000e+00, 1.0245e-08, ..., 9.3132e-10, 6.5193e-09, 0.0000e+00], ..., [ 3.0920e-06, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, -2.6077e-08, 0.0000e+00], [ 1.8626e-08, 0.0000e+00, 9.4995e-08, ..., 8.3819e-09, 2.1420e-08, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 3.7253e-08, ..., 1.3970e-08, 2.0489e-08, 0.0000e+00]], device='cuda:0') Epoch 323, bias, value: tensor([ 0.0059, 0.0114, 0.0135, -0.0034, 0.0271, -0.0005, -0.0059, -0.0059, 0.0069, -0.0203], device='cuda:0'), grad: tensor([ 2.2296e-06, 3.2783e-07, -9.9391e-06, 4.0047e-08, -6.2399e-08, 3.9600e-06, -6.4559e-06, 9.4771e-06, 2.4214e-07, 1.4622e-07], device='cuda:0') 100 0.0001 changing lr epoch 322, time 245.84, cls_loss 0.0007 cls_loss_mapping 0.0023 cls_loss_causal 0.5112 re_mapping 0.0030 re_causal 0.0098 /// teacc 99.20 lr 0.00010000 Epoch 324, weight, value: tensor([[-0.2243, -0.0572, 0.2177, ..., 0.0442, -0.2281, -0.1712], [-0.1793, -0.2617, -0.0696, ..., -0.1304, -0.3152, -0.0984], [ 0.2901, -0.0752, -0.2991, ..., -0.2291, -0.2576, -0.2197], ..., [-0.1182, -0.0332, -0.3759, ..., -0.1292, 0.1184, 0.0341], [-0.2135, -0.1416, -0.0530, ..., -0.2237, -0.0531, -0.0596], [-0.1020, -0.0989, -0.1492, ..., 0.0976, 0.0254, -0.1724]], device='cuda:0'), grad: tensor([[-1.9092e-08, 4.6566e-10, -2.4913e-07, ..., -1.4808e-07, 6.0536e-09, 4.6566e-10], [ 1.0245e-08, 9.3132e-10, 1.4901e-08, ..., 8.8476e-09, 3.8184e-08, 0.0000e+00], [-6.1002e-08, 1.3970e-09, 1.7695e-08, ..., 1.0245e-08, 5.4017e-08, 0.0000e+00], ..., [ 1.1642e-08, 2.3283e-09, 1.8626e-08, ..., 5.1223e-09, -2.3935e-07, 0.0000e+00], [ 6.5193e-09, 9.3132e-10, 1.7229e-08, ..., 3.3528e-08, 4.9360e-08, 0.0000e+00], [ 3.2131e-08, 1.3970e-09, 2.8405e-08, ..., -3.2596e-09, 1.6298e-08, 0.0000e+00]], device='cuda:0') Epoch 324, bias, value: tensor([ 0.0025, 0.0109, 0.0135, -0.0035, 0.0273, -0.0003, -0.0026, -0.0055, 0.0066, -0.0205], device='cuda:0'), grad: tensor([-7.3714e-07, 1.9884e-07, 1.5367e-07, 3.0687e-07, 3.4925e-08, 2.8964e-07, 8.3819e-08, -7.9069e-07, 3.0082e-07, 1.7416e-07], device='cuda:0') 100 0.0001 changing lr epoch 323, time 246.09, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4898 re_mapping 0.0030 re_causal 0.0105 /// teacc 99.15 lr 0.00010000 Epoch 325, weight, value: tensor([[-0.2244, -0.0572, 0.2178, ..., 0.0443, -0.2283, -0.1712], [-0.1796, -0.2620, -0.0699, ..., -0.1306, -0.3171, -0.0985], [ 0.2902, -0.0754, -0.2991, ..., -0.2292, -0.2583, -0.2197], ..., [-0.1183, -0.0334, -0.3787, ..., -0.1293, 0.1186, 0.0341], [-0.2138, -0.1418, -0.0530, ..., -0.2241, -0.0532, -0.0596], [-0.1022, -0.0990, -0.1497, ..., 0.0975, 0.0253, -0.1724]], device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, 1.3970e-09, ..., -9.3132e-10, 2.7940e-09, 1.8626e-09], [ 6.5193e-09, 0.0000e+00, 1.3970e-09, ..., 2.3283e-09, 1.3970e-09, 4.6566e-10], [-2.1420e-08, 0.0000e+00, -0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 5.5879e-09, 4.6566e-10, 0.0000e+00, ..., 6.9849e-09, 9.3132e-10, 0.0000e+00], [ 1.3970e-09, 0.0000e+00, 1.1176e-08, ..., 4.6566e-09, 3.7253e-09, 1.8626e-09], [ 4.6566e-10, 0.0000e+00, 8.3819e-09, ..., -6.5891e-07, -2.2119e-07, 0.0000e+00]], device='cuda:0') Epoch 325, bias, value: tensor([ 0.0026, 0.0109, 0.0135, -0.0035, 0.0274, -0.0003, -0.0026, -0.0055, 0.0067, -0.0206], device='cuda:0'), grad: tensor([ 1.5367e-08, -1.0710e-08, -3.4459e-08, -1.1642e-08, 1.4054e-06, 5.0757e-08, -6.4261e-08, 4.5635e-08, 3.1199e-08, -1.4119e-06], device='cuda:0') 100 0.0001 changing lr epoch 324, time 246.32, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4760 re_mapping 0.0030 re_causal 0.0103 /// teacc 99.11 lr 0.00010000 Epoch 326, weight, value: tensor([[-0.2246, -0.0573, 0.2178, ..., 0.0444, -0.2290, -0.1712], [-0.1799, -0.2627, -0.0699, ..., -0.1307, -0.3179, -0.0986], [ 0.2902, -0.0758, -0.2999, ..., -0.2293, -0.2588, -0.2202], ..., [-0.1183, -0.0338, -0.3802, ..., -0.1293, 0.1187, 0.0341], [-0.2141, -0.1427, -0.0532, ..., -0.2243, -0.0534, -0.0600], [-0.1027, -0.0994, -0.1501, ..., 0.0973, 0.0251, -0.1724]], device='cuda:0'), grad: tensor([[ 4.5635e-08, 3.5390e-08, 1.8626e-09, ..., 2.1094e-07, 9.3132e-10, 4.6566e-10], [ 2.7940e-08, 2.1420e-08, 9.3132e-10, ..., 2.7008e-08, 1.8626e-09, 0.0000e+00], [ 2.4680e-08, 2.5611e-08, 0.0000e+00, ..., 3.2596e-09, 1.3970e-09, 0.0000e+00], ..., [ 4.3772e-08, 2.7474e-08, 0.0000e+00, ..., 4.2375e-07, 5.1223e-09, 0.0000e+00], [ 9.6858e-08, 7.5437e-08, -1.3970e-09, ..., 4.6566e-09, 1.3970e-09, 4.6566e-10], [ 6.2659e-06, 4.8950e-06, 1.3970e-09, ..., -8.6520e-07, 1.2992e-07, 0.0000e+00]], device='cuda:0') Epoch 326, bias, value: tensor([ 0.0026, 0.0109, 0.0134, -0.0034, 0.0280, -0.0004, -0.0028, -0.0055, 0.0069, -0.0209], device='cuda:0'), grad: tensor([ 1.5572e-06, 3.2783e-07, 2.1141e-07, -5.7369e-05, 2.1644e-06, 1.9610e-05, 1.4296e-07, 2.8163e-06, 3.8743e-07, 3.0175e-05], device='cuda:0') 100 0.0001 changing lr epoch 325, time 246.38, cls_loss 0.0005 cls_loss_mapping 0.0015 cls_loss_causal 0.4840 re_mapping 0.0030 re_causal 0.0098 /// teacc 99.13 lr 0.00010000 Epoch 327, weight, value: tensor([[-0.2247, -0.0575, 0.2178, ..., 0.0444, -0.2293, -0.1712], [-0.1829, -0.2640, -0.0700, ..., -0.1308, -0.3188, -0.0987], [ 0.2909, -0.0760, -0.2997, ..., -0.2294, -0.2597, -0.2202], ..., [-0.1184, -0.0341, -0.3811, ..., -0.1294, 0.1189, 0.0341], [-0.2176, -0.1439, -0.0535, ..., -0.2245, -0.0537, -0.0601], [-0.1053, -0.1017, -0.1503, ..., 0.0972, 0.0249, -0.1724]], device='cuda:0'), grad: tensor([[ 1.1642e-08, 0.0000e+00, -8.7079e-08, ..., -3.6787e-08, 7.4506e-09, 0.0000e+00], [ 1.8161e-08, 0.0000e+00, 6.0536e-09, ..., 1.0245e-08, 1.4435e-08, 0.0000e+00], [-1.0338e-07, -4.6566e-10, 1.5367e-08, ..., 2.3283e-09, 1.4901e-08, 0.0000e+00], ..., [ 4.6566e-09, 0.0000e+00, 1.3970e-09, ..., 2.3283e-08, -3.5856e-08, 0.0000e+00], [ 1.6298e-08, 0.0000e+00, 7.9162e-09, ..., 4.4238e-08, 1.8626e-09, 0.0000e+00], [ 1.3970e-09, 0.0000e+00, 7.8231e-08, ..., -1.6578e-07, 8.3819e-09, 0.0000e+00]], device='cuda:0') Epoch 327, bias, value: tensor([ 0.0026, 0.0107, 0.0140, -0.0031, 0.0281, -0.0005, -0.0028, -0.0056, 0.0069, -0.0212], device='cuda:0'), grad: tensor([-8.1025e-08, 1.5739e-07, -1.3178e-07, 1.3132e-07, 5.3644e-07, 4.2375e-08, -8.0094e-08, -3.6787e-08, 1.0524e-07, -6.5146e-07], device='cuda:0') 100 0.0001 changing lr epoch 326, time 246.25, cls_loss 0.0006 cls_loss_mapping 0.0010 cls_loss_causal 0.4721 re_mapping 0.0030 re_causal 0.0096 /// teacc 99.14 lr 0.00010000 Epoch 328, weight, value: tensor([[-0.2250, -0.0576, 0.2180, ..., 0.0446, -0.2299, -0.1713], [-0.1835, -0.2644, -0.0701, ..., -0.1311, -0.3204, -0.0988], [ 0.2913, -0.0760, -0.3001, ..., -0.2298, -0.2607, -0.2203], ..., [-0.1184, -0.0342, -0.3849, ..., -0.1296, 0.1191, 0.0341], [-0.2203, -0.1442, -0.0539, ..., -0.2253, -0.0542, -0.0602], [-0.1056, -0.1017, -0.1509, ..., 0.0971, 0.0247, -0.1724]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 5.5879e-09, ..., 6.5193e-09, 5.1223e-09, 9.3132e-10], [ 9.3132e-10, 4.6566e-10, -3.9581e-08, ..., -1.3504e-08, 9.0338e-08, 4.6566e-10], [ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, 3.2596e-09, 0.0000e+00], ..., [ 2.3283e-09, 9.3132e-10, 2.2352e-08, ..., 1.3039e-08, -3.4040e-07, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.8626e-08, 5.1223e-09, 4.6566e-10], [ 9.3132e-10, 9.3132e-10, 6.5193e-09, ..., -4.1910e-08, 1.7695e-08, 0.0000e+00]], device='cuda:0') Epoch 328, bias, value: tensor([ 0.0028, 0.0108, 0.0142, -0.0032, 0.0283, -0.0004, -0.0029, -0.0057, 0.0066, -0.0213], device='cuda:0'), grad: tensor([ 8.3819e-08, -3.8696e-07, 3.0268e-08, 7.3714e-07, 6.7055e-08, 7.6834e-08, 1.8626e-09, -6.9756e-07, 6.1002e-08, 4.9360e-08], device='cuda:0') 100 0.0001 changing lr epoch 327, time 246.47, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4553 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.10 lr 0.00010000 Epoch 329, weight, value: tensor([[-0.2251, -0.0578, 0.2180, ..., 0.0447, -0.2305, -0.1713], [-0.1836, -0.2651, -0.0703, ..., -0.1313, -0.3218, -0.1001], [ 0.2913, -0.0761, -0.3003, ..., -0.2299, -0.2614, -0.2204], ..., [-0.1184, -0.0345, -0.3860, ..., -0.1302, 0.1190, 0.0340], [-0.2205, -0.1448, -0.0521, ..., -0.2256, -0.0539, -0.0606], [-0.1064, -0.1044, -0.1525, ..., 0.0970, 0.0248, -0.1724]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., 4.6566e-10, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 4.6566e-10, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.1223e-09, -3.2596e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -8.3819e-09, 2.7940e-09, 0.0000e+00]], device='cuda:0') Epoch 329, bias, value: tensor([ 0.0028, 0.0108, 0.0142, -0.0030, 0.0285, -0.0002, -0.0030, -0.0058, 0.0075, -0.0220], device='cuda:0'), grad: tensor([ 2.3283e-09, -9.3132e-10, 3.2596e-09, -1.2573e-08, 8.8476e-09, 9.3132e-09, 4.1910e-09, 6.9849e-09, 2.3283e-09, -6.5193e-09], device='cuda:0') 100 0.0001 changing lr epoch 328, time 246.20, cls_loss 0.0006 cls_loss_mapping 0.0025 cls_loss_causal 0.4736 re_mapping 0.0030 re_causal 0.0097 /// teacc 99.06 lr 0.00010000 Epoch 330, weight, value: tensor([[-0.2252, -0.0579, 0.2181, ..., 0.0448, -0.2312, -0.1713], [-0.1838, -0.2656, -0.0733, ..., -0.1315, -0.3231, -0.1003], [ 0.2913, -0.0764, -0.3009, ..., -0.2301, -0.2629, -0.2205], ..., [-0.1184, -0.0348, -0.3876, ..., -0.1304, 0.1212, 0.0340], [-0.2207, -0.1454, -0.0527, ..., -0.2259, -0.0546, -0.0609], [-0.1065, -0.1044, -0.1530, ..., 0.0969, 0.0229, -0.1724]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-08, ..., -8.3819e-09, 3.7253e-09, 9.3132e-10], [ 9.3132e-10, 0.0000e+00, -2.8871e-08, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [-2.7940e-09, -9.3132e-10, 4.6566e-09, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 9.3132e-10, 1.7695e-08, ..., 1.8626e-09, -9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.7940e-09, -1.8626e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., -5.5879e-09, -0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 330, bias, value: tensor([ 0.0028, 0.0103, 0.0140, -0.0029, 0.0286, -0.0002, -0.0021, -0.0053, 0.0076, -0.0233], device='cuda:0'), grad: tensor([ 0.0000e+00, -2.4214e-07, 3.4459e-08, 2.0489e-08, 2.8871e-08, 6.4261e-08, -1.0245e-08, 1.4529e-07, -5.8673e-08, 2.1420e-08], device='cuda:0') 100 0.0001 changing lr epoch 329, time 246.53, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4883 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.15 lr 0.00010000 Epoch 331, weight, value: tensor([[-0.2254, -0.0580, 0.2184, ..., 0.0451, -0.2315, -0.1713], [-0.1840, -0.2661, -0.0735, ..., -0.1322, -0.3239, -0.1003], [ 0.2914, -0.0764, -0.3012, ..., -0.2304, -0.2634, -0.2205], ..., [-0.1183, -0.0350, -0.3888, ..., -0.1308, 0.1222, 0.0340], [-0.2210, -0.1454, -0.0531, ..., -0.2261, -0.0548, -0.0611], [-0.1068, -0.1044, -0.1534, ..., 0.0971, 0.0221, -0.1724]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.2841e-08, ..., -1.3039e-08, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, -0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [-1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, -9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.2596e-08, ..., 5.5879e-09, 9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 331, bias, value: tensor([ 0.0030, 0.0089, 0.0139, -0.0029, 0.0284, -0.0002, -0.0023, -0.0039, 0.0073, -0.0233], device='cuda:0'), grad: tensor([-5.7742e-08, -2.8592e-07, 7.4506e-09, -3.4273e-07, 9.3132e-09, 7.8231e-08, 3.3528e-08, 2.5891e-07, 1.7881e-07, 1.2387e-07], device='cuda:0') 100 0.0001 changing lr epoch 330, time 246.57, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4691 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.13 lr 0.00010000 Epoch 332, weight, value: tensor([[-0.2255, -0.0580, 0.2184, ..., 0.0451, -0.2319, -0.1714], [-0.1841, -0.2670, -0.0735, ..., -0.1322, -0.3253, -0.1003], [ 0.2915, -0.0765, -0.3014, ..., -0.2305, -0.2654, -0.2206], ..., [-0.1184, -0.0353, -0.3894, ..., -0.1313, 0.1223, 0.0340], [-0.2213, -0.1458, -0.0534, ..., -0.2279, -0.0557, -0.0613], [-0.1069, -0.1046, -0.1538, ..., 0.0973, 0.0222, -0.1724]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.6729e-07, ..., -3.8184e-08, 9.3132e-10, 9.3132e-10], [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 1.8906e-07, 0.0000e+00], [-6.5193e-09, -0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], ..., [ 2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, -1.9278e-07, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, -1.8626e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 8.3819e-09, ..., 1.8626e-09, 5.5879e-09, 0.0000e+00]], device='cuda:0') Epoch 332, bias, value: tensor([ 3.0687e-03, 9.0623e-03, 1.3859e-02, -3.0083e-03, 2.8432e-02, -3.8579e-05, -2.2786e-03, -4.1072e-03, 7.0482e-03, -2.3238e-02], device='cuda:0'), grad: tensor([-3.4925e-07, 1.2955e-06, -9.3132e-10, 5.3085e-08, -5.5879e-09, -2.0862e-07, 4.3586e-07, -1.2955e-06, 9.3132e-10, 7.7300e-08], device='cuda:0') 100 0.0001 changing lr epoch 331, time 246.09, cls_loss 0.0006 cls_loss_mapping 0.0021 cls_loss_causal 0.4880 re_mapping 0.0029 re_causal 0.0096 /// teacc 99.06 lr 0.00010000 Epoch 333, weight, value: tensor([[-0.2257, -0.0580, 0.2185, ..., 0.0424, -0.2331, -0.1714], [-0.1842, -0.2685, -0.0735, ..., -0.1335, -0.3298, -0.1005], [ 0.2916, -0.0768, -0.3017, ..., -0.2307, -0.2667, -0.2210], ..., [-0.1184, -0.0360, -0.3902, ..., -0.1317, 0.1229, 0.0342], [-0.2218, -0.1473, -0.0538, ..., -0.2281, -0.0559, -0.0614], [-0.1070, -0.1047, -0.1547, ..., 0.0995, 0.0222, -0.1724]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -9.3132e-10, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., -1.8626e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-10, 1.8626e-09, 0.0000e+00, ..., 1.8626e-09, -9.3132e-10, 0.0000e+00], [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, -1.8626e-09, 0.0000e+00], [ 0.0000e+00, 1.8626e-09, 2.7940e-09, ..., -0.0000e+00, 2.7940e-09, 0.0000e+00]], device='cuda:0') Epoch 333, bias, value: tensor([ 1.7280e-03, 8.3839e-03, 1.3836e-02, -2.9308e-03, 2.7949e-02, -9.5762e-05, -1.9174e-03, -3.5790e-03, 7.2022e-03, -2.1743e-02], device='cuda:0'), grad: tensor([ 9.3132e-10, -3.9116e-08, 5.5879e-09, -7.9162e-08, 7.4506e-09, 3.7253e-09, 2.3283e-08, 3.4459e-08, -8.3819e-09, 4.6566e-08], device='cuda:0') 100 0.0001 changing lr epoch 332, time 246.19, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4518 re_mapping 0.0033 re_causal 0.0100 /// teacc 99.16 lr 0.00010000 Epoch 334, weight, value: tensor([[-0.2259, -0.0582, 0.2186, ..., 0.0424, -0.2353, -0.1715], [-0.1843, -0.2695, -0.0744, ..., -0.1333, -0.3304, -0.1003], [ 0.2921, -0.0768, -0.3019, ..., -0.2307, -0.2672, -0.2197], ..., [-0.1186, -0.0380, -0.3907, ..., -0.1321, 0.1228, 0.0328], [-0.2219, -0.1449, -0.0553, ..., -0.2305, -0.0587, -0.0649], [-0.1080, -0.1048, -0.1551, ..., 0.0987, 0.0225, -0.1724]], device='cuda:0'), grad: tensor([[ 3.2596e-08, 0.0000e+00, -1.0245e-07, ..., -4.8429e-08, 2.7940e-09, 1.8626e-09], [ 2.7940e-09, 9.3132e-10, 3.7253e-09, ..., 9.3132e-10, 1.8626e-09, 9.3132e-10], [-2.7381e-07, -4.8429e-08, 1.0245e-08, ..., 6.5193e-09, 9.3132e-10, 9.3132e-10], ..., [ 2.0117e-07, 5.8673e-08, 9.3132e-10, ..., 9.3132e-10, 3.7253e-09, 5.5879e-09], [ 1.1176e-08, 1.8626e-09, 1.0245e-08, ..., 1.8626e-09, 1.8626e-09, 1.8626e-09], [ 1.8626e-09, 0.0000e+00, 6.5193e-09, ..., 4.6566e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 334, bias, value: tensor([ 0.0018, 0.0063, 0.0141, -0.0027, 0.0291, -0.0003, -0.0015, -0.0017, 0.0065, -0.0228], device='cuda:0'), grad: tensor([ 7.1712e-08, -3.9265e-06, -4.3772e-07, -1.7043e-07, 6.1467e-08, 2.3954e-06, 7.1712e-07, 1.1846e-06, 4.1910e-08, 4.4703e-08], device='cuda:0') 100 0.0001 changing lr epoch 333, time 246.41, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4719 re_mapping 0.0030 re_causal 0.0097 /// teacc 99.12 lr 0.00010000 Epoch 335, weight, value: tensor([[-0.2263, -0.0586, 0.2187, ..., 0.0424, -0.2359, -0.1715], [-0.1847, -0.2715, -0.0744, ..., -0.1333, -0.3315, -0.1014], [ 0.2928, -0.0758, -0.3016, ..., -0.2308, -0.2678, -0.2183], ..., [-0.1188, -0.0422, -0.3915, ..., -0.1329, 0.1225, 0.0315], [-0.2222, -0.1447, -0.0559, ..., -0.2331, -0.0595, -0.0651], [-0.1081, -0.1048, -0.1553, ..., 0.0989, 0.0232, -0.1724]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.1514e-07, ..., -4.6566e-08, 9.3132e-10, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 2.7940e-09, 9.3132e-10, 0.0000e+00], [ 9.3132e-10, 9.3132e-10, 9.3132e-09, ..., 4.6566e-09, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 7.4506e-09, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., 2.7940e-09, -0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.8429e-08, ..., 3.7253e-08, 4.6566e-09, 0.0000e+00]], device='cuda:0') Epoch 335, bias, value: tensor([ 0.0018, 0.0065, 0.0144, -0.0025, 0.0290, -0.0005, -0.0017, -0.0020, 0.0066, -0.0222], device='cuda:0'), grad: tensor([-5.8673e-07, 1.1176e-08, 2.8871e-08, -1.8626e-08, -6.5193e-08, 2.6170e-07, 1.2200e-07, 3.1665e-08, 2.7008e-08, 1.9837e-07], device='cuda:0') 100 0.0001 changing lr epoch 334, time 246.47, cls_loss 0.0004 cls_loss_mapping 0.0018 cls_loss_causal 0.4679 re_mapping 0.0034 re_causal 0.0105 /// teacc 99.11 lr 0.00010000 Epoch 336, weight, value: tensor([[-0.2264, -0.0586, 0.2188, ..., 0.0424, -0.2360, -0.1715], [-0.1849, -0.2721, -0.0745, ..., -0.1336, -0.3322, -0.1017], [ 0.2929, -0.0759, -0.3020, ..., -0.2309, -0.2687, -0.2184], ..., [-0.1189, -0.0425, -0.3920, ..., -0.1331, 0.1226, 0.0315], [-0.2224, -0.1449, -0.0568, ..., -0.2334, -0.0597, -0.0654], [-0.1082, -0.1048, -0.1559, ..., 0.0988, 0.0232, -0.1725]], device='cuda:0'), grad: tensor([[ 9.3132e-09, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, 9.3132e-10, 0.0000e+00], [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 1.1176e-08, 0.0000e+00], [-1.9558e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, -4.6566e-09, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, 4.6566e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.5146e-08, -1.1176e-08, 0.0000e+00]], device='cuda:0') Epoch 336, bias, value: tensor([ 0.0019, 0.0066, 0.0144, -0.0024, 0.0293, -0.0006, -0.0017, -0.0021, 0.0065, -0.0224], device='cuda:0'), grad: tensor([ 5.1223e-08, -3.1013e-06, -3.6322e-08, 7.3574e-08, -1.6857e-07, -6.7987e-08, 2.2259e-07, 2.9542e-06, 6.7987e-08, -5.5879e-09], device='cuda:0') 100 0.0001 changing lr epoch 335, time 246.57, cls_loss 0.0005 cls_loss_mapping 0.0017 cls_loss_causal 0.4808 re_mapping 0.0029 re_causal 0.0097 /// teacc 98.96 lr 0.00010000 Epoch 337, weight, value: tensor([[-0.2265, -0.0587, 0.2189, ..., 0.0424, -0.2371, -0.1715], [-0.1852, -0.2727, -0.0746, ..., -0.1338, -0.3313, -0.1017], [ 0.2930, -0.0761, -0.3027, ..., -0.2311, -0.2697, -0.2185], ..., [-0.1189, -0.0427, -0.3932, ..., -0.1338, 0.1220, 0.0314], [-0.2226, -0.1453, -0.0578, ..., -0.2336, -0.0599, -0.0653], [-0.1084, -0.1049, -0.1564, ..., 0.0988, 0.0235, -0.1725]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 9.3132e-10, -1.4901e-08, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 2.7940e-09, 3.7253e-09, ..., 4.6566e-09, -4.6566e-09, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-10, -0.0000e+00, -1.8626e-09], [ 9.3132e-10, 9.3132e-10, 1.8626e-09, ..., -1.3970e-08, -2.7940e-09, 0.0000e+00]], device='cuda:0') Epoch 337, bias, value: tensor([ 0.0019, 0.0067, 0.0144, -0.0023, 0.0295, -0.0007, -0.0016, -0.0023, 0.0066, -0.0224], device='cuda:0'), grad: tensor([ 5.5879e-09, -5.4017e-08, 1.0245e-08, -1.3327e-06, 3.7253e-08, 1.3094e-06, 2.4214e-08, 5.3085e-08, -1.3039e-08, -2.8871e-08], device='cuda:0') 100 0.0001 changing lr epoch 336, time 246.71, cls_loss 0.0011 cls_loss_mapping 0.0018 cls_loss_causal 0.4586 re_mapping 0.0031 re_causal 0.0094 /// teacc 99.06 lr 0.00010000 Epoch 338, weight, value: tensor([[-0.2266, -0.0589, 0.2196, ..., 0.0408, -0.2396, -0.1716], [-0.1861, -0.2736, -0.0748, ..., -0.1347, -0.3339, -0.1025], [ 0.2934, -0.0761, -0.3033, ..., -0.2317, -0.2738, -0.2199], ..., [-0.1192, -0.0429, -0.3944, ..., -0.1354, 0.1224, 0.0354], [-0.2228, -0.1456, -0.0558, ..., -0.2378, -0.0600, -0.0662], [-0.1086, -0.1048, -0.1589, ..., 0.1035, 0.0280, -0.1729]], device='cuda:0'), grad: tensor([[ 7.4506e-09, 0.0000e+00, -3.3993e-07, ..., -1.0524e-07, 8.3819e-09, 3.7253e-09], [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, 6.5193e-09, 0.0000e+00], [-2.2352e-08, -0.0000e+00, 7.4506e-09, ..., 3.7253e-09, 7.4506e-09, 2.7940e-09], ..., [ 1.1176e-08, 0.0000e+00, 5.5879e-09, ..., 2.7940e-09, -1.0245e-08, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 2.0489e-08, ..., 6.5193e-09, 5.5879e-09, 2.7940e-09], [ 0.0000e+00, 0.0000e+00, 2.8964e-07, ..., 8.9407e-08, 1.8626e-09, 0.0000e+00]], device='cuda:0') Epoch 338, bias, value: tensor([ 0.0010, 0.0067, 0.0144, -0.0038, 0.0250, -0.0030, -0.0004, -0.0024, 0.0102, -0.0173], device='cuda:0'), grad: tensor([-4.8708e-07, 2.3283e-08, -5.5879e-09, 2.4214e-08, -1.3039e-08, -2.6450e-07, 2.4214e-07, -4.6566e-09, 4.3772e-08, 4.4983e-07], device='cuda:0') 100 0.0001 changing lr epoch 337, time 246.45, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4826 re_mapping 0.0032 re_causal 0.0103 /// teacc 99.03 lr 0.00010000 Epoch 339, weight, value: tensor([[-0.2267, -0.0592, 0.2193, ..., 0.0406, -0.2429, -0.1727], [-0.1863, -0.2746, -0.0751, ..., -0.1352, -0.3349, -0.1038], [ 0.2936, -0.0760, -0.3037, ..., -0.2320, -0.2745, -0.2201], ..., [-0.1193, -0.0431, -0.3952, ..., -0.1353, 0.1249, 0.0357], [-0.2230, -0.1460, -0.0557, ..., -0.2380, -0.0598, -0.0667], [-0.1095, -0.1050, -0.1595, ..., 0.1036, 0.0279, -0.1732]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -7.1339e-07, ..., -2.8498e-07, 0.0000e+00, 9.3132e-10], [ 1.8626e-09, 9.3132e-10, 1.7695e-08, ..., 9.3132e-09, 1.2107e-08, 2.7940e-09], [-4.6566e-09, 9.3132e-10, 7.4506e-09, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], ..., [ 2.7940e-09, 9.3132e-10, 2.7940e-08, ..., 9.3132e-09, -1.2107e-08, -2.7940e-09], [ 9.3132e-10, 0.0000e+00, 8.3819e-09, ..., 2.7940e-09, 1.8626e-09, 1.8626e-09], [ 1.8626e-09, 0.0000e+00, 4.3586e-07, ..., 1.7043e-07, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 339, bias, value: tensor([ 0.0005, 0.0067, 0.0144, -0.0032, 0.0249, -0.0030, -0.0009, -0.0022, 0.0107, -0.0172], device='cuda:0'), grad: tensor([-1.3588e-06, 6.4261e-08, 2.7940e-09, -3.4459e-08, 2.6077e-08, -1.2107e-07, 5.2154e-07, 5.0291e-08, 2.5146e-08, 8.3353e-07], device='cuda:0') 100 0.0001 changing lr epoch 338, time 246.62, cls_loss 0.0006 cls_loss_mapping 0.0019 cls_loss_causal 0.4748 re_mapping 0.0032 re_causal 0.0100 /// teacc 99.18 lr 0.00010000 Epoch 340, weight, value: tensor([[-0.2267, -0.0591, 0.2198, ..., 0.0412, -0.2429, -0.1727], [-0.1864, -0.2769, -0.0751, ..., -0.1356, -0.3357, -0.1035], [ 0.2937, -0.0763, -0.3044, ..., -0.2332, -0.2754, -0.2211], ..., [-0.1193, -0.0437, -0.3966, ..., -0.1353, 0.1255, 0.0373], [-0.2237, -0.1463, -0.0561, ..., -0.2382, -0.0599, -0.0669], [-0.1103, -0.1055, -0.1614, ..., 0.1036, 0.0278, -0.1740]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., 2.7940e-09, 1.8626e-09, 1.8626e-09], [ 9.3132e-10, 0.0000e+00, -2.7940e-09, ..., -5.5879e-09, 1.5460e-07, 2.7940e-09], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, 2.7940e-09, 2.7940e-09], ..., [ 9.3132e-10, 9.3132e-10, 1.8626e-09, ..., 7.4506e-09, -2.7847e-07, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 5.5879e-09, 5.5879e-09], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.0489e-08, 1.1455e-07, 2.7940e-09]], device='cuda:0') Epoch 340, bias, value: tensor([ 1.1821e-03, 6.7544e-03, 1.4147e-02, 1.4303e-05, 2.4928e-02, -5.7604e-03, -9.4102e-04, -2.1872e-03, 1.0707e-02, -1.7279e-02], device='cuda:0'), grad: tensor([ 3.2596e-08, 3.1199e-07, 2.7008e-08, -9.3132e-10, -1.9465e-07, 1.0803e-07, 1.6764e-08, -8.0466e-07, 2.7008e-08, 4.8708e-07], device='cuda:0') 100 0.0001 changing lr epoch 339, time 246.49, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4536 re_mapping 0.0030 re_causal 0.0094 /// teacc 99.13 lr 0.00010000 Epoch 341, weight, value: tensor([[-0.2269, -0.0592, 0.2201, ..., 0.0415, -0.2430, -0.1727], [-0.1890, -0.2777, -0.0751, ..., -0.1369, -0.3363, -0.1037], [ 0.2944, -0.0765, -0.3048, ..., -0.2337, -0.2758, -0.2212], ..., [-0.1195, -0.0439, -0.3980, ..., -0.1355, 0.1257, 0.0374], [-0.2241, -0.1464, -0.0564, ..., -0.2384, -0.0600, -0.0674], [-0.1105, -0.1055, -0.1623, ..., 0.1036, 0.0278, -0.1741]], device='cuda:0'), grad: tensor([[ 5.5879e-09, 1.7695e-08, -8.0094e-08, ..., -3.6322e-08, 0.0000e+00, 0.0000e+00], [ 1.0245e-08, 3.0734e-08, 9.3132e-10, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [ 2.6524e-06, 9.0376e-06, 2.0489e-08, ..., 9.3132e-09, 0.0000e+00, 0.0000e+00], ..., [ 2.0117e-07, 6.5472e-07, 3.7253e-09, ..., 5.5879e-09, -1.8626e-09, 0.0000e+00], [ 3.7253e-08, 2.1420e-08, 2.0489e-08, ..., 9.3132e-09, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 2.7940e-09, 2.9802e-08, ..., 2.0489e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 341, bias, value: tensor([ 1.5533e-03, 6.6092e-03, 1.4545e-02, -7.4696e-05, 2.4801e-02, -5.6534e-03, 5.3902e-04, -2.1868e-03, 1.0739e-02, -1.7307e-02], device='cuda:0'), grad: tensor([-1.5460e-07, 7.0781e-08, 2.3246e-05, -2.5496e-05, -2.7940e-08, 3.6601e-07, 3.0734e-08, 1.7267e-06, 1.6298e-07, 9.8720e-08], device='cuda:0') 100 0.0001 changing lr epoch 340, time 246.37, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4867 re_mapping 0.0028 re_causal 0.0094 /// teacc 99.11 lr 0.00010000 Epoch 342, weight, value: tensor([[-0.2270, -0.0595, 0.2202, ..., 0.0415, -0.2431, -0.1727], [-0.1892, -0.2784, -0.0751, ..., -0.1342, -0.3370, -0.1038], [ 0.2945, -0.0800, -0.3050, ..., -0.2346, -0.2765, -0.2211], ..., [-0.1196, -0.0446, -0.3988, ..., -0.1386, 0.1260, 0.0373], [-0.2244, -0.1468, -0.0566, ..., -0.2387, -0.0602, -0.0692], [-0.1106, -0.1055, -0.1624, ..., 0.1036, 0.0278, -0.1743]], device='cuda:0'), grad: tensor([[ 1.2107e-08, 0.0000e+00, -2.1420e-08, ..., -1.3039e-08, 0.0000e+00, 0.0000e+00], [ 1.1176e-08, 0.0000e+00, -0.0000e+00, ..., 9.3132e-10, 1.8626e-09, 0.0000e+00], [-9.8720e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], ..., [ 2.5146e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -3.7253e-09, 0.0000e+00], [ 4.4703e-08, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, 1.8626e-09, 2.7940e-09], [ 1.8626e-09, 0.0000e+00, 1.8626e-08, ..., 2.0489e-08, 2.7940e-09, 0.0000e+00]], device='cuda:0') Epoch 342, bias, value: tensor([ 0.0016, 0.0082, 0.0143, -0.0003, 0.0248, -0.0054, 0.0007, -0.0036, 0.0106, -0.0173], device='cuda:0'), grad: tensor([ 0.0000e+00, 1.3970e-08, -2.8405e-07, 2.7940e-09, -1.8626e-08, 1.8626e-09, 4.6566e-09, 7.4506e-08, 1.4435e-07, 7.0781e-08], device='cuda:0') 100 0.0001 changing lr epoch 341, time 246.33, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4715 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.07 lr 0.00010000 Epoch 343, weight, value: tensor([[-0.2272, -0.0596, 0.2202, ..., 0.0415, -0.2431, -0.1727], [-0.1911, -0.2785, -0.0751, ..., -0.1342, -0.3388, -0.1046], [ 0.2952, -0.0800, -0.3051, ..., -0.2349, -0.2770, -0.2206], ..., [-0.1198, -0.0447, -0.3990, ..., -0.1386, 0.1266, 0.0377], [-0.2248, -0.1470, -0.0568, ..., -0.2390, -0.0603, -0.0696], [-0.1109, -0.1054, -0.1625, ..., 0.1035, 0.0278, -0.1744]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.5390e-08, ..., -2.7008e-08, 9.3132e-10, 9.3132e-10], [ 1.8626e-09, 0.0000e+00, -1.3039e-08, ..., 9.3132e-10, 1.8626e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 9.3132e-09, ..., 6.5193e-09, -9.3132e-10, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 8.3819e-09, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 5.5879e-09, -9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 343, bias, value: tensor([ 0.0016, 0.0082, 0.0147, -0.0004, 0.0248, -0.0054, 0.0008, -0.0036, 0.0106, -0.0174], device='cuda:0'), grad: tensor([-1.0990e-07, -2.1793e-07, 1.0245e-08, -1.7695e-08, 2.3283e-08, 4.0047e-08, 1.8626e-08, 1.4994e-07, 8.8476e-08, 2.6077e-08], device='cuda:0') 100 0.0001 changing lr epoch 342, time 245.84, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4783 re_mapping 0.0031 re_causal 0.0101 /// teacc 99.09 lr 0.00010000 Epoch 344, weight, value: tensor([[-0.2274, -0.0594, 0.2203, ..., 0.0416, -0.2432, -0.1728], [-0.1914, -0.2791, -0.0754, ..., -0.1342, -0.3398, -0.1058], [ 0.2954, -0.0800, -0.3053, ..., -0.2349, -0.2772, -0.2206], ..., [-0.1198, -0.0449, -0.3997, ..., -0.1387, 0.1268, 0.0377], [-0.2253, -0.1474, -0.0572, ..., -0.2392, -0.0604, -0.0712], [-0.1110, -0.1055, -0.1626, ..., 0.1035, 0.0278, -0.1745]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5646e-07, ..., -1.1548e-07, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, 1.3039e-08, 2.7940e-09], [-1.8626e-09, 0.0000e+00, 3.7253e-09, ..., 2.7940e-09, 8.3819e-09, 1.8626e-09], ..., [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., -2.7940e-09, -4.4703e-08, -9.3132e-09], [ 9.3132e-10, 0.0000e+00, 3.7253e-09, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 8.7544e-08, ..., 6.9849e-08, 1.8626e-08, 3.7253e-09]], device='cuda:0') Epoch 344, bias, value: tensor([ 0.0016, 0.0081, 0.0148, -0.0003, 0.0248, -0.0054, 0.0009, -0.0036, 0.0104, -0.0174], device='cuda:0'), grad: tensor([-3.5111e-07, 1.0617e-07, 1.7043e-07, 1.9558e-08, -9.3132e-09, 3.8184e-08, 1.4529e-07, -1.8533e-07, -2.1886e-07, 2.7567e-07], device='cuda:0') 100 0.0001 changing lr epoch 343, time 246.16, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4727 re_mapping 0.0031 re_causal 0.0102 /// teacc 99.14 lr 0.00010000 Epoch 345, weight, value: tensor([[-0.2275, -0.0595, 0.2202, ..., 0.0416, -0.2436, -0.1729], [-0.1915, -0.2793, -0.0754, ..., -0.1342, -0.3405, -0.1060], [ 0.2955, -0.0800, -0.3056, ..., -0.2352, -0.2780, -0.2208], ..., [-0.1199, -0.0450, -0.4000, ..., -0.1387, 0.1269, 0.0378], [-0.2257, -0.1477, -0.0588, ..., -0.2398, -0.0609, -0.0724], [-0.1111, -0.1054, -0.1627, ..., 0.1035, 0.0277, -0.1746]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -7.4506e-09, ..., -5.5879e-09, 1.8626e-09, 0.0000e+00], [ 2.6077e-08, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, 1.8626e-09, 0.0000e+00], [-4.8429e-08, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], ..., [ 1.3970e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, -9.3132e-10, 0.0000e+00], [ 2.7940e-09, 0.0000e+00, 1.4901e-08, ..., 0.0000e+00, 5.5879e-09, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 1.1176e-08, ..., 7.4506e-09, 9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 345, bias, value: tensor([ 0.0015, 0.0083, 0.0148, -0.0003, 0.0249, -0.0054, 0.0011, -0.0038, 0.0103, -0.0174], device='cuda:0'), grad: tensor([-7.4506e-09, -4.5449e-07, -7.8231e-08, -3.7253e-09, 1.1176e-08, 4.3772e-08, -7.6368e-08, 4.7684e-07, 5.0291e-08, 4.4703e-08], device='cuda:0') 100 0.0001 changing lr epoch 344, time 246.23, cls_loss 0.0005 cls_loss_mapping 0.0018 cls_loss_causal 0.4732 re_mapping 0.0031 re_causal 0.0095 /// teacc 99.14 lr 0.00010000 Epoch 346, weight, value: tensor([[-0.2278, -0.0596, 0.2209, ..., 0.0419, -0.2434, -0.1728], [-0.1920, -0.2796, -0.0754, ..., -0.1351, -0.3422, -0.1063], [ 0.2958, -0.0800, -0.3064, ..., -0.2361, -0.2782, -0.2208], ..., [-0.1200, -0.0451, -0.4008, ..., -0.1389, 0.1268, 0.0379], [-0.2260, -0.1478, -0.0591, ..., -0.2403, -0.0611, -0.0727], [-0.1115, -0.1054, -0.1635, ..., 0.1036, 0.0278, -0.1748]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -4.6566e-09, ..., 4.6566e-09, 9.3132e-10, -0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 3.2596e-08, 0.0000e+00], [-3.3807e-07, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], ..., [ 3.2969e-07, 1.8626e-09, 0.0000e+00, ..., 2.7940e-09, -5.4948e-08, -9.3132e-10], [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., -1.5832e-08, 1.3039e-08, 1.8626e-09]], device='cuda:0') Epoch 346, bias, value: tensor([ 0.0022, 0.0081, 0.0149, -0.0004, 0.0249, -0.0051, 0.0003, -0.0039, 0.0102, -0.0173], device='cuda:0'), grad: tensor([ 9.3132e-09, 1.2480e-07, -7.4040e-07, 3.7253e-09, 2.0489e-08, -9.3132e-10, 7.4506e-09, 5.2247e-07, 2.0489e-08, 3.9116e-08], device='cuda:0') 100 0.0001 changing lr epoch 345, time 246.59, cls_loss 0.0006 cls_loss_mapping 0.0017 cls_loss_causal 0.4657 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.14 lr 0.00010000 Epoch 347, weight, value: tensor([[-0.2283, -0.0603, 0.2211, ..., 0.0420, -0.2439, -0.1729], [-0.1923, -0.2805, -0.0756, ..., -0.1352, -0.3440, -0.1070], [ 0.2964, -0.0792, -0.3067, ..., -0.2364, -0.2766, -0.2208], ..., [-0.1202, -0.0455, -0.4017, ..., -0.1390, 0.1272, 0.0381], [-0.2268, -0.1484, -0.0599, ..., -0.2412, -0.0620, -0.0731], [-0.1117, -0.1055, -0.1639, ..., 0.1035, 0.0277, -0.1750]], device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, -9.6858e-08, ..., -3.0734e-08, 3.7253e-09, 9.3132e-10], [ 2.7940e-08, 0.0000e+00, 2.7940e-09, ..., 5.5879e-09, 1.0245e-08, 9.3132e-10], [-2.5798e-07, 0.0000e+00, 8.3819e-09, ..., 2.7940e-09, -5.4948e-08, 0.0000e+00], ..., [ 1.7695e-07, 0.0000e+00, 2.7940e-09, ..., 5.5879e-09, 3.3528e-08, -9.3132e-10], [ 1.0245e-08, 0.0000e+00, 1.0245e-08, ..., 9.3132e-09, 1.3039e-08, 0.0000e+00], [ 2.7940e-09, 0.0000e+00, 2.0489e-08, ..., 3.1013e-07, 7.0781e-08, 9.3132e-10]], device='cuda:0') Epoch 347, bias, value: tensor([ 0.0022, 0.0080, 0.0153, -0.0005, 0.0250, -0.0050, 0.0004, -0.0038, 0.0099, -0.0174], device='cuda:0'), grad: tensor([-1.9837e-07, 6.4261e-08, -7.3574e-07, 8.0094e-08, -6.5193e-07, 7.9162e-08, 4.9360e-08, 5.3924e-07, 1.0803e-07, 6.6031e-07], device='cuda:0') 100 0.0001 changing lr epoch 346, time 246.07, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4685 re_mapping 0.0030 re_causal 0.0097 /// teacc 99.15 lr 0.00010000 Epoch 348, weight, value: tensor([[-0.2285, -0.0604, 0.2214, ..., 0.0422, -0.2441, -0.1729], [-0.1931, -0.2812, -0.0756, ..., -0.1352, -0.3456, -0.1078], [ 0.2972, -0.0791, -0.3071, ..., -0.2370, -0.2755, -0.2206], ..., [-0.1207, -0.0458, -0.4037, ..., -0.1391, 0.1276, 0.0382], [-0.2275, -0.1486, -0.0603, ..., -0.2416, -0.0623, -0.0732], [-0.1119, -0.1057, -0.1643, ..., 0.1034, 0.0276, -0.1751]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.4506e-08, ..., -3.1665e-08, 9.3132e-10, -9.3132e-10], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, 1.3970e-08, 0.0000e+00], [-2.7940e-09, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, 9.3132e-10, -0.0000e+00], ..., [ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 2.6077e-08, -1.8626e-09, 0.0000e+00], [ 0.0000e+00, -3.0734e-08, 5.5879e-09, ..., 2.7940e-09, -3.2596e-08, -2.7940e-09], [ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., -2.0489e-08, -1.3970e-08, 0.0000e+00]], device='cuda:0') Epoch 348, bias, value: tensor([ 0.0026, 0.0079, 0.0157, -0.0006, 0.0250, -0.0050, 0.0003, -0.0038, 0.0098, -0.0175], device='cuda:0'), grad: tensor([-1.3597e-07, 3.6322e-08, 8.3819e-09, 3.4180e-07, 4.6566e-09, 9.5926e-08, 1.0058e-07, 5.4948e-08, -4.6473e-07, -3.7253e-08], device='cuda:0') 100 0.0001 changing lr epoch 347, time 246.11, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4518 re_mapping 0.0030 re_causal 0.0095 /// teacc 99.17 lr 0.00010000 Epoch 349, weight, value: tensor([[-0.2298, -0.0608, 0.2214, ..., 0.0422, -0.2449, -0.1733], [-0.1932, -0.2818, -0.0755, ..., -0.1353, -0.3468, -0.1081], [ 0.2975, -0.0791, -0.3069, ..., -0.2375, -0.2762, -0.2207], ..., [-0.1207, -0.0460, -0.4051, ..., -0.1392, 0.1279, 0.0385], [-0.2280, -0.1482, -0.0607, ..., -0.2422, -0.0626, -0.0733], [-0.1120, -0.1057, -0.1646, ..., 0.1034, 0.0276, -0.1753]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.4506e-09, ..., 1.1176e-08, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8720e-07, 2.7940e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, -1.8626e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 1.8626e-09, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.5146e-08, 2.7940e-09, 9.3132e-10]], device='cuda:0') Epoch 349, bias, value: tensor([ 0.0024, 0.0079, 0.0158, -0.0006, 0.0251, -0.0050, 0.0002, -0.0038, 0.0098, -0.0176], device='cuda:0'), grad: tensor([ 1.4901e-08, 6.2212e-07, 1.8626e-09, 2.6077e-08, -7.2177e-07, -6.8918e-08, 2.9802e-08, 2.7940e-09, 9.3132e-09, 8.2888e-08], device='cuda:0') 100 0.0001 changing lr epoch 348, time 245.89, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4748 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.09 lr 0.00010000 Epoch 350, weight, value: tensor([[-0.2301, -0.0614, 0.2216, ..., 0.0422, -0.2450, -0.1733], [-0.1934, -0.2823, -0.0755, ..., -0.1353, -0.3476, -0.1092], [ 0.2976, -0.0791, -0.3071, ..., -0.2378, -0.2766, -0.2209], ..., [-0.1208, -0.0462, -0.4058, ..., -0.1392, 0.1282, 0.0385], [-0.2284, -0.1485, -0.0610, ..., -0.2429, -0.0627, -0.0741], [-0.1134, -0.1058, -0.1647, ..., 0.1034, 0.0276, -0.1755]], device='cuda:0'), grad: tensor([[ 4.6566e-09, 0.0000e+00, -2.3283e-09, ..., 9.3132e-10, 1.3970e-09, 4.6566e-10], [ 3.0268e-08, 0.0000e+00, -6.9849e-09, ..., 2.6543e-08, 1.1642e-07, -0.0000e+00], [ 2.6543e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-10, 0.0000e+00], ..., [ 2.0023e-08, 0.0000e+00, 4.6566e-10, ..., -2.7474e-08, -1.2573e-07, -4.6566e-10], [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, 0.0000e+00, 0.0000e+00], [ 6.9849e-09, 0.0000e+00, 1.8626e-09, ..., -3.7253e-09, 6.0536e-09, 4.6566e-10]], device='cuda:0') Epoch 350, bias, value: tensor([ 0.0025, 0.0082, 0.0158, -0.0007, 0.0250, -0.0046, -0.0001, -0.0041, 0.0097, -0.0175], device='cuda:0'), grad: tensor([ 2.7008e-08, 3.9907e-07, 1.3411e-07, -4.8056e-07, -1.3970e-09, 5.7742e-08, 1.0803e-07, -3.1339e-07, 2.3283e-08, 3.6787e-08], device='cuda:0') 100 0.0001 changing lr epoch 349, time 246.09, cls_loss 0.0007 cls_loss_mapping 0.0015 cls_loss_causal 0.4948 re_mapping 0.0027 re_causal 0.0090 /// teacc 99.17 lr 0.00010000 Epoch 351, weight, value: tensor([[-0.2303, -0.0619, 0.2220, ..., 0.0425, -0.2451, -0.1733], [-0.1935, -0.2829, -0.0756, ..., -0.1354, -0.3491, -0.1101], [ 0.2978, -0.0791, -0.3077, ..., -0.2386, -0.2781, -0.2211], ..., [-0.1209, -0.0466, -0.4075, ..., -0.1392, 0.1286, 0.0388], [-0.2288, -0.1488, -0.0618, ..., -0.2436, -0.0636, -0.0748], [-0.1137, -0.1058, -0.1650, ..., 0.1033, 0.0276, -0.1758]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 1.8626e-09, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -6.5193e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 351, bias, value: tensor([ 0.0028, 0.0084, 0.0156, -0.0007, 0.0251, -0.0045, -0.0004, -0.0043, 0.0095, -0.0177], device='cuda:0'), grad: tensor([ 8.3819e-09, -4.5449e-07, 3.7998e-07, 5.5879e-09, 3.5390e-08, 2.7940e-09, 1.8626e-08, 2.0489e-08, -2.4214e-08, 6.5193e-09], device='cuda:0') 100 0.0001 changing lr epoch 350, time 246.52, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4642 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.09 lr 0.00010000 Epoch 352, weight, value: tensor([[-0.2306, -0.0621, 0.2224, ..., 0.0426, -0.2451, -0.1732], [-0.1943, -0.2836, -0.0756, ..., -0.1355, -0.3504, -0.1103], [ 0.2982, -0.0792, -0.3080, ..., -0.2379, -0.2791, -0.2212], ..., [-0.1210, -0.0471, -0.4086, ..., -0.1395, 0.1286, 0.0388], [-0.2298, -0.1492, -0.0621, ..., -0.2440, -0.0641, -0.0751], [-0.1156, -0.1059, -0.1653, ..., 0.1032, 0.0276, -0.1760]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 1.8626e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., 2.7940e-09, 2.7940e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 1.8626e-09, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.3819e-09, 1.8626e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, -1.2107e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.5193e-09, -0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 352, bias, value: tensor([ 0.0030, 0.0076, 0.0158, -0.0009, 0.0252, -0.0042, -0.0007, -0.0036, 0.0096, -0.0177], device='cuda:0'), grad: tensor([ 2.1420e-08, -6.4746e-06, 2.7008e-08, 2.7940e-09, -1.7695e-08, 1.0245e-08, 5.7742e-08, 5.6438e-06, -2.2352e-08, 7.5903e-07], device='cuda:0') 100 0.0001 changing lr epoch 351, time 246.40, cls_loss 0.0005 cls_loss_mapping 0.0010 cls_loss_causal 0.4495 re_mapping 0.0028 re_causal 0.0090 /// teacc 99.20 lr 0.00010000 Epoch 353, weight, value: tensor([[-0.2308, -0.0625, 0.2224, ..., 0.0426, -0.2454, -0.1733], [-0.1946, -0.2846, -0.0756, ..., -0.1360, -0.3516, -0.1106], [ 0.2987, -0.0792, -0.3082, ..., -0.2375, -0.2799, -0.2211], ..., [-0.1211, -0.0474, -0.4094, ..., -0.1396, 0.1286, 0.0388], [-0.2303, -0.1495, -0.0625, ..., -0.2450, -0.0645, -0.0756], [-0.1154, -0.1057, -0.1654, ..., 0.1032, 0.0276, -0.1761]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.5832e-07, ..., -6.6124e-08, 1.8626e-09, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 9.3132e-10, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 1.8626e-09, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, -3.7253e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 5.6811e-08, ..., 2.6077e-08, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.1665e-08, ..., 2.7940e-09, -3.7253e-09, 0.0000e+00]], device='cuda:0') Epoch 353, bias, value: tensor([ 0.0030, 0.0075, 0.0160, -0.0009, 0.0253, -0.0042, -0.0007, -0.0036, 0.0094, -0.0177], device='cuda:0'), grad: tensor([-2.8964e-07, -9.3132e-10, 1.2107e-08, -5.5879e-09, 7.4506e-09, 1.1176e-08, 1.3039e-07, -2.7940e-09, 1.1362e-07, 3.0734e-08], device='cuda:0') 100 0.0001 changing lr epoch 352, time 246.32, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4719 re_mapping 0.0028 re_causal 0.0096 /// teacc 99.16 lr 0.00010000 Epoch 354, weight, value: tensor([[-0.2310, -0.0627, 0.2225, ..., 0.0427, -0.2456, -0.1733], [-0.1947, -0.2850, -0.0756, ..., -0.1360, -0.3527, -0.1115], [ 0.2991, -0.0793, -0.3084, ..., -0.2363, -0.2806, -0.2212], ..., [-0.1212, -0.0476, -0.4100, ..., -0.1397, 0.1289, 0.0388], [-0.2295, -0.1501, -0.0627, ..., -0.2447, -0.0646, -0.0762], [-0.1164, -0.1057, -0.1656, ..., 0.1032, 0.0275, -0.1765]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-09, ..., -1.8626e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [-1.7695e-08, -3.7253e-09, 0.0000e+00, ..., 0.0000e+00, -9.3132e-10, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, -9.3132e-10, 0.0000e+00], [ 0.0000e+00, 9.3132e-10, -9.3132e-10, ..., 9.3132e-10, -9.3132e-10, 0.0000e+00], [ 1.6764e-08, 1.1176e-08, 4.6566e-09, ..., -3.6228e-07, -2.4214e-08, 0.0000e+00]], device='cuda:0') Epoch 354, bias, value: tensor([ 0.0030, 0.0058, 0.0159, -0.0022, 0.0253, -0.0060, 0.0023, -0.0019, 0.0095, -0.0178], device='cuda:0'), grad: tensor([-3.7253e-09, -2.5146e-08, -2.6077e-08, 2.5146e-08, 6.9849e-07, -1.2945e-07, 2.5146e-08, 2.8871e-08, 1.8626e-09, -5.9418e-07], device='cuda:0') 100 0.0001 changing lr epoch 353, time 245.83, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4554 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.21 lr 0.00010000 Epoch 355, weight, value: tensor([[-0.2311, -0.0628, 0.2226, ..., 0.0428, -0.2457, -0.1733], [-0.1949, -0.2851, -0.0756, ..., -0.1362, -0.3537, -0.1120], [ 0.2993, -0.0792, -0.3087, ..., -0.2365, -0.2816, -0.2214], ..., [-0.1212, -0.0477, -0.4115, ..., -0.1398, 0.1293, 0.0392], [-0.2295, -0.1504, -0.0640, ..., -0.2450, -0.0660, -0.0771], [-0.1172, -0.1058, -0.1658, ..., 0.1031, 0.0275, -0.1769]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, 4.6566e-09, 0.0000e+00], [-2.0489e-08, 0.0000e+00, -9.3132e-10, ..., 9.3132e-10, 9.3132e-10, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, -9.3132e-09, -0.0000e+00], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 4.6566e-09, 4.6566e-09], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.7695e-08, 3.7253e-09, 0.0000e+00]], device='cuda:0') Epoch 355, bias, value: tensor([ 0.0031, 0.0060, 0.0157, -0.0027, 0.0253, -0.0056, 0.0024, -0.0021, 0.0088, -0.0178], device='cuda:0'), grad: tensor([ 1.8626e-09, 1.4901e-08, -2.4214e-08, 1.2107e-08, -4.3772e-08, -5.8673e-08, 3.0734e-08, -2.0489e-08, 3.7253e-08, 4.7497e-08], device='cuda:0') 100 0.0001 changing lr epoch 354, time 246.04, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4621 re_mapping 0.0030 re_causal 0.0095 /// teacc 99.13 lr 0.00010000 Epoch 356, weight, value: tensor([[-0.2313, -0.0630, 0.2232, ..., 0.0410, -0.2459, -0.1734], [-0.1951, -0.2854, -0.0771, ..., -0.1364, -0.3557, -0.1124], [ 0.2994, -0.0792, -0.3102, ..., -0.2372, -0.2826, -0.2214], ..., [-0.1213, -0.0478, -0.4121, ..., -0.1398, 0.1307, 0.0393], [-0.2300, -0.1507, -0.0643, ..., -0.2462, -0.0664, -0.0774], [-0.1176, -0.1058, -0.1659, ..., 0.1033, 0.0274, -0.1770]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 0.0000e+00, 4.2841e-08, 1.8626e-09], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 1.5832e-08, 9.3132e-10], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, -1.3597e-07, -6.5193e-09], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.7253e-08, 4.5635e-08, 2.7940e-09]], device='cuda:0') Epoch 356, bias, value: tensor([ 0.0022, 0.0059, 0.0156, -0.0029, 0.0254, -0.0057, 0.0024, -0.0020, 0.0086, -0.0177], device='cuda:0'), grad: tensor([ 2.0489e-08, 2.2817e-07, 1.0990e-07, 1.8626e-08, 1.3225e-07, 1.5832e-08, -4.6566e-09, -6.4727e-07, -1.5926e-07, 2.9057e-07], device='cuda:0') 100 0.0001 changing lr epoch 355, time 246.46, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4676 re_mapping 0.0034 re_causal 0.0103 /// teacc 99.14 lr 0.00010000 Epoch 357, weight, value: tensor([[-0.2317, -0.0631, 0.2233, ..., 0.0410, -0.2463, -0.1734], [-0.1954, -0.2859, -0.0772, ..., -0.1364, -0.3566, -0.1125], [ 0.2997, -0.0795, -0.3108, ..., -0.2376, -0.2848, -0.2216], ..., [-0.1214, -0.0486, -0.4127, ..., -0.1398, 0.1310, 0.0395], [-0.2317, -0.1513, -0.0648, ..., -0.2466, -0.0666, -0.0775], [-0.1192, -0.1060, -0.1663, ..., 0.1032, 0.0274, -0.1771]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.6764e-08, ..., -6.5193e-09, 9.3132e-10, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [-2.1420e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.4901e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 9.3132e-09, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 1.4901e-08, ..., 5.5879e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 357, bias, value: tensor([ 0.0021, 0.0059, 0.0156, -0.0053, 0.0254, -0.0040, 0.0027, -0.0020, 0.0085, -0.0178], device='cuda:0'), grad: tensor([-2.5146e-08, -4.7497e-08, -4.6566e-08, -1.9558e-08, 9.3132e-10, 1.3039e-08, -6.5193e-09, 7.9162e-08, 2.6077e-08, 3.5390e-08], device='cuda:0') 100 0.0001 changing lr epoch 356, time 245.86, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4664 re_mapping 0.0033 re_causal 0.0100 /// teacc 99.21 lr 0.00010000 Epoch 358, weight, value: tensor([[-0.2322, -0.0633, 0.2234, ..., 0.0411, -0.2469, -0.1735], [-0.1956, -0.2865, -0.0771, ..., -0.1364, -0.3580, -0.1129], [ 0.3003, -0.0795, -0.3111, ..., -0.2377, -0.2876, -0.2217], ..., [-0.1216, -0.0488, -0.4131, ..., -0.1400, 0.1315, 0.0397], [-0.2338, -0.1536, -0.0652, ..., -0.2472, -0.0669, -0.0776], [-0.1194, -0.1060, -0.1664, ..., 0.1032, 0.0274, -0.1771]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, -2.7940e-09, ..., 1.1176e-08, 5.5879e-09, 0.0000e+00], [ 5.5879e-09, 1.8626e-09, 0.0000e+00, ..., 8.3819e-09, 1.8626e-09, 0.0000e+00], [-3.5390e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 2.5146e-08, 9.3132e-10, 9.3132e-10, ..., 1.0245e-08, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 5.5879e-09, 9.3132e-10, ..., 3.2596e-08, 4.6566e-09, 0.0000e+00], [ 0.0000e+00, -7.8231e-08, 2.7940e-09, ..., -6.9756e-07, -1.4249e-07, 0.0000e+00]], device='cuda:0') Epoch 358, bias, value: tensor([ 0.0021, 0.0061, 0.0151, -0.0055, 0.0254, -0.0038, 0.0027, -0.0021, 0.0083, -0.0178], device='cuda:0'), grad: tensor([ 3.4459e-08, 1.3039e-08, -7.9162e-08, 1.1176e-07, 1.0254e-06, 2.9150e-07, -1.3970e-08, 9.3132e-08, 6.6124e-08, -1.5413e-06], device='cuda:0') 100 0.0001 changing lr epoch 357, time 245.96, cls_loss 0.0004 cls_loss_mapping 0.0010 cls_loss_causal 0.4517 re_mapping 0.0029 re_causal 0.0092 /// teacc 99.19 lr 0.00010000 Epoch 359, weight, value: tensor([[-0.2327, -0.0635, 0.2235, ..., 0.0411, -0.2471, -0.1735], [-0.1958, -0.2867, -0.0771, ..., -0.1364, -0.3590, -0.1137], [ 0.3005, -0.0795, -0.3112, ..., -0.2378, -0.2878, -0.2217], ..., [-0.1217, -0.0490, -0.4139, ..., -0.1399, 0.1329, 0.0399], [-0.2347, -0.1554, -0.0654, ..., -0.2473, -0.0656, -0.0778], [-0.1194, -0.1059, -0.1665, ..., 0.1032, 0.0272, -0.1772]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., -6.5193e-09, 0.0000e+00, -9.3132e-10], [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 1.8626e-09, 0.0000e+00, ..., 4.6566e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.6077e-08, 4.6566e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.0245e-08, ..., -4.9360e-08, -1.0245e-08, 0.0000e+00]], device='cuda:0') Epoch 359, bias, value: tensor([ 0.0021, 0.0061, 0.0151, -0.0056, 0.0254, -0.0038, 0.0026, -0.0020, 0.0091, -0.0179], device='cuda:0'), grad: tensor([-1.9558e-08, -6.7055e-08, 7.4506e-09, -3.2596e-08, 5.5879e-08, 1.6764e-08, 1.3039e-08, 6.6124e-08, 9.7789e-08, -1.2759e-07], device='cuda:0') 100 0.0001 changing lr epoch 358, time 246.02, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4759 re_mapping 0.0028 re_causal 0.0091 /// teacc 99.19 lr 0.00010000 Epoch 360, weight, value: tensor([[-0.2333, -0.0638, 0.2236, ..., 0.0410, -0.2474, -0.1735], [-0.1961, -0.2873, -0.0771, ..., -0.1366, -0.3611, -0.1138], [ 0.3008, -0.0796, -0.3114, ..., -0.2401, -0.2883, -0.2217], ..., [-0.1218, -0.0493, -0.4146, ..., -0.1401, 0.1330, 0.0399], [-0.2352, -0.1557, -0.0656, ..., -0.2479, -0.0657, -0.0779], [-0.1197, -0.1059, -0.1668, ..., 0.1032, 0.0273, -0.1773]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.4703e-08, ..., -9.3132e-10, 2.4214e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -3.7253e-09, 5.5879e-09, 0.0000e+00], [-4.6566e-09, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], ..., [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -1.5832e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, -9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.7940e-09, 9.3132e-09, 0.0000e+00]], device='cuda:0') Epoch 360, bias, value: tensor([ 0.0020, 0.0062, 0.0150, -0.0055, 0.0255, -0.0038, 0.0026, -0.0021, 0.0090, -0.0179], device='cuda:0'), grad: tensor([ 1.1548e-07, -4.7497e-08, -2.7940e-09, 3.7253e-09, 1.7695e-08, 5.2154e-08, -1.7136e-07, 2.7940e-09, -5.5879e-09, 3.9116e-08], device='cuda:0') 100 0.0001 changing lr epoch 359, time 246.68, cls_loss 0.0004 cls_loss_mapping 0.0015 cls_loss_causal 0.4721 re_mapping 0.0030 re_causal 0.0096 /// teacc 99.13 lr 0.00010000 Epoch 361, weight, value: tensor([[-0.2337, -0.0643, 0.2237, ..., 0.0410, -0.2475, -0.1736], [-0.1974, -0.2879, -0.0771, ..., -0.1366, -0.3621, -0.1143], [ 0.3008, -0.0796, -0.3118, ..., -0.2408, -0.2883, -0.2218], ..., [-0.1214, -0.0498, -0.4148, ..., -0.1402, 0.1332, 0.0400], [-0.2358, -0.1576, -0.0657, ..., -0.2481, -0.0658, -0.0780], [-0.1200, -0.1090, -0.1669, ..., 0.1032, 0.0273, -0.1773]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1176e-08, ..., 2.7940e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.9914e-06, 2.7940e-09, 0.0000e+00], [-2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 8.3819e-09, 1.8626e-09, 0.0000e+00], ..., [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 3.4925e-07, 9.3132e-10, 0.0000e+00], [-9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 3.3528e-08, 7.4506e-09, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 7.4506e-09, ..., 2.3264e-06, -1.8626e-08, 0.0000e+00]], device='cuda:0') Epoch 361, bias, value: tensor([ 0.0020, 0.0060, 0.0137, -0.0049, 0.0255, -0.0038, 0.0026, -0.0018, 0.0084, -0.0180], device='cuda:0'), grad: tensor([ 6.5193e-09, -1.4573e-05, 7.0781e-08, 3.8650e-07, 1.1148e-06, -4.0606e-07, 3.0734e-08, 2.4233e-06, 5.2154e-08, 1.0878e-05], device='cuda:0') 100 0.0001 changing lr epoch 360, time 246.19, cls_loss 0.0003 cls_loss_mapping 0.0013 cls_loss_causal 0.4674 re_mapping 0.0031 re_causal 0.0098 /// teacc 99.18 lr 0.00010000 Epoch 362, weight, value: tensor([[-0.2338, -0.0644, 0.2239, ..., 0.0410, -0.2476, -0.1736], [-0.1975, -0.2882, -0.0771, ..., -0.1365, -0.3636, -0.1149], [ 0.3010, -0.0797, -0.3121, ..., -0.2410, -0.2884, -0.2218], ..., [-0.1215, -0.0501, -0.4153, ..., -0.1402, 0.1341, 0.0405], [-0.2361, -0.1583, -0.0660, ..., -0.2486, -0.0661, -0.0783], [-0.1201, -0.1090, -0.1675, ..., 0.1031, 0.0272, -0.1779]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 9.3132e-10, -5.5879e-09, ..., -6.5193e-09, 5.5879e-09, 1.8626e-09], [ 1.8626e-09, 1.8626e-09, 8.3819e-09, ..., 7.4506e-09, 2.4214e-08, 1.8626e-09], [ 3.7253e-09, 3.7253e-09, 4.6566e-09, ..., 2.7940e-09, 3.7253e-09, 9.3132e-10], ..., [ 2.7940e-09, 3.7253e-09, 9.3132e-10, ..., 1.1176e-08, -3.2596e-08, 0.0000e+00], [ 2.7940e-09, 2.7940e-09, 6.5193e-09, ..., 6.5193e-09, 1.0245e-08, 9.3132e-10], [ 9.3132e-10, 9.3132e-10, 1.2107e-08, ..., 3.9581e-07, 6.0536e-08, 0.0000e+00]], device='cuda:0') Epoch 362, bias, value: tensor([ 0.0021, 0.0060, 0.0136, -0.0048, 0.0255, -0.0038, 0.0026, -0.0017, 0.0081, -0.0180], device='cuda:0'), grad: tensor([ 1.1176e-08, 6.5193e-09, 4.0978e-08, -1.0803e-07, -8.9966e-07, 1.4063e-07, -1.9558e-07, 2.8871e-08, 6.7987e-08, 9.1270e-07], device='cuda:0') 100 0.0001 changing lr epoch 361, time 245.75, cls_loss 0.0006 cls_loss_mapping 0.0016 cls_loss_causal 0.4743 re_mapping 0.0029 re_causal 0.0093 /// teacc 99.13 lr 0.00010000 Epoch 363, weight, value: tensor([[-0.2351, -0.0649, 0.2237, ..., 0.0410, -0.2488, -0.1742], [-0.1982, -0.2887, -0.0767, ..., -0.1365, -0.3656, -0.1155], [ 0.3016, -0.0797, -0.3123, ..., -0.2413, -0.2888, -0.2219], ..., [-0.1217, -0.0503, -0.4168, ..., -0.1403, 0.1352, 0.0412], [-0.2368, -0.1586, -0.0663, ..., -0.2495, -0.0663, -0.0784], [-0.1203, -0.1090, -0.1677, ..., 0.1031, 0.0271, -0.1784]], device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 9.3132e-09, 9.3132e-10, 0.0000e+00], [-1.6764e-08, -0.0000e+00, 4.6566e-09, ..., 2.1420e-08, 0.0000e+00, 0.0000e+00], ..., [ 2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, -5.5879e-09, 0.0000e+00], [ 2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, 9.3132e-10, 9.3132e-10], [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 5.5879e-09, 9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 363, bias, value: tensor([ 0.0017, 0.0051, 0.0136, -0.0048, 0.0256, -0.0037, 0.0024, -0.0007, 0.0081, -0.0181], device='cuda:0'), grad: tensor([ 2.8871e-08, 8.3819e-09, 1.5832e-08, -2.5518e-07, -1.4808e-07, 2.2538e-07, 6.9849e-08, 1.8626e-08, 1.2107e-08, 2.5146e-08], device='cuda:0') 100 0.0001 changing lr epoch 362, time 246.08, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4764 re_mapping 0.0029 re_causal 0.0094 /// teacc 99.17 lr 0.00010000 Epoch 364, weight, value: tensor([[-0.2353, -0.0651, 0.2241, ..., 0.0412, -0.2492, -0.1742], [-0.1983, -0.2890, -0.0774, ..., -0.1366, -0.3679, -0.1158], [ 0.3018, -0.0797, -0.3128, ..., -0.2418, -0.2898, -0.2220], ..., [-0.1219, -0.0510, -0.4177, ..., -0.1404, 0.1366, 0.0416], [-0.2370, -0.1587, -0.0664, ..., -0.2494, -0.0664, -0.0786], [-0.1205, -0.1090, -0.1680, ..., 0.1031, 0.0270, -0.1786]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, 0.0000e+00, 0.0000e+00], [ 1.2107e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [-3.4459e-08, -1.8626e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.1176e-08, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 364, bias, value: tensor([ 0.0020, 0.0077, 0.0133, -0.0048, 0.0256, -0.0038, 0.0024, -0.0032, 0.0083, -0.0182], device='cuda:0'), grad: tensor([-2.7940e-09, 1.7695e-08, -5.6811e-08, -4.6566e-09, -3.7253e-09, 4.6566e-09, 0.0000e+00, 2.3283e-08, 1.4901e-08, 9.3132e-09], device='cuda:0') 100 0.0001 changing lr epoch 363, time 246.11, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4607 re_mapping 0.0030 re_causal 0.0093 /// teacc 99.15 lr 0.00010000 Epoch 365, weight, value: tensor([[-0.2359, -0.0654, 0.2241, ..., 0.0412, -0.2498, -0.1744], [-0.1990, -0.2893, -0.0773, ..., -0.1367, -0.3687, -0.1164], [ 0.3027, -0.0803, -0.3130, ..., -0.2426, -0.2921, -0.2238], ..., [-0.1224, -0.0514, -0.4189, ..., -0.1410, 0.1360, 0.0436], [-0.2374, -0.1591, -0.0670, ..., -0.2497, -0.0654, -0.0793], [-0.1207, -0.1090, -0.1683, ..., 0.1031, 0.0272, -0.1787]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.9802e-08, ..., 7.4506e-09, 1.3039e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, 2.7940e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 4.6566e-09, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, -3.7253e-09, -0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -1.0245e-08, -1.3039e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 365, bias, value: tensor([ 0.0019, 0.0078, 0.0134, -0.0048, 0.0257, -0.0038, 0.0025, -0.0034, 0.0086, -0.0182], device='cuda:0'), grad: tensor([ 6.3330e-08, -8.3819e-09, 1.8626e-08, 5.5879e-09, 2.1420e-08, 6.2399e-08, -1.0058e-07, 1.3039e-08, -1.6484e-07, 8.8476e-08], device='cuda:0') 100 0.0001 changing lr epoch 364, time 245.89, cls_loss 0.0007 cls_loss_mapping 0.0018 cls_loss_causal 0.5158 re_mapping 0.0030 re_causal 0.0094 /// teacc 99.22 lr 0.00010000 Epoch 366, weight, value: tensor([[-0.2364, -0.0663, 0.2243, ..., 0.0410, -0.2515, -0.1752], [-0.1996, -0.2900, -0.0779, ..., -0.1367, -0.3704, -0.1202], [ 0.3029, -0.0806, -0.3137, ..., -0.2441, -0.2940, -0.2252], ..., [-0.1225, -0.0517, -0.4196, ..., -0.1416, 0.1368, 0.0466], [-0.2385, -0.1598, -0.0691, ..., -0.2523, -0.0671, -0.0802], [-0.1198, -0.1090, -0.1692, ..., 0.1028, 0.0270, -0.1793]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., 3.7253e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 6.7987e-08, 2.0489e-08, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 1.0245e-08, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, -4.7497e-08, -9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, 4.6566e-09, 0.0000e+00], [ 0.0000e+00, -3.7253e-09, 4.6566e-09, ..., -2.4587e-07, 7.4506e-09, 0.0000e+00]], device='cuda:0') Epoch 366, bias, value: tensor([ 0.0017, 0.0078, 0.0129, -0.0048, 0.0260, -0.0040, 0.0028, -0.0034, 0.0084, -0.0185], device='cuda:0'), grad: tensor([ 1.5832e-08, 1.4715e-07, 8.2888e-08, 9.3132e-09, 4.4517e-07, 2.7008e-08, -1.3970e-08, -9.4064e-08, 1.8626e-08, -6.3144e-07], device='cuda:0') 100 0.0001 changing lr epoch 365, time 246.30, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4835 re_mapping 0.0030 re_causal 0.0094 /// teacc 99.15 lr 0.00010000 Epoch 367, weight, value: tensor([[-0.2368, -0.0667, 0.2244, ..., 0.0411, -0.2520, -0.1753], [-0.2001, -0.2903, -0.0780, ..., -0.1368, -0.3713, -0.1210], [ 0.3033, -0.0806, -0.3141, ..., -0.2444, -0.2943, -0.2251], ..., [-0.1226, -0.0520, -0.4200, ..., -0.1417, 0.1371, 0.0469], [-0.2399, -0.1600, -0.0694, ..., -0.2528, -0.0671, -0.0805], [-0.1203, -0.1090, -0.1695, ..., 0.1027, 0.0269, -0.1795]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, 7.4506e-09, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.7742e-08, 3.7253e-09, -9.3132e-10], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.1910e-08, 5.5879e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.6787e-07, 1.8440e-07, 0.0000e+00]], device='cuda:0') Epoch 367, bias, value: tensor([ 0.0017, 0.0084, 0.0124, -0.0047, 0.0261, -0.0040, 0.0029, -0.0040, 0.0083, -0.0186], device='cuda:0'), grad: tensor([ 4.6566e-09, 4.5635e-08, 9.3132e-10, 6.5193e-09, -8.7544e-07, 1.8626e-09, 1.3039e-08, 1.1548e-07, 1.1735e-07, 5.7369e-07], device='cuda:0') 100 0.0001 changing lr epoch 366, time 246.61, cls_loss 0.0006 cls_loss_mapping 0.0013 cls_loss_causal 0.4498 re_mapping 0.0030 re_causal 0.0093 /// teacc 99.23 lr 0.00010000 Epoch 368, weight, value: tensor([[-0.2376, -0.0668, 0.2244, ..., 0.0402, -0.2537, -0.1762], [-0.2017, -0.2904, -0.0782, ..., -0.1369, -0.3719, -0.1212], [ 0.3059, -0.0806, -0.3146, ..., -0.2453, -0.2934, -0.2250], ..., [-0.1249, -0.0521, -0.4203, ..., -0.1425, 0.1358, 0.0469], [-0.2406, -0.1599, -0.0700, ..., -0.2526, -0.0673, -0.0809], [-0.1206, -0.1091, -0.1697, ..., 0.1028, 0.0271, -0.1795]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, 1.8626e-09, 9.3132e-10], [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-09, 2.7940e-09], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 9.3132e-10], ..., [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -7.4506e-09, -4.6566e-09], [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 3.7253e-09, 2.7940e-09, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., -8.3819e-09, 2.7940e-09, 1.8626e-09]], device='cuda:0') Epoch 368, bias, value: tensor([ 0.0009, 0.0084, 0.0133, -0.0046, 0.0262, -0.0043, 0.0028, -0.0042, 0.0100, -0.0185], device='cuda:0'), grad: tensor([ 2.6077e-08, -4.0978e-08, 2.7008e-08, -3.3528e-08, 1.9558e-08, -1.1083e-07, 3.0734e-08, -2.7940e-09, 6.5193e-09, 5.9605e-08], device='cuda:0') 100 0.0001 changing lr epoch 367, time 246.58, cls_loss 0.0005 cls_loss_mapping 0.0011 cls_loss_causal 0.4869 re_mapping 0.0031 re_causal 0.0100 /// teacc 99.17 lr 0.00010000 Epoch 369, weight, value: tensor([[-0.2381, -0.0669, 0.2246, ..., 0.0400, -0.2540, -0.1763], [-0.2023, -0.2907, -0.0782, ..., -0.1370, -0.3757, -0.1204], [ 0.3075, -0.0803, -0.3149, ..., -0.2455, -0.2938, -0.2245], ..., [-0.1258, -0.0525, -0.4208, ..., -0.1425, 0.1378, 0.0466], [-0.2441, -0.1602, -0.0702, ..., -0.2528, -0.0674, -0.0812], [-0.1213, -0.1091, -0.1708, ..., 0.1029, 0.0270, -0.1797]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., -1.7695e-08, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 9.3132e-10], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, -9.3132e-10], [ 0.0000e+00, 0.0000e+00, 2.2352e-08, ..., 2.5146e-07, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 369, bias, value: tensor([ 0.0007, 0.0081, 0.0142, -0.0045, 0.0261, -0.0044, 0.0030, -0.0040, 0.0098, -0.0186], device='cuda:0'), grad: tensor([-4.1910e-08, -1.1548e-07, 4.6566e-08, 1.8626e-09, -4.4890e-07, 1.4901e-08, 1.3970e-08, 1.1735e-07, -6.5193e-08, 4.9360e-07], device='cuda:0') 100 0.0001 changing lr epoch 368, time 246.51, cls_loss 0.0005 cls_loss_mapping 0.0016 cls_loss_causal 0.5057 re_mapping 0.0032 re_causal 0.0101 /// teacc 99.14 lr 0.00010000 Epoch 370, weight, value: tensor([[-0.2387, -0.0670, 0.2249, ..., 0.0402, -0.2547, -0.1763], [-0.2035, -0.2909, -0.0782, ..., -0.1370, -0.3766, -0.1221], [ 0.3080, -0.0807, -0.3154, ..., -0.2457, -0.2969, -0.2275], ..., [-0.1258, -0.0529, -0.4214, ..., -0.1426, 0.1396, 0.0496], [-0.2445, -0.1603, -0.0707, ..., -0.2532, -0.0676, -0.0806], [-0.1215, -0.1091, -0.1717, ..., 0.1027, 0.0269, -0.1800]], device='cuda:0'), grad: tensor([[ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 5.5879e-08, 0.0000e+00, -0.0000e+00, ..., 4.6566e-09, 1.8626e-09, 0.0000e+00], [-1.1548e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.2107e-08, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, -0.0000e+00, 0.0000e+00], [ 8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, 9.3132e-10, -0.0000e+00], [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., -2.8871e-08, -3.7253e-09, 0.0000e+00]], device='cuda:0') Epoch 370, bias, value: tensor([ 0.0008, 0.0081, 0.0133, -0.0042, 0.0263, -0.0047, 0.0031, -0.0038, 0.0098, -0.0188], device='cuda:0'), grad: tensor([ 1.3970e-08, 1.2387e-07, -2.4866e-07, 6.0536e-08, 4.5635e-08, 1.6764e-08, 0.0000e+00, 5.8673e-08, 1.3039e-08, -7.6368e-08], device='cuda:0') 100 0.0001 changing lr epoch 369, time 246.47, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4463 re_mapping 0.0031 re_causal 0.0092 /// teacc 99.06 lr 0.00010000 Epoch 371, weight, value: tensor([[-0.2397, -0.0682, 0.2252, ..., 0.0404, -0.2548, -0.1763], [-0.2040, -0.2911, -0.0784, ..., -0.1372, -0.3770, -0.1232], [ 0.3098, -0.0806, -0.3160, ..., -0.2459, -0.2971, -0.2274], ..., [-0.1262, -0.0531, -0.4228, ..., -0.1427, 0.1398, 0.0498], [-0.2477, -0.1589, -0.0713, ..., -0.2535, -0.0677, -0.0809], [-0.1218, -0.1091, -0.1723, ..., 0.1026, 0.0268, -0.1804]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.9849e-09, ..., 3.7253e-09, 4.6566e-10, -0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 3.3993e-08, 4.6566e-10, 0.0000e+00], [-3.3062e-08, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, 0.0000e+00, 0.0000e+00], ..., [ 3.0268e-08, 0.0000e+00, 0.0000e+00, ..., 4.7637e-07, -9.3132e-10, 0.0000e+00], [ 4.6566e-10, -9.3132e-10, 0.0000e+00, ..., 1.5832e-08, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., -1.1111e-06, 4.6566e-10, 0.0000e+00]], device='cuda:0') Epoch 371, bias, value: tensor([ 0.0008, 0.0082, 0.0143, -0.0045, 0.0264, -0.0048, 0.0031, -0.0039, 0.0100, -0.0188], device='cuda:0'), grad: tensor([ 9.3132e-09, 9.0338e-08, -5.2154e-08, 9.7789e-09, 1.7984e-06, -8.3819e-09, 2.7940e-08, 1.3635e-06, 3.4925e-08, -3.2652e-06], device='cuda:0') 100 0.0001 changing lr epoch 370, time 246.58, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4653 re_mapping 0.0030 re_causal 0.0094 /// teacc 99.20 lr 0.00010000 Epoch 372, weight, value: tensor([[-0.2399, -0.0683, 0.2254, ..., 0.0403, -0.2549, -0.1763], [-0.2045, -0.2913, -0.0785, ..., -0.1373, -0.3788, -0.1239], [ 0.3101, -0.0804, -0.3164, ..., -0.2462, -0.2974, -0.2275], ..., [-0.1263, -0.0532, -0.4233, ..., -0.1430, 0.1405, 0.0499], [-0.2478, -0.1592, -0.0717, ..., -0.2542, -0.0678, -0.0810], [-0.1220, -0.1091, -0.1725, ..., 0.1025, 0.0268, -0.1805]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.4506e-09, ..., -3.2596e-09, 6.9849e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 9.3132e-10, 1.3504e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 5.1223e-09, ..., 9.3132e-10, 1.3970e-09, 0.0000e+00], ..., [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 5.1223e-09, -2.9337e-08, 0.0000e+00], [ 4.6566e-10, -0.0000e+00, 1.5367e-08, ..., 1.3970e-09, 6.9849e-09, 4.6566e-10], [ 0.0000e+00, -4.6566e-10, 6.9849e-09, ..., -1.2573e-08, 6.5193e-09, 0.0000e+00]], device='cuda:0') Epoch 372, bias, value: tensor([ 0.0007, 0.0081, 0.0134, -0.0045, 0.0265, -0.0048, 0.0030, -0.0039, 0.0107, -0.0189], device='cuda:0'), grad: tensor([ 4.9826e-08, 5.7276e-08, 1.7229e-08, 2.3283e-08, 3.7579e-07, 3.9116e-08, -5.4436e-07, -9.3598e-08, 6.8918e-08, 1.8626e-08], device='cuda:0') 100 0.0001 changing lr epoch 371, time 246.36, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4545 re_mapping 0.0030 re_causal 0.0098 /// teacc 99.13 lr 0.00010000 Epoch 373, weight, value: tensor([[-0.2401, -0.0687, 0.2255, ..., 0.0403, -0.2550, -0.1763], [-0.2046, -0.2915, -0.0785, ..., -0.1374, -0.3796, -0.1247], [ 0.3105, -0.0796, -0.3168, ..., -0.2466, -0.2975, -0.2275], ..., [-0.1264, -0.0534, -0.4235, ..., -0.1431, 0.1409, 0.0500], [-0.2483, -0.1604, -0.0721, ..., -0.2549, -0.0680, -0.0811], [-0.1227, -0.1090, -0.1729, ..., 0.1025, 0.0268, -0.1806]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 9.3132e-10, -1.8626e-09, ..., 6.0070e-08, 2.3283e-09, 0.0000e+00], [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., 2.9337e-08, 1.8626e-09, 4.6566e-10], [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 3.7253e-09, 4.6566e-10, 0.0000e+00], ..., [ 4.6566e-10, 1.8626e-09, 0.0000e+00, ..., 2.2072e-07, 9.7789e-09, 4.6566e-10], [ 9.3132e-10, 1.0710e-08, 9.3132e-10, ..., 9.1270e-08, 9.3132e-10, 1.3970e-09], [ 4.6566e-10, -1.7229e-08, 4.1910e-09, ..., -6.2911e-07, -1.7229e-08, 1.3970e-09]], device='cuda:0') Epoch 373, bias, value: tensor([ 0.0008, 0.0081, 0.0135, -0.0045, 0.0265, -0.0048, 0.0030, -0.0038, 0.0105, -0.0190], device='cuda:0'), grad: tensor([ 2.2398e-07, 2.1933e-07, 3.0547e-07, -5.7705e-06, 6.4587e-07, 5.6922e-06, 4.6566e-09, 3.5390e-07, 3.3993e-07, -2.0135e-06], device='cuda:0') 100 0.0001 changing lr epoch 372, time 246.35, cls_loss 0.0006 cls_loss_mapping 0.0021 cls_loss_causal 0.4791 re_mapping 0.0031 re_causal 0.0094 /// teacc 99.11 lr 0.00010000 Epoch 374, weight, value: tensor([[-0.2404, -0.0691, 0.2267, ..., 0.0406, -0.2551, -0.1763], [-0.2051, -0.2924, -0.0799, ..., -0.1377, -0.3835, -0.1229], [ 0.3110, -0.0802, -0.3178, ..., -0.2471, -0.2985, -0.2280], ..., [-0.1267, -0.0531, -0.4245, ..., -0.1432, 0.1434, 0.0506], [-0.2485, -0.1607, -0.0735, ..., -0.2561, -0.0682, -0.0817], [-0.1228, -0.1089, -0.1737, ..., 0.1015, 0.0266, -0.1807]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -9.1363e-07, ..., -5.4063e-07, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 4.6566e-10, 4.6566e-10, ..., 3.0268e-08, 9.3132e-09, 9.3132e-10], [-3.7253e-09, 0.0000e+00, 3.7253e-09, ..., 3.7253e-09, 2.3283e-09, 0.0000e+00], ..., [ 2.7940e-09, 4.6566e-10, 9.3132e-10, ..., 1.3039e-08, -2.1886e-08, -1.8626e-09], [ 1.8626e-09, 0.0000e+00, 2.5611e-08, ..., 2.6543e-08, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 8.7498e-07, ..., 1.9856e-06, 6.5193e-09, 0.0000e+00]], device='cuda:0') Epoch 374, bias, value: tensor([ 0.0018, 0.0073, 0.0133, -0.0044, 0.0274, -0.0047, 0.0026, -0.0031, 0.0102, -0.0198], device='cuda:0'), grad: tensor([-1.6857e-06, 8.4750e-08, 1.9092e-08, 1.5832e-08, -2.5630e-06, 2.4633e-07, 8.4285e-08, -4.6566e-08, -2.9430e-07, 4.1351e-06], device='cuda:0') 100 0.0001 changing lr epoch 373, time 246.19, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4709 re_mapping 0.0030 re_causal 0.0093 /// teacc 99.12 lr 0.00010000 Epoch 375, weight, value: tensor([[-0.2406, -0.0692, 0.2271, ..., 0.0408, -0.2554, -0.1763], [-0.2057, -0.2929, -0.0799, ..., -0.1378, -0.3837, -0.1231], [ 0.3115, -0.0807, -0.3185, ..., -0.2475, -0.2991, -0.2284], ..., [-0.1269, -0.0524, -0.4258, ..., -0.1434, 0.1440, 0.0511], [-0.2487, -0.1611, -0.0754, ..., -0.2572, -0.0685, -0.0819], [-0.1239, -0.1086, -0.1745, ..., 0.1014, 0.0265, -0.1810]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.9828e-06, ..., -1.0841e-06, 4.6566e-10, -1.0245e-08], [ 0.0000e+00, 0.0000e+00, 1.2573e-08, ..., 8.3819e-09, 4.6566e-10, 0.0000e+00], [-2.3283e-09, 0.0000e+00, 1.8626e-09, ..., 4.6566e-10, 4.6566e-10, 0.0000e+00], ..., [ 1.3970e-09, 0.0000e+00, 3.7253e-09, ..., 2.1886e-08, -4.6566e-10, 0.0000e+00], [ 4.6566e-10, 1.8626e-09, 2.4214e-08, ..., 2.3283e-08, 1.8626e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 5.8673e-08, ..., -2.1560e-07, -5.1223e-09, 0.0000e+00]], device='cuda:0') Epoch 375, bias, value: tensor([ 0.0021, 0.0072, 0.0133, -0.0044, 0.0275, -0.0046, 0.0023, -0.0029, 0.0100, -0.0199], device='cuda:0'), grad: tensor([-1.3568e-05, 3.3993e-08, 1.2573e-08, 1.2200e-07, 4.1164e-07, 1.7788e-06, 1.1414e-05, 6.6590e-08, 8.1491e-08, -3.7206e-07], device='cuda:0') 100 0.0001 changing lr epoch 374, time 246.30, cls_loss 0.0006 cls_loss_mapping 0.0011 cls_loss_causal 0.4551 re_mapping 0.0028 re_causal 0.0089 /// teacc 99.10 lr 0.00010000 Epoch 376, weight, value: tensor([[-0.2405, -0.0693, 0.2301, ..., 0.0413, -0.2552, -0.1764], [-0.2062, -0.2934, -0.0800, ..., -0.1380, -0.3839, -0.1233], [ 0.3120, -0.0812, -0.3211, ..., -0.2485, -0.2994, -0.2284], ..., [-0.1271, -0.0523, -0.4269, ..., -0.1435, 0.1442, 0.0511], [-0.2487, -0.1611, -0.0768, ..., -0.2576, -0.0688, -0.0822], [-0.1261, -0.1100, -0.1749, ..., 0.1011, 0.0265, -0.1810]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, 9.3132e-10, 4.6566e-10], [ 4.6566e-10, 0.0000e+00, -0.0000e+00, ..., 3.2596e-09, 7.4506e-09, 2.3283e-09], [-7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.7940e-09, 9.3132e-10], ..., [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.6764e-08, -1.2573e-08, -4.6566e-09], [ 2.3283e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 1.8626e-09, 1.3970e-09], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -3.8650e-08, -1.8626e-09, 9.3132e-10]], device='cuda:0') Epoch 376, bias, value: tensor([ 0.0050, 0.0066, 0.0132, -0.0044, 0.0278, -0.0046, 0.0011, -0.0023, 0.0099, -0.0201], device='cuda:0'), grad: tensor([ 1.3039e-08, -7.3109e-08, 1.3504e-08, 3.8184e-08, 7.9162e-08, -1.2061e-07, 1.0896e-07, 4.6100e-08, 2.9337e-08, -1.1409e-07], device='cuda:0') 100 0.0001 changing lr epoch 375, time 246.39, cls_loss 0.0004 cls_loss_mapping 0.0010 cls_loss_causal 0.4495 re_mapping 0.0028 re_causal 0.0093 /// teacc 99.12 lr 0.00010000 Epoch 377, weight, value: tensor([[-0.2407, -0.0695, 0.2332, ..., 0.0386, -0.2555, -0.1764], [-0.2064, -0.2937, -0.0803, ..., -0.1383, -0.3840, -0.1236], [ 0.3121, -0.0813, -0.3219, ..., -0.2488, -0.2995, -0.2285], ..., [-0.1272, -0.0527, -0.4274, ..., -0.1437, 0.1444, 0.0512], [-0.2488, -0.1612, -0.0770, ..., -0.2577, -0.0691, -0.0823], [-0.1262, -0.1104, -0.1752, ..., 0.1019, 0.0265, -0.1811]], device='cuda:0'), grad: tensor([[-0.0000e+00, 0.0000e+00, -4.6566e-10, ..., 4.6566e-10, 4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., 1.3970e-09, 4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, -2.7940e-09, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 1.8859e-07, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 377, bias, value: tensor([ 0.0059, 0.0065, 0.0131, -0.0041, 0.0278, -0.0048, -0.0011, -0.0023, 0.0099, -0.0196], device='cuda:0'), grad: tensor([ 2.3283e-09, -3.7253e-08, 3.2596e-09, 4.6566e-09, -4.0419e-07, 4.3772e-08, -3.4459e-08, 2.1420e-08, 3.7253e-09, 4.0093e-07], device='cuda:0') 100 0.0001 changing lr epoch 376, time 246.59, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4695 re_mapping 0.0027 re_causal 0.0089 /// teacc 99.11 lr 0.00010000 Epoch 378, weight, value: tensor([[-0.2410, -0.0696, 0.2332, ..., 0.0386, -0.2557, -0.1764], [-0.2068, -0.2941, -0.0804, ..., -0.1392, -0.3841, -0.1237], [ 0.3125, -0.0810, -0.3223, ..., -0.2491, -0.2996, -0.2285], ..., [-0.1274, -0.0531, -0.4280, ..., -0.1439, 0.1446, 0.0512], [-0.2490, -0.1615, -0.0776, ..., -0.2585, -0.0696, -0.0825], [-0.1276, -0.1107, -0.1755, ..., 0.1008, 0.0262, -0.1811]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.3027e-07, ..., 0.0000e+00, 4.1910e-09, -6.5658e-08], [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, 2.7940e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-10, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, -5.5879e-09, 0.0000e+00], [ 0.0000e+00, -4.6566e-10, 1.8626e-09, ..., 9.3132e-10, 9.3132e-10, 4.6566e-10], [ 0.0000e+00, 4.6566e-10, 5.1223e-09, ..., 1.3737e-07, 4.6566e-10, 9.3132e-10]], device='cuda:0') Epoch 378, bias, value: tensor([ 0.0059, 0.0064, 0.0133, -0.0042, 0.0290, -0.0044, -0.0013, -0.0024, 0.0097, -0.0206], device='cuda:0'), grad: tensor([-8.1351e-07, 3.1665e-08, 1.0710e-08, 4.1910e-09, -2.7334e-07, 8.0047e-07, 1.8161e-08, -6.5193e-08, -5.7742e-08, 3.4552e-07], device='cuda:0') 100 0.0001 changing lr epoch 377, time 246.48, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4634 re_mapping 0.0028 re_causal 0.0089 /// teacc 99.13 lr 0.00010000 Epoch 379, weight, value: tensor([[-0.2412, -0.0699, 0.2337, ..., 0.0384, -0.2555, -0.1757], [-0.2073, -0.2946, -0.0805, ..., -0.1401, -0.3843, -0.1238], [ 0.3129, -0.0810, -0.3232, ..., -0.2496, -0.2997, -0.2285], ..., [-0.1275, -0.0533, -0.4305, ..., -0.1441, 0.1448, 0.0512], [-0.2491, -0.1617, -0.0784, ..., -0.2590, -0.0697, -0.0826], [-0.1278, -0.1107, -0.1768, ..., 0.1008, 0.0262, -0.1812]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -2.3749e-08, ..., -7.9162e-09, 0.0000e+00, 0.0000e+00], [ 3.2596e-09, 0.0000e+00, 9.3132e-10, ..., 2.7940e-09, 4.6566e-10, 0.0000e+00], [-8.5682e-08, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], ..., [ 8.1025e-08, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, 0.0000e+00, 0.0000e+00], [ 1.3970e-09, 0.0000e+00, 7.9162e-09, ..., 5.5879e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., -1.5367e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 379, bias, value: tensor([ 0.0060, 0.0057, 0.0133, -0.0032, 0.0291, -0.0054, -0.0013, -0.0018, 0.0096, -0.0206], device='cuda:0'), grad: tensor([-4.1444e-08, -3.3993e-08, -2.1700e-07, 4.6566e-09, 2.9337e-08, 2.3283e-09, 1.8626e-08, 2.6776e-07, 2.5611e-08, -3.9581e-08], device='cuda:0') 100 0.0001 changing lr epoch 378, time 246.63, cls_loss 0.0005 cls_loss_mapping 0.0009 cls_loss_causal 0.4514 re_mapping 0.0028 re_causal 0.0089 /// teacc 99.12 lr 0.00010000 Epoch 380, weight, value: tensor([[-0.2414, -0.0700, 0.2337, ..., 0.0384, -0.2556, -0.1756], [-0.2072, -0.2948, -0.0803, ..., -0.1405, -0.3846, -0.1239], [ 0.3131, -0.0811, -0.3236, ..., -0.2502, -0.2998, -0.2285], ..., [-0.1277, -0.0536, -0.4322, ..., -0.1443, 0.1452, 0.0513], [-0.2493, -0.1618, -0.0791, ..., -0.2611, -0.0702, -0.0827], [-0.1278, -0.1106, -0.1770, ..., 0.1007, 0.0262, -0.1815]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.3132e-10, ..., -1.8626e-09, 2.7940e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-09, 3.7253e-09], [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 9.3132e-10], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -1.9558e-08, -8.3819e-09], [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -5.5879e-09, 3.7253e-09, 1.8626e-09]], device='cuda:0') Epoch 380, bias, value: tensor([ 0.0061, 0.0058, 0.0126, -0.0031, 0.0293, -0.0055, -0.0013, -0.0018, 0.0093, -0.0208], device='cuda:0'), grad: tensor([ 8.3819e-09, -2.6822e-07, 1.3970e-08, 2.1420e-08, 2.6077e-08, 2.1420e-08, -6.5193e-09, 1.0617e-07, 3.6322e-08, 4.1910e-08], device='cuda:0') 100 0.0001 changing lr epoch 379, time 246.45, cls_loss 0.0006 cls_loss_mapping 0.0012 cls_loss_causal 0.4482 re_mapping 0.0029 re_causal 0.0089 /// teacc 99.15 lr 0.00010000 Epoch 381, weight, value: tensor([[-0.2420, -0.0703, 0.2338, ..., 0.0384, -0.2559, -0.1756], [-0.2092, -0.2952, -0.0806, ..., -0.1426, -0.3858, -0.1273], [ 0.3140, -0.0806, -0.3240, ..., -0.2504, -0.3000, -0.2286], ..., [-0.1280, -0.0540, -0.4326, ..., -0.1443, 0.1470, 0.0523], [-0.2498, -0.1625, -0.0794, ..., -0.2619, -0.0707, -0.0839], [-0.1280, -0.1108, -0.1774, ..., 0.1006, 0.0262, -0.1825]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.4901e-08, ..., -3.7253e-09, 9.3132e-10, -9.3132e-10], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., -9.3132e-10, 5.5879e-09, 9.3132e-10], [-2.7940e-09, -0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-09, 0.0000e+00], ..., [ 2.7940e-09, 0.0000e+00, 5.5879e-09, ..., 2.7940e-09, -3.0734e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -4.6566e-09, 1.8626e-09, 0.0000e+00]], device='cuda:0') Epoch 381, bias, value: tensor([ 0.0061, 0.0054, 0.0129, -0.0032, 0.0296, -0.0054, -0.0011, -0.0016, 0.0089, -0.0210], device='cuda:0'), grad: tensor([-1.0245e-08, -1.6857e-07, 2.0489e-08, 2.2352e-08, 8.1956e-08, 2.2352e-08, 3.1665e-08, -4.1910e-08, 4.9360e-08, -9.3132e-09], device='cuda:0') 100 0.0001 changing lr epoch 380, time 246.13, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4206 re_mapping 0.0029 re_causal 0.0089 /// teacc 99.17 lr 0.00010000 Epoch 382, weight, value: tensor([[-0.2425, -0.0707, 0.2338, ..., 0.0384, -0.2562, -0.1756], [-0.2096, -0.2956, -0.0806, ..., -0.1427, -0.3859, -0.1273], [ 0.3145, -0.0805, -0.3241, ..., -0.2505, -0.3000, -0.2285], ..., [-0.1285, -0.0547, -0.4329, ..., -0.1444, 0.1474, 0.0524], [-0.2498, -0.1627, -0.0797, ..., -0.2620, -0.0708, -0.0840], [-0.1281, -0.1108, -0.1776, ..., 0.1005, 0.0261, -0.1826]], device='cuda:0'), grad: tensor([[ 2.7940e-09, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, 2.7940e-09, 0.0000e+00], [ 8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [-8.5682e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 4.4703e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 9.3132e-10, 0.0000e+00], [ 1.0245e-08, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, 8.3819e-09, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., -0.0000e+00, 2.7940e-09, 0.0000e+00]], device='cuda:0') Epoch 382, bias, value: tensor([ 0.0060, 0.0055, 0.0128, -0.0028, 0.0297, -0.0056, -0.0011, -0.0017, 0.0088, -0.0211], device='cuda:0'), grad: tensor([ 1.7695e-08, 6.5193e-09, -1.8720e-07, 8.6613e-08, 1.5832e-08, -2.0117e-07, 5.9605e-08, 1.1548e-07, 7.5437e-08, 1.6764e-08], device='cuda:0') 100 0.0001 changing lr epoch 381, time 245.86, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4788 re_mapping 0.0031 re_causal 0.0096 /// teacc 99.10 lr 0.00010000 Epoch 383, weight, value: tensor([[-0.2433, -0.0709, 0.2338, ..., 0.0384, -0.2567, -0.1757], [-0.2097, -0.2959, -0.0806, ..., -0.1428, -0.3861, -0.1277], [ 0.3149, -0.0806, -0.3244, ..., -0.2508, -0.3001, -0.2285], ..., [-0.1287, -0.0552, -0.4335, ..., -0.1446, 0.1477, 0.0523], [-0.2500, -0.1628, -0.0800, ..., -0.2627, -0.0706, -0.0843], [-0.1282, -0.1108, -0.1778, ..., 0.1005, 0.0262, -0.1827]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 9.3132e-10, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, -2.7940e-09, -9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -6.5193e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 383, bias, value: tensor([ 0.0060, 0.0054, 0.0119, -0.0029, 0.0297, -0.0054, -0.0011, -0.0015, 0.0090, -0.0211], device='cuda:0'), grad: tensor([ 2.7940e-08, 3.5204e-07, 3.9116e-08, 6.5193e-09, 2.7008e-08, 4.6566e-09, 9.3132e-10, -4.2096e-07, 3.7253e-09, -2.5146e-08], device='cuda:0') 100 0.0001 changing lr epoch 382, time 245.85, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4317 re_mapping 0.0030 re_causal 0.0093 /// teacc 99.20 lr 0.00010000 Epoch 384, weight, value: tensor([[-0.2439, -0.0711, 0.2338, ..., 0.0384, -0.2572, -0.1757], [-0.2137, -0.2967, -0.0811, ..., -0.1431, -0.3863, -0.1283], [ 0.3174, -0.0805, -0.3251, ..., -0.2544, -0.3002, -0.2289], ..., [-0.1291, -0.0554, -0.4343, ..., -0.1447, 0.1480, 0.0529], [-0.2501, -0.1631, -0.0809, ..., -0.2639, -0.0707, -0.0845], [-0.1286, -0.1109, -0.1781, ..., 0.1004, 0.0262, -0.1828]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, -0.0000e+00, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 4.0047e-08, ..., 1.3970e-08, 1.1176e-08, 2.7940e-09], [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., -1.9558e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 384, bias, value: tensor([ 0.0059, 0.0050, 0.0128, -0.0031, 0.0298, -0.0054, -0.0008, -0.0014, 0.0090, -0.0212], device='cuda:0'), grad: tensor([ 2.4214e-08, 3.7253e-09, 1.8626e-08, 2.5444e-06, 6.5193e-09, -3.2075e-06, 5.2992e-07, 1.0245e-08, 8.2888e-08, -1.1176e-08], device='cuda:0') 100 0.0001 changing lr epoch 383, time 246.00, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4789 re_mapping 0.0028 re_causal 0.0093 /// teacc 99.17 lr 0.00010000 Epoch 385, weight, value: tensor([[-0.2442, -0.0712, 0.2338, ..., 0.0384, -0.2573, -0.1757], [-0.2144, -0.2972, -0.0805, ..., -0.1431, -0.3864, -0.1283], [ 0.3183, -0.0807, -0.3257, ..., -0.2546, -0.3002, -0.2289], ..., [-0.1296, -0.0560, -0.4358, ..., -0.1452, 0.1478, 0.0529], [-0.2505, -0.1633, -0.0818, ..., -0.2649, -0.0707, -0.0846], [-0.1287, -0.1109, -0.1785, ..., 0.1005, 0.0262, -0.1828]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 385, bias, value: tensor([ 0.0059, 0.0055, 0.0129, -0.0029, 0.0297, -0.0055, -0.0007, -0.0019, 0.0091, -0.0211], device='cuda:0'), grad: tensor([ 9.3132e-10, 6.5193e-09, -5.5879e-09, 1.3970e-08, -1.5832e-08, -1.1176e-08, 2.0862e-07, 9.3132e-09, -2.1420e-07, 1.3970e-08], device='cuda:0') 100 0.0001 changing lr epoch 384, time 246.00, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4564 re_mapping 0.0029 re_causal 0.0090 /// teacc 99.09 lr 0.00010000 Epoch 386, weight, value: tensor([[-0.2445, -0.0713, 0.2338, ..., 0.0384, -0.2576, -0.1757], [-0.2145, -0.2975, -0.0801, ..., -0.1432, -0.3866, -0.1292], [ 0.3186, -0.0811, -0.3263, ..., -0.2547, -0.3003, -0.2290], ..., [-0.1297, -0.0563, -0.4366, ..., -0.1453, 0.1481, 0.0532], [-0.2507, -0.1633, -0.0825, ..., -0.2649, -0.0707, -0.0848], [-0.1304, -0.1109, -0.1787, ..., 0.1004, 0.0261, -0.1832]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 9.3132e-09, 0.0000e+00, -9.3132e-10, ..., -6.5193e-09, 6.5193e-09, -0.0000e+00], [-2.1420e-08, -0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, -0.0000e+00], ..., [ 5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, -1.6764e-08, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 3.7253e-09, 0.0000e+00, 9.3132e-10, ..., -6.5193e-09, 5.5879e-09, 0.0000e+00]], device='cuda:0') Epoch 386, bias, value: tensor([ 0.0058, 0.0058, 0.0128, -0.0036, 0.0298, -0.0047, -0.0007, -0.0022, 0.0094, -0.0213], device='cuda:0'), grad: tensor([ 3.1665e-08, -1.0058e-07, -1.7695e-08, 9.9652e-08, 1.4994e-07, -1.3132e-07, 7.5437e-08, -3.6880e-07, 2.2352e-08, 2.3935e-07], device='cuda:0') 100 0.0001 changing lr epoch 385, time 245.89, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4538 re_mapping 0.0029 re_causal 0.0090 /// teacc 99.16 lr 0.00010000 Epoch 387, weight, value: tensor([[-0.2455, -0.0714, 0.2339, ..., 0.0384, -0.2580, -0.1758], [-0.2147, -0.2976, -0.0785, ..., -0.1430, -0.3868, -0.1280], [ 0.3189, -0.0811, -0.3269, ..., -0.2552, -0.3004, -0.2290], ..., [-0.1298, -0.0565, -0.4375, ..., -0.1461, 0.1483, 0.0534], [-0.2508, -0.1634, -0.0839, ..., -0.2654, -0.0709, -0.0849], [-0.1306, -0.1109, -0.1792, ..., 0.1007, 0.0262, -0.1836]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., -1.3970e-08, 0.0000e+00, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -4.6566e-09, -9.3132e-10, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 4.0978e-08], [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 2.7940e-09, 0.0000e+00, 1.3970e-08]], device='cuda:0') Epoch 387, bias, value: tensor([ 0.0058, 0.0060, 0.0126, -0.0037, 0.0295, -0.0048, -0.0010, -0.0023, 0.0095, -0.0211], device='cuda:0'), grad: tensor([-4.1910e-08, 6.3330e-08, 2.2352e-08, 3.5390e-08, 1.6764e-08, -5.9418e-07, 1.2573e-07, -1.9092e-07, 3.4366e-07, 2.1979e-07], device='cuda:0') 100 0.0001 changing lr epoch 386, time 246.17, cls_loss 0.0005 cls_loss_mapping 0.0012 cls_loss_causal 0.4660 re_mapping 0.0028 re_causal 0.0090 /// teacc 99.08 lr 0.00010000 Epoch 388, weight, value: tensor([[-0.2459, -0.0714, 0.2339, ..., 0.0385, -0.2584, -0.1760], [-0.2151, -0.2977, -0.0786, ..., -0.1430, -0.3870, -0.1284], [ 0.3196, -0.0811, -0.3275, ..., -0.2554, -0.3004, -0.2291], ..., [-0.1303, -0.0567, -0.4378, ..., -0.1463, 0.1486, 0.0537], [-0.2510, -0.1634, -0.0843, ..., -0.2656, -0.0710, -0.0853], [-0.1311, -0.1109, -0.1794, ..., 0.1007, 0.0261, -0.1838]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -0.0000e+00, 0.0000e+00, 0.0000e+00], [-1.8626e-09, -0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 388, bias, value: tensor([ 0.0057, 0.0052, 0.0124, -0.0036, 0.0295, -0.0048, -0.0010, -0.0014, 0.0101, -0.0211], device='cuda:0'), grad: tensor([ 2.7940e-09, -1.4901e-08, 1.8626e-09, 1.8626e-09, -2.0489e-08, -3.7253e-09, 4.6566e-09, 1.3970e-08, 0.0000e+00, 1.3970e-08], device='cuda:0') 100 0.0001 changing lr epoch 387, time 246.15, cls_loss 0.0004 cls_loss_mapping 0.0010 cls_loss_causal 0.4448 re_mapping 0.0029 re_causal 0.0093 /// teacc 99.04 lr 0.00010000 Epoch 389, weight, value: tensor([[-0.2471, -0.0715, 0.2352, ..., 0.0395, -0.2585, -0.1759], [-0.2155, -0.2978, -0.0786, ..., -0.1433, -0.3870, -0.1286], [ 0.3204, -0.0811, -0.3288, ..., -0.2557, -0.3004, -0.2291], ..., [-0.1308, -0.0574, -0.4382, ..., -0.1468, 0.1486, 0.0538], [-0.2511, -0.1635, -0.0846, ..., -0.2667, -0.0711, -0.0853], [-0.1313, -0.1110, -0.1818, ..., 0.1003, 0.0262, -0.1839]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.7229e-07, ..., -1.1362e-07, 1.8626e-08, 5.5879e-09], [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 5.5879e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 5.5879e-09, ..., 4.6566e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.4901e-08, ..., 9.3132e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.1921e-07, ..., 6.0536e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 389, bias, value: tensor([ 0.0068, 0.0051, 0.0128, -0.0036, 0.0294, -0.0048, -0.0011, -0.0014, 0.0102, -0.0214], device='cuda:0'), grad: tensor([-4.0047e-07, 1.6764e-08, 8.3819e-09, 4.5635e-08, 1.5832e-08, -3.9116e-08, 2.7940e-09, 2.7008e-08, 4.3772e-08, 2.8592e-07], device='cuda:0') 100 0.0001 changing lr epoch 388, time 246.05, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4711 re_mapping 0.0028 re_causal 0.0089 /// teacc 99.08 lr 0.00010000 Epoch 390, weight, value: tensor([[-0.2477, -0.0716, 0.2365, ..., 0.0409, -0.2595, -0.1760], [-0.2157, -0.2981, -0.0786, ..., -0.1435, -0.3871, -0.1286], [ 0.3211, -0.0810, -0.3299, ..., -0.2561, -0.3005, -0.2291], ..., [-0.1314, -0.0583, -0.4388, ..., -0.1472, 0.1487, 0.0539], [-0.2512, -0.1636, -0.0849, ..., -0.2679, -0.0712, -0.0859], [-0.1325, -0.1109, -0.1835, ..., 0.0998, 0.0262, -0.1834]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 9.3132e-10], [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 9.3132e-10, 0.0000e+00], [-5.7742e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.7695e-08, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, -0.0000e+00, 9.3132e-10], [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.4214e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 390, bias, value: tensor([ 0.0082, 0.0049, 0.0131, -0.0039, 0.0294, -0.0046, -0.0011, -0.0012, 0.0101, -0.0220], device='cuda:0'), grad: tensor([ 1.2107e-08, -1.1176e-08, -1.1269e-07, 7.5437e-08, 6.2399e-08, -3.9116e-08, 7.4506e-09, 4.3772e-08, 9.3132e-09, -4.3772e-08], device='cuda:0') 100 0.0001 changing lr epoch 389, time 246.21, cls_loss 0.0004 cls_loss_mapping 0.0016 cls_loss_causal 0.4525 re_mapping 0.0030 re_causal 0.0094 /// teacc 99.08 lr 0.00010000 Epoch 391, weight, value: tensor([[-0.2500, -0.0717, 0.2364, ..., 0.0408, -0.2600, -0.1770], [-0.2159, -0.2982, -0.0785, ..., -0.1436, -0.3872, -0.1288], [ 0.3215, -0.0812, -0.3302, ..., -0.2564, -0.3005, -0.2291], ..., [-0.1317, -0.0584, -0.4393, ..., -0.1473, 0.1488, 0.0538], [-0.2512, -0.1636, -0.0848, ..., -0.2680, -0.0709, -0.0860], [-0.1329, -0.1109, -0.1835, ..., 0.0997, 0.0262, -0.1838]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.9558e-08, ..., -2.7940e-09, 0.0000e+00, -3.7253e-09], [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 6.5193e-09, 9.3132e-10, 9.3132e-10], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 2.7940e-09, 9.3132e-10, 3.7253e-09, ..., 5.5879e-09, -9.3132e-10, 1.8626e-09], [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., 4.0047e-08, 0.0000e+00, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., 2.6077e-08, 0.0000e+00, 9.3132e-10]], device='cuda:0') Epoch 391, bias, value: tensor([ 0.0079, 0.0049, 0.0133, -0.0040, 0.0296, -0.0048, -0.0008, -0.0013, 0.0107, -0.0221], device='cuda:0'), grad: tensor([-3.3528e-08, -2.0415e-05, 1.0617e-07, 9.3132e-09, -3.3714e-07, 2.4214e-08, 1.2107e-08, 2.0280e-05, 1.7881e-07, 1.4435e-07], device='cuda:0') 100 0.0001 changing lr epoch 390, time 246.29, cls_loss 0.0013 cls_loss_mapping 0.0015 cls_loss_causal 0.4735 re_mapping 0.0028 re_causal 0.0087 /// teacc 99.13 lr 0.00010000 Epoch 392, weight, value: tensor([[-0.2519, -0.0717, 0.2338, ..., 0.0360, -0.2606, -0.1787], [-0.2159, -0.2983, -0.0789, ..., -0.1438, -0.3877, -0.1298], [ 0.3218, -0.0813, -0.3307, ..., -0.2569, -0.3007, -0.2295], ..., [-0.1319, -0.0586, -0.4400, ..., -0.1475, 0.1495, 0.0553], [-0.2513, -0.1636, -0.0852, ..., -0.2687, -0.0710, -0.0867], [-0.1332, -0.1110, -0.1808, ..., 0.1043, 0.0267, -0.1855]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.4459e-08, ..., 1.9558e-08, 0.0000e+00, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, 9.3132e-10, 2.7940e-09], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.1420e-08, 1.8626e-09, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 1.3039e-08, 2.3283e-08], [ 0.0000e+00, 0.0000e+00, 3.0734e-08, ..., -5.8673e-08, 0.0000e+00, 2.7940e-09]], device='cuda:0') Epoch 392, bias, value: tensor([ 0.0034, 0.0050, 0.0132, -0.0042, 0.0286, -0.0045, -0.0006, -0.0014, 0.0108, -0.0179], device='cuda:0'), grad: tensor([ 2.4214e-08, 2.7008e-08, 1.3039e-08, 2.1607e-07, -7.4506e-08, -8.4005e-07, 5.6531e-07, 8.0094e-08, 3.3528e-08, -4.6566e-08], device='cuda:0') 100 0.0001 changing lr epoch 391, time 246.29, cls_loss 0.0004 cls_loss_mapping 0.0011 cls_loss_causal 0.4634 re_mapping 0.0031 re_causal 0.0099 /// teacc 99.11 lr 0.00010000 Epoch 393, weight, value: tensor([[-0.2527, -0.0717, 0.2337, ..., 0.0357, -0.2607, -0.1786], [-0.2160, -0.2984, -0.0789, ..., -0.1440, -0.3878, -0.1300], [ 0.3220, -0.0814, -0.3317, ..., -0.2574, -0.3008, -0.2297], ..., [-0.1322, -0.0587, -0.4403, ..., -0.1478, 0.1498, 0.0558], [-0.2515, -0.1637, -0.0861, ..., -0.2691, -0.0711, -0.0877], [-0.1333, -0.1110, -0.1806, ..., 0.1046, 0.0267, -0.1869]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -5.5879e-08, ..., 5.3085e-08, 0.0000e+00, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, 2.6077e-08, 1.8626e-09], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, -3.2596e-08, -1.8626e-09], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, 3.7253e-09, 4.6566e-09], [ 2.7940e-09, 0.0000e+00, 5.4017e-08, ..., -2.3004e-07, 1.8626e-09, 2.7940e-09]], device='cuda:0') Epoch 393, bias, value: tensor([ 0.0031, 0.0050, 0.0131, -0.0041, 0.0286, -0.0046, -0.0006, -0.0014, 0.0107, -0.0176], device='cuda:0'), grad: tensor([ 7.3574e-08, 1.2573e-07, 1.0245e-08, 1.7136e-07, 4.7125e-07, -3.4738e-07, 1.2480e-07, -1.4901e-07, 6.0536e-08, -5.3830e-07], device='cuda:0') 100 0.0001 changing lr epoch 392, time 246.44, cls_loss 0.0004 cls_loss_mapping 0.0010 cls_loss_causal 0.4572 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.14 lr 0.00010000 Epoch 394, weight, value: tensor([[-0.2538, -0.0719, 0.2337, ..., 0.0357, -0.2611, -0.1786], [-0.2161, -0.2989, -0.0788, ..., -0.1443, -0.3881, -0.1301], [ 0.3226, -0.0814, -0.3335, ..., -0.2584, -0.3009, -0.2298], ..., [-0.1325, -0.0591, -0.4406, ..., -0.1482, 0.1500, 0.0560], [-0.2521, -0.1638, -0.0862, ..., -0.2697, -0.0712, -0.0879], [-0.1335, -0.1110, -0.1806, ..., 0.1046, 0.0267, -0.1873]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.7695e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 394, bias, value: tensor([ 0.0031, 0.0050, 0.0132, -0.0038, 0.0287, -0.0047, -0.0005, -0.0015, 0.0106, -0.0176], device='cuda:0'), grad: tensor([ 4.6566e-09, -5.7742e-08, 7.4506e-09, 2.9802e-08, -6.7987e-08, -1.3039e-08, -1.5832e-08, 4.0978e-08, -1.0245e-08, 8.1025e-08], device='cuda:0') 100 0.0001 changing lr epoch 393, time 246.38, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4630 re_mapping 0.0029 re_causal 0.0087 /// teacc 99.12 lr 0.00010000 Epoch 395, weight, value: tensor([[-0.2545, -0.0721, 0.2337, ..., 0.0357, -0.2616, -0.1784], [-0.2163, -0.2993, -0.0790, ..., -0.1448, -0.3886, -0.1307], [ 0.3235, -0.0833, -0.3353, ..., -0.2593, -0.3010, -0.2287], ..., [-0.1332, -0.0597, -0.4411, ..., -0.1490, 0.1508, 0.0564], [-0.2525, -0.1638, -0.0872, ..., -0.2700, -0.0716, -0.0877], [-0.1337, -0.1110, -0.1807, ..., 0.1046, 0.0266, -0.1891]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -6.6124e-08, ..., -2.2352e-08, 9.3132e-10, -2.0489e-08], [ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 1.8626e-09, 9.3132e-10, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 3.7253e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, -1.8626e-09, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 4.6566e-09, 2.7940e-09, 4.6566e-09], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -3.2596e-08, 9.3132e-10, 9.3132e-10]], device='cuda:0') Epoch 395, bias, value: tensor([ 0.0031, 0.0050, 0.0134, -0.0029, 0.0289, -0.0049, -0.0004, -0.0015, 0.0107, -0.0176], device='cuda:0'), grad: tensor([-1.0151e-07, -2.1420e-08, 7.4506e-09, -6.2399e-08, 5.9605e-08, 1.8813e-07, -8.0094e-08, 4.7497e-08, 3.0734e-08, -6.8918e-08], device='cuda:0') 100 0.0001 changing lr epoch 394, time 246.03, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.5023 re_mapping 0.0027 re_causal 0.0091 /// teacc 99.11 lr 0.00010000 Epoch 396, weight, value: tensor([[-0.2550, -0.0726, 0.2337, ..., 0.0357, -0.2621, -0.1786], [-0.2188, -0.3003, -0.0790, ..., -0.1439, -0.3891, -0.1308], [ 0.3258, -0.0844, -0.3359, ..., -0.2594, -0.3012, -0.2289], ..., [-0.1335, -0.0608, -0.4417, ..., -0.1503, 0.1525, 0.0568], [-0.2529, -0.1646, -0.0876, ..., -0.2707, -0.0718, -0.0880], [-0.1342, -0.1111, -0.1807, ..., 0.1046, 0.0264, -0.1898]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [-2.7940e-09, -0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 396, bias, value: tensor([ 0.0031, 0.0052, 0.0149, -0.0027, 0.0289, -0.0049, -0.0004, -0.0017, 0.0109, -0.0176], device='cuda:0'), grad: tensor([ 1.8626e-09, 1.8626e-09, -4.6566e-09, -1.2107e-08, -1.0245e-08, 4.6566e-09, 9.3132e-10, 1.0245e-08, 1.8626e-09, 1.2107e-08], device='cuda:0') 100 0.0001 changing lr epoch 395, time 246.59, cls_loss 0.0005 cls_loss_mapping 0.0013 cls_loss_causal 0.4716 re_mapping 0.0028 re_causal 0.0090 /// teacc 99.10 lr 0.00010000 Epoch 397, weight, value: tensor([[-0.2567, -0.0728, 0.2338, ..., 0.0357, -0.2626, -0.1786], [-0.2189, -0.3006, -0.0791, ..., -0.1443, -0.3896, -0.1316], [ 0.3261, -0.0851, -0.3364, ..., -0.2601, -0.3014, -0.2290], ..., [-0.1338, -0.0607, -0.4421, ..., -0.1509, 0.1533, 0.0581], [-0.2531, -0.1650, -0.0883, ..., -0.2741, -0.0720, -0.0886], [-0.1347, -0.1111, -0.1807, ..., 0.1046, 0.0263, -0.1908]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -7.4506e-08, ..., -1.3970e-08, 0.0000e+00, -1.8626e-09], [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [-1.1176e-08, 0.0000e+00, 4.6566e-09, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.6764e-08, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.2107e-08, ..., 5.5879e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 397, bias, value: tensor([ 0.0031, 0.0056, 0.0148, -0.0028, 0.0282, -0.0047, -0.0002, -0.0021, 0.0103, -0.0176], device='cuda:0'), grad: tensor([-1.5553e-07, -2.7940e-09, -1.7695e-08, 0.0000e+00, -4.6566e-09, 1.2107e-08, 7.5437e-08, 3.2596e-08, 3.5390e-08, 2.9802e-08], device='cuda:0') 100 0.0001 changing lr epoch 396, time 246.26, cls_loss 0.0004 cls_loss_mapping 0.0013 cls_loss_causal 0.4759 re_mapping 0.0028 re_causal 0.0093 /// teacc 99.23 lr 0.00010000 Epoch 398, weight, value: tensor([[-0.2567, -0.0730, 0.2338, ..., 0.0357, -0.2629, -0.1786], [-0.2190, -0.3009, -0.0791, ..., -0.1445, -0.3897, -0.1317], [ 0.3263, -0.0855, -0.3372, ..., -0.2606, -0.3015, -0.2290], ..., [-0.1343, -0.0607, -0.4422, ..., -0.1511, 0.1507, 0.0581], [-0.2533, -0.1653, -0.0886, ..., -0.2751, -0.0692, -0.0888], [-0.1358, -0.1111, -0.1807, ..., 0.1046, 0.0263, -0.1911]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-09, ..., -1.8626e-09, 0.0000e+00, -0.0000e+00], [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [-1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 398, bias, value: tensor([ 0.0031, 0.0056, 0.0148, -0.0034, 0.0279, -0.0041, -0.0004, -0.0042, 0.0131, -0.0176], device='cuda:0'), grad: tensor([-7.4506e-09, 0.0000e+00, -9.3132e-10, -4.4703e-08, -2.7940e-09, 2.6077e-08, 0.0000e+00, 1.0245e-08, 4.6566e-09, 9.3132e-09], device='cuda:0') 100 0.0001 changing lr epoch 397, time 246.32, cls_loss 0.0003 cls_loss_mapping 0.0006 cls_loss_causal 0.4783 re_mapping 0.0029 re_causal 0.0095 /// teacc 99.10 lr 0.00010000 Epoch 399, weight, value: tensor([[-0.2568, -0.0730, 0.2338, ..., 0.0357, -0.2634, -0.1787], [-0.2190, -0.3011, -0.0791, ..., -0.1445, -0.3899, -0.1320], [ 0.3268, -0.0855, -0.3378, ..., -0.2610, -0.3016, -0.2291], ..., [-0.1349, -0.0609, -0.4424, ..., -0.1512, 0.1505, 0.0585], [-0.2535, -0.1654, -0.0902, ..., -0.2756, -0.0690, -0.0891], [-0.1360, -0.1112, -0.1807, ..., 0.1046, 0.0263, -0.1915]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, 3.4459e-08, ..., 0.0000e+00, 1.9558e-08, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [-1.4901e-08, -9.3132e-10, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, -0.0000e+00], ..., [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -9.3132e-10, -0.0000e+00], [ 9.3132e-10, 0.0000e+00, 2.5146e-08, ..., 1.8626e-09, 1.3970e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, 9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 399, bias, value: tensor([ 0.0031, 0.0056, 0.0149, -0.0034, 0.0279, -0.0041, -0.0003, -0.0043, 0.0134, -0.0176], device='cuda:0'), grad: tensor([ 6.3330e-08, 6.5193e-09, -3.0734e-08, 1.8626e-09, -6.5193e-09, 1.8626e-09, -1.0990e-07, 1.3970e-08, 5.4948e-08, 5.5879e-09], device='cuda:0') 100 0.0001 changing lr epoch 398, time 246.87, cls_loss 0.0003 cls_loss_mapping 0.0007 cls_loss_causal 0.4623 re_mapping 0.0028 re_causal 0.0092 /// teacc 99.09 lr 0.00010000 Epoch 400, weight, value: tensor([[-0.2571, -0.0730, 0.2338, ..., 0.0357, -0.2635, -0.1788], [-0.2191, -0.3012, -0.0791, ..., -0.1446, -0.3900, -0.1320], [ 0.3275, -0.0855, -0.3379, ..., -0.2612, -0.3016, -0.2287], ..., [-0.1358, -0.0611, -0.4427, ..., -0.1531, 0.1505, 0.0581], [-0.2536, -0.1654, -0.0905, ..., -0.2757, -0.0690, -0.0892], [-0.1360, -0.1112, -0.1808, ..., 0.1046, 0.0264, -0.1916]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.7940e-08, ..., -1.0245e-08, 9.3132e-10, 0.0000e+00], [ 4.6566e-10, 4.6566e-10, 1.3970e-09, ..., 9.3132e-10, 7.9162e-09, 2.7940e-09], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.7940e-09, 0.0000e+00], ..., [ 4.6566e-10, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, -4.2841e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, 4.6566e-10, 4.6566e-10], [ 4.6566e-10, 4.6566e-10, 2.4214e-08, ..., 1.2480e-07, 8.3819e-09, 0.0000e+00]], device='cuda:0') Epoch 400, bias, value: tensor([ 0.0031, 0.0056, 0.0154, -0.0035, 0.0280, -0.0041, -0.0002, -0.0045, 0.0134, -0.0176], device='cuda:0'), grad: tensor([-6.2864e-08, 4.4703e-08, 8.8476e-09, 5.0757e-08, -2.1048e-07, -2.1886e-08, 2.4680e-08, -1.2573e-07, -1.4435e-08, 3.1060e-07], device='cuda:0') 100 0.0001 changing lr epoch 399, time 247.07, cls_loss 0.0005 cls_loss_mapping 0.0014 cls_loss_causal 0.4547 re_mapping 0.0025 re_causal 0.0084 /// teacc 99.11 lr 0.00001000 Epoch 401, weight, value: tensor([[-0.2572, -0.0731, 0.2339, ..., 0.0357, -0.2638, -0.1785], [-0.2192, -0.3014, -0.0794, ..., -0.1449, -0.3903, -0.1322], [ 0.3278, -0.0856, -0.3385, ..., -0.2616, -0.3019, -0.2290], ..., [-0.1361, -0.0625, -0.4433, ..., -0.1545, 0.1503, 0.0589], [-0.2535, -0.1655, -0.0912, ..., -0.2762, -0.0690, -0.0895], [-0.1370, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1929]], device='cuda:0'), grad: tensor([[ 2.3283e-09, 0.0000e+00, -1.3970e-09, ..., -9.3132e-10, 0.0000e+00, 4.6566e-10], [ 2.3283e-09, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, 0.0000e+00, -4.6566e-10], [-3.9116e-08, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 3.2596e-08, 1.8626e-09, 0.0000e+00, ..., 0.0000e+00, 2.3283e-09, 4.6566e-10], [ 3.7253e-09, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, -3.2596e-09, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 401, bias, value: tensor([ 3.0899e-03, 5.6636e-03, 1.5365e-02, -4.5598e-03, 2.8849e-02, -3.6655e-03, -6.2230e-05, -4.6025e-03, 1.3379e-02, -1.7585e-02], device='cuda:0'), grad: tensor([ 2.7940e-09, -1.3504e-08, -6.3330e-08, -1.3970e-08, 2.2817e-08, 2.3283e-08, 6.5658e-08, 1.1129e-07, -1.4808e-07, 9.3132e-09], device='cuda:0') 100 1e-05 changing lr epoch 400, time 246.58, cls_loss 0.0006 cls_loss_mapping 0.0009 cls_loss_causal 0.4189 re_mapping 0.0026 re_causal 0.0079 /// teacc 99.15 lr 0.00001000 Epoch 402, weight, value: tensor([[-0.2572, -0.0731, 0.2339, ..., 0.0357, -0.2638, -0.1785], [-0.2192, -0.3015, -0.0794, ..., -0.1449, -0.3904, -0.1322], [ 0.3278, -0.0856, -0.3386, ..., -0.2616, -0.3019, -0.2290], ..., [-0.1361, -0.0626, -0.4433, ..., -0.1545, 0.1503, 0.0589], [-0.2536, -0.1656, -0.0912, ..., -0.2763, -0.0690, -0.0895], [-0.1370, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1929]], device='cuda:0'), grad: tensor([[ 6.5193e-09, 0.0000e+00, -2.6077e-08, ..., -1.8161e-08, 4.6566e-10, 4.6566e-10], [ 1.3970e-09, 0.0000e+00, 4.6566e-09, ..., 3.2596e-09, 4.6566e-10, 4.6566e-10], [-2.8871e-08, 0.0000e+00, 4.1910e-09, ..., 2.3283e-09, 4.6566e-10, 0.0000e+00], ..., [ 8.8476e-09, 0.0000e+00, 1.3970e-09, ..., 9.3132e-10, 4.6566e-10, -1.3970e-09], [ 9.3132e-10, -0.0000e+00, 3.7253e-09, ..., 2.3283e-09, -3.2596e-09, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 4.6566e-09, ..., 4.1910e-09, 4.6566e-10, 0.0000e+00]], device='cuda:0') Epoch 402, bias, value: tensor([ 3.0888e-03, 5.6677e-03, 1.5326e-02, -4.4647e-03, 2.8847e-02, -3.7790e-03, -4.4756e-05, -4.6102e-03, 1.3386e-02, -1.7584e-02], device='cuda:0'), grad: tensor([-6.8452e-08, 2.3283e-09, -6.5658e-08, 2.5146e-08, 3.2596e-09, 2.3749e-08, 1.3970e-08, 5.2154e-08, -1.3970e-08, 2.7940e-08], device='cuda:0') 100 1e-05 changing lr epoch 401, time 245.89, cls_loss 0.0005 cls_loss_mapping 0.0008 cls_loss_causal 0.4372 re_mapping 0.0024 re_causal 0.0080 /// teacc 99.14 lr 0.00001000 Epoch 403, weight, value: tensor([[-0.2573, -0.0731, 0.2339, ..., 0.0357, -0.2639, -0.1784], [-0.2192, -0.3015, -0.0795, ..., -0.1449, -0.3904, -0.1322], [ 0.3279, -0.0857, -0.3385, ..., -0.2616, -0.3019, -0.2290], ..., [-0.1361, -0.0626, -0.4434, ..., -0.1545, 0.1503, 0.0590], [-0.2536, -0.1656, -0.0913, ..., -0.2763, -0.0690, -0.0895], [-0.1370, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1930]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.4261e-08, ..., -1.8626e-09, 9.3132e-10, -2.6077e-08], [ 0.0000e+00, 4.6566e-10, 4.6566e-10, ..., 2.3283e-09, 2.2817e-08, 9.3132e-09], [-1.8626e-09, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, 4.6566e-10, 4.6566e-10], ..., [ 4.6566e-10, 9.3132e-10, 4.6566e-10, ..., 4.1910e-09, -2.7940e-08, -1.1642e-08], [ 1.8626e-09, 4.6566e-10, 9.3132e-10, ..., 9.3132e-10, 4.6566e-10, 4.6566e-10], [ 4.6566e-10, 4.6566e-10, 4.6566e-09, ..., 5.5879e-09, 1.3970e-09, 2.3283e-09]], device='cuda:0') Epoch 403, bias, value: tensor([ 3.0889e-03, 5.6687e-03, 1.5227e-02, -4.1460e-03, 2.8843e-02, -4.0865e-03, -4.0300e-05, -4.6057e-03, 1.3388e-02, -1.7584e-02], device='cuda:0'), grad: tensor([-1.4668e-07, 1.1455e-07, 3.7253e-09, -3.9581e-08, -4.6100e-08, 1.1502e-07, 6.0536e-08, -1.2154e-07, 1.3504e-08, 6.4261e-08], device='cuda:0') 100 1e-05 changing lr epoch 402, time 246.63, cls_loss 0.0004 cls_loss_mapping 0.0007 cls_loss_causal 0.4439 re_mapping 0.0025 re_causal 0.0081 /// teacc 99.14 lr 0.00001000 Epoch 404, weight, value: tensor([[-0.2573, -0.0731, 0.2339, ..., 0.0357, -0.2640, -0.1785], [-0.2192, -0.3016, -0.0795, ..., -0.1450, -0.3904, -0.1322], [ 0.3279, -0.0857, -0.3385, ..., -0.2617, -0.3019, -0.2290], ..., [-0.1362, -0.0627, -0.4434, ..., -0.1545, 0.1503, 0.0590], [-0.2536, -0.1656, -0.0913, ..., -0.2763, -0.0690, -0.0894], [-0.1371, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1931]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.1910e-09, ..., -2.3283e-09, 1.1176e-08, 1.0245e-08], [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 9.3132e-10], [-6.8918e-08, 4.6566e-10, 4.6566e-10, ..., 4.6566e-10, 2.7940e-09, 2.7940e-09], ..., [ 7.2643e-08, 2.3283e-09, 0.0000e+00, ..., 0.0000e+00, -1.8626e-08, -1.6764e-08], [ 3.2596e-09, 6.0536e-09, 9.3132e-10, ..., 4.6566e-10, 4.6566e-10, 4.6566e-10], [ 4.6566e-10, 4.6566e-10, 9.3132e-10, ..., 2.3283e-09, 9.3132e-10, 9.3132e-10]], device='cuda:0') Epoch 404, bias, value: tensor([ 3.0889e-03, 5.6740e-03, 1.5234e-02, -4.1376e-03, 2.8839e-02, -4.1107e-03, -1.0064e-05, -4.6119e-03, 1.3391e-02, -1.7584e-02], device='cuda:0'), grad: tensor([ 5.2154e-08, 1.1967e-07, -6.5658e-08, -1.4994e-07, 1.8626e-09, 3.4459e-08, -9.3132e-10, -1.8766e-07, 9.0804e-08, 1.0896e-07], device='cuda:0') 100 1e-05 changing lr epoch 403, time 246.09, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4582 re_mapping 0.0024 re_causal 0.0083 /// teacc 99.17 lr 0.00001000 Epoch 405, weight, value: tensor([[-0.2573, -0.0731, 0.2339, ..., 0.0357, -0.2640, -0.1785], [-0.2192, -0.3016, -0.0795, ..., -0.1450, -0.3904, -0.1322], [ 0.3279, -0.0857, -0.3385, ..., -0.2617, -0.3019, -0.2290], ..., [-0.1362, -0.0628, -0.4434, ..., -0.1546, 0.1503, 0.0591], [-0.2536, -0.1656, -0.0913, ..., -0.2763, -0.0690, -0.0894], [-0.1372, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1932]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -9.7789e-09, ..., -5.5879e-09, 0.0000e+00, -9.3132e-10], [ 1.3970e-09, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, 0.0000e+00, -0.0000e+00], [-8.8476e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, -4.6566e-10], ..., [ 5.1223e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 9.3132e-10], [ 2.3283e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 5.1223e-09, ..., 3.2596e-09, 0.0000e+00, 4.6566e-10]], device='cuda:0') Epoch 405, bias, value: tensor([ 3.0889e-03, 5.6746e-03, 1.5228e-02, -4.0974e-03, 2.8843e-02, -4.1383e-03, -2.2912e-05, -4.6139e-03, 1.3391e-02, -1.7584e-02], device='cuda:0'), grad: tensor([-2.0955e-08, -1.0394e-06, 5.5879e-09, 6.0536e-08, 1.6298e-08, -4.6566e-09, 1.3504e-08, 9.6392e-07, 8.8476e-09, 1.4435e-08], device='cuda:0') 100 1e-05 changing lr epoch 404, time 246.23, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4607 re_mapping 0.0024 re_causal 0.0082 /// teacc 99.16 lr 0.00001000 Epoch 406, weight, value: tensor([[-0.2573, -0.0731, 0.2339, ..., 0.0357, -0.2640, -0.1785], [-0.2192, -0.3016, -0.0795, ..., -0.1450, -0.3905, -0.1323], [ 0.3280, -0.0857, -0.3386, ..., -0.2619, -0.3019, -0.2288], ..., [-0.1364, -0.0628, -0.4435, ..., -0.1546, 0.1503, 0.0589], [-0.2536, -0.1656, -0.0914, ..., -0.2764, -0.0690, -0.0895], [-0.1374, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1932]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -1.3970e-09, ..., 9.3132e-09, 0.0000e+00, 0.0000e+00], [ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 3.2596e-09, 4.6566e-10, -4.6566e-10], [-1.8626e-09, 4.6566e-10, 0.0000e+00, ..., 3.2596e-09, 0.0000e+00, 0.0000e+00], ..., [ 6.9849e-09, 1.3970e-09, 0.0000e+00, ..., 1.1642e-08, 9.3132e-10, 4.6566e-10], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 9.3132e-10], [ 4.6566e-10, 0.0000e+00, 1.8626e-09, ..., 4.2841e-08, 1.3970e-09, 0.0000e+00]], device='cuda:0') Epoch 406, bias, value: tensor([ 3.0890e-03, 5.6753e-03, 1.5285e-02, -4.0789e-03, 2.8846e-02, -4.1272e-03, -2.5131e-05, -4.6234e-03, 1.3391e-02, -1.7584e-02], device='cuda:0'), grad: tensor([ 3.0268e-08, -1.8626e-09, 1.1176e-08, -3.6787e-08, -2.9057e-07, -3.9116e-08, 8.1956e-08, 7.9628e-08, 3.6787e-08, 1.4156e-07], device='cuda:0') 100 1e-05 changing lr epoch 405, time 246.31, cls_loss 0.0004 cls_loss_mapping 0.0007 cls_loss_causal 0.4252 re_mapping 0.0023 re_causal 0.0077 /// teacc 99.15 lr 0.00001000 Epoch 407, weight, value: tensor([[-0.2573, -0.0731, 0.2339, ..., 0.0357, -0.2641, -0.1785], [-0.2192, -0.3016, -0.0795, ..., -0.1450, -0.3905, -0.1323], [ 0.3281, -0.0857, -0.3386, ..., -0.2619, -0.3019, -0.2288], ..., [-0.1365, -0.0629, -0.4435, ..., -0.1546, 0.1503, 0.0589], [-0.2537, -0.1656, -0.0914, ..., -0.2764, -0.0690, -0.0894], [-0.1374, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1933]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1176e-08, ..., -4.6566e-09, 0.0000e+00, -0.0000e+00], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 9.3132e-09, -1.9558e-08], [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, -0.0000e+00], ..., [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, -1.3970e-08, 1.8626e-09], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 3.7253e-09], [ 0.0000e+00, 0.0000e+00, 6.5193e-09, ..., -0.0000e+00, 3.7253e-09, 0.0000e+00]], device='cuda:0') Epoch 407, bias, value: tensor([ 3.0889e-03, 5.6765e-03, 1.5312e-02, -4.0540e-03, 2.8864e-02, -4.1526e-03, -2.6312e-05, -4.6281e-03, 1.3393e-02, -1.7584e-02], device='cuda:0'), grad: tensor([-2.4214e-08, -2.3004e-07, 3.7253e-08, 9.3132e-10, 1.1176e-08, 2.9802e-08, 7.4506e-08, 8.4750e-08, -8.3819e-09, 3.1665e-08], device='cuda:0') 100 1e-05 changing lr epoch 406, time 246.78, cls_loss 0.0005 cls_loss_mapping 0.0005 cls_loss_causal 0.4331 re_mapping 0.0023 re_causal 0.0079 /// teacc 99.15 lr 0.00001000 Epoch 408, weight, value: tensor([[-0.2574, -0.0731, 0.2339, ..., 0.0357, -0.2641, -0.1784], [-0.2192, -0.3016, -0.0796, ..., -0.1450, -0.3906, -0.1323], [ 0.3282, -0.0857, -0.3387, ..., -0.2619, -0.3020, -0.2288], ..., [-0.1366, -0.0629, -0.4435, ..., -0.1546, 0.1503, 0.0589], [-0.2537, -0.1656, -0.0915, ..., -0.2764, -0.0690, -0.0894], [-0.1374, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1933]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1176e-08, ..., -4.6566e-09, 0.0000e+00, -3.7253e-09], [ 9.3132e-10, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, 3.7253e-09, 3.7253e-09], [-9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 1.8626e-09, 1.8626e-09], ..., [-2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 1.8626e-09, -1.5832e-08, -1.5832e-08], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 5.5879e-09, 0.0000e+00, 1.8626e-09]], device='cuda:0') Epoch 408, bias, value: tensor([ 3.0890e-03, 5.6792e-03, 1.5310e-02, -4.0430e-03, 2.8865e-02, -4.1699e-03, -1.0583e-05, -4.6305e-03, 1.3394e-02, -1.7585e-02], device='cuda:0'), grad: tensor([-3.1665e-08, 4.0978e-08, 1.3039e-08, 3.7253e-09, 5.2154e-08, -8.3819e-09, 2.4214e-08, -1.1176e-07, 1.8626e-09, 2.0489e-08], device='cuda:0') 100 1e-05 changing lr epoch 407, time 246.23, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4672 re_mapping 0.0022 re_causal 0.0080 /// teacc 99.15 lr 0.00001000 Epoch 409, weight, value: tensor([[-0.2574, -0.0731, 0.2339, ..., 0.0357, -0.2641, -0.1784], [-0.2192, -0.3016, -0.0796, ..., -0.1450, -0.3906, -0.1323], [ 0.3282, -0.0857, -0.3386, ..., -0.2619, -0.3020, -0.2288], ..., [-0.1367, -0.0629, -0.4435, ..., -0.1546, 0.1504, 0.0590], [-0.2537, -0.1656, -0.0915, ..., -0.2765, -0.0690, -0.0894], [-0.1376, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1934]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3039e-08, ..., -5.5879e-09, 1.8626e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 1.0245e-08, -0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 3.0734e-08, 2.1607e-07, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.7940e-09, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 9.3132e-09, ..., 4.6566e-09, 2.7940e-09, 0.0000e+00]], device='cuda:0') Epoch 409, bias, value: tensor([ 3.0890e-03, 5.7177e-03, 1.5336e-02, -4.0331e-03, 2.8868e-02, -4.1572e-03, -1.1627e-05, -4.6578e-03, 1.3392e-02, -1.7585e-02], device='cuda:0'), grad: tensor([-1.4901e-08, 4.5635e-08, 9.3132e-10, -1.2508e-06, 0.0000e+00, 1.7881e-07, -8.1025e-08, 1.0785e-06, 1.3970e-08, 3.1665e-08], device='cuda:0') 100 1e-05 changing lr epoch 408, time 246.67, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4375 re_mapping 0.0023 re_causal 0.0080 /// teacc 99.19 lr 0.00001000 Epoch 410, weight, value: tensor([[-0.2574, -0.0731, 0.2339, ..., 0.0357, -0.2641, -0.1784], [-0.2193, -0.3017, -0.0796, ..., -0.1451, -0.3907, -0.1324], [ 0.3284, -0.0857, -0.3386, ..., -0.2619, -0.3020, -0.2287], ..., [-0.1368, -0.0629, -0.4436, ..., -0.1547, 0.1504, 0.0589], [-0.2538, -0.1656, -0.0915, ..., -0.2765, -0.0690, -0.0894], [-0.1376, -0.1113, -0.1808, ..., 0.1046, 0.0261, -0.1935]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, 0.0000e+00, -0.0000e+00], [ 9.3132e-10, 1.8626e-09, -1.8626e-09, ..., -9.3132e-10, 9.3132e-10, -9.3132e-10], [-1.8626e-09, 4.6566e-09, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 3.7253e-09, 5.5879e-09, 0.0000e+00, ..., 9.3132e-10, -1.8626e-09, 9.3132e-10], [ 1.8626e-09, 9.3132e-10, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.7253e-09, ..., -6.5193e-09, 1.8626e-09, 0.0000e+00]], device='cuda:0') Epoch 410, bias, value: tensor([ 3.0891e-03, 5.7314e-03, 1.5391e-02, -4.0099e-03, 2.8875e-02, -4.1750e-03, -2.3204e-05, -4.6684e-03, 1.3391e-02, -1.7585e-02], device='cuda:0'), grad: tensor([-9.3132e-10, -1.0245e-08, 1.9558e-08, -1.0990e-07, 1.3970e-08, 2.8871e-08, 4.6566e-09, 2.6077e-08, 2.3283e-08, 7.4506e-09], device='cuda:0') 100 1e-05 changing lr epoch 409, time 246.56, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4500 re_mapping 0.0022 re_causal 0.0081 /// teacc 99.20 lr 0.00001000 Epoch 411, weight, value: tensor([[-0.2575, -0.0731, 0.2339, ..., 0.0357, -0.2641, -0.1784], [-0.2193, -0.3017, -0.0797, ..., -0.1451, -0.3907, -0.1325], [ 0.3284, -0.0857, -0.3386, ..., -0.2619, -0.3020, -0.2287], ..., [-0.1369, -0.0629, -0.4436, ..., -0.1547, 0.1504, 0.0590], [-0.2538, -0.1656, -0.0916, ..., -0.2765, -0.0690, -0.0894], [-0.1377, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1935]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.2165e-07, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 9.3132e-10, -0.0000e+00], [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.2107e-08, -1.8626e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -2.5053e-07, 9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 411, bias, value: tensor([ 3.0892e-03, 5.7306e-03, 1.5395e-02, -3.9917e-03, 2.8873e-02, -4.1854e-03, -3.4837e-05, -4.6672e-03, 1.3390e-02, -1.7585e-02], device='cuda:0'), grad: tensor([ 7.0222e-07, 3.7253e-09, -2.7940e-09, 1.6764e-08, 4.8429e-08, -1.3970e-08, 1.8626e-09, 2.8871e-08, 4.6566e-09, -7.8045e-07], device='cuda:0') 100 1e-05 changing lr epoch 410, time 246.40, cls_loss 0.0004 cls_loss_mapping 0.0005 cls_loss_causal 0.4305 re_mapping 0.0022 re_causal 0.0079 /// teacc 99.19 lr 0.00001000 Epoch 412, weight, value: tensor([[-0.2575, -0.0731, 0.2339, ..., 0.0357, -0.2642, -0.1784], [-0.2193, -0.3017, -0.0798, ..., -0.1451, -0.3908, -0.1325], [ 0.3285, -0.0857, -0.3387, ..., -0.2620, -0.3020, -0.2287], ..., [-0.1369, -0.0630, -0.4436, ..., -0.1547, 0.1504, 0.0590], [-0.2538, -0.1656, -0.0916, ..., -0.2766, -0.0690, -0.0894], [-0.1377, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1936]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.3085e-08, ..., -3.3528e-08, 0.0000e+00, -6.5193e-09], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, 3.7253e-09, -0.0000e+00], [-3.7253e-09, -0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 2.7940e-09, 0.0000e+00, 9.3132e-10, ..., 3.7253e-09, -6.5193e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.5146e-08, ..., 8.9407e-08, 1.8626e-09, 2.7940e-09]], device='cuda:0') Epoch 412, bias, value: tensor([ 3.0893e-03, 5.7400e-03, 1.5396e-02, -4.0086e-03, 2.8873e-02, -4.1713e-03, -3.4275e-05, -4.6747e-03, 1.3390e-02, -1.7585e-02], device='cuda:0'), grad: tensor([-1.4529e-07, -3.2596e-08, -1.8626e-09, 3.9116e-08, -2.0023e-07, -3.0734e-08, 6.6124e-08, 3.5390e-08, 1.1176e-08, 2.6450e-07], device='cuda:0') 100 1e-05 changing lr epoch 411, time 246.76, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4451 re_mapping 0.0022 re_causal 0.0080 /// teacc 99.17 lr 0.00001000 Epoch 413, weight, value: tensor([[-0.2575, -0.0731, 0.2339, ..., 0.0357, -0.2642, -0.1784], [-0.2193, -0.3017, -0.0798, ..., -0.1451, -0.3909, -0.1326], [ 0.3285, -0.0857, -0.3387, ..., -0.2620, -0.3020, -0.2287], ..., [-0.1370, -0.0630, -0.4436, ..., -0.1547, 0.1504, 0.0590], [-0.2538, -0.1656, -0.0917, ..., -0.2766, -0.0690, -0.0894], [-0.1377, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1936]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., 0.0000e+00, 8.3819e-09, 1.0245e-08], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, -0.0000e+00], [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, 0.0000e+00, 9.3132e-10], [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 1.8626e-09], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.9558e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 413, bias, value: tensor([ 3.0893e-03, 5.7327e-03, 1.5421e-02, -4.0923e-03, 2.8888e-02, -4.0947e-03, -2.6335e-05, -4.6718e-03, 1.3388e-02, -1.7586e-02], device='cuda:0'), grad: tensor([ 5.6811e-08, -5.3085e-08, -5.5879e-09, 1.9558e-08, -6.7055e-08, -6.6124e-08, -2.6077e-08, 4.9360e-08, 2.0489e-08, 7.6368e-08], device='cuda:0') 100 1e-05 changing lr epoch 412, time 246.59, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4111 re_mapping 0.0022 re_causal 0.0080 /// teacc 99.18 lr 0.00001000 Epoch 414, weight, value: tensor([[-0.2575, -0.0731, 0.2340, ..., 0.0357, -0.2642, -0.1784], [-0.2193, -0.3017, -0.0798, ..., -0.1451, -0.3910, -0.1325], [ 0.3286, -0.0857, -0.3387, ..., -0.2620, -0.3020, -0.2286], ..., [-0.1371, -0.0630, -0.4437, ..., -0.1547, 0.1504, 0.0589], [-0.2539, -0.1656, -0.0917, ..., -0.2766, -0.0690, -0.0894], [-0.1377, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1937]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -1.0245e-08, ..., -7.4506e-09, 9.3132e-10, 9.3132e-10], [ 9.3132e-10, 0.0000e+00, 2.7940e-09, ..., 1.8626e-09, 5.4948e-08, 0.0000e+00], [-3.7253e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, -5.7742e-08, 0.0000e+00], [ 1.8626e-09, 1.8626e-09, 1.8626e-09, ..., 9.3132e-10, 9.3132e-10, -9.3132e-10], [ 9.3132e-10, 9.3132e-10, 3.7253e-09, ..., 9.3132e-10, 9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 414, bias, value: tensor([ 3.0893e-03, 5.7312e-03, 1.5461e-02, -4.0985e-03, 2.8898e-02, -4.0841e-03, -2.6809e-05, -4.6730e-03, 1.3386e-02, -1.7586e-02], device='cuda:0'), grad: tensor([-2.7008e-08, 1.2098e-06, 0.0000e+00, 1.2107e-08, 9.3132e-09, 1.3970e-08, 0.0000e+00, -1.2405e-06, 2.1420e-08, 2.8871e-08], device='cuda:0') 100 1e-05 changing lr epoch 413, time 246.50, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4214 re_mapping 0.0022 re_causal 0.0079 /// teacc 99.23 lr 0.00001000 Epoch 415, weight, value: tensor([[-0.2575, -0.0731, 0.2340, ..., 0.0357, -0.2642, -0.1784], [-0.2193, -0.3017, -0.0799, ..., -0.1451, -0.3911, -0.1326], [ 0.3287, -0.0857, -0.3387, ..., -0.2620, -0.3021, -0.2286], ..., [-0.1372, -0.0630, -0.4437, ..., -0.1547, 0.1504, 0.0589], [-0.2539, -0.1657, -0.0918, ..., -0.2767, -0.0690, -0.0895], [-0.1379, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1937]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], [-2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, -0.0000e+00], ..., [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, -0.0000e+00, 0.0000e+00], [ 9.3132e-10, 9.3132e-10, 0.0000e+00, ..., 3.7253e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 9.3132e-10, 0.0000e+00, ..., -4.1910e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 415, bias, value: tensor([ 3.0893e-03, 5.7281e-03, 1.5516e-02, -4.0983e-03, 2.8898e-02, -4.0690e-03, -9.1731e-06, -4.6711e-03, 1.3385e-02, -1.7586e-02], device='cuda:0'), grad: tensor([ 1.1176e-08, 1.3970e-08, -5.5879e-09, 5.5879e-08, 8.1025e-08, -6.7987e-08, -2.8871e-08, 2.6077e-08, 5.4017e-08, -1.2107e-07], device='cuda:0') 100 1e-05 changing lr epoch 414, time 246.75, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4306 re_mapping 0.0022 re_causal 0.0081 /// teacc 99.21 lr 0.00001000 Epoch 416, weight, value: tensor([[-0.2575, -0.0731, 0.2340, ..., 0.0357, -0.2643, -0.1784], [-0.2193, -0.3017, -0.0799, ..., -0.1452, -0.3911, -0.1326], [ 0.3287, -0.0857, -0.3387, ..., -0.2621, -0.3021, -0.2286], ..., [-0.1372, -0.0631, -0.4437, ..., -0.1547, 0.1505, 0.0590], [-0.2540, -0.1657, -0.0918, ..., -0.2767, -0.0690, -0.0895], [-0.1379, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1938]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., 3.7253e-09, 0.0000e+00, -4.6566e-09], [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], [-1.1176e-07, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 9.3132e-10], [ 1.1176e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -0.0000e+00, -9.3132e-10], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., -1.4901e-08, -0.0000e+00, 9.3132e-10]], device='cuda:0') Epoch 416, bias, value: tensor([ 3.0894e-03, 5.7204e-03, 1.5551e-02, -4.1018e-03, 2.8903e-02, -4.0634e-03, -3.9225e-06, -4.6679e-03, 1.3382e-02, -1.7586e-02], device='cuda:0'), grad: tensor([-1.6764e-08, 1.0245e-08, -2.6543e-07, 1.1176e-08, 9.3132e-10, -2.0489e-08, 3.1665e-08, 2.4866e-07, 2.6077e-08, -2.1420e-08], device='cuda:0') 100 1e-05 changing lr epoch 415, time 246.63, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4529 re_mapping 0.0022 re_causal 0.0081 /// teacc 99.16 lr 0.00001000 Epoch 417, weight, value: tensor([[-0.2576, -0.0731, 0.2340, ..., 0.0357, -0.2643, -0.1784], [-0.2193, -0.3017, -0.0799, ..., -0.1452, -0.3912, -0.1326], [ 0.3288, -0.0857, -0.3388, ..., -0.2621, -0.3021, -0.2286], ..., [-0.1373, -0.0631, -0.4437, ..., -0.1547, 0.1505, 0.0590], [-0.2540, -0.1657, -0.0919, ..., -0.2767, -0.0690, -0.0895], [-0.1379, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1938]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -9.3132e-10, 9.3132e-10, 1.8626e-09], [ 6.5193e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [-1.1176e-08, 0.0000e+00, 2.7940e-09, ..., 0.0000e+00, 9.3132e-10, 9.3132e-10], ..., [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 2.7940e-09, 0.0000e+00, 1.8626e-09, ..., -9.3132e-10, -0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 417, bias, value: tensor([ 3.0894e-03, 5.7139e-03, 1.5585e-02, -4.1038e-03, 2.8905e-02, -4.0572e-03, 7.7366e-06, -4.6640e-03, 1.3378e-02, -1.7586e-02], device='cuda:0'), grad: tensor([ 4.6566e-09, -2.4214e-08, -1.2107e-08, -2.2352e-08, 1.1176e-08, 2.0489e-08, -2.3283e-08, 6.2399e-08, -2.1420e-08, 6.5193e-09], device='cuda:0') 100 1e-05 changing lr epoch 416, time 246.66, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4284 re_mapping 0.0022 re_causal 0.0080 /// teacc 99.18 lr 0.00001000 Epoch 418, weight, value: tensor([[-0.2576, -0.0731, 0.2340, ..., 0.0357, -0.2643, -0.1784], [-0.2193, -0.3017, -0.0800, ..., -0.1452, -0.3913, -0.1327], [ 0.3289, -0.0857, -0.3388, ..., -0.2621, -0.3021, -0.2286], ..., [-0.1374, -0.0631, -0.4438, ..., -0.1548, 0.1505, 0.0590], [-0.2541, -0.1657, -0.0920, ..., -0.2768, -0.0690, -0.0895], [-0.1380, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1938]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3039e-08, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 9.3132e-10, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, -3.7253e-09, -1.8626e-09], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.2573e-07, 9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 418, bias, value: tensor([ 3.0894e-03, 5.7157e-03, 1.5603e-02, -4.1047e-03, 2.8909e-02, -4.0366e-03, 1.5412e-05, -4.6663e-03, 1.3377e-02, -1.7587e-02], device='cuda:0'), grad: tensor([ 3.7253e-08, 2.0489e-08, 7.4506e-09, 3.3807e-07, -3.5111e-07, -3.7346e-07, 1.1176e-08, -7.4506e-09, 5.5879e-09, 3.2783e-07], device='cuda:0') 100 1e-05 changing lr epoch 417, time 246.86, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4226 re_mapping 0.0021 re_causal 0.0077 /// teacc 99.17 lr 0.00001000 Epoch 419, weight, value: tensor([[-0.2576, -0.0731, 0.2340, ..., 0.0357, -0.2644, -0.1784], [-0.2194, -0.3017, -0.0800, ..., -0.1452, -0.3913, -0.1327], [ 0.3289, -0.0857, -0.3389, ..., -0.2622, -0.3021, -0.2286], ..., [-0.1374, -0.0631, -0.4438, ..., -0.1548, 0.1505, 0.0590], [-0.2541, -0.1657, -0.0920, ..., -0.2768, -0.0690, -0.0895], [-0.1380, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1939]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-08, ..., 9.3132e-10, 1.0245e-08, 1.3039e-08], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.5193e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -2.4214e-08, -0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 419, bias, value: tensor([ 3.0894e-03, 5.7150e-03, 1.5618e-02, -4.1364e-03, 2.8925e-02, -3.9985e-03, 7.3957e-06, -4.6687e-03, 1.3374e-02, -1.7587e-02], device='cuda:0'), grad: tensor([ 6.8918e-08, 5.5879e-09, 0.0000e+00, 7.4506e-09, 3.2596e-08, -7.4506e-09, -7.1712e-08, 2.5146e-08, 9.3132e-10, -6.9849e-08], device='cuda:0') 100 1e-05 changing lr epoch 418, time 246.31, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4228 re_mapping 0.0022 re_causal 0.0077 /// teacc 99.19 lr 0.00001000 Epoch 420, weight, value: tensor([[-0.2576, -0.0731, 0.2340, ..., 0.0357, -0.2644, -0.1784], [-0.2194, -0.3017, -0.0801, ..., -0.1453, -0.3914, -0.1328], [ 0.3290, -0.0857, -0.3389, ..., -0.2623, -0.3021, -0.2285], ..., [-0.1376, -0.0631, -0.4438, ..., -0.1548, 0.1505, 0.0590], [-0.2541, -0.1657, -0.0921, ..., -0.2769, -0.0690, -0.0895], [-0.1380, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1939]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 8.3819e-09, 0.0000e+00], [-2.7940e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 2.2352e-08, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, -1.7695e-08, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.4506e-09, 6.5193e-09, 0.0000e+00]], device='cuda:0') Epoch 420, bias, value: tensor([ 3.0895e-03, 5.7084e-03, 1.5689e-02, -4.1365e-03, 2.8935e-02, -4.0021e-03, 1.2542e-05, -4.6701e-03, 1.3374e-02, -1.7587e-02], device='cuda:0'), grad: tensor([ 3.7253e-09, 4.0978e-08, -5.4017e-08, 1.6764e-08, 1.7695e-08, -1.8626e-09, 7.3574e-08, -2.4214e-08, -8.6613e-08, 1.4901e-08], device='cuda:0') 100 1e-05 changing lr epoch 419, time 246.92, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4243 re_mapping 0.0021 re_causal 0.0078 /// teacc 99.21 lr 0.00001000 Epoch 421, weight, value: tensor([[-0.2576, -0.0731, 0.2340, ..., 0.0357, -0.2645, -0.1784], [-0.2194, -0.3017, -0.0801, ..., -0.1453, -0.3915, -0.1328], [ 0.3291, -0.0857, -0.3389, ..., -0.2623, -0.3021, -0.2285], ..., [-0.1376, -0.0632, -0.4438, ..., -0.1548, 0.1505, 0.0590], [-0.2542, -0.1657, -0.0922, ..., -0.2769, -0.0690, -0.0895], [-0.1380, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1940]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-08, ..., -6.5193e-09, 0.0000e+00, 0.0000e+00], [ 7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [-1.0245e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, -0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.3039e-08, ..., -2.5146e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 421, bias, value: tensor([ 3.0895e-03, 5.7111e-03, 1.5691e-02, -4.1334e-03, 2.8948e-02, -4.0063e-03, 5.6771e-05, -4.6741e-03, 1.3371e-02, -1.7587e-02], device='cuda:0'), grad: tensor([-2.5146e-08, -1.8626e-09, -1.4901e-08, 9.3132e-10, 6.4261e-08, 6.5193e-09, -9.3132e-10, 1.8626e-08, 2.7940e-09, -5.6811e-08], device='cuda:0') 100 1e-05 changing lr epoch 420, time 246.60, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3887 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.20 lr 0.00001000 Epoch 422, weight, value: tensor([[-0.2576, -0.0731, 0.2340, ..., 0.0357, -0.2645, -0.1784], [-0.2194, -0.3017, -0.0801, ..., -0.1453, -0.3915, -0.1328], [ 0.3291, -0.0857, -0.3390, ..., -0.2624, -0.3022, -0.2285], ..., [-0.1377, -0.0632, -0.4439, ..., -0.1549, 0.1506, 0.0590], [-0.2542, -0.1657, -0.0923, ..., -0.2770, -0.0690, -0.0895], [-0.1380, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1940]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, 0.0000e+00, -0.0000e+00], [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 9.3132e-10, 0.0000e+00], [-7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, -2.7940e-09, 0.0000e+00], [ 3.7253e-09, 0.0000e+00, 0.0000e+00, ..., -9.3132e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -0.0000e+00, 1.8626e-09, 0.0000e+00]], device='cuda:0') Epoch 422, bias, value: tensor([ 0.0031, 0.0057, 0.0157, -0.0041, 0.0289, -0.0040, 0.0001, -0.0047, 0.0134, -0.0176], device='cuda:0'), grad: tensor([ 0.0000e+00, 1.7695e-08, -1.9558e-08, 5.5879e-09, 1.8626e-09, -1.8626e-08, 2.0489e-08, 2.7940e-09, -1.2387e-07, 1.1921e-07], device='cuda:0') 100 1e-05 changing lr epoch 421, time 246.18, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4038 re_mapping 0.0021 re_causal 0.0074 /// teacc 99.20 lr 0.00001000 Epoch 423, weight, value: tensor([[-0.2577, -0.0731, 0.2340, ..., 0.0357, -0.2646, -0.1784], [-0.2194, -0.3018, -0.0801, ..., -0.1453, -0.3916, -0.1329], [ 0.3292, -0.0857, -0.3390, ..., -0.2624, -0.3022, -0.2284], ..., [-0.1378, -0.0632, -0.4439, ..., -0.1549, 0.1506, 0.0589], [-0.2542, -0.1657, -0.0924, ..., -0.2770, -0.0690, -0.0896], [-0.1380, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1941]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.7253e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.5437e-08, 3.7253e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 7.4506e-09, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 8.0094e-08, -1.3970e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 1.8626e-09, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -7.2643e-08, -3.7253e-09, 0.0000e+00]], device='cuda:0') Epoch 423, bias, value: tensor([ 0.0031, 0.0057, 0.0157, -0.0041, 0.0290, -0.0040, 0.0002, -0.0047, 0.0134, -0.0176], device='cuda:0'), grad: tensor([ 2.0489e-08, -1.1073e-06, 4.9360e-08, 1.8626e-09, 3.6415e-07, 1.3039e-08, 4.6566e-09, 7.5810e-07, 2.7940e-09, -1.0431e-07], device='cuda:0') 100 1e-05 changing lr epoch 422, time 246.54, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4227 re_mapping 0.0021 re_causal 0.0076 /// teacc 99.22 lr 0.00001000 Epoch 424, weight, value: tensor([[-0.2577, -0.0731, 0.2340, ..., 0.0357, -0.2646, -0.1784], [-0.2194, -0.3018, -0.0801, ..., -0.1454, -0.3917, -0.1329], [ 0.3294, -0.0857, -0.3391, ..., -0.2625, -0.3022, -0.2283], ..., [-0.1380, -0.0632, -0.4439, ..., -0.1549, 0.1506, 0.0588], [-0.2542, -0.1657, -0.0925, ..., -0.2771, -0.0690, -0.0896], [-0.1381, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1941]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -3.7253e-09, ..., -9.3132e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [-8.3819e-09, -9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, -0.0000e+00], ..., [ 5.5879e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, -9.3132e-10, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 9.3132e-10, ..., 4.6566e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -6.5193e-09, 0.0000e+00, 1.8626e-09]], device='cuda:0') Epoch 424, bias, value: tensor([ 0.0031, 0.0057, 0.0158, -0.0041, 0.0290, -0.0040, 0.0002, -0.0047, 0.0134, -0.0176], device='cuda:0'), grad: tensor([-3.7253e-09, 3.7253e-09, -1.4901e-08, 2.7940e-09, -2.7940e-09, -9.3132e-10, 5.5879e-09, 1.2107e-08, 1.5832e-08, -1.4901e-08], device='cuda:0') 100 1e-05 changing lr epoch 423, time 246.62, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4283 re_mapping 0.0021 re_causal 0.0077 /// teacc 99.21 lr 0.00001000 Epoch 425, weight, value: tensor([[-0.2577, -0.0731, 0.2340, ..., 0.0357, -0.2646, -0.1784], [-0.2194, -0.3018, -0.0802, ..., -0.1454, -0.3917, -0.1329], [ 0.3295, -0.0857, -0.3391, ..., -0.2625, -0.3022, -0.2283], ..., [-0.1381, -0.0633, -0.4440, ..., -0.1550, 0.1506, 0.0588], [-0.2542, -0.1657, -0.0926, ..., -0.2771, -0.0690, -0.0896], [-0.1381, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1941]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., 5.5879e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], [-4.6566e-09, -0.0000e+00, 0.0000e+00, ..., 5.5879e-09, -0.0000e+00, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 7.4506e-09, 9.3132e-10, 0.0000e+00], [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., -2.1420e-08, -2.7940e-09, 0.0000e+00]], device='cuda:0') Epoch 425, bias, value: tensor([ 0.0031, 0.0057, 0.0158, -0.0041, 0.0290, -0.0040, 0.0002, -0.0047, 0.0134, -0.0176], device='cuda:0'), grad: tensor([ 1.4901e-08, 5.5879e-09, 3.7253e-09, 3.7253e-09, 1.1176e-08, 1.8626e-09, 1.1176e-08, 2.8871e-08, 7.4506e-09, -7.0781e-08], device='cuda:0') 100 1e-05 changing lr epoch 424, time 246.11, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3980 re_mapping 0.0021 re_causal 0.0077 /// teacc 99.17 lr 0.00001000 Epoch 426, weight, value: tensor([[-0.2577, -0.0731, 0.2340, ..., 0.0357, -0.2646, -0.1784], [-0.2195, -0.3018, -0.0802, ..., -0.1454, -0.3918, -0.1329], [ 0.3295, -0.0857, -0.3392, ..., -0.2625, -0.3022, -0.2283], ..., [-0.1381, -0.0633, -0.4440, ..., -0.1550, 0.1506, 0.0588], [-0.2543, -0.1658, -0.0927, ..., -0.2772, -0.0690, -0.0896], [-0.1381, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1941]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, 9.3132e-10, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.7940e-09, 0.0000e+00], [-5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, -9.3132e-09, 0.0000e+00], [ 4.6566e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -5.5879e-09, 5.5879e-09, 0.0000e+00]], device='cuda:0') Epoch 426, bias, value: tensor([ 0.0031, 0.0057, 0.0158, -0.0041, 0.0290, -0.0040, 0.0001, -0.0047, 0.0134, -0.0176], device='cuda:0'), grad: tensor([ 9.3132e-09, 2.7940e-08, -1.2107e-08, -6.5193e-09, 6.5193e-09, 2.2352e-08, -1.3970e-08, -9.4064e-08, 1.5832e-08, 5.4017e-08], device='cuda:0') 100 1e-05 changing lr epoch 425, time 246.26, cls_loss 0.0003 cls_loss_mapping 0.0004 cls_loss_causal 0.4118 re_mapping 0.0021 re_causal 0.0077 /// teacc 99.20 lr 0.00001000 Epoch 427, weight, value: tensor([[-0.2578, -0.0731, 0.2340, ..., 0.0357, -0.2647, -0.1784], [-0.2195, -0.3018, -0.0802, ..., -0.1454, -0.3918, -0.1329], [ 0.3296, -0.0857, -0.3392, ..., -0.2625, -0.3023, -0.2282], ..., [-0.1382, -0.0633, -0.4440, ..., -0.1550, 0.1506, 0.0588], [-0.2543, -0.1658, -0.0928, ..., -0.2773, -0.0690, -0.0896], [-0.1381, -0.1113, -0.1808, ..., 0.1046, 0.0260, -0.1942]], device='cuda:0'), grad: tensor([[ 1.1642e-08, 0.0000e+00, -9.3132e-10, ..., -4.6566e-10, 4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [-1.3970e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [ 2.3283e-09, 4.6566e-10, 4.6566e-10, ..., 0.0000e+00, 4.6566e-10, 4.6566e-10], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -4.1910e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 427, bias, value: tensor([ 0.0031, 0.0057, 0.0159, -0.0041, 0.0290, -0.0040, 0.0002, -0.0047, 0.0134, -0.0176], device='cuda:0'), grad: tensor([ 2.3749e-08, 1.3970e-09, -2.9802e-08, -6.6124e-08, 9.7789e-09, 6.9384e-08, -1.3970e-08, 5.1223e-09, 1.3504e-08, -7.9162e-09], device='cuda:0') 100 1e-05 changing lr epoch 426, time 246.71, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4215 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 Epoch 428, weight, value: tensor([[-0.2578, -0.0731, 0.2340, ..., 0.0357, -0.2647, -0.1784], [-0.2195, -0.3018, -0.0803, ..., -0.1455, -0.3919, -0.1329], [ 0.3297, -0.0857, -0.3393, ..., -0.2626, -0.3023, -0.2281], ..., [-0.1384, -0.0634, -0.4441, ..., -0.1551, 0.1506, 0.0587], [-0.2544, -0.1658, -0.0930, ..., -0.2774, -0.0690, -0.0897], [-0.1381, -0.1114, -0.1809, ..., 0.1046, 0.0260, -0.1942]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -8.1025e-08, ..., -5.1223e-08, 1.3970e-09, 1.3970e-09], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 2.3283e-09, 9.3132e-10, 4.6566e-10], [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 0.0000e+00, 2.3283e-09, ..., 2.7940e-09, -9.3132e-10, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 3.2596e-09, ..., 2.3283e-09, 3.7253e-09, 4.6566e-09], [ 0.0000e+00, 0.0000e+00, 3.1199e-08, ..., 7.9162e-09, 4.6566e-10, 0.0000e+00]], device='cuda:0') Epoch 428, bias, value: tensor([ 0.0031, 0.0057, 0.0160, -0.0041, 0.0290, -0.0040, 0.0002, -0.0047, 0.0134, -0.0176], device='cuda:0'), grad: tensor([-2.6915e-07, 1.2107e-08, 1.1176e-08, 2.3283e-09, 3.7719e-08, 1.7602e-07, -7.4971e-08, 1.1176e-08, 2.8871e-08, 6.7521e-08], device='cuda:0') 100 1e-05 changing lr epoch 427, time 245.81, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.3944 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.20 lr 0.00001000 Epoch 429, weight, value: tensor([[-0.2578, -0.0731, 0.2341, ..., 0.0357, -0.2647, -0.1784], [-0.2195, -0.3018, -0.0804, ..., -0.1455, -0.3920, -0.1330], [ 0.3298, -0.0858, -0.3394, ..., -0.2626, -0.3023, -0.2281], ..., [-0.1384, -0.0634, -0.4441, ..., -0.1551, 0.1507, 0.0587], [-0.2544, -0.1658, -0.0931, ..., -0.2775, -0.0691, -0.0897], [-0.1382, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1942]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, 4.6566e-10, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], ..., [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -1.4435e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, 1.3970e-09, 0.0000e+00]], device='cuda:0') Epoch 429, bias, value: tensor([ 0.0031, 0.0057, 0.0160, -0.0041, 0.0290, -0.0041, 0.0003, -0.0047, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 5.5879e-09, 8.7544e-08, 6.9849e-09, 2.5146e-08, 5.1223e-09, -1.3504e-08, -6.0536e-09, -1.2945e-07, 1.1176e-08, 1.6764e-08], device='cuda:0') 100 1e-05 changing lr epoch 428, time 245.56, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4130 re_mapping 0.0020 re_causal 0.0077 /// teacc 99.21 lr 0.00001000 Epoch 430, weight, value: tensor([[-0.2578, -0.0731, 0.2341, ..., 0.0357, -0.2648, -0.1784], [-0.2195, -0.3018, -0.0804, ..., -0.1455, -0.3921, -0.1330], [ 0.3298, -0.0858, -0.3394, ..., -0.2627, -0.3023, -0.2281], ..., [-0.1384, -0.0634, -0.4442, ..., -0.1551, 0.1507, 0.0587], [-0.2544, -0.1658, -0.0932, ..., -0.2775, -0.0691, -0.0897], [-0.1382, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1943]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., -9.3132e-10, 1.8626e-09, 1.8626e-09], [ 1.7229e-08, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 3.7253e-09, 4.1910e-09], [-2.7474e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 1.8626e-09, 1.3970e-09], ..., [ 1.0710e-08, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, -2.2352e-08, -2.2817e-08], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-09, 2.7940e-09], [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 6.9849e-09, 4.6566e-10, 1.3970e-09]], device='cuda:0') Epoch 430, bias, value: tensor([ 0.0031, 0.0057, 0.0160, -0.0041, 0.0290, -0.0041, 0.0003, -0.0047, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.2573e-08, 1.8440e-07, -4.8429e-08, 1.4435e-07, 3.0268e-08, -5.7556e-07, 4.6426e-07, -2.6077e-07, 3.3528e-08, 2.9802e-08], device='cuda:0') 100 1e-05 changing lr epoch 429, time 245.88, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4063 re_mapping 0.0020 re_causal 0.0078 /// teacc 99.21 lr 0.00001000 Epoch 431, weight, value: tensor([[-0.2578, -0.0731, 0.2341, ..., 0.0357, -0.2648, -0.1784], [-0.2195, -0.3019, -0.0805, ..., -0.1455, -0.3922, -0.1330], [ 0.3299, -0.0858, -0.3395, ..., -0.2627, -0.3023, -0.2281], ..., [-0.1385, -0.0635, -0.4442, ..., -0.1551, 0.1507, 0.0587], [-0.2544, -0.1658, -0.0933, ..., -0.2776, -0.0691, -0.0898], [-0.1382, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1943]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.9849e-09, ..., -9.3132e-10, 0.0000e+00, -0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, -9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 6.9849e-09, 4.6566e-10, 0.0000e+00], [ 0.0000e+00, 4.6566e-10, 6.5193e-09, ..., -9.3132e-10, 4.6566e-10, 0.0000e+00]], device='cuda:0') Epoch 431, bias, value: tensor([ 0.0031, 0.0057, 0.0160, -0.0041, 0.0290, -0.0041, 0.0003, -0.0047, 0.0133, -0.0176], device='cuda:0'), grad: tensor([-7.4506e-09, 7.9162e-09, 2.3283e-09, -1.0151e-07, -3.3993e-08, 3.9116e-08, 3.1665e-08, 4.1910e-09, 4.0978e-08, 2.0955e-08], device='cuda:0') 100 1e-05 changing lr epoch 430, time 246.03, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4198 re_mapping 0.0021 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 Epoch 432, weight, value: tensor([[-0.2579, -0.0731, 0.2341, ..., 0.0357, -0.2649, -0.1784], [-0.2195, -0.3019, -0.0805, ..., -0.1456, -0.3923, -0.1330], [ 0.3299, -0.0858, -0.3396, ..., -0.2628, -0.3024, -0.2281], ..., [-0.1385, -0.0635, -0.4442, ..., -0.1552, 0.1507, 0.0588], [-0.2545, -0.1658, -0.0934, ..., -0.2776, -0.0691, -0.0898], [-0.1382, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1943]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, 4.6566e-10, 4.6566e-10], [ 4.1910e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, -0.0000e+00], [-2.2817e-08, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, 4.6566e-10, 4.6566e-10], ..., [ 1.5367e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -9.3132e-10, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 432, bias, value: tensor([ 0.0031, 0.0058, 0.0160, -0.0041, 0.0290, -0.0041, 0.0003, -0.0047, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 5.5879e-09, -9.7789e-09, -4.3772e-08, 3.2596e-09, 1.2573e-08, 1.1176e-08, -1.3970e-08, 4.2375e-08, 5.5879e-09, -1.3970e-09], device='cuda:0') 100 1e-05 changing lr epoch 431, time 246.18, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4277 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.22 lr 0.00001000 Epoch 433, weight, value: tensor([[-0.2579, -0.0731, 0.2341, ..., 0.0357, -0.2649, -0.1784], [-0.2196, -0.3019, -0.0805, ..., -0.1456, -0.3925, -0.1330], [ 0.3301, -0.0858, -0.3396, ..., -0.2628, -0.3024, -0.2280], ..., [-0.1388, -0.0635, -0.4442, ..., -0.1552, 0.1508, 0.0586], [-0.2545, -0.1658, -0.0934, ..., -0.2777, -0.0691, -0.0898], [-0.1383, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1944]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -1.6298e-08, ..., -1.1642e-08, 0.0000e+00, 0.0000e+00], [ 4.6566e-09, 0.0000e+00, 3.2596e-09, ..., 1.3970e-09, 1.3970e-09, 0.0000e+00], [-2.3283e-08, 0.0000e+00, 2.7940e-09, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], ..., [ 1.1642e-08, 0.0000e+00, 1.8626e-09, ..., 2.7940e-09, -2.7940e-09, 0.0000e+00], [ 1.3970e-09, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 7.9162e-09, ..., 1.3644e-07, 1.3970e-09, 0.0000e+00]], device='cuda:0') Epoch 433, bias, value: tensor([ 0.0031, 0.0058, 0.0161, -0.0041, 0.0290, -0.0041, 0.0003, -0.0048, 0.0133, -0.0176], device='cuda:0'), grad: tensor([-4.5169e-08, -9.1270e-08, -5.2154e-08, 7.9162e-09, -2.9569e-07, 5.1223e-09, 6.5193e-09, 1.1316e-07, 1.0710e-08, 3.5344e-07], device='cuda:0') 100 1e-05 changing lr epoch 432, time 246.41, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4178 re_mapping 0.0020 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 Epoch 434, weight, value: tensor([[-0.2579, -0.0731, 0.2341, ..., 0.0357, -0.2650, -0.1784], [-0.2196, -0.3019, -0.0806, ..., -0.1456, -0.3926, -0.1330], [ 0.3302, -0.0858, -0.3397, ..., -0.2629, -0.3024, -0.2280], ..., [-0.1388, -0.0635, -0.4443, ..., -0.1553, 0.1508, 0.0586], [-0.2545, -0.1658, -0.0935, ..., -0.2778, -0.0691, -0.0899], [-0.1383, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1944]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.7695e-08, ..., 9.6858e-08, 0.0000e+00, -4.6566e-10], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 0.0000e+00, 2.3283e-09, ..., 1.8626e-09, -1.3970e-09, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 4.6566e-09, ..., 4.6566e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 6.9849e-09, ..., -1.2945e-07, 4.6566e-10, 0.0000e+00]], device='cuda:0') Epoch 434, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0041, 0.0290, -0.0041, 0.0004, -0.0048, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.7649e-07, 1.2573e-08, 2.3283e-09, 3.7253e-09, 6.7521e-08, -9.3132e-10, 7.4506e-09, 9.3132e-09, 2.4680e-08, -2.8266e-07], device='cuda:0') 100 1e-05 changing lr epoch 433, time 246.03, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4126 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 Epoch 435, weight, value: tensor([[-0.2579, -0.0731, 0.2341, ..., 0.0357, -0.2650, -0.1784], [-0.2196, -0.3019, -0.0806, ..., -0.1456, -0.3927, -0.1330], [ 0.3303, -0.0858, -0.3397, ..., -0.2629, -0.3024, -0.2280], ..., [-0.1388, -0.0636, -0.4443, ..., -0.1553, 0.1508, 0.0586], [-0.2546, -0.1658, -0.0936, ..., -0.2779, -0.0691, -0.0899], [-0.1383, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1944]], device='cuda:0'), grad: tensor([[ 4.1910e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 8.8476e-09, 9.3132e-10], [-7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-10, 0.0000e+00], ..., [ 2.3283e-09, 9.3132e-10, 0.0000e+00, ..., 9.3132e-10, -3.3062e-08, -1.3970e-09], [ 4.6566e-09, 9.3132e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., -1.3970e-09, 2.0489e-08, 0.0000e+00]], device='cuda:0') Epoch 435, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0040, 0.0291, -0.0041, 0.0004, -0.0048, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.5367e-08, 3.2131e-08, -1.5367e-08, -7.0315e-08, 1.1642e-08, 2.8405e-08, 9.3132e-10, -1.0803e-07, 2.8871e-08, 7.9628e-08], device='cuda:0') 100 1e-05 changing lr epoch 434, time 246.07, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4203 re_mapping 0.0019 re_causal 0.0074 /// teacc 99.20 lr 0.00001000 Epoch 436, weight, value: tensor([[-0.2580, -0.0731, 0.2341, ..., 0.0357, -0.2652, -0.1785], [-0.2196, -0.3019, -0.0806, ..., -0.1457, -0.3928, -0.1331], [ 0.3303, -0.0858, -0.3398, ..., -0.2629, -0.3025, -0.2280], ..., [-0.1389, -0.0636, -0.4444, ..., -0.1554, 0.1509, 0.0586], [-0.2546, -0.1658, -0.0938, ..., -0.2780, -0.0691, -0.0899], [-0.1383, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1944]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.3970e-09, ..., -4.6566e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, 0.0000e+00, 4.6566e-10], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 9.3132e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 436, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0040, 0.0291, -0.0041, 0.0005, -0.0048, 0.0133, -0.0176], device='cuda:0'), grad: tensor([-3.7253e-09, 2.3283e-09, 4.6566e-10, -1.8626e-09, -1.7229e-08, 4.1910e-09, 7.9162e-09, 6.9849e-09, 3.2596e-09, 1.9558e-08], device='cuda:0') 100 1e-05 changing lr epoch 435, time 246.07, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4230 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.20 lr 0.00001000 Epoch 437, weight, value: tensor([[-0.2580, -0.0731, 0.2341, ..., 0.0357, -0.2652, -0.1785], [-0.2196, -0.3019, -0.0806, ..., -0.1457, -0.3930, -0.1331], [ 0.3304, -0.0858, -0.3398, ..., -0.2630, -0.3025, -0.2279], ..., [-0.1389, -0.0636, -0.4444, ..., -0.1555, 0.1509, 0.0586], [-0.2546, -0.1659, -0.0938, ..., -0.2781, -0.0691, -0.0899], [-0.1384, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1945]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.0489e-08, ..., -1.0245e-08, 2.3283e-09, -9.3132e-10], [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, 9.3132e-10, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, 0.0000e+00, 7.9162e-09], ..., [ 1.3970e-09, 0.0000e+00, 1.8626e-09, ..., 1.3970e-09, -9.3132e-10, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 5.1223e-09, ..., 4.6566e-10, 2.3283e-09, -6.6124e-08], [ 0.0000e+00, 0.0000e+00, 1.3504e-08, ..., -1.7695e-08, 4.6566e-10, 9.3132e-10]], device='cuda:0') Epoch 437, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0040, 0.0291, -0.0041, 0.0005, -0.0048, 0.0133, -0.0176], device='cuda:0'), grad: tensor([-4.3306e-08, -5.8394e-07, 1.4482e-07, -4.0047e-08, 6.6124e-08, 9.8813e-07, 8.3819e-08, 5.5926e-07, -1.1902e-06, 2.0023e-08], device='cuda:0') 100 1e-05 changing lr epoch 436, time 245.94, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4444 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 Epoch 438, weight, value: tensor([[-0.2580, -0.0731, 0.2341, ..., 0.0357, -0.2653, -0.1785], [-0.2197, -0.3019, -0.0806, ..., -0.1457, -0.3931, -0.1331], [ 0.3304, -0.0858, -0.3399, ..., -0.2630, -0.3025, -0.2279], ..., [-0.1389, -0.0636, -0.4444, ..., -0.1555, 0.1509, 0.0586], [-0.2547, -0.1659, -0.0939, ..., -0.2783, -0.0691, -0.0899], [-0.1384, -0.1114, -0.1809, ..., 0.1046, 0.0259, -0.1945]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, -8.8476e-09, ..., -4.6566e-09, 0.0000e+00, 0.0000e+00], [ 6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 4.6566e-10, 0.0000e+00], [-3.8650e-08, 0.0000e+00, 9.3132e-10, ..., -6.5193e-09, 0.0000e+00, 0.0000e+00], ..., [ 1.1642e-08, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, -9.3132e-10, 0.0000e+00], [ 6.9849e-09, 0.0000e+00, 0.0000e+00, ..., 3.2596e-09, 0.0000e+00, 0.0000e+00], [ 1.3039e-08, 0.0000e+00, 6.9849e-09, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 438, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0041, 0.0291, -0.0041, 0.0005, -0.0048, 0.0133, -0.0176], device='cuda:0'), grad: tensor([-1.8161e-08, 1.5832e-08, -8.2888e-08, 2.4214e-08, 1.1642e-08, -1.9558e-08, 2.3283e-09, 3.6787e-08, 2.7474e-08, 1.3970e-08], device='cuda:0') 100 1e-05 changing lr epoch 437, time 246.07, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3979 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.21 lr 0.00001000 Epoch 439, weight, value: tensor([[-0.2581, -0.0731, 0.2341, ..., 0.0357, -0.2653, -0.1785], [-0.2197, -0.3019, -0.0807, ..., -0.1458, -0.3932, -0.1332], [ 0.3305, -0.0858, -0.3400, ..., -0.2631, -0.3025, -0.2280], ..., [-0.1390, -0.0637, -0.4444, ..., -0.1556, 0.1510, 0.0587], [-0.2547, -0.1659, -0.0940, ..., -0.2784, -0.0691, -0.0900], [-0.1384, -0.1114, -0.1809, ..., 0.1046, 0.0258, -0.1945]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 2.4214e-08, ..., 9.3132e-09, 0.0000e+00, 2.3283e-09], [ 1.3970e-09, 1.3970e-09, -2.6543e-08, ..., -5.1223e-09, 0.0000e+00, -2.3283e-09], [-2.3749e-08, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.9558e-08, 4.6566e-09, 9.3132e-10, ..., 2.7940e-09, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 2.3283e-08, 0.0000e+00, 0.0000e+00], [-2.7940e-09, 4.6566e-10, 0.0000e+00, ..., -7.1246e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 439, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0041, 0.0291, -0.0041, 0.0005, -0.0048, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.0710e-07, -8.0094e-08, -6.1002e-08, -1.6391e-07, 1.2806e-07, 8.7079e-08, 8.8476e-09, 1.3830e-07, 9.5461e-08, -2.5565e-07], device='cuda:0') 100 1e-05 changing lr epoch 438, time 245.89, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4197 re_mapping 0.0019 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 Epoch 440, weight, value: tensor([[-0.2581, -0.0731, 0.2342, ..., 0.0357, -0.2654, -0.1785], [-0.2197, -0.3020, -0.0807, ..., -0.1458, -0.3933, -0.1332], [ 0.3305, -0.0858, -0.3401, ..., -0.2631, -0.3026, -0.2280], ..., [-0.1390, -0.0637, -0.4445, ..., -0.1557, 0.1510, 0.0587], [-0.2547, -0.1659, -0.0941, ..., -0.2785, -0.0691, -0.0900], [-0.1384, -0.1114, -0.1809, ..., 0.1046, 0.0258, -0.1945]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.8626e-09, 4.6566e-10, 4.6566e-10], [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.6298e-08, 4.6566e-10, -1.3970e-09], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.3749e-08, 1.8626e-09, 4.6566e-10], [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.5146e-08, -6.9849e-09, 0.0000e+00]], device='cuda:0') Epoch 440, bias, value: tensor([ 0.0031, 0.0058, 0.0162, -0.0041, 0.0291, -0.0041, 0.0005, -0.0048, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.8626e-08, -5.6345e-08, 6.5193e-09, 9.3132e-09, -6.1467e-08, 1.3039e-08, 1.4435e-08, 1.3132e-07, 2.5146e-08, -7.6368e-08], device='cuda:0') 100 1e-05 changing lr epoch 439, time 245.97, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4420 re_mapping 0.0019 re_causal 0.0076 /// teacc 99.18 lr 0.00001000 Epoch 441, weight, value: tensor([[-0.2581, -0.0731, 0.2342, ..., 0.0357, -0.2654, -0.1785], [-0.2197, -0.3020, -0.0807, ..., -0.1459, -0.3934, -0.1332], [ 0.3307, -0.0858, -0.3402, ..., -0.2631, -0.3026, -0.2279], ..., [-0.1391, -0.0637, -0.4445, ..., -0.1558, 0.1511, 0.0586], [-0.2547, -0.1659, -0.0942, ..., -0.2786, -0.0691, -0.0900], [-0.1384, -0.1114, -0.1809, ..., 0.1046, 0.0258, -0.1946]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 4.6566e-10, 0.0000e+00]], device='cuda:0') Epoch 441, bias, value: tensor([ 0.0031, 0.0059, 0.0162, -0.0041, 0.0291, -0.0042, 0.0005, -0.0048, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 2.7940e-09, -6.9849e-09, 3.7253e-09, 4.6566e-10, 1.8626e-09, 7.9162e-09, -9.3132e-10, 6.0536e-09, -1.3039e-08, 1.2107e-08], device='cuda:0') 100 1e-05 changing lr epoch 440, time 246.41, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4306 re_mapping 0.0019 re_causal 0.0076 /// teacc 99.17 lr 0.00001000 Epoch 442, weight, value: tensor([[-0.2581, -0.0731, 0.2342, ..., 0.0357, -0.2655, -0.1785], [-0.2197, -0.3020, -0.0808, ..., -0.1459, -0.3935, -0.1332], [ 0.3307, -0.0858, -0.3402, ..., -0.2632, -0.3027, -0.2279], ..., [-0.1392, -0.0637, -0.4445, ..., -0.1559, 0.1511, 0.0586], [-0.2548, -0.1659, -0.0943, ..., -0.2787, -0.0691, -0.0900], [-0.1384, -0.1114, -0.1810, ..., 0.1046, 0.0258, -0.1946]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 4.6566e-10], [-1.1642e-08, -4.6566e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, -1.8626e-09], ..., [ 1.1642e-08, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, -9.3132e-10, 1.8626e-09], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 2.7940e-09, 4.6566e-10, 0.0000e+00]], device='cuda:0') Epoch 442, bias, value: tensor([ 0.0031, 0.0059, 0.0162, -0.0041, 0.0291, -0.0042, 0.0005, -0.0049, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.3970e-09, -1.3039e-08, -2.5611e-08, 1.3970e-09, -3.7253e-09, 2.1886e-08, -1.1642e-08, 3.0734e-08, -1.3504e-08, 1.5367e-08], device='cuda:0') 100 1e-05 changing lr epoch 441, time 246.07, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4056 re_mapping 0.0020 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 Epoch 443, weight, value: tensor([[-0.2581, -0.0731, 0.2342, ..., 0.0357, -0.2655, -0.1784], [-0.2197, -0.3020, -0.0808, ..., -0.1460, -0.3936, -0.1332], [ 0.3307, -0.0858, -0.3403, ..., -0.2632, -0.3027, -0.2279], ..., [-0.1392, -0.0638, -0.4446, ..., -0.1559, 0.1511, 0.0587], [-0.2548, -0.1659, -0.0943, ..., -0.2788, -0.0691, -0.0900], [-0.1385, -0.1114, -0.1810, ..., 0.1046, 0.0258, -0.1946]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., -9.3132e-10, 4.6566e-10, 4.6566e-10], [ 4.6566e-09, 4.6566e-10, 9.3132e-10, ..., 1.3970e-09, 1.8626e-09, 0.0000e+00], [-1.1642e-08, 0.0000e+00, 1.8626e-09, ..., 0.0000e+00, 4.6566e-10, 0.0000e+00], ..., [ 4.1910e-09, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, -2.0955e-08, -0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 6.5193e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., -1.7695e-08, 1.9092e-08, 0.0000e+00]], device='cuda:0') Epoch 443, bias, value: tensor([ 0.0031, 0.0059, 0.0162, -0.0041, 0.0291, -0.0042, 0.0005, -0.0049, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.2107e-08, 1.2107e-08, -4.6566e-09, -8.3819e-09, 4.3772e-08, 2.3283e-09, -2.5146e-08, -5.6811e-08, -6.0536e-09, 3.0734e-08], device='cuda:0') 100 1e-05 changing lr epoch 442, time 246.29, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4119 re_mapping 0.0020 re_causal 0.0077 /// teacc 99.19 lr 0.00001000 Epoch 444, weight, value: tensor([[-0.2582, -0.0731, 0.2343, ..., 0.0357, -0.2655, -0.1784], [-0.2198, -0.3020, -0.0809, ..., -0.1460, -0.3936, -0.1333], [ 0.3308, -0.0858, -0.3404, ..., -0.2633, -0.3027, -0.2279], ..., [-0.1392, -0.0638, -0.4446, ..., -0.1560, 0.1511, 0.0587], [-0.2548, -0.1659, -0.0944, ..., -0.2788, -0.0691, -0.0900], [-0.1385, -0.1114, -0.1810, ..., 0.1046, 0.0258, -0.1947]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -2.7940e-09, 5.1223e-09, 2.7940e-09], [ 1.3970e-09, 0.0000e+00, 4.6566e-10, ..., 2.7940e-09, 9.3132e-10, 4.6566e-10], [-1.3970e-09, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], ..., [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 3.4925e-08, -9.3132e-10, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 8.8476e-09, ..., 2.3283e-09, 9.3132e-10, 4.6566e-10], [ 0.0000e+00, 0.0000e+00, 2.7940e-09, ..., 4.5169e-08, 4.6566e-10, 4.6566e-10]], device='cuda:0') Epoch 444, bias, value: tensor([ 0.0031, 0.0059, 0.0162, -0.0041, 0.0291, -0.0042, 0.0005, -0.0049, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 6.0536e-09, 1.4435e-08, -1.3970e-09, 2.0489e-08, -2.5285e-07, -1.1176e-08, -4.4238e-08, 8.9873e-08, 2.4214e-08, 1.6438e-07], device='cuda:0') 100 1e-05 changing lr epoch 443, time 246.37, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4123 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.18 lr 0.00001000 Epoch 445, weight, value: tensor([[-0.2582, -0.0732, 0.2343, ..., 0.0357, -0.2656, -0.1784], [-0.2198, -0.3020, -0.0809, ..., -0.1461, -0.3937, -0.1333], [ 0.3309, -0.0858, -0.3405, ..., -0.2634, -0.3027, -0.2279], ..., [-0.1393, -0.0638, -0.4447, ..., -0.1561, 0.1511, 0.0587], [-0.2548, -0.1659, -0.0944, ..., -0.2789, -0.0691, -0.0900], [-0.1385, -0.1114, -0.1810, ..., 0.1046, 0.0258, -0.1947]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 2.7940e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 4.6566e-10, 0.0000e+00], [-3.7253e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, -9.3132e-10, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., -7.4506e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 445, bias, value: tensor([ 0.0031, 0.0059, 0.0163, -0.0041, 0.0291, -0.0042, 0.0005, -0.0049, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 3.2596e-09, -2.7940e-09, -2.7940e-09, -1.7229e-08, 2.1886e-08, 1.3504e-08, -7.4506e-09, 8.8476e-09, 5.1223e-09, -1.5832e-08], device='cuda:0') 100 1e-05 changing lr epoch 444, time 245.96, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.3998 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.20 lr 0.00001000 Epoch 446, weight, value: tensor([[-0.2582, -0.0732, 0.2343, ..., 0.0357, -0.2656, -0.1784], [-0.2199, -0.3020, -0.0810, ..., -0.1461, -0.3938, -0.1333], [ 0.3310, -0.0858, -0.3405, ..., -0.2636, -0.3027, -0.2279], ..., [-0.1394, -0.0638, -0.4447, ..., -0.1562, 0.1512, 0.0587], [-0.2549, -0.1659, -0.0945, ..., -0.2791, -0.0691, -0.0901], [-0.1385, -0.1114, -0.1810, ..., 0.1046, 0.0258, -0.1947]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, 0.0000e+00, 0.0000e+00], [ 1.3970e-09, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [-9.7789e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, -9.3132e-10], ..., [ 6.0536e-09, 4.6566e-10, 0.0000e+00, ..., 3.2596e-09, 4.6566e-10, 4.6566e-10], [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 1.3970e-09, ..., -6.5193e-09, -9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 446, bias, value: tensor([ 0.0031, 0.0059, 0.0163, -0.0041, 0.0292, -0.0042, 0.0005, -0.0049, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 9.3132e-10, -2.9337e-08, -2.7474e-08, -6.9849e-09, 1.3504e-08, -4.1910e-09, 1.1642e-08, 4.6566e-08, 8.3819e-09, -1.0710e-08], device='cuda:0') 100 1e-05 changing lr epoch 445, time 246.24, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4244 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.18 lr 0.00001000 Epoch 447, weight, value: tensor([[-0.2583, -0.0732, 0.2343, ..., 0.0357, -0.2657, -0.1784], [-0.2199, -0.3020, -0.0810, ..., -0.1462, -0.3939, -0.1333], [ 0.3311, -0.0858, -0.3406, ..., -0.2636, -0.3028, -0.2279], ..., [-0.1395, -0.0638, -0.4447, ..., -0.1563, 0.1512, 0.0587], [-0.2549, -0.1659, -0.0946, ..., -0.2792, -0.0691, -0.0901], [-0.1386, -0.1114, -0.1810, ..., 0.1046, 0.0258, -0.1948]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, -7.2177e-09, ..., -2.0955e-09, 0.0000e+00, 0.0000e+00], [ 4.6566e-10, 2.3283e-10, 2.3283e-10, ..., -1.3970e-09, 2.3283e-09, 2.3283e-10], [-5.5879e-09, -9.3132e-10, 4.6566e-10, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], ..., [ 3.2596e-09, 2.3283e-10, 2.3283e-10, ..., 4.6566e-10, -1.2340e-08, -1.6298e-09], [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 5.8208e-09, ..., -2.0955e-09, 8.1491e-09, 1.1642e-09]], device='cuda:0') Epoch 447, bias, value: tensor([ 0.0031, 0.0059, 0.0163, -0.0041, 0.0292, -0.0042, 0.0006, -0.0049, 0.0133, -0.0176], device='cuda:0'), grad: tensor([-6.9849e-10, -4.0280e-08, -6.9849e-09, 3.4692e-08, 2.5611e-08, -2.9802e-08, 7.2177e-09, -6.5193e-09, 8.8476e-09, 2.7707e-08], device='cuda:0') 100 1e-05 changing lr epoch 446, time 246.60, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4281 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.18 lr 0.00001000 Epoch 448, weight, value: tensor([[-0.2583, -0.0732, 0.2343, ..., 0.0357, -0.2657, -0.1784], [-0.2199, -0.3020, -0.0810, ..., -0.1462, -0.3940, -0.1333], [ 0.3312, -0.0858, -0.3406, ..., -0.2636, -0.3028, -0.2279], ..., [-0.1395, -0.0639, -0.4448, ..., -0.1564, 0.1512, 0.0587], [-0.2550, -0.1659, -0.0947, ..., -0.2793, -0.0691, -0.0901], [-0.1386, -0.1114, -0.1810, ..., 0.1046, 0.0257, -0.1948]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.0151e-07, ..., -4.1444e-08, 4.6566e-10, -6.9849e-10], [ 0.0000e+00, 0.0000e+00, 2.1188e-08, ..., 1.1176e-08, 1.8626e-09, 1.3970e-09], [ 0.0000e+00, 0.0000e+00, 1.9092e-08, ..., 7.9162e-09, 1.6298e-09, 6.9849e-10], ..., [ 2.3283e-10, 0.0000e+00, 9.3132e-10, ..., 6.2864e-09, -3.2596e-09, -2.0955e-09], [ 0.0000e+00, 0.0000e+00, 4.8894e-09, ..., 6.0536e-09, 6.9849e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.5157e-08, ..., -1.4435e-08, -3.7253e-09, 2.3283e-10]], device='cuda:0') Epoch 448, bias, value: tensor([ 0.0031, 0.0059, 0.0164, -0.0041, 0.0292, -0.0042, 0.0005, -0.0049, 0.0133, -0.0176], device='cuda:0'), grad: tensor([-2.2491e-07, 6.6590e-08, 5.4482e-08, 3.4925e-09, 6.3796e-08, 8.6147e-09, 4.9127e-08, -3.7253e-09, 2.8871e-08, -2.9104e-08], device='cuda:0') 100 1e-05 changing lr epoch 447, time 246.65, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4175 re_mapping 0.0019 re_causal 0.0076 /// teacc 99.15 lr 0.00001000 Epoch 449, weight, value: tensor([[-0.2583, -0.0732, 0.2344, ..., 0.0357, -0.2658, -0.1784], [-0.2200, -0.3020, -0.0811, ..., -0.1462, -0.3940, -0.1333], [ 0.3313, -0.0858, -0.3407, ..., -0.2637, -0.3028, -0.2279], ..., [-0.1396, -0.0639, -0.4448, ..., -0.1565, 0.1513, 0.0586], [-0.2550, -0.1660, -0.0948, ..., -0.2793, -0.0691, -0.0901], [-0.1386, -0.1114, -0.1810, ..., 0.1046, 0.0257, -0.1948]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, 2.3283e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 2.3283e-10, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, -0.0000e+00, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., -0.0000e+00, 2.3283e-10, 0.0000e+00]], device='cuda:0') Epoch 449, bias, value: tensor([ 0.0031, 0.0059, 0.0164, -0.0041, 0.0292, -0.0042, 0.0006, -0.0050, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 1.6298e-09, 1.2619e-07, 2.5146e-08, -2.3283e-09, 1.3970e-09, 7.2177e-09, 3.4925e-09, -1.3015e-07, -2.0256e-08, 6.5193e-09], device='cuda:0') 100 1e-05 changing lr epoch 448, time 246.33, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3970 re_mapping 0.0019 re_causal 0.0073 /// teacc 99.18 lr 0.00001000 Epoch 450, weight, value: tensor([[-0.2583, -0.0732, 0.2344, ..., 0.0357, -0.2659, -0.1784], [-0.2200, -0.3020, -0.0811, ..., -0.1462, -0.3941, -0.1333], [ 0.3314, -0.0858, -0.3408, ..., -0.2638, -0.3028, -0.2279], ..., [-0.1397, -0.0639, -0.4449, ..., -0.1566, 0.1513, 0.0586], [-0.2550, -0.1660, -0.0949, ..., -0.2794, -0.0691, -0.0901], [-0.1387, -0.1114, -0.1811, ..., 0.1046, 0.0257, -0.1949]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.5611e-09, ..., 2.5611e-09, 4.6566e-10, 2.3283e-10], [ 2.3283e-10, 0.0000e+00, 9.3132e-10, ..., 5.8208e-09, 2.3283e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.0955e-09, 0.0000e+00, 0.0000e+00], ..., [ 2.3283e-10, 0.0000e+00, 4.6566e-10, ..., 7.4506e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 3.9581e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., -7.2177e-09, 2.3283e-10, 0.0000e+00]], device='cuda:0') Epoch 450, bias, value: tensor([ 0.0031, 0.0059, 0.0164, -0.0040, 0.0292, -0.0043, 0.0007, -0.0050, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 4.1910e-09, -7.2643e-08, 5.3551e-09, 1.8626e-08, -6.5193e-08, 9.5461e-09, 1.1874e-08, 9.3365e-08, 1.2573e-08, -2.0955e-09], device='cuda:0') 100 1e-05 changing lr epoch 449, time 245.83, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3925 re_mapping 0.0019 re_causal 0.0072 /// teacc 99.20 lr 0.00001000 Epoch 451, weight, value: tensor([[-0.2584, -0.0732, 0.2344, ..., 0.0357, -0.2659, -0.1784], [-0.2200, -0.3020, -0.0811, ..., -0.1463, -0.3942, -0.1334], [ 0.3315, -0.0858, -0.3409, ..., -0.2638, -0.3029, -0.2279], ..., [-0.1397, -0.0640, -0.4449, ..., -0.1567, 0.1513, 0.0586], [-0.2551, -0.1660, -0.0950, ..., -0.2795, -0.0691, -0.0901], [-0.1387, -0.1114, -0.1811, ..., 0.1046, 0.0257, -0.1949]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -9.3132e-10, 0.0000e+00, -2.3283e-10], [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, 1.8626e-09, 9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.3283e-10, 2.3283e-10], ..., [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, -3.0268e-09, -9.3132e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.5611e-09, ..., 2.3283e-10, 1.1642e-09, 2.3283e-10]], device='cuda:0') Epoch 451, bias, value: tensor([ 0.0031, 0.0059, 0.0164, -0.0039, 0.0293, -0.0044, 0.0007, -0.0050, 0.0133, -0.0176], device='cuda:0'), grad: tensor([-2.5611e-09, 9.0804e-09, 6.7521e-09, 1.6298e-09, 6.5193e-09, 3.0268e-09, 3.2596e-09, -3.5856e-08, 1.8626e-09, 2.7940e-08], device='cuda:0') 100 1e-05 changing lr epoch 450, time 245.95, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4114 re_mapping 0.0019 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 Epoch 452, weight, value: tensor([[-0.2584, -0.0732, 0.2344, ..., 0.0357, -0.2660, -0.1784], [-0.2200, -0.3020, -0.0812, ..., -0.1463, -0.3942, -0.1334], [ 0.3316, -0.0858, -0.3410, ..., -0.2638, -0.3029, -0.2279], ..., [-0.1398, -0.0640, -0.4450, ..., -0.1568, 0.1513, 0.0586], [-0.2551, -0.1660, -0.0951, ..., -0.2796, -0.0691, -0.0901], [-0.1387, -0.1114, -0.1811, ..., 0.1046, 0.0257, -0.1949]], device='cuda:0'), grad: tensor([[-4.6566e-10, 0.0000e+00, -5.5879e-09, ..., -3.4925e-09, 2.3283e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, 3.7253e-09, 2.3283e-10], [ 2.3283e-10, 0.0000e+00, 6.9849e-10, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], ..., [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, -4.6566e-09, -2.3283e-10], [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, 0.0000e+00, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 3.2596e-09, ..., 2.7940e-09, 6.9849e-10, 0.0000e+00]], device='cuda:0') Epoch 452, bias, value: tensor([ 0.0031, 0.0060, 0.0164, -0.0040, 0.0293, -0.0044, 0.0008, -0.0050, 0.0133, -0.0176], device='cuda:0'), grad: tensor([-1.3737e-08, 2.0955e-08, 2.7940e-09, 0.0000e+00, -2.3283e-10, -2.5611e-09, 2.7940e-09, -1.9791e-08, -1.8859e-08, 2.9569e-08], device='cuda:0') 100 1e-05 changing lr epoch 451, time 245.81, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4426 re_mapping 0.0019 re_causal 0.0078 /// teacc 99.22 lr 0.00001000 Epoch 453, weight, value: tensor([[-0.2584, -0.0732, 0.2345, ..., 0.0357, -0.2661, -0.1785], [-0.2201, -0.3021, -0.0812, ..., -0.1464, -0.3944, -0.1334], [ 0.3316, -0.0858, -0.3410, ..., -0.2639, -0.3029, -0.2279], ..., [-0.1399, -0.0640, -0.4450, ..., -0.1569, 0.1514, 0.0586], [-0.2552, -0.1660, -0.0952, ..., -0.2797, -0.0691, -0.0901], [-0.1387, -0.1114, -0.1811, ..., 0.1046, 0.0257, -0.1950]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.3283e-10, 0.0000e+00], [-1.3737e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.2340e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 1.8626e-09, 2.3283e-10, 0.0000e+00]], device='cuda:0') Epoch 453, bias, value: tensor([ 0.0031, 0.0060, 0.0164, -0.0040, 0.0294, -0.0043, 0.0007, -0.0050, 0.0133, -0.0176], device='cuda:0'), grad: tensor([ 2.3283e-10, -4.8894e-09, -2.7241e-08, -3.2363e-08, 4.6566e-10, 3.5390e-08, 1.8626e-09, 3.3993e-08, 2.7940e-09, 9.5461e-09], device='cuda:0') 100 1e-05 changing lr epoch 452, time 245.88, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4301 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 Epoch 454, weight, value: tensor([[-0.2584, -0.0732, 0.2345, ..., 0.0357, -0.2661, -0.1784], [-0.2201, -0.3021, -0.0813, ..., -0.1464, -0.3945, -0.1334], [ 0.3318, -0.0858, -0.3411, ..., -0.2641, -0.3029, -0.2278], ..., [-0.1400, -0.0641, -0.4451, ..., -0.1570, 0.1514, 0.0586], [-0.2552, -0.1660, -0.0953, ..., -0.2798, -0.0691, -0.0901], [-0.1387, -0.1114, -0.1811, ..., 0.1046, 0.0257, -0.1950]], device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, -6.5193e-09, ..., -3.9581e-09, 0.0000e+00, -2.3283e-10], [ 1.3970e-09, 4.6566e-10, 4.6566e-10, ..., 6.9849e-10, 2.3283e-10, 2.3283e-10], [-1.2107e-08, -2.3283e-10, 2.3283e-10, ..., 2.3283e-10, 0.0000e+00, 0.0000e+00], ..., [ 5.3551e-09, 9.3132e-10, 2.3283e-10, ..., 3.0268e-09, 6.9849e-10, 0.0000e+00], [ 8.3819e-09, 1.1642e-09, 4.6566e-10, ..., 9.3132e-10, 0.0000e+00, -6.9849e-10], [ 2.3283e-10, 2.3283e-10, 6.5193e-09, ..., -2.2119e-08, -6.9849e-10, 4.6566e-10]], device='cuda:0') Epoch 454, bias, value: tensor([ 0.0031, 0.0060, 0.0164, -0.0039, 0.0294, -0.0043, 0.0007, -0.0050, 0.0133, -0.0176], device='cuda:0'), grad: tensor([-1.2107e-08, 7.4506e-09, -2.8871e-08, -6.3563e-08, 5.3085e-08, 4.0745e-08, 6.9849e-10, 3.7486e-08, 1.8394e-08, -3.7951e-08], device='cuda:0') 100 1e-05 changing lr epoch 453, time 245.87, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4589 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 Epoch 455, weight, value: tensor([[-0.2585, -0.0732, 0.2346, ..., 0.0357, -0.2663, -0.1785], [-0.2201, -0.3021, -0.0813, ..., -0.1465, -0.3946, -0.1335], [ 0.3319, -0.0858, -0.3412, ..., -0.2642, -0.3030, -0.2278], ..., [-0.1402, -0.0641, -0.4452, ..., -0.1572, 0.1515, 0.0586], [-0.2553, -0.1660, -0.0955, ..., -0.2799, -0.0691, -0.0902], [-0.1388, -0.1114, -0.1812, ..., 0.1046, 0.0256, -0.1950]], device='cuda:0'), grad: tensor([[ 1.1642e-09, 2.3283e-10, -5.0478e-07, ..., -1.2945e-07, 1.6298e-09, -4.0047e-08], [ 1.3970e-09, 4.6566e-10, 2.5611e-09, ..., 9.3132e-10, 2.3283e-10, 2.3283e-10], [-1.1642e-09, 2.3283e-10, 1.4668e-08, ..., 3.7253e-09, 4.6566e-10, 1.1642e-09], ..., [ 3.0268e-09, 1.1642e-09, 1.1642e-09, ..., 4.6566e-10, 0.0000e+00, 2.3283e-10], [ 9.3132e-10, 2.3283e-10, 2.5146e-08, ..., 6.5193e-09, 2.3283e-10, 1.8626e-09], [ 2.3283e-10, 2.3283e-10, 1.2503e-07, ..., 3.4692e-08, 2.3283e-10, 8.8476e-09]], device='cuda:0') Epoch 455, bias, value: tensor([ 0.0031, 0.0059, 0.0165, -0.0039, 0.0294, -0.0044, 0.0008, -0.0050, 0.0132, -0.0176], device='cuda:0'), grad: tensor([-1.0440e-06, 1.2340e-08, 3.3295e-08, -3.3295e-08, 9.4995e-08, 2.9104e-08, 5.7789e-07, 1.8626e-08, 5.8673e-08, 2.6939e-07], device='cuda:0') 100 1e-05 changing lr epoch 454, time 245.79, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4093 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.19 lr 0.00001000 Epoch 456, weight, value: tensor([[-0.2585, -0.0732, 0.2347, ..., 0.0357, -0.2663, -0.1784], [-0.2202, -0.3021, -0.0814, ..., -0.1465, -0.3948, -0.1335], [ 0.3320, -0.0858, -0.3414, ..., -0.2642, -0.3030, -0.2278], ..., [-0.1402, -0.0642, -0.4453, ..., -0.1573, 0.1515, 0.0586], [-0.2554, -0.1661, -0.0958, ..., -0.2801, -0.0691, -0.0902], [-0.1389, -0.1114, -0.1812, ..., 0.1046, 0.0256, -0.1951]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-10, 2.3283e-10], [ 0.0000e+00, 0.0000e+00, -0.0000e+00, ..., 2.3283e-10, 1.8626e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 1.3970e-09, 4.6566e-10], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 9.3132e-10, -2.3283e-09, 4.6566e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.5611e-09, 1.1642e-09], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -3.9581e-09, 1.3970e-09, 2.3283e-10]], device='cuda:0') Epoch 456, bias, value: tensor([ 0.0031, 0.0059, 0.0164, -0.0039, 0.0294, -0.0043, 0.0006, -0.0050, 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 5.1223e-09, 1.2107e-08, 1.3039e-08, 2.4005e-07, 7.0082e-08, -1.6550e-06, 1.3029e-06, -1.2340e-08, 2.4214e-08, 6.2864e-09], device='cuda:0') 100 1e-05 changing lr epoch 455, time 246.50, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4372 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.17 lr 0.00001000 Epoch 457, weight, value: tensor([[-0.2585, -0.0732, 0.2348, ..., 0.0357, -0.2664, -0.1784], [-0.2202, -0.3021, -0.0815, ..., -0.1466, -0.3949, -0.1335], [ 0.3321, -0.0858, -0.3415, ..., -0.2643, -0.3031, -0.2278], ..., [-0.1403, -0.0642, -0.4453, ..., -0.1574, 0.1516, 0.0585], [-0.2554, -0.1661, -0.0959, ..., -0.2802, -0.0692, -0.0902], [-0.1389, -0.1114, -0.1813, ..., 0.1046, 0.0256, -0.1951]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 9.3132e-10, -3.7253e-09, ..., 1.8626e-09, 5.1223e-09, 4.6566e-10], [ 2.3283e-09, 4.1910e-09, 4.1910e-09, ..., 2.7940e-09, 3.7253e-09, 1.3970e-09], [ 1.3970e-09, 2.7940e-09, 3.7253e-09, ..., 0.0000e+00, 9.3132e-10, 1.3970e-09], ..., [ 1.3504e-08, 2.6077e-08, 0.0000e+00, ..., 4.6566e-09, -1.1176e-08, 0.0000e+00], [ 9.3132e-10, 1.3970e-09, 9.3132e-10, ..., 9.3132e-10, 4.6566e-10, 9.3132e-10], [ 4.6566e-10, 9.3132e-10, 1.8626e-09, ..., 2.1840e-07, 1.3970e-09, 4.6566e-10]], device='cuda:0') Epoch 457, bias, value: tensor([ 0.0031, 0.0059, 0.0164, -0.0039, 0.0294, -0.0043, 0.0006, -0.0050, 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 1.3597e-07, 1.4110e-07, 4.7032e-08, -2.6310e-07, -4.6706e-07, 1.0710e-07, -5.8208e-08, -1.4994e-07, 7.9162e-09, 4.9453e-07], device='cuda:0') 100 1e-05 changing lr epoch 456, time 246.30, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4274 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.16 lr 0.00001000 Epoch 458, weight, value: tensor([[-0.2586, -0.0732, 0.2349, ..., 0.0357, -0.2665, -0.1784], [-0.2203, -0.3021, -0.0815, ..., -0.1466, -0.3950, -0.1336], [ 0.3323, -0.0858, -0.3416, ..., -0.2644, -0.3031, -0.2278], ..., [-0.1404, -0.0643, -0.4454, ..., -0.1575, 0.1517, 0.0586], [-0.2555, -0.1661, -0.0960, ..., -0.2803, -0.0692, -0.0902], [-0.1389, -0.1114, -0.1813, ..., 0.1046, 0.0256, -0.1952]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, -9.3132e-10, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, 4.6566e-10, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 1.8626e-09, ..., 3.2596e-09, 4.6566e-10, 0.0000e+00]], device='cuda:0') Epoch 458, bias, value: tensor([ 0.0031, 0.0059, 0.0165, -0.0039, 0.0295, -0.0043, 0.0006, -0.0050, 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 9.3132e-10, -4.6566e-10, 1.8626e-09, -1.1642e-08, -2.3283e-09, 1.4435e-08, -9.3132e-09, 2.3283e-09, 4.6566e-10, 1.5367e-08], device='cuda:0') 100 1e-05 changing lr epoch 457, time 246.40, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4260 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 Epoch 459, weight, value: tensor([[-0.2586, -0.0732, 0.2349, ..., 0.0357, -0.2666, -0.1784], [-0.2203, -0.3022, -0.0816, ..., -0.1468, -0.3951, -0.1336], [ 0.3325, -0.0858, -0.3417, ..., -0.2644, -0.3031, -0.2278], ..., [-0.1405, -0.0643, -0.4454, ..., -0.1576, 0.1517, 0.0586], [-0.2556, -0.1661, -0.0961, ..., -0.2804, -0.0692, -0.0903], [-0.1390, -0.1114, -0.1813, ..., 0.1046, 0.0255, -0.1952]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.6566e-10, ..., -4.6566e-10, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 1.3970e-09, 0.0000e+00, 3.2596e-09, ..., 0.0000e+00, 2.3283e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., -7.4506e-09, 0.0000e+00, 4.6566e-10]], device='cuda:0') Epoch 459, bias, value: tensor([ 0.0031, 0.0059, 0.0166, -0.0040, 0.0296, -0.0042, 0.0005, -0.0051, 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 5.1223e-09, -4.6566e-10, 2.3283e-09, 3.0734e-08, 1.8161e-08, 3.7253e-08, -7.4506e-08, 1.0710e-08, -3.3993e-08, 1.9092e-08], device='cuda:0') 100 1e-05 changing lr epoch 458, time 246.54, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4030 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 Epoch 460, weight, value: tensor([[-0.2587, -0.0732, 0.2350, ..., 0.0357, -0.2666, -0.1784], [-0.2204, -0.3022, -0.0816, ..., -0.1467, -0.3952, -0.1336], [ 0.3326, -0.0859, -0.3418, ..., -0.2645, -0.3032, -0.2278], ..., [-0.1406, -0.0644, -0.4455, ..., -0.1578, 0.1518, 0.0586], [-0.2556, -0.1661, -0.0963, ..., -0.2805, -0.0692, -0.0903], [-0.1390, -0.1114, -0.1814, ..., 0.1046, 0.0255, -0.1953]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 7.9162e-09, ..., 4.6566e-10, 3.2596e-09, 0.0000e+00], [ 1.8626e-09, 4.6566e-10, 1.3970e-09, ..., 4.6566e-10, 1.3970e-09, 0.0000e+00], [-2.7940e-09, -4.6566e-10, 4.6566e-10, ..., 0.0000e+00, 4.6566e-10, 0.0000e+00], ..., [ 3.2596e-09, 4.6566e-10, 0.0000e+00, ..., 9.3132e-10, -1.3970e-09, -0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, -6.0536e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-09, ..., -3.2596e-09, 2.7940e-09, 0.0000e+00]], device='cuda:0') Epoch 460, bias, value: tensor([ 0.0031, 0.0059, 0.0166, -0.0040, 0.0295, -0.0041, 0.0004, -0.0051, 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 3.9116e-08, 1.9092e-08, 0.0000e+00, -1.8626e-09, 5.3551e-08, 6.2864e-08, -1.4296e-07, 3.2596e-09, -3.2131e-08, 1.1642e-08], device='cuda:0') 100 1e-05 changing lr epoch 459, time 246.63, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4065 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 Epoch 461, weight, value: tensor([[-0.2587, -0.0732, 0.2351, ..., 0.0357, -0.2666, -0.1784], [-0.2204, -0.3022, -0.0818, ..., -0.1468, -0.3954, -0.1337], [ 0.3327, -0.0858, -0.3419, ..., -0.2645, -0.3032, -0.2278], ..., [-0.1407, -0.0644, -0.4455, ..., -0.1579, 0.1519, 0.0586], [-0.2557, -0.1661, -0.0964, ..., -0.2806, -0.0692, -0.0903], [-0.1390, -0.1114, -0.1814, ..., 0.1046, 0.0255, -0.1953]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.7253e-09, ..., -1.3970e-09, 0.0000e+00, 0.0000e+00], [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 1.3970e-09, 0.0000e+00, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 9.3132e-10, 0.0000e+00, ..., 1.1176e-08, 4.6566e-10, 4.6566e-10], [ 1.3970e-09, 9.3132e-10, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [ 4.6566e-10, 4.6566e-10, 3.2596e-09, ..., 1.1642e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 461, bias, value: tensor([ 0.0031, 0.0059, 0.0166, -0.0039, 0.0295, -0.0041, 0.0003, -0.0051, 0.0132, -0.0176], device='cuda:0'), grad: tensor([-4.6566e-09, 2.0023e-08, 3.7253e-09, -8.5216e-08, -7.5903e-08, 4.0047e-08, 1.8626e-09, 4.6100e-08, 1.3970e-08, 4.5635e-08], device='cuda:0') 100 1e-05 changing lr epoch 460, time 246.58, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4236 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.22 lr 0.00001000 Epoch 462, weight, value: tensor([[-0.2587, -0.0732, 0.2352, ..., 0.0357, -0.2667, -0.1784], [-0.2204, -0.3022, -0.0819, ..., -0.1469, -0.3955, -0.1337], [ 0.3328, -0.0859, -0.3420, ..., -0.2646, -0.3033, -0.2278], ..., [-0.1408, -0.0644, -0.4456, ..., -0.1580, 0.1519, 0.0586], [-0.2557, -0.1661, -0.0965, ..., -0.2807, -0.0692, -0.0903], [-0.1390, -0.1114, -0.1815, ..., 0.1046, 0.0255, -0.1953]], device='cuda:0'), grad: tensor([[0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, 0.0000e+00, 0.0000e+00], [0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 1.1642e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 462, bias, value: tensor([ 0.0031, 0.0059, 0.0166, -0.0039, 0.0296, -0.0041, 0.0003, -0.0051, 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 4.6566e-10, -9.7789e-09, 2.3283e-09, -9.3132e-10, -3.5390e-08, 4.6566e-10, 2.7940e-09, 1.6764e-08, 6.0536e-09, 3.4459e-08], device='cuda:0') 100 1e-05 changing lr epoch 461, time 246.45, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4600 re_mapping 0.0018 re_causal 0.0075 /// teacc 99.21 lr 0.00001000 Epoch 463, weight, value: tensor([[-0.2587, -0.0732, 0.2353, ..., 0.0357, -0.2667, -0.1783], [-0.2205, -0.3022, -0.0821, ..., -0.1470, -0.3956, -0.1338], [ 0.3329, -0.0859, -0.3421, ..., -0.2646, -0.3033, -0.2278], ..., [-0.1409, -0.0645, -0.4456, ..., -0.1581, 0.1520, 0.0586], [-0.2558, -0.1661, -0.0965, ..., -0.2807, -0.0692, -0.0903], [-0.1391, -0.1114, -0.1815, ..., 0.1046, 0.0254, -0.1954]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.1874e-07, ..., -3.1199e-08, 0.0000e+00, -2.1886e-08], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.1910e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 4.6566e-10, 9.3132e-10, ..., 6.0536e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.5367e-08, ..., 2.7940e-08, 0.0000e+00, 1.8626e-09]], device='cuda:0') Epoch 463, bias, value: tensor([ 0.0031, 0.0059, 0.0167, -0.0039, 0.0296, -0.0041, 0.0003, -0.0051, 0.0132, -0.0176], device='cuda:0'), grad: tensor([-3.0687e-07, 1.2573e-08, 9.3132e-10, 1.7695e-08, -7.7765e-08, 4.4703e-08, 2.0629e-07, 2.3749e-08, 4.1910e-09, 8.5216e-08], device='cuda:0') 100 1e-05 changing lr epoch 462, time 246.57, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4144 re_mapping 0.0018 re_causal 0.0073 /// teacc 99.22 lr 0.00001000 Epoch 464, weight, value: tensor([[-0.2587, -0.0732, 0.2353, ..., 0.0357, -0.2668, -0.1783], [-0.2205, -0.3022, -0.0821, ..., -0.1472, -0.3957, -0.1338], [ 0.3330, -0.0859, -0.3421, ..., -0.2647, -0.3033, -0.2278], ..., [-0.1410, -0.0645, -0.4457, ..., -0.1583, 0.1521, 0.0586], [-0.2559, -0.1661, -0.0966, ..., -0.2809, -0.0692, -0.0904], [-0.1391, -0.1114, -0.1816, ..., 0.1046, 0.0254, -0.1954]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.2596e-09, ..., -1.3970e-09, 0.0000e+00, -0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 4.6566e-10, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 9.3132e-10, 0.0000e+00, ..., 1.8626e-09, -4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., -4.6566e-10, -2.7940e-09, 0.0000e+00], [ 0.0000e+00, 4.6566e-10, 2.3283e-09, ..., -1.2573e-08, 1.3970e-09, 0.0000e+00]], device='cuda:0') Epoch 464, bias, value: tensor([ 0.0031, 0.0059, 0.0166, -0.0040, 0.0297, -0.0041, 0.0003, -0.0051, 0.0132, -0.0176], device='cuda:0'), grad: tensor([-5.1223e-09, 1.0245e-08, 9.3132e-09, -1.2200e-07, 3.4925e-08, 1.0943e-07, 4.1910e-09, 1.9092e-08, -1.9092e-08, -2.6077e-08], device='cuda:0') 100 1e-05 changing lr epoch 463, time 246.36, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4420 re_mapping 0.0018 re_causal 0.0076 /// teacc 99.20 lr 0.00001000 Epoch 465, weight, value: tensor([[-0.2588, -0.0732, 0.2354, ..., 0.0357, -0.2669, -0.1783], [-0.2206, -0.3022, -0.0821, ..., -0.1472, -0.3959, -0.1338], [ 0.3331, -0.0859, -0.3422, ..., -0.2647, -0.3033, -0.2278], ..., [-0.1410, -0.0645, -0.4457, ..., -0.1584, 0.1521, 0.0587], [-0.2559, -0.1662, -0.0967, ..., -0.2810, -0.0692, -0.0904], [-0.1391, -0.1114, -0.1816, ..., 0.1046, 0.0254, -0.1954]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.9116e-08, ..., -7.9162e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [-3.7253e-09, 0.0000e+00, 9.3132e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 1.8626e-09, 0.0000e+00, 0.0000e+00], [ 4.1910e-09, 0.0000e+00, 3.7253e-09, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.7695e-08, ..., -2.1886e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 465, bias, value: tensor([ 0.0031, 0.0058, 0.0166, -0.0039, 0.0297, -0.0041, 0.0003, -0.0051, 0.0131, -0.0176], device='cuda:0'), grad: tensor([-1.2713e-07, -8.0559e-08, -9.3132e-10, 6.5193e-09, 6.1933e-08, 2.2817e-08, 2.0023e-08, 2.7008e-08, 8.6147e-08, -6.0536e-09], device='cuda:0') 100 1e-05 changing lr epoch 464, time 246.23, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4153 re_mapping 0.0018 re_causal 0.0075 /// teacc 99.20 lr 0.00001000 Epoch 466, weight, value: tensor([[-0.2588, -0.0732, 0.2355, ..., 0.0357, -0.2669, -0.1783], [-0.2206, -0.3022, -0.0822, ..., -0.1472, -0.3960, -0.1338], [ 0.3332, -0.0859, -0.3422, ..., -0.2647, -0.3034, -0.2279], ..., [-0.1411, -0.0645, -0.4457, ..., -0.1585, 0.1522, 0.0587], [-0.2560, -0.1662, -0.0968, ..., -0.2810, -0.0692, -0.0904], [-0.1392, -0.1114, -0.1816, ..., 0.1046, 0.0254, -0.1955]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 2.2817e-08, 0.0000e+00, 0.0000e+00], [ 3.5390e-08, 0.0000e+00, -4.6566e-10, ..., 5.1223e-09, 4.6566e-10, 0.0000e+00], [-3.8184e-08, 0.0000e+00, 0.0000e+00, ..., 7.9162e-09, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 4.6566e-10, 4.6566e-10, ..., 7.4506e-09, -9.3132e-10, 0.0000e+00], [ 1.3970e-09, 4.6566e-10, 0.0000e+00, ..., 1.3970e-09, 0.0000e+00, 0.0000e+00], [ 4.6566e-10, 4.6566e-10, 9.3132e-10, ..., 3.7719e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 466, bias, value: tensor([ 0.0031, 0.0058, 0.0166, -0.0039, 0.0298, -0.0041, 0.0003, -0.0051, 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 6.0536e-08, 1.0198e-07, -7.6368e-08, -6.5193e-09, -2.2957e-07, 1.4435e-08, 3.4925e-08, 2.6543e-08, -6.0536e-09, 9.7789e-08], device='cuda:0') 100 1e-05 changing lr epoch 465, time 246.27, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4099 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 Epoch 467, weight, value: tensor([[-0.2588, -0.0732, 0.2355, ..., 0.0357, -0.2670, -0.1783], [-0.2206, -0.3022, -0.0822, ..., -0.1473, -0.3961, -0.1338], [ 0.3333, -0.0859, -0.3423, ..., -0.2649, -0.3034, -0.2278], ..., [-0.1411, -0.0646, -0.4457, ..., -0.1586, 0.1523, 0.0587], [-0.2560, -0.1662, -0.0968, ..., -0.2811, -0.0692, -0.0904], [-0.1392, -0.1114, -0.1816, ..., 0.1046, 0.0253, -0.1955]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.3993e-08, ..., -1.7695e-08, 0.0000e+00, -4.6566e-10], [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 5.5879e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.3283e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.7789e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.7789e-09, ..., -7.4506e-09, 1.8626e-09, 4.6566e-10]], device='cuda:0') Epoch 467, bias, value: tensor([ 0.0031, 0.0058, 0.0166, -0.0040, 0.0298, -0.0041, 0.0002, -0.0051, 0.0131, -0.0176], device='cuda:0'), grad: tensor([-7.2177e-08, 1.6298e-08, 3.7253e-09, 3.2596e-09, 3.7253e-09, 9.7789e-09, 4.2375e-08, 8.8476e-09, 3.6787e-08, -3.5856e-08], device='cuda:0') 100 1e-05 changing lr epoch 466, time 246.74, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4247 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.20 lr 0.00001000 Epoch 468, weight, value: tensor([[-0.2588, -0.0732, 0.2355, ..., 0.0357, -0.2671, -0.1783], [-0.2207, -0.3022, -0.0822, ..., -0.1474, -0.3962, -0.1338], [ 0.3334, -0.0859, -0.3423, ..., -0.2649, -0.3034, -0.2279], ..., [-0.1412, -0.0646, -0.4458, ..., -0.1587, 0.1523, 0.0587], [-0.2560, -0.1662, -0.0969, ..., -0.2812, -0.0692, -0.0905], [-0.1392, -0.1114, -0.1817, ..., 0.1046, 0.0253, -0.1955]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-09, 4.6566e-10, 0.0000e+00], [-6.0536e-09, 0.0000e+00, 0.0000e+00, ..., 2.9802e-08, 0.0000e+00, 0.0000e+00], ..., [ 1.4901e-08, 0.0000e+00, 0.0000e+00, ..., 4.1910e-09, 0.0000e+00, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 3.7253e-09, 0.0000e+00, 4.6566e-10, ..., -1.5367e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 468, bias, value: tensor([ 0.0031, 0.0058, 0.0166, -0.0040, 0.0299, -0.0041, 0.0002, -0.0051, 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 1.4435e-08, 1.0710e-08, 3.4459e-08, -5.1223e-09, -3.2596e-08, 6.5193e-09, 9.3132e-10, 4.7963e-08, 3.2596e-09, -7.1712e-08], device='cuda:0') 100 1e-05 changing lr epoch 467, time 246.33, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3952 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.16 lr 0.00001000 Epoch 469, weight, value: tensor([[-0.2589, -0.0732, 0.2356, ..., 0.0357, -0.2671, -0.1783], [-0.2207, -0.3022, -0.0823, ..., -0.1475, -0.3963, -0.1338], [ 0.3335, -0.0859, -0.3423, ..., -0.2650, -0.3035, -0.2279], ..., [-0.1412, -0.0646, -0.4458, ..., -0.1588, 0.1524, 0.0587], [-0.2560, -0.1662, -0.0970, ..., -0.2813, -0.0692, -0.0905], [-0.1392, -0.1114, -0.1817, ..., 0.1046, 0.0253, -0.1955]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 4.6566e-10, 4.6566e-10, 4.6566e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 4.6566e-10, -0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., -3.7253e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 469, bias, value: tensor([ 0.0031, 0.0058, 0.0166, -0.0040, 0.0299, -0.0041, 0.0003, -0.0051, 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 4.1910e-09, -5.1223e-09, 0.0000e+00, 1.1176e-08, 7.4506e-09, -1.0710e-08, -1.8626e-09, 8.3819e-09, 4.6566e-10, -7.4506e-09], device='cuda:0') 100 1e-05 changing lr epoch 468, time 246.26, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4212 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 Epoch 470, weight, value: tensor([[-0.2589, -0.0732, 0.2357, ..., 0.0357, -0.2672, -0.1784], [-0.2207, -0.3022, -0.0823, ..., -0.1476, -0.3965, -0.1338], [ 0.3336, -0.0859, -0.3424, ..., -0.2650, -0.3035, -0.2278], ..., [-0.1413, -0.0646, -0.4458, ..., -0.1591, 0.1525, 0.0587], [-0.2561, -0.1662, -0.0970, ..., -0.2814, -0.0692, -0.0905], [-0.1393, -0.1114, -0.1817, ..., 0.1046, 0.0253, -0.1955]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 4.6566e-10, 4.6566e-10, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 9.3132e-10, -0.0000e+00], [-1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.8626e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -9.3132e-10, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 4.6566e-10, ..., 4.1910e-09, 4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -1.0245e-08, 9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 470, bias, value: tensor([ 0.0031, 0.0058, 0.0166, -0.0040, 0.0299, -0.0042, 0.0003, -0.0051, 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 3.2596e-09, 3.2596e-09, -1.8626e-09, 2.7940e-09, 1.3039e-08, 3.5856e-08, -2.9802e-08, 3.2596e-09, 7.9162e-09, -2.6543e-08], device='cuda:0') 100 1e-05 changing lr epoch 469, time 245.93, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4597 re_mapping 0.0018 re_causal 0.0075 /// teacc 99.17 lr 0.00001000 Epoch 471, weight, value: tensor([[-0.2589, -0.0732, 0.2357, ..., 0.0357, -0.2673, -0.1784], [-0.2208, -0.3023, -0.0823, ..., -0.1477, -0.3966, -0.1338], [ 0.3337, -0.0859, -0.3424, ..., -0.2650, -0.3035, -0.2278], ..., [-0.1414, -0.0647, -0.4458, ..., -0.1592, 0.1526, 0.0587], [-0.2561, -0.1662, -0.0971, ..., -0.2815, -0.0693, -0.0905], [-0.1393, -0.1114, -0.1818, ..., 0.1046, 0.0252, -0.1956]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-09, ..., 9.3132e-10, 9.3132e-10, 4.6566e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8626e-09, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -2.7940e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 5.5879e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, -4.6566e-10, 0.0000e+00, ..., 1.4901e-08, 9.3132e-10, 0.0000e+00]], device='cuda:0') Epoch 471, bias, value: tensor([ 0.0031, 0.0058, 0.0165, -0.0040, 0.0299, -0.0043, 0.0003, -0.0051, 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 7.4506e-09, 6.5193e-09, 4.6566e-10, 2.9337e-08, -8.8010e-08, -3.5390e-08, 1.6764e-08, -7.4506e-09, 1.8161e-08, 5.8673e-08], device='cuda:0') 100 1e-05 changing lr epoch 470, time 246.46, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4012 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 Epoch 472, weight, value: tensor([[-0.2590, -0.0732, 0.2358, ..., 0.0357, -0.2673, -0.1784], [-0.2208, -0.3023, -0.0823, ..., -0.1477, -0.3967, -0.1339], [ 0.3338, -0.0859, -0.3424, ..., -0.2651, -0.3036, -0.2278], ..., [-0.1415, -0.0647, -0.4459, ..., -0.1594, 0.1526, 0.0587], [-0.2561, -0.1662, -0.0971, ..., -0.2816, -0.0693, -0.0905], [-0.1393, -0.1114, -0.1818, ..., 0.1046, 0.0252, -0.1956]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.8673e-08, ..., -3.5856e-08, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.8626e-09, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 0.0000e+00, 2.3283e-09, ..., 2.3283e-09, -4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 8.8476e-09, ..., 5.5879e-09, 0.0000e+00, 4.6566e-10], [ 0.0000e+00, 0.0000e+00, 3.5390e-08, ..., 2.2352e-08, 4.6566e-10, 4.6566e-10]], device='cuda:0') Epoch 472, bias, value: tensor([ 0.0031, 0.0059, 0.0165, -0.0039, 0.0300, -0.0043, 0.0002, -0.0052, 0.0131, -0.0176], device='cuda:0'), grad: tensor([-1.4808e-07, 6.5193e-09, 4.6566e-09, -9.7789e-09, -1.8626e-09, 3.2596e-09, 2.4680e-08, 1.0245e-08, 2.4680e-08, 9.2201e-08], device='cuda:0') 100 1e-05 changing lr epoch 471, time 246.39, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4211 re_mapping 0.0019 re_causal 0.0077 /// teacc 99.21 lr 0.00001000 Epoch 473, weight, value: tensor([[-0.2590, -0.0732, 0.2358, ..., 0.0357, -0.2674, -0.1785], [-0.2208, -0.3023, -0.0823, ..., -0.1478, -0.3968, -0.1339], [ 0.3340, -0.0859, -0.3424, ..., -0.2651, -0.3036, -0.2278], ..., [-0.1416, -0.0647, -0.4459, ..., -0.1595, 0.1527, 0.0587], [-0.2562, -0.1662, -0.0972, ..., -0.2817, -0.0693, -0.0905], [-0.1393, -0.1113, -0.1818, ..., 0.1046, 0.0252, -0.1956]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 9.3132e-10, 4.6566e-10, 4.6566e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.8161e-08, -2.3283e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., -2.8871e-08, 1.8626e-09, 0.0000e+00]], device='cuda:0') Epoch 473, bias, value: tensor([ 0.0031, 0.0058, 0.0167, -0.0039, 0.0300, -0.0043, 0.0003, -0.0052, 0.0131, -0.0176], device='cuda:0'), grad: tensor([-9.3132e-10, 5.1223e-09, 0.0000e+00, 9.3132e-10, 2.8405e-08, 8.3819e-09, -1.2107e-08, 1.3970e-08, 1.3970e-09, -3.3062e-08], device='cuda:0') 100 1e-05 changing lr epoch 472, time 246.70, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4410 re_mapping 0.0018 re_causal 0.0074 /// teacc 99.20 lr 0.00001000 Epoch 474, weight, value: tensor([[-0.2590, -0.0732, 0.2359, ..., 0.0357, -0.2675, -0.1785], [-0.2209, -0.3023, -0.0824, ..., -0.1478, -0.3969, -0.1339], [ 0.3342, -0.0860, -0.3425, ..., -0.2651, -0.3036, -0.2278], ..., [-0.1417, -0.0648, -0.4459, ..., -0.1596, 0.1528, 0.0587], [-0.2562, -0.1662, -0.0972, ..., -0.2817, -0.0693, -0.0906], [-0.1393, -0.1113, -0.1819, ..., 0.1046, 0.0252, -0.1956]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -6.5193e-09, ..., -1.8626e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.3970e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], ..., [ 4.6566e-10, 4.6566e-10, 0.0000e+00, ..., 2.7940e-09, 4.6566e-10, 0.0000e+00], [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 1.8626e-09, 4.6566e-10, 0.0000e+00], [ 0.0000e+00, -1.3970e-09, 2.7940e-09, ..., -1.0245e-08, -2.3283e-09, 0.0000e+00]], device='cuda:0') Epoch 474, bias, value: tensor([ 0.0031, 0.0058, 0.0167, -0.0038, 0.0301, -0.0043, 0.0002, -0.0052, 0.0131, -0.0176], device='cuda:0'), grad: tensor([-5.5879e-09, -3.5856e-08, 4.1910e-09, -1.3504e-08, -7.4506e-09, -2.3935e-07, 2.4308e-07, 5.3085e-08, 1.3504e-08, -7.4506e-09], device='cuda:0') 100 1e-05 changing lr epoch 473, time 246.20, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4242 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.19 lr 0.00001000 Epoch 475, weight, value: tensor([[-0.2591, -0.0732, 0.2360, ..., 0.0357, -0.2675, -0.1785], [-0.2209, -0.3023, -0.0824, ..., -0.1479, -0.3970, -0.1339], [ 0.3343, -0.0859, -0.3425, ..., -0.2652, -0.3036, -0.2278], ..., [-0.1418, -0.0648, -0.4460, ..., -0.1597, 0.1529, 0.0587], [-0.2563, -0.1662, -0.0973, ..., -0.2819, -0.0693, -0.0906], [-0.1394, -0.1113, -0.1819, ..., 0.1046, 0.0251, -0.1956]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.3062e-08, ..., -1.6298e-08, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., 2.3283e-09, 1.3970e-09, 0.0000e+00], [-1.8626e-09, 0.0000e+00, 2.3283e-09, ..., 9.3132e-10, 0.0000e+00, 0.0000e+00], ..., [ 1.3970e-09, 0.0000e+00, 4.6566e-10, ..., 1.8626e-09, -1.2573e-08, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.3970e-09, ..., 2.3283e-09, 2.3283e-09, 4.6566e-10], [ 0.0000e+00, 0.0000e+00, 2.4214e-08, ..., 1.4901e-08, 1.0710e-08, 1.8626e-09]], device='cuda:0') Epoch 475, bias, value: tensor([ 3.1182e-03, 5.8945e-03, 1.6783e-02, -3.8210e-03, 3.0269e-02, -4.2500e-03, -1.5174e-06, -5.2641e-03, 1.3048e-02, -1.7632e-02], device='cuda:0'), grad: tensor([-6.7987e-08, 1.3504e-08, 2.3283e-09, 2.3283e-09, -8.8476e-09, -3.6322e-08, 2.5611e-08, -4.5169e-08, 3.1665e-08, 1.0245e-07], device='cuda:0') 100 1e-05 changing lr epoch 474, time 246.48, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4262 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 Epoch 476, weight, value: tensor([[-0.2591, -0.0732, 0.2361, ..., 0.0357, -0.2676, -0.1786], [-0.2210, -0.3023, -0.0824, ..., -0.1479, -0.3971, -0.1339], [ 0.3345, -0.0860, -0.3426, ..., -0.2652, -0.3036, -0.2278], ..., [-0.1419, -0.0648, -0.4460, ..., -0.1600, 0.1529, 0.0587], [-0.2564, -0.1663, -0.0974, ..., -0.2820, -0.0693, -0.0906], [-0.1394, -0.1114, -0.1820, ..., 0.1046, 0.0251, -0.1957]], device='cuda:0'), grad: tensor([[ 9.3132e-10, 0.0000e+00, -3.1665e-08, ..., -2.0023e-08, 0.0000e+00, -0.0000e+00], [ 4.6566e-10, 0.0000e+00, 1.8626e-09, ..., 3.7253e-09, 4.6566e-10, 0.0000e+00], [-6.5193e-09, 0.0000e+00, 4.1910e-09, ..., 2.7940e-09, 4.6566e-10, 0.0000e+00], ..., [ 3.7253e-09, 0.0000e+00, 1.3970e-09, ..., 3.7253e-09, -1.3970e-09, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 7.4506e-09, ..., 5.5879e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.1176e-08, ..., 4.5169e-08, -1.3970e-09, 0.0000e+00]], device='cuda:0') Epoch 476, bias, value: tensor([ 3.1198e-03, 6.0766e-03, 1.6831e-02, -3.7820e-03, 3.0331e-02, -4.2687e-03, -2.8614e-05, -5.4375e-03, 1.3033e-02, -1.7635e-02], device='cuda:0'), grad: tensor([-8.9873e-08, 1.4901e-08, -4.6566e-10, 1.4901e-08, -1.0012e-07, -4.6566e-10, 6.5193e-09, 1.6298e-08, 2.8405e-08, 1.2387e-07], device='cuda:0') 100 1e-05 changing lr epoch 475, time 246.37, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4242 re_mapping 0.0018 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 Epoch 477, weight, value: tensor([[-0.2592, -0.0732, 0.2361, ..., 0.0357, -0.2677, -0.1786], [-0.2211, -0.3023, -0.0824, ..., -0.1480, -0.3972, -0.1339], [ 0.3347, -0.0860, -0.3426, ..., -0.2652, -0.3036, -0.2278], ..., [-0.1420, -0.0648, -0.4461, ..., -0.1601, 0.1530, 0.0587], [-0.2564, -0.1663, -0.0974, ..., -0.2820, -0.0693, -0.0906], [-0.1394, -0.1113, -0.1821, ..., 0.1046, 0.0251, -0.1957]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [ 4.6566e-10, 2.3283e-10, 0.0000e+00, ..., 9.3132e-10, 0.0000e+00, 2.3283e-10], [-5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.3970e-09, 2.3283e-10, 0.0000e+00, ..., 3.9581e-09, 0.0000e+00, 2.3283e-10], [ 4.6566e-10, 0.0000e+00, 2.3283e-10, ..., 4.6566e-10, 0.0000e+00, -6.9849e-10], [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., -1.0477e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 477, bias, value: tensor([ 3.1199e-03, 6.0670e-03, 1.6961e-02, -3.6773e-03, 3.0383e-02, -4.4487e-03, 6.0416e-05, -5.4583e-03, 1.3052e-02, -1.7636e-02], device='cuda:0'), grad: tensor([ 2.3283e-09, 4.4238e-09, -9.7789e-09, 3.2596e-08, 1.9325e-08, 7.9162e-09, 3.2596e-09, 3.0501e-08, -2.7707e-08, -3.8184e-08], device='cuda:0') 100 1e-05 changing lr epoch 476, time 246.74, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4072 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.20 lr 0.00001000 Epoch 478, weight, value: tensor([[-0.2592, -0.0732, 0.2362, ..., 0.0357, -0.2677, -0.1786], [-0.2211, -0.3023, -0.0824, ..., -0.1480, -0.3972, -0.1339], [ 0.3349, -0.0860, -0.3426, ..., -0.2652, -0.3037, -0.2278], ..., [-0.1421, -0.0649, -0.4461, ..., -0.1602, 0.1530, 0.0587], [-0.2565, -0.1663, -0.0974, ..., -0.2821, -0.0693, -0.0906], [-0.1394, -0.1113, -0.1821, ..., 0.1046, 0.0250, -0.1957]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 6.2864e-09, ..., 4.6566e-10, 2.7940e-09, 1.3970e-09], [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.7474e-08, 2.3283e-10, -2.3283e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.6298e-09, 0.0000e+00, 2.3283e-10], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 2.3283e-10, -2.3283e-10, 2.3283e-10], [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., -4.7497e-08, 2.3283e-10, 0.0000e+00]], device='cuda:0') Epoch 478, bias, value: tensor([ 3.1208e-03, 6.2690e-03, 1.7072e-02, -3.6802e-03, 3.0329e-02, -4.4400e-03, -2.1296e-05, -5.6580e-03, 1.3060e-02, -1.7636e-02], device='cuda:0'), grad: tensor([ 1.6764e-08, 7.5670e-08, 2.3283e-10, 6.2864e-09, 5.2387e-08, 8.8243e-08, -9.8720e-08, 9.5461e-09, -5.5879e-09, -1.3295e-07], device='cuda:0') 100 1e-05 changing lr epoch 477, time 248.26, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3914 re_mapping 0.0017 re_causal 0.0070 /// teacc 99.22 lr 0.00001000 Epoch 479, weight, value: tensor([[-0.2593, -0.0732, 0.2363, ..., 0.0357, -0.2678, -0.1786], [-0.2212, -0.3023, -0.0825, ..., -0.1481, -0.3973, -0.1339], [ 0.3350, -0.0859, -0.3427, ..., -0.2653, -0.3037, -0.2278], ..., [-0.1422, -0.0649, -0.4461, ..., -0.1603, 0.1531, 0.0587], [-0.2565, -0.1663, -0.0974, ..., -0.2821, -0.0693, -0.0906], [-0.1394, -0.1113, -0.1821, ..., 0.1046, 0.0250, -0.1957]], device='cuda:0'), grad: tensor([[ 6.9849e-10, 0.0000e+00, -1.6298e-09, ..., -9.3132e-10, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 0.0000e+00, 0.0000e+00, ..., 1.1642e-08, 2.3283e-10, 0.0000e+00], [-7.9162e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 4.8894e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [ 1.8626e-09, 2.3283e-10, 0.0000e+00, ..., 4.6566e-10, 0.0000e+00, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 1.3970e-09, ..., -1.8394e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 479, bias, value: tensor([ 3.1219e-03, 6.3710e-03, 1.7141e-02, -3.6522e-03, 3.0344e-02, -4.5717e-03, -4.7893e-05, -5.7705e-03, 1.3113e-02, -1.7637e-02], device='cuda:0'), grad: tensor([-1.1642e-09, 4.3539e-08, -2.3749e-08, 7.4506e-09, 2.4680e-08, -1.1642e-08, 5.3551e-09, 1.9092e-08, 1.0710e-08, -5.4017e-08], device='cuda:0') 100 1e-05 changing lr epoch 478, time 248.08, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4356 re_mapping 0.0018 re_causal 0.0076 /// teacc 99.18 lr 0.00001000 Epoch 480, weight, value: tensor([[-0.2593, -0.0732, 0.2364, ..., 0.0357, -0.2678, -0.1786], [-0.2212, -0.3024, -0.0825, ..., -0.1482, -0.3974, -0.1339], [ 0.3352, -0.0859, -0.3427, ..., -0.2653, -0.3037, -0.2278], ..., [-0.1424, -0.0649, -0.4462, ..., -0.1604, 0.1532, 0.0587], [-0.2566, -0.1663, -0.0974, ..., -0.2821, -0.0693, -0.0906], [-0.1395, -0.1113, -0.1822, ..., 0.1046, 0.0250, -0.1957]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.5379e-08, ..., -1.6531e-08, 0.0000e+00, 0.0000e+00], [ 2.3283e-10, 2.3283e-10, 6.9849e-10, ..., 9.3132e-10, 0.0000e+00, -6.9849e-10], [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, 0.0000e+00, 0.0000e+00], ..., [ 6.9849e-10, 4.6566e-10, 2.3283e-10, ..., 6.9849e-10, 0.0000e+00, 9.3132e-10], [ 2.3283e-10, 2.3283e-10, 6.9849e-10, ..., -9.5461e-09, 2.3283e-10, 0.0000e+00], [ 2.3283e-10, 2.3283e-10, 2.4447e-08, ..., 2.9569e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 480, bias, value: tensor([ 0.0031, 0.0064, 0.0173, -0.0036, 0.0303, -0.0046, -0.0001, -0.0058, 0.0132, -0.0176], device='cuda:0'), grad: tensor([-5.7975e-08, -3.4925e-09, 1.2806e-08, -1.1642e-08, -1.0477e-08, 4.4471e-08, 4.1910e-09, 2.7241e-08, -1.5995e-07, 1.6717e-07], device='cuda:0') 100 1e-05 changing lr epoch 479, time 247.75, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4094 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.21 lr 0.00001000 Epoch 481, weight, value: tensor([[-0.2593, -0.0732, 0.2365, ..., 0.0357, -0.2679, -0.1786], [-0.2213, -0.3024, -0.0826, ..., -0.1482, -0.3974, -0.1339], [ 0.3355, -0.0859, -0.3427, ..., -0.2653, -0.3037, -0.2278], ..., [-0.1425, -0.0650, -0.4462, ..., -0.1605, 0.1532, 0.0587], [-0.2566, -0.1663, -0.0975, ..., -0.2822, -0.0693, -0.0906], [-0.1395, -0.1113, -0.1822, ..., 0.1046, 0.0250, -0.1957]], device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.3283e-10, 0.0000e+00], [ 6.2864e-09, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, 2.0955e-09, 0.0000e+00], [-8.3819e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 9.3132e-10, 0.0000e+00], ..., [ 3.7253e-09, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, -3.0268e-09, 0.0000e+00], [ 6.9849e-10, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, 2.3283e-10, 0.0000e+00], [ 2.3283e-10, 2.3283e-10, 4.6566e-10, ..., -2.5611e-09, 4.6566e-10, 0.0000e+00]], device='cuda:0') Epoch 481, bias, value: tensor([ 0.0031, 0.0064, 0.0175, -0.0036, 0.0304, -0.0047, -0.0002, -0.0058, 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 2.5611e-09, 3.4925e-08, -1.0477e-08, -3.0268e-08, 7.6834e-09, 1.1642e-08, 1.1642e-09, -3.9581e-09, 6.0536e-09, 0.0000e+00], device='cuda:0') 100 1e-05 changing lr epoch 480, time 248.28, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4267 re_mapping 0.0018 re_causal 0.0077 /// teacc 99.20 lr 0.00001000 Epoch 482, weight, value: tensor([[-0.2594, -0.0732, 0.2366, ..., 0.0357, -0.2679, -0.1786], [-0.2214, -0.3024, -0.0826, ..., -0.1482, -0.3975, -0.1339], [ 0.3357, -0.0859, -0.3427, ..., -0.2653, -0.3037, -0.2278], ..., [-0.1426, -0.0650, -0.4462, ..., -0.1607, 0.1533, 0.0587], [-0.2567, -0.1663, -0.0975, ..., -0.2822, -0.0693, -0.0906], [-0.1395, -0.1113, -0.1823, ..., 0.1046, 0.0249, -0.1957]], device='cuda:0'), grad: tensor([[ 1.6298e-09, 0.0000e+00, -3.7253e-09, ..., -1.1642e-09, 0.0000e+00, 0.0000e+00], [ 6.9849e-10, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [-7.6834e-09, 0.0000e+00, -1.6298e-09, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 2.3283e-10, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, 0.0000e+00, 0.0000e+00], [ 6.9849e-10, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 4.6566e-09, ..., 1.0710e-08, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 482, bias, value: tensor([ 0.0031, 0.0065, 0.0176, -0.0035, 0.0303, -0.0048, -0.0003, -0.0059, 0.0132, -0.0176], device='cuda:0'), grad: tensor([-3.9581e-09, -3.7253e-09, -1.4435e-08, -3.2596e-09, -2.1188e-08, 3.7253e-09, 7.4506e-09, 8.6147e-09, 2.7940e-09, 3.1432e-08], device='cuda:0') 100 1e-05 changing lr epoch 481, time 248.09, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4102 re_mapping 0.0018 re_causal 0.0075 /// teacc 99.21 lr 0.00001000 Epoch 483, weight, value: tensor([[-0.2594, -0.0732, 0.2366, ..., 0.0357, -0.2680, -0.1786], [-0.2215, -0.3024, -0.0826, ..., -0.1482, -0.3976, -0.1339], [ 0.3359, -0.0860, -0.3428, ..., -0.2654, -0.3037, -0.2278], ..., [-0.1429, -0.0650, -0.4462, ..., -0.1608, 0.1534, 0.0587], [-0.2567, -0.1663, -0.0975, ..., -0.2822, -0.0693, -0.0906], [-0.1395, -0.1113, -0.1823, ..., 0.1046, 0.0249, -0.1957]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, 2.3283e-10, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 2.3283e-10, ..., 1.3970e-09, 2.3283e-10, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 5.3551e-09, 1.1642e-09, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, -1.1642e-09, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 4.6566e-10, ..., -2.4680e-08, -4.8894e-09, 0.0000e+00]], device='cuda:0') Epoch 483, bias, value: tensor([ 0.0031, 0.0065, 0.0178, -0.0035, 0.0306, -0.0049, -0.0002, -0.0060, 0.0132, -0.0176], device='cuda:0'), grad: tensor([ 3.0268e-09, -3.9581e-09, 1.3970e-09, -1.7462e-08, 5.2154e-08, 1.7462e-08, 0.0000e+00, 3.1665e-08, -2.9802e-08, -4.3074e-08], device='cuda:0') 100 1e-05 changing lr epoch 482, time 247.64, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4022 re_mapping 0.0018 re_causal 0.0073 /// teacc 99.19 lr 0.00001000 Epoch 484, weight, value: tensor([[-0.2594, -0.0732, 0.2367, ..., 0.0357, -0.2680, -0.1786], [-0.2215, -0.3024, -0.0827, ..., -0.1483, -0.3977, -0.1339], [ 0.3360, -0.0860, -0.3428, ..., -0.2654, -0.3038, -0.2278], ..., [-0.1429, -0.0651, -0.4463, ..., -0.1609, 0.1535, 0.0587], [-0.2568, -0.1663, -0.0976, ..., -0.2823, -0.0693, -0.0906], [-0.1395, -0.1113, -0.1824, ..., 0.1046, 0.0249, -0.1958]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.3050e-08, ..., -6.0536e-09, 0.0000e+00, -9.3132e-10], [ 2.3283e-10, 0.0000e+00, 4.6566e-10, ..., 2.3283e-10, 1.1642e-09, 0.0000e+00], [-1.6298e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.3283e-10, 0.0000e+00], ..., [ 1.1642e-09, 0.0000e+00, 9.3132e-10, ..., 2.3283e-10, -8.6147e-09, 2.3283e-10], [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -2.3283e-10, -2.3283e-10], [ 0.0000e+00, 0.0000e+00, 1.7229e-08, ..., 4.8894e-09, 6.5193e-09, 6.9849e-10]], device='cuda:0') Epoch 484, bias, value: tensor([ 3.1262e-03, 6.4709e-03, 1.7825e-02, -3.4399e-03, 3.0611e-02, -4.9969e-03, 3.0793e-06, -5.9801e-03, 1.3163e-02, -1.7649e-02], device='cuda:0'), grad: tensor([-5.3551e-08, 3.7253e-09, -1.6298e-09, -2.0955e-09, 2.3283e-09, 6.2864e-09, 7.9162e-09, -2.0256e-08, -1.6298e-09, 6.6124e-08], device='cuda:0') 100 1e-05 changing lr epoch 483, time 247.72, cls_loss 0.0003 cls_loss_mapping 0.0002 cls_loss_causal 0.4330 re_mapping 0.0018 re_causal 0.0076 /// teacc 99.21 lr 0.00001000 Epoch 485, weight, value: tensor([[-0.2594, -0.0732, 0.2368, ..., 0.0357, -0.2681, -0.1786], [-0.2216, -0.3024, -0.0827, ..., -0.1483, -0.3979, -0.1339], [ 0.3361, -0.0860, -0.3428, ..., -0.2655, -0.3038, -0.2278], ..., [-0.1430, -0.0651, -0.4463, ..., -0.1610, 0.1536, 0.0587], [-0.2568, -0.1663, -0.0977, ..., -0.2824, -0.0693, -0.0906], [-0.1395, -0.1113, -0.1824, ..., 0.1046, 0.0248, -0.1958]], device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.7940e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, -2.3283e-09, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 3.7253e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 485, bias, value: tensor([ 3.1284e-03, 6.5911e-03, 1.7799e-02, -3.3711e-03, 3.0659e-02, -5.0840e-03, -3.9424e-05, -6.0899e-03, 1.3146e-02, -1.7653e-02], device='cuda:0'), grad: tensor([ 6.9849e-10, 1.3271e-08, 4.6566e-10, 4.6566e-10, -6.2864e-09, 4.6566e-10, 1.3970e-09, -2.0955e-09, 9.3132e-10, 1.0477e-08], device='cuda:0') 100 1e-05 changing lr epoch 484, time 247.28, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4447 re_mapping 0.0018 re_causal 0.0077 /// teacc 99.17 lr 0.00001000 Epoch 486, weight, value: tensor([[-0.2594, -0.0733, 0.2369, ..., 0.0357, -0.2681, -0.1786], [-0.2216, -0.3024, -0.0827, ..., -0.1483, -0.3980, -0.1339], [ 0.3363, -0.0860, -0.3429, ..., -0.2655, -0.3038, -0.2278], ..., [-0.1431, -0.0651, -0.4463, ..., -0.1611, 0.1537, 0.0587], [-0.2569, -0.1664, -0.0977, ..., -0.2825, -0.0693, -0.0906], [-0.1396, -0.1113, -0.1825, ..., 0.1046, 0.0248, -0.1958]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -2.4913e-08, ..., -6.7521e-09, 0.0000e+00, -4.8894e-09], [ 2.3283e-10, 0.0000e+00, 6.9849e-10, ..., 2.3283e-10, 6.9849e-10, 9.3132e-10], [-1.6298e-09, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, 2.3283e-10, 0.0000e+00], ..., [ 1.1642e-09, 2.3283e-10, 4.6566e-10, ..., 0.0000e+00, -2.3982e-08, -6.9849e-10], [ 2.3283e-10, 0.0000e+00, 6.9849e-10, ..., 2.3283e-10, 0.0000e+00, 2.3283e-10], [ 0.0000e+00, 0.0000e+00, 3.4925e-09, ..., 1.1642e-09, 2.0955e-09, 6.9849e-10]], device='cuda:0') Epoch 486, bias, value: tensor([ 3.1297e-03, 6.5977e-03, 1.7871e-02, -3.3474e-03, 3.0687e-02, -5.0980e-03, -6.1985e-05, -6.1066e-03, 1.3108e-02, -1.7655e-02], device='cuda:0'), grad: tensor([-4.7497e-08, 1.2573e-08, -9.3132e-10, -3.2131e-08, 1.0966e-07, 5.5647e-08, 4.0745e-08, -1.3178e-07, -1.2340e-08, 1.8859e-08], device='cuda:0') 100 1e-05 changing lr epoch 485, time 246.56, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.3865 re_mapping 0.0018 re_causal 0.0071 /// teacc 99.19 lr 0.00001000 Epoch 487, weight, value: tensor([[-0.2595, -0.0733, 0.2369, ..., 0.0357, -0.2682, -0.1786], [-0.2217, -0.3024, -0.0828, ..., -0.1484, -0.3981, -0.1339], [ 0.3363, -0.0860, -0.3429, ..., -0.2655, -0.3038, -0.2278], ..., [-0.1432, -0.0652, -0.4463, ..., -0.1612, 0.1539, 0.0587], [-0.2569, -0.1664, -0.0977, ..., -0.2826, -0.0693, -0.0906], [-0.1396, -0.1113, -0.1825, ..., 0.1046, 0.0248, -0.1958]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 6.9849e-10, 2.3283e-10, 2.3283e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, 2.3283e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, 0.0000e+00, 0.0000e+00], ..., [ 1.6298e-09, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, 4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., -2.3283e-10, -2.0955e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 6.9849e-10, ..., 4.1910e-09, 1.1642e-09, 0.0000e+00]], device='cuda:0') Epoch 487, bias, value: tensor([ 0.0031, 0.0066, 0.0178, -0.0032, 0.0307, -0.0051, -0.0001, -0.0061, 0.0131, -0.0177], device='cuda:0'), grad: tensor([ 4.4238e-09, -3.1898e-08, 3.4925e-09, 5.8208e-09, -1.9092e-08, -1.6298e-08, 2.5844e-08, 4.1677e-08, -1.2340e-08, 2.3982e-08], device='cuda:0') 100 1e-05 changing lr epoch 486, time 246.15, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4571 re_mapping 0.0017 re_causal 0.0074 /// teacc 99.19 lr 0.00001000 Epoch 488, weight, value: tensor([[-0.2595, -0.0733, 0.2370, ..., 0.0357, -0.2682, -0.1786], [-0.2217, -0.3025, -0.0828, ..., -0.1485, -0.3983, -0.1339], [ 0.3365, -0.0860, -0.3429, ..., -0.2655, -0.3039, -0.2278], ..., [-0.1432, -0.0652, -0.4464, ..., -0.1613, 0.1540, 0.0587], [-0.2569, -0.1664, -0.0978, ..., -0.2826, -0.0693, -0.0906], [-0.1396, -0.1113, -0.1826, ..., 0.1046, 0.0247, -0.1958]], device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], [ 4.6566e-10, 2.3283e-10, 2.3283e-10, ..., 2.3283e-10, 0.0000e+00, 0.0000e+00], [-6.0536e-09, -6.9849e-10, 0.0000e+00, ..., 0.0000e+00, 4.6566e-10, 4.6566e-10], ..., [ 4.4238e-09, 9.3132e-10, 0.0000e+00, ..., 2.3283e-10, -4.6566e-10, -2.3283e-10], [ 3.2596e-09, 1.1642e-09, 0.0000e+00, ..., 0.0000e+00, -2.3283e-10, 0.0000e+00], [ 1.1642e-09, 4.6566e-10, 6.9849e-10, ..., 8.1491e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 488, bias, value: tensor([ 0.0031, 0.0066, 0.0177, -0.0032, 0.0306, -0.0051, -0.0001, -0.0062, 0.0131, -0.0177], device='cuda:0'), grad: tensor([ 1.8626e-09, 3.0268e-09, -7.9162e-09, -1.0012e-06, -1.7229e-08, 9.8441e-07, 3.7253e-09, 1.3271e-08, -2.3283e-09, 2.6543e-08], device='cuda:0') 100 1e-05 changing lr epoch 487, time 246.45, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4235 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.20 lr 0.00001000 Epoch 489, weight, value: tensor([[-0.2595, -0.0733, 0.2371, ..., 0.0357, -0.2682, -0.1786], [-0.2218, -0.3025, -0.0828, ..., -0.1486, -0.3984, -0.1339], [ 0.3366, -0.0860, -0.3430, ..., -0.2656, -0.3039, -0.2278], ..., [-0.1433, -0.0652, -0.4464, ..., -0.1615, 0.1541, 0.0587], [-0.2569, -0.1664, -0.0978, ..., -0.2827, -0.0693, -0.0907], [-0.1396, -0.1113, -0.1827, ..., 0.1046, 0.0247, -0.1958]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.5623e-08, ..., -3.1898e-08, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.2596e-09, ..., -3.0268e-09, 4.1910e-09, 2.3283e-10], [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 2.3283e-10, 0.0000e+00, 0.0000e+00], ..., [ 2.3283e-10, 0.0000e+00, 4.6566e-10, ..., 3.0268e-09, -6.9849e-09, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 6.9849e-10, ..., 4.6566e-10, 9.3132e-10, 6.9849e-10], [ 0.0000e+00, 0.0000e+00, 3.5623e-08, ..., 2.9569e-08, 3.7253e-09, 0.0000e+00]], device='cuda:0') Epoch 489, bias, value: tensor([ 0.0031, 0.0066, 0.0176, -0.0032, 0.0304, -0.0052, -0.0002, -0.0062, 0.0131, -0.0176], device='cuda:0'), grad: tensor([-8.4052e-08, -7.4506e-09, 4.8894e-09, 5.8208e-09, 2.6543e-08, 1.1874e-08, -4.2841e-08, -1.1642e-08, 2.7940e-09, 1.1642e-07], device='cuda:0') 100 1e-05 changing lr epoch 488, time 246.44, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4426 re_mapping 0.0017 re_causal 0.0073 /// teacc 99.18 lr 0.00001000 Epoch 490, weight, value: tensor([[-0.2595, -0.0733, 0.2373, ..., 0.0357, -0.2683, -0.1787], [-0.2218, -0.3025, -0.0829, ..., -0.1487, -0.3984, -0.1339], [ 0.3366, -0.0860, -0.3431, ..., -0.2656, -0.3039, -0.2278], ..., [-0.1434, -0.0652, -0.4464, ..., -0.1617, 0.1542, 0.0587], [-0.2570, -0.1664, -0.0978, ..., -0.2828, -0.0694, -0.0907], [-0.1396, -0.1113, -0.1828, ..., 0.1046, 0.0247, -0.1958]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 2.3283e-10, 6.9849e-10, ..., 6.9849e-10, 4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, 9.3132e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.3283e-10, 0.0000e+00], ..., [ 0.0000e+00, 2.3283e-10, 0.0000e+00, ..., 5.8208e-09, 1.3970e-09, 2.3283e-10], [ 0.0000e+00, 4.6566e-10, 0.0000e+00, ..., 2.3283e-10, -1.6298e-09, 0.0000e+00], [ 0.0000e+00, 4.6566e-10, 3.0268e-09, ..., -2.8871e-08, -7.2177e-09, 0.0000e+00]], device='cuda:0') Epoch 490, bias, value: tensor([ 0.0031, 0.0068, 0.0174, -0.0032, 0.0302, -0.0053, -0.0001, -0.0064, 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 5.3551e-09, 6.2864e-09, 9.3132e-10, 4.4238e-09, 6.0303e-08, -2.0256e-08, 4.6566e-09, 1.9558e-08, -4.4238e-09, -5.6811e-08], device='cuda:0') 100 1e-05 changing lr epoch 489, time 246.52, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4015 re_mapping 0.0017 re_causal 0.0068 /// teacc 99.18 lr 0.00001000 Epoch 491, weight, value: tensor([[-0.2595, -0.0733, 0.2374, ..., 0.0357, -0.2684, -0.1787], [-0.2218, -0.3025, -0.0829, ..., -0.1488, -0.3986, -0.1339], [ 0.3367, -0.0860, -0.3432, ..., -0.2657, -0.3040, -0.2278], ..., [-0.1435, -0.0653, -0.4465, ..., -0.1618, 0.1543, 0.0587], [-0.2570, -0.1664, -0.0979, ..., -0.2829, -0.0694, -0.0907], [-0.1397, -0.1113, -0.1829, ..., 0.1046, 0.0247, -0.1958]], device='cuda:0'), grad: tensor([[ 1.8626e-09, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 9.3132e-10], [ 4.6566e-10, 2.3283e-10, -2.3283e-10, ..., 2.3283e-10, 6.9849e-10, 0.0000e+00], [-4.6566e-09, -1.1642e-09, 0.0000e+00, ..., 0.0000e+00, 2.3283e-10, 0.0000e+00], ..., [ 9.3132e-10, 4.6566e-10, 0.0000e+00, ..., 0.0000e+00, -4.6566e-10, 0.0000e+00], [ 1.6298e-09, 4.6566e-10, 0.0000e+00, ..., 6.9849e-10, 6.9849e-10, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., -1.1642e-09, 6.9849e-10, 2.3283e-10]], device='cuda:0') Epoch 491, bias, value: tensor([ 3.1347e-03, 6.8227e-03, 1.7222e-02, -3.2285e-03, 3.0162e-02, -5.3294e-03, 1.7088e-05, -6.3709e-03, 1.3122e-02, -1.7644e-02], device='cuda:0'), grad: tensor([ 1.1409e-08, -1.1642e-08, -8.8476e-09, 1.1874e-08, 6.0536e-09, -6.8918e-08, 1.7462e-08, 1.9092e-08, 2.2585e-08, 7.4506e-09], device='cuda:0') 100 1e-05 changing lr epoch 490, time 246.35, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4216 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 Epoch 492, weight, value: tensor([[-0.2595, -0.0733, 0.2376, ..., 0.0357, -0.2684, -0.1787], [-0.2219, -0.3025, -0.0830, ..., -0.1489, -0.3987, -0.1339], [ 0.3368, -0.0860, -0.3433, ..., -0.2657, -0.3040, -0.2278], ..., [-0.1435, -0.0653, -0.4465, ..., -0.1619, 0.1545, 0.0587], [-0.2570, -0.1664, -0.0980, ..., -0.2830, -0.0694, -0.0907], [-0.1397, -0.1113, -0.1830, ..., 0.1046, 0.0246, -0.1959]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -5.1223e-09, ..., -3.2596e-09, 0.0000e+00, -2.3283e-10], [ 2.7940e-09, 0.0000e+00, 4.6566e-10, ..., 2.3283e-09, 6.9849e-10, 2.7940e-09], [-3.2596e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.6298e-09, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, -9.3132e-10, 0.0000e+00], [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., 4.6566e-10, 0.0000e+00, 2.3283e-10], [ 0.0000e+00, 0.0000e+00, 3.4925e-09, ..., 7.2177e-09, 2.3283e-10, 4.6566e-10]], device='cuda:0') Epoch 492, bias, value: tensor([ 3.1399e-03, 6.8837e-03, 1.7148e-02, -3.1274e-03, 3.0145e-02, -5.4663e-03, 7.6682e-05, -6.4609e-03, 1.3102e-02, -1.7647e-02], device='cuda:0'), grad: tensor([-9.5461e-09, 1.0477e-07, -6.5193e-09, 5.3318e-08, -2.4214e-08, -1.6182e-07, 7.6834e-09, 1.0943e-08, 7.6834e-09, 3.3760e-08], device='cuda:0') 100 1e-05 changing lr epoch 491, time 246.76, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4164 re_mapping 0.0017 re_causal 0.0070 /// teacc 99.23 lr 0.00001000 Epoch 493, weight, value: tensor([[-0.2596, -0.0733, 0.2377, ..., 0.0357, -0.2685, -0.1787], [-0.2219, -0.3025, -0.0830, ..., -0.1490, -0.3988, -0.1339], [ 0.3369, -0.0860, -0.3433, ..., -0.2657, -0.3040, -0.2278], ..., [-0.1436, -0.0653, -0.4466, ..., -0.1620, 0.1546, 0.0587], [-0.2570, -0.1664, -0.0981, ..., -0.2831, -0.0694, -0.0907], [-0.1397, -0.1113, -0.1831, ..., 0.1046, 0.0246, -0.1959]], device='cuda:0'), grad: tensor([[ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, 0.0000e+00, 0.0000e+00], [ 6.2864e-09, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, 0.0000e+00, 0.0000e+00], [-2.1188e-08, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, -0.0000e+00, 0.0000e+00], ..., [ 1.2573e-08, 6.9849e-10, 0.0000e+00, ..., 2.3283e-10, 0.0000e+00, 0.0000e+00], [ 9.3132e-10, 2.3283e-10, 0.0000e+00, ..., 6.9849e-10, 0.0000e+00, 0.0000e+00], [ 2.0955e-09, -1.8626e-09, 9.3132e-10, ..., -7.9162e-09, -0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 493, bias, value: tensor([ 0.0031, 0.0069, 0.0171, -0.0030, 0.0301, -0.0056, 0.0002, -0.0065, 0.0131, -0.0176], device='cuda:0'), grad: tensor([ 3.7253e-09, 1.1176e-08, -4.9127e-08, 1.1874e-08, 8.6147e-09, -5.0291e-08, 8.8476e-09, 4.2608e-08, 3.1432e-08, -6.5193e-09], device='cuda:0') 100 1e-05 changing lr epoch 492, time 246.54, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4069 re_mapping 0.0017 re_causal 0.0068 /// teacc 99.19 lr 0.00001000 Epoch 494, weight, value: tensor([[-0.2596, -0.0733, 0.2378, ..., 0.0357, -0.2686, -0.1788], [-0.2220, -0.3025, -0.0831, ..., -0.1490, -0.3989, -0.1339], [ 0.3371, -0.0860, -0.3434, ..., -0.2657, -0.3041, -0.2278], ..., [-0.1437, -0.0653, -0.4466, ..., -0.1621, 0.1548, 0.0588], [-0.2571, -0.1664, -0.0982, ..., -0.2832, -0.0694, -0.0907], [-0.1397, -0.1113, -0.1832, ..., 0.1046, 0.0245, -0.1959]], device='cuda:0'), grad: tensor([[ 3.0268e-09, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, 2.3283e-10, 0.0000e+00], [ 4.6566e-10, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 6.9849e-10, 0.0000e+00], [-7.4506e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.1642e-09, 2.3283e-10, 0.0000e+00, ..., 2.3283e-10, -6.9849e-10, 0.0000e+00], [ 3.0268e-09, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, 2.3283e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 1.1642e-09, 0.0000e+00, 0.0000e+00]], device='cuda:0') Epoch 494, bias, value: tensor([ 0.0031, 0.0069, 0.0171, -0.0030, 0.0300, -0.0057, 0.0003, -0.0065, 0.0130, -0.0176], device='cuda:0'), grad: tensor([ 9.3132e-09, 9.3132e-10, -2.0023e-08, 1.4203e-08, 4.1910e-09, -1.4668e-08, 8.8476e-09, 4.6566e-09, 6.2864e-09, 6.0536e-09], device='cuda:0') 100 1e-05 changing lr epoch 493, time 246.71, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4303 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.21 lr 0.00001000 Epoch 495, weight, value: tensor([[-0.2596, -0.0733, 0.2380, ..., 0.0357, -0.2687, -0.1788], [-0.2220, -0.3025, -0.0831, ..., -0.1491, -0.3991, -0.1339], [ 0.3373, -0.0860, -0.3434, ..., -0.2658, -0.3041, -0.2278], ..., [-0.1438, -0.0654, -0.4466, ..., -0.1623, 0.1550, 0.0588], [-0.2572, -0.1665, -0.0982, ..., -0.2832, -0.0694, -0.0908], [-0.1397, -0.1113, -0.1833, ..., 0.1046, 0.0245, -0.1959]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -4.3772e-08, ..., -3.0035e-08, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 9.3132e-10, 1.8626e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.3283e-10, 1.3970e-09, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 1.8626e-09, -4.1910e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 1.1642e-09, ..., 1.1642e-09, 2.3283e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.7951e-08, ..., -5.5879e-09, 1.3970e-09, 0.0000e+00]], device='cuda:0') Epoch 495, bias, value: tensor([ 0.0031, 0.0069, 0.0171, -0.0030, 0.0299, -0.0057, 0.0004, -0.0066, 0.0130, -0.0176], device='cuda:0'), grad: tensor([-1.1059e-07, 1.6065e-08, 1.0943e-08, 3.7253e-09, 9.0105e-08, -1.3970e-09, 9.3132e-09, -2.0489e-08, 6.2864e-09, 3.7253e-09], device='cuda:0') 100 1e-05 changing lr epoch 494, time 246.80, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4218 re_mapping 0.0017 re_causal 0.0070 /// teacc 99.20 lr 0.00001000 Epoch 496, weight, value: tensor([[-0.2596, -0.0733, 0.2381, ..., 0.0357, -0.2688, -0.1788], [-0.2221, -0.3025, -0.0832, ..., -0.1492, -0.3992, -0.1340], [ 0.3375, -0.0860, -0.3435, ..., -0.2658, -0.3041, -0.2278], ..., [-0.1439, -0.0654, -0.4467, ..., -0.1624, 0.1552, 0.0588], [-0.2572, -0.1665, -0.0983, ..., -0.2833, -0.0695, -0.0908], [-0.1397, -0.1113, -0.1834, ..., 0.1046, 0.0244, -0.1959]], device='cuda:0'), grad: tensor([[ 6.9849e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, 2.3283e-10, 0.0000e+00], [ 6.9849e-10, 0.0000e+00, 2.3283e-10, ..., 6.9849e-10, 9.3132e-10, -2.3283e-09], [-3.2596e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 2.3283e-10, 0.0000e+00], ..., [ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, -9.3132e-10, 2.3283e-10], [ 1.1642e-09, 0.0000e+00, 0.0000e+00, ..., 2.5611e-09, 6.2864e-09, 1.6298e-09], [ 2.3283e-10, 0.0000e+00, 2.3283e-10, ..., -1.6298e-08, -4.6566e-10, 0.0000e+00]], device='cuda:0') Epoch 496, bias, value: tensor([ 0.0031, 0.0070, 0.0172, -0.0030, 0.0298, -0.0057, 0.0005, -0.0066, 0.0130, -0.0176], device='cuda:0'), grad: tensor([ 5.1223e-09, -6.5193e-08, -4.8894e-09, 1.6997e-08, 3.8184e-08, -1.2550e-07, 4.6799e-08, 8.1491e-09, 1.2130e-07, -2.8405e-08], device='cuda:0') 100 1e-05 changing lr epoch 495, time 246.16, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4127 re_mapping 0.0017 re_causal 0.0070 /// teacc 99.19 lr 0.00001000 Epoch 497, weight, value: tensor([[-0.2597, -0.0733, 0.2382, ..., 0.0357, -0.2689, -0.1788], [-0.2222, -0.3025, -0.0832, ..., -0.1494, -0.3994, -0.1340], [ 0.3376, -0.0860, -0.3435, ..., -0.2658, -0.3042, -0.2278], ..., [-0.1440, -0.0654, -0.4467, ..., -0.1626, 0.1554, 0.0588], [-0.2572, -0.1665, -0.0984, ..., -0.2834, -0.0695, -0.0908], [-0.1398, -0.1113, -0.1835, ..., 0.1046, 0.0244, -0.1960]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, 4.6566e-10, ..., 2.3283e-10, 8.1491e-10, 4.6566e-10], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 3.4925e-10, 1.0477e-09, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-10, 1.1642e-10, 0.0000e+00], ..., [ 5.8208e-10, 0.0000e+00, 0.0000e+00, ..., 4.6566e-10, -1.2806e-09, 0.0000e+00], [ 5.8208e-10, 0.0000e+00, 1.1642e-10, ..., 6.9849e-10, 5.8208e-10, 2.3283e-10], [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 2.5611e-09, 8.1491e-10, 0.0000e+00]], device='cuda:0') Epoch 497, bias, value: tensor([ 0.0031, 0.0069, 0.0172, -0.0030, 0.0299, -0.0057, 0.0005, -0.0065, 0.0129, -0.0177], device='cuda:0'), grad: tensor([ 4.6566e-09, 5.4715e-09, 8.1491e-10, 8.0327e-09, -1.2224e-08, -1.9092e-08, -3.9581e-09, -3.6089e-09, 7.6834e-09, 1.4901e-08], device='cuda:0') 100 1e-05 changing lr epoch 496, time 246.51, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4403 re_mapping 0.0017 re_causal 0.0069 /// teacc 99.19 lr 0.00001000 Epoch 498, weight, value: tensor([[-0.2597, -0.0733, 0.2383, ..., 0.0357, -0.2689, -0.1788], [-0.2223, -0.3025, -0.0833, ..., -0.1495, -0.3995, -0.1340], [ 0.3379, -0.0861, -0.3435, ..., -0.2659, -0.3042, -0.2278], ..., [-0.1442, -0.0655, -0.4467, ..., -0.1627, 0.1555, 0.0588], [-0.2573, -0.1665, -0.0984, ..., -0.2834, -0.0695, -0.0908], [-0.1398, -0.1113, -0.1836, ..., 0.1046, 0.0243, -0.1960]], device='cuda:0'), grad: tensor([[ 2.3283e-10, 0.0000e+00, 0.0000e+00, ..., 2.3283e-10, 4.6566e-10, 4.6566e-10], [ 1.3970e-09, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, 1.3970e-09, 2.3283e-10], [-5.5879e-09, 0.0000e+00, 0.0000e+00, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 1.1642e-09, 9.3132e-10, 0.0000e+00, ..., 2.5611e-09, -6.9849e-10, 0.0000e+00], [ 1.8626e-09, 2.3283e-10, 0.0000e+00, ..., 0.0000e+00, 1.3970e-09, 6.9849e-10], [ 0.0000e+00, 0.0000e+00, 9.3132e-10, ..., 1.8161e-08, 6.9849e-10, 2.3283e-10]], device='cuda:0') Epoch 498, bias, value: tensor([ 0.0032, 0.0069, 0.0173, -0.0030, 0.0298, -0.0058, 0.0006, -0.0066, 0.0130, -0.0177], device='cuda:0'), grad: tensor([ 4.1910e-09, -1.7928e-08, -1.7229e-08, -2.3283e-09, -4.0978e-08, 4.8894e-09, -2.1653e-08, 3.4925e-08, 1.6298e-08, 5.1456e-08], device='cuda:0') 100 1e-05 changing lr epoch 497, time 246.15, cls_loss 0.0004 cls_loss_mapping 0.0004 cls_loss_causal 0.4404 re_mapping 0.0016 re_causal 0.0071 /// teacc 99.19 lr 0.00001000 Epoch 499, weight, value: tensor([[-0.2597, -0.0733, 0.2386, ..., 0.0357, -0.2691, -0.1788], [-0.2224, -0.3026, -0.0833, ..., -0.1495, -0.3997, -0.1340], [ 0.3382, -0.0861, -0.3436, ..., -0.2659, -0.3043, -0.2278], ..., [-0.1444, -0.0655, -0.4468, ..., -0.1628, 0.1557, 0.0588], [-0.2574, -0.1665, -0.0985, ..., -0.2835, -0.0695, -0.0908], [-0.1398, -0.1113, -0.1837, ..., 0.1046, 0.0243, -0.1960]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -1.8626e-09, ..., -2.3283e-10, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 6.9849e-10, 4.6566e-10, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 0.0000e+00, 0.0000e+00, 0.0000e+00], ..., [ 0.0000e+00, 0.0000e+00, 0.0000e+00, ..., 1.1642e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 2.3283e-10, ..., 1.1642e-09, 0.0000e+00, 0.0000e+00], [ 0.0000e+00, 0.0000e+00, 3.0268e-09, ..., -2.3749e-08, 2.3283e-10, 0.0000e+00]], device='cuda:0') Epoch 499, bias, value: tensor([ 0.0032, 0.0070, 0.0174, -0.0031, 0.0297, -0.0055, 0.0005, -0.0068, 0.0129, -0.0177], device='cuda:0'), grad: tensor([-1.6298e-09, -9.7789e-09, 4.6566e-10, 1.1176e-08, 7.3574e-08, -1.8394e-08, 7.9162e-09, 2.0489e-08, 5.5879e-09, -7.3807e-08], device='cuda:0') 100 1e-05 changing lr epoch 498, time 246.47, cls_loss 0.0004 cls_loss_mapping 0.0003 cls_loss_causal 0.4229 re_mapping 0.0017 re_causal 0.0071 /// teacc 99.22 lr 0.00001000 Epoch 500, weight, value: tensor([[-0.2598, -0.0733, 0.2387, ..., 0.0357, -0.2691, -0.1788], [-0.2225, -0.3026, -0.0833, ..., -0.1497, -0.3999, -0.1340], [ 0.3383, -0.0861, -0.3436, ..., -0.2659, -0.3043, -0.2278], ..., [-0.1445, -0.0655, -0.4468, ..., -0.1629, 0.1560, 0.0588], [-0.2574, -0.1665, -0.0985, ..., -0.2836, -0.0696, -0.0908], [-0.1398, -0.1113, -0.1838, ..., 0.1046, 0.0242, -0.1960]], device='cuda:0'), grad: tensor([[ 0.0000e+00, 0.0000e+00, -3.0734e-08, ..., -1.3039e-08, 6.9849e-10, -0.0000e+00], [ 2.3283e-10, 4.6566e-10, 1.3970e-09, ..., 9.3132e-10, 2.3283e-10, 0.0000e+00], [ 1.1642e-09, 3.4925e-09, 1.1642e-09, ..., 4.6566e-10, 6.9849e-10, 2.3283e-10], ..., [ 4.6566e-10, 1.1642e-09, 2.3283e-09, ..., 2.5611e-09, 6.9849e-10, 0.0000e+00], [ 0.0000e+00, -1.1642e-09, 6.9849e-09, ..., 4.6566e-10, -1.6298e-08, 2.3283e-10], [ 0.0000e+00, 0.0000e+00, 1.2340e-08, ..., 2.6310e-08, 2.3283e-10, 2.3283e-10]], device='cuda:0') Epoch 500, bias, value: tensor([ 0.0032, 0.0070, 0.0173, -0.0032, 0.0296, -0.0054, 0.0005, -0.0067, 0.0130, -0.0177], device='cuda:0'), grad: tensor([-7.4040e-08, 6.7521e-09, 2.3283e-08, 3.3760e-08, -8.1258e-08, 1.2806e-08, 1.3970e-08, 2.7474e-08, -7.1945e-08, 1.1921e-07], device='cuda:0') 100 1e-05 changing lr epoch 499, time 246.31, cls_loss 0.0003 cls_loss_mapping 0.0003 cls_loss_causal 0.4185 re_mapping 0.0017 re_causal 0.0072 /// teacc 99.17 lr 0.00001000 ---------------------saving last model at epoch 499---------------------------------------------------- /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_onlyblock2', '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([[-0.0169, -0.0253, 0.0165, ..., 0.0308, -0.0245, 0.0071], [ 0.0253, -0.0156, -0.0133, ..., -0.0190, 0.0099, -0.0189], [-0.0034, 0.0289, -0.0063, ..., -0.0258, -0.0044, -0.0031], ..., [-0.0209, 0.0037, 0.0139, ..., -0.0167, -0.0230, 0.0121], [-0.0203, 0.0144, 0.0223, ..., -0.0020, 0.0059, -0.0016], [-0.0243, -0.0195, 0.0228, ..., 0.0235, -0.0155, 0.0110]], device='cuda:0'), grad: None Epoch 1, bias, value: tensor([ 0.0295, -0.0169, -0.0055, 0.0009, 0.0004, -0.0044, -0.0083, -0.0051, 0.0301, 0.0197], device='cuda:0'), grad: None 100 0.0001 changing lr ---------------------saving model at epoch 0---------------------------------------------------- epoch 0, time 276.71, cls_loss 1.2511 cls_loss_mapping 1.7983 cls_loss_causal 2.2017 re_mapping 0.1599 re_causal 0.1691 /// teacc 81.97 lr 0.00010000 Epoch 2, weight, value: tensor([[-0.0159, -0.0265, 0.0146, ..., 0.0314, -0.0300, 0.0096], [ 0.0175, -0.0222, -0.0215, ..., -0.0196, 0.0111, -0.0269], [ 0.0010, 0.0316, -0.0102, ..., -0.0265, -0.0117, -0.0075], ..., [-0.0158, 0.0036, 0.0140, ..., -0.0174, -0.0271, 0.0187], [-0.0228, 0.0152, 0.0244, ..., -0.0026, 0.0058, -0.0085], [-0.0246, -0.0182, 0.0245, ..., 0.0228, -0.0146, 0.0131]], device='cuda:0'), grad: tensor([[ 2.3232e-03, 8.7690e-04, -2.7943e-03, ..., -7.7844e-05, -7.9041e-03, -1.5823e-02], [ 1.0414e-02, 4.3106e-03, 8.9417e-03, ..., 3.8580e-07, 2.2644e-02, 1.2611e-02], [ 6.0177e-04, 5.2605e-03, 1.0315e-02, ..., 6.9402e-06, 2.1088e-02, 8.8196e-03], ..., [-4.5746e-02, -3.5515e-03, -3.8166e-03, ..., 1.3569e-06, 1.0201e-02, -6.1707e-02], [ 1.4015e-02, 1.4467e-03, 9.7351e-03, ..., 6.0350e-06, 2.3148e-02, 2.8137e-02], [ 3.6682e-02, 3.5610e-03, 4.6753e-02, ..., 2.8834e-06, 4.0314e-02, 8.4595e-02]], device='cuda:0') Epoch 2, bias, value: tensor([ 0.0273, -0.0155, -0.0061, 0.0006, -0.0003, -0.0041, -0.0080, -0.0044, 0.0297, 0.0196], device='cuda:0'), grad: tensor([-0.0194, 0.0277, 0.0198, -0.0734, -0.0612, 0.0463, -0.0369, -0.0263, 0.0376, 0.0858], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 1---------------------------------------------------- epoch 1, time 275.56, cls_loss 0.4374 cls_loss_mapping 0.7801 cls_loss_causal 1.9032 re_mapping 0.2112 re_causal 0.2743 /// teacc 91.82 lr 0.00010000 Epoch 3, weight, value: tensor([[-0.0167, -0.0268, 0.0123, ..., 0.0305, -0.0334, 0.0104], [ 0.0147, -0.0253, -0.0245, ..., -0.0167, 0.0124, -0.0288], [ 0.0039, 0.0329, -0.0128, ..., -0.0289, -0.0159, -0.0076], ..., [-0.0139, 0.0030, 0.0116, ..., -0.0196, -0.0330, 0.0209], [-0.0220, 0.0185, 0.0271, ..., -0.0052, 0.0076, -0.0138], [-0.0237, -0.0164, 0.0242, ..., 0.0211, -0.0172, 0.0153]], device='cuda:0'), grad: tensor([[-2.0981e-03, 1.9484e-03, 2.2566e-04, ..., 3.1758e-07, 1.1930e-03, -6.1750e-04], [ 1.6602e-02, 9.0027e-03, 5.2147e-03, ..., -5.6237e-05, 1.8692e-02, 1.8402e-02], [-1.0246e-02, -7.2250e-03, 2.3499e-03, ..., 1.0781e-05, 3.9558e-03, 1.5459e-03], ..., [-6.8893e-03, 1.4181e-03, 4.1733e-03, ..., 1.3402e-06, 7.2403e-03, -2.3590e-02], [ 4.2419e-03, 1.5480e-02, 2.5665e-02, ..., 3.6806e-05, 2.1729e-02, -3.9635e-03], [-2.6150e-03, -1.1200e-02, -8.7585e-03, ..., 3.9255e-07, -6.9809e-03, -6.5193e-03]], device='cuda:0') Epoch 3, bias, value: tensor([ 0.0270, -0.0152, -0.0063, 0.0003, -0.0003, -0.0032, -0.0089, -0.0048, 0.0301, 0.0200], device='cuda:0'), grad: tensor([ 0.0004, 0.0277, -0.0048, -0.0062, -0.0087, -0.0186, 0.0092, -0.0090, 0.0159, -0.0059], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 2---------------------------------------------------- epoch 2, time 275.31, cls_loss 0.2946 cls_loss_mapping 0.4847 cls_loss_causal 1.7281 re_mapping 0.1530 re_causal 0.2479 /// teacc 94.34 lr 0.00010000 Epoch 4, weight, value: tensor([[-0.0169, -0.0273, 0.0102, ..., 0.0296, -0.0365, 0.0103], [ 0.0126, -0.0277, -0.0265, ..., -0.0140, 0.0128, -0.0307], [ 0.0067, 0.0341, -0.0145, ..., -0.0334, -0.0197, -0.0077], ..., [-0.0129, 0.0024, 0.0107, ..., -0.0329, -0.0363, 0.0225], [-0.0222, 0.0209, 0.0286, ..., -0.0074, 0.0086, -0.0162], [-0.0245, -0.0163, 0.0237, ..., 0.0107, -0.0183, 0.0161]], device='cuda:0'), grad: tensor([[-8.3313e-03, 1.1253e-03, 6.3658e-04, ..., -1.9073e-04, -3.5820e-03, -8.5144e-03], [ 2.1706e-03, 3.8319e-03, 4.9095e-03, ..., 1.6415e-04, 9.4299e-03, 6.4516e-04], [ 1.0557e-03, 3.4866e-03, 4.3869e-03, ..., 1.8330e-03, 6.3858e-03, 1.3914e-03], ..., [ 1.3371e-03, 2.2907e-03, 1.6737e-03, ..., 6.1393e-05, 2.7962e-03, 2.3329e-04], [-1.7746e-02, -1.9470e-02, -6.0120e-03, ..., 2.0428e-03, -1.1093e-02, -4.0932e-03], [ 8.6136e-03, 1.0872e-02, 1.6754e-02, ..., 1.2791e-04, 2.9449e-02, 6.5613e-03]], device='cuda:0') Epoch 4, bias, value: tensor([ 0.0267, -0.0154, -0.0060, 0.0005, -0.0002, -0.0028, -0.0093, -0.0052, 0.0303, 0.0199], device='cuda:0'), grad: tensor([-0.0111, 0.0097, 0.0061, 0.0003, -0.0330, 0.0131, -0.0012, 0.0043, -0.0292, 0.0409], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 3---------------------------------------------------- epoch 3, time 275.17, cls_loss 0.2248 cls_loss_mapping 0.3588 cls_loss_causal 1.5569 re_mapping 0.1233 re_causal 0.2187 /// teacc 94.49 lr 0.00010000 Epoch 5, weight, value: tensor([[-0.0167, -0.0278, 0.0086, ..., 0.0312, -0.0387, 0.0103], [ 0.0115, -0.0291, -0.0284, ..., -0.0160, 0.0131, -0.0324], [ 0.0083, 0.0344, -0.0164, ..., -0.0371, -0.0224, -0.0089], ..., [-0.0120, 0.0027, 0.0096, ..., -0.0394, -0.0391, 0.0235], [-0.0226, 0.0225, 0.0301, ..., -0.0079, 0.0088, -0.0190], [-0.0251, -0.0160, 0.0233, ..., 0.0091, -0.0188, 0.0174]], device='cuda:0'), grad: tensor([[-1.3962e-03, 1.3437e-03, 9.4366e-04, ..., -5.6791e-04, 7.7915e-04, -5.2567e-03], [ 1.2985e-02, 9.7885e-03, 6.7787e-03, ..., 3.8099e-04, 1.1681e-02, 4.8447e-03], [ 1.0635e-02, -5.5199e-03, 1.2922e-03, ..., 2.0492e-04, -3.4885e-03, 1.6418e-02], ..., [-4.4739e-02, -1.3489e-02, 2.5439e-04, ..., 2.2233e-05, 6.1417e-04, -3.6621e-02], [ 4.3564e-03, 5.6076e-04, 4.6349e-03, ..., 1.3723e-03, 3.5515e-03, 3.1452e-03], [ 5.5428e-03, 3.4943e-03, 1.2045e-03, ..., 1.3709e-04, 1.0433e-03, 6.0501e-03]], device='cuda:0') Epoch 5, bias, value: tensor([ 0.0268, -0.0152, -0.0061, 0.0004, -0.0001, -0.0028, -0.0095, -0.0051, 0.0300, 0.0202], device='cuda:0'), grad: tensor([-0.0038, 0.0334, -0.0076, 0.0182, 0.0019, -0.0165, 0.0015, -0.0421, 0.0070, 0.0080], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 4---------------------------------------------------- epoch 4, time 275.33, cls_loss 0.1651 cls_loss_mapping 0.2728 cls_loss_causal 1.4585 re_mapping 0.1028 re_causal 0.2002 /// teacc 96.07 lr 0.00010000 Epoch 6, weight, value: tensor([[-0.0164, -0.0282, 0.0070, ..., 0.0306, -0.0408, 0.0110], [ 0.0100, -0.0311, -0.0302, ..., -0.0184, 0.0131, -0.0339], [ 0.0103, 0.0358, -0.0182, ..., -0.0392, -0.0255, -0.0099], ..., [-0.0110, 0.0029, 0.0091, ..., -0.0416, -0.0409, 0.0245], [-0.0223, 0.0240, 0.0310, ..., -0.0101, 0.0093, -0.0211], [-0.0265, -0.0161, 0.0225, ..., 0.0083, -0.0197, 0.0178]], device='cuda:0'), grad: tensor([[ 4.1351e-03, 9.9106e-03, 4.4174e-03, ..., 6.6042e-05, 1.3571e-03, 1.6756e-03], [-9.1267e-04, -5.9471e-03, -3.2749e-03, ..., 2.9579e-06, -5.1689e-03, 3.9434e-04], [ 2.2793e-03, 5.3329e-03, 3.4237e-03, ..., 2.2739e-05, 3.2387e-03, 9.9754e-04], ..., [ 5.0316e-03, 5.3329e-03, 1.4944e-03, ..., 5.1269e-07, 5.6152e-03, 1.1543e-02], [-1.4236e-02, -3.2684e-02, -5.1666e-02, ..., 1.6034e-05, -3.2288e-02, -8.8453e-04], [ 8.9264e-03, 9.1705e-03, 2.9278e-03, ..., 5.0664e-06, 5.1041e-03, 1.6739e-02]], device='cuda:0') Epoch 6, bias, value: tensor([ 2.6959e-02, -1.5339e-02, -5.8197e-03, 1.9760e-04, -3.5663e-05, -2.6938e-03, -9.9802e-03, -4.9399e-03, 3.0016e-02, 2.0198e-02], device='cuda:0'), grad: tensor([ 0.0058, -0.0184, 0.0090, -0.0263, 0.0036, 0.0185, 0.0043, 0.0122, -0.0270, 0.0182], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 5---------------------------------------------------- epoch 5, time 275.76, cls_loss 0.1362 cls_loss_mapping 0.2158 cls_loss_causal 1.2994 re_mapping 0.0852 re_causal 0.1719 /// teacc 96.45 lr 0.00010000 Epoch 7, weight, value: tensor([[-0.0162, -0.0287, 0.0053, ..., 0.0312, -0.0428, 0.0110], [ 0.0088, -0.0325, -0.0323, ..., -0.0203, 0.0123, -0.0347], [ 0.0118, 0.0360, -0.0202, ..., -0.0399, -0.0283, -0.0110], ..., [-0.0097, 0.0040, 0.0083, ..., -0.0435, -0.0432, 0.0254], [-0.0218, 0.0257, 0.0320, ..., -0.0124, 0.0099, -0.0227], [-0.0276, -0.0165, 0.0222, ..., 0.0070, -0.0200, 0.0186]], device='cuda:0'), grad: tensor([[-5.1842e-03, -2.6017e-05, 9.9182e-04, ..., 6.9261e-05, 9.1028e-04, -1.3313e-03], [ 1.1246e-02, 1.3443e-02, 2.9488e-03, ..., 1.2599e-05, 1.8740e-03, 2.2755e-03], [ 5.3329e-03, 1.4496e-02, 6.5842e-03, ..., 4.1097e-05, 5.6343e-03, 1.4210e-03], ..., [ 5.6915e-03, 5.5923e-03, 5.3501e-04, ..., 2.8126e-06, 1.5879e-03, 1.9398e-03], [-2.5360e-02, -4.3243e-02, -1.6205e-02, ..., -9.5218e-06, -1.3893e-02, -4.1237e-03], [ 7.5493e-03, 1.8549e-03, 8.3303e-04, ..., 1.6302e-05, 5.2834e-03, 6.4583e-03]], device='cuda:0') Epoch 7, bias, value: tensor([ 2.7292e-02, -1.5538e-02, -6.1149e-03, 1.5834e-04, -5.4673e-05, -2.4777e-03, -1.0435e-02, -4.7028e-03, 3.0156e-02, 2.0233e-02], device='cuda:0'), grad: tensor([-0.0057, 0.0148, 0.0147, 0.0049, -0.0163, 0.0052, 0.0021, 0.0110, -0.0480, 0.0173], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 6---------------------------------------------------- epoch 6, time 275.37, cls_loss 0.1168 cls_loss_mapping 0.1849 cls_loss_causal 1.2598 re_mapping 0.0758 re_causal 0.1636 /// teacc 96.88 lr 0.00010000 Epoch 8, weight, value: tensor([[-0.0158, -0.0294, 0.0044, ..., 0.0310, -0.0444, 0.0110], [ 0.0074, -0.0339, -0.0338, ..., -0.0220, 0.0116, -0.0350], [ 0.0132, 0.0368, -0.0214, ..., -0.0410, -0.0301, -0.0114], ..., [-0.0093, 0.0043, 0.0080, ..., -0.0435, -0.0451, 0.0257], [-0.0216, 0.0272, 0.0327, ..., -0.0151, 0.0102, -0.0243], [-0.0284, -0.0169, 0.0219, ..., 0.0055, -0.0199, 0.0193]], device='cuda:0'), grad: tensor([[ 1.6117e-04, 1.7271e-03, 3.1128e-03, ..., 1.2932e-03, 2.5558e-03, 1.3103e-03], [ 5.2977e-04, 3.5954e-04, 4.4060e-03, ..., 1.4436e-04, 9.2163e-03, 2.8019e-03], [-2.1954e-03, -2.5630e-04, 7.5340e-04, ..., 2.2316e-04, 7.7629e-04, 3.6025e-04], ..., [-6.0368e-04, -7.8964e-04, 8.6546e-04, ..., 6.8605e-05, 1.1816e-03, -7.6103e-03], [ 5.0640e-04, -1.1462e-04, 1.9646e-03, ..., 3.8719e-04, 3.6583e-03, 2.8534e-03], [ 4.0674e-04, -8.3542e-04, 3.5229e-03, ..., 1.3733e-04, 4.4632e-03, 3.6087e-03]], device='cuda:0') Epoch 8, bias, value: tensor([ 2.7674e-02, -1.5673e-02, -6.1870e-03, 6.2555e-04, 1.7484e-05, -2.7191e-03, -1.0810e-02, -4.7630e-03, 3.0345e-02, 2.0064e-02], device='cuda:0'), grad: tensor([ 3.7212e-03, 1.4145e-02, -8.8692e-05, -2.6760e-03, 4.8904e-03, -2.1713e-02, -4.9896e-03, -4.4518e-03, 5.8861e-03, 5.2910e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 7---------------------------------------------------- epoch 7, time 275.10, cls_loss 0.1027 cls_loss_mapping 0.1569 cls_loss_causal 1.2471 re_mapping 0.0680 re_causal 0.1488 /// teacc 97.03 lr 0.00010000 Epoch 9, weight, value: tensor([[-0.0156, -0.0299, 0.0031, ..., 0.0297, -0.0468, 0.0114], [ 0.0063, -0.0354, -0.0353, ..., -0.0237, 0.0108, -0.0350], [ 0.0142, 0.0367, -0.0226, ..., -0.0419, -0.0326, -0.0122], ..., [-0.0083, 0.0053, 0.0078, ..., -0.0443, -0.0463, 0.0264], [-0.0212, 0.0285, 0.0337, ..., -0.0165, 0.0109, -0.0268], [-0.0298, -0.0172, 0.0215, ..., 0.0049, -0.0202, 0.0196]], device='cuda:0'), grad: tensor([[ 0.0045, 0.0041, 0.0011, ..., 0.0008, 0.0015, 0.0043], [ 0.0033, 0.0057, 0.0042, ..., 0.0010, 0.0025, 0.0004], [-0.0218, -0.0076, 0.0026, ..., 0.0016, 0.0008, 0.0013], ..., [-0.0017, -0.0050, 0.0027, ..., -0.0013, 0.0028, -0.0038], [ 0.0105, -0.0063, -0.0068, ..., -0.0011, -0.0013, 0.0034], [ 0.0126, 0.0043, 0.0100, ..., 0.0003, 0.0115, 0.0250]], device='cuda:0') Epoch 9, bias, value: tensor([ 0.0275, -0.0158, -0.0063, 0.0006, 0.0002, -0.0028, -0.0108, -0.0046, 0.0307, 0.0198], device='cuda:0'), grad: tensor([ 0.0071, 0.0078, -0.0177, -0.0286, 0.0060, 0.0042, 0.0003, -0.0055, 0.0024, 0.0240], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 8---------------------------------------------------- epoch 8, time 275.44, cls_loss 0.0948 cls_loss_mapping 0.1475 cls_loss_causal 1.1582 re_mapping 0.0588 re_causal 0.1363 /// teacc 97.49 lr 0.00010000 Epoch 10, weight, value: tensor([[-0.0159, -0.0307, 0.0018, ..., 0.0284, -0.0487, 0.0112], [ 0.0049, -0.0366, -0.0364, ..., -0.0258, 0.0102, -0.0351], [ 0.0153, 0.0370, -0.0241, ..., -0.0430, -0.0346, -0.0128], ..., [-0.0077, 0.0052, 0.0073, ..., -0.0423, -0.0474, 0.0269], [-0.0211, 0.0294, 0.0339, ..., -0.0176, 0.0107, -0.0284], [-0.0296, -0.0169, 0.0213, ..., 0.0035, -0.0204, 0.0202]], device='cuda:0'), grad: tensor([[-9.4080e-04, 6.1035e-04, 1.6289e-03, ..., 2.5864e-03, 1.5965e-03, 1.2102e-03], [ 2.3592e-04, -1.4675e-04, -1.8759e-03, ..., 9.8467e-05, -3.3894e-03, 8.9183e-06], [-1.1005e-03, 1.7726e-04, 1.0118e-03, ..., 5.4932e-04, 1.4162e-03, 1.1158e-04], ..., [ 4.2230e-05, 4.5359e-05, 1.6916e-04, ..., 6.6698e-05, 3.3689e-04, -6.2513e-04], [ 1.0502e-04, -3.4027e-03, -3.3131e-03, ..., 1.8907e-04, -4.8180e-03, 2.4939e-04], [ 5.8460e-04, 2.6584e-04, 4.5967e-04, ..., 1.7285e-04, 1.3924e-04, -2.4247e-04]], device='cuda:0') Epoch 10, bias, value: tensor([ 0.0272, -0.0162, -0.0062, 0.0007, 0.0005, -0.0029, -0.0110, -0.0046, 0.0307, 0.0201], device='cuda:0'), grad: tensor([ 3.6240e-05, -1.0551e-02, 3.1033e-03, 2.2335e-03, 6.9542e-03, 2.2984e-03, -2.6703e-03, 4.8518e-04, -3.0575e-03, 1.1654e-03], device='cuda:0') 100 0.0001 changing lr epoch 9, time 259.07, cls_loss 0.0863 cls_loss_mapping 0.1368 cls_loss_causal 1.1418 re_mapping 0.0532 re_causal 0.1240 /// teacc 97.38 lr 0.00010000 Epoch 11, weight, value: tensor([[-0.0160, -0.0312, 0.0003, ..., 0.0278, -0.0507, 0.0104], [ 0.0042, -0.0381, -0.0380, ..., -0.0272, 0.0094, -0.0347], [ 0.0164, 0.0377, -0.0252, ..., -0.0448, -0.0364, -0.0136], ..., [-0.0072, 0.0056, 0.0071, ..., -0.0394, -0.0488, 0.0273], [-0.0214, 0.0300, 0.0349, ..., -0.0188, 0.0113, -0.0301], [-0.0303, -0.0171, 0.0210, ..., 0.0034, -0.0205, 0.0206]], device='cuda:0'), grad: tensor([[-2.1534e-03, -5.0116e-04, 4.2892e-04, ..., 1.6057e-04, 5.1308e-04, -2.2049e-03], [ 8.5831e-04, 6.3992e-04, 1.8988e-03, ..., -2.0790e-04, 2.4452e-03, 7.0620e-04], [-1.3695e-03, -2.2292e-04, 2.9240e-03, ..., 4.0960e-04, 3.6659e-03, 2.9430e-03], ..., [ 1.2608e-03, 7.4530e-04, 2.9731e-04, ..., 6.2227e-05, 8.0490e-04, 1.6470e-03], [ 3.3307e-04, -2.5196e-03, -1.0853e-03, ..., -1.3084e-03, -2.2049e-03, 1.9054e-03], [ 1.6022e-04, -1.5306e-04, 8.3590e-04, ..., 8.1420e-05, 3.0041e-04, -3.6693e-04]], device='cuda:0') Epoch 11, bias, value: tensor([ 0.0272, -0.0163, -0.0059, 0.0012, 0.0007, -0.0031, -0.0112, -0.0047, 0.0305, 0.0200], device='cuda:0'), grad: tensor([-0.0053, 0.0016, 0.0048, -0.0148, 0.0081, 0.0005, 0.0014, 0.0035, -0.0003, 0.0005], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 10---------------------------------------------------- epoch 10, time 276.13, cls_loss 0.0698 cls_loss_mapping 0.1133 cls_loss_causal 1.0684 re_mapping 0.0501 re_causal 0.1200 /// teacc 97.53 lr 0.00010000 Epoch 12, weight, value: tensor([[-0.0157, -0.0318, -0.0004, ..., 0.0270, -0.0518, 0.0098], [ 0.0035, -0.0390, -0.0397, ..., -0.0275, 0.0083, -0.0343], [ 0.0171, 0.0380, -0.0266, ..., -0.0450, -0.0378, -0.0143], ..., [-0.0067, 0.0056, 0.0067, ..., -0.0390, -0.0499, 0.0280], [-0.0209, 0.0313, 0.0360, ..., -0.0191, 0.0123, -0.0307], [-0.0311, -0.0173, 0.0208, ..., 0.0027, -0.0206, 0.0210]], device='cuda:0'), grad: tensor([[-1.9658e-04, 5.9307e-05, 6.4611e-05, ..., -2.8804e-05, 7.4029e-05, -7.6115e-05], [ 4.1604e-04, 2.2936e-04, 7.2896e-05, ..., 1.5704e-07, 1.4246e-04, 2.1124e-04], [-1.3828e-04, -2.3472e-04, 2.4080e-04, ..., 3.2429e-06, 2.5415e-04, 3.6740e-04], ..., [-4.5443e-04, -5.1737e-04, 1.2994e-04, ..., 4.3027e-07, 2.0766e-04, -8.0395e-04], [ 3.2616e-04, 2.5225e-04, 4.6062e-04, ..., 1.8766e-06, 4.5991e-04, 4.1175e-04], [ 2.9492e-04, 1.1909e-04, 2.9817e-05, ..., 1.6894e-06, 7.3314e-06, 4.8965e-05]], device='cuda:0') Epoch 12, bias, value: tensor([ 0.0273, -0.0163, -0.0058, 0.0009, 0.0006, -0.0032, -0.0116, -0.0046, 0.0311, 0.0200], device='cuda:0'), grad: tensor([-4.8041e-04, 3.2449e-04, 7.5960e-04, -1.2484e-03, -7.0906e-04, -7.9393e-05, 5.6791e-04, -2.0337e-04, 8.4162e-04, 2.2495e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 11---------------------------------------------------- epoch 11, time 275.13, cls_loss 0.0733 cls_loss_mapping 0.1179 cls_loss_causal 1.0573 re_mapping 0.0450 re_causal 0.1138 /// teacc 97.79 lr 0.00010000 Epoch 13, weight, value: tensor([[-0.0158, -0.0325, -0.0011, ..., 0.0252, -0.0530, 0.0094], [ 0.0027, -0.0400, -0.0405, ..., -0.0281, 0.0080, -0.0346], [ 0.0181, 0.0382, -0.0282, ..., -0.0455, -0.0396, -0.0149], ..., [-0.0066, 0.0057, 0.0059, ..., -0.0382, -0.0510, 0.0284], [-0.0208, 0.0320, 0.0364, ..., -0.0200, 0.0124, -0.0319], [-0.0319, -0.0176, 0.0207, ..., 0.0017, -0.0208, 0.0211]], device='cuda:0'), grad: tensor([[-4.0293e-05, 1.3294e-03, 6.8521e-04, ..., 2.5063e-03, 1.4372e-03, 1.1206e-03], [ 3.0756e-04, 1.4949e-04, 1.5473e-04, ..., 2.2709e-05, 2.1458e-05, -5.8031e-04], [-2.1782e-03, 4.8614e-04, 2.3127e-04, ..., 8.4758e-05, 2.5344e-04, 2.5105e-04], ..., [-1.8482e-03, -2.4815e-03, 2.4700e-04, ..., 1.1981e-05, 2.2495e-04, -3.7346e-03], [-1.3113e-04, -1.0214e-03, 1.0830e-04, ..., 1.1653e-04, -5.3406e-04, 9.6178e-04], [ 1.0929e-03, 1.1120e-03, 6.8951e-04, ..., 9.6202e-05, 4.7135e-04, 1.5631e-03]], device='cuda:0') Epoch 13, bias, value: tensor([ 0.0275, -0.0162, -0.0058, 0.0013, 0.0008, -0.0034, -0.0119, -0.0047, 0.0310, 0.0198], device='cuda:0'), grad: tensor([ 2.3193e-03, -1.3065e-03, -5.6601e-04, 3.1261e-03, 1.6251e-03, -2.4681e-03, -1.7252e-03, -2.4624e-03, -2.5451e-05, 1.4830e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 12---------------------------------------------------- epoch 12, time 275.01, cls_loss 0.0677 cls_loss_mapping 0.1051 cls_loss_causal 1.0156 re_mapping 0.0419 re_causal 0.1046 /// teacc 97.96 lr 0.00010000 Epoch 14, weight, value: tensor([[-0.0159, -0.0332, -0.0023, ..., 0.0248, -0.0545, 0.0091], [ 0.0016, -0.0416, -0.0421, ..., -0.0291, 0.0072, -0.0346], [ 0.0189, 0.0386, -0.0294, ..., -0.0465, -0.0412, -0.0155], ..., [-0.0064, 0.0063, 0.0051, ..., -0.0377, -0.0522, 0.0287], [-0.0206, 0.0332, 0.0372, ..., -0.0213, 0.0130, -0.0323], [-0.0323, -0.0176, 0.0201, ..., 0.0008, -0.0214, 0.0216]], device='cuda:0'), grad: tensor([[ 1.3990e-03, 7.7133e-03, 4.5547e-03, ..., 3.2444e-03, 4.1695e-03, 1.7941e-04], [ 6.0129e-04, 6.5851e-04, 2.9945e-04, ..., 8.0943e-05, 2.5225e-04, 2.2936e-04], [ 1.3189e-03, 2.4948e-03, 6.9475e-04, ..., 3.8910e-04, 6.5422e-04, 1.1539e-03], ..., [-3.8185e-03, -6.2895e-04, 3.1328e-04, ..., 2.8044e-05, 3.4261e-04, -4.0932e-03], [-1.9951e-03, 2.5024e-03, 1.0941e-02, ..., 5.8289e-03, 1.0803e-02, 8.0681e-04], [-9.2220e-04, -1.0592e-04, -1.7118e-03, ..., 1.9836e-04, -3.2444e-03, -1.8850e-05]], device='cuda:0') Epoch 14, bias, value: tensor([ 0.0275, -0.0166, -0.0059, 0.0011, 0.0013, -0.0034, -0.0121, -0.0048, 0.0314, 0.0198], device='cuda:0'), grad: tensor([ 0.0086, 0.0004, 0.0040, 0.0004, 0.0041, -0.0027, -0.0134, -0.0040, 0.0061, -0.0036], device='cuda:0') 100 0.0001 changing lr epoch 13, time 259.21, cls_loss 0.0587 cls_loss_mapping 0.0894 cls_loss_causal 0.9640 re_mapping 0.0414 re_causal 0.1019 /// teacc 97.94 lr 0.00010000 Epoch 15, weight, value: tensor([[-0.0162, -0.0341, -0.0034, ..., 0.0234, -0.0562, 0.0086], [ 0.0008, -0.0426, -0.0431, ..., -0.0305, 0.0063, -0.0338], [ 0.0196, 0.0385, -0.0310, ..., -0.0478, -0.0428, -0.0163], ..., [-0.0059, 0.0071, 0.0048, ..., -0.0376, -0.0526, 0.0290], [-0.0201, 0.0340, 0.0378, ..., -0.0226, 0.0137, -0.0334], [-0.0327, -0.0175, 0.0196, ..., 0.0003, -0.0218, 0.0219]], device='cuda:0'), grad: tensor([[-2.4796e-03, -1.9350e-03, -4.6396e-04, ..., 5.8442e-05, 1.6081e-04, -2.4204e-03], [ 4.5705e-04, 1.6272e-04, 1.8334e-04, ..., 2.9579e-05, 1.8096e-04, 2.1964e-05], [ 1.2755e-04, 6.9666e-04, 3.7432e-04, ..., 3.8892e-05, 2.8539e-04, 3.1447e-04], ..., [ 3.2496e-04, 3.2276e-05, 1.6832e-04, ..., 1.0371e-05, 2.5487e-04, -1.5993e-03], [-1.0300e-03, -1.4858e-03, -7.5102e-04, ..., 7.2062e-05, -7.7200e-04, 2.6107e-04], [ 3.6240e-04, 4.4346e-04, 2.7990e-04, ..., 3.4750e-05, 2.0468e-04, 1.4162e-03]], device='cuda:0') Epoch 15, bias, value: tensor([ 0.0273, -0.0165, -0.0061, 0.0010, 0.0010, -0.0032, -0.0121, -0.0047, 0.0318, 0.0198], device='cuda:0'), grad: tensor([-0.0072, 0.0001, 0.0013, 0.0072, -0.0003, -0.0020, 0.0002, -0.0007, -0.0009, 0.0023], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 14---------------------------------------------------- epoch 14, time 273.71, cls_loss 0.0544 cls_loss_mapping 0.0873 cls_loss_causal 0.9824 re_mapping 0.0371 re_causal 0.0995 /// teacc 98.09 lr 0.00010000 Epoch 16, weight, value: tensor([[-0.0161, -0.0346, -0.0045, ..., 0.0224, -0.0576, 0.0081], [-0.0001, -0.0436, -0.0440, ..., -0.0319, 0.0057, -0.0328], [ 0.0200, 0.0387, -0.0328, ..., -0.0487, -0.0445, -0.0171], ..., [-0.0051, 0.0074, 0.0043, ..., -0.0373, -0.0536, 0.0296], [-0.0196, 0.0353, 0.0387, ..., -0.0229, 0.0144, -0.0343], [-0.0337, -0.0178, 0.0194, ..., -0.0004, -0.0220, 0.0218]], device='cuda:0'), grad: tensor([[ 2.3508e-04, 3.7694e-04, 5.9843e-04, ..., 1.8251e-04, 5.4359e-04, 2.1160e-04], [ 1.4219e-03, 4.7922e-04, 1.7297e-04, ..., 1.2279e-05, 1.6761e-04, 1.3409e-03], [-1.8711e-03, -6.0415e-04, 2.1315e-04, ..., 1.8567e-05, 1.6785e-04, 1.6475e-04], ..., [-2.1744e-03, -8.3065e-04, 5.0735e-04, ..., 3.1330e-06, 4.6039e-04, -3.9368e-03], [-3.4885e-03, 1.0394e-05, -8.5020e-04, ..., 1.0908e-04, 8.2016e-04, 7.2479e-04], [ 8.4543e-04, -1.8082e-03, -3.4943e-03, ..., 1.7852e-05, -6.1531e-03, 6.4754e-04]], device='cuda:0') Epoch 16, bias, value: tensor([ 0.0275, -0.0163, -0.0066, 0.0013, 0.0011, -0.0034, -0.0123, -0.0043, 0.0319, 0.0194], device='cuda:0'), grad: tensor([ 0.0008, 0.0022, -0.0034, 0.0082, 0.0017, -0.0023, 0.0025, -0.0035, -0.0007, -0.0055], device='cuda:0') 100 0.0001 changing lr epoch 15, time 255.52, cls_loss 0.0535 cls_loss_mapping 0.0839 cls_loss_causal 0.9495 re_mapping 0.0369 re_causal 0.0959 /// teacc 98.03 lr 0.00010000 Epoch 17, weight, value: tensor([[-0.0161, -0.0355, -0.0055, ..., 0.0219, -0.0588, 0.0076], [-0.0012, -0.0451, -0.0455, ..., -0.0341, 0.0048, -0.0325], [ 0.0209, 0.0394, -0.0337, ..., -0.0489, -0.0461, -0.0179], ..., [-0.0050, 0.0077, 0.0038, ..., -0.0371, -0.0539, 0.0299], [-0.0194, 0.0363, 0.0394, ..., -0.0231, 0.0149, -0.0351], [-0.0334, -0.0175, 0.0193, ..., -0.0017, -0.0222, 0.0220]], device='cuda:0'), grad: tensor([[ 8.1062e-04, 6.0886e-05, 7.7844e-05, ..., 4.9502e-05, 6.5386e-05, 7.7784e-05], [ 9.8646e-05, 1.2493e-04, 9.1553e-05, ..., 4.6730e-05, 9.1672e-05, -7.5054e-04], [-5.7869e-03, 1.0622e-04, 5.1498e-05, ..., 3.4809e-05, 6.2287e-05, 4.8709e-04], ..., [-2.0826e-04, -6.9618e-04, 2.2948e-04, ..., -5.6401e-06, 7.0989e-05, 4.9925e-04], [ 2.9016e-04, 1.0705e-04, 2.0409e-04, ..., 6.8903e-05, 1.7440e-04, 2.5558e-04], [ 2.2678e-03, 7.3075e-05, 2.8000e-03, ..., 1.7270e-05, 1.6174e-03, 5.2071e-03]], device='cuda:0') Epoch 17, bias, value: tensor([ 0.0273, -0.0166, -0.0061, 0.0014, 0.0010, -0.0038, -0.0124, -0.0043, 0.0318, 0.0197], device='cuda:0'), grad: tensor([ 0.0008, -0.0011, -0.0052, 0.0041, -0.0101, -0.0021, -0.0001, 0.0015, 0.0007, 0.0117], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 16---------------------------------------------------- epoch 16, time 266.73, cls_loss 0.0541 cls_loss_mapping 0.0815 cls_loss_causal 0.9760 re_mapping 0.0351 re_causal 0.0910 /// teacc 98.33 lr 0.00010000 Epoch 18, weight, value: tensor([[-0.0158, -0.0360, -0.0063, ..., 0.0210, -0.0598, 0.0075], [-0.0024, -0.0461, -0.0466, ..., -0.0349, 0.0040, -0.0329], [ 0.0219, 0.0398, -0.0343, ..., -0.0492, -0.0470, -0.0183], ..., [-0.0045, 0.0082, 0.0036, ..., -0.0362, -0.0550, 0.0306], [-0.0193, 0.0369, 0.0397, ..., -0.0231, 0.0153, -0.0363], [-0.0340, -0.0176, 0.0191, ..., -0.0032, -0.0221, 0.0221]], device='cuda:0'), grad: tensor([[-1.6904e-04, 2.6059e-04, 1.0478e-04, ..., 1.0207e-05, 1.5473e-04, -2.7204e-04], [ 1.5950e-04, 2.2149e-04, 1.3673e-04, ..., 1.3456e-05, -3.1638e-04, -5.9843e-04], [-1.1053e-03, -4.2844e-04, 1.3351e-04, ..., 1.8314e-05, 3.8862e-04, 5.2977e-04], ..., [ 1.2720e-04, 4.7088e-04, 1.1027e-04, ..., 2.7604e-06, 2.4080e-04, 7.0286e-04], [ 2.1291e-04, 1.5929e-05, 8.6010e-05, ..., 2.0254e-04, 1.2672e-04, 5.3930e-04], [-9.3341e-05, -2.0313e-03, 6.1846e-04, ..., 1.6630e-05, -1.5008e-04, -2.4643e-03]], device='cuda:0') Epoch 18, bias, value: tensor([ 0.0274, -0.0172, -0.0061, 0.0014, 0.0010, -0.0037, -0.0125, -0.0038, 0.0319, 0.0197], device='cuda:0'), grad: tensor([-1.1120e-03, -1.1749e-03, -3.9399e-05, 4.1175e-04, 2.1667e-03, -2.9516e-04, 1.1129e-03, 1.4925e-03, 9.9182e-04, -3.5572e-03], device='cuda:0') 100 0.0001 changing lr epoch 17, time 248.92, cls_loss 0.0422 cls_loss_mapping 0.0719 cls_loss_causal 0.9324 re_mapping 0.0328 re_causal 0.0915 /// teacc 98.31 lr 0.00010000 Epoch 19, weight, value: tensor([[-0.0159, -0.0364, -0.0071, ..., 0.0199, -0.0609, 0.0074], [-0.0032, -0.0469, -0.0471, ..., -0.0355, 0.0037, -0.0324], [ 0.0227, 0.0400, -0.0353, ..., -0.0498, -0.0484, -0.0187], ..., [-0.0047, 0.0080, 0.0031, ..., -0.0358, -0.0558, 0.0310], [-0.0188, 0.0376, 0.0405, ..., -0.0242, 0.0158, -0.0372], [-0.0344, -0.0175, 0.0184, ..., -0.0034, -0.0229, 0.0222]], device='cuda:0'), grad: tensor([[-1.3781e-03, 4.9293e-05, 1.1772e-04, ..., -8.6010e-05, 9.2566e-05, -7.7009e-04], [ 5.9456e-05, 1.5724e-04, 2.1446e-04, ..., 3.7163e-05, 1.6594e-04, -1.8701e-05], [ 2.9325e-04, 8.0764e-05, 5.5403e-05, ..., 3.3051e-05, 4.4674e-05, 1.9526e-04], ..., [ 3.6764e-04, -1.5700e-04, 2.8923e-05, ..., 1.3009e-05, 3.3170e-05, 3.9268e-04], [ 6.3539e-05, -3.9291e-04, -3.4118e-04, ..., 1.3554e-04, -1.8382e-04, 1.4913e-04], [ 7.1228e-05, -6.2287e-05, 1.4293e-04, ..., 4.0114e-05, 7.3671e-05, -1.8253e-03]], device='cuda:0') Epoch 19, bias, value: tensor([ 0.0275, -0.0172, -0.0063, 0.0017, 0.0011, -0.0038, -0.0127, -0.0041, 0.0320, 0.0197], device='cuda:0'), grad: tensor([-2.4071e-03, 1.0598e-04, 6.0320e-04, 1.0967e-03, 1.0214e-03, -9.9301e-05, 5.5981e-04, 1.1187e-03, -2.9898e-04, -1.6975e-03], device='cuda:0') 100 0.0001 changing lr epoch 18, time 248.69, cls_loss 0.0388 cls_loss_mapping 0.0684 cls_loss_causal 0.9245 re_mapping 0.0321 re_causal 0.0886 /// teacc 97.92 lr 0.00010000 Epoch 20, weight, value: tensor([[-0.0160, -0.0373, -0.0083, ..., 0.0188, -0.0621, 0.0071], [-0.0038, -0.0477, -0.0477, ..., -0.0358, 0.0035, -0.0314], [ 0.0234, 0.0403, -0.0359, ..., -0.0505, -0.0495, -0.0199], ..., [-0.0044, 0.0083, 0.0027, ..., -0.0351, -0.0565, 0.0316], [-0.0189, 0.0384, 0.0411, ..., -0.0240, 0.0163, -0.0382], [-0.0349, -0.0175, 0.0182, ..., -0.0042, -0.0230, 0.0225]], device='cuda:0'), grad: tensor([[-1.2422e-04, 2.4155e-05, 3.8862e-05, ..., 2.3797e-05, 4.2737e-05, 7.8380e-06], [ 8.8334e-05, -1.2636e-04, 1.9193e-04, ..., 1.5363e-05, 3.2043e-04, -4.8685e-04], [ 2.4915e-04, 2.8706e-04, 2.3627e-04, ..., 3.6955e-05, 4.0126e-04, 5.2261e-04], ..., [-2.1803e-04, 1.2048e-05, 2.0432e-04, ..., 1.1012e-05, 4.2081e-04, 4.2653e-04], [ 1.6057e-04, 3.1185e-04, 2.1636e-04, ..., 1.0586e-04, 2.8443e-04, 3.0160e-04], [-1.1463e-03, -1.4400e-03, -7.0453e-05, ..., 2.1845e-05, -7.5626e-04, -7.3242e-04]], device='cuda:0') Epoch 20, bias, value: tensor([ 0.0271, -0.0169, -0.0065, 0.0018, 0.0014, -0.0037, -0.0127, -0.0039, 0.0320, 0.0195], device='cuda:0'), grad: tensor([-0.0004, -0.0009, 0.0012, -0.0014, 0.0010, 0.0003, 0.0009, 0.0013, 0.0007, -0.0028], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 19---------------------------------------------------- epoch 19, time 265.54, cls_loss 0.0465 cls_loss_mapping 0.0705 cls_loss_causal 0.9298 re_mapping 0.0297 re_causal 0.0807 /// teacc 98.35 lr 0.00010000 Epoch 21, weight, value: tensor([[-0.0163, -0.0379, -0.0095, ..., 0.0180, -0.0633, 0.0068], [-0.0052, -0.0483, -0.0485, ..., -0.0367, 0.0028, -0.0321], [ 0.0241, 0.0399, -0.0372, ..., -0.0516, -0.0506, -0.0204], ..., [-0.0037, 0.0087, 0.0021, ..., -0.0348, -0.0574, 0.0326], [-0.0186, 0.0392, 0.0417, ..., -0.0247, 0.0166, -0.0391], [-0.0353, -0.0175, 0.0176, ..., -0.0050, -0.0234, 0.0225]], device='cuda:0'), grad: tensor([[-9.7036e-04, -3.2640e-04, 6.1870e-05, ..., 3.1501e-05, -1.2577e-04, 5.6565e-05], [ 5.3465e-05, 1.5843e-04, 1.9825e-04, ..., 1.5843e-04, 8.2016e-05, 8.2180e-06], [ 1.0210e-04, 6.5029e-05, 1.0157e-04, ..., 4.9949e-05, 6.1154e-05, 8.2195e-05], ..., [ 3.2544e-05, 4.5681e-04, 8.6784e-04, ..., 3.0518e-04, 3.1543e-04, 1.0329e-04], [ 9.1553e-05, -4.6587e-04, -4.8685e-04, ..., 7.5698e-05, -1.9121e-04, 1.8632e-04], [ 1.8704e-04, -1.8001e-04, 2.7370e-04, ..., 1.1414e-04, 3.3522e-04, -8.9264e-04]], device='cuda:0') Epoch 21, bias, value: tensor([ 0.0268, -0.0176, -0.0064, 0.0013, 0.0015, -0.0030, -0.0129, -0.0034, 0.0319, 0.0196], device='cuda:0'), grad: tensor([-1.6756e-03, 5.9652e-04, 4.3035e-04, 8.8632e-05, -6.3782e-03, 1.2884e-03, 3.1605e-03, 1.9875e-03, 7.6199e-04, -2.6464e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 20---------------------------------------------------- epoch 20, time 265.11, cls_loss 0.0410 cls_loss_mapping 0.0690 cls_loss_causal 0.8732 re_mapping 0.0294 re_causal 0.0804 /// teacc 98.45 lr 0.00010000 Epoch 22, weight, value: tensor([[-0.0159, -0.0382, -0.0100, ..., 0.0174, -0.0642, 0.0067], [-0.0061, -0.0493, -0.0493, ..., -0.0380, 0.0025, -0.0327], [ 0.0245, 0.0401, -0.0379, ..., -0.0521, -0.0516, -0.0204], ..., [-0.0034, 0.0086, 0.0019, ..., -0.0343, -0.0583, 0.0327], [-0.0182, 0.0399, 0.0419, ..., -0.0259, 0.0168, -0.0396], [-0.0352, -0.0175, 0.0175, ..., -0.0057, -0.0237, 0.0234]], device='cuda:0'), grad: tensor([[ 3.5262e-04, 8.0156e-04, 1.1501e-03, ..., 4.7684e-04, 7.1669e-04, 4.0317e-04], [ 4.8614e-04, 4.9162e-04, 1.4758e-04, ..., 6.2771e-06, 1.7190e-04, 1.9085e-04], [ 7.7209e-03, 6.3438e-03, 1.5914e-04, ..., 5.8934e-06, 1.8752e-04, 5.5504e-03], ..., [-1.5793e-02, -1.2459e-02, 6.7353e-05, ..., 1.3132e-06, -2.8777e-04, -1.1002e-02], [ 6.0511e-04, 6.8724e-05, -4.5848e-04, ..., 1.1516e-04, -4.6206e-04, 8.1968e-04], [ 6.4087e-04, 3.4547e-04, 2.5797e-04, ..., 3.8058e-05, 2.8110e-04, 2.1100e-04]], device='cuda:0') Epoch 22, bias, value: tensor([ 0.0270, -0.0180, -0.0064, 0.0011, 0.0014, -0.0029, -0.0132, -0.0034, 0.0319, 0.0201], device='cuda:0'), grad: tensor([ 0.0016, 0.0005, 0.0100, 0.0068, 0.0002, 0.0020, -0.0024, -0.0201, 0.0010, 0.0003], device='cuda:0') 100 0.0001 changing lr epoch 21, time 248.71, cls_loss 0.0329 cls_loss_mapping 0.0601 cls_loss_causal 0.8950 re_mapping 0.0283 re_causal 0.0801 /// teacc 98.33 lr 0.00010000 Epoch 23, weight, value: tensor([[-0.0154, -0.0385, -0.0106, ..., 0.0164, -0.0652, 0.0066], [-0.0070, -0.0502, -0.0500, ..., -0.0389, 0.0021, -0.0320], [ 0.0254, 0.0405, -0.0382, ..., -0.0527, -0.0527, -0.0211], ..., [-0.0031, 0.0091, 0.0014, ..., -0.0345, -0.0588, 0.0330], [-0.0182, 0.0404, 0.0422, ..., -0.0264, 0.0171, -0.0407], [-0.0357, -0.0178, 0.0168, ..., -0.0051, -0.0241, 0.0238]], device='cuda:0'), grad: tensor([[ 6.5923e-05, 3.4118e-04, 1.9467e-04, ..., 9.7334e-05, 7.2300e-05, 7.4983e-05], [ 2.6393e-04, 2.5797e-04, 1.5342e-04, ..., 4.2140e-05, 5.9158e-05, -1.8568e-03], [-1.5659e-03, -2.5787e-03, -8.6737e-04, ..., -7.2336e-04, 4.8757e-05, 2.1279e-04], ..., [-4.4560e-04, -6.5851e-04, 1.0610e-04, ..., 3.5703e-05, 5.7638e-05, -6.3753e-04], [-5.2834e-04, -1.8835e-03, -1.9093e-03, ..., 1.7679e-04, -1.7004e-03, 1.1510e-04], [ 1.0519e-03, 2.4071e-03, 1.5182e-03, ..., 2.9802e-05, 1.1587e-03, 7.0286e-04]], device='cuda:0') Epoch 23, bias, value: tensor([ 0.0273, -0.0177, -0.0065, 0.0010, 0.0019, -0.0029, -0.0136, -0.0036, 0.0317, 0.0201], device='cuda:0'), grad: tensor([ 0.0002, -0.0016, -0.0040, 0.0045, 0.0010, -0.0024, 0.0012, -0.0007, -0.0017, 0.0036], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 22---------------------------------------------------- epoch 22, time 265.08, cls_loss 0.0302 cls_loss_mapping 0.0552 cls_loss_causal 0.8820 re_mapping 0.0280 re_causal 0.0819 /// teacc 98.52 lr 0.00010000 Epoch 24, weight, value: tensor([[-0.0156, -0.0393, -0.0121, ..., 0.0154, -0.0668, 0.0063], [-0.0079, -0.0510, -0.0506, ..., -0.0395, 0.0017, -0.0316], [ 0.0258, 0.0405, -0.0390, ..., -0.0535, -0.0539, -0.0217], ..., [-0.0031, 0.0093, 0.0009, ..., -0.0326, -0.0594, 0.0335], [-0.0181, 0.0411, 0.0426, ..., -0.0263, 0.0175, -0.0417], [-0.0355, -0.0175, 0.0166, ..., -0.0067, -0.0242, 0.0242]], device='cuda:0'), grad: tensor([[ 1.9288e-04, 3.1853e-04, 4.6849e-04, ..., 2.2495e-04, 3.4189e-04, 1.9097e-04], [ 1.0002e-04, 9.2566e-05, 6.3598e-05, ..., 1.1608e-05, 6.3777e-05, -1.6987e-04], [-3.0441e-03, -2.3632e-03, -6.2180e-04, ..., 1.3955e-05, -3.0875e-04, 5.2392e-05], ..., [-1.4620e-03, -1.3533e-03, 4.2647e-05, ..., 2.0321e-06, 4.6700e-05, -1.4238e-03], [ 1.9369e-03, 1.5869e-03, 6.4135e-04, ..., 8.5175e-05, 4.6968e-04, 1.0777e-04], [ 1.5059e-03, 1.3914e-03, 2.5177e-04, ..., 1.3500e-05, 3.3808e-04, 1.3018e-03]], device='cuda:0') Epoch 24, bias, value: tensor([ 0.0270, -0.0176, -0.0065, 0.0010, 0.0020, -0.0025, -0.0137, -0.0037, 0.0315, 0.0203], device='cuda:0'), grad: tensor([ 6.6519e-04, -7.0751e-05, -3.8586e-03, 2.3317e-04, -4.5433e-03, 6.1846e-04, -8.1968e-04, -1.2159e-03, 3.2139e-03, 5.7793e-03], device='cuda:0') 100 0.0001 changing lr epoch 23, time 248.33, cls_loss 0.0266 cls_loss_mapping 0.0512 cls_loss_causal 0.8695 re_mapping 0.0272 re_causal 0.0773 /// teacc 98.31 lr 0.00010000 Epoch 25, weight, value: tensor([[-0.0154, -0.0398, -0.0129, ..., 0.0145, -0.0677, 0.0059], [-0.0084, -0.0520, -0.0520, ..., -0.0406, 0.0009, -0.0303], [ 0.0262, 0.0408, -0.0399, ..., -0.0541, -0.0552, -0.0226], ..., [-0.0026, 0.0099, 0.0003, ..., -0.0311, -0.0600, 0.0339], [-0.0181, 0.0417, 0.0429, ..., -0.0270, 0.0178, -0.0423], [-0.0359, -0.0175, 0.0166, ..., -0.0070, -0.0242, 0.0242]], device='cuda:0'), grad: tensor([[-1.1568e-03, 1.0633e-04, -1.7405e-03, ..., -2.3818e-04, -5.8746e-04, -1.0653e-03], [ 1.2720e-04, 1.5497e-04, 1.9348e-04, ..., 3.5197e-05, 2.1064e-04, -1.2266e-06], [ 8.9884e-04, 1.0729e-03, 8.0633e-04, ..., 9.1672e-05, 8.7547e-04, 8.9169e-05], ..., [ 2.8992e-04, 2.1660e-04, 2.5606e-04, ..., 9.4593e-05, 1.8406e-04, 3.6263e-04], [-1.6766e-03, -2.2373e-03, -1.6623e-03, ..., 4.2647e-05, -1.6565e-03, -8.2910e-05], [ 1.9705e-04, 1.9133e-04, 7.5531e-04, ..., 5.4926e-05, 5.6982e-04, -3.4118e-04]], device='cuda:0') Epoch 25, bias, value: tensor([ 0.0270, -0.0174, -0.0067, 0.0014, 0.0019, -0.0025, -0.0139, -0.0035, 0.0314, 0.0200], device='cuda:0'), grad: tensor([-3.1815e-03, 3.1400e-04, 1.5593e-03, 1.1311e-03, 6.0111e-05, 1.6904e-04, 1.0319e-03, 8.1491e-04, -2.4586e-03, 5.5885e-04], device='cuda:0') 100 0.0001 changing lr epoch 24, time 248.55, cls_loss 0.0304 cls_loss_mapping 0.0529 cls_loss_causal 0.8876 re_mapping 0.0261 re_causal 0.0764 /// teacc 97.92 lr 0.00010000 Epoch 26, weight, value: tensor([[-0.0151, -0.0408, -0.0136, ..., 0.0138, -0.0688, 0.0056], [-0.0094, -0.0527, -0.0527, ..., -0.0415, 0.0006, -0.0300], [ 0.0267, 0.0410, -0.0407, ..., -0.0547, -0.0564, -0.0229], ..., [-0.0025, 0.0103, -0.0002, ..., -0.0309, -0.0605, 0.0341], [-0.0174, 0.0425, 0.0436, ..., -0.0273, 0.0184, -0.0432], [-0.0362, -0.0181, 0.0160, ..., -0.0077, -0.0247, 0.0243]], device='cuda:0'), grad: tensor([[ 7.2837e-05, 1.9276e-04, 2.7373e-05, ..., 5.1379e-05, 6.6459e-05, 8.2374e-05], [ 2.9278e-04, 4.7493e-04, 1.5289e-05, ..., 6.7428e-06, 1.0240e-04, 2.4915e-04], [ 9.8133e-04, 1.3819e-03, 1.5542e-05, ..., 1.0543e-05, 4.1127e-04, 1.1082e-03], ..., [-7.2575e-04, -1.4744e-03, 1.0200e-05, ..., 2.6100e-07, 2.3472e-04, -9.4080e-04], [ 4.9353e-04, 6.6566e-04, 9.0718e-05, ..., 5.1647e-05, 4.2319e-04, 4.2772e-04], [ 1.7190e-04, -8.9854e-06, 1.4269e-04, ..., 2.7977e-06, 5.5456e-04, -1.9205e-04]], device='cuda:0') Epoch 26, bias, value: tensor([ 0.0270, -0.0174, -0.0068, 0.0013, 0.0025, -0.0030, -0.0137, -0.0035, 0.0318, 0.0196], device='cuda:0'), grad: tensor([ 1.9646e-04, 6.7043e-04, 2.3174e-03, -3.3455e-03, -1.5345e-03, 3.9905e-05, 2.0593e-05, -1.4534e-03, 1.4467e-03, 1.6403e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 25---------------------------------------------------- epoch 25, time 264.65, cls_loss 0.0275 cls_loss_mapping 0.0484 cls_loss_causal 0.8157 re_mapping 0.0243 re_causal 0.0696 /// teacc 98.56 lr 0.00010000 Epoch 27, weight, value: tensor([[-0.0147, -0.0416, -0.0145, ..., 0.0127, -0.0697, 0.0053], [-0.0105, -0.0538, -0.0533, ..., -0.0422, 0.0002, -0.0298], [ 0.0275, 0.0416, -0.0414, ..., -0.0551, -0.0572, -0.0236], ..., [-0.0024, 0.0104, -0.0006, ..., -0.0310, -0.0613, 0.0346], [-0.0170, 0.0430, 0.0441, ..., -0.0270, 0.0189, -0.0434], [-0.0366, -0.0178, 0.0156, ..., -0.0074, -0.0251, 0.0244]], device='cuda:0'), grad: tensor([[ 2.6870e-04, 2.1863e-04, 2.1017e-04, ..., 6.9797e-05, 1.2970e-04, 1.6117e-04], [-3.7789e-04, -2.7180e-04, -2.0385e-04, ..., 4.2319e-05, 8.8096e-05, -2.9888e-03], [ 3.2020e-04, -2.2876e-04, 2.8992e-04, ..., -6.6817e-05, 4.7612e-04, 1.0691e-03], ..., [ 2.7313e-03, 2.8419e-03, 2.5892e-04, ..., 1.5944e-05, 3.6430e-04, 2.2526e-03], [ 1.8959e-03, 2.5539e-03, 3.0518e-03, ..., 1.5783e-03, 2.0542e-03, 1.4973e-03], [ 1.5383e-03, 1.3189e-03, -9.0122e-04, ..., 1.4193e-05, -3.7217e-04, 4.2319e-04]], device='cuda:0') Epoch 27, bias, value: tensor([ 0.0269, -0.0175, -0.0065, 0.0010, 0.0022, -0.0025, -0.0137, -0.0033, 0.0319, 0.0193], device='cuda:0'), grad: tensor([ 0.0006, -0.0063, 0.0019, -0.0033, -0.0092, 0.0023, -0.0027, 0.0066, 0.0058, 0.0042], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 26---------------------------------------------------- epoch 26, time 266.96, cls_loss 0.0283 cls_loss_mapping 0.0480 cls_loss_causal 0.8021 re_mapping 0.0240 re_causal 0.0671 /// teacc 98.64 lr 0.00010000 Epoch 28, weight, value: tensor([[-0.0148, -0.0420, -0.0153, ..., 0.0122, -0.0707, 0.0052], [-0.0114, -0.0547, -0.0539, ..., -0.0427, -0.0003, -0.0297], [ 0.0283, 0.0424, -0.0412, ..., -0.0560, -0.0576, -0.0238], ..., [-0.0019, 0.0105, -0.0013, ..., -0.0304, -0.0627, 0.0350], [-0.0171, 0.0436, 0.0447, ..., -0.0264, 0.0194, -0.0441], [-0.0370, -0.0178, 0.0154, ..., -0.0079, -0.0253, 0.0244]], device='cuda:0'), grad: tensor([[-1.4496e-03, 4.0442e-05, 1.0258e-04, ..., 2.4110e-05, 3.5912e-05, -5.0211e-04], [ 9.3102e-05, 3.8326e-05, 3.6180e-05, ..., 3.6508e-06, 4.6134e-05, 1.7464e-05], [ 2.7275e-04, -1.7929e-04, 3.0443e-05, ..., -6.2771e-07, 5.2184e-05, 3.0565e-04], ..., [ 5.5611e-05, 3.9697e-05, 5.5224e-05, ..., 7.2420e-06, 6.0439e-05, -1.5378e-04], [ 4.2105e-04, 8.3804e-05, 3.6240e-04, ..., 6.2287e-05, 3.2258e-04, 2.5177e-04], [-4.5240e-05, -1.0157e-04, 6.4731e-05, ..., 1.6332e-05, 4.2379e-05, -7.9423e-06]], device='cuda:0') Epoch 28, bias, value: tensor([ 0.0267, -0.0177, -0.0063, 0.0009, 0.0024, -0.0023, -0.0141, -0.0030, 0.0317, 0.0193], device='cuda:0'), grad: tensor([-1.6499e-03, 8.7619e-05, 3.5143e-04, 2.0771e-03, 4.3511e-04, -2.7428e-03, 5.1260e-04, 1.5521e-04, 7.5006e-04, 2.3097e-05], device='cuda:0') 100 0.0001 changing lr epoch 27, time 249.27, cls_loss 0.0229 cls_loss_mapping 0.0445 cls_loss_causal 0.7922 re_mapping 0.0240 re_causal 0.0684 /// teacc 98.44 lr 0.00010000 Epoch 29, weight, value: tensor([[-0.0150, -0.0426, -0.0161, ..., 0.0116, -0.0715, 0.0048], [-0.0124, -0.0554, -0.0541, ..., -0.0433, -0.0003, -0.0296], [ 0.0288, 0.0426, -0.0423, ..., -0.0566, -0.0592, -0.0245], ..., [-0.0017, 0.0108, -0.0019, ..., -0.0298, -0.0635, 0.0355], [-0.0168, 0.0441, 0.0452, ..., -0.0270, 0.0197, -0.0449], [-0.0371, -0.0179, 0.0154, ..., -0.0085, -0.0253, 0.0247]], device='cuda:0'), grad: tensor([[ 1.4019e-03, 1.5793e-03, 8.0729e-04, ..., 1.2316e-05, 6.7234e-04, 1.0834e-03], [ 7.3624e-04, 5.3453e-04, 1.8799e-04, ..., 9.7454e-06, 1.7715e-04, -3.5495e-05], [-1.0681e-03, -6.6805e-04, 3.8552e-04, ..., 1.4432e-05, 3.1400e-04, 3.3045e-04], ..., [ 3.8433e-04, 4.2272e-04, 2.9731e-04, ..., 1.2010e-05, 2.8872e-04, 2.0826e-04], [ 2.9297e-03, 2.1114e-03, 3.6774e-03, ..., 2.8539e-04, 2.6093e-03, 2.2125e-03], [-6.7940e-03, -7.9346e-03, -1.9550e-03, ..., 7.2479e-05, -1.2817e-03, -9.5520e-03]], device='cuda:0') Epoch 29, bias, value: tensor([ 0.0265, -0.0177, -0.0063, 0.0012, 0.0027, -0.0022, -0.0144, -0.0029, 0.0317, 0.0191], device='cuda:0'), grad: tensor([ 0.0025, 0.0012, -0.0020, 0.0033, 0.0146, -0.0112, 0.0032, 0.0008, 0.0064, -0.0189], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 28---------------------------------------------------- epoch 28, time 266.38, cls_loss 0.0319 cls_loss_mapping 0.0554 cls_loss_causal 0.8528 re_mapping 0.0215 re_causal 0.0646 /// teacc 98.72 lr 0.00010000 Epoch 30, weight, value: tensor([[-0.0147, -0.0433, -0.0163, ..., 0.0112, -0.0725, 0.0047], [-0.0132, -0.0565, -0.0553, ..., -0.0439, -0.0012, -0.0297], [ 0.0293, 0.0429, -0.0431, ..., -0.0569, -0.0604, -0.0250], ..., [-0.0015, 0.0109, -0.0024, ..., -0.0297, -0.0640, 0.0355], [-0.0168, 0.0445, 0.0453, ..., -0.0274, 0.0200, -0.0457], [-0.0371, -0.0177, 0.0151, ..., -0.0088, -0.0257, 0.0251]], device='cuda:0'), grad: tensor([[-1.7309e-03, -1.4868e-03, 3.8117e-05, ..., 5.2340e-06, 5.0843e-05, -1.2608e-03], [ 2.5272e-04, 1.0693e-04, -3.0351e-04, ..., -1.0028e-05, -3.1924e-04, -5.7411e-04], [ 2.2106e-03, 1.3380e-03, 7.4089e-05, ..., 1.9129e-06, 6.6161e-05, 1.3008e-03], ..., [-2.4853e-03, -1.7023e-03, 1.7571e-04, ..., 1.7239e-06, 1.3912e-04, -2.2411e-03], [ 4.1652e-04, 2.5964e-04, 2.9016e-04, ..., 7.9051e-06, 2.1601e-04, 8.4209e-04], [ 7.5912e-04, 5.1165e-04, 1.7381e-04, ..., 7.4133e-07, 6.4611e-05, 4.2558e-04]], device='cuda:0') Epoch 30, bias, value: tensor([ 0.0262, -0.0181, -0.0066, 0.0013, 0.0025, -0.0019, -0.0146, -0.0027, 0.0317, 0.0195], device='cuda:0'), grad: tensor([-0.0037, -0.0011, 0.0052, 0.0029, -0.0037, -0.0008, 0.0009, -0.0035, 0.0020, 0.0019], device='cuda:0') 100 0.0001 changing lr epoch 29, time 250.53, cls_loss 0.0205 cls_loss_mapping 0.0383 cls_loss_causal 0.7987 re_mapping 0.0232 re_causal 0.0671 /// teacc 98.57 lr 0.00010000 Epoch 31, weight, value: tensor([[-0.0142, -0.0437, -0.0172, ..., 0.0101, -0.0734, 0.0042], [-0.0140, -0.0576, -0.0563, ..., -0.0442, -0.0019, -0.0296], [ 0.0299, 0.0428, -0.0443, ..., -0.0577, -0.0614, -0.0256], ..., [-0.0016, 0.0106, -0.0029, ..., -0.0297, -0.0651, 0.0356], [-0.0160, 0.0459, 0.0462, ..., -0.0272, 0.0208, -0.0456], [-0.0373, -0.0176, 0.0148, ..., -0.0091, -0.0259, 0.0253]], device='cuda:0'), grad: tensor([[-3.7994e-03, -3.8643e-03, 1.5986e-04, ..., 9.9361e-05, -6.7949e-04, 6.1095e-05], [ 1.6994e-03, 2.1534e-03, 2.0921e-04, ..., 1.9982e-05, 6.5136e-04, 7.4685e-05], [ 6.5041e-04, 7.1239e-04, 1.2374e-04, ..., 5.1975e-05, 2.5201e-04, 6.1154e-05], ..., [-6.6710e-04, -5.5075e-04, 9.0301e-05, ..., 3.2671e-06, 1.2982e-04, -6.3133e-04], [-8.6665e-05, -3.4595e-04, -7.9060e-04, ..., 6.2585e-05, -8.4734e-04, 1.5366e-04], [ 6.9046e-04, 5.4932e-04, 5.0306e-04, ..., 1.6898e-05, 6.4993e-04, 2.5511e-04]], device='cuda:0') Epoch 31, bias, value: tensor([ 0.0261, -0.0186, -0.0062, 0.0012, 0.0027, -0.0018, -0.0148, -0.0030, 0.0323, 0.0194], device='cuda:0'), grad: tensor([-1.3702e-02, -4.4405e-05, 2.4719e-03, -5.6601e-04, 6.9923e-03, 1.5545e-03, 1.6031e-03, -3.6335e-04, 6.5804e-05, 1.9855e-03], device='cuda:0') 100 0.0001 changing lr epoch 30, time 251.03, cls_loss 0.0239 cls_loss_mapping 0.0422 cls_loss_causal 0.8037 re_mapping 0.0220 re_causal 0.0649 /// teacc 98.72 lr 0.00010000 Epoch 32, weight, value: tensor([[-0.0129, -0.0440, -0.0177, ..., 0.0088, -0.0743, 0.0045], [-0.0147, -0.0583, -0.0575, ..., -0.0459, -0.0028, -0.0292], [ 0.0302, 0.0431, -0.0445, ..., -0.0582, -0.0615, -0.0261], ..., [-0.0015, 0.0108, -0.0034, ..., -0.0290, -0.0658, 0.0358], [-0.0158, 0.0465, 0.0465, ..., -0.0277, 0.0212, -0.0461], [-0.0376, -0.0175, 0.0147, ..., -0.0097, -0.0260, 0.0253]], device='cuda:0'), grad: tensor([[ 7.9334e-05, 1.2732e-04, 1.9228e-04, ..., 8.9258e-06, 2.3615e-04, 2.9588e-04], [-5.2738e-04, -6.5184e-04, -8.2636e-04, ..., 3.5614e-06, -1.3819e-03, -2.2545e-03], [ 1.5426e-04, 2.7609e-04, 4.4918e-04, ..., 6.3181e-06, 6.6805e-04, 9.7179e-04], ..., [ 5.4568e-05, 4.9531e-05, 4.8786e-05, ..., 3.7197e-06, 6.6280e-05, 1.0383e-04], [ 1.6880e-04, 1.9109e-04, 3.6120e-04, ..., 1.4991e-05, 4.9162e-04, 7.6056e-04], [-1.0097e-04, -2.1362e-04, 5.7459e-05, ..., 2.2333e-06, 5.5015e-05, -3.4356e-04]], device='cuda:0') Epoch 32, bias, value: tensor([ 0.0269, -0.0186, -0.0063, 0.0013, 0.0025, -0.0020, -0.0143, -0.0029, 0.0319, 0.0191], device='cuda:0'), grad: tensor([ 4.7994e-04, -4.2839e-03, 1.9703e-03, 6.7174e-05, 5.9509e-04, -5.3310e-04, 6.2895e-04, 3.3593e-04, 1.5316e-03, -7.9203e-04], device='cuda:0') 100 0.0001 changing lr epoch 31, time 251.47, cls_loss 0.0263 cls_loss_mapping 0.0437 cls_loss_causal 0.8222 re_mapping 0.0223 re_causal 0.0630 /// teacc 98.68 lr 0.00010000 Epoch 33, weight, value: tensor([[-0.0131, -0.0445, -0.0185, ..., 0.0083, -0.0753, 0.0041], [-0.0158, -0.0585, -0.0577, ..., -0.0467, -0.0025, -0.0296], [ 0.0309, 0.0432, -0.0455, ..., -0.0586, -0.0628, -0.0269], ..., [-0.0011, 0.0115, -0.0038, ..., -0.0288, -0.0668, 0.0362], [-0.0155, 0.0471, 0.0468, ..., -0.0280, 0.0215, -0.0467], [-0.0375, -0.0178, 0.0144, ..., -0.0101, -0.0265, 0.0262]], device='cuda:0'), grad: tensor([[ 5.7966e-05, 1.4746e-04, 1.8775e-04, ..., 1.1194e-04, 1.4865e-04, 5.0634e-05], [ 3.4928e-04, 4.8161e-04, 5.2303e-05, ..., 1.7190e-04, 8.5473e-05, 4.6700e-05], [-6.1178e-04, 1.1939e-04, 1.7214e-04, ..., 1.0347e-03, 5.6839e-04, -4.9680e-05], ..., [ 9.8884e-05, 1.5020e-04, 6.4671e-05, ..., 4.8369e-05, 6.4790e-05, 3.9518e-05], [ 7.7665e-05, -2.1935e-05, 1.4663e-04, ..., 1.0633e-04, 3.0696e-05, 8.3089e-05], [-7.5221e-05, -3.4988e-05, 6.0940e-04, ..., 3.2216e-05, 5.8889e-04, 9.5367e-05]], device='cuda:0') Epoch 33, bias, value: tensor([ 0.0267, -0.0186, -0.0063, 0.0010, 0.0025, -0.0019, -0.0142, -0.0026, 0.0316, 0.0193], device='cuda:0'), grad: tensor([ 0.0005, 0.0013, 0.0008, -0.0310, -0.0047, 0.0296, 0.0013, 0.0005, 0.0007, 0.0009], device='cuda:0') 100 0.0001 changing lr epoch 32, time 251.11, cls_loss 0.0226 cls_loss_mapping 0.0419 cls_loss_causal 0.7983 re_mapping 0.0215 re_causal 0.0620 /// teacc 98.54 lr 0.00010000 Epoch 34, weight, value: tensor([[-0.0128, -0.0455, -0.0195, ..., 0.0078, -0.0766, 0.0038], [-0.0167, -0.0597, -0.0584, ..., -0.0472, -0.0032, -0.0297], [ 0.0312, 0.0435, -0.0464, ..., -0.0590, -0.0634, -0.0269], ..., [-0.0006, 0.0120, -0.0041, ..., -0.0292, -0.0676, 0.0366], [-0.0153, 0.0477, 0.0470, ..., -0.0285, 0.0216, -0.0477], [-0.0375, -0.0175, 0.0141, ..., -0.0103, -0.0270, 0.0266]], device='cuda:0'), grad: tensor([[-1.7989e-04, 1.4842e-04, 8.3387e-05, ..., 4.9360e-08, 5.9754e-05, -7.1824e-05], [ 9.0837e-05, -1.9226e-03, -8.7643e-04, ..., 1.0304e-05, 7.2300e-05, -3.7899e-03], [ 1.1578e-03, 3.1300e-03, 2.2526e-03, ..., 3.7253e-05, 1.7557e-03, 4.4274e-04], ..., [ 5.5820e-05, 1.6403e-04, 8.2791e-05, ..., 1.8673e-06, 3.7730e-05, 1.8179e-04], [-1.3781e-03, -2.2449e-03, -2.0199e-03, ..., 9.4891e-05, -2.0885e-03, 2.3060e-03], [ 2.1935e-03, 3.1013e-03, 8.6427e-05, ..., 1.4596e-05, 5.7250e-05, 5.2786e-04]], device='cuda:0') Epoch 34, bias, value: tensor([ 0.0267, -0.0190, -0.0059, 0.0012, 0.0025, -0.0015, -0.0148, -0.0027, 0.0314, 0.0195], device='cuda:0'), grad: tensor([-8.5235e-05, -9.4452e-03, 3.9291e-03, 1.4210e-04, -7.2784e-03, 9.6941e-04, 8.6355e-04, 6.4516e-04, 2.9011e-03, 7.3662e-03], device='cuda:0') 100 0.0001 changing lr epoch 33, time 251.02, cls_loss 0.0200 cls_loss_mapping 0.0398 cls_loss_causal 0.7622 re_mapping 0.0206 re_causal 0.0601 /// teacc 98.57 lr 0.00010000 Epoch 35, weight, value: tensor([[-1.2679e-02, -4.6142e-02, -2.0089e-02, ..., 7.3313e-03, -7.7372e-02, 3.5660e-03], [-1.7384e-02, -6.0510e-02, -5.9200e-02, ..., -4.7669e-02, -4.2675e-03, -2.9394e-02], [ 3.1480e-02, 4.3955e-02, -4.6198e-02, ..., -5.9361e-02, -6.2164e-02, -2.7430e-02], ..., [-1.9993e-05, 1.2685e-02, -4.4425e-03, ..., -2.9124e-02, -6.8417e-02, 3.7467e-02], [-1.5319e-02, 4.8000e-02, 4.7081e-02, ..., -2.9701e-02, 2.1834e-02, -4.8492e-02], [-3.7499e-02, -1.6989e-02, 1.4206e-02, ..., -9.1683e-03, -2.7018e-02, 2.6614e-02]], device='cuda:0'), grad: tensor([[ 6.2823e-05, 2.7204e-04, 1.8573e-04, ..., 1.8120e-04, 2.0158e-04, 8.3745e-05], [ 4.1455e-05, 7.1764e-05, 3.7253e-05, ..., 2.6122e-05, 3.6806e-05, -1.2141e-04], [ 1.0478e-04, 3.6240e-04, 2.5749e-04, ..., 2.4378e-04, 2.8491e-04, 1.3089e-04], ..., [-5.9032e-04, -7.3433e-04, 2.6986e-05, ..., 1.8150e-05, 2.9907e-05, -4.7827e-04], [-1.3769e-04, -1.0377e-04, -2.8396e-04, ..., 1.1867e-04, -1.3876e-04, 8.0764e-05], [ 3.4022e-04, 5.6887e-04, 1.9240e-04, ..., 1.5438e-04, 1.9348e-04, 2.6679e-04]], device='cuda:0') Epoch 35, bias, value: tensor([ 0.0264, -0.0192, -0.0055, 0.0012, 0.0021, -0.0016, -0.0149, -0.0018, 0.0309, 0.0196], device='cuda:0'), grad: tensor([ 3.8791e-04, -1.3006e-04, 6.0892e-04, 1.3661e-04, 3.9506e-04, 1.4029e-03, -2.8305e-03, -7.1478e-04, 1.5618e-06, 7.4005e-04], device='cuda:0') 100 0.0001 changing lr epoch 34, time 251.40, cls_loss 0.0185 cls_loss_mapping 0.0336 cls_loss_causal 0.7600 re_mapping 0.0202 re_causal 0.0584 /// teacc 98.59 lr 0.00010000 Epoch 36, weight, value: tensor([[-1.2800e-02, -4.7033e-02, -2.0935e-02, ..., 6.7311e-03, -7.8751e-02, 3.1389e-03], [-1.8015e-02, -6.1389e-02, -6.0033e-02, ..., -4.8466e-02, -4.9184e-03, -2.9077e-02], [ 3.1717e-02, 4.4053e-02, -4.6511e-02, ..., -5.9664e-02, -6.2287e-02, -2.8211e-02], ..., [ 6.1960e-05, 1.3001e-02, -5.2057e-03, ..., -2.9075e-02, -6.9221e-02, 3.7479e-02], [-1.4722e-02, 4.9189e-02, 4.7954e-02, ..., -2.9769e-02, 2.2954e-02, -4.8978e-02], [-3.7085e-02, -1.6689e-02, 1.3799e-02, ..., -9.4828e-03, -2.7587e-02, 2.6985e-02]], device='cuda:0'), grad: tensor([[ 2.6608e-04, 3.3617e-04, 3.5977e-04, ..., 2.9063e-04, 3.2592e-04, 1.6820e-04], [ 7.5579e-05, 8.8096e-05, 8.0764e-05, ..., 5.7369e-05, 7.5877e-05, -2.7871e-04], [ 4.7994e-04, 6.4230e-04, 5.6505e-04, ..., 5.5885e-04, 5.9366e-04, 3.1090e-04], ..., [-7.2837e-05, -1.3340e-04, 3.3212e-04, ..., 4.5896e-05, 6.5029e-05, 2.4393e-05], [ 4.1342e-04, 4.2081e-04, 5.2071e-04, ..., 5.2214e-04, 4.6301e-04, 3.5715e-04], [ 6.6328e-04, 7.9775e-04, 8.7500e-04, ..., 6.0701e-04, 7.3051e-04, 4.5109e-04]], device='cuda:0') Epoch 36, bias, value: tensor([ 0.0261, -0.0187, -0.0057, 0.0011, 0.0021, -0.0018, -0.0152, -0.0019, 0.0314, 0.0197], device='cuda:0'), grad: tensor([ 0.0008, -0.0005, 0.0015, -0.0005, 0.0015, 0.0010, -0.0077, 0.0004, 0.0014, 0.0020], device='cuda:0') 100 0.0001 changing lr epoch 35, time 253.75, cls_loss 0.0181 cls_loss_mapping 0.0343 cls_loss_causal 0.7322 re_mapping 0.0199 re_causal 0.0576 /// teacc 98.67 lr 0.00010000 Epoch 37, weight, value: tensor([[-0.0122, -0.0469, -0.0215, ..., 0.0064, -0.0794, 0.0031], [-0.0188, -0.0624, -0.0602, ..., -0.0497, -0.0052, -0.0285], [ 0.0325, 0.0445, -0.0469, ..., -0.0605, -0.0628, -0.0290], ..., [ 0.0007, 0.0136, -0.0054, ..., -0.0284, -0.0693, 0.0380], [-0.0152, 0.0493, 0.0479, ..., -0.0302, 0.0229, -0.0499], [-0.0374, -0.0168, 0.0131, ..., -0.0101, -0.0281, 0.0271]], device='cuda:0'), grad: tensor([[ 9.4622e-06, 1.4439e-05, 3.3885e-05, ..., 2.2519e-06, 3.1680e-05, 4.7423e-06], [ 2.0659e-04, 1.1808e-04, 3.4392e-05, ..., 4.1537e-07, 4.9800e-05, 3.1944e-06], [-2.2233e-05, -1.5414e-04, 1.3196e-04, ..., 3.9954e-07, 2.0897e-04, -2.1085e-05], ..., [ 2.4700e-04, 1.0395e-04, 1.1617e-04, ..., 3.7765e-07, 1.5485e-04, 1.0115e-04], [ 2.7442e-04, 5.7459e-05, 3.2711e-04, ..., 7.0743e-06, 3.4308e-04, 4.3303e-05], [ 1.4877e-04, -1.0198e-04, 2.4843e-04, ..., 3.4906e-06, 2.4509e-04, -1.0592e-04]], device='cuda:0') Epoch 37, bias, value: tensor([ 0.0263, -0.0183, -0.0056, 0.0009, 0.0020, -0.0014, -0.0153, -0.0017, 0.0304, 0.0199], device='cuda:0'), grad: tensor([ 4.1544e-05, 2.1851e-04, 2.0683e-04, -2.0447e-03, 1.3149e-04, -8.3303e-04, 6.1321e-04, 5.7411e-04, 7.4244e-04, 3.4833e-04], device='cuda:0') 100 0.0001 changing lr epoch 36, time 256.61, cls_loss 0.0205 cls_loss_mapping 0.0349 cls_loss_causal 0.7607 re_mapping 0.0193 re_causal 0.0562 /// teacc 98.63 lr 0.00010000 Epoch 38, weight, value: tensor([[-0.0117, -0.0471, -0.0222, ..., 0.0063, -0.0800, 0.0031], [-0.0197, -0.0636, -0.0607, ..., -0.0495, -0.0059, -0.0284], [ 0.0327, 0.0442, -0.0482, ..., -0.0610, -0.0630, -0.0298], ..., [ 0.0010, 0.0140, -0.0058, ..., -0.0287, -0.0703, 0.0383], [-0.0145, 0.0501, 0.0484, ..., -0.0310, 0.0234, -0.0505], [-0.0380, -0.0174, 0.0126, ..., -0.0106, -0.0284, 0.0270]], device='cuda:0'), grad: tensor([[-2.7776e-05, 5.8860e-05, 5.5015e-05, ..., 3.9637e-05, 4.3839e-05, 8.3596e-06], [ 2.5943e-05, 5.2214e-05, 5.8770e-05, ..., 2.5898e-05, 5.2840e-05, 4.7803e-05], [ 1.1581e-04, 1.3947e-04, 2.1696e-04, ..., 2.2516e-05, 1.6391e-04, 2.6509e-05], ..., [-2.2388e-04, -9.3579e-05, 5.4061e-05, ..., 2.5108e-06, 7.6890e-05, -4.8661e-04], [-1.8537e-04, -1.7571e-04, -2.0087e-04, ..., 8.6904e-05, -9.6738e-05, 7.8976e-05], [ 2.3377e-04, -4.4256e-05, -5.6982e-05, ..., 7.5027e-06, -1.0657e-04, 3.4881e-04]], device='cuda:0') Epoch 38, bias, value: tensor([ 0.0264, -0.0186, -0.0056, 0.0014, 0.0021, -0.0014, -0.0160, -0.0016, 0.0307, 0.0196], device='cuda:0'), grad: tensor([-4.0317e-04, 3.7599e-04, 5.1880e-04, -2.1923e-04, -2.3861e-03, 4.3750e-04, -2.6792e-05, -7.9155e-04, -2.6926e-05, 2.5215e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 37---------------------------------------------------- epoch 37, time 273.34, cls_loss 0.0145 cls_loss_mapping 0.0319 cls_loss_causal 0.7050 re_mapping 0.0182 re_causal 0.0542 /// teacc 98.73 lr 0.00010000 Epoch 39, weight, value: tensor([[-0.0113, -0.0477, -0.0228, ..., 0.0058, -0.0804, 0.0031], [-0.0207, -0.0652, -0.0615, ..., -0.0502, -0.0067, -0.0284], [ 0.0330, 0.0443, -0.0494, ..., -0.0615, -0.0635, -0.0297], ..., [ 0.0011, 0.0138, -0.0061, ..., -0.0288, -0.0712, 0.0386], [-0.0140, 0.0512, 0.0489, ..., -0.0313, 0.0241, -0.0509], [-0.0381, -0.0172, 0.0123, ..., -0.0110, -0.0286, 0.0272]], device='cuda:0'), grad: tensor([[-8.1003e-05, 5.2750e-05, 6.4135e-05, ..., 2.6643e-05, 4.6432e-05, -3.3118e-06], [ 9.6440e-05, 7.8082e-05, 9.4473e-05, ..., 1.5110e-05, 1.2422e-04, -2.8944e-04], [-2.6932e-03, -3.7327e-03, -2.2850e-03, ..., 1.1243e-05, 5.6297e-05, 3.1519e-04], ..., [ 5.9426e-05, 4.1336e-05, 3.6895e-05, ..., 2.5108e-06, 3.2753e-05, 1.2118e-04], [ 3.4885e-03, 3.7060e-03, 2.7046e-03, ..., 7.4983e-05, 1.1702e-03, 1.3316e-04], [ 4.2051e-05, -1.9118e-05, 2.0429e-05, ..., 6.6608e-06, 1.2144e-05, -2.5868e-05]], device='cuda:0') Epoch 39, bias, value: tensor([ 0.0257, -0.0192, -0.0055, 0.0012, 0.0022, -0.0012, -0.0157, -0.0016, 0.0312, 0.0196], device='cuda:0'), grad: tensor([-1.5604e-04, -3.9148e-04, -2.4567e-03, -1.6499e-03, 1.9455e-04, 3.0136e-04, -3.6097e-04, 3.0994e-04, 4.1313e-03, 7.3910e-05], device='cuda:0') 100 0.0001 changing lr epoch 38, time 256.61, cls_loss 0.0142 cls_loss_mapping 0.0298 cls_loss_causal 0.6949 re_mapping 0.0183 re_causal 0.0540 /// teacc 98.67 lr 0.00010000 Epoch 40, weight, value: tensor([[-0.0107, -0.0478, -0.0233, ..., 0.0060, -0.0813, 0.0033], [-0.0216, -0.0662, -0.0620, ..., -0.0504, -0.0069, -0.0282], [ 0.0330, 0.0442, -0.0499, ..., -0.0620, -0.0642, -0.0310], ..., [ 0.0016, 0.0144, -0.0062, ..., -0.0289, -0.0715, 0.0391], [-0.0135, 0.0518, 0.0489, ..., -0.0320, 0.0244, -0.0516], [-0.0385, -0.0173, 0.0122, ..., -0.0106, -0.0286, 0.0273]], device='cuda:0'), grad: tensor([[-1.3697e-04, 3.8207e-05, 8.3208e-05, ..., 4.3958e-05, 7.2658e-05, 3.4243e-05], [ 4.6402e-05, 2.9728e-05, 2.2829e-05, ..., 2.8789e-05, -6.9916e-05, -5.2691e-04], [-7.7665e-05, -6.2250e-06, 1.2094e-04, ..., 5.2303e-05, 1.0484e-04, 5.8860e-05], ..., [ 2.8759e-06, 2.1234e-05, 1.9386e-05, ..., 5.0813e-06, 3.0518e-05, 3.5226e-05], [ 5.5879e-05, 3.1114e-04, 6.4039e-04, ..., 3.3450e-04, 5.2977e-04, 2.0945e-04], [-2.5854e-05, -8.9526e-05, 2.2963e-05, ..., 5.5805e-06, -1.6227e-05, -1.1110e-04]], device='cuda:0') Epoch 40, bias, value: tensor([ 0.0260, -0.0192, -0.0060, 0.0013, 0.0025, -0.0012, -0.0159, -0.0013, 0.0311, 0.0194], device='cuda:0'), grad: tensor([-2.2686e-04, -8.0061e-04, 4.2260e-05, 1.0628e-04, 2.9016e-04, 9.3508e-04, -1.4267e-03, 2.0897e-04, 9.7322e-04, -9.9063e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 39---------------------------------------------------- epoch 39, time 272.93, cls_loss 0.0165 cls_loss_mapping 0.0317 cls_loss_causal 0.7442 re_mapping 0.0176 re_causal 0.0529 /// teacc 98.76 lr 0.00010000 Epoch 41, weight, value: tensor([[-0.0107, -0.0485, -0.0241, ..., 0.0051, -0.0821, 0.0031], [-0.0227, -0.0674, -0.0625, ..., -0.0517, -0.0072, -0.0270], [ 0.0341, 0.0450, -0.0504, ..., -0.0622, -0.0646, -0.0316], ..., [ 0.0016, 0.0144, -0.0065, ..., -0.0289, -0.0718, 0.0391], [-0.0138, 0.0518, 0.0491, ..., -0.0327, 0.0245, -0.0525], [-0.0382, -0.0171, 0.0117, ..., -0.0111, -0.0289, 0.0278]], device='cuda:0'), grad: tensor([[-2.6170e-06, 1.1522e-04, 1.7273e-04, ..., 3.4004e-05, 1.3566e-04, 5.2631e-05], [ 2.0599e-04, 2.9659e-04, 3.7283e-05, ..., 4.6864e-06, 3.6240e-05, 1.9026e-04], [-2.2388e-04, -4.4227e-04, 1.6242e-05, ..., -1.0580e-04, 7.5579e-05, 1.0252e-04], ..., [-6.3229e-04, -9.3508e-04, 7.0333e-05, ..., 9.2685e-06, -1.4758e-04, -9.5224e-04], [ 4.4751e-04, 6.7091e-04, 4.3416e-04, ..., 9.0718e-05, 4.6086e-04, 6.0320e-04], [ 1.1599e-04, 8.4937e-05, 4.0197e-04, ..., 9.3430e-06, 3.3188e-04, 1.3852e-04]], device='cuda:0') Epoch 41, bias, value: tensor([ 0.0259, -0.0188, -0.0057, 0.0011, 0.0026, -0.0009, -0.0154, -0.0021, 0.0305, 0.0196], device='cuda:0'), grad: tensor([ 0.0002, 0.0003, -0.0005, 0.0009, 0.0002, -0.0018, 0.0004, -0.0017, 0.0015, 0.0006], device='cuda:0') 100 0.0001 changing lr epoch 40, time 256.57, cls_loss 0.0183 cls_loss_mapping 0.0334 cls_loss_causal 0.7049 re_mapping 0.0178 re_causal 0.0517 /// teacc 98.66 lr 0.00010000 Epoch 42, weight, value: tensor([[-0.0100, -0.0491, -0.0239, ..., 0.0041, -0.0825, 0.0025], [-0.0234, -0.0682, -0.0629, ..., -0.0523, -0.0075, -0.0267], [ 0.0349, 0.0456, -0.0511, ..., -0.0620, -0.0652, -0.0307], ..., [ 0.0019, 0.0148, -0.0069, ..., -0.0290, -0.0723, 0.0394], [-0.0136, 0.0525, 0.0494, ..., -0.0330, 0.0248, -0.0531], [-0.0389, -0.0168, 0.0112, ..., -0.0095, -0.0293, 0.0281]], device='cuda:0'), grad: tensor([[-9.2201e-07, 7.0858e-04, 1.1358e-03, ..., 4.7135e-04, 9.0742e-04, 3.2711e-04], [ 3.4213e-05, 5.4747e-05, 5.0545e-05, ..., 3.0294e-05, 4.3780e-05, -4.4137e-05], [-6.5279e-04, -2.3568e-04, 8.0347e-05, ..., 1.3769e-05, 5.4002e-05, 4.3124e-05], ..., [ 6.3777e-05, 8.8289e-06, 1.9193e-05, ..., 8.0764e-06, 1.1280e-05, -6.9588e-06], [ 9.6679e-05, 4.3392e-04, 5.6171e-04, ..., 3.1805e-04, 4.8566e-04, 2.3580e-04], [ 6.8545e-05, 7.4208e-05, 1.4043e-04, ..., 4.2707e-05, 7.1883e-05, 8.9556e-06]], device='cuda:0') Epoch 42, bias, value: tensor([ 0.0262, -0.0188, -0.0053, 0.0014, 0.0028, -0.0015, -0.0160, -0.0023, 0.0306, 0.0198], device='cuda:0'), grad: tensor([ 1.2884e-03, -3.8099e-04, -7.9203e-04, 8.6164e-04, -1.0061e-04, -9.1732e-05, -2.5196e-03, 2.1863e-04, 9.6893e-04, 5.4741e-04], device='cuda:0') 100 0.0001 changing lr epoch 41, time 257.11, cls_loss 0.0179 cls_loss_mapping 0.0310 cls_loss_causal 0.7320 re_mapping 0.0169 re_causal 0.0499 /// teacc 98.73 lr 0.00010000 Epoch 43, weight, value: tensor([[-0.0101, -0.0499, -0.0248, ..., 0.0037, -0.0840, 0.0023], [-0.0240, -0.0697, -0.0634, ..., -0.0529, -0.0079, -0.0266], [ 0.0353, 0.0461, -0.0515, ..., -0.0621, -0.0655, -0.0309], ..., [ 0.0014, 0.0146, -0.0073, ..., -0.0289, -0.0728, 0.0397], [-0.0129, 0.0536, 0.0499, ..., -0.0327, 0.0252, -0.0538], [-0.0386, -0.0166, 0.0106, ..., -0.0101, -0.0298, 0.0287]], device='cuda:0'), grad: tensor([[ 6.9320e-05, 1.3149e-04, 4.0245e-04, ..., 1.5891e-04, 3.0327e-04, 3.6812e-04], [ 9.9659e-05, 1.2457e-04, 1.3404e-05, ..., 4.8690e-06, 1.4193e-05, -1.2293e-05], [ 4.9442e-05, 9.4473e-05, 2.6554e-05, ..., -3.8892e-06, 2.5630e-05, 1.9264e-04], ..., [-4.7517e-04, -5.9319e-04, 1.4558e-05, ..., 4.2729e-06, 1.7673e-05, -4.1389e-04], [-9.1502e-07, 2.7433e-05, -1.7118e-04, ..., 5.8621e-05, -9.1791e-05, 5.2959e-05], [-1.5509e-04, -3.1042e-04, 1.4997e-04, ..., 9.4716e-07, 1.0884e-04, -9.1255e-05]], device='cuda:0') Epoch 43, bias, value: tensor([ 0.0258, -0.0191, -0.0053, 0.0012, 0.0028, -0.0015, -0.0155, -0.0023, 0.0308, 0.0200], device='cuda:0'), grad: tensor([ 8.7309e-04, 4.9531e-05, 1.9717e-04, 1.3733e-04, 1.2531e-03, 7.5400e-05, -9.4032e-04, -8.5163e-04, 1.2052e-04, -9.1362e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 42---------------------------------------------------- epoch 42, time 273.33, cls_loss 0.0146 cls_loss_mapping 0.0286 cls_loss_causal 0.6979 re_mapping 0.0176 re_causal 0.0504 /// teacc 98.80 lr 0.00010000 Epoch 44, weight, value: tensor([[-0.0098, -0.0507, -0.0253, ..., 0.0032, -0.0846, 0.0024], [-0.0248, -0.0711, -0.0639, ..., -0.0538, -0.0084, -0.0267], [ 0.0357, 0.0464, -0.0520, ..., -0.0619, -0.0658, -0.0314], ..., [ 0.0020, 0.0156, -0.0076, ..., -0.0290, -0.0733, 0.0402], [-0.0125, 0.0542, 0.0505, ..., -0.0335, 0.0257, -0.0543], [-0.0388, -0.0165, 0.0102, ..., -0.0105, -0.0301, 0.0287]], device='cuda:0'), grad: tensor([[ 6.1207e-06, 1.1277e-04, 2.5019e-05, ..., 5.8115e-05, 2.9460e-05, 2.7061e-05], [ 6.5006e-06, 1.9848e-05, 1.5721e-05, ..., 3.5595e-06, 1.5423e-05, -4.9978e-05], [-3.9518e-05, -7.9861e-07, 1.9386e-05, ..., 3.9935e-06, 2.9102e-05, 2.5615e-05], ..., [-2.5854e-05, -4.6134e-05, 1.1757e-05, ..., 3.5856e-07, 1.3135e-05, -2.2396e-05], [ 1.1124e-05, -7.3351e-06, 3.2812e-05, ..., 3.0041e-05, 3.6925e-05, 1.6406e-05], [ 2.1875e-05, 3.7879e-05, 3.0577e-05, ..., 1.1206e-05, 2.6032e-05, 1.9744e-05]], device='cuda:0') Epoch 44, bias, value: tensor([ 0.0259, -0.0193, -0.0051, 0.0011, 0.0027, -0.0018, -0.0156, -0.0016, 0.0308, 0.0197], device='cuda:0'), grad: tensor([ 1.2028e-04, -9.7930e-05, 5.1945e-05, -3.4899e-05, 4.3511e-06, -3.0786e-05, -1.9515e-04, -4.2804e-06, 7.6771e-05, 1.0955e-04], device='cuda:0') 100 0.0001 changing lr epoch 43, time 256.20, cls_loss 0.0152 cls_loss_mapping 0.0291 cls_loss_causal 0.7285 re_mapping 0.0167 re_causal 0.0498 /// teacc 98.64 lr 0.00010000 Epoch 45, weight, value: tensor([[-0.0094, -0.0513, -0.0258, ..., 0.0025, -0.0851, 0.0022], [-0.0254, -0.0716, -0.0645, ..., -0.0543, -0.0089, -0.0256], [ 0.0359, 0.0464, -0.0526, ..., -0.0626, -0.0662, -0.0323], ..., [ 0.0027, 0.0164, -0.0078, ..., -0.0291, -0.0738, 0.0404], [-0.0123, 0.0547, 0.0506, ..., -0.0339, 0.0259, -0.0548], [-0.0396, -0.0170, 0.0096, ..., -0.0109, -0.0305, 0.0289]], device='cuda:0'), grad: tensor([[ 7.0632e-05, 1.0157e-04, 7.3075e-05, ..., 1.4700e-05, 5.3883e-05, 2.2873e-05], [ 2.8372e-04, 3.0708e-04, 1.5509e-04, ..., 3.9972e-06, 1.2553e-04, 9.0003e-05], [-6.8951e-04, -2.7418e-04, 1.2147e-04, ..., 1.8533e-06, 9.9719e-05, 1.7092e-05], ..., [ 3.5763e-04, 5.2834e-04, 8.1420e-05, ..., 1.6484e-07, 6.0260e-05, 7.4530e-04], [-3.1519e-04, -1.2045e-03, -8.5545e-04, ..., -6.0070e-07, -6.8903e-04, -1.0735e-04], [-2.0969e-04, -2.4652e-04, 7.6246e-04, ..., 1.7248e-06, 5.6934e-04, -7.3624e-04]], device='cuda:0') Epoch 45, bias, value: tensor([ 0.0256, -0.0187, -0.0055, 0.0011, 0.0024, -0.0015, -0.0156, -0.0014, 0.0306, 0.0197], device='cuda:0'), grad: tensor([ 2.0015e-04, 6.0654e-04, -6.4421e-04, 1.8921e-02, -2.9549e-05, -2.0172e-02, 2.5439e-04, 1.5249e-03, -1.1635e-03, 4.9400e-04], device='cuda:0') 100 0.0001 changing lr epoch 44, time 256.47, cls_loss 0.0120 cls_loss_mapping 0.0246 cls_loss_causal 0.6910 re_mapping 0.0160 re_causal 0.0475 /// teacc 98.79 lr 0.00010000 Epoch 46, weight, value: tensor([[-0.0095, -0.0518, -0.0262, ..., 0.0017, -0.0857, 0.0017], [-0.0265, -0.0722, -0.0649, ..., -0.0550, -0.0090, -0.0249], [ 0.0361, 0.0463, -0.0533, ..., -0.0638, -0.0668, -0.0330], ..., [ 0.0024, 0.0165, -0.0082, ..., -0.0293, -0.0743, 0.0402], [-0.0123, 0.0550, 0.0509, ..., -0.0339, 0.0261, -0.0555], [-0.0397, -0.0169, 0.0092, ..., -0.0113, -0.0309, 0.0291]], device='cuda:0'), grad: tensor([[-1.9986e-06, 3.4451e-05, 1.0818e-05, ..., 1.8980e-06, 1.0133e-05, 2.2873e-05], [ 3.0056e-05, -1.7440e-04, 4.5061e-05, ..., 9.4762e-07, 4.8310e-05, -4.3821e-04], [-3.9995e-05, -1.6894e-06, 2.2337e-05, ..., 4.8568e-07, 2.5377e-05, 3.0547e-05], ..., [-1.5521e-04, -2.6584e-04, 2.1443e-05, ..., 9.5926e-08, 1.7643e-05, 1.3612e-05], [-5.9366e-05, -9.1970e-05, -8.6725e-05, ..., 6.9924e-06, -1.0097e-04, 2.2545e-05], [ 1.5426e-04, 9.2387e-05, -2.8968e-04, ..., 9.3551e-07, -2.2638e-04, -1.1003e-04]], device='cuda:0') Epoch 46, bias, value: tensor([ 0.0250, -0.0185, -0.0058, 0.0017, 0.0026, -0.0014, -0.0148, -0.0018, 0.0302, 0.0196], device='cuda:0'), grad: tensor([-2.5146e-06, -9.7847e-04, 2.3454e-05, 3.3379e-04, 2.7132e-04, 3.4118e-04, 9.8869e-06, 2.4140e-04, -3.2574e-05, -2.0766e-04], device='cuda:0') 100 0.0001 changing lr epoch 45, time 256.69, cls_loss 0.0123 cls_loss_mapping 0.0253 cls_loss_causal 0.6936 re_mapping 0.0158 re_causal 0.0458 /// teacc 98.80 lr 0.00010000 Epoch 47, weight, value: tensor([[-0.0091, -0.0522, -0.0268, ..., 0.0012, -0.0867, 0.0019], [-0.0273, -0.0732, -0.0661, ..., -0.0553, -0.0098, -0.0241], [ 0.0367, 0.0470, -0.0526, ..., -0.0639, -0.0666, -0.0338], ..., [ 0.0025, 0.0163, -0.0086, ..., -0.0293, -0.0754, 0.0400], [-0.0124, 0.0551, 0.0513, ..., -0.0343, 0.0267, -0.0559], [-0.0398, -0.0164, 0.0090, ..., -0.0112, -0.0312, 0.0294]], device='cuda:0'), grad: tensor([[ 7.3835e-06, -1.1559e-03, 5.9485e-05, ..., 2.4527e-05, 6.0916e-05, 7.0743e-06], [ 3.5852e-05, 7.3075e-05, 1.3304e-04, ..., 6.3926e-06, 1.6928e-04, -2.6926e-05], [ 1.7658e-05, 2.9111e-04, 7.8201e-05, ..., 1.3158e-05, 9.2864e-05, 2.6435e-05], ..., [-1.3590e-04, -6.6400e-05, 5.2810e-05, ..., 1.2759e-06, 5.8562e-05, -1.3173e-04], [-1.6391e-04, -2.2674e-04, -5.1689e-04, ..., 2.3618e-05, -5.3883e-04, 1.7956e-05], [ 8.0824e-05, 2.5558e-04, 3.0017e-04, ..., 3.7868e-06, 3.2163e-04, -7.0296e-06]], device='cuda:0') Epoch 47, bias, value: tensor([ 0.0255, -0.0185, -0.0056, 0.0020, 0.0024, -0.0017, -0.0151, -0.0022, 0.0302, 0.0197], device='cuda:0'), grad: tensor([-7.8659e-03, 3.8719e-04, 1.7910e-03, -6.4039e-04, 2.7561e-03, 1.2074e-03, 9.9754e-04, 4.8995e-05, 2.9898e-04, 1.0157e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 46---------------------------------------------------- epoch 46, time 272.31, cls_loss 0.0122 cls_loss_mapping 0.0228 cls_loss_causal 0.7073 re_mapping 0.0153 re_causal 0.0470 /// teacc 98.83 lr 0.00010000 Epoch 48, weight, value: tensor([[-0.0091, -0.0527, -0.0275, ..., 0.0009, -0.0875, 0.0015], [-0.0283, -0.0735, -0.0665, ..., -0.0558, -0.0099, -0.0238], [ 0.0369, 0.0466, -0.0539, ..., -0.0641, -0.0675, -0.0344], ..., [ 0.0027, 0.0167, -0.0089, ..., -0.0294, -0.0758, 0.0401], [-0.0119, 0.0557, 0.0517, ..., -0.0348, 0.0270, -0.0566], [-0.0397, -0.0165, 0.0086, ..., -0.0107, -0.0316, 0.0298]], device='cuda:0'), grad: tensor([[ 6.6310e-07, 1.0431e-04, 4.8280e-05, ..., 2.0992e-06, 2.7753e-06, 5.3018e-05], [ 1.6361e-05, 3.4332e-05, 2.0549e-05, ..., 4.6706e-07, 1.1303e-05, -5.0694e-05], [-6.3237e-07, 1.0878e-04, 8.8871e-05, ..., 1.4659e-06, 1.6659e-05, 9.3520e-05], ..., [-1.0872e-04, -1.1641e-04, 2.5198e-05, ..., 3.0152e-07, 2.9206e-06, -1.2839e-04], [ 6.0489e-07, 1.4651e-04, 1.7488e-04, ..., 4.4107e-06, -5.6326e-05, 4.3333e-05], [ 5.3138e-05, -3.6907e-04, 2.4006e-05, ..., 7.4785e-07, 5.1260e-06, -3.3832e-04]], device='cuda:0') Epoch 48, bias, value: tensor([ 0.0255, -0.0182, -0.0061, 0.0016, 0.0032, -0.0012, -0.0152, -0.0023, 0.0302, 0.0196], device='cuda:0'), grad: tensor([ 6.7592e-05, -1.3933e-05, 2.4581e-04, 3.3927e-04, 9.8288e-05, -5.4407e-04, 1.1808e-04, -2.2745e-04, 3.1018e-04, -3.9482e-04], device='cuda:0') 100 0.0001 changing lr epoch 47, time 256.65, cls_loss 0.0097 cls_loss_mapping 0.0216 cls_loss_causal 0.7042 re_mapping 0.0155 re_causal 0.0493 /// teacc 98.82 lr 0.00010000 Epoch 49, weight, value: tensor([[-0.0087, -0.0531, -0.0281, ..., 0.0003, -0.0882, 0.0012], [-0.0290, -0.0746, -0.0670, ..., -0.0565, -0.0102, -0.0237], [ 0.0374, 0.0468, -0.0544, ..., -0.0643, -0.0679, -0.0350], ..., [ 0.0029, 0.0174, -0.0093, ..., -0.0293, -0.0759, 0.0406], [-0.0119, 0.0558, 0.0518, ..., -0.0346, 0.0271, -0.0573], [-0.0400, -0.0160, 0.0083, ..., -0.0108, -0.0319, 0.0301]], device='cuda:0'), grad: tensor([[-1.3798e-05, 3.2634e-06, 3.7253e-06, ..., 0.0000e+00, 4.0568e-06, 2.5425e-06], [ 1.8835e-05, 3.1680e-05, 2.4676e-05, ..., 0.0000e+00, 3.1382e-05, 2.1040e-05], [-9.5740e-07, 1.1036e-06, 7.0967e-06, ..., 0.0000e+00, 9.3952e-06, 3.6899e-06], ..., [-2.6584e-05, -1.0416e-05, 7.3425e-06, ..., 0.0000e+00, 1.0148e-05, -2.9504e-05], [ 4.2133e-06, -1.7866e-05, 2.0564e-06, ..., 0.0000e+00, -6.3796e-07, 9.8720e-06], [-4.3735e-06, -4.8399e-05, 2.9266e-05, ..., 0.0000e+00, 3.1829e-05, -3.8564e-05]], device='cuda:0') Epoch 49, bias, value: tensor([ 0.0255, -0.0185, -0.0062, 0.0016, 0.0029, -0.0009, -0.0150, -0.0019, 0.0299, 0.0196], device='cuda:0'), grad: tensor([-2.6718e-05, 7.0333e-05, 3.0085e-05, 4.2224e-04, -3.3408e-05, -5.4407e-04, 4.3958e-05, 5.9754e-06, 4.1693e-05, -1.1235e-05], device='cuda:0') 100 0.0001 changing lr epoch 48, time 256.99, cls_loss 0.0119 cls_loss_mapping 0.0245 cls_loss_causal 0.7021 re_mapping 0.0152 re_causal 0.0464 /// teacc 98.81 lr 0.00010000 Epoch 50, weight, value: tensor([[-0.0086, -0.0534, -0.0285, ..., -0.0005, -0.0887, 0.0009], [-0.0303, -0.0756, -0.0676, ..., -0.0569, -0.0104, -0.0240], [ 0.0379, 0.0469, -0.0551, ..., -0.0648, -0.0686, -0.0351], ..., [ 0.0029, 0.0172, -0.0096, ..., -0.0293, -0.0771, 0.0410], [-0.0115, 0.0564, 0.0522, ..., -0.0346, 0.0275, -0.0580], [-0.0400, -0.0158, 0.0081, ..., -0.0093, -0.0320, 0.0304]], device='cuda:0'), grad: tensor([[ 1.8954e-05, 4.3571e-05, 5.6028e-05, ..., 3.0309e-05, 2.8744e-05, 2.5511e-05], [ 2.0850e-04, 1.3244e-04, 1.8671e-05, ..., 2.2482e-06, 1.1221e-05, 4.6182e-04], [-2.7847e-04, -1.5426e-04, -7.3731e-05, ..., 5.2340e-06, -1.8999e-05, 1.0952e-05], ..., [-1.0328e-03, -6.6185e-04, 6.1952e-06, ..., 8.9547e-07, -4.3437e-06, -2.7447e-03], [ 1.9670e-04, 1.1468e-04, 8.1241e-05, ..., 9.3803e-06, 2.7940e-05, 2.7514e-04], [ 7.0238e-04, 3.8576e-04, 2.4900e-05, ..., 2.4643e-06, 1.0148e-05, 1.6832e-03]], device='cuda:0') Epoch 50, bias, value: tensor([ 0.0249, -0.0190, -0.0060, 0.0017, 0.0036, -0.0010, -0.0152, -0.0018, 0.0298, 0.0197], device='cuda:0'), grad: tensor([ 1.0157e-04, 8.9121e-04, -4.4584e-04, 2.4486e-04, 6.1214e-05, 9.5665e-05, -1.4670e-05, -4.7073e-03, 7.1716e-04, 3.0556e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 49---------------------------------------------------- epoch 49, time 273.89, cls_loss 0.0141 cls_loss_mapping 0.0260 cls_loss_causal 0.6992 re_mapping 0.0151 re_causal 0.0427 /// teacc 98.85 lr 0.00010000 Epoch 51, weight, value: tensor([[-0.0082, -0.0544, -0.0297, ..., -0.0014, -0.0907, 0.0010], [-0.0314, -0.0765, -0.0682, ..., -0.0574, -0.0109, -0.0244], [ 0.0385, 0.0468, -0.0558, ..., -0.0650, -0.0694, -0.0357], ..., [ 0.0032, 0.0178, -0.0099, ..., -0.0293, -0.0774, 0.0419], [-0.0107, 0.0572, 0.0523, ..., -0.0357, 0.0280, -0.0588], [-0.0405, -0.0162, 0.0077, ..., -0.0098, -0.0325, 0.0304]], device='cuda:0'), grad: tensor([[ 8.0392e-06, 5.5671e-05, 4.0799e-05, ..., 7.5065e-06, 2.8789e-05, 1.3210e-05], [ 2.5535e-04, 1.0996e-03, 8.5974e-04, ..., 3.6985e-05, 6.6376e-04, -1.7607e-04], [ 1.3292e-04, 1.6165e-04, 1.6522e-04, ..., 1.3299e-05, 1.4603e-04, 2.0945e-04], ..., [-4.4727e-04, -3.4690e-04, 7.5161e-05, ..., 6.4559e-06, 3.3259e-05, -5.6219e-04], [-3.4857e-04, -1.8396e-03, -1.5440e-03, ..., -1.0914e-04, -1.1425e-03, 2.5749e-05], [ 1.1563e-04, 2.7728e-04, 9.8884e-05, ..., 3.8296e-06, 7.1943e-05, 2.4283e-04]], device='cuda:0') Epoch 51, bias, value: tensor([ 0.0248, -0.0198, -0.0057, 0.0014, 0.0033, -0.0002, -0.0154, -0.0009, 0.0296, 0.0195], device='cuda:0'), grad: tensor([ 7.3731e-05, 2.0199e-03, 7.0620e-04, 4.5347e-04, 1.9097e-04, 1.8513e-04, 4.8923e-04, -9.7656e-04, -3.8376e-03, 6.9189e-04], device='cuda:0') 100 0.0001 changing lr epoch 50, time 256.43, cls_loss 0.0106 cls_loss_mapping 0.0214 cls_loss_causal 0.6583 re_mapping 0.0148 re_causal 0.0445 /// teacc 98.76 lr 0.00010000 Epoch 52, weight, value: tensor([[-0.0083, -0.0547, -0.0303, ..., -0.0014, -0.0914, 0.0009], [-0.0325, -0.0774, -0.0688, ..., -0.0576, -0.0115, -0.0239], [ 0.0396, 0.0476, -0.0563, ..., -0.0652, -0.0699, -0.0357], ..., [ 0.0031, 0.0175, -0.0103, ..., -0.0295, -0.0777, 0.0416], [-0.0107, 0.0576, 0.0527, ..., -0.0358, 0.0284, -0.0594], [-0.0403, -0.0155, 0.0074, ..., -0.0099, -0.0328, 0.0309]], device='cuda:0'), grad: tensor([[ 1.0028e-05, 2.6822e-05, 1.8314e-05, ..., 1.4879e-05, 1.6928e-05, 1.5274e-05], [ 1.1563e-05, 1.6734e-05, 5.5581e-06, ..., 1.7397e-06, 7.2457e-06, -2.5332e-05], [ 5.0020e-04, 5.7507e-04, 1.0826e-05, ..., 2.4009e-06, 1.5423e-05, 2.5988e-04], ..., [-5.8603e-04, -6.4993e-04, 3.2574e-05, ..., 1.8277e-07, 3.3528e-05, -2.7657e-04], [-1.8311e-04, -3.5596e-04, -2.3615e-04, ..., 6.9439e-06, -3.0899e-04, -5.0664e-05], [ 1.2207e-04, 2.1958e-04, 2.0635e-04, ..., 1.1735e-06, 2.4438e-04, -2.8491e-05]], device='cuda:0') Epoch 52, bias, value: tensor([ 2.4986e-02, -1.9963e-02, -5.0843e-03, 1.1135e-03, 2.9460e-03, -5.9599e-05, -1.5603e-02, -8.7716e-04, 2.9459e-02, 1.9553e-02], device='cuda:0'), grad: tensor([ 4.4823e-05, -5.0008e-05, 6.0368e-04, 4.2343e-04, 2.8685e-05, -2.7871e-04, -2.4393e-05, -5.8556e-04, -2.8205e-04, 1.1867e-04], device='cuda:0') 100 0.0001 changing lr epoch 51, time 256.54, cls_loss 0.0110 cls_loss_mapping 0.0227 cls_loss_causal 0.6552 re_mapping 0.0143 re_causal 0.0419 /// teacc 98.58 lr 0.00010000 Epoch 53, weight, value: tensor([[-0.0076, -0.0550, -0.0312, ..., -0.0019, -0.0925, 0.0008], [-0.0331, -0.0779, -0.0695, ..., -0.0580, -0.0122, -0.0238], [ 0.0402, 0.0480, -0.0569, ..., -0.0656, -0.0701, -0.0360], ..., [ 0.0027, 0.0173, -0.0107, ..., -0.0295, -0.0787, 0.0417], [-0.0097, 0.0587, 0.0532, ..., -0.0367, 0.0291, -0.0597], [-0.0405, -0.0156, 0.0069, ..., -0.0101, -0.0332, 0.0311]], device='cuda:0'), grad: tensor([[-8.5682e-08, 3.6657e-05, 6.1929e-05, ..., 9.8571e-06, 3.7283e-05, 1.6969e-06], [ 9.6932e-06, 3.4630e-05, 5.5969e-05, ..., 1.2003e-05, 3.0905e-05, -3.0184e-04], [ 3.8564e-05, 1.0413e-04, 1.5998e-04, ..., 2.4408e-05, 9.6262e-05, 4.4078e-05], ..., [-1.8239e-04, -2.2411e-04, 2.9728e-05, ..., 3.5241e-06, 5.0068e-06, -1.2219e-04], [ 3.3593e-04, -9.4986e-04, -2.3289e-03, ..., -9.5320e-04, -9.9850e-04, 1.2770e-05], [ 4.9204e-05, 8.2552e-05, 6.8784e-05, ..., 9.1270e-06, 3.9369e-05, 4.9561e-05]], device='cuda:0') Epoch 53, bias, value: tensor([ 2.4774e-02, -1.9746e-02, -5.0063e-03, 1.0615e-03, 3.1965e-03, -6.8045e-05, -1.5259e-02, -1.4891e-03, 2.9790e-02, 1.9421e-02], device='cuda:0'), grad: tensor([ 4.3690e-05, -8.3351e-04, 2.7442e-04, 3.4595e-04, 4.8113e-04, -1.6823e-03, 2.7122e-03, -1.7488e-04, -1.3523e-03, 1.9014e-04], device='cuda:0') 100 0.0001 changing lr epoch 52, time 256.48, cls_loss 0.0099 cls_loss_mapping 0.0242 cls_loss_causal 0.6785 re_mapping 0.0151 re_causal 0.0448 /// teacc 98.82 lr 0.00010000 Epoch 54, weight, value: tensor([[-0.0065, -0.0554, -0.0315, ..., -0.0036, -0.0928, 0.0015], [-0.0338, -0.0795, -0.0703, ..., -0.0578, -0.0129, -0.0237], [ 0.0403, 0.0479, -0.0574, ..., -0.0660, -0.0708, -0.0364], ..., [ 0.0028, 0.0178, -0.0110, ..., -0.0298, -0.0795, 0.0419], [-0.0095, 0.0594, 0.0538, ..., -0.0368, 0.0300, -0.0599], [-0.0409, -0.0158, 0.0065, ..., -0.0102, -0.0337, 0.0310]], device='cuda:0'), grad: tensor([[-6.2466e-05, 8.2612e-05, 3.8356e-05, ..., 9.6142e-05, 3.5435e-05, -1.6363e-06], [ 8.9705e-06, -2.7633e-04, 2.9117e-05, ..., 1.0389e-04, 2.9534e-05, -2.2483e-04], [-7.6532e-05, 7.1406e-05, 3.2693e-05, ..., 6.3300e-05, 3.2514e-05, 2.2441e-05], ..., [ 3.3557e-05, 1.3602e-04, 1.1429e-05, ..., 1.6406e-05, 1.1422e-05, 1.4913e-04], [-2.2903e-05, 1.3232e-04, -1.6525e-05, ..., 8.2672e-05, -1.8671e-05, 8.4519e-05], [ 1.4566e-05, -1.1331e-04, 3.2634e-05, ..., 3.5286e-05, 2.5839e-05, -2.1899e-04]], device='cuda:0') Epoch 54, bias, value: tensor([ 0.0254, -0.0202, -0.0054, 0.0010, 0.0031, 0.0003, -0.0152, -0.0012, 0.0303, 0.0189], device='cuda:0'), grad: tensor([ 0.0002, -0.0014, 0.0002, 0.0004, 0.0015, 0.0007, -0.0027, 0.0005, 0.0008, -0.0002], device='cuda:0') 100 0.0001 changing lr epoch 53, time 256.63, cls_loss 0.0097 cls_loss_mapping 0.0218 cls_loss_causal 0.6871 re_mapping 0.0138 re_causal 0.0428 /// teacc 98.57 lr 0.00010000 Epoch 55, weight, value: tensor([[-0.0062, -0.0556, -0.0319, ..., -0.0041, -0.0933, 0.0016], [-0.0343, -0.0795, -0.0705, ..., -0.0584, -0.0132, -0.0221], [ 0.0406, 0.0479, -0.0579, ..., -0.0661, -0.0715, -0.0370], ..., [ 0.0033, 0.0181, -0.0113, ..., -0.0297, -0.0798, 0.0417], [-0.0093, 0.0597, 0.0541, ..., -0.0374, 0.0303, -0.0607], [-0.0411, -0.0157, 0.0061, ..., -0.0100, -0.0341, 0.0316]], device='cuda:0'), grad: tensor([[-2.1920e-05, -4.6613e-07, -1.9129e-06, ..., -3.6191e-06, 8.2254e-06, -6.3628e-06], [ 3.3021e-05, 2.6911e-05, 1.0483e-05, ..., 1.4789e-06, 1.3120e-05, -4.0196e-06], [-1.4627e-04, -1.3947e-04, 7.1824e-06, ..., 1.9912e-06, 9.2685e-06, -2.7511e-06], ..., [ 6.3255e-06, 1.3866e-05, 7.0892e-06, ..., 4.2561e-07, 1.9697e-07, -1.9163e-05], [ 8.0705e-05, 8.4162e-05, 3.7313e-05, ..., 7.6517e-06, 2.9400e-05, 1.3083e-05], [ 5.4352e-06, 2.8871e-06, 2.2084e-05, ..., 1.0598e-06, 2.2009e-05, 1.4221e-06]], device='cuda:0') Epoch 55, bias, value: tensor([ 0.0257, -0.0192, -0.0056, 0.0016, 0.0026, -0.0003, -0.0152, -0.0015, 0.0300, 0.0190], device='cuda:0'), grad: tensor([-8.0884e-05, 5.7131e-05, -1.8907e-04, 1.3804e-04, 1.9237e-05, -2.2602e-04, 3.6478e-05, 8.7023e-06, 1.9240e-04, 4.3482e-05], device='cuda:0') 100 0.0001 changing lr epoch 54, time 256.27, cls_loss 0.0101 cls_loss_mapping 0.0190 cls_loss_causal 0.6973 re_mapping 0.0147 re_causal 0.0428 /// teacc 98.75 lr 0.00010000 Epoch 56, weight, value: tensor([[-0.0057, -0.0559, -0.0323, ..., -0.0044, -0.0937, 0.0016], [-0.0362, -0.0805, -0.0708, ..., -0.0582, -0.0134, -0.0224], [ 0.0415, 0.0480, -0.0584, ..., -0.0663, -0.0722, -0.0368], ..., [ 0.0036, 0.0187, -0.0119, ..., -0.0301, -0.0804, 0.0423], [-0.0092, 0.0601, 0.0542, ..., -0.0379, 0.0305, -0.0615], [-0.0413, -0.0159, 0.0057, ..., -0.0103, -0.0345, 0.0319]], device='cuda:0'), grad: tensor([[-4.1239e-06, 4.3660e-06, 6.6943e-06, ..., 4.4890e-06, 4.7050e-06, 4.2059e-06], [ 9.5665e-06, 1.1556e-05, 5.8338e-06, ..., 8.9081e-07, 8.4862e-06, -2.4050e-05], [ 1.0058e-05, 9.9987e-06, 9.0748e-06, ..., 1.1185e-06, 1.0848e-05, 1.1101e-05], ..., [-5.5879e-05, -2.5779e-05, 3.1758e-06, ..., 8.0792e-08, 6.6496e-06, -3.6031e-05], [ 6.3935e-07, -9.0823e-06, -2.8089e-06, ..., 3.5092e-06, 1.9581e-07, 1.0215e-05], [ 1.6494e-06, -3.6992e-06, 5.6140e-06, ..., 4.5146e-07, 6.4746e-06, 4.1560e-07]], device='cuda:0') Epoch 56, bias, value: tensor([ 2.6025e-02, -1.9512e-02, -5.2482e-03, 1.1847e-03, 2.9592e-03, 5.2053e-05, -1.5376e-02, -1.5131e-03, 2.9576e-02, 1.8909e-02], device='cuda:0'), grad: tensor([ 6.3963e-06, -4.5359e-05, 5.3108e-05, -8.3670e-06, -1.3128e-05, 3.2276e-05, -1.3664e-05, -6.8188e-05, 4.5329e-05, 1.1414e-05], device='cuda:0') 100 0.0001 changing lr epoch 55, time 256.56, cls_loss 0.0080 cls_loss_mapping 0.0194 cls_loss_causal 0.6532 re_mapping 0.0137 re_causal 0.0410 /// teacc 98.75 lr 0.00010000 Epoch 57, weight, value: tensor([[-0.0053, -0.0566, -0.0332, ..., -0.0050, -0.0945, 0.0015], [-0.0367, -0.0806, -0.0716, ..., -0.0593, -0.0137, -0.0221], [ 0.0418, 0.0482, -0.0590, ..., -0.0669, -0.0730, -0.0372], ..., [ 0.0038, 0.0189, -0.0122, ..., -0.0301, -0.0808, 0.0427], [-0.0090, 0.0605, 0.0544, ..., -0.0383, 0.0307, -0.0623], [-0.0414, -0.0159, 0.0054, ..., -0.0106, -0.0348, 0.0319]], device='cuda:0'), grad: tensor([[-3.6687e-05, 2.8327e-05, 6.9439e-05, ..., 8.3819e-06, 8.4221e-05, 3.3051e-05], [ 4.7177e-05, 6.3717e-05, 3.5584e-05, ..., 5.2564e-06, 4.6641e-05, -3.5256e-05], [-1.0079e-04, -1.1605e-04, 6.1154e-05, ..., 9.8869e-06, 5.4270e-05, 3.4094e-05], ..., [ 2.5681e-07, 7.9200e-06, 2.3872e-05, ..., 1.0813e-06, 4.1336e-05, 3.3170e-05], [-7.2658e-05, -9.3341e-05, -3.3069e-04, ..., 1.4879e-05, -3.2592e-04, 2.5108e-05], [ 3.5733e-05, -9.9838e-06, 1.3411e-04, ..., 5.1595e-06, 4.7922e-05, -3.4523e-04]], device='cuda:0') Epoch 57, bias, value: tensor([ 0.0261, -0.0193, -0.0053, 0.0011, 0.0034, 0.0001, -0.0154, -0.0016, 0.0293, 0.0187], device='cuda:0'), grad: tensor([ 9.3222e-05, 9.5248e-05, -6.8426e-05, 4.0894e-03, 7.0953e-04, -3.4027e-03, -4.0531e-04, 2.0909e-04, -1.4853e-04, -1.1683e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 56---------------------------------------------------- epoch 56, time 272.32, cls_loss 0.0099 cls_loss_mapping 0.0215 cls_loss_causal 0.6489 re_mapping 0.0136 re_causal 0.0397 /// teacc 98.87 lr 0.00010000 Epoch 58, weight, value: tensor([[-0.0047, -0.0571, -0.0346, ..., -0.0052, -0.0960, 0.0012], [-0.0378, -0.0816, -0.0732, ..., -0.0597, -0.0148, -0.0221], [ 0.0421, 0.0484, -0.0596, ..., -0.0672, -0.0737, -0.0378], ..., [ 0.0042, 0.0192, -0.0126, ..., -0.0301, -0.0811, 0.0430], [-0.0090, 0.0610, 0.0551, ..., -0.0386, 0.0315, -0.0632], [-0.0414, -0.0157, 0.0048, ..., -0.0107, -0.0353, 0.0323]], device='cuda:0'), grad: tensor([[ 2.0996e-05, 1.5855e-05, 2.0802e-05, ..., 5.8077e-06, 1.9774e-05, 1.2510e-05], [ 1.3843e-05, 8.2254e-06, 6.7763e-06, ..., 1.4007e-06, 9.0972e-06, -6.4313e-05], [ 4.0412e-05, 3.1769e-05, 1.5095e-05, ..., 1.1250e-06, 2.6554e-05, 4.6581e-05], ..., [-5.6934e-04, -4.5687e-05, 1.9491e-05, ..., 3.7402e-06, 2.3618e-05, -5.8413e-04], [ 1.1575e-04, 4.2111e-05, 1.4508e-04, ..., 4.2439e-05, 1.4496e-04, 4.2945e-05], [-8.4519e-05, -2.3031e-04, 2.5883e-05, ..., 1.5534e-06, -6.5491e-06, -5.8860e-05]], device='cuda:0') Epoch 58, bias, value: tensor([ 0.0261, -0.0199, -0.0057, 0.0014, 0.0035, 0.0008, -0.0158, -0.0011, 0.0292, 0.0186], device='cuda:0'), grad: tensor([ 5.3555e-05, -1.2040e-04, 1.2493e-04, 9.1374e-05, 9.8419e-04, -1.3363e-04, 1.4663e-04, -9.8515e-04, 2.8419e-04, -4.4441e-04], device='cuda:0') 100 0.0001 changing lr epoch 57, time 257.23, cls_loss 0.0075 cls_loss_mapping 0.0161 cls_loss_causal 0.6540 re_mapping 0.0135 re_causal 0.0404 /// teacc 98.85 lr 0.00010000 Epoch 59, weight, value: tensor([[-0.0043, -0.0574, -0.0350, ..., -0.0056, -0.0964, 0.0010], [-0.0386, -0.0820, -0.0736, ..., -0.0602, -0.0151, -0.0220], [ 0.0425, 0.0487, -0.0600, ..., -0.0674, -0.0742, -0.0382], ..., [ 0.0040, 0.0192, -0.0132, ..., -0.0303, -0.0815, 0.0433], [-0.0089, 0.0611, 0.0552, ..., -0.0393, 0.0316, -0.0638], [-0.0412, -0.0152, 0.0045, ..., -0.0109, -0.0355, 0.0326]], device='cuda:0'), grad: tensor([[-4.0866e-06, 1.8269e-05, 1.7509e-05, ..., 7.7561e-06, 8.9705e-06, -6.4410e-06], [ 5.1886e-05, 6.9439e-05, 2.7969e-05, ..., 1.0476e-05, 1.3188e-05, 5.6207e-05], [ 5.7489e-05, 7.0870e-05, 3.5673e-05, ..., 1.1899e-05, 1.6958e-05, 6.2287e-05], ..., [-3.2258e-04, -4.3583e-04, 4.8019e-06, ..., 1.1846e-06, 2.5313e-06, -5.1641e-04], [-5.5695e-04, -6.6614e-04, -9.9468e-04, ..., -3.0732e-04, -4.1223e-04, 3.7432e-05], [ 1.9050e-04, 2.0289e-04, 8.7738e-05, ..., 9.8124e-06, 1.6958e-05, 2.3067e-04]], device='cuda:0') Epoch 59, bias, value: tensor([ 0.0260, -0.0197, -0.0057, 0.0018, 0.0029, 0.0007, -0.0155, -0.0014, 0.0289, 0.0189], device='cuda:0'), grad: tensor([-1.1377e-05, 1.4377e-04, 1.6296e-04, 1.6797e-04, -3.3259e-04, 2.3234e-04, 9.4175e-04, -9.5749e-04, -1.0929e-03, 7.4482e-04], device='cuda:0') 100 0.0001 changing lr epoch 58, time 256.63, cls_loss 0.0075 cls_loss_mapping 0.0174 cls_loss_causal 0.6943 re_mapping 0.0128 re_causal 0.0410 /// teacc 98.68 lr 0.00010000 Epoch 60, weight, value: tensor([[-0.0033, -0.0574, -0.0356, ..., -0.0059, -0.0973, 0.0013], [-0.0390, -0.0822, -0.0740, ..., -0.0599, -0.0155, -0.0215], [ 0.0427, 0.0484, -0.0607, ..., -0.0675, -0.0750, -0.0385], ..., [ 0.0044, 0.0196, -0.0135, ..., -0.0309, -0.0820, 0.0435], [-0.0084, 0.0619, 0.0557, ..., -0.0397, 0.0323, -0.0643], [-0.0413, -0.0151, 0.0041, ..., -0.0112, -0.0359, 0.0326]], device='cuda:0'), grad: tensor([[-7.5758e-05, 1.5259e-05, 1.0878e-05, ..., 4.6156e-06, 3.9414e-06, -1.5825e-05], [ 1.4119e-05, 1.7986e-05, -2.9624e-05, ..., 1.5777e-06, 8.2627e-06, -1.0288e-04], [-1.2779e-04, -4.1306e-05, 1.6987e-05, ..., -2.5094e-05, -3.5524e-05, 1.6376e-05], ..., [-1.2165e-04, -2.0647e-04, 3.8128e-06, ..., 1.2880e-06, 4.6603e-06, -1.7536e-04], [-3.0939e-06, -2.5183e-05, -7.7710e-06, ..., 9.4697e-06, -3.1739e-05, 5.2720e-05], [ 9.8467e-05, 1.5581e-04, 1.2472e-05, ..., 2.3041e-06, 7.9051e-06, 1.4436e-04]], device='cuda:0') Epoch 60, bias, value: tensor([ 0.0263, -0.0194, -0.0060, 0.0016, 0.0022, 0.0005, -0.0152, -0.0016, 0.0292, 0.0192], device='cuda:0'), grad: tensor([-9.4235e-05, -2.4128e-04, -2.2650e-04, 3.3760e-04, -1.3268e-04, 9.8169e-05, 6.7532e-05, -2.4462e-04, 1.0645e-04, 3.2949e-04], device='cuda:0') 100 0.0001 changing lr epoch 59, time 256.55, cls_loss 0.0119 cls_loss_mapping 0.0243 cls_loss_causal 0.6653 re_mapping 0.0133 re_causal 0.0384 /// teacc 98.83 lr 0.00010000 Epoch 61, weight, value: tensor([[-0.0025, -0.0581, -0.0366, ..., -0.0061, -0.0985, 0.0009], [-0.0402, -0.0831, -0.0746, ..., -0.0614, -0.0161, -0.0213], [ 0.0433, 0.0485, -0.0611, ..., -0.0679, -0.0753, -0.0390], ..., [ 0.0048, 0.0202, -0.0138, ..., -0.0311, -0.0824, 0.0438], [-0.0084, 0.0626, 0.0563, ..., -0.0403, 0.0327, -0.0653], [-0.0416, -0.0149, 0.0036, ..., -0.0117, -0.0364, 0.0323]], device='cuda:0'), grad: tensor([[-4.7255e-04, -1.9872e-04, -2.6539e-05, ..., 1.3527e-07, 5.0059e-07, -8.3625e-05], [ 1.5914e-05, 4.2506e-06, 5.8413e-06, ..., 3.1199e-08, 6.6347e-06, -3.2097e-05], [-1.3459e-04, -3.9315e-04, 6.2026e-06, ..., 4.9360e-08, -4.7177e-05, -8.6784e-05], ..., [ 8.6963e-05, 1.3745e-04, 2.9169e-06, ..., 3.8883e-08, 2.1473e-05, 4.9531e-05], [ 3.6359e-04, 2.5964e-04, 5.4955e-05, ..., 7.1339e-06, 4.6432e-05, 9.1076e-05], [ 1.2159e-05, -1.3433e-05, 9.2462e-06, ..., 2.0419e-07, 8.3521e-06, -1.6347e-05]], device='cuda:0') Epoch 61, bias, value: tensor([ 0.0256, -0.0197, -0.0064, 0.0016, 0.0039, 0.0004, -0.0149, -0.0014, 0.0291, 0.0187], device='cuda:0'), grad: tensor([-7.3814e-04, -5.1796e-05, -3.6168e-04, 2.4116e-04, -1.7130e-04, -2.6122e-05, 1.1581e-04, 2.5010e-04, 6.6566e-04, 7.7009e-05], device='cuda:0') 100 0.0001 changing lr epoch 60, time 256.64, cls_loss 0.0105 cls_loss_mapping 0.0198 cls_loss_causal 0.6604 re_mapping 0.0135 re_causal 0.0376 /// teacc 98.67 lr 0.00010000 Epoch 62, weight, value: tensor([[-0.0020, -0.0589, -0.0371, ..., -0.0070, -0.0995, 0.0003], [-0.0412, -0.0838, -0.0751, ..., -0.0620, -0.0167, -0.0210], [ 0.0437, 0.0489, -0.0618, ..., -0.0684, -0.0758, -0.0395], ..., [ 0.0050, 0.0201, -0.0142, ..., -0.0312, -0.0836, 0.0437], [-0.0082, 0.0630, 0.0564, ..., -0.0404, 0.0327, -0.0657], [-0.0418, -0.0146, 0.0030, ..., -0.0120, -0.0370, 0.0328]], device='cuda:0'), grad: tensor([[ 1.8597e-05, 2.8044e-05, 1.5989e-05, ..., 1.3925e-05, 9.0748e-06, 2.0713e-05], [ 3.1519e-04, 2.5225e-04, 5.1111e-06, ..., 2.6617e-06, 3.2961e-05, 3.4285e-04], [ 9.5558e-04, 1.7715e-04, 8.6501e-06, ..., 1.3057e-06, 3.3379e-05, 5.5170e-04], ..., [-2.3117e-03, -1.0376e-03, 7.4059e-06, ..., 2.7730e-07, -5.0694e-05, -1.8702e-03], [ 2.5988e-04, 1.9896e-04, 2.8070e-06, ..., 6.0759e-06, 3.0383e-05, 2.9850e-04], [ 2.2262e-05, 7.5325e-06, 1.2174e-05, ..., 8.8429e-07, 3.0473e-05, 3.4064e-05]], device='cuda:0') Epoch 62, bias, value: tensor([ 0.0256, -0.0197, -0.0066, 0.0023, 0.0036, 0.0005, -0.0154, -0.0013, 0.0289, 0.0189], device='cuda:0'), grad: tensor([ 7.1347e-05, 5.7030e-04, 1.3256e-03, -2.6989e-04, 6.0987e-04, 7.8773e-04, -3.0375e-04, -3.3932e-03, 5.3120e-04, 6.8486e-05], device='cuda:0') 100 0.0001 changing lr epoch 61, time 256.87, cls_loss 0.0075 cls_loss_mapping 0.0179 cls_loss_causal 0.6402 re_mapping 0.0133 re_causal 0.0395 /// teacc 98.83 lr 0.00010000 Epoch 63, weight, value: tensor([[-0.0015, -0.0593, -0.0376, ..., -0.0071, -0.1002, 0.0002], [-0.0419, -0.0842, -0.0755, ..., -0.0623, -0.0170, -0.0208], [ 0.0436, 0.0486, -0.0628, ..., -0.0686, -0.0765, -0.0403], ..., [ 0.0059, 0.0209, -0.0145, ..., -0.0312, -0.0840, 0.0442], [-0.0077, 0.0637, 0.0572, ..., -0.0406, 0.0336, -0.0664], [-0.0418, -0.0148, 0.0028, ..., -0.0119, -0.0373, 0.0332]], device='cuda:0'), grad: tensor([[-9.2626e-05, -1.3493e-05, 1.6704e-05, ..., 1.5367e-08, 1.1615e-05, -1.6997e-06], [ 1.8314e-05, 2.9549e-05, 7.3910e-06, ..., 2.7940e-09, 7.7859e-06, -3.8520e-06], [-1.3441e-05, -6.7912e-06, 7.4394e-06, ..., 1.8626e-09, 9.2983e-06, 1.1206e-05], ..., [-7.1228e-05, -1.7297e-04, 1.1593e-05, ..., 1.2340e-08, 8.9630e-06, -9.3758e-05], [-1.2845e-05, -4.3303e-05, -6.6519e-05, ..., 9.4064e-08, -5.2661e-05, 5.9903e-06], [ 1.1164e-04, 1.6224e-04, 1.0425e-04, ..., 7.7300e-08, 7.6830e-05, 6.1631e-05]], device='cuda:0') Epoch 63, bias, value: tensor([ 2.5724e-02, -1.9717e-02, -7.1135e-03, 2.3144e-03, 4.1779e-03, 7.1365e-05, -1.5483e-02, -6.1219e-04, 2.9038e-02, 1.8492e-02], device='cuda:0'), grad: tensor([-3.6120e-04, 1.3590e-05, 3.4958e-05, 3.6329e-05, 2.3365e-05, -1.4639e-04, 1.8370e-04, -1.6606e-04, 3.0324e-06, 3.7885e-04], device='cuda:0') 100 0.0001 changing lr epoch 62, time 256.34, cls_loss 0.0081 cls_loss_mapping 0.0177 cls_loss_causal 0.6572 re_mapping 0.0129 re_causal 0.0375 /// teacc 98.83 lr 0.00010000 Epoch 64, weight, value: tensor([[-1.2839e-03, -5.9813e-02, -3.8527e-02, ..., -7.4525e-03, -1.0122e-01, -4.0430e-05], [-4.2566e-02, -8.5597e-02, -7.6964e-02, ..., -6.4692e-02, -1.8765e-02, -2.0637e-02], [ 4.3938e-02, 4.8839e-02, -6.3455e-02, ..., -6.8890e-02, -7.6510e-02, -4.0566e-02], ..., [ 6.4619e-03, 2.1703e-02, -1.4420e-02, ..., -3.1293e-02, -8.3626e-02, 4.4538e-02], [-7.8719e-03, 6.4043e-02, 5.7459e-02, ..., -4.0561e-02, 3.3759e-02, -6.7257e-02], [-4.2224e-02, -1.4898e-02, 2.3034e-03, ..., -1.2204e-02, -3.7643e-02, 3.3122e-02]], device='cuda:0'), grad: tensor([[-6.7055e-06, 6.2548e-06, 9.1493e-06, ..., 1.3206e-06, 8.6129e-06, 5.8077e-06], [ 1.0341e-05, 1.2919e-05, 7.0855e-06, ..., -4.9360e-07, 1.6138e-05, -4.0472e-05], [ 1.9118e-05, 2.0444e-05, 1.1280e-05, ..., 7.3947e-07, 4.4405e-05, 6.6876e-05], ..., [ 1.8284e-05, 1.0476e-05, 5.0813e-06, ..., 1.6298e-07, 1.8612e-05, 2.9549e-05], [ 3.1646e-06, -1.4147e-06, 4.5896e-05, ..., 7.4729e-06, 3.3557e-05, 2.8133e-05], [ 7.2047e-06, 1.6708e-06, 1.4082e-05, ..., 1.9511e-07, 1.3314e-05, -1.1194e-06]], device='cuda:0') Epoch 64, bias, value: tensor([ 2.5741e-02, -2.0571e-02, -6.7446e-03, 1.7965e-03, 4.4808e-03, 7.8124e-04, -1.5096e-02, -3.4100e-05, 2.8628e-02, 1.7988e-02], device='cuda:0'), grad: tensor([ 5.0999e-06, -3.8147e-05, 1.3852e-04, -2.4605e-04, 2.1905e-05, -1.4853e-04, 8.8751e-05, 7.4267e-05, 7.9036e-05, 2.5004e-05], device='cuda:0') 100 0.0001 changing lr epoch 63, time 255.97, cls_loss 0.0073 cls_loss_mapping 0.0165 cls_loss_causal 0.6767 re_mapping 0.0126 re_causal 0.0384 /// teacc 98.69 lr 0.00010000 Epoch 65, weight, value: tensor([[-0.0009, -0.0603, -0.0389, ..., -0.0079, -0.1018, -0.0001], [-0.0431, -0.0852, -0.0774, ..., -0.0651, -0.0188, -0.0193], [ 0.0445, 0.0494, -0.0641, ..., -0.0692, -0.0769, -0.0410], ..., [ 0.0058, 0.0214, -0.0163, ..., -0.0314, -0.0843, 0.0438], [-0.0076, 0.0649, 0.0584, ..., -0.0404, 0.0344, -0.0680], [-0.0422, -0.0144, 0.0020, ..., -0.0123, -0.0379, 0.0337]], device='cuda:0'), grad: tensor([[-2.0906e-05, -4.0770e-04, -4.5323e-04, ..., 2.8908e-05, 2.9922e-05, -6.9427e-04], [ 5.1916e-05, 7.2598e-05, 1.4171e-05, ..., 2.0694e-06, 7.3947e-06, -8.3745e-05], [ 8.6650e-06, 1.5512e-05, 1.0177e-05, ..., 8.3540e-07, 1.6272e-05, 4.1366e-05], ..., [-7.0274e-05, -9.0241e-05, 7.7188e-06, ..., 2.7288e-07, 1.4760e-05, -1.2565e-04], [ 2.7746e-05, 9.0003e-05, 1.3483e-04, ..., 4.6045e-05, 9.2268e-05, 8.0764e-05], [ 3.3118e-06, 3.3993e-06, 2.6301e-05, ..., 3.0231e-06, 9.0972e-06, 2.4036e-05]], device='cuda:0') Epoch 65, bias, value: tensor([ 0.0258, -0.0199, -0.0064, 0.0017, 0.0040, 0.0016, -0.0159, -0.0013, 0.0288, 0.0185], device='cuda:0'), grad: tensor([-2.8629e-03, -2.4819e-04, 1.1051e-04, -6.5088e-05, 4.2701e-04, 2.5635e-03, -6.2108e-05, -2.2542e-04, 2.5368e-04, 1.0800e-04], device='cuda:0') 100 0.0001 changing lr epoch 64, time 256.28, cls_loss 0.0082 cls_loss_mapping 0.0175 cls_loss_causal 0.6467 re_mapping 0.0116 re_causal 0.0363 /// teacc 98.81 lr 0.00010000 Epoch 66, weight, value: tensor([[-0.0007, -0.0610, -0.0394, ..., -0.0085, -0.1027, -0.0002], [-0.0435, -0.0846, -0.0780, ..., -0.0653, -0.0195, -0.0190], [ 0.0446, 0.0491, -0.0648, ..., -0.0694, -0.0772, -0.0409], ..., [ 0.0057, 0.0212, -0.0170, ..., -0.0314, -0.0852, 0.0437], [-0.0069, 0.0657, 0.0591, ..., -0.0409, 0.0351, -0.0693], [-0.0418, -0.0139, 0.0020, ..., -0.0125, -0.0379, 0.0344]], device='cuda:0'), grad: tensor([[-2.6189e-06, 5.5842e-06, 2.4159e-06, ..., 2.0806e-06, 1.8552e-06, -8.0094e-06], [ 1.2003e-05, 1.6242e-05, 9.6671e-07, ..., 1.6950e-07, 1.6708e-06, -1.3029e-06], [-6.3837e-05, -1.9446e-05, 7.3528e-07, ..., 9.6625e-08, 2.2929e-06, 8.6650e-06], ..., [ 1.8282e-06, -2.2113e-05, 2.1467e-07, ..., 1.3271e-08, 1.2062e-05, -3.2187e-05], [ 6.0052e-06, 2.3115e-06, -7.0687e-07, ..., 9.7509e-07, 1.1614e-06, 7.3835e-06], [ 1.7628e-05, 3.7521e-05, 7.5856e-07, ..., 2.1746e-07, 1.1818e-06, 3.8218e-04]], device='cuda:0') Epoch 66, bias, value: tensor([ 0.0259, -0.0201, -0.0060, 0.0016, 0.0039, 0.0014, -0.0165, -0.0016, 0.0289, 0.0191], device='cuda:0'), grad: tensor([ 3.6389e-05, 1.6540e-05, -1.8668e-04, 5.8562e-05, -2.2316e-03, 1.5795e-05, 1.5661e-05, 4.3154e-05, 3.8087e-05, 2.1915e-03], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 65---------------------------------------------------- epoch 65, time 273.92, cls_loss 0.0075 cls_loss_mapping 0.0167 cls_loss_causal 0.6222 re_mapping 0.0125 re_causal 0.0367 /// teacc 98.88 lr 0.00010000 Epoch 67, weight, value: tensor([[-2.0603e-05, -6.1639e-02, -4.0704e-02, ..., -8.9565e-03, -1.0458e-01, -2.6783e-04], [-4.4231e-02, -8.5306e-02, -7.8906e-02, ..., -6.6170e-02, -2.0162e-02, -1.9085e-02], [ 4.4850e-02, 4.9048e-02, -6.5618e-02, ..., -6.9612e-02, -7.8054e-02, -4.1381e-02], ..., [ 5.4832e-03, 2.1248e-02, -1.7342e-02, ..., -3.1538e-02, -8.6029e-02, 4.3753e-02], [-6.6382e-03, 6.6384e-02, 5.9615e-02, ..., -4.1311e-02, 3.5609e-02, -6.9969e-02], [-4.2128e-02, -1.3857e-02, 1.6435e-03, ..., -1.2707e-02, -3.8092e-02, 3.4588e-02]], device='cuda:0'), grad: tensor([[ 3.7514e-06, 8.4639e-06, 1.0662e-05, ..., 2.0713e-06, 1.2428e-05, 4.7274e-06], [ 5.3912e-05, 5.1826e-05, 7.1079e-06, ..., 5.4203e-07, 8.7842e-06, 5.4896e-05], [-2.5645e-05, 4.3124e-05, 1.4111e-05, ..., 1.7486e-07, -2.0936e-05, 2.5034e-05], ..., [-1.1146e-04, -1.0026e-04, 1.0125e-05, ..., 1.9092e-07, 2.3052e-05, -1.4460e-04], [ 1.1402e-04, 3.2282e-04, 1.0359e-04, ..., 3.8818e-06, 1.3340e-04, 1.5092e-04], [-1.3745e-04, -5.0020e-04, -1.6367e-04, ..., 5.4622e-07, -1.2362e-04, -2.3615e-04]], device='cuda:0') Epoch 67, bias, value: tensor([ 0.0258, -0.0206, -0.0063, 0.0020, 0.0039, 0.0017, -0.0163, -0.0016, 0.0291, 0.0191], device='cuda:0'), grad: tensor([ 3.0458e-05, 7.0810e-05, -1.8072e-04, -4.7088e-04, 1.7321e-04, 5.4169e-04, -2.5779e-06, -1.3900e-04, 5.0306e-04, -5.2595e-04], device='cuda:0') 100 0.0001 changing lr epoch 66, time 256.83, cls_loss 0.0078 cls_loss_mapping 0.0165 cls_loss_causal 0.6682 re_mapping 0.0121 re_causal 0.0377 /// teacc 98.65 lr 0.00010000 Epoch 68, weight, value: tensor([[ 0.0007, -0.0622, -0.0405, ..., -0.0097, -0.1043, -0.0007], [-0.0450, -0.0860, -0.0793, ..., -0.0666, -0.0204, -0.0186], [ 0.0449, 0.0487, -0.0662, ..., -0.0697, -0.0787, -0.0422], ..., [ 0.0064, 0.0224, -0.0175, ..., -0.0316, -0.0862, 0.0442], [-0.0064, 0.0670, 0.0602, ..., -0.0415, 0.0361, -0.0708], [-0.0423, -0.0137, 0.0009, ..., -0.0130, -0.0387, 0.0351]], device='cuda:0'), grad: tensor([[ 6.2436e-06, 9.9391e-06, 5.4985e-05, ..., 3.0044e-06, 3.3170e-05, 3.3796e-05], [ 1.5870e-05, -1.7166e-05, 2.4900e-05, ..., 6.3470e-07, 1.4082e-05, -2.7999e-05], [ 9.5785e-05, 1.3685e-04, 1.2703e-05, ..., 4.7497e-07, 1.0140e-05, 2.4140e-05], ..., [-1.1319e-04, -1.0139e-04, 1.1958e-05, ..., 1.8231e-07, 7.1637e-06, 1.6809e-04], [ 2.2322e-05, 2.5719e-05, 8.5771e-05, ..., 2.9653e-06, 4.6939e-05, 5.4091e-05], [-1.1510e-04, -1.6177e-04, 2.9221e-05, ..., 1.0515e-06, 1.7464e-05, -2.9731e-04]], device='cuda:0') Epoch 68, bias, value: tensor([ 0.0267, -0.0206, -0.0065, 0.0020, 0.0034, 0.0011, -0.0169, -0.0007, 0.0291, 0.0190], device='cuda:0'), grad: tensor([ 1.2094e-04, 5.5122e-04, 2.7823e-04, 5.2243e-05, -1.8225e-03, -3.1877e-04, 2.3103e-04, 9.1267e-04, 2.0647e-04, -2.1327e-04], device='cuda:0') 100 0.0001 changing lr epoch 67, time 255.57, cls_loss 0.0074 cls_loss_mapping 0.0166 cls_loss_causal 0.6581 re_mapping 0.0116 re_causal 0.0356 /// teacc 98.85 lr 0.00010000 Epoch 69, weight, value: tensor([[ 8.7428e-04, -6.3210e-02, -4.1182e-02, ..., -1.0532e-02, -1.0498e-01, -1.1126e-03], [-4.5573e-02, -8.6667e-02, -7.9679e-02, ..., -6.7155e-02, -2.0986e-02, -1.7908e-02], [ 4.5019e-02, 4.8490e-02, -6.6956e-02, ..., -7.0138e-02, -7.9265e-02, -4.2981e-02], ..., [ 6.9094e-03, 2.2966e-02, -1.7682e-02, ..., -3.1659e-02, -8.6639e-02, 4.4504e-02], [-6.0203e-03, 6.7481e-02, 6.0290e-02, ..., -4.2125e-02, 3.6301e-02, -7.1359e-02], [-4.2797e-02, -1.3476e-02, 2.9767e-05, ..., -1.3296e-02, -3.9486e-02, 3.5437e-02]], device='cuda:0'), grad: tensor([[-2.0504e-05, 3.8669e-06, 2.4773e-06, ..., 5.5460e-07, 2.3134e-06, 1.7704e-06], [ 7.6517e-06, 4.7572e-06, 2.8946e-06, ..., 2.8964e-07, 2.8685e-06, -4.8161e-05], [-2.6986e-05, -2.1253e-06, 5.9754e-06, ..., 2.1909e-07, 5.9344e-06, 5.1521e-06], ..., [-2.2084e-05, -4.3899e-05, 1.1936e-05, ..., 8.4983e-09, 1.0431e-05, -5.8979e-05], [-2.7046e-05, -4.3154e-05, -4.9353e-05, ..., 1.2079e-06, -4.1157e-05, 1.0788e-05], [ 5.0217e-05, 4.1306e-05, 1.9535e-05, ..., 1.1164e-07, 1.7881e-05, 5.5730e-05]], device='cuda:0') Epoch 69, bias, value: tensor([ 0.0263, -0.0202, -0.0072, 0.0020, 0.0043, 0.0013, -0.0179, -0.0003, 0.0288, 0.0190], device='cuda:0'), grad: tensor([-5.9426e-05, -7.1466e-05, -3.1263e-05, 3.6865e-05, 1.7673e-05, 2.6003e-05, 2.4095e-05, -6.1393e-05, -3.4779e-05, 1.5354e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 68---------------------------------------------------- epoch 68, time 273.95, cls_loss 0.0060 cls_loss_mapping 0.0145 cls_loss_causal 0.6212 re_mapping 0.0112 re_causal 0.0345 /// teacc 98.94 lr 0.00010000 Epoch 70, weight, value: tensor([[ 0.0015, -0.0638, -0.0416, ..., -0.0112, -0.1055, -0.0016], [-0.0462, -0.0871, -0.0801, ..., -0.0674, -0.0214, -0.0176], [ 0.0452, 0.0485, -0.0676, ..., -0.0705, -0.0800, -0.0434], ..., [ 0.0071, 0.0231, -0.0181, ..., -0.0317, -0.0870, 0.0447], [-0.0055, 0.0683, 0.0607, ..., -0.0429, 0.0368, -0.0719], [-0.0431, -0.0139, -0.0005, ..., -0.0134, -0.0397, 0.0354]], device='cuda:0'), grad: tensor([[-1.0282e-06, 1.9986e-06, 1.3806e-05, ..., 2.5239e-07, 1.3217e-05, 2.0247e-06], [ 2.4959e-05, 2.0340e-05, 1.5572e-05, ..., 7.7649e-08, 2.0251e-05, 1.4327e-05], [ 7.6652e-05, 2.4602e-05, 1.6600e-05, ..., 6.3912e-08, 1.2863e-04, 1.0020e-04], ..., [-1.5661e-05, -6.1244e-06, 9.0450e-06, ..., 1.9791e-09, 1.5542e-05, -4.1485e-05], [ 7.3947e-06, -2.0470e-06, 2.0280e-05, ..., 5.2433e-07, 2.3544e-05, 9.7081e-06], [ 2.0593e-05, 1.1557e-04, 4.8935e-05, ..., 3.5972e-08, 4.3243e-05, 1.5497e-05]], device='cuda:0') Epoch 70, bias, value: tensor([ 0.0260, -0.0201, -0.0076, 0.0015, 0.0039, 0.0014, -0.0168, -0.0002, 0.0291, 0.0190], device='cuda:0'), grad: tensor([ 1.1653e-05, 7.4923e-05, 2.2936e-04, -2.4533e-04, -3.4332e-04, -2.4843e-04, 8.2612e-05, 7.6964e-06, 5.7071e-05, 3.7432e-04], device='cuda:0') 100 0.0001 changing lr epoch 69, time 256.70, cls_loss 0.0069 cls_loss_mapping 0.0145 cls_loss_causal 0.6239 re_mapping 0.0116 re_causal 0.0351 /// teacc 98.88 lr 0.00010000 Epoch 71, weight, value: tensor([[ 0.0017, -0.0644, -0.0422, ..., -0.0119, -0.1060, -0.0017], [-0.0473, -0.0880, -0.0806, ..., -0.0676, -0.0217, -0.0179], [ 0.0456, 0.0488, -0.0681, ..., -0.0708, -0.0806, -0.0432], ..., [ 0.0072, 0.0229, -0.0183, ..., -0.0318, -0.0874, 0.0449], [-0.0051, 0.0695, 0.0614, ..., -0.0429, 0.0377, -0.0723], [-0.0432, -0.0140, -0.0009, ..., -0.0138, -0.0402, 0.0358]], device='cuda:0'), grad: tensor([[ 1.4484e-05, 8.9407e-05, 9.5606e-05, ..., 1.7672e-07, 4.8637e-05, 1.4476e-05], [ 4.5508e-05, 1.1390e-04, 2.9877e-05, ..., 8.5449e-08, 5.1051e-05, 8.2135e-05], [ 2.1994e-05, 9.1076e-05, 7.3850e-05, ..., 9.1502e-08, 6.4552e-05, 4.5240e-05], ..., [ 7.3090e-03, 1.6937e-02, 1.4432e-05, ..., 2.7940e-09, 7.0686e-03, 1.3145e-02], [-5.8603e-04, -3.4809e-03, -4.4937e-03, ..., 2.7800e-07, -1.4191e-03, 2.2054e-04], [ 1.1945e-04, 4.0030e-04, 2.9755e-04, ..., 3.4226e-08, 2.0528e-04, 1.5044e-04]], device='cuda:0') Epoch 71, bias, value: tensor([ 0.0259, -0.0202, -0.0076, 0.0009, 0.0041, 0.0019, -0.0169, -0.0003, 0.0298, 0.0188], device='cuda:0'), grad: tensor([ 1.3840e-04, 2.4796e-04, 2.1267e-04, -2.6596e-02, 2.3142e-05, 5.4283e-03, 1.2010e-04, 2.5330e-02, -5.6419e-03, 7.3528e-04], device='cuda:0') 100 0.0001 changing lr epoch 70, time 256.88, cls_loss 0.0064 cls_loss_mapping 0.0142 cls_loss_causal 0.6033 re_mapping 0.0112 re_causal 0.0343 /// teacc 98.80 lr 0.00010000 Epoch 72, weight, value: tensor([[ 0.0021, -0.0647, -0.0426, ..., -0.0120, -0.1066, -0.0017], [-0.0478, -0.0884, -0.0809, ..., -0.0679, -0.0220, -0.0174], [ 0.0459, 0.0490, -0.0687, ..., -0.0711, -0.0814, -0.0435], ..., [ 0.0072, 0.0228, -0.0187, ..., -0.0319, -0.0884, 0.0443], [-0.0050, 0.0700, 0.0618, ..., -0.0434, 0.0383, -0.0731], [-0.0437, -0.0147, -0.0016, ..., -0.0143, -0.0410, 0.0358]], device='cuda:0'), grad: tensor([[-1.7032e-05, 7.7263e-06, 7.0892e-06, ..., 5.2229e-06, 5.9865e-06, 2.4401e-06], [ 1.7108e-06, 8.0019e-06, 2.4755e-06, ..., 2.1271e-06, 6.4336e-06, 4.3847e-06], [ 1.3977e-05, -3.2457e-07, 2.6003e-06, ..., 1.3616e-06, 3.4738e-06, 4.4443e-06], ..., [ 2.7090e-05, 1.5450e-04, 5.5227e-07, ..., 1.8929e-07, 1.6296e-04, 2.8348e-04], [ 3.8259e-06, 2.4527e-05, 1.9684e-05, ..., 1.2547e-05, 1.6481e-05, 1.1839e-05], [-3.6135e-06, -2.2091e-06, -1.3160e-06, ..., 1.0068e-06, 9.8199e-06, 1.2092e-05]], device='cuda:0') Epoch 72, bias, value: tensor([ 0.0262, -0.0200, -0.0073, 0.0015, 0.0044, 0.0019, -0.0172, -0.0008, 0.0297, 0.0183], device='cuda:0'), grad: tensor([-2.9549e-05, 8.9332e-06, 4.4644e-05, -4.8447e-04, 3.3587e-05, 2.8446e-05, -8.8096e-05, 4.3106e-04, 4.4107e-05, 1.1973e-05], device='cuda:0') 100 0.0001 changing lr epoch 71, time 255.86, cls_loss 0.0074 cls_loss_mapping 0.0130 cls_loss_causal 0.6604 re_mapping 0.0108 re_causal 0.0352 /// teacc 98.85 lr 0.00010000 Epoch 73, weight, value: tensor([[ 0.0026, -0.0649, -0.0431, ..., -0.0123, -0.1072, -0.0019], [-0.0488, -0.0896, -0.0815, ..., -0.0688, -0.0242, -0.0180], [ 0.0462, 0.0495, -0.0692, ..., -0.0714, -0.0803, -0.0425], ..., [ 0.0072, 0.0226, -0.0188, ..., -0.0317, -0.0904, 0.0441], [-0.0050, 0.0702, 0.0617, ..., -0.0442, 0.0384, -0.0741], [-0.0439, -0.0147, -0.0019, ..., -0.0145, -0.0414, 0.0360]], device='cuda:0'), grad: tensor([[ 1.0550e-05, 1.4983e-05, 6.8322e-06, ..., 1.4342e-06, 8.1956e-06, 9.9167e-06], [ 3.3826e-05, 3.9607e-05, 5.6744e-05, ..., -8.9593e-07, 7.1585e-05, 6.8247e-05], [-7.0632e-05, -3.1948e-05, 2.3812e-05, ..., 1.2377e-06, 3.0071e-05, 4.9889e-05], ..., [-1.8217e-06, -6.9261e-05, 4.4733e-05, ..., 7.1945e-08, 5.6416e-05, -2.5734e-05], [ 5.1826e-05, 8.3148e-05, 4.1157e-05, ..., 2.7753e-06, 5.4926e-05, 7.0691e-05], [-6.0588e-05, -1.1861e-04, 2.6926e-05, ..., 1.6647e-07, 1.0200e-05, -3.5137e-05]], device='cuda:0') Epoch 73, bias, value: tensor([ 0.0265, -0.0208, -0.0062, 0.0015, 0.0045, 0.0027, -0.0173, -0.0014, 0.0292, 0.0180], device='cuda:0'), grad: tensor([ 4.5210e-05, 2.2721e-04, -2.2158e-05, -8.1635e-04, 4.3750e-05, 2.5940e-04, 3.6746e-05, 2.9325e-05, 2.2578e-04, -2.9221e-05], device='cuda:0') 100 0.0001 changing lr epoch 72, time 256.37, cls_loss 0.0081 cls_loss_mapping 0.0160 cls_loss_causal 0.6303 re_mapping 0.0113 re_causal 0.0316 /// teacc 98.84 lr 0.00010000 Epoch 74, weight, value: tensor([[ 0.0026, -0.0662, -0.0443, ..., -0.0127, -0.1088, -0.0027], [-0.0494, -0.0903, -0.0823, ..., -0.0696, -0.0250, -0.0175], [ 0.0468, 0.0500, -0.0696, ..., -0.0720, -0.0807, -0.0435], ..., [ 0.0073, 0.0231, -0.0192, ..., -0.0318, -0.0904, 0.0444], [-0.0049, 0.0706, 0.0620, ..., -0.0443, 0.0389, -0.0748], [-0.0439, -0.0144, -0.0025, ..., -0.0148, -0.0422, 0.0364]], device='cuda:0'), grad: tensor([[ 1.8859e-08, 4.0799e-05, 2.1458e-06, ..., 1.6941e-06, 3.2596e-06, 8.6948e-06], [ 2.5351e-06, 9.7677e-06, 3.2783e-06, ..., 4.9500e-07, 3.4049e-06, -4.7288e-07], [-6.9253e-06, -2.4238e-07, 8.0466e-07, ..., 4.6776e-07, 9.9465e-07, 3.3323e-06], ..., [-5.3160e-06, -2.4691e-05, 8.1677e-07, ..., 1.3039e-08, 1.5786e-06, -2.3335e-05], [ 6.6347e-06, 3.8207e-05, 5.6505e-05, ..., 1.7043e-06, 4.3333e-05, 2.5302e-05], [-5.5619e-06, -1.8537e-05, 1.3858e-05, ..., 1.2619e-07, 2.6766e-06, -1.2152e-05]], device='cuda:0') Epoch 74, bias, value: tensor([ 0.0256, -0.0203, -0.0064, 0.0017, 0.0051, 0.0023, -0.0172, -0.0011, 0.0289, 0.0179], device='cuda:0'), grad: tensor([ 6.4433e-05, 7.8082e-06, -3.1898e-07, 3.0518e-05, 1.3580e-03, -9.1493e-05, -1.4238e-03, -4.1038e-05, 1.1528e-04, -1.7866e-05], device='cuda:0') 100 0.0001 changing lr epoch 73, time 256.54, cls_loss 0.0065 cls_loss_mapping 0.0154 cls_loss_causal 0.5874 re_mapping 0.0112 re_causal 0.0323 /// teacc 98.89 lr 0.00010000 Epoch 75, weight, value: tensor([[ 0.0028, -0.0667, -0.0446, ..., -0.0130, -0.1092, -0.0029], [-0.0501, -0.0908, -0.0825, ..., -0.0702, -0.0253, -0.0179], [ 0.0472, 0.0500, -0.0699, ..., -0.0727, -0.0810, -0.0438], ..., [ 0.0074, 0.0231, -0.0194, ..., -0.0318, -0.0911, 0.0447], [-0.0048, 0.0704, 0.0620, ..., -0.0448, 0.0391, -0.0754], [-0.0438, -0.0128, -0.0027, ..., -0.0151, -0.0420, 0.0372]], device='cuda:0'), grad: tensor([[-8.6753e-07, 2.6479e-05, 9.8944e-05, ..., 1.6749e-05, 7.8201e-05, 2.8193e-05], [ 1.7034e-06, 5.9903e-06, 1.1668e-05, ..., 2.5090e-06, 1.1027e-05, -1.7151e-05], [ 1.2536e-06, 6.1616e-06, 1.2062e-05, ..., 1.8859e-06, 1.1817e-05, 5.5432e-06], ..., [ 1.4156e-06, 1.3383e-06, 9.1195e-06, ..., 1.9395e-07, 9.6560e-06, 2.5705e-06], [ 7.3947e-06, -1.6373e-06, 1.3924e-04, ..., 1.3694e-05, 1.0824e-04, 2.0355e-05], [ 1.7926e-05, 1.5065e-05, 7.8619e-05, ..., 1.5935e-06, 7.4565e-05, 4.1090e-06]], device='cuda:0') Epoch 75, bias, value: tensor([ 0.0256, -0.0207, -0.0066, 0.0016, 0.0046, 0.0019, -0.0169, -0.0011, 0.0282, 0.0195], device='cuda:0'), grad: tensor([ 1.1432e-04, -2.7210e-05, 2.9504e-05, -1.0484e-04, 3.6478e-05, -1.0926e-04, -2.5773e-04, 2.4974e-05, 1.6916e-04, 1.2481e-04], device='cuda:0') 100 0.0001 changing lr epoch 74, time 256.30, cls_loss 0.0083 cls_loss_mapping 0.0149 cls_loss_causal 0.6236 re_mapping 0.0110 re_causal 0.0311 /// teacc 98.92 lr 0.00010000 Epoch 76, weight, value: tensor([[ 0.0028, -0.0675, -0.0455, ..., -0.0135, -0.1100, -0.0032], [-0.0510, -0.0914, -0.0836, ..., -0.0712, -0.0259, -0.0177], [ 0.0477, 0.0498, -0.0708, ..., -0.0732, -0.0816, -0.0443], ..., [ 0.0075, 0.0233, -0.0197, ..., -0.0318, -0.0914, 0.0449], [-0.0044, 0.0716, 0.0629, ..., -0.0460, 0.0397, -0.0760], [-0.0435, -0.0122, -0.0035, ..., -0.0154, -0.0429, 0.0377]], device='cuda:0'), grad: tensor([[-9.0227e-06, 9.3579e-06, 2.3954e-06, ..., 1.5553e-06, 9.4548e-06, -8.4490e-06], [ 5.0701e-06, 5.4762e-06, 1.8002e-06, ..., 5.7183e-07, 1.1966e-05, -3.4682e-06], [-2.1422e-04, -1.5175e-04, 1.3448e-06, ..., -4.9965e-07, -2.2354e-03, -5.9515e-05], ..., [ 2.7210e-05, -3.8967e-06, 6.5984e-07, ..., -3.3788e-06, 1.2495e-05, -9.9540e-06], [ 4.1097e-05, 4.4674e-05, 7.3053e-06, ..., 2.8592e-06, 2.2843e-05, 2.4699e-06], [ 2.8953e-05, 1.4096e-05, -3.7234e-06, ..., 3.2187e-06, 2.1381e-03, 2.9370e-05]], device='cuda:0') Epoch 76, bias, value: tensor([ 0.0242, -0.0205, -0.0066, 0.0014, 0.0045, 0.0018, -0.0167, -0.0010, 0.0285, 0.0201], device='cuda:0'), grad: tensor([-5.1945e-05, 3.4511e-05, -1.2840e-02, 3.0589e-04, 3.7789e-05, 1.1724e-04, 3.7923e-06, 5.9754e-05, 1.3220e-04, 1.2207e-02], device='cuda:0') 100 0.0001 changing lr epoch 75, time 256.19, cls_loss 0.0061 cls_loss_mapping 0.0124 cls_loss_causal 0.6168 re_mapping 0.0116 re_causal 0.0338 /// teacc 98.74 lr 0.00010000 Epoch 77, weight, value: tensor([[ 0.0033, -0.0679, -0.0471, ..., -0.0140, -0.1113, -0.0036], [-0.0517, -0.0918, -0.0845, ..., -0.0718, -0.0263, -0.0174], [ 0.0476, 0.0497, -0.0714, ..., -0.0731, -0.0817, -0.0448], ..., [ 0.0077, 0.0238, -0.0206, ..., -0.0318, -0.0916, 0.0452], [-0.0038, 0.0723, 0.0634, ..., -0.0463, 0.0402, -0.0762], [-0.0436, -0.0117, -0.0036, ..., -0.0157, -0.0435, 0.0375]], device='cuda:0'), grad: tensor([[-6.2659e-06, 1.4305e-06, 1.5544e-06, ..., 7.1060e-07, 1.0161e-06, 5.4808e-07], [-1.3057e-06, 3.3434e-06, -1.6037e-06, ..., 4.6054e-07, 2.1253e-06, -1.4819e-05], [ 3.8818e-06, 5.9977e-06, 8.3447e-06, ..., 2.6505e-06, 6.9551e-06, 3.1255e-06], ..., [-9.2804e-05, -1.5569e-04, 3.4310e-06, ..., 1.9791e-08, 3.8370e-06, -1.1081e-04], [ 5.3756e-06, 1.7673e-05, 2.9624e-05, ..., 1.0885e-05, 2.1175e-05, 8.5384e-06], [ 8.6606e-05, 1.3816e-04, 2.4103e-06, ..., 1.1758e-07, 1.6652e-06, 1.0437e-04]], device='cuda:0') Epoch 77, bias, value: tensor([ 0.0242, -0.0207, -0.0066, 0.0009, 0.0043, 0.0026, -0.0162, -0.0011, 0.0288, 0.0198], device='cuda:0'), grad: tensor([ 6.5938e-07, -3.6836e-05, 2.4170e-05, -2.3648e-05, -9.3699e-05, 1.8567e-05, -2.8938e-05, -1.5020e-04, 5.3316e-05, 2.3627e-04], device='cuda:0') 100 0.0001 changing lr epoch 76, time 256.68, cls_loss 0.0078 cls_loss_mapping 0.0145 cls_loss_causal 0.6116 re_mapping 0.0112 re_causal 0.0331 /// teacc 98.78 lr 0.00010000 Epoch 78, weight, value: tensor([[ 0.0039, -0.0682, -0.0474, ..., -0.0142, -0.1115, -0.0036], [-0.0528, -0.0926, -0.0865, ..., -0.0727, -0.0272, -0.0174], [ 0.0474, 0.0489, -0.0728, ..., -0.0732, -0.0823, -0.0456], ..., [ 0.0081, 0.0244, -0.0210, ..., -0.0319, -0.0921, 0.0454], [-0.0028, 0.0726, 0.0634, ..., -0.0471, 0.0402, -0.0769], [-0.0438, -0.0119, -0.0048, ..., -0.0158, -0.0446, 0.0379]], device='cuda:0'), grad: tensor([[ 3.1561e-05, 2.9370e-05, 2.5988e-05, ..., 1.4693e-05, 2.0579e-05, 5.6103e-06], [ 1.5748e-04, 1.4305e-04, 4.6563e-04, ..., 6.9067e-06, 4.0507e-04, 6.4135e-05], [-3.0351e-04, -8.1435e-06, 2.0266e-05, ..., 1.0505e-05, 1.9982e-05, 6.6422e-06], ..., [ 1.0103e-04, 9.6485e-06, 2.8655e-05, ..., 4.1840e-07, 2.8476e-05, -8.5831e-06], [-1.4520e-04, -1.6940e-04, -6.2895e-04, ..., 1.8090e-05, -4.8018e-04, -6.8188e-05], [ 1.7092e-05, 1.3001e-05, 5.8621e-05, ..., 2.0675e-06, 5.7399e-05, 1.1079e-05]], device='cuda:0') Epoch 78, bias, value: tensor([ 2.4505e-02, -2.1379e-02, -7.4757e-03, 1.2477e-03, 5.5767e-03, 2.9672e-03, -1.6983e-02, -9.5242e-05, 2.8568e-02, 1.8986e-02], device='cuda:0'), grad: tensor([ 1.3101e-04, 8.4686e-04, -5.9843e-04, -7.6115e-05, 1.0449e-04, -1.3582e-05, -1.3614e-04, 2.5225e-04, -7.0381e-04, 1.9383e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 77---------------------------------------------------- epoch 77, time 273.90, cls_loss 0.0068 cls_loss_mapping 0.0121 cls_loss_causal 0.5957 re_mapping 0.0109 re_causal 0.0316 /// teacc 98.95 lr 0.00010000 Epoch 79, weight, value: tensor([[ 0.0040, -0.0688, -0.0480, ..., -0.0147, -0.1121, -0.0039], [-0.0537, -0.0929, -0.0865, ..., -0.0730, -0.0275, -0.0170], [ 0.0490, 0.0502, -0.0735, ..., -0.0735, -0.0827, -0.0457], ..., [ 0.0070, 0.0231, -0.0213, ..., -0.0320, -0.0926, 0.0452], [-0.0026, 0.0733, 0.0648, ..., -0.0470, 0.0412, -0.0779], [-0.0427, -0.0109, -0.0055, ..., -0.0161, -0.0451, 0.0391]], device='cuda:0'), grad: tensor([[-3.3307e-04, -1.1379e-04, 1.5467e-05, ..., 6.0536e-07, 7.7859e-06, -9.9897e-05], [ 1.9884e-04, 6.9022e-05, 9.1642e-06, ..., 1.1711e-07, 5.9381e-06, -4.9162e-04], [ 3.1590e-05, 1.2085e-05, 5.1558e-06, ..., 1.2340e-07, 3.1516e-06, 2.3913e-04], ..., [ 1.1824e-05, -1.3895e-06, 7.2233e-06, ..., 9.0338e-08, 4.1462e-06, 1.9681e-04], [ 6.5923e-05, 3.1814e-06, 3.0398e-04, ..., 3.2969e-06, 1.4913e-04, 1.0476e-05], [ 1.2957e-05, 2.9225e-06, 3.2932e-05, ..., 7.5949e-07, 1.8641e-05, 4.2260e-05]], device='cuda:0') Epoch 79, bias, value: tensor([ 0.0245, -0.0211, -0.0066, 0.0014, 0.0050, 0.0024, -0.0175, -0.0003, 0.0284, 0.0195], device='cuda:0'), grad: tensor([-0.0036, 0.0012, 0.0008, 0.0004, -0.0002, -0.0005, 0.0006, 0.0006, 0.0005, 0.0002], device='cuda:0') 100 0.0001 changing lr epoch 78, time 255.43, cls_loss 0.0055 cls_loss_mapping 0.0140 cls_loss_causal 0.6005 re_mapping 0.0110 re_causal 0.0323 /// teacc 98.90 lr 0.00010000 Epoch 80, weight, value: tensor([[ 0.0052, -0.0692, -0.0486, ..., -0.0152, -0.1127, -0.0041], [-0.0542, -0.0932, -0.0867, ..., -0.0732, -0.0277, -0.0157], [ 0.0490, 0.0501, -0.0746, ..., -0.0738, -0.0834, -0.0460], ..., [ 0.0068, 0.0231, -0.0215, ..., -0.0320, -0.0927, 0.0445], [-0.0027, 0.0738, 0.0649, ..., -0.0479, 0.0414, -0.0785], [-0.0430, -0.0111, -0.0060, ..., -0.0164, -0.0455, 0.0388]], device='cuda:0'), grad: tensor([[ 1.4435e-07, 2.7344e-06, -5.7742e-06, ..., 8.7311e-07, 1.2117e-06, -9.5665e-06], [ 3.7730e-05, 4.0382e-05, 1.2200e-06, ..., 1.4692e-07, 5.6624e-07, 1.1629e-04], [ 1.7941e-05, 1.9625e-05, 5.7183e-07, ..., 1.4459e-07, 4.8475e-07, 3.0130e-05], ..., [-1.5247e-04, -1.7619e-04, 4.5262e-07, ..., 3.6787e-08, 2.2026e-07, -2.7776e-04], [ 7.1861e-06, 9.6262e-06, 5.3607e-06, ..., 1.6559e-06, 2.2240e-06, 5.5581e-06], [ 1.6406e-05, 1.8388e-05, 3.8370e-06, ..., 4.7171e-07, 1.9222e-06, 2.9400e-05]], device='cuda:0') Epoch 80, bias, value: tensor([ 0.0249, -0.0203, -0.0073, 0.0022, 0.0051, 0.0019, -0.0167, -0.0009, 0.0282, 0.0190], device='cuda:0'), grad: tensor([-5.2780e-05, 2.2399e-04, 5.7489e-05, 1.9169e-04, 2.1055e-05, 1.9632e-06, 2.9922e-05, -5.7650e-04, 3.0428e-05, 7.3016e-05], device='cuda:0') 100 0.0001 changing lr epoch 79, time 256.76, cls_loss 0.0057 cls_loss_mapping 0.0145 cls_loss_causal 0.6349 re_mapping 0.0101 re_causal 0.0322 /// teacc 98.74 lr 0.00010000 Epoch 81, weight, value: tensor([[ 0.0056, -0.0693, -0.0491, ..., -0.0155, -0.1134, -0.0039], [-0.0554, -0.0940, -0.0872, ..., -0.0739, -0.0279, -0.0163], [ 0.0496, 0.0502, -0.0756, ..., -0.0740, -0.0835, -0.0455], ..., [ 0.0073, 0.0237, -0.0216, ..., -0.0320, -0.0928, 0.0449], [-0.0030, 0.0740, 0.0652, ..., -0.0481, 0.0416, -0.0790], [-0.0432, -0.0113, -0.0065, ..., -0.0164, -0.0459, 0.0391]], device='cuda:0'), grad: tensor([[-2.2367e-05, 3.9250e-05, 5.2124e-05, ..., 2.6613e-05, 1.7852e-05, 6.2063e-06], [ 6.1840e-06, 4.4592e-06, 1.7006e-06, ..., 3.7323e-07, 1.7090e-06, -6.3144e-06], [-2.5249e-04, -1.2970e-04, 9.9242e-06, ..., 6.6543e-07, 1.1824e-05, 1.4640e-06], ..., [ 2.1172e-04, 1.2958e-04, 6.1933e-07, ..., 3.1199e-08, 1.1558e-06, 5.6531e-07], [ 1.0990e-05, -3.4243e-05, -2.8655e-05, ..., 1.1846e-06, -3.8147e-05, 3.6042e-06], [-4.3539e-07, -8.0094e-06, 1.5600e-06, ..., 3.7323e-07, 1.5297e-07, -6.8769e-06]], device='cuda:0') Epoch 81, bias, value: tensor([ 0.0251, -0.0214, -0.0070, 0.0020, 0.0051, 0.0021, -0.0162, -0.0003, 0.0279, 0.0190], device='cuda:0'), grad: tensor([ 5.7101e-05, 2.9220e-07, -4.1199e-04, 9.9480e-05, 1.1683e-05, 2.5392e-05, -8.4519e-05, 3.2520e-04, -2.5362e-05, 3.0212e-06], device='cuda:0') 100 0.0001 changing lr epoch 80, time 257.05, cls_loss 0.0047 cls_loss_mapping 0.0117 cls_loss_causal 0.6047 re_mapping 0.0107 re_causal 0.0331 /// teacc 98.91 lr 0.00010000 Epoch 82, weight, value: tensor([[ 0.0058, -0.0700, -0.0509, ..., -0.0161, -0.1155, -0.0043], [-0.0558, -0.0938, -0.0879, ..., -0.0741, -0.0284, -0.0161], [ 0.0502, 0.0502, -0.0761, ..., -0.0741, -0.0835, -0.0460], ..., [ 0.0073, 0.0240, -0.0221, ..., -0.0320, -0.0929, 0.0451], [-0.0031, 0.0742, 0.0653, ..., -0.0483, 0.0416, -0.0795], [-0.0436, -0.0114, -0.0067, ..., -0.0167, -0.0457, 0.0391]], device='cuda:0'), grad: tensor([[-7.2531e-06, -6.3656e-07, 2.3074e-07, ..., 8.2655e-08, 6.2818e-07, 3.6322e-08], [ 3.0152e-07, 2.8359e-07, 4.5006e-07, ..., 1.3504e-08, 7.4226e-07, -4.6372e-05], [ 1.1642e-06, 3.1688e-07, 1.1176e-06, ..., 6.7754e-08, 1.6280e-06, 2.0385e-05], ..., [ 5.7416e-07, 3.8813e-07, 8.5123e-07, ..., 1.6298e-09, 1.7975e-06, 4.2729e-06], [-2.1346e-06, -5.0552e-06, -6.4485e-06, ..., 2.3725e-07, -2.4326e-06, 1.3430e-06], [ 1.5153e-06, -2.7940e-07, 1.7881e-06, ..., 1.5367e-08, 3.3453e-06, 5.7602e-07]], device='cuda:0') Epoch 82, bias, value: tensor([ 0.0244, -0.0211, -0.0063, 0.0012, 0.0054, 0.0028, -0.0160, -0.0004, 0.0273, 0.0188], device='cuda:0'), grad: tensor([-1.9729e-05, -1.2100e-04, 6.4075e-05, -3.9816e-05, -1.2949e-05, 5.8651e-05, 2.7552e-05, 1.9163e-05, 6.4783e-06, 1.7777e-05], device='cuda:0') 100 0.0001 changing lr epoch 81, time 256.66, cls_loss 0.0057 cls_loss_mapping 0.0124 cls_loss_causal 0.6072 re_mapping 0.0104 re_causal 0.0312 /// teacc 98.95 lr 0.00010000 Epoch 83, weight, value: tensor([[ 0.0058, -0.0707, -0.0526, ..., -0.0166, -0.1172, -0.0049], [-0.0562, -0.0953, -0.0879, ..., -0.0742, -0.0285, -0.0158], [ 0.0504, 0.0510, -0.0771, ..., -0.0742, -0.0842, -0.0461], ..., [ 0.0077, 0.0243, -0.0226, ..., -0.0320, -0.0931, 0.0454], [-0.0028, 0.0746, 0.0656, ..., -0.0487, 0.0419, -0.0802], [-0.0436, -0.0111, -0.0072, ..., -0.0163, -0.0461, 0.0392]], device='cuda:0'), grad: tensor([[-3.4302e-05, 2.2091e-06, 2.2631e-06, ..., 4.0862e-07, 1.8161e-06, -2.2780e-06], [ 4.3958e-06, 3.1162e-06, 2.2631e-06, ..., 1.0058e-07, 1.8440e-06, -7.8082e-06], [ 4.5389e-05, 3.8177e-05, 2.2985e-06, ..., 2.3399e-07, 2.0973e-06, 7.4804e-05], ..., [ 2.1562e-05, 2.1249e-05, 1.1921e-06, ..., 6.7521e-09, 2.5500e-06, 3.3051e-05], [ 2.2575e-06, -3.7886e-06, -2.5332e-05, ..., 1.1567e-06, -1.6585e-05, 1.1422e-05], [-1.1310e-05, -3.5137e-05, 2.5686e-06, ..., 5.9139e-08, -4.9919e-06, -1.6406e-05]], device='cuda:0') Epoch 83, bias, value: tensor([ 0.0239, -0.0209, -0.0060, 0.0013, 0.0054, 0.0029, -0.0163, -0.0004, 0.0272, 0.0188], device='cuda:0'), grad: tensor([-9.9242e-05, -1.6749e-05, 4.7326e-04, -6.6519e-04, 1.4797e-05, 4.3571e-05, 3.8356e-05, 2.1040e-04, 2.7895e-05, -2.8282e-05], device='cuda:0') 100 0.0001 changing lr epoch 82, time 257.17, cls_loss 0.0060 cls_loss_mapping 0.0123 cls_loss_causal 0.6140 re_mapping 0.0100 re_causal 0.0301 /// teacc 98.77 lr 0.00010000 Epoch 84, weight, value: tensor([[ 0.0060, -0.0711, -0.0534, ..., -0.0169, -0.1180, -0.0051], [-0.0567, -0.0966, -0.0901, ..., -0.0747, -0.0305, -0.0138], [ 0.0504, 0.0506, -0.0778, ..., -0.0747, -0.0847, -0.0467], ..., [ 0.0080, 0.0247, -0.0230, ..., -0.0321, -0.0933, 0.0448], [-0.0025, 0.0758, 0.0665, ..., -0.0498, 0.0432, -0.0813], [-0.0435, -0.0111, -0.0075, ..., -0.0165, -0.0462, 0.0393]], device='cuda:0'), grad: tensor([[-1.1422e-05, 4.6454e-06, 8.7842e-06, ..., 4.7572e-06, 5.6848e-06, 6.3516e-07], [ 1.1651e-06, -8.1509e-06, 3.5055e-06, ..., 1.9968e-06, 3.6471e-06, -7.1406e-05], [ 5.3160e-06, 1.8850e-05, 6.4299e-06, ..., 1.7434e-06, 9.3728e-06, 6.8486e-05], ..., [ 5.6513e-06, -5.5246e-06, 6.7614e-07, ..., 6.3097e-08, 1.8328e-05, 3.3498e-05], [ 2.0228e-06, 7.9572e-06, 2.5764e-05, ..., 1.5721e-05, 9.6411e-06, 7.5698e-06], [ 4.5076e-06, 4.5672e-06, 3.6396e-06, ..., 8.7265e-07, 4.6566e-06, 8.7395e-06]], device='cuda:0') Epoch 84, bias, value: tensor([ 0.0240, -0.0208, -0.0063, 0.0013, 0.0050, 0.0026, -0.0158, -0.0010, 0.0282, 0.0188], device='cuda:0'), grad: tensor([-2.5347e-05, -1.2565e-04, 1.3733e-04, -5.2810e-05, -2.9728e-06, 9.7334e-05, -1.5128e-04, 4.7415e-05, 3.6031e-05, 3.9637e-05], device='cuda:0') 100 0.0001 changing lr epoch 83, time 257.23, cls_loss 0.0053 cls_loss_mapping 0.0114 cls_loss_causal 0.6007 re_mapping 0.0107 re_causal 0.0318 /// teacc 98.86 lr 0.00010000 Epoch 85, weight, value: tensor([[ 0.0066, -0.0714, -0.0538, ..., -0.0179, -0.1178, -0.0056], [-0.0572, -0.0969, -0.0903, ..., -0.0749, -0.0306, -0.0137], [ 0.0507, 0.0507, -0.0787, ..., -0.0754, -0.0854, -0.0471], ..., [ 0.0081, 0.0249, -0.0233, ..., -0.0321, -0.0935, 0.0450], [-0.0026, 0.0758, 0.0664, ..., -0.0503, 0.0431, -0.0820], [-0.0434, -0.0110, -0.0079, ..., -0.0169, -0.0465, 0.0396]], device='cuda:0'), grad: tensor([[-4.0770e-05, 1.4463e-06, 2.3078e-06, ..., 1.8785e-06, -4.5099e-07, 5.0478e-07], [ 8.6650e-06, 8.9407e-06, 6.4224e-06, ..., 1.0328e-06, 7.6443e-06, -1.6391e-05], [-1.4044e-06, -2.2855e-06, 2.4989e-05, ..., 5.0664e-06, 4.0263e-05, 1.7002e-05], ..., [-1.8895e-05, 5.7332e-06, 3.8520e-06, ..., 2.1607e-07, 6.0014e-06, -2.2322e-05], [-7.8902e-06, -1.4104e-05, -1.0066e-05, ..., 3.0082e-06, -8.9034e-06, 4.7162e-06], [-3.7197e-06, -3.7760e-05, -3.6806e-05, ..., 3.6927e-07, -2.7165e-05, -1.1280e-05]], device='cuda:0') Epoch 85, bias, value: tensor([ 0.0249, -0.0210, -0.0063, 0.0012, 0.0050, 0.0026, -0.0158, -0.0008, 0.0277, 0.0188], device='cuda:0'), grad: tensor([-8.9109e-05, -1.0990e-05, 3.5644e-05, 2.4378e-05, 1.4067e-04, 4.1038e-05, 5.3972e-05, -2.2814e-05, 5.3167e-05, -2.2602e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 84---------------------------------------------------- epoch 84, time 275.12, cls_loss 0.0053 cls_loss_mapping 0.0102 cls_loss_causal 0.5924 re_mapping 0.0103 re_causal 0.0303 /// teacc 98.96 lr 0.00010000 Epoch 86, weight, value: tensor([[ 0.0070, -0.0717, -0.0543, ..., -0.0179, -0.1184, -0.0057], [-0.0575, -0.0971, -0.0905, ..., -0.0754, -0.0306, -0.0120], [ 0.0508, 0.0503, -0.0795, ..., -0.0765, -0.0866, -0.0475], ..., [ 0.0083, 0.0246, -0.0236, ..., -0.0321, -0.0942, 0.0438], [-0.0024, 0.0762, 0.0666, ..., -0.0507, 0.0434, -0.0825], [-0.0430, -0.0102, -0.0084, ..., -0.0172, -0.0470, 0.0403]], device='cuda:0'), grad: tensor([[-9.4017e-07, 5.5939e-05, 2.5019e-05, ..., 4.4614e-05, 1.8910e-05, 1.5959e-05], [ 4.1798e-06, 1.1057e-05, 3.2336e-06, ..., 2.3395e-06, 2.2743e-06, -1.3340e-04], [-4.6566e-08, 8.4341e-06, 4.6529e-06, ..., 4.6566e-06, 3.3397e-06, 5.7034e-06], ..., [-2.2173e-05, -4.5985e-05, 1.3784e-06, ..., 2.2375e-07, 9.5367e-07, 2.0079e-06], [ 1.6941e-06, 2.3782e-05, 3.0965e-05, ..., 1.8567e-05, 2.2218e-05, 9.6038e-06], [ 1.1131e-05, 2.5049e-05, 6.3144e-06, ..., 3.1963e-06, 4.1984e-06, 8.0645e-05]], device='cuda:0') Epoch 86, bias, value: tensor([ 0.0252, -0.0200, -0.0070, 0.0019, 0.0048, 0.0021, -0.0157, -0.0015, 0.0276, 0.0189], device='cuda:0'), grad: tensor([ 9.3222e-05, -2.2554e-04, 2.1383e-05, 1.7357e-04, 1.0103e-05, -1.7464e-04, -1.8120e-04, -1.0133e-06, 7.8857e-05, 2.0480e-04], device='cuda:0') 100 0.0001 changing lr epoch 85, time 257.03, cls_loss 0.0047 cls_loss_mapping 0.0103 cls_loss_causal 0.6027 re_mapping 0.0100 re_causal 0.0296 /// teacc 98.87 lr 0.00010000 Epoch 87, weight, value: tensor([[ 0.0074, -0.0719, -0.0550, ..., -0.0179, -0.1191, -0.0059], [-0.0581, -0.0974, -0.0908, ..., -0.0758, -0.0308, -0.0119], [ 0.0510, 0.0498, -0.0805, ..., -0.0771, -0.0874, -0.0481], ..., [ 0.0087, 0.0254, -0.0240, ..., -0.0321, -0.0946, 0.0446], [-0.0015, 0.0768, 0.0669, ..., -0.0518, 0.0437, -0.0834], [-0.0434, -0.0107, -0.0091, ..., -0.0176, -0.0476, 0.0400]], device='cuda:0'), grad: tensor([[ 1.8209e-05, 9.8571e-06, 6.1810e-05, ..., 1.5460e-07, 7.0393e-05, 8.2180e-06], [ 1.1139e-05, 1.7524e-05, 1.6719e-05, ..., 3.1991e-07, 1.5318e-05, -1.1083e-07], [ 1.3776e-05, 7.6592e-06, 2.9370e-05, ..., 9.4436e-07, 3.3885e-05, 5.8785e-06], ..., [ 1.6451e-05, 4.1783e-05, 8.6427e-06, ..., 3.5856e-08, 7.6964e-06, 3.2932e-05], [ 9.7990e-05, 1.2052e-04, 1.3061e-05, ..., 1.1027e-06, 3.3110e-05, 9.6262e-05], [-2.0170e-04, -3.5286e-04, 2.6822e-05, ..., 3.3597e-07, 2.9147e-05, -2.3282e-04]], device='cuda:0') Epoch 87, bias, value: tensor([ 0.0253, -0.0200, -0.0075, 0.0015, 0.0051, 0.0025, -0.0160, -0.0009, 0.0283, 0.0181], device='cuda:0'), grad: tensor([ 2.6369e-04, 5.3525e-05, 1.3244e-04, -4.3559e-04, 5.5313e-04, 6.6817e-05, 5.7817e-05, 1.0705e-04, 3.7074e-04, -1.1692e-03], device='cuda:0') 100 0.0001 changing lr epoch 86, time 257.45, cls_loss 0.0064 cls_loss_mapping 0.0138 cls_loss_causal 0.5793 re_mapping 0.0107 re_causal 0.0297 /// teacc 98.88 lr 0.00010000 Epoch 88, weight, value: tensor([[ 0.0081, -0.0726, -0.0554, ..., -0.0182, -0.1196, -0.0060], [-0.0586, -0.0976, -0.0912, ..., -0.0766, -0.0310, -0.0117], [ 0.0514, 0.0498, -0.0810, ..., -0.0770, -0.0880, -0.0485], ..., [ 0.0102, 0.0280, -0.0242, ..., -0.0318, -0.0946, 0.0458], [-0.0015, 0.0768, 0.0669, ..., -0.0525, 0.0438, -0.0842], [-0.0447, -0.0126, -0.0095, ..., -0.0181, -0.0478, 0.0393]], device='cuda:0'), grad: tensor([[ 1.6671e-06, 3.6508e-06, 1.3914e-06, ..., 4.3819e-07, 2.1569e-06, 1.5553e-07], [ 3.0398e-06, 5.0813e-06, 1.4743e-06, ..., 2.1863e-07, 3.5856e-06, 8.1258e-08], [-9.0837e-05, -8.3506e-05, 1.7285e-06, ..., 3.9418e-07, -1.7896e-05, 1.4100e-06], ..., [ 2.8983e-05, 2.1815e-05, 5.3598e-07, ..., 2.2119e-08, 1.3039e-05, -1.0118e-05], [ 8.7991e-06, 7.7784e-06, -6.3032e-06, ..., 4.4890e-06, -5.5693e-06, 5.7742e-07], [ 7.8380e-06, 1.4693e-05, 8.1509e-06, ..., 1.1828e-07, 1.1489e-05, 3.1553e-06]], device='cuda:0') Epoch 88, bias, value: tensor([ 0.0254, -0.0199, -0.0074, 0.0012, 0.0050, 0.0031, -0.0170, 0.0007, 0.0277, 0.0170], device='cuda:0'), grad: tensor([ 3.9116e-06, 1.3597e-05, -1.4043e-04, 4.7028e-05, -4.4137e-05, 2.3782e-05, -1.3774e-06, 4.9621e-05, 1.3024e-05, 3.5167e-05], device='cuda:0') 100 0.0001 changing lr epoch 87, time 256.87, cls_loss 0.0054 cls_loss_mapping 0.0110 cls_loss_causal 0.6113 re_mapping 0.0096 re_causal 0.0296 /// teacc 98.94 lr 0.00010000 Epoch 89, weight, value: tensor([[ 0.0095, -0.0721, -0.0558, ..., -0.0184, -0.1200, -0.0076], [-0.0596, -0.0982, -0.0913, ..., -0.0769, -0.0312, -0.0114], [ 0.0519, 0.0505, -0.0815, ..., -0.0772, -0.0885, -0.0490], ..., [ 0.0098, 0.0278, -0.0245, ..., -0.0319, -0.0951, 0.0456], [-0.0013, 0.0774, 0.0674, ..., -0.0526, 0.0444, -0.0847], [-0.0451, -0.0128, -0.0100, ..., -0.0184, -0.0483, 0.0399]], device='cuda:0'), grad: tensor([[-7.8022e-05, 2.4624e-06, 1.3001e-06, ..., 1.0999e-06, 1.1856e-06, 8.2050e-07], [ 2.6021e-06, 2.0992e-06, 6.4960e-07, ..., -1.1222e-07, 6.0257e-07, -4.1097e-05], [ 8.0094e-06, -1.8803e-06, 1.0775e-06, ..., 2.7008e-07, 9.6206e-07, 6.5267e-06], ..., [-3.0790e-06, -1.8030e-05, 2.0722e-07, ..., 9.5228e-08, 1.9209e-07, -1.0334e-05], [ 3.6955e-05, -3.8445e-06, -6.8396e-06, ..., 1.6456e-06, -5.3421e-06, 5.1968e-06], [ 9.4622e-06, 3.3472e-06, 4.6636e-07, ..., 2.1723e-07, 4.2235e-07, 4.8243e-06]], device='cuda:0') Epoch 89, bias, value: tensor([ 0.0246, -0.0201, -0.0069, 0.0014, 0.0055, 0.0032, -0.0174, 0.0002, 0.0277, 0.0174], device='cuda:0'), grad: tensor([-1.3626e-04, -7.7486e-05, 1.7196e-05, 2.8387e-05, 5.2869e-05, 2.2590e-05, 1.2659e-05, -2.6777e-05, 7.7367e-05, 2.9311e-05], device='cuda:0') 100 0.0001 changing lr epoch 88, time 257.51, cls_loss 0.0048 cls_loss_mapping 0.0120 cls_loss_causal 0.6052 re_mapping 0.0094 re_causal 0.0286 /// teacc 98.92 lr 0.00010000 Epoch 90, weight, value: tensor([[ 0.0098, -0.0727, -0.0564, ..., -0.0188, -0.1206, -0.0076], [-0.0599, -0.0984, -0.0916, ..., -0.0773, -0.0317, -0.0111], [ 0.0527, 0.0513, -0.0823, ..., -0.0776, -0.0891, -0.0482], ..., [ 0.0097, 0.0278, -0.0249, ..., -0.0320, -0.0947, 0.0453], [-0.0011, 0.0777, 0.0674, ..., -0.0526, 0.0443, -0.0851], [-0.0451, -0.0126, -0.0100, ..., -0.0158, -0.0482, 0.0400]], device='cuda:0'), grad: tensor([[-3.3118e-06, 5.0850e-06, 9.6206e-07, ..., 5.9837e-07, 1.1129e-06, 1.0923e-05], [-5.0932e-05, -2.0787e-05, 1.0887e-06, ..., 8.2888e-08, 1.5553e-06, -6.7532e-05], [ 1.5995e-07, -5.0589e-06, 2.6133e-06, ..., 1.7486e-07, 3.5428e-06, 2.7344e-05], ..., [ 8.9049e-05, 1.1885e-04, 8.3726e-07, ..., 1.0245e-08, 1.2498e-06, 2.2125e-04], [ 3.4064e-05, 1.7330e-05, 1.5413e-06, ..., 7.1106e-07, 2.3916e-06, 4.3988e-05], [-8.4341e-05, -1.2708e-04, 1.3821e-06, ..., 8.0094e-08, 2.1067e-06, -8.7380e-05]], device='cuda:0') Epoch 90, bias, value: tensor([ 2.4833e-02, -2.0306e-02, -6.4494e-03, 2.0833e-03, 5.4204e-03, 2.9967e-03, -1.7723e-02, -4.4118e-05, 2.7293e-02, 1.7605e-02], device='cuda:0'), grad: tensor([ 1.2584e-05, -2.6798e-04, 1.2493e-04, -3.7029e-06, -9.7752e-04, 7.1466e-05, 6.3479e-05, 7.0906e-04, 2.0468e-04, 6.2466e-05], device='cuda:0') 100 0.0001 changing lr epoch 89, time 255.45, cls_loss 0.0050 cls_loss_mapping 0.0111 cls_loss_causal 0.6131 re_mapping 0.0096 re_causal 0.0292 /// teacc 98.93 lr 0.00010000 Epoch 91, weight, value: tensor([[ 0.0110, -0.0724, -0.0570, ..., -0.0185, -0.1212, -0.0073], [-0.0603, -0.0987, -0.0920, ..., -0.0782, -0.0320, -0.0106], [ 0.0524, 0.0504, -0.0833, ..., -0.0780, -0.0903, -0.0486], ..., [ 0.0098, 0.0280, -0.0251, ..., -0.0320, -0.0949, 0.0453], [-0.0004, 0.0785, 0.0676, ..., -0.0533, 0.0446, -0.0862], [-0.0451, -0.0122, -0.0103, ..., -0.0159, -0.0482, 0.0401]], device='cuda:0'), grad: tensor([[ 3.0901e-06, 8.2105e-06, 1.6719e-05, ..., 3.6098e-06, 1.2934e-05, 1.3029e-06], [ 1.2465e-05, 2.6911e-05, 1.9252e-05, ..., 3.3248e-07, 1.8179e-05, 1.6451e-05], [ 2.5287e-05, 4.7773e-05, 8.4937e-05, ..., 5.4296e-07, 7.3493e-05, 1.1764e-05], ..., [-1.8850e-05, -6.2525e-05, 5.9456e-05, ..., 2.5262e-08, 4.4465e-05, -7.3910e-05], [ 1.8930e-04, 2.8062e-04, 7.6771e-04, ..., 6.6385e-06, 6.6566e-04, 1.5292e-06], [ 1.3709e-05, 1.9491e-05, 2.2516e-05, ..., 2.2701e-07, 2.0847e-05, 1.1906e-05]], device='cuda:0') Epoch 91, bias, value: tensor([ 2.5816e-02, -2.0139e-02, -7.2502e-03, 2.1645e-03, 5.9563e-03, 2.8759e-03, -1.7688e-02, -6.0538e-05, 2.7556e-02, 1.7025e-02], device='cuda:0'), grad: tensor([ 2.8461e-05, 6.8963e-05, 1.7428e-04, -1.7128e-03, -1.3532e-06, 1.0622e-04, -1.6004e-05, -5.2303e-05, 1.3285e-03, 7.4327e-05], device='cuda:0') 100 0.0001 changing lr epoch 90, time 257.27, cls_loss 0.0045 cls_loss_mapping 0.0112 cls_loss_causal 0.6029 re_mapping 0.0095 re_causal 0.0295 /// teacc 98.84 lr 0.00010000 Epoch 92, weight, value: tensor([[ 0.0113, -0.0731, -0.0579, ..., -0.0192, -0.1221, -0.0074], [-0.0607, -0.0989, -0.0925, ..., -0.0785, -0.0325, -0.0103], [ 0.0529, 0.0507, -0.0830, ..., -0.0785, -0.0903, -0.0492], ..., [ 0.0099, 0.0281, -0.0255, ..., -0.0320, -0.0951, 0.0453], [-0.0007, 0.0788, 0.0678, ..., -0.0537, 0.0449, -0.0866], [-0.0449, -0.0119, -0.0104, ..., -0.0161, -0.0482, 0.0405]], device='cuda:0'), grad: tensor([[-3.0875e-04, 4.2245e-06, 3.0613e-04, ..., -4.0799e-05, 2.5582e-04, -5.1148e-06], [ 1.1846e-05, 3.2280e-06, 1.8582e-05, ..., 5.3691e-07, 1.4096e-05, -7.4983e-05], [ 2.5678e-04, 1.9550e-05, 5.4657e-05, ..., 2.2091e-06, 4.2409e-05, 5.7161e-05], ..., [-5.9545e-05, -1.0759e-04, 1.2529e-04, ..., 9.1270e-08, 9.8705e-05, 2.6040e-06], [ 3.0220e-05, -9.9465e-07, 3.0056e-05, ..., 3.8221e-06, 2.1845e-05, 1.3426e-05], [ 8.0466e-05, 7.8380e-05, 3.7044e-05, ..., 8.7591e-07, 2.8268e-05, 4.6819e-05]], device='cuda:0') Epoch 92, bias, value: tensor([ 0.0256, -0.0202, -0.0072, 0.0022, 0.0070, 0.0026, -0.0177, -0.0002, 0.0273, 0.0170], device='cuda:0'), grad: tensor([-4.6492e-04, -4.9353e-05, 6.1893e-04, 4.4250e-04, 1.5572e-05, -1.4229e-03, 4.3154e-04, 4.5061e-05, 1.2141e-04, 2.6250e-04], device='cuda:0') 100 0.0001 changing lr epoch 91, time 257.00, cls_loss 0.0045 cls_loss_mapping 0.0102 cls_loss_causal 0.6137 re_mapping 0.0093 re_causal 0.0289 /// teacc 98.81 lr 0.00010000 Epoch 93, weight, value: tensor([[ 0.0113, -0.0739, -0.0586, ..., -0.0199, -0.1228, -0.0077], [-0.0616, -0.0992, -0.0928, ..., -0.0787, -0.0328, -0.0101], [ 0.0539, 0.0508, -0.0837, ..., -0.0792, -0.0908, -0.0501], ..., [ 0.0101, 0.0281, -0.0264, ..., -0.0322, -0.0956, 0.0457], [-0.0005, 0.0790, 0.0679, ..., -0.0544, 0.0451, -0.0874], [-0.0446, -0.0115, -0.0106, ..., -0.0163, -0.0483, 0.0408]], device='cuda:0'), grad: tensor([[ 1.4007e-06, 9.2387e-06, 1.0490e-05, ..., 1.3404e-05, 1.3746e-05, 7.4692e-07], [ 1.2793e-05, 1.6123e-05, 8.9332e-06, ..., 8.5980e-06, 1.3441e-05, 3.2075e-06], [-1.9252e-05, -7.9721e-06, 3.0130e-05, ..., 1.0110e-05, 5.1618e-05, 1.9088e-05], ..., [-2.9191e-05, -5.7280e-05, 1.3657e-05, ..., 6.5891e-07, 2.5198e-05, -1.0565e-05], [ 2.1890e-05, 1.6257e-05, 8.2552e-06, ..., 4.4741e-06, 1.6093e-05, 6.4969e-06], [ 3.2365e-05, 4.0054e-05, 8.3596e-06, ..., 1.9539e-06, 1.7479e-05, 1.0639e-05]], device='cuda:0') Epoch 93, bias, value: tensor([ 0.0250, -0.0202, -0.0069, 0.0019, 0.0067, 0.0027, -0.0185, 0.0003, 0.0270, 0.0175], device='cuda:0'), grad: tensor([ 4.8071e-05, 7.9393e-05, -1.7196e-05, -1.4818e-04, 1.6057e-04, 4.0114e-05, -3.6526e-04, -4.8995e-05, 8.1897e-05, 1.6904e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 92---------------------------------------------------- epoch 92, time 273.97, cls_loss 0.0064 cls_loss_mapping 0.0119 cls_loss_causal 0.5649 re_mapping 0.0093 re_causal 0.0264 /// teacc 98.99 lr 0.00010000 Epoch 94, weight, value: tensor([[ 0.0122, -0.0732, -0.0591, ..., -0.0195, -0.1232, -0.0071], [-0.0621, -0.0994, -0.0930, ..., -0.0782, -0.0331, -0.0104], [ 0.0545, 0.0511, -0.0841, ..., -0.0806, -0.0911, -0.0506], ..., [ 0.0103, 0.0284, -0.0270, ..., -0.0324, -0.0960, 0.0458], [-0.0018, 0.0788, 0.0683, ..., -0.0550, 0.0455, -0.0885], [-0.0441, -0.0113, -0.0115, ..., -0.0167, -0.0490, 0.0415]], device='cuda:0'), grad: tensor([[-2.8148e-05, 3.6787e-06, 1.0431e-06, ..., 2.5216e-07, 7.7346e-07, 4.1239e-06], [ 4.2289e-05, 2.9564e-05, 1.3551e-06, ..., 6.7055e-08, 1.0254e-06, 3.0726e-05], [-1.0036e-05, -2.0042e-05, 1.9278e-06, ..., 5.4482e-08, 1.0738e-06, 2.1264e-05], ..., [-2.6250e-04, -1.7989e-04, 1.6307e-06, ..., 1.8626e-09, 1.3346e-06, -2.3746e-04], [ 2.5734e-05, 1.7896e-05, -6.7838e-06, ..., 4.3260e-07, -1.7392e-07, 2.2769e-05], [ 5.7966e-05, 6.5006e-06, -2.5049e-05, ..., 6.4727e-08, -3.0756e-05, 5.0485e-05]], device='cuda:0') Epoch 94, bias, value: tensor([ 0.0252, -0.0203, -0.0072, 0.0020, 0.0060, 0.0027, -0.0190, 0.0004, 0.0260, 0.0188], device='cuda:0'), grad: tensor([-7.7248e-05, 8.5592e-05, -4.9919e-05, 2.9063e-04, 2.1243e-04, 4.2289e-05, 1.4670e-05, -5.8174e-04, 1.0294e-04, -4.0084e-05], device='cuda:0') 100 0.0001 changing lr epoch 93, time 256.99, cls_loss 0.0038 cls_loss_mapping 0.0080 cls_loss_causal 0.5764 re_mapping 0.0097 re_causal 0.0288 /// teacc 98.97 lr 0.00010000 Epoch 95, weight, value: tensor([[ 0.0144, -0.0725, -0.0596, ..., -0.0198, -0.1238, -0.0067], [-0.0624, -0.0998, -0.0936, ..., -0.0792, -0.0337, -0.0103], [ 0.0546, 0.0511, -0.0846, ..., -0.0791, -0.0914, -0.0508], ..., [ 0.0105, 0.0286, -0.0274, ..., -0.0324, -0.0965, 0.0460], [-0.0014, 0.0800, 0.0689, ..., -0.0552, 0.0466, -0.0891], [-0.0450, -0.0116, -0.0122, ..., -0.0161, -0.0495, 0.0414]], device='cuda:0'), grad: tensor([[-2.0146e-05, 3.1870e-06, 2.9150e-06, ..., 1.9204e-06, 2.8405e-06, -6.5677e-06], [ 4.3996e-06, 4.4703e-06, 3.7830e-06, ..., 2.6822e-07, 3.4589e-06, -3.7074e-05], [ 1.5810e-05, 1.2875e-05, 1.0170e-05, ..., 5.0338e-07, 8.0764e-06, 8.5309e-06], ..., [ 8.4788e-06, 6.8732e-06, 5.2899e-06, ..., 1.0012e-08, 4.1574e-06, 2.0862e-05], [-6.5148e-05, -7.3195e-05, -5.6654e-05, ..., 1.8021e-06, -4.2975e-05, 3.9339e-06], [ 8.1360e-06, 5.9139e-07, 2.7288e-06, ..., 9.1037e-08, 3.0696e-06, 1.7788e-06]], device='cuda:0') Epoch 95, bias, value: tensor([ 0.0261, -0.0204, -0.0070, 0.0019, 0.0064, 0.0029, -0.0197, 0.0004, 0.0265, 0.0179], device='cuda:0'), grad: tensor([-6.9976e-05, -1.3566e-04, 6.0797e-05, 2.7239e-05, 7.1824e-06, 8.1062e-05, -1.0971e-06, 9.6738e-05, -9.9063e-05, 3.2872e-05], device='cuda:0') 100 0.0001 changing lr epoch 94, time 257.19, cls_loss 0.0046 cls_loss_mapping 0.0116 cls_loss_causal 0.5844 re_mapping 0.0093 re_causal 0.0279 /// teacc 98.94 lr 0.00010000 Epoch 96, weight, value: tensor([[ 0.0143, -0.0731, -0.0609, ..., -0.0203, -0.1257, -0.0071], [-0.0625, -0.1001, -0.0941, ..., -0.0799, -0.0343, -0.0091], [ 0.0548, 0.0511, -0.0852, ..., -0.0797, -0.0914, -0.0513], ..., [ 0.0107, 0.0290, -0.0277, ..., -0.0325, -0.0967, 0.0459], [-0.0010, 0.0807, 0.0696, ..., -0.0558, 0.0472, -0.0900], [-0.0452, -0.0117, -0.0126, ..., -0.0160, -0.0498, 0.0415]], device='cuda:0'), grad: tensor([[-2.1413e-05, 1.5348e-06, 2.1663e-06, ..., 3.8557e-07, 1.6065e-06, 3.1888e-06], [ 2.6338e-06, 2.8759e-06, 3.2131e-06, ..., 1.1176e-07, 2.4047e-06, 9.1612e-05], [-9.3728e-06, -5.4762e-06, 2.3320e-06, ..., 1.3178e-07, 1.8589e-06, -5.4538e-06], ..., [ 2.2706e-06, 4.3679e-07, 1.7751e-06, ..., 9.3132e-09, 1.2433e-06, 2.5436e-05], [-1.3955e-05, -4.2140e-05, -5.5611e-05, ..., 2.5202e-06, -4.5955e-05, 1.0423e-05], [ 1.3299e-05, 4.4480e-06, 2.9635e-06, ..., 8.1956e-08, 2.2296e-06, 2.7037e-04]], device='cuda:0') Epoch 96, bias, value: tensor([ 0.0253, -0.0198, -0.0070, 0.0021, 0.0047, 0.0028, -0.0200, 0.0004, 0.0267, 0.0191], device='cuda:0'), grad: tensor([-1.4794e-04, 2.2864e-04, -1.4052e-05, 2.6435e-05, -1.0004e-03, 2.4036e-05, 1.2338e-04, 6.9439e-05, -1.6183e-05, 7.0620e-04], device='cuda:0') 100 0.0001 changing lr epoch 95, time 256.68, cls_loss 0.0086 cls_loss_mapping 0.0156 cls_loss_causal 0.5567 re_mapping 0.0106 re_causal 0.0284 /// teacc 98.92 lr 0.00010000 Epoch 97, weight, value: tensor([[ 0.0148, -0.0736, -0.0620, ..., -0.0211, -0.1267, -0.0072], [-0.0645, -0.1019, -0.0952, ..., -0.0807, -0.0350, -0.0100], [ 0.0548, 0.0505, -0.0864, ..., -0.0803, -0.0920, -0.0524], ..., [ 0.0118, 0.0301, -0.0281, ..., -0.0326, -0.0971, 0.0472], [-0.0012, 0.0807, 0.0698, ..., -0.0574, 0.0472, -0.0909], [-0.0450, -0.0112, -0.0122, ..., -0.0163, -0.0487, 0.0417]], device='cuda:0'), grad: tensor([[ 5.3132e-07, 5.0813e-06, 8.4639e-06, ..., 3.6228e-06, 6.8992e-06, 1.9744e-07], [ 5.3551e-07, 1.2601e-06, 1.6494e-06, ..., 3.0873e-07, 9.2899e-07, -1.4931e-05], [ 1.5991e-06, 4.2431e-06, 6.0201e-06, ..., 1.8859e-07, 5.0738e-06, 5.0440e-06], ..., [-7.5717e-07, -4.3353e-07, 2.6375e-06, ..., 1.3504e-08, 3.6675e-06, 2.9970e-06], [-1.0647e-05, -1.8880e-05, -1.4782e-05, ..., 3.5241e-06, -1.3173e-05, 1.1623e-06], [ 2.3749e-06, 7.3165e-06, 7.2680e-06, ..., 6.0629e-07, 7.9274e-06, 7.2457e-07]], device='cuda:0') Epoch 97, bias, value: tensor([ 0.0248, -0.0211, -0.0082, 0.0011, 0.0051, 0.0025, -0.0201, 0.0018, 0.0263, 0.0203], device='cuda:0'), grad: tensor([ 1.0841e-05, -2.3291e-05, 1.7405e-05, 1.9923e-05, 7.0781e-06, 2.7996e-06, -2.9474e-05, 1.2688e-05, -3.2961e-05, 1.5058e-05], device='cuda:0') 100 0.0001 changing lr epoch 96, time 256.92, cls_loss 0.0035 cls_loss_mapping 0.0094 cls_loss_causal 0.5791 re_mapping 0.0101 re_causal 0.0300 /// teacc 98.93 lr 0.00010000 Epoch 98, weight, value: tensor([[ 0.0154, -0.0739, -0.0635, ..., -0.0213, -0.1279, -0.0073], [-0.0648, -0.1023, -0.0955, ..., -0.0812, -0.0353, -0.0094], [ 0.0553, 0.0506, -0.0869, ..., -0.0804, -0.0924, -0.0527], ..., [ 0.0118, 0.0302, -0.0285, ..., -0.0326, -0.0974, 0.0470], [-0.0009, 0.0810, 0.0693, ..., -0.0583, 0.0471, -0.0913], [-0.0454, -0.0115, -0.0124, ..., -0.0163, -0.0491, 0.0415]], device='cuda:0'), grad: tensor([[-4.6417e-06, 6.5565e-06, 3.1944e-06, ..., 3.1805e-07, 3.6024e-06, 3.9563e-06], [ 7.1339e-06, 9.2089e-06, 5.7071e-06, ..., 3.7346e-07, 6.4522e-06, -1.7751e-06], [-5.3108e-05, -3.3379e-05, 5.5432e-06, ..., 2.0908e-07, 7.9572e-06, 1.3653e-06], ..., [ 3.0518e-05, 3.7611e-05, 1.7464e-05, ..., 1.3039e-08, 1.8194e-05, -1.7151e-05], [-3.3885e-05, -1.2672e-04, -1.0574e-04, ..., -5.1036e-07, -1.0061e-04, 1.5264e-06], [ 2.0564e-05, 1.7792e-05, 1.3813e-05, ..., 5.6811e-08, 1.3947e-05, 6.9328e-06]], device='cuda:0') Epoch 98, bias, value: tensor([ 0.0248, -0.0207, -0.0079, 0.0008, 0.0062, 0.0024, -0.0192, 0.0016, 0.0259, 0.0192], device='cuda:0'), grad: tensor([-5.9605e-06, 1.9252e-05, -8.2552e-05, 2.2903e-05, 7.1060e-07, 6.2585e-05, 1.1779e-05, 7.8261e-05, -1.6856e-04, 6.1393e-05], device='cuda:0') 100 0.0001 changing lr epoch 97, time 256.86, cls_loss 0.0047 cls_loss_mapping 0.0110 cls_loss_causal 0.5596 re_mapping 0.0089 re_causal 0.0259 /// teacc 98.99 lr 0.00010000 Epoch 99, weight, value: tensor([[ 0.0160, -0.0747, -0.0641, ..., -0.0215, -0.1287, -0.0074], [-0.0642, -0.1016, -0.0960, ..., -0.0815, -0.0359, -0.0083], [ 0.0558, 0.0510, -0.0869, ..., -0.0807, -0.0918, -0.0542], ..., [ 0.0116, 0.0299, -0.0289, ..., -0.0327, -0.0986, 0.0463], [-0.0007, 0.0817, 0.0698, ..., -0.0587, 0.0479, -0.0917], [-0.0452, -0.0111, -0.0128, ..., -0.0163, -0.0492, 0.0421]], device='cuda:0'), grad: tensor([[ 4.3549e-06, 6.0983e-06, 1.8254e-06, ..., 1.0189e-06, 1.9521e-06, 1.0036e-05], [ 4.2468e-06, 5.3160e-06, 2.4941e-06, ..., 1.3318e-07, 2.6468e-06, 5.1707e-06], [-5.3272e-06, -2.7884e-06, 1.2303e-06, ..., 1.3364e-07, 1.9930e-06, 1.3681e-06], ..., [ 1.3304e-04, 1.5509e-04, 1.2740e-06, ..., 1.5367e-08, 2.0396e-06, 3.4547e-04], [-5.5768e-06, -1.1988e-05, -1.2890e-05, ..., 3.3295e-07, -9.4846e-06, 1.5385e-06], [-1.4365e-04, -1.6940e-04, 4.1723e-06, ..., 1.6298e-07, 3.7123e-06, -3.8719e-04]], device='cuda:0') Epoch 99, bias, value: tensor([ 0.0250, -0.0192, -0.0083, 0.0002, 0.0062, 0.0025, -0.0194, 0.0005, 0.0262, 0.0193], device='cuda:0'), grad: tensor([ 3.1471e-05, 3.6985e-05, 1.7416e-04, 7.0073e-06, 3.8815e-04, 7.9691e-05, 1.0356e-05, -2.6703e-04, -5.1968e-06, -4.5586e-04], device='cuda:0') 100 0.0001 changing lr epoch 98, time 256.61, cls_loss 0.0044 cls_loss_mapping 0.0093 cls_loss_causal 0.5607 re_mapping 0.0093 re_causal 0.0274 /// teacc 98.96 lr 0.00010000 Epoch 100, weight, value: tensor([[ 0.0161, -0.0753, -0.0652, ..., -0.0220, -0.1299, -0.0078], [-0.0645, -0.1017, -0.0968, ..., -0.0823, -0.0375, -0.0067], [ 0.0566, 0.0514, -0.0866, ..., -0.0801, -0.0905, -0.0535], ..., [ 0.0116, 0.0300, -0.0295, ..., -0.0327, -0.0989, 0.0453], [-0.0008, 0.0828, 0.0707, ..., -0.0571, 0.0489, -0.0928], [-0.0451, -0.0110, -0.0132, ..., -0.0165, -0.0494, 0.0420]], device='cuda:0'), grad: tensor([[-5.7416e-07, 6.7567e-07, 3.0976e-06, ..., 2.9989e-07, 2.8722e-06, 1.0589e-06], [ 4.5784e-06, 1.0796e-05, 8.1584e-07, ..., 5.2620e-08, 8.6147e-07, 2.3544e-06], [-5.7928e-07, 3.4496e-06, 6.5286e-07, ..., 1.2992e-07, 7.3807e-07, 2.7809e-06], ..., [-2.5019e-05, -6.4790e-05, 7.3994e-07, ..., 1.3970e-09, 4.6892e-07, -2.7344e-05], [ 1.2130e-05, 2.9102e-05, 3.4589e-06, ..., 1.1874e-07, 3.1963e-06, 1.1735e-05], [ 4.4256e-06, 9.7007e-06, 1.0923e-05, ..., 3.0268e-08, 1.0289e-05, 8.0392e-06]], device='cuda:0') Epoch 100, bias, value: tensor([ 0.0249, -0.0186, -0.0071, 0.0006, 0.0059, 0.0027, -0.0200, -0.0004, 0.0263, 0.0193], device='cuda:0'), grad: tensor([ 5.8301e-06, -6.7949e-05, 1.3128e-05, 4.0412e-05, 9.4026e-06, -7.1406e-05, 3.5077e-05, -1.1075e-04, 1.0264e-04, 4.3541e-05], device='cuda:0') 100 0.0001 changing lr epoch 99, time 256.68, cls_loss 0.0039 cls_loss_mapping 0.0095 cls_loss_causal 0.5947 re_mapping 0.0090 re_causal 0.0282 /// teacc 98.83 lr 0.00010000 Epoch 101, weight, value: tensor([[ 0.0167, -0.0754, -0.0666, ..., -0.0219, -0.1311, -0.0081], [-0.0660, -0.1025, -0.0972, ..., -0.0826, -0.0378, -0.0062], [ 0.0576, 0.0517, -0.0874, ..., -0.0811, -0.0913, -0.0538], ..., [ 0.0114, 0.0299, -0.0301, ..., -0.0328, -0.0992, 0.0451], [-0.0008, 0.0830, 0.0707, ..., -0.0570, 0.0491, -0.0939], [-0.0452, -0.0109, -0.0138, ..., -0.0166, -0.0498, 0.0423]], device='cuda:0'), grad: tensor([[-1.3178e-06, 7.0110e-06, 6.7465e-06, ..., 4.9211e-06, 3.5670e-06, 1.1250e-06], [ 6.8638e-07, 1.8328e-06, 6.1095e-07, ..., 3.7160e-07, 9.2527e-07, -1.6401e-06], [ 1.3607e-06, 3.1125e-06, 1.5460e-06, ..., 3.8603e-07, 4.1313e-06, 1.8850e-06], ..., [-1.5926e-06, -5.3644e-06, 2.3609e-07, ..., 1.9092e-08, -7.1526e-07, 6.2492e-07], [ 1.0859e-06, 4.4368e-06, 3.4608e-06, ..., 2.2035e-06, 2.9430e-06, 1.9111e-06], [-2.8219e-06, -3.6396e-06, 1.3225e-06, ..., 4.1956e-07, 1.4193e-06, -5.4091e-06]], device='cuda:0') Epoch 101, bias, value: tensor([ 0.0249, -0.0184, -0.0066, 0.0004, 0.0061, 0.0029, -0.0197, -0.0007, 0.0258, 0.0189], device='cuda:0'), grad: tensor([ 8.7395e-06, -1.5050e-06, 2.4989e-05, -3.3736e-05, 8.3819e-06, 1.9580e-05, -3.6627e-05, 6.0461e-06, 1.1176e-05, -7.0594e-06], device='cuda:0') 100 0.0001 changing lr epoch 100, time 255.70, cls_loss 0.0036 cls_loss_mapping 0.0083 cls_loss_causal 0.6052 re_mapping 0.0092 re_causal 0.0270 /// teacc 98.86 lr 0.00010000 Epoch 102, weight, value: tensor([[ 0.0169, -0.0760, -0.0672, ..., -0.0223, -0.1318, -0.0080], [-0.0669, -0.1029, -0.0977, ..., -0.0831, -0.0382, -0.0061], [ 0.0578, 0.0512, -0.0880, ..., -0.0816, -0.0919, -0.0546], ..., [ 0.0117, 0.0301, -0.0305, ..., -0.0327, -0.0995, 0.0456], [-0.0006, 0.0832, 0.0710, ..., -0.0575, 0.0493, -0.0945], [-0.0451, -0.0104, -0.0140, ..., -0.0168, -0.0500, 0.0425]], device='cuda:0'), grad: tensor([[-9.8944e-06, -2.3916e-06, 3.2373e-06, ..., 1.7723e-06, 1.9241e-06, -3.9898e-06], [ 2.1793e-06, 3.6489e-06, 2.7269e-06, ..., 2.4680e-08, 3.2596e-06, 6.7567e-07], [ 4.8652e-06, 5.5619e-06, 4.1984e-06, ..., 2.9802e-08, 5.2564e-06, 1.3905e-06], ..., [-1.0289e-05, -8.6278e-06, 9.8441e-07, ..., -3.5390e-08, 1.3234e-06, -9.4250e-06], [-3.5353e-06, -7.4357e-06, -1.9912e-06, ..., 2.7334e-07, -1.8030e-06, 1.5087e-06], [ 1.3359e-05, 1.1161e-05, 4.4778e-06, ..., 6.2864e-08, 3.9972e-06, 1.4849e-05]], device='cuda:0') Epoch 102, bias, value: tensor([ 0.0250, -0.0186, -0.0070, 0.0002, 0.0063, 0.0026, -0.0193, -0.0004, 0.0257, 0.0189], device='cuda:0'), grad: tensor([-3.1620e-05, 1.0096e-05, 1.9312e-05, -1.1399e-05, -4.1038e-05, -2.5421e-05, 1.5661e-05, -1.8001e-05, 3.6415e-06, 7.8738e-05], device='cuda:0') 100 0.0001 changing lr epoch 101, time 256.73, cls_loss 0.0056 cls_loss_mapping 0.0105 cls_loss_causal 0.5695 re_mapping 0.0091 re_causal 0.0260 /// teacc 98.94 lr 0.00010000 Epoch 103, weight, value: tensor([[ 1.7405e-02, -7.6313e-02, -6.7860e-02, ..., -2.2697e-02, -1.3247e-01, -7.9666e-03], [-6.8657e-02, -1.0403e-01, -9.8322e-02, ..., -8.4685e-02, -3.8752e-02, -6.2386e-03], [ 5.8280e-02, 5.1710e-02, -8.8969e-02, ..., -8.2148e-02, -9.2701e-02, -5.5193e-02], ..., [ 1.0156e-02, 2.8522e-02, -3.1229e-02, ..., -3.2869e-02, -1.0060e-01, 4.5693e-02], [ 9.1587e-06, 8.3869e-02, 7.1720e-02, ..., -5.8579e-02, 5.0215e-02, -9.5644e-02], [-4.2508e-02, -8.0927e-03, -1.4605e-02, ..., -1.7101e-02, -5.0279e-02, 4.3528e-02]], device='cuda:0'), grad: tensor([[-6.1728e-06, 1.3663e-06, 1.5479e-06, ..., 5.5134e-07, 1.5553e-06, 5.3551e-08], [ 5.6624e-07, 4.8662e-07, 5.6252e-07, ..., 7.2177e-08, 8.0513e-07, -1.3551e-06], [ 1.9930e-06, 2.5686e-06, 2.5164e-06, ..., 2.2165e-07, 3.3956e-06, 9.9652e-08], ..., [ 5.7975e-07, 3.9907e-07, 3.9954e-07, ..., 5.1223e-09, 6.3051e-07, 3.8557e-07], [-4.1202e-06, -6.9328e-06, -4.1015e-06, ..., 7.5251e-07, -5.7220e-06, 2.1048e-07], [ 2.4587e-06, 1.1958e-06, 2.6878e-06, ..., 2.4214e-08, 3.2578e-06, -6.7009e-07]], device='cuda:0') Epoch 103, bias, value: tensor([ 2.5178e-02, -1.9103e-02, -6.9718e-03, 4.0717e-05, 6.0884e-03, 1.9722e-03, -1.8974e-02, -1.1598e-03, 2.5889e-02, 2.0023e-02], device='cuda:0'), grad: tensor([-1.1913e-05, 1.1679e-06, 8.9183e-06, -2.5481e-06, -4.6760e-05, 6.6217e-07, 3.8184e-06, 1.1869e-05, -5.4836e-06, 4.0174e-05], device='cuda:0') 100 0.0001 changing lr epoch 102, time 256.70, cls_loss 0.0035 cls_loss_mapping 0.0079 cls_loss_causal 0.5910 re_mapping 0.0089 re_causal 0.0272 /// teacc 98.92 lr 0.00010000 Epoch 104, weight, value: tensor([[ 0.0177, -0.0773, -0.0692, ..., -0.0253, -0.1336, -0.0085], [-0.0695, -0.1046, -0.0989, ..., -0.0857, -0.0392, -0.0061], [ 0.0585, 0.0518, -0.0895, ..., -0.0824, -0.0932, -0.0555], ..., [ 0.0105, 0.0287, -0.0316, ..., -0.0330, -0.1009, 0.0461], [ 0.0002, 0.0841, 0.0716, ..., -0.0594, 0.0501, -0.0962], [-0.0424, -0.0081, -0.0153, ..., -0.0172, -0.0509, 0.0434]], device='cuda:0'), grad: tensor([[-1.5218e-06, 3.3621e-06, 2.4214e-06, ..., 9.0525e-07, 2.1514e-06, 1.7826e-06], [ 1.3579e-06, 2.1439e-06, 1.7136e-06, ..., 3.1106e-07, 2.2799e-06, -2.7108e-04], [ 4.1500e-06, 6.6943e-06, 4.6007e-06, ..., 3.7299e-07, 5.4650e-06, 4.3064e-06], ..., [ 3.6508e-06, 5.4985e-06, 3.8743e-06, ..., 1.6764e-08, 4.9099e-06, 2.3234e-04], [-3.1114e-05, -4.6581e-05, -2.0593e-05, ..., 9.9558e-07, -2.6599e-05, -9.2238e-06], [ 1.8552e-05, 2.3693e-05, 1.4335e-05, ..., 1.8487e-07, 1.8373e-05, 1.9178e-05]], device='cuda:0') Epoch 104, bias, value: tensor([ 0.0248, -0.0192, -0.0071, -0.0003, 0.0061, 0.0028, -0.0185, -0.0009, 0.0255, 0.0198], device='cuda:0'), grad: tensor([-1.7527e-06, -4.6420e-04, 1.6883e-05, -1.1176e-05, 3.6061e-05, 9.5442e-06, -5.8673e-06, 4.1080e-04, -4.4316e-05, 5.3585e-05], device='cuda:0') 100 0.0001 changing lr epoch 103, time 256.78, cls_loss 0.0045 cls_loss_mapping 0.0084 cls_loss_causal 0.5878 re_mapping 0.0090 re_causal 0.0270 /// teacc 98.92 lr 0.00010000 Epoch 105, weight, value: tensor([[ 0.0190, -0.0777, -0.0698, ..., -0.0269, -0.1333, -0.0088], [-0.0701, -0.1067, -0.0998, ..., -0.0884, -0.0402, -0.0049], [ 0.0586, 0.0519, -0.0902, ..., -0.0821, -0.0938, -0.0562], ..., [ 0.0107, 0.0296, -0.0319, ..., -0.0331, -0.1014, 0.0458], [ 0.0004, 0.0845, 0.0721, ..., -0.0599, 0.0507, -0.0967], [-0.0428, -0.0082, -0.0159, ..., -0.0182, -0.0513, 0.0433]], device='cuda:0'), grad: tensor([[-4.5866e-05, -3.3021e-05, -1.3784e-07, ..., 1.9092e-07, -6.0759e-06, -2.3395e-05], [ 3.4273e-07, 2.8545e-07, 9.2667e-08, ..., -2.1420e-06, 3.1665e-07, -7.7710e-06], [ 1.3802e-06, 9.6764e-07, 7.2177e-08, ..., 1.5507e-07, 4.6333e-07, 1.6121e-06], ..., [-5.6438e-07, -5.9325e-07, 5.7742e-08, ..., 3.2736e-07, -3.5018e-07, 3.1348e-06], [ 2.3656e-06, 1.8319e-06, 1.8161e-07, ..., 1.4296e-07, 6.8359e-07, 1.9725e-06], [ 3.6687e-05, 2.6166e-05, 7.8697e-08, ..., 9.1735e-08, 4.9695e-06, 2.3425e-05]], device='cuda:0') Epoch 105, bias, value: tensor([ 0.0253, -0.0193, -0.0073, -0.0006, 0.0051, 0.0029, -0.0176, -0.0004, 0.0256, 0.0198], device='cuda:0'), grad: tensor([-1.5068e-04, -2.1309e-05, 7.6592e-06, 8.5384e-06, -4.6879e-05, 5.0999e-06, 1.1161e-05, 1.1846e-05, 1.0766e-05, 1.6379e-04], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 104---------------------------------------------------- epoch 104, time 273.50, cls_loss 0.0037 cls_loss_mapping 0.0098 cls_loss_causal 0.5752 re_mapping 0.0088 re_causal 0.0268 /// teacc 99.01 lr 0.00010000 Epoch 106, weight, value: tensor([[ 0.0195, -0.0782, -0.0697, ..., -0.0270, -0.1333, -0.0088], [-0.0708, -0.1078, -0.1009, ..., -0.0895, -0.0416, -0.0046], [ 0.0587, 0.0511, -0.0916, ..., -0.0839, -0.0947, -0.0568], ..., [ 0.0112, 0.0303, -0.0324, ..., -0.0329, -0.1013, 0.0459], [ 0.0003, 0.0850, 0.0725, ..., -0.0602, 0.0515, -0.0972], [-0.0432, -0.0088, -0.0164, ..., -0.0186, -0.0516, 0.0429]], device='cuda:0'), grad: tensor([[ 9.6709e-06, 8.7470e-06, 2.9337e-06, ..., 8.7544e-06, 1.8487e-07, 1.1653e-05], [ 3.8333e-06, 2.3693e-06, 1.9018e-06, ..., 1.6261e-06, 1.5404e-06, 4.8196e-07], [-1.2898e-04, -6.1274e-05, 9.1130e-07, ..., 1.5413e-07, 5.9418e-07, -1.2629e-05], ..., [ 1.2137e-05, 3.9637e-06, -6.6832e-06, ..., -3.3051e-05, 6.4634e-07, -4.3601e-05], [ 1.1988e-05, 6.4634e-06, 1.0349e-05, ..., 4.2096e-06, 9.1642e-06, 2.6822e-06], [ 2.2590e-05, 8.3297e-06, 5.7332e-06, ..., 1.0177e-05, 2.2911e-06, 1.5661e-05]], device='cuda:0') Epoch 106, bias, value: tensor([ 0.0255, -0.0197, -0.0080, -0.0008, 0.0057, 0.0027, -0.0171, 0.0002, 0.0258, 0.0191], device='cuda:0'), grad: tensor([ 3.0667e-05, 8.1509e-06, -2.2018e-04, 1.0735e-04, 1.5721e-06, -7.8976e-06, 1.5840e-05, -5.1826e-05, 4.4167e-05, 7.2122e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 105---------------------------------------------------- epoch 105, time 273.16, cls_loss 0.0034 cls_loss_mapping 0.0088 cls_loss_causal 0.5901 re_mapping 0.0081 re_causal 0.0265 /// teacc 99.02 lr 0.00010000 Epoch 107, weight, value: tensor([[ 0.0200, -0.0789, -0.0702, ..., -0.0277, -0.1334, -0.0090], [-0.0713, -0.1083, -0.1013, ..., -0.0900, -0.0419, -0.0043], [ 0.0596, 0.0518, -0.0926, ..., -0.0837, -0.0953, -0.0562], ..., [ 0.0110, 0.0303, -0.0329, ..., -0.0344, -0.1014, 0.0456], [-0.0002, 0.0847, 0.0721, ..., -0.0609, 0.0509, -0.0980], [-0.0432, -0.0088, -0.0159, ..., -0.0189, -0.0510, 0.0433]], device='cuda:0'), grad: tensor([[-1.5497e-05, 4.7907e-06, 5.9307e-06, ..., 6.1989e-06, 3.7365e-06, -1.9027e-06], [ 2.2482e-06, 1.1735e-06, 2.0675e-06, ..., 9.7789e-07, 1.6270e-06, -1.3085e-06], [-2.1324e-05, -3.1712e-07, 1.2144e-06, ..., 7.9954e-07, -4.7423e-06, 2.6636e-07], ..., [ 1.5616e-05, -1.7779e-06, 1.5451e-06, ..., 1.2433e-06, 5.0440e-06, -8.9593e-07], [ 4.5821e-06, 2.6956e-05, 3.5435e-05, ..., 3.2961e-05, 2.8044e-05, 7.8883e-07], [ 3.9972e-06, 1.0282e-06, 4.5523e-06, ..., 1.6820e-06, 3.7123e-06, 1.2107e-07]], device='cuda:0') Epoch 107, bias, value: tensor([ 0.0256, -0.0195, -0.0077, 0.0003, 0.0060, 0.0023, -0.0171, -0.0001, 0.0245, 0.0191], device='cuda:0'), grad: tensor([-2.4259e-05, 4.6790e-06, -4.7415e-05, 7.5102e-05, -1.8090e-05, -6.4075e-05, -5.9336e-05, 5.1439e-05, 6.1572e-05, 2.0310e-05], device='cuda:0') 100 0.0001 changing lr epoch 106, time 257.01, cls_loss 0.0028 cls_loss_mapping 0.0072 cls_loss_causal 0.5731 re_mapping 0.0083 re_causal 0.0263 /// teacc 99.02 lr 0.00010000 Epoch 108, weight, value: tensor([[ 0.0205, -0.0792, -0.0706, ..., -0.0278, -0.1339, -0.0089], [-0.0716, -0.1086, -0.1016, ..., -0.0906, -0.0423, -0.0041], [ 0.0596, 0.0516, -0.0932, ..., -0.0839, -0.0962, -0.0567], ..., [ 0.0112, 0.0307, -0.0334, ..., -0.0342, -0.1017, 0.0457], [-0.0002, 0.0850, 0.0722, ..., -0.0611, 0.0511, -0.0985], [-0.0434, -0.0090, -0.0165, ..., -0.0190, -0.0515, 0.0432]], device='cuda:0'), grad: tensor([[-1.1601e-05, 7.0296e-06, 7.9572e-06, ..., 3.9004e-06, 5.8487e-06, 1.2992e-06], [ 4.3884e-06, 5.9642e-06, 2.8219e-06, ..., 6.4680e-07, 3.1609e-06, 4.1686e-06], [ 5.2229e-06, 8.6352e-06, 5.3458e-06, ..., 2.4885e-06, 4.2766e-06, 3.3490e-06], ..., [-6.6698e-05, -7.4804e-05, 6.0303e-07, ..., -1.0021e-06, 6.8825e-07, -6.7830e-05], [-3.5673e-05, -2.7359e-05, -7.1228e-06, ..., 2.4498e-05, -1.4782e-05, 9.2527e-07], [ 1.0365e-04, 1.3018e-04, 5.5969e-05, ..., 2.8256e-06, 4.9055e-05, 5.2899e-05]], device='cuda:0') Epoch 108, bias, value: tensor([ 0.0258, -0.0192, -0.0080, 0.0005, 0.0065, 0.0025, -0.0173, -0.0003, 0.0244, 0.0187], device='cuda:0'), grad: tensor([-3.3051e-05, 1.6749e-05, 1.9372e-05, -3.1348e-06, -9.6187e-06, 6.5506e-05, -1.3018e-04, -1.4198e-04, -6.7472e-05, 2.8372e-04], device='cuda:0') 100 0.0001 changing lr epoch 107, time 257.19, cls_loss 0.0053 cls_loss_mapping 0.0116 cls_loss_causal 0.5863 re_mapping 0.0083 re_causal 0.0246 /// teacc 98.93 lr 0.00010000 Epoch 109, weight, value: tensor([[ 0.0211, -0.0779, -0.0709, ..., -0.0273, -0.1344, -0.0089], [-0.0710, -0.1072, -0.1022, ..., -0.0918, -0.0426, -0.0032], [ 0.0600, 0.0516, -0.0938, ..., -0.0843, -0.0969, -0.0570], ..., [ 0.0113, 0.0298, -0.0338, ..., -0.0336, -0.1022, 0.0444], [-0.0009, 0.0843, 0.0724, ..., -0.0614, 0.0510, -0.1004], [-0.0431, -0.0080, -0.0170, ..., -0.0195, -0.0518, 0.0451]], device='cuda:0'), grad: tensor([[ 9.3458e-07, 2.0694e-06, 1.4678e-06, ..., 1.7118e-06, 1.2554e-06, 1.6252e-07], [ 5.7183e-06, 7.6741e-06, 1.4901e-06, ..., 1.6466e-06, 1.4966e-06, 3.6024e-06], [-1.8859e-06, 9.5461e-07, 2.3413e-06, ..., 1.6186e-06, 2.0247e-06, 3.6694e-07], ..., [-7.4431e-06, -9.6112e-06, 9.7230e-07, ..., 1.6680e-06, 5.9418e-07, -7.2531e-06], [-8.4415e-06, -1.0036e-05, -1.2800e-05, ..., 8.8941e-07, -1.7270e-05, 3.2224e-07], [ 2.9821e-06, 9.1419e-06, 1.4566e-05, ..., 8.4415e-06, 6.6124e-06, -1.2591e-06]], device='cuda:0') Epoch 109, bias, value: tensor([ 0.0265, -0.0181, -0.0078, 0.0003, 0.0065, 0.0025, -0.0178, -0.0019, 0.0226, 0.0200], device='cuda:0'), grad: tensor([ 1.0751e-05, 2.7359e-05, 6.8396e-06, 1.6689e-05, -2.6894e-04, 1.2182e-05, 1.2505e-04, -1.6227e-05, -8.9034e-06, 9.5189e-05], device='cuda:0') 100 0.0001 changing lr epoch 108, time 256.96, cls_loss 0.0054 cls_loss_mapping 0.0108 cls_loss_causal 0.5704 re_mapping 0.0086 re_causal 0.0242 /// teacc 98.95 lr 0.00010000 Epoch 110, weight, value: tensor([[ 0.0232, -0.0784, -0.0715, ..., -0.0268, -0.1349, -0.0090], [-0.0723, -0.1080, -0.1031, ..., -0.0923, -0.0435, -0.0032], [ 0.0596, 0.0522, -0.0947, ..., -0.0846, -0.0973, -0.0568], ..., [ 0.0117, 0.0303, -0.0344, ..., -0.0333, -0.1026, 0.0444], [-0.0003, 0.0851, 0.0731, ..., -0.0616, 0.0520, -0.1007], [-0.0434, -0.0080, -0.0183, ..., -0.0212, -0.0533, 0.0453]], device='cuda:0'), grad: tensor([[-4.6566e-10, 8.5235e-06, 9.0748e-06, ..., 4.7870e-07, 8.4415e-06, -1.0133e-06], [ 6.1952e-06, 7.4320e-06, 7.5214e-06, ..., 4.7032e-08, 8.4788e-06, -6.4913e-07], [ 2.7046e-05, 3.4004e-05, 3.4034e-05, ..., 7.9162e-08, 3.1978e-05, 2.0470e-06], ..., [-9.0823e-06, -8.6427e-06, 6.6869e-06, ..., 2.7940e-09, 8.5011e-06, -1.4223e-05], [-8.8811e-05, -1.1885e-04, -1.3411e-04, ..., 3.5530e-07, -1.2827e-04, 1.2498e-06], [ 1.3769e-05, 1.1198e-05, 1.0908e-05, ..., 4.8429e-08, 1.2711e-05, 5.0217e-06]], device='cuda:0') Epoch 110, bias, value: tensor([ 0.0248, -0.0191, -0.0057, -0.0003, 0.0069, 0.0032, -0.0197, -0.0018, 0.0231, 0.0212], device='cuda:0'), grad: tensor([-2.0579e-05, 1.2346e-05, 8.1837e-05, 2.5406e-05, -2.1684e-04, 7.5758e-05, 2.7150e-05, -2.4483e-05, -2.5010e-04, 2.8944e-04], device='cuda:0') 100 0.0001 changing lr epoch 109, time 256.77, cls_loss 0.0035 cls_loss_mapping 0.0094 cls_loss_causal 0.5697 re_mapping 0.0083 re_causal 0.0254 /// teacc 98.97 lr 0.00010000 Epoch 111, weight, value: tensor([[ 0.0240, -0.0788, -0.0723, ..., -0.0270, -0.1357, -0.0089], [-0.0733, -0.1082, -0.1039, ..., -0.0931, -0.0438, -0.0036], [ 0.0602, 0.0523, -0.0958, ..., -0.0849, -0.0977, -0.0573], ..., [ 0.0120, 0.0310, -0.0348, ..., -0.0332, -0.1031, 0.0454], [-0.0002, 0.0850, 0.0734, ..., -0.0622, 0.0523, -0.1017], [-0.0438, -0.0086, -0.0188, ..., -0.0215, -0.0538, 0.0444]], device='cuda:0'), grad: tensor([[-1.2191e-06, 1.2983e-06, 7.2271e-07, ..., 1.9837e-07, 6.7661e-07, -1.2945e-07], [ 3.1060e-07, 8.1118e-07, 2.8312e-07, ..., 7.3574e-08, 2.8312e-07, -2.5649e-06], [ 7.9907e-07, 1.1325e-06, 1.1632e-06, ..., 6.3796e-08, 1.1362e-06, 4.2841e-07], ..., [-1.5460e-07, -1.7649e-07, 1.1083e-07, ..., 1.8626e-09, 1.1595e-07, 6.3982e-07], [-3.7085e-06, -3.4682e-06, -4.5151e-06, ..., 4.3586e-07, -4.3884e-06, 6.1514e-07], [ 1.3197e-06, -1.5823e-06, 8.7963e-07, ..., 1.4435e-08, 8.5449e-07, -8.5495e-07]], device='cuda:0') Epoch 111, bias, value: tensor([ 0.0250, -0.0193, -0.0058, -0.0006, 0.0077, 0.0029, -0.0191, -0.0009, 0.0226, 0.0201], device='cuda:0'), grad: tensor([-2.4959e-06, -4.9099e-06, 4.3772e-06, 2.9281e-06, -1.3947e-05, 3.2969e-06, 1.0729e-05, 3.4086e-06, -8.5980e-06, 5.1968e-06], device='cuda:0') 100 0.0001 changing lr epoch 110, time 255.11, cls_loss 0.0028 cls_loss_mapping 0.0067 cls_loss_causal 0.5605 re_mapping 0.0078 re_causal 0.0245 /// teacc 98.98 lr 0.00010000 Epoch 112, weight, value: tensor([[ 0.0245, -0.0792, -0.0732, ..., -0.0265, -0.1366, -0.0092], [-0.0736, -0.1086, -0.1044, ..., -0.0944, -0.0445, -0.0035], [ 0.0604, 0.0523, -0.0961, ..., -0.0850, -0.0982, -0.0576], ..., [ 0.0123, 0.0314, -0.0351, ..., -0.0330, -0.1032, 0.0456], [-0.0003, 0.0853, 0.0736, ..., -0.0625, 0.0523, -0.1021], [-0.0441, -0.0087, -0.0192, ..., -0.0217, -0.0542, 0.0442]], device='cuda:0'), grad: tensor([[ 1.5097e-06, 3.1255e-06, 4.4219e-06, ..., 1.4389e-06, 4.4666e-06, 2.3935e-07], [ 3.0864e-06, 4.9435e-06, 3.4086e-06, ..., 2.8405e-07, 3.5334e-06, -1.0999e-06], [-9.7789e-07, 1.3420e-06, 2.1383e-06, ..., 3.7719e-07, 3.4850e-06, 1.1502e-06], ..., [-1.0990e-05, -2.2769e-05, 1.8254e-06, ..., 1.4435e-08, 1.8422e-06, -9.6858e-06], [-1.3039e-07, -1.0384e-06, 2.1592e-05, ..., 6.8545e-07, 1.8314e-05, 1.0366e-06], [ 5.3532e-06, 7.0296e-06, 3.0044e-06, ..., 1.0198e-07, 2.8703e-06, 3.6675e-06]], device='cuda:0') Epoch 112, bias, value: tensor([ 0.0253, -0.0194, -0.0059, -0.0002, 0.0076, 0.0033, -0.0193, -0.0007, 0.0222, 0.0199], device='cuda:0'), grad: tensor([ 1.0759e-05, 5.9903e-06, 2.3562e-06, 2.8923e-05, 3.6061e-06, -2.6155e-04, 1.9228e-04, -2.8670e-05, 3.3051e-05, 1.3657e-05], device='cuda:0') 100 0.0001 changing lr ---------------------saving model at epoch 111---------------------------------------------------- epoch 111, time 273.89, cls_loss 0.0029 cls_loss_mapping 0.0077 cls_loss_causal 0.5681 re_mapping 0.0080 re_causal 0.0244 /// teacc 99.08 lr 0.00010000 Epoch 113, weight, value: tensor([[ 0.0253, -0.0790, -0.0737, ..., -0.0269, -0.1370, -0.0092], [-0.0739, -0.1084, -0.1053, ..., -0.0949, -0.0454, -0.0032], [ 0.0603, 0.0520, -0.0972, ..., -0.0851, -0.0994, -0.0580], ..., [ 0.0125, 0.0315, -0.0355, ..., -0.0330, -0.1034, 0.0455], [ 0.0002, 0.0861, 0.0744, ..., -0.0628, 0.0533, -0.1024], [-0.0443, -0.0087, -0.0188, ..., -0.0215, -0.0540, 0.0441]], device='cuda:0'), grad: tensor([[-5.1744e-06, 2.6729e-06, 1.9148e-06, ..., 1.4920e-06, 1.7341e-06, 8.4471e-07], [ 2.8666e-06, 4.3027e-06, 1.1884e-06, ..., 8.4564e-07, 1.2415e-06, -7.2597e-07], [-3.6247e-06, -1.5432e-06, 1.1381e-06, ..., 7.0082e-07, 1.1399e-06, 1.1874e-06], ..., [ 1.9446e-06, 1.7062e-06, 1.9697e-07, ..., 4.7497e-08, 2.3702e-07, 1.9837e-06], [-3.0641e-06, -3.8780e-06, -3.5856e-06, ..., 3.8408e-06, -2.4829e-06, 7.4413e-07], [-3.7868e-06, -1.2405e-05, 6.5863e-06, ..., 1.8300e-07, 5.5507e-06, -1.8224e-05]], device='cuda:0') Epoch 113, bias, value: tensor([ 0.0259, -0.0190, -0.0066, 0.0002, 0.0076, 0.0026, -0.0194, -0.0008, 0.0227, 0.0196], device='cuda:0'), grad: tensor([-1.0170e-05, 9.3952e-06, -2.1402e-06, 9.8497e-06, 4.5836e-05, 2.2203e-05, -2.0653e-05, 2.0877e-05, 7.6462e-07, -7.5996e-05], device='cuda:0') 100 0.0001 changing lr epoch 112, time 256.48, cls_loss 0.0034 cls_loss_mapping 0.0073 cls_loss_causal 0.5579 re_mapping 0.0082 re_causal 0.0245 /// teacc 98.97 lr 0.00010000 Epoch 114, weight, value: tensor([[ 0.0267, -0.0787, -0.0744, ..., -0.0265, -0.1378, -0.0090], [-0.0745, -0.1092, -0.1061, ..., -0.0958, -0.0462, -0.0031], [ 0.0608, 0.0518, -0.0983, ..., -0.0859, -0.0998, -0.0582], ..., [ 0.0126, 0.0317, -0.0361, ..., -0.0329, -0.1040, 0.0457], [ 0.0003, 0.0865, 0.0746, ..., -0.0633, 0.0536, -0.1029], [-0.0441, -0.0084, -0.0194, ..., -0.0216, -0.0546, 0.0442]], device='cuda:0'), grad: tensor([[ 1.5413e-07, 3.0436e-06, 2.2370e-06, ..., 1.4259e-06, 1.9390e-06, 1.0580e-06], [ 4.6333e-07, 5.5246e-06, 1.2666e-06, ..., 8.4750e-07, 1.3672e-06, 1.8617e-06], [-1.9912e-06, 4.0652e-07, 1.4268e-06, ..., 1.0254e-06, 1.8189e-06, 3.1432e-07], ..., [ 4.0047e-08, 1.7835e-06, 1.1865e-06, ..., 9.6392e-08, 1.3309e-06, 1.7276e-06], [ 1.1669e-06, 3.3583e-06, 4.4443e-06, ..., 1.2796e-06, 4.3549e-06, 1.7593e-06], [ 2.8498e-07, -5.2094e-05, 1.4184e-06, ..., -4.7088e-06, 1.4585e-06, -2.9936e-05]], device='cuda:0') Epoch 114, bias, value: tensor([ 0.0267, -0.0193, -0.0066, 0.0005, 0.0072, 0.0026, -0.0192, -0.0008, 0.0225, 0.0194], device='cuda:0'), grad: tensor([ 6.1132e-06, 1.2338e-05, 8.7991e-06, 7.0520e-06, 1.1700e-04, -8.4490e-06, -2.2978e-05, 9.9540e-06, 1.1787e-05, -1.4162e-04], device='cuda:0') 100 0.0001 changing lr epoch 113, time 256.84, cls_loss 0.0031 cls_loss_mapping 0.0083 cls_loss_causal 0.5639 re_mapping 0.0081 re_causal 0.0242 /// teacc 99.06 lr 0.00010000 Epoch 115, weight, value: tensor([[ 0.0270, -0.0790, -0.0753, ..., -0.0268, -0.1389, -0.0088], [-0.0753, -0.1103, -0.1072, ..., -0.0960, -0.0475, -0.0031], [ 0.0623, 0.0522, -0.0990, ..., -0.0863, -0.0998, -0.0587], ..., [ 0.0130, 0.0324, -0.0364, ..., -0.0319, -0.1043, 0.0459], [ 0.0004, 0.0874, 0.0752, ..., -0.0639, 0.0545, -0.1031], [-0.0443, -0.0087, -0.0199, ..., -0.0221, -0.0547, 0.0441]], device='cuda:0'), grad: tensor([[-2.2613e-06, 5.0804e-07, 1.8412e-06, ..., 6.7661e-07, 1.7146e-06, -2.4531e-06], [ 5.4622e-07, 8.9081e-07, 8.0001e-07, ..., 1.1455e-07, 6.8918e-07, -6.3982e-07], [ 9.3132e-08, 3.7905e-07, 5.6531e-07, ..., 1.2852e-07, 5.2107e-07, 5.2201e-07], ..., [-1.4501e-06, -3.4161e-06, 8.2003e-07, ..., 1.6810e-07, 7.2224e-07, -1.7732e-06], [ 6.5267e-06, 2.6226e-06, 2.6673e-05, ..., 1.0751e-05, 2.8729e-05, 2.4680e-07], [ 1.8496e-06, 1.6764e-06, 1.9241e-06, ..., 4.5681e-07, 1.7472e-06, 2.3879e-06]], device='cuda:0') Epoch 115, bias, value: tensor([ 0.0269, -0.0196, -0.0060, -0.0002, 0.0071, 0.0026, -0.0195, -0.0002, 0.0230, 0.0189], device='cuda:0'), grad: tensor([-2.5406e-05, 2.4959e-07, 2.2911e-06, 4.4852e-05, -1.0002e-06, -1.0097e-04, 2.4125e-05, -1.5656e-06, 3.7283e-05, 1.9982e-05], device='cuda:0') 100 0.0001 changing lr epoch 114, time 257.04, cls_loss 0.0032 cls_loss_mapping 0.0070 cls_loss_causal 0.5699 re_mapping 0.0079 re_causal 0.0250 /// teacc 98.99 lr 0.00010000 Epoch 116, weight, value: tensor([[ 0.0277, -0.0787, -0.0762, ..., -0.0265, -0.1399, -0.0090], [-0.0760, -0.1111, -0.1090, ..., -0.0962, -0.0496, -0.0028], [ 0.0625, 0.0516, -0.0999, ..., -0.0872, -0.1005, -0.0591], ..., [ 0.0135, 0.0330, -0.0367, ..., -0.0312, -0.1048, 0.0461], [ 0.0002, 0.0876, 0.0751, ..., -0.0649, 0.0551, -0.1036], [-0.0448, -0.0090, -0.0204, ..., -0.0224, -0.0552, 0.0440]], device='cuda:0'), grad: tensor([[-3.9786e-05, 1.7583e-06, -6.7987e-06, ..., 1.0524e-06, 1.5870e-06, -5.5693e-06], [ 2.1253e-06, 1.4678e-06, 1.1828e-06, ..., 3.0268e-07, 1.3309e-06, 4.2026e-07], [-1.8077e-06, -1.6401e-06, 1.5441e-06, ..., 2.7684e-07, 1.0626e-06, 3.7253e-07], ..., [ 2.2110e-06, 1.0282e-06, 3.1572e-07, ..., -2.2305e-07, 1.0859e-06, 1.5623e-07], [ 2.2929e-06, -1.7853e-06, -1.8533e-07, ..., 1.1390e-06, -2.6673e-06, 5.7137e-07], [ 2.5362e-05, 4.0256e-07, 6.5342e-06, ..., 1.5623e-07, 1.1465e-06, 3.2410e-06]], device='cuda:0') Epoch 116, bias, value: tensor([ 2.7209e-02, -1.9717e-02, -6.2022e-03, -5.1442e-04, 7.5631e-03, 1.9853e-03, -1.8267e-02, 7.6748e-05, 2.2704e-02, 1.8295e-02], device='cuda:0'), grad: tensor([-1.1653e-04, 8.9854e-06, 6.8918e-07, -3.3192e-06, -6.4224e-06, 1.6972e-05, -4.1886e-07, 1.0379e-05, 8.8289e-06, 8.0764e-05], device='cuda:0') 100 0.0001 changing lr epoch 115, time 256.70, cls_loss 0.0032 cls_loss_mapping 0.0086 cls_loss_causal 0.5699 re_mapping 0.0082 re_causal 0.0250 /// teacc 98.97 lr 0.00010000 Epoch 117, weight, value: tensor([[ 0.0276, -0.0796, -0.0773, ..., -0.0270, -0.1417, -0.0088], [-0.0766, -0.1112, -0.1098, ..., -0.0965, -0.0502, -0.0027], [ 0.0624, 0.0512, -0.1013, ..., -0.0878, -0.1020, -0.0597], ..., [ 0.0130, 0.0329, -0.0388, ..., -0.0310, -0.1059, 0.0460], [ 0.0011, 0.0882, 0.0752, ..., -0.0653, 0.0553, -0.1041], [-0.0448, -0.0086, -0.0205, ..., -0.0227, -0.0551, 0.0443]], device='cuda:0'), grad: tensor([[ 6.4122e-07, 6.0629e-07, 4.5425e-07, ..., 2.1420e-07, 3.7882e-07, 1.1735e-07], [ 4.3120e-07, 6.1700e-07, 2.8429e-07, ..., 1.8394e-08, 3.0850e-07, -7.7998e-07], [-1.3458e-06, -5.2946e-07, 4.8010e-07, ..., 2.0489e-08, 6.0257e-07, 5.6205e-07], ..., [-7.6462e-07, -1.5926e-06, 5.6764e-07, ..., -1.1642e-09, 7.2783e-07, -1.3607e-06], [ 3.6950e-07, 4.4936e-08, 1.1791e-06, ..., 1.3015e-07, 9.2946e-07, 1.8044e-07], [ 8.6520e-07, 6.4960e-07, 6.5938e-07, ..., 1.3271e-08, 7.9582e-07, 6.1002e-07]], device='cuda:0') Epoch 117, bias, value: tensor([ 0.0268, -0.0197, -0.0067, 0.0008, 0.0077, 0.0021, -0.0187, -0.0004, 0.0229, 0.0187], device='cuda:0'), grad: tensor([ 2.1700e-06, -1.3579e-06, -1.1157e-06, -3.1050e-06, 2.6613e-07, -1.5767e-06, -5.2620e-08, -4.6939e-07, 2.5351e-06, 2.6990e-06], device='cuda:0') 100 0.0001 changing lr epoch 116, time 257.28, cls_loss 0.0029 cls_loss_mapping 0.0064 cls_loss_causal 0.5882 re_mapping 0.0080 re_causal 0.0249 /// teacc 98.96 lr 0.00010000 Epoch 118, weight, value: tensor([[ 0.0278, -0.0799, -0.0782, ..., -0.0273, -0.1424, -0.0089], [-0.0771, -0.1117, -0.1100, ..., -0.0966, -0.0503, -0.0032], [ 0.0626, 0.0509, -0.1018, ..., -0.0880, -0.1024, -0.0603], ..., [ 0.0134, 0.0335, -0.0393, ..., -0.0307, -0.1061, 0.0466], [ 0.0019, 0.0890, 0.0763, ..., -0.0658, 0.0563, -0.1045], [-0.0449, -0.0087, -0.0210, ..., -0.0227, -0.0556, 0.0451]], device='cuda:0'), grad: tensor([[-1.1779e-05, 1.0384e-06, 2.1756e-05, ..., 3.4925e-07, 1.6272e-05, 3.8650e-07], [ 6.6683e-07, 5.3551e-07, 2.0027e-05, ..., 6.2864e-08, 1.4983e-05, -2.4140e-06], [-3.2559e-06, -3.0361e-06, 1.4558e-05, ..., 2.1537e-07, 1.0908e-05, 4.1770e-07], ..., [ 1.6168e-06, 1.2014e-06, 2.4140e-05, ..., 8.4052e-08, 1.8045e-05, 1.3392e-06], [ 2.4475e-06, 2.4531e-06, 1.5628e-04, ..., 4.5006e-07, 1.1635e-04, 9.6485e-07], [ 7.0855e-06, -3.6769e-06, 2.6464e-05, ..., -2.0047e-07, 1.9699e-05, -3.3099e-06]], device='cuda:0') Epoch 118, bias, value: tensor([ 2.6891e-02, -2.0213e-02, -6.7734e-03, 8.9989e-05, 7.5903e-03, 1.3307e-03, -1.8470e-02, 5.3537e-04, 2.3243e-02, 1.8801e-02], device='cuda:0'), grad: tensor([-1.0617e-06, 2.5317e-05, 1.6421e-05, 1.2236e-03, 1.0014e-05, -1.7233e-03, 1.1778e-04, 4.3124e-05, 2.3699e-04, 5.2124e-05], device='cuda:0') 100 0.0001 changing lr epoch 117, time 256.93, cls_loss 0.0031 cls_loss_mapping 0.0067 cls_loss_causal 0.5500 re_mapping 0.0078 re_causal 0.0241 /// teacc 98.91 lr 0.00010000 Epoch 119, weight, value: tensor([[ 0.0269, -0.0821, -0.0780, ..., -0.0279, -0.1430, -0.0090], [-0.0776, -0.1122, -0.1103, ..., -0.0968, -0.0505, -0.0032], [ 0.0628, 0.0509, -0.1022, ..., -0.0881, -0.1029, -0.0609], ..., [ 0.0135, 0.0339, -0.0398, ..., -0.0301, -0.1068, 0.0468], [ 0.0021, 0.0889, 0.0758, ..., -0.0672, 0.0559, -0.1050], [-0.0442, -0.0076, -0.0217, ..., -0.0231, -0.0563, 0.0450]], device='cuda:0'), grad: tensor([[ 1.1558e-06, 4.6007e-06, 1.8608e-06, ..., 7.0687e-07, 3.2894e-06, 7.4273e-08], [ 1.1241e-06, 6.0648e-06, 3.0156e-06, ..., 3.0790e-06, 4.8280e-06, -1.0142e-06], [-2.3976e-05, -3.2485e-05, 1.7211e-06, ..., 5.7882e-07, -4.8988e-06, 7.0175e-07], ..., [ 1.6950e-06, 2.6412e-06, 8.9314e-07, ..., 2.8405e-08, 1.9576e-06, 2.3982e-08], [-1.2048e-05, -6.4075e-05, -5.2959e-05, ..., 1.7826e-06, -9.6679e-05, 2.1025e-07], [ 8.9640e-07, 2.5821e-07, 5.8580e-07, ..., 1.1222e-07, 8.4797e-07, -4.1421e-07]], device='cuda:0') Epoch 119, bias, value: tensor([ 0.0264, -0.0199, -0.0073, 0.0003, 0.0074, 0.0014, -0.0182, 0.0006, 0.0227, 0.0192], device='cuda:0'), grad: tensor([ 7.0669e-06, 1.0587e-05, -6.8307e-05, 5.1975e-05, 4.2608e-08, 9.6738e-05, -1.1072e-05, 7.9870e-06, -9.8288e-05, 3.2261e-06], device='cuda:0') 100 0.0001 changing lr epoch 118, time 257.14, cls_loss 0.0030 cls_loss_mapping 0.0060 cls_loss_causal 0.5426 re_mapping 0.0078 re_causal 0.0238 /// teacc 99.01 lr 0.00010000 Epoch 120, weight, value: tensor([[ 0.0291, -0.0799, -0.0756, ..., -0.0230, -0.1395, -0.0076], [-0.0779, -0.1121, -0.1106, ..., -0.0971, -0.0508, -0.0028], [ 0.0630, 0.0511, -0.1027, ..., -0.0886, -0.1034, -0.0609], ..., [ 0.0134, 0.0339, -0.0401, ..., -0.0302, -0.1073, 0.0468], [ 0.0020, 0.0885, 0.0753, ..., -0.0691, 0.0556, -0.1055], [-0.0454, -0.0082, -0.0220, ..., -0.0259, -0.0561, 0.0449]], device='cuda:0'), grad: tensor([[-2.4125e-05, 5.3465e-05, 7.1287e-05, ..., 1.1867e-04, 1.5283e-04, 5.9465e-07], [ 8.9407e-07, 1.3513e-06, 1.7080e-06, ..., 2.2128e-06, 3.8855e-06, 8.9407e-08], [ 1.3024e-05, -1.8403e-06, 1.5320e-06, ..., 1.8887e-06, 4.1611e-06, 2.1863e-07], ..., [ 1.2089e-06, -9.3132e-09, 9.1409e-07, ..., 3.6764e-07, 1.7369e-06, -8.7777e-08], [ 1.1241e-06, 5.3532e-06, 9.2015e-06, ..., 8.8587e-06, 1.0863e-05, 3.5693e-07], [ 4.1984e-06, 1.8487e-06, 1.3858e-05, ..., 4.0606e-06, 1.4447e-05, 5.9279e-07]], device='cuda:0') Epoch 120, bias, value: tensor([ 0.0284, -0.0196, -0.0076, -0.0009, 0.0071, 0.0021, -0.0187, 0.0005, 0.0220, 0.0183], device='cuda:0'), grad: tensor([ 2.8205e-04, 1.2092e-05, 3.6716e-05, -1.5236e-05, 8.7619e-06, 4.6754e-04, -8.7547e-04, 8.1733e-06, 2.7046e-05, 4.8399e-05], device='cuda:0') 100 0.0001 changing lr epoch 119, time 257.04, cls_loss 0.0027 cls_loss_mapping 0.0075 cls_loss_causal 0.5665 re_mapping 0.0077 re_causal 0.0247 /// teacc 98.83 lr 0.00010000 Epoch 121, weight, value: tensor([[ 0.0295, -0.0799, -0.0762, ..., -0.0234, -0.1402, -0.0077], [-0.0783, -0.1128, -0.1117, ..., -0.0975, -0.0520, -0.0021], [ 0.0630, 0.0513, -0.1036, ..., -0.0888, -0.1043, -0.0612], ..., [ 0.0136, 0.0341, -0.0406, ..., -0.0297, -0.1081, 0.0462], [ 0.0022, 0.0896, 0.0759, ..., -0.0694, 0.0565, -0.1059], [-0.0457, -0.0086, -0.0227, ..., -0.0262, -0.0557, 0.0453]], device='cuda:0'), grad: tensor([[-3.0752e-06, 2.3339e-06, 1.3532e-06, ..., 4.8429e-07, 1.1418e-06, 3.2596e-09], [ 9.4175e-06, 2.9191e-05, 1.3374e-05, ..., 5.9158e-06, 8.4415e-06, 1.3746e-05], [ 2.5053e-06, 7.1116e-06, 7.5176e-06, ..., 1.9260e-06, 6.9737e-06, 1.2545e-06], ..., [ 7.3016e-06, 2.5481e-05, 1.2601e-06, ..., 3.7136e-07, 1.0906e-06, 3.6091e-05], [-1.3679e-05, -5.0157e-05, -2.2516e-05, ..., -1.5013e-05, -7.7561e-06, -3.1367e-06], [-7.6294e-06, -3.7432e-05, 1.4240e-06, ..., 1.5926e-07, 1.5292e-06, -5.5075e-05]], device='cuda:0') Epoch 121, bias, value: tensor([ 2.8732e-02, -1.9494e-02, -7.7390e-03, -1.5409e-03, 7.1881e-03, 2.3704e-03, -1.8773e-02, 4.5337e-05, 2.2548e-02, 1.8267e-02], device='cuda:0'), grad: tensor([-6.5491e-06, 4.8429e-05, 1.7241e-05, -3.4302e-05, 5.8599e-06, 4.2170e-06, 2.9862e-05, 6.1214e-05, -4.6462e-05, -7.9632e-05], device='cuda:0') 100 0.0001 changing lr epoch 120, time 257.02, cls_loss 0.0037 cls_loss_mapping 0.0070 cls_loss_causal 0.5394 re_mapping 0.0080 re_causal 0.0239 /// teacc 98.85 lr 0.00010000 Epoch 122, weight, value: tensor([[ 0.0304, -0.0795, -0.0772, ..., -0.0222, -0.1397, -0.0077], [-0.0805, -0.1151, -0.1123, ..., -0.0980, -0.0525, -0.0039], [ 0.0626, 0.0508, -0.1045, ..., -0.0890, -0.1052, -0.0618], ..., [ 0.0149, 0.0359, -0.0411, ..., -0.0292, -0.1086, 0.0482], [ 0.0028, 0.0906, 0.0760, ..., -0.0699, 0.0572, -0.1066], [-0.0460, -0.0088, -0.0233, ..., -0.0263, -0.0558, 0.0452]], device='cuda:0'), grad: tensor([[ 1.5059e-06, 1.3635e-06, 1.0198e-06, ..., 2.6124e-07, 1.0030e-06, 6.4401e-07], [ 1.6503e-06, 1.5441e-06, 1.0654e-06, ..., 1.4179e-07, 1.1194e-06, -1.7034e-06], [-1.2077e-05, -2.6245e-06, 1.3886e-06, ..., 2.2072e-07, 1.7574e-06, 4.9314e-07], ..., [ 1.3802e-06, 2.1104e-06, 1.2042e-06, ..., 6.2864e-09, 2.0936e-06, 2.7400e-06], [ 6.8471e-06, 2.0098e-06, -9.2667e-08, ..., 7.3342e-08, -6.0070e-08, 2.2780e-06], [-2.8219e-06, -8.4043e-06, 1.1735e-06, ..., 1.5600e-08, -9.1316e-07, -7.7188e-06]], device='cuda:0') Epoch 122, bias, value: tensor([ 0.0294, -0.0206, -0.0086, -0.0022, 0.0071, 0.0033, -0.0199, 0.0017, 0.0227, 0.0178], device='cuda:0'), grad: tensor([ 6.0573e-06, 7.0594e-07, -1.7151e-05, -8.0988e-06, -3.9935e-05, 7.8836e-07, 4.4733e-05, 9.8646e-06, 1.7598e-05, -1.4536e-05], device='cuda:0') 100 0.0001 changing lr epoch 121, time 255.30, cls_loss 0.0031 cls_loss_mapping 0.0050 cls_loss_causal 0.5733 re_mapping 0.0072 re_causal 0.0240 /// teacc 98.91 lr 0.00010000 Epoch 123, weight, value: tensor([[ 0.0304, -0.0798, -0.0781, ..., -0.0226, -0.1404, -0.0080], [-0.0812, -0.1167, -0.1142, ..., -0.1008, -0.0547, -0.0041], [ 0.0659, 0.0541, -0.1050, ..., -0.0892, -0.1056, -0.0608], ..., [ 0.0130, 0.0340, -0.0416, ..., -0.0290, -0.1089, 0.0482], [ 0.0026, 0.0901, 0.0758, ..., -0.0711, 0.0568, -0.1079], [-0.0462, -0.0087, -0.0236, ..., -0.0263, -0.0562, 0.0450]], device='cuda:0'), grad: tensor([[ 3.2093e-06, 2.6107e-05, 2.6971e-05, ..., 2.0206e-05, 2.0102e-05, 7.3165e-06], [ 4.0093e-07, 1.6866e-06, 2.1327e-06, ..., 2.6217e-07, 3.5223e-06, -1.3500e-05], [-1.0185e-05, -1.0431e-05, 3.8184e-06, ..., 8.0233e-07, 5.9456e-06, 1.0267e-05], ..., [-9.3970e-07, -5.5656e-06, 2.6021e-06, ..., 1.7323e-06, 1.7295e-06, 4.2915e-06], [ 4.5486e-06, 8.7097e-06, 6.6683e-06, ..., 2.4475e-06, 5.4911e-06, 1.2089e-06], [-8.7637e-07, -6.2361e-06, 5.4277e-06, ..., -3.6769e-06, 6.7130e-06, -7.7635e-06]], device='cuda:0') Epoch 123, bias, value: tensor([ 0.0292, -0.0214, -0.0061, -0.0026, 0.0088, 0.0034, -0.0185, 0.0007, 0.0219, 0.0168], device='cuda:0'), grad: tensor([ 6.5327e-05, -2.1291e-04, 1.3638e-04, -9.4771e-05, 7.5251e-06, 3.7718e-04, -3.6025e-04, 4.6968e-05, 3.2455e-05, 1.8384e-06], device='cuda:0') 100 0.0001 changing lr epoch 122, time 256.79, cls_loss 0.0033 cls_loss_mapping 0.0054 cls_loss_causal 0.5703 re_mapping 0.0075 re_causal 0.0238 /// teacc 98.95 lr 0.00010000 Epoch 124, weight, value: tensor([[ 0.0304, -0.0800, -0.0787, ..., -0.0228, -0.1407, -0.0082], [-0.0818, -0.1172, -0.1147, ..., -0.1011, -0.0550, -0.0041], [ 0.0668, 0.0547, -0.1058, ..., -0.0895, -0.1061, -0.0614], ..., [ 0.0130, 0.0340, -0.0421, ..., -0.0279, -0.1093, 0.0481], [ 0.0015, 0.0896, 0.0761, ..., -0.0713, 0.0571, -0.1084], [-0.0458, -0.0080, -0.0230, ..., -0.0254, -0.0562, 0.0463]], device='cuda:0'), grad: tensor([[ 1.7919e-06, 6.1132e-06, 1.0096e-05, ..., 4.0978e-08, 6.6757e-06, 1.0151e-07], [ 3.7998e-07, 5.7649e-07, 4.2235e-07, ..., -3.4925e-08, 3.3760e-07, -1.5339e-06], [ 3.2540e-06, 3.5521e-06, 2.9188e-06, ..., 6.5193e-09, 2.5760e-06, 4.8103e-07], ..., [-4.8988e-07, -9.6019e-07, 7.6275e-07, ..., 3.2596e-08, 7.0455e-07, 3.1386e-07], [-1.2681e-05, -1.5378e-05, -1.7032e-05, ..., 4.8708e-07, -1.1958e-05, 9.1270e-08], [ 2.9784e-06, 3.7672e-07, 1.8030e-06, ..., 3.7253e-08, 1.0645e-06, -2.0955e-08]], device='cuda:0') Epoch 124, bias, value: tensor([ 0.0292, -0.0216, -0.0057, -0.0030, 0.0082, 0.0039, -0.0190, 0.0007, 0.0208, 0.0177], device='cuda:0'), grad: tensor([ 2.9951e-06, -2.9076e-06, 8.4490e-06, 3.6396e-06, 2.7008e-08, -3.9749e-06, 5.7556e-06, 4.8382e-07, -2.6643e-05, 1.2159e-05], device='cuda:0') 100 0.0001 changing lr epoch 123, time 256.60, cls_loss 0.0030 cls_loss_mapping 0.0072 cls_loss_causal 0.5619 re_mapping 0.0083 re_causal 0.0242 /// teacc 99.00 lr 0.00010000 Epoch 125, weight, value: tensor([[ 0.0304, -0.0803, -0.0796, ..., -0.0232, -0.1414, -0.0084], [-0.0822, -0.1175, -0.1151, ..., -0.1011, -0.0553, -0.0040], [ 0.0669, 0.0546, -0.1069, ..., -0.0899, -0.1073, -0.0619], ..., [ 0.0124, 0.0336, -0.0428, ..., -0.0278, -0.1100, 0.0475], [ 0.0019, 0.0906, 0.0766, ..., -0.0713, 0.0575, -0.1090], [-0.0448, -0.0064, -0.0228, ..., -0.0251, -0.0560, 0.0478]], device='cuda:0'), grad: tensor([[ 4.5970e-06, 5.6662e-06, 1.2685e-06, ..., 3.0547e-07, 1.5348e-06, 3.6880e-07], [ 6.2957e-07, -2.3052e-05, 3.5716e-07, ..., -8.7917e-07, 1.7509e-07, -3.2093e-06], [-1.5022e-06, 5.4128e-06, 6.6608e-06, ..., 2.4820e-07, 6.7800e-06, 9.3738e-07], ..., [-5.0478e-06, -4.4666e-06, 1.0030e-06, ..., 1.4668e-07, 7.2643e-08, -1.4575e-07], [-3.4105e-06, 6.4485e-06, -6.4522e-06, ..., 8.3027e-07, -5.7369e-06, 7.5344e-07], [ 9.9186e-07, 3.2280e-06, 8.3074e-07, ..., 1.3039e-07, 1.0654e-06, 1.9651e-07]], device='cuda:0') Epoch 125, bias, value: tensor([ 0.0291, -0.0213, -0.0063, -0.0026, 0.0076, 0.0043, -0.0193, 0.0002, 0.0209, 0.0185], device='cuda:0'), grad: tensor([ 2.4214e-05, -2.5368e-04, 2.2128e-05, 5.9558e-07, 7.2755e-06, 7.7784e-06, 1.0543e-05, 1.9759e-05, 1.2994e-04, 3.1173e-05], device='cuda:0') 100 0.0001 changing lr epoch 124, time 256.54, cls_loss 0.0029 cls_loss_mapping 0.0075 cls_loss_causal 0.5420 re_mapping 0.0075 re_causal 0.0225 /// teacc 99.02 lr 0.00010000 Epoch 126, weight, value: tensor([[ 0.0303, -0.0807, -0.0808, ..., -0.0240, -0.1425, -0.0087], [-0.0826, -0.1175, -0.1169, ..., -0.1014, -0.0564, -0.0035], [ 0.0669, 0.0544, -0.1088, ..., -0.0907, -0.1083, -0.0624], ..., [ 0.0127, 0.0338, -0.0432, ..., -0.0280, -0.1103, 0.0474], [ 0.0030, 0.0927, 0.0782, ..., -0.0710, 0.0591, -0.1095], [-0.0449, -0.0068, -0.0238, ..., -0.0253, -0.0570, 0.0479]], device='cuda:0'), grad: tensor([[ 2.2631e-07, 6.0022e-05, 2.0176e-05, ..., 6.0678e-05, 1.3836e-05, 1.5631e-05], [ 8.0168e-06, 5.3197e-05, 2.3805e-06, ..., 8.0988e-06, 9.9279e-07, 2.1122e-06], [-5.5134e-06, -4.1097e-05, 1.7313e-06, ..., 8.3679e-07, 1.1483e-06, 2.0536e-07], ..., [-2.0815e-07, 2.7120e-06, 2.8703e-06, ..., 6.0629e-07, 1.6093e-06, 3.7719e-08], [-2.4274e-05, -4.0203e-05, -3.4839e-05, ..., 8.1211e-06, -1.6987e-05, 7.3807e-07], [ 9.5665e-06, 3.0756e-05, 1.8328e-05, ..., 7.1377e-06, 1.0915e-05, 1.3178e-06]], device='cuda:0') Epoch 126, bias, value: tensor([ 0.0288, -0.0208, -0.0068, -0.0037, 0.0072, 0.0051, -0.0191, 0.0002, 0.0225, 0.0183], device='cuda:0'), grad: tensor([ 1.7715e-04, 3.3641e-04, -2.9922e-04, 3.2373e-06, 2.6274e-04, 1.2863e-04, -6.8998e-04, 1.8284e-05, -4.8786e-05, 1.1164e-04], device='cuda:0') 100 0.0001 changing lr epoch 125, time 256.99, cls_loss 0.0023 cls_loss_mapping 0.0056 cls_loss_causal 0.5572 re_mapping 0.0069 re_causal 0.0229 /// teacc 98.95 lr 0.00010000 Epoch 127, weight, value: tensor([[ 0.0306, -0.0809, -0.0816, ..., -0.0243, -0.1430, -0.0087], [-0.0833, -0.1181, -0.1174, ..., -0.1025, -0.0568, -0.0035], [ 0.0668, 0.0543, -0.1092, ..., -0.0909, -0.1088, -0.0633], ..., [ 0.0130, 0.0340, -0.0436, ..., -0.0281, -0.1107, 0.0474], [ 0.0034, 0.0933, 0.0787, ..., -0.0717, 0.0596, -0.1098], [-0.0450, -0.0069, -0.0245, ..., -0.0255, -0.0577, 0.0484]], device='cuda:0'), grad: tensor([[-1.1520e-06, 6.8638e-07, 2.7288e-07, ..., 1.4668e-07, 1.5041e-07, 2.2678e-07], [ 3.8696e-07, 1.0887e-06, 6.1933e-08, ..., 1.8161e-08, 4.5635e-08, -2.2333e-06], [-5.5395e-06, -2.1815e-05, 5.3085e-08, ..., 1.7229e-08, 4.6566e-08, 1.4193e-06], ..., [ 5.6736e-06, 2.1473e-05, 1.4575e-07, ..., 4.6566e-10, 8.4750e-08, -1.4296e-06], [ 1.2955e-06, 2.9318e-06, 2.6869e-07, ..., 3.9581e-08, 1.4203e-07, 6.9244e-07], [ 1.1427e-06, 1.2340e-06, 3.9535e-07, ..., 6.9849e-09, 3.1805e-07, 1.4761e-07]], device='cuda:0') Epoch 127, bias, value: tensor([ 0.0286, -0.0208, -0.0071, -0.0038, 0.0075, 0.0048, -0.0188, 0.0002, 0.0227, 0.0184], device='cuda:0'), grad: tensor([-1.2554e-06, 3.1926e-06, -1.9699e-05, 2.2426e-06, -1.6797e-04, 4.3400e-06, 1.3018e-04, 3.0994e-05, 8.0094e-06, 9.9763e-06], device='cuda:0') 100 0.0001 changing lr epoch 126, time 256.88, cls_loss 0.0024 cls_loss_mapping 0.0058 cls_loss_causal 0.5396 re_mapping 0.0070 re_causal 0.0220 /// teacc 99.02 lr 0.00010000 Epoch 128, weight, value: tensor([[ 0.0307, -0.0812, -0.0825, ..., -0.0247, -0.1436, -0.0088], [-0.0836, -0.1180, -0.1178, ..., -0.1027, -0.0568, -0.0033], [ 0.0670, 0.0543, -0.1101, ..., -0.0910, -0.1095, -0.0635], ..., [ 0.0129, 0.0341, -0.0447, ..., -0.0281, -0.1113, 0.0474], [ 0.0042, 0.0947, 0.0806, ..., -0.0716, 0.0610, -0.1115], [-0.0450, -0.0067, -0.0248, ..., -0.0255, -0.0576, 0.0486]], device='cuda:0'), grad: tensor([[ 2.8387e-06, 2.7716e-06, 3.6899e-06, ..., 5.3551e-08, 7.1377e-06, 1.9446e-06], [ 3.2801e-06, 1.0626e-06, 2.5723e-06, ..., -9.7789e-09, 3.3919e-06, 1.7649e-06], [ 4.3064e-06, 1.1614e-06, 4.7572e-06, ..., 9.3132e-09, 6.4336e-06, 3.4086e-06], ..., [ 1.1168e-05, 1.4696e-06, 4.1164e-06, ..., 2.7940e-09, 5.8785e-06, 1.8263e-06], [ 9.4846e-06, 6.6534e-06, 9.7975e-06, ..., 3.0268e-08, 2.0668e-05, 5.6252e-06], [ 4.5337e-06, -1.4156e-05, -2.4110e-05, ..., 1.3970e-09, -5.3495e-05, -1.1809e-05]], device='cuda:0') Epoch 128, bias, value: tensor([ 0.0287, -0.0206, -0.0071, -0.0035, 0.0074, 0.0038, -0.0188, 0.0001, 0.0234, 0.0183], device='cuda:0'), grad: tensor([ 1.1697e-05, 2.0474e-05, 2.2188e-05, -5.2154e-06, -4.6182e-04, 2.2441e-05, 1.2465e-05, 1.1665e-04, 5.6863e-05, 2.0468e-04], device='cuda:0') 100 0.0001 changing lr epoch 127, time 256.87, cls_loss 0.0030 cls_loss_mapping 0.0057 cls_loss_causal 0.5374 re_mapping 0.0076 re_causal 0.0220 /// teacc 99.06 lr 0.00010000 Epoch 129, weight, value: tensor([[ 0.0307, -0.0815, -0.0833, ..., -0.0250, -0.1442, -0.0091], [-0.0842, -0.1183, -0.1182, ..., -0.1028, -0.0571, -0.0033], [ 0.0676, 0.0545, -0.1109, ..., -0.0912, -0.1098, -0.0637], ..., [ 0.0133, 0.0345, -0.0454, ..., -0.0280, -0.1118, 0.0477], [ 0.0043, 0.0950, 0.0807, ..., -0.0723, 0.0614, -0.1123], [-0.0450, -0.0069, -0.0249, ..., -0.0254, -0.0578, 0.0486]], device='cuda:0'), grad: tensor([[ 5.1036e-06, 1.7926e-05, 1.3039e-07, ..., 7.2643e-08, 7.1200e-07, 7.6517e-06], [ 3.6415e-07, 3.9581e-07, 7.2177e-08, ..., 5.5879e-09, 7.3574e-08, -4.3586e-07], [-2.3516e-07, -3.0752e-06, 4.2701e-07, ..., 2.0955e-08, 3.4412e-07, 7.3574e-07], ..., [-1.2759e-07, 2.6915e-06, 1.0105e-07, ..., 9.3132e-10, 5.3784e-07, -5.9325e-07], [ 8.0019e-06, 1.0677e-05, 9.3598e-08, ..., -1.2107e-08, 1.1707e-06, 4.8392e-06], [-2.5183e-05, -3.5286e-05, 8.4424e-07, ..., 1.4435e-08, -3.4552e-06, -1.5736e-05]], device='cuda:0') Epoch 129, bias, value: tensor([ 0.0287, -0.0207, -0.0068, -0.0036, 0.0070, 0.0038, -0.0192, 0.0005, 0.0229, 0.0185], device='cuda:0'), grad: tensor([ 1.3113e-06, -6.5239e-07, -3.2671e-06, 1.4126e-05, 6.8685e-07, -4.9919e-06, 2.4080e-05, 1.7732e-06, 1.9535e-05, -5.2601e-05], device='cuda:0') 100 0.0001 changing lr epoch 128, time 257.15, cls_loss 0.0036 cls_loss_mapping 0.0073 cls_loss_causal 0.5400 re_mapping 0.0071 re_causal 0.0216 /// teacc 98.95 lr 0.00010000 Epoch 130, weight, value: tensor([[ 0.0309, -0.0817, -0.0855, ..., -0.0252, -0.1465, -0.0093], [-0.0844, -0.1190, -0.1184, ..., -0.1029, -0.0584, -0.0027], [ 0.0674, 0.0547, -0.1118, ..., -0.0914, -0.1086, -0.0663], ..., [ 0.0138, 0.0349, -0.0460, ..., -0.0275, -0.1123, 0.0477], [ 0.0041, 0.0959, 0.0827, ..., -0.0726, 0.0624, -0.1130], [-0.0449, -0.0067, -0.0253, ..., -0.0259, -0.0583, 0.0486]], device='cuda:0'), grad: tensor([[-4.6045e-06, 1.0999e-06, -3.2689e-07, ..., 5.0385e-07, -3.9442e-07, 6.2305e-07], [ 5.6578e-07, 5.0552e-06, 1.1940e-06, ..., -9.3598e-08, 1.0505e-06, -6.0489e-07], [ 3.6135e-07, 2.9970e-06, 7.2923e-07, ..., 1.9139e-07, 7.6042e-07, 7.3062e-07], ..., [-4.9621e-06, -1.8001e-05, 6.6170e-07, ..., 2.8405e-08, 5.7556e-07, -1.1049e-05], [ 1.4035e-06, 1.0677e-05, 2.3112e-05, ..., 1.0729e-06, 1.9580e-05, 6.2166e-07], [ 3.7011e-06, 5.5060e-06, 1.1260e-06, ..., 5.8673e-08, 9.9000e-07, 5.7593e-06]], device='cuda:0') Epoch 130, bias, value: tensor([ 0.0285, -0.0192, -0.0083, -0.0033, 0.0067, 0.0042, -0.0202, 0.0006, 0.0233, 0.0184], device='cuda:0'), grad: tensor([-1.9372e-05, 1.1824e-05, 1.1109e-05, -5.2124e-05, 1.6242e-06, 2.9832e-05, 3.2894e-06, -5.7608e-05, 4.4584e-05, 2.6658e-05], device='cuda:0') 100 0.0001 changing lr epoch 129, time 256.82, cls_loss 0.0039 cls_loss_mapping 0.0078 cls_loss_causal 0.5529 re_mapping 0.0071 re_causal 0.0211 /// teacc 99.00 lr 0.00010000 Epoch 131, weight, value: tensor([[ 0.0311, -0.0821, -0.0862, ..., -0.0260, -0.1470, -0.0095], [-0.0850, -0.1188, -0.1191, ..., -0.1034, -0.0588, -0.0017], [ 0.0675, 0.0546, -0.1125, ..., -0.0918, -0.1092, -0.0675], ..., [ 0.0139, 0.0349, -0.0465, ..., -0.0276, -0.1128, 0.0471], [ 0.0047, 0.0957, 0.0822, ..., -0.0743, 0.0622, -0.1135], [-0.0443, -0.0053, -0.0263, ..., -0.0264, -0.0588, 0.0489]], device='cuda:0'), grad: tensor([[-2.8266e-07, 2.7474e-06, 2.6729e-06, ..., 1.7397e-06, 2.3674e-06, 8.7544e-07], [ 1.4976e-06, 3.5260e-06, 5.0887e-06, ..., 2.5667e-06, 4.2692e-06, -7.3165e-06], [-1.8952e-06, 5.0627e-06, 5.5693e-06, ..., 3.9339e-06, 5.0254e-06, 1.2340e-06], ..., [-5.8673e-08, 7.2736e-07, 2.5760e-06, ..., 1.3895e-06, 2.1979e-06, 5.2378e-06], [ 6.3796e-08, 6.0499e-05, 8.8573e-05, ..., 4.9144e-05, 7.6413e-05, 2.3609e-07], [-1.5162e-06, 1.1623e-06, 6.8061e-06, ..., 3.3565e-06, 5.6028e-06, -3.3975e-06]], device='cuda:0') Epoch 131, bias, value: tensor([ 2.8567e-02, -1.7859e-02, -9.0745e-03, -2.2606e-03, 5.1761e-03, 3.1119e-03, -1.8991e-02, -8.3274e-05, 2.2161e-02, 1.9299e-02], device='cuda:0'), grad: tensor([ 1.3098e-05, -1.4096e-05, 1.1221e-05, 5.0336e-05, 5.0180e-06, -4.6825e-04, 1.8024e-04, 3.4422e-05, 1.9813e-04, -1.0580e-05], device='cuda:0') 100 0.0001 changing lr epoch 130, time 257.20, cls_loss 0.0023 cls_loss_mapping 0.0069 cls_loss_causal 0.5395 re_mapping 0.0075 re_causal 0.0228 /// teacc 99.02 lr 0.00010000 Epoch 132, weight, value: tensor([[ 0.0311, -0.0825, -0.0873, ..., -0.0264, -0.1477, -0.0098], [-0.0854, -0.1189, -0.1194, ..., -0.1039, -0.0590, -0.0015], [ 0.0677, 0.0546, -0.1128, ..., -0.0920, -0.1095, -0.0677], ..., [ 0.0139, 0.0349, -0.0468, ..., -0.0278, -0.1131, 0.0471], [ 0.0058, 0.0964, 0.0828, ..., -0.0748, 0.0627, -0.1140], [-0.0440, -0.0048, -0.0279, ..., -0.0270, -0.0598, 0.0492]], device='cuda:0'), grad: tensor([[ 4.1025e-07, 8.4192e-07, 3.5036e-06, ..., 2.0117e-07, 1.6894e-06, 1.2955e-06], [ 3.7309e-06, -4.0606e-06, 5.0031e-06, ..., 6.7521e-08, 3.9600e-06, -1.9446e-05], [ 1.3569e-06, 3.0603e-06, 3.4329e-06, ..., 8.3819e-08, 2.6934e-06, 3.6210e-06], ..., [-4.4405e-06, -2.9355e-06, 2.7250e-06, ..., 3.0734e-08, 1.7928e-06, 1.3553e-05], [ 5.4948e-08, -1.6717e-06, 2.3887e-05, ..., 5.5367e-07, 1.4618e-05, 1.4799e-06], [ 2.6897e-06, -5.6438e-07, 5.7779e-06, ..., 1.0151e-07, 1.0610e-05, -3.1497e-06]], device='cuda:0') Epoch 132, bias, value: tensor([ 0.0282, -0.0178, -0.0090, -0.0017, 0.0044, 0.0020, -0.0180, -0.0001, 0.0226, 0.0198], device='cuda:0'), grad: tensor([ 1.7121e-05, -2.8148e-05, 2.8133e-05, -6.3300e-05, -7.6056e-05, 9.8720e-06, 1.2398e-05, 4.8667e-05, 6.2287e-05, -1.1228e-05], device='cuda:0') 100 0.0001 changing lr epoch 131, time 256.57, cls_loss 0.0034 cls_loss_mapping 0.0062 cls_loss_causal 0.5661 re_mapping 0.0069 re_causal 0.0204 /// teacc 99.00 lr 0.00010000 Epoch 133, weight, value: tensor([[ 0.0315, -0.0828, -0.0877, ..., -0.0264, -0.1482, -0.0101], [-0.0861, -0.1192, -0.1198, ..., -0.1042, -0.0593, -0.0009], [ 0.0679, 0.0546, -0.1140, ..., -0.0926, -0.1102, -0.0682], ..., [ 0.0143, 0.0353, -0.0476, ..., -0.0263, -0.1132, 0.0468], [ 0.0060, 0.0966, 0.0830, ..., -0.0751, 0.0629, -0.1148], [-0.0442, -0.0049, -0.0290, ..., -0.0280, -0.0604, 0.0492]], device='cuda:0'), grad: tensor([[ 4.9770e-05, 9.2341e-07, 5.9837e-07, ..., 8.1398e-07, 6.6578e-05, 5.9575e-05], [ 4.0010e-06, 1.8589e-06, 5.6950e-07, ..., 9.0338e-08, 5.6736e-06, 4.3698e-06], [ 2.5555e-05, -1.3178e-06, 4.3632e-07, ..., 5.9605e-08, 3.6091e-05, 3.1233e-05], ..., [ 1.0364e-05, 5.6345e-07, 2.4913e-07, ..., 4.1910e-09, 1.4193e-05, 1.1995e-05], [ 1.7462e-06, -2.3041e-06, -1.0794e-06, ..., 3.2177e-07, 1.0505e-06, 2.1476e-06], [-7.7295e-04, 1.3085e-07, 1.2852e-07, ..., 4.3306e-08, -1.0061e-03, -9.0694e-04]], device='cuda:0') Epoch 133, bias, value: tensor([ 0.0286, -0.0174, -0.0090, -0.0016, 0.0045, 0.0019, -0.0181, -0.0003, 0.0221, 0.0195], device='cuda:0'), grad: tensor([ 6.7282e-04, 5.5850e-05, 3.6430e-04, 8.7051e-03, 1.5661e-05, 5.8413e-04, 7.4245e-06, 1.4424e-04, 2.4527e-05, -1.0574e-02], device='cuda:0') 100 0.0001 changing lr epoch 132, time 255.35, cls_loss 0.0024 cls_loss_mapping 0.0060 cls_loss_causal 0.5313 re_mapping 0.0072 re_causal 0.0219 /// teacc 99.02 lr 0.00010000 Epoch 134, weight, value: tensor([[ 0.0316, -0.0830, -0.0880, ..., -0.0269, -0.1487, -0.0104], [-0.0867, -0.1205, -0.1211, ..., -0.1047, -0.0600, -0.0012], [ 0.0683, 0.0547, -0.1145, ..., -0.0928, -0.1107, -0.0683], ..., [ 0.0143, 0.0358, -0.0481, ..., -0.0264, -0.1139, 0.0471], [ 0.0061, 0.0963, 0.0827, ..., -0.0756, 0.0624, -0.1151], [-0.0445, -0.0051, -0.0295, ..., -0.0280, -0.0594, 0.0494]], device='cuda:0'), grad: tensor([[ 3.1944e-06, 1.4775e-05, 1.4201e-05, ..., 3.1106e-06, 9.4399e-06, 2.6431e-06], [ 2.6390e-05, 5.1647e-05, 3.5256e-05, ..., 9.5041e-07, 3.1263e-05, -1.3551e-06], [ 5.2005e-05, 1.2863e-04, 1.1271e-04, ..., 2.6301e-05, 7.8797e-05, 9.3058e-06], ..., [ 1.5378e-05, 2.4542e-05, 1.6317e-05, ..., 3.3099e-06, 1.3076e-05, 8.3074e-06], [-1.1051e-04, -2.2984e-04, -1.2672e-04, ..., -6.2846e-06, -1.1253e-04, 9.0012e-07], [ 5.2191e-06, 1.7300e-05, 1.8179e-05, ..., 3.7719e-06, 1.3039e-05, -8.7395e-06]], device='cuda:0') Epoch 134, bias, value: tensor([ 0.0288, -0.0181, -0.0089, -0.0022, 0.0055, 0.0020, -0.0184, 0.0003, 0.0214, 0.0196], device='cuda:0'), grad: tensor([ 3.7968e-05, 1.0812e-04, 3.4761e-04, -1.8673e-03, 1.2003e-05, 1.6394e-03, 7.0453e-05, 7.0453e-05, -3.8981e-04, -2.8506e-05], device='cuda:0') 100 0.0001 changing lr epoch 133, time 256.79, cls_loss 0.0024 cls_loss_mapping 0.0047 cls_loss_causal 0.5320 re_mapping 0.0074 re_causal 0.0221 /// teacc 98.94 lr 0.00010000 Epoch 135, weight, value: tensor([[ 0.0318, -0.0835, -0.0895, ..., -0.0287, -0.1497, -0.0105], [-0.0871, -0.1209, -0.1215, ..., -0.1050, -0.0604, -0.0011], [ 0.0685, 0.0545, -0.1156, ..., -0.0935, -0.1117, -0.0685], ..., [ 0.0146, 0.0363, -0.0487, ..., -0.0263, -0.1146, 0.0472], [ 0.0052, 0.0955, 0.0828, ..., -0.0763, 0.0622, -0.1177], [-0.0444, -0.0044, -0.0292, ..., -0.0287, -0.0584, 0.0497]], device='cuda:0'), grad: tensor([[ 4.1053e-06, 5.7295e-06, 1.0896e-07, ..., 6.2864e-08, 1.5227e-07, 3.4738e-07], [ 4.6268e-06, 8.4117e-06, 8.1956e-08, ..., 7.4506e-09, 2.6729e-07, -2.4401e-07], [-1.8254e-06, 2.8402e-05, 2.6496e-07, ..., 1.7695e-08, 6.7987e-07, 1.6019e-07], ..., [-2.6673e-05, -6.1393e-05, 1.4622e-07, ..., 4.6566e-10, 4.7591e-07, 2.2762e-06], [ 1.2070e-06, 1.9092e-06, -5.2387e-07, ..., 5.3085e-08, -6.9011e-07, 9.0292e-07], [-1.0831e-06, -9.2089e-06, 7.5437e-08, ..., 4.1910e-09, 3.5577e-07, -6.1132e-06]], device='cuda:0') Epoch 135, bias, value: tensor([ 0.0287, -0.0189, -0.0089, -0.0019, 0.0058, 0.0019, -0.0182, 0.0014, 0.0206, 0.0195], device='cuda:0'), grad: tensor([ 1.7866e-05, 2.4021e-05, 5.8174e-05, 5.5671e-05, 9.9018e-06, 3.8296e-06, 1.2167e-05, -1.7285e-04, 8.6725e-06, -1.7688e-05], device='cuda:0') 100 0.0001 changing lr epoch 134, time 256.90, cls_loss 0.0030 cls_loss_mapping 0.0059 cls_loss_causal 0.5268 re_mapping 0.0077 re_causal 0.0220 /// teacc 99.02 lr 0.00010000 Epoch 136, weight, value: tensor([[ 0.0319, -0.0840, -0.0903, ..., -0.0294, -0.1502, -0.0110], [-0.0881, -0.1217, -0.1220, ..., -0.1055, -0.0608, -0.0016], [ 0.0686, 0.0542, -0.1164, ..., -0.0937, -0.1124, -0.0692], ..., [ 0.0154, 0.0371, -0.0491, ..., -0.0253, -0.1154, 0.0486], [ 0.0062, 0.0971, 0.0834, ..., -0.0758, 0.0633, -0.1176], [-0.0448, -0.0057, -0.0296, ..., -0.0291, -0.0591, 0.0483]], device='cuda:0'), grad: tensor([[-8.1509e-06, 2.7791e-06, 2.6375e-06, ..., 3.0715e-06, 2.1271e-06, 8.4937e-07], [ 1.3728e-06, 1.3765e-06, 1.1967e-07, ..., 4.8801e-07, 1.3923e-07, -2.1569e-06], [-2.2184e-06, -2.3581e-06, 1.3225e-07, ..., 1.7649e-07, -1.4901e-07, 3.0780e-07], ..., [-1.8068e-06, -1.2340e-06, 1.3504e-08, ..., -1.2526e-07, 2.6543e-08, -1.8012e-06], [ 1.6559e-06, 2.9150e-06, 4.4191e-07, ..., 4.2003e-07, 4.1351e-07, 7.8417e-07], [ 6.3926e-06, 3.9488e-06, 1.7276e-07, ..., 4.3120e-07, 1.4342e-07, 9.5740e-07]], device='cuda:0') Epoch 136, bias, value: tensor([ 0.0285, -0.0193, -0.0093, -0.0021, 0.0066, 0.0016, -0.0181, 0.0029, 0.0216, 0.0183], device='cuda:0'), grad: tensor([-4.8548e-05, -6.7614e-06, -8.1351e-07, 5.7407e-06, -5.7697e-05, 3.7216e-06, 1.6704e-05, 1.0140e-05, 1.3337e-05, 6.4135e-05], device='cuda:0') 100 0.0001 changing lr epoch 135, time 256.81, cls_loss 0.0030 cls_loss_mapping 0.0062 cls_loss_causal 0.5643 re_mapping 0.0071 re_causal 0.0218 /// teacc 98.97 lr 0.00010000 Epoch 137, weight, value: tensor([[ 0.0318, -0.0845, -0.0916, ..., -0.0300, -0.1512, -0.0115], [-0.0889, -0.1220, -0.1227, ..., -0.1062, -0.0615, -0.0014], [ 0.0696, 0.0547, -0.1176, ..., -0.0940, -0.1136, -0.0693], ..., [ 0.0152, 0.0375, -0.0500, ..., -0.0251, -0.1178, 0.0487], [ 0.0062, 0.0984, 0.0843, ..., -0.0755, 0.0640, -0.1179], [-0.0443, -0.0063, -0.0295, ..., -0.0294, -0.0572, 0.0477]], device='cuda:0'), grad: tensor([[ 4.3735e-06, 5.3532e-06, 3.7234e-06, ..., 2.6189e-06, 3.6433e-06, 7.9535e-07], [ 5.0068e-06, 4.4629e-06, 3.0342e-06, ..., 2.0657e-06, 3.2373e-06, 2.1551e-06], [-1.9848e-05, -6.0629e-07, 1.5227e-06, ..., -3.3621e-06, 1.5618e-06, 4.0652e-07], ..., [ 3.5437e-07, 5.0478e-07, 5.3737e-07, ..., 1.7323e-07, 5.7556e-07, -8.8476e-09], [ 1.3905e-06, -2.1104e-06, -1.0848e-05, ..., 5.6773e-06, -8.5011e-06, 2.3395e-06], [-5.1968e-06, 1.0423e-05, 1.4566e-06, ..., 2.0815e-07, 1.4324e-06, 4.1366e-05]], device='cuda:0') Epoch 137, bias, value: tensor([ 0.0281, -0.0194, -0.0089, -0.0019, 0.0064, 0.0013, -0.0182, 0.0026, 0.0215, 0.0189], device='cuda:0'), grad: tensor([ 2.5555e-05, 6.2287e-05, -9.2447e-05, 4.1544e-05, -7.2336e-04, 1.9774e-05, -1.0923e-05, 8.4043e-06, -5.5227e-07, 6.6996e-04], device='cuda:0') 100 0.0001 changing lr epoch 136, time 257.41, cls_loss 0.0021 cls_loss_mapping 0.0049 cls_loss_causal 0.5401 re_mapping 0.0068 re_causal 0.0215 /// teacc 99.00 lr 0.00010000 Epoch 138, weight, value: tensor([[ 0.0318, -0.0847, -0.0921, ..., -0.0303, -0.1517, -0.0117], [-0.0896, -0.1223, -0.1230, ..., -0.1063, -0.0620, -0.0014], [ 0.0699, 0.0548, -0.1185, ..., -0.0945, -0.1143, -0.0695], ..., [ 0.0155, 0.0376, -0.0505, ..., -0.0252, -0.1188, 0.0488], [ 0.0062, 0.0986, 0.0842, ..., -0.0758, 0.0641, -0.1185], [-0.0444, -0.0063, -0.0298, ..., -0.0295, -0.0571, 0.0475]], device='cuda:0'), grad: tensor([[-7.3854e-07, 1.3756e-06, 1.7658e-06, ..., 1.1865e-06, 1.6522e-06, 9.3598e-08], [ 6.5425e-07, 7.0361e-07, 3.3993e-07, ..., 8.1025e-08, 5.2201e-07, 6.3796e-08], [ 4.0000e-07, 1.6531e-06, 1.8300e-06, ..., 1.9092e-07, 2.5369e-06, 2.1746e-07], ..., [-7.4767e-06, -9.0376e-06, 6.6496e-07, ..., 5.1223e-09, 1.0338e-06, -4.9807e-06], [-1.2601e-06, -3.2559e-06, 3.1516e-06, ..., 6.7428e-07, 2.0228e-06, 6.4261e-08], [ 7.6815e-06, 8.4043e-06, 5.7407e-06, ..., 8.1025e-08, 5.6513e-06, 4.1388e-06]], device='cuda:0') Epoch 138, bias, value: tensor([ 0.0276, -0.0197, -0.0087, -0.0017, 0.0071, 0.0011, -0.0179, 0.0025, 0.0212, 0.0189], device='cuda:0'), grad: tensor([ 7.5623e-07, 1.7006e-06, 4.1686e-06, 5.8897e-06, 8.9174e-07, -2.2963e-05, -4.2394e-06, -1.2919e-05, 2.9933e-06, 2.3693e-05], device='cuda:0') 100 0.0001 changing lr epoch 137, time 257.34, cls_loss 0.0022 cls_loss_mapping 0.0060 cls_loss_causal 0.5614 re_mapping 0.0069 re_causal 0.0218 /// teacc 98.90 lr 0.00010000 Epoch 139, weight, value: tensor([[ 0.0316, -0.0855, -0.0929, ..., -0.0307, -0.1531, -0.0128], [-0.0903, -0.1224, -0.1233, ..., -0.1067, -0.0629, -0.0014], [ 0.0702, 0.0549, -0.1193, ..., -0.0949, -0.1140, -0.0692], ..., [ 0.0159, 0.0379, -0.0509, ..., -0.0253, -0.1190, 0.0489], [ 0.0064, 0.0989, 0.0843, ..., -0.0763, 0.0644, -0.1194], [-0.0445, -0.0063, -0.0298, ..., -0.0293, -0.0567, 0.0474]], device='cuda:0'), grad: tensor([[-3.1386e-07, 4.8950e-06, 4.1127e-06, ..., 2.3786e-06, 5.7742e-06, 7.4506e-08], [ 8.4162e-05, 8.0168e-05, 1.4659e-06, ..., 4.5914e-07, 2.1793e-06, 5.2363e-05], [ 1.3560e-06, 3.3248e-06, 2.1998e-06, ..., 4.6287e-07, 3.4049e-06, 7.6182e-07], ..., [-9.9421e-05, -9.4354e-05, 1.6652e-06, ..., 3.1944e-07, 2.2482e-06, -6.2227e-05], [ 2.4810e-06, 6.5416e-06, 5.3868e-06, ..., 1.5842e-06, 7.8455e-06, 1.0803e-06], [ 1.0870e-05, 1.1332e-05, 4.5784e-06, ..., 9.4716e-07, 6.3851e-06, 5.7966e-06]], device='cuda:0') Epoch 139, bias, value: tensor([ 0.0274, -0.0199, -0.0079, -0.0017, 0.0073, 0.0012, -0.0179, 0.0024, 0.0211, 0.0186], device='cuda:0'), grad: tensor([ 1.0066e-05, 2.1124e-04, 1.1422e-05, -3.6627e-05, 8.1360e-06, -1.5214e-05, -1.7639e-06, -2.4652e-04, 2.2039e-05, 3.7134e-05], device='cuda:0') 100 0.0001 changing lr epoch 138, time 251.92, cls_loss 0.0022 cls_loss_mapping 0.0062 cls_loss_causal 0.5142 re_mapping 0.0070 re_causal 0.0204 /// teacc 98.98 lr 0.00010000 Epoch 140, weight, value: tensor([[ 0.0317, -0.0858, -0.0936, ..., -0.0312, -0.1537, -0.0129], [-0.0910, -0.1233, -0.1240, ..., -0.1081, -0.0638, -0.0012], [ 0.0711, 0.0551, -0.1200, ..., -0.0956, -0.1148, -0.0688], ..., [ 0.0159, 0.0383, -0.0514, ..., -0.0254, -0.1189, 0.0490], [ 0.0064, 0.0991, 0.0843, ..., -0.0766, 0.0645, -0.1197], [-0.0446, -0.0064, -0.0305, ..., -0.0297, -0.0571, 0.0473]], device='cuda:0'), grad: tensor([[-6.4149e-06, 1.4165e-06, 9.3598e-07, ..., 6.3702e-07, 8.5030e-07, 3.5763e-07], [ 1.2163e-06, 2.3060e-06, 2.3842e-07, ..., 8.9407e-08, 3.1013e-07, -6.5044e-06], [ 6.5453e-06, 1.5028e-05, 4.2468e-07, ..., 1.3318e-07, 5.2154e-07, 3.3807e-06], ..., [-7.9945e-06, -2.0698e-05, 1.1362e-07, ..., -1.3225e-07, 1.3970e-07, -2.3283e-06], [ 7.9814e-07, 5.0254e-06, 6.3740e-06, ..., 4.5747e-06, 4.4815e-06, 8.5589e-07], [ 1.5264e-06, 1.7229e-07, 3.7532e-07, ..., 9.2201e-08, 4.3772e-07, 1.2815e-06]], device='cuda:0') Epoch 140, bias, value: tensor([ 0.0275, -0.0199, -0.0077, -0.0017, 0.0074, 0.0020, -0.0190, 0.0026, 0.0207, 0.0181], device='cuda:0'), grad: tensor([-1.7047e-05, -1.9982e-05, 2.6599e-05, 6.9067e-06, 3.7476e-06, 7.1302e-06, -6.5118e-06, -2.0206e-05, 9.5516e-06, 9.7975e-06], device='cuda:0') 100 0.0001 changing lr epoch 139, time 251.57, cls_loss 0.0033 cls_loss_mapping 0.0067 cls_loss_causal 0.5489 re_mapping 0.0071 re_causal 0.0205 /// teacc 98.92 lr 0.00010000 Epoch 141, weight, value: tensor([[ 0.0318, -0.0865, -0.0949, ..., -0.0322, -0.1545, -0.0132], [-0.0920, -0.1243, -0.1243, ..., -0.1084, -0.0644, -0.0014], [ 0.0719, 0.0536, -0.1220, ..., -0.0961, -0.1197, -0.0669], ..., [ 0.0133, 0.0365, -0.0518, ..., -0.0239, -0.1223, 0.0489], [ 0.0112, 0.1047, 0.0848, ..., -0.0768, 0.0697, -0.1181], [-0.0447, -0.0063, -0.0303, ..., -0.0294, -0.0573, 0.0470]], device='cuda:0'), grad: tensor([[ 6.6031e-07, 2.0228e-06, 2.2464e-06, ..., 2.7940e-08, 3.0231e-06, 1.7947e-06], [ 2.1569e-06, 5.2620e-07, 2.2817e-07, ..., 2.7940e-09, -2.2985e-06, -1.0580e-06], [-1.2107e-07, 5.3924e-07, 4.0047e-07, ..., 6.5193e-09, 3.6508e-07, 9.2387e-07], ..., [-4.1164e-06, -8.4043e-06, 3.7439e-07, ..., 7.4506e-09, 1.0654e-06, -6.4522e-06], [ 6.3516e-07, 2.4345e-06, 6.9812e-06, ..., 4.0047e-08, 7.0706e-06, 1.9763e-06], [-2.0508e-06, -1.6764e-06, 1.0077e-06, ..., 1.0245e-08, -2.5164e-06, -2.0768e-06]], device='cuda:0') Epoch 141, bias, value: tensor([ 0.0275, -0.0205, -0.0078, -0.0022, 0.0072, 0.0020, -0.0192, 0.0016, 0.0250, 0.0179], device='cuda:0'), grad: tensor([ 4.7460e-06, -6.9924e-06, 2.0750e-06, 8.4713e-06, 6.5081e-06, -2.5049e-05, 1.4566e-05, -2.3186e-05, 1.9163e-05, -3.6135e-07], device='cuda:0') 100 0.0001 changing lr epoch 140, time 257.00, cls_loss 0.0024 cls_loss_mapping 0.0059 cls_loss_causal 0.5502 re_mapping 0.0079 re_causal 0.0230 /// teacc 98.95 lr 0.00010000 Epoch 142, weight, value: tensor([[ 0.0326, -0.0865, -0.0952, ..., -0.0323, -0.1542, -0.0132], [-0.0926, -0.1267, -0.1245, ..., -0.1087, -0.0645, -0.0018], [ 0.0719, 0.0534, -0.1232, ..., -0.0965, -0.1204, -0.0674], ..., [ 0.0134, 0.0378, -0.0521, ..., -0.0236, -0.1225, 0.0496], [ 0.0113, 0.1048, 0.0847, ..., -0.0771, 0.0698, -0.1185], [-0.0452, -0.0065, -0.0306, ..., -0.0295, -0.0581, 0.0470]], device='cuda:0'), grad: tensor([[ 3.5055e-06, 4.0904e-06, 1.5162e-06, ..., 1.0850e-06, 1.6959e-06, 1.6578e-07], [ 2.4494e-07, 6.2585e-07, 2.3749e-07, ..., 1.3411e-07, 3.6880e-07, -5.2154e-08], [ 6.6683e-07, -4.1164e-06, 2.0396e-07, ..., 5.6811e-08, -1.7043e-07, 8.5682e-08], ..., [ 2.4531e-06, 3.2131e-06, 3.3341e-07, ..., 6.5193e-09, 6.1095e-07, 1.1735e-07], [ 1.1642e-07, 6.4168e-07, 2.5313e-06, ..., 4.0699e-07, 2.5053e-06, 3.4459e-08], [-1.2293e-05, -1.0923e-05, 7.8790e-07, ..., 2.5146e-08, 1.4836e-06, -5.4948e-07]], device='cuda:0') Epoch 142, bias, value: tensor([ 0.0284, -0.0209, -0.0086, -0.0024, 0.0094, 0.0019, -0.0191, 0.0028, 0.0247, 0.0158], device='cuda:0'), grad: tensor([ 1.8433e-05, 2.6859e-06, -8.1956e-06, 1.0431e-05, -2.5705e-06, -7.8380e-06, -6.7614e-07, 1.4670e-05, 8.2329e-06, -3.5197e-05], device='cuda:0') 100 0.0001 changing lr epoch 141, time 256.90, cls_loss 0.0031 cls_loss_mapping 0.0071 cls_loss_causal 0.5561 re_mapping 0.0074 re_causal 0.0213 /// teacc 98.99 lr 0.00010000 Epoch 143, weight, value: tensor([[ 0.0327, -0.0869, -0.0961, ..., -0.0327, -0.1552, -0.0136], [-0.0928, -0.1264, -0.1247, ..., -0.1093, -0.0646, -0.0009], [ 0.0729, 0.0534, -0.1259, ..., -0.0979, -0.1198, -0.0684], ..., [ 0.0133, 0.0380, -0.0528, ..., -0.0237, -0.1225, 0.0493], [ 0.0116, 0.1053, 0.0860, ..., -0.0755, 0.0705, -0.1189], [-0.0450, -0.0066, -0.0314, ..., -0.0296, -0.0586, 0.0471]], device='cuda:0'), grad: tensor([[ 8.4937e-07, 1.4780e-06, 8.0373e-07, ..., 4.0699e-07, 1.3392e-06, 5.4017e-07], [ 4.5821e-07, 1.6363e-06, 5.7742e-07, ..., 1.2387e-07, 1.3467e-06, 3.6322e-07], [-3.1162e-06, -3.1237e-06, 2.6543e-07, ..., 1.3039e-07, -1.7872e-06, 5.4482e-07], ..., [-6.4448e-07, -4.6492e-06, 1.7742e-06, ..., 7.4506e-09, 2.6170e-06, -9.6019e-07], [ 1.0999e-06, 5.2005e-06, 2.0713e-06, ..., 7.3854e-07, 4.4145e-06, 1.2070e-06], [-3.7253e-07, 2.0955e-07, 3.7514e-06, ..., 5.7742e-08, 6.3963e-06, 1.4259e-06]], device='cuda:0') Epoch 143, bias, value: tensor([ 0.0282, -0.0190, -0.0093, -0.0029, 0.0089, 0.0014, -0.0191, 0.0025, 0.0249, 0.0158], device='cuda:0'), grad: tensor([ 8.2925e-06, 5.7705e-06, -2.5496e-05, -2.2829e-05, 5.2005e-06, 7.4394e-06, -5.3868e-06, -4.8894e-07, 1.6272e-05, 1.1191e-05], device='cuda:0') 100 0.0001 changing lr epoch 142, time 257.07, cls_loss 0.0021 cls_loss_mapping 0.0052 cls_loss_causal 0.5403 re_mapping 0.0071 re_causal 0.0215 /// teacc 98.94 lr 0.00010000 Epoch 144, weight, value: tensor([[ 0.0329, -0.0872, -0.0968, ..., -0.0330, -0.1559, -0.0133], [-0.0931, -0.1266, -0.1258, ..., -0.1100, -0.0660, -0.0007], [ 0.0731, 0.0536, -0.1271, ..., -0.0981, -0.1195, -0.0686], ..., [ 0.0133, 0.0382, -0.0538, ..., -0.0236, -0.1226, 0.0494], [ 0.0117, 0.1053, 0.0858, ..., -0.0761, 0.0701, -0.1198], [-0.0452, -0.0070, -0.0326, ..., -0.0297, -0.0592, 0.0471]], device='cuda:0'), grad: tensor([[ 7.0781e-08, 4.5914e-07, 2.5239e-07, ..., 1.2852e-07, 2.2352e-07, 1.7695e-07], [ 1.2014e-07, 5.2620e-07, 2.8871e-07, ..., 1.9558e-08, 3.0641e-07, -5.0012e-07], [ 2.4214e-07, 1.1390e-06, 4.1537e-07, ..., 3.5390e-08, 4.4331e-07, 5.3924e-07], ..., [-7.2550e-07, -3.6806e-06, 3.6415e-07, ..., 0.0000e+00, 3.8929e-07, -8.0653e-07], [ 2.6822e-07, 1.6708e-06, 3.0715e-06, ..., 2.7381e-07, 3.1758e-06, 3.8370e-07], [-9.4529e-07, -6.3702e-07, 1.3681e-06, ..., 6.5193e-09, 1.4566e-06, -1.3355e-06]], device='cuda:0') Epoch 144, bias, value: tensor([ 0.0283, -0.0194, -0.0084, -0.0028, 0.0089, 0.0014, -0.0190, 0.0027, 0.0245, 0.0154], device='cuda:0'), grad: tensor([ 7.2736e-07, -1.9558e-07, 2.4661e-06, -9.1866e-06, -1.2510e-05, 2.4587e-06, 5.4948e-07, -9.2387e-07, 6.4000e-06, 1.0192e-05], device='cuda:0') 100 0.0001 changing lr epoch 143, time 256.97, cls_loss 0.0024 cls_loss_mapping 0.0059 cls_loss_causal 0.5401 re_mapping 0.0071 re_causal 0.0209 /// teacc 98.96 lr 0.00010000 Epoch 145, weight, value: tensor([[ 0.0331, -0.0879, -0.0990, ..., -0.0343, -0.1574, -0.0138], [-0.0934, -0.1268, -0.1260, ..., -0.1104, -0.0660, -0.0004], [ 0.0734, 0.0538, -0.1281, ..., -0.0983, -0.1196, -0.0689], ..., [ 0.0133, 0.0382, -0.0546, ..., -0.0238, -0.1227, 0.0493], [ 0.0117, 0.1053, 0.0861, ..., -0.0765, 0.0703, -0.1206], [-0.0451, -0.0067, -0.0334, ..., -0.0301, -0.0597, 0.0467]], device='cuda:0'), grad: tensor([[ 2.7847e-06, 1.3430e-06, 8.1956e-08, ..., 5.0291e-08, 7.3574e-08, 2.6058e-06], [ 2.0303e-07, 2.5947e-06, 1.0896e-07, ..., 3.7253e-09, 1.0803e-07, -1.0327e-05], [-3.3155e-07, -2.6524e-06, 1.6391e-07, ..., 3.7253e-09, 1.4901e-07, 4.6939e-06], ..., [ 1.9178e-05, 8.0243e-06, 2.0303e-07, ..., 0.0000e+00, 1.8347e-07, 2.0221e-05], [-4.0233e-07, -1.0729e-06, -2.0023e-06, ..., 1.9558e-08, -1.6028e-06, 4.8801e-07], [-2.3156e-05, -1.0513e-05, 1.7043e-07, ..., 9.3132e-10, 1.4808e-07, -2.1279e-05]], device='cuda:0') Epoch 145, bias, value: tensor([ 0.0281, -0.0193, -0.0081, -0.0028, 0.0101, 0.0015, -0.0186, 0.0025, 0.0242, 0.0144], device='cuda:0'), grad: tensor([ 8.6948e-06, -6.3896e-05, 2.3529e-05, 3.4794e-06, 8.7842e-06, 6.5081e-06, 1.4948e-06, 8.0347e-05, 6.3982e-07, -6.9499e-05], device='cuda:0') 100 0.0001 changing lr epoch 144, time 256.94, cls_loss 0.0023 cls_loss_mapping 0.0059 cls_loss_causal 0.5761 re_mapping 0.0072 re_causal 0.0222 /// teacc 98.81 lr 0.00010000 Epoch 146, weight, value: tensor([[ 0.0334, -0.0882, -0.0995, ..., -0.0346, -0.1580, -0.0138], [-0.0937, -0.1268, -0.1263, ..., -0.1107, -0.0662, -0.0003], [ 0.0728, 0.0527, -0.1285, ..., -0.0986, -0.1198, -0.0693], ..., [ 0.0140, 0.0391, -0.0552, ..., -0.0238, -0.1227, 0.0494], [ 0.0117, 0.1053, 0.0860, ..., -0.0770, 0.0702, -0.1210], [-0.0453, -0.0066, -0.0345, ..., -0.0302, -0.0599, 0.0468]], device='cuda:0'), grad: tensor([[-2.7567e-06, 8.9500e-07, 7.9628e-07, ..., 1.3318e-07, 8.7731e-07, 1.8626e-08], [ 4.7777e-07, 1.1232e-06, 1.0096e-06, ..., 5.7742e-08, 1.1316e-06, 1.8626e-09], [ 5.8673e-08, 1.4476e-05, 1.3523e-05, ..., 3.8184e-08, 1.5102e-05, 1.6764e-08], ..., [ 1.0803e-07, 5.5507e-06, 5.1893e-06, ..., 1.8626e-09, 5.7966e-06, 8.0094e-08], [ 2.6915e-07, 1.3165e-05, 1.2331e-05, ..., 9.4995e-08, 1.3709e-05, 1.2293e-07], [ 1.0664e-06, 2.3376e-07, 1.9297e-06, ..., 2.8871e-08, 2.1383e-06, -7.7486e-07]], device='cuda:0') Epoch 146, bias, value: tensor([ 0.0277, -0.0191, -0.0095, -0.0028, 0.0101, 0.0017, -0.0185, 0.0035, 0.0238, 0.0146], device='cuda:0'), grad: tensor([-4.3124e-05, 1.6361e-05, 1.8179e-05, -1.4150e-04, -4.5002e-06, 5.8770e-05, 6.1952e-06, 2.5228e-05, 4.6164e-05, 1.8224e-05], device='cuda:0') 100 0.0001 changing lr epoch 145, time 255.67, cls_loss 0.0029 cls_loss_mapping 0.0066 cls_loss_causal 0.5153 re_mapping 0.0069 re_causal 0.0196 /// teacc 99.02 lr 0.00010000 Epoch 147, weight, value: tensor([[ 3.3642e-02, -8.8174e-02, -1.0058e-01, ..., -3.4658e-02, -1.5901e-01, -1.3827e-02], [-9.4089e-02, -1.2732e-01, -1.2684e-01, ..., -1.1022e-01, -6.6740e-02, -5.8788e-05], [ 7.3148e-02, 5.2653e-02, -1.2904e-01, ..., -9.8792e-02, -1.2040e-01, -6.9431e-02], ..., [ 1.3571e-02, 3.9036e-02, -5.6408e-02, ..., -2.3869e-02, -1.2401e-01, 4.9015e-02], [ 1.1777e-02, 1.0562e-01, 8.6505e-02, ..., -7.6968e-02, 7.0496e-02, -1.2166e-01], [-4.5356e-02, -6.4570e-03, -3.6120e-02, ..., -3.0314e-02, -6.1014e-02, 4.7254e-02]], device='cuda:0'), grad: tensor([[-1.1735e-06, 1.2852e-06, 1.2834e-06, ..., 6.7335e-07, 1.0589e-06, 6.7055e-08], [ 3.8464e-07, 3.7309e-06, 4.0885e-07, ..., 1.4435e-07, 3.8184e-07, -1.1638e-05], [-1.8468e-06, -1.0608e-06, 2.5146e-07, ..., 4.4703e-08, -6.0257e-07, 6.6049e-06], ..., [-9.3319e-07, -4.5896e-06, 1.0859e-06, ..., 6.5193e-09, 7.1619e-07, 2.3656e-06], [ 9.8906e-07, 3.5651e-06, 1.1273e-05, ..., 2.2352e-06, 7.0482e-06, 9.6858e-08], [ 1.1167e-06, 2.5146e-07, 1.9576e-06, ..., 1.1548e-07, 1.0347e-06, 9.2201e-08]], device='cuda:0') Epoch 147, bias, value: tensor([ 0.0293, -0.0199, -0.0101, -0.0015, 0.0102, 0.0014, -0.0192, 0.0035, 0.0240, 0.0144], device='cuda:0'), grad: tensor([-1.6503e-06, -5.6028e-05, 2.7627e-05, 1.2852e-05, -3.1292e-05, -2.0713e-05, -3.2447e-06, 4.2647e-05, 2.0862e-05, 8.8662e-06], device='cuda:0') 100 0.0001 changing lr epoch 146, time 256.98, cls_loss 0.0021 cls_loss_mapping 0.0045 cls_loss_causal 0.5277 re_mapping 0.0069 re_causal 0.0200 /// teacc 99.03 lr 0.00010000 Epoch 148, weight, value: tensor([[ 0.0345, -0.0885, -0.1015, ..., -0.0355, -0.1596, -0.0142], [-0.0947, -0.1283, -0.1272, ..., -0.1105, -0.0671, -0.0008], [ 0.0739, 0.0531, -0.1294, ..., -0.0994, -0.1206, -0.0694], ..., [ 0.0132, 0.0391, -0.0574, ..., -0.0239, -0.1241, 0.0495], [ 0.0115, 0.1055, 0.0859, ..., -0.0779, 0.0701, -0.1222], [-0.0453, -0.0056, -0.0364, ..., -0.0304, -0.0615, 0.0483]], device='cuda:0'), grad: tensor([[ 4.6566e-08, 5.5227e-07, 5.1782e-07, ..., 3.1572e-07, 5.9791e-07, -2.1607e-07], [ 3.5577e-07, 1.5274e-07, 1.8161e-07, ..., 8.3819e-09, 2.2911e-07, -2.1011e-05], [ 1.5553e-07, 6.5193e-09, 9.3319e-07, ..., 7.8231e-08, 1.2210e-06, 9.2983e-06], ..., [ 1.0245e-08, -1.2387e-07, 1.3970e-07, ..., 0.0000e+00, 1.7323e-07, 8.6650e-06], [-1.2945e-06, -1.3700e-06, -1.7509e-07, ..., 4.3772e-08, -6.9756e-07, 4.2375e-07], [ 9.9652e-08, 7.2643e-08, 2.4401e-07, ..., 2.7940e-09, 2.6729e-07, 2.6543e-07]], device='cuda:0') Epoch 148, bias, value: tensor([ 0.0295, -0.0207, -0.0095, -0.0017, 0.0101, 0.0018, -0.0188, 0.0036, 0.0234, 0.0147], device='cuda:0'), grad: tensor([-1.8440e-07, -1.0800e-04, 4.4465e-05, -3.1684e-06, 9.3356e-06, 4.8280e-06, 4.2059e-06, 4.5836e-05, 7.7765e-07, 1.9092e-06], device='cuda:0') 100 0.0001 changing lr epoch 147, time 257.16, cls_loss 0.0020 cls_loss_mapping 0.0055 cls_loss_causal 0.5517 re_mapping 0.0066 re_causal 0.0202 /// teacc 98.93 lr 0.00010000 Epoch 149, weight, value: tensor([[ 0.0347, -0.0888, -0.1026, ..., -0.0363, -0.1602, -0.0146], [-0.0952, -0.1286, -0.1278, ..., -0.1110, -0.0676, -0.0003], [ 0.0741, 0.0532, -0.1303, ..., -0.1000, -0.1211, -0.0700], ..., [ 0.0133, 0.0391, -0.0585, ..., -0.0241, -0.1242, 0.0492], [ 0.0115, 0.1048, 0.0844, ..., -0.0806, 0.0686, -0.1229], [-0.0453, -0.0055, -0.0367, ..., -0.0306, -0.0616, 0.0486]], device='cuda:0'), grad: tensor([[ 5.0571e-07, 2.2817e-07, 4.0978e-07, ..., 2.0955e-07, 3.1199e-07, 4.9267e-07], [ 3.6042e-07, 9.4064e-08, 1.3784e-07, ..., 1.9558e-08, 1.2107e-07, -5.5768e-06], [-3.7253e-08, 1.9558e-08, 1.1921e-07, ..., 3.2596e-08, 1.6671e-07, 1.0943e-06], ..., [-4.3139e-06, -2.6543e-07, 3.2783e-07, ..., 4.6566e-09, 1.3690e-07, -6.1188e-07], [-1.7416e-07, -5.9605e-08, 1.6391e-07, ..., 2.1886e-07, -2.5891e-07, 6.1467e-07], [ 3.0119e-06, 2.3469e-07, 2.6766e-06, ..., 2.2352e-08, 6.8825e-07, 2.7344e-06]], device='cuda:0') Epoch 149, bias, value: tensor([ 0.0295, -0.0204, -0.0097, -0.0019, 0.0100, 0.0022, -0.0173, 0.0032, 0.0223, 0.0148], device='cuda:0'), grad: tensor([ 2.2855e-06, -1.7434e-05, 2.9206e-06, 7.0669e-06, 3.6731e-06, -1.0647e-05, 4.8988e-07, -3.9078e-06, 1.9148e-06, 1.3612e-05], device='cuda:0') 100 0.0001 changing lr epoch 148, time 257.00, cls_loss 0.0025 cls_loss_mapping 0.0055 cls_loss_causal 0.5375 re_mapping 0.0066 re_causal 0.0201 /// teacc 99.04 lr 0.00010000 Epoch 150, weight, value: tensor([[ 0.0349, -0.0890, -0.1034, ..., -0.0370, -0.1609, -0.0147], [-0.0958, -0.1292, -0.1298, ..., -0.1128, -0.0687, -0.0005], [ 0.0740, 0.0526, -0.1314, ..., -0.1004, -0.1215, -0.0703], ..., [ 0.0134, 0.0397, -0.0595, ..., -0.0244, -0.1243, 0.0493], [ 0.0115, 0.1044, 0.0820, ..., -0.0833, 0.0673, -0.1242], [-0.0449, -0.0050, -0.0372, ..., -0.0308, -0.0608, 0.0498]], device='cuda:0'), grad: tensor([[ 1.4054e-06, 6.5472e-07, 5.3458e-07, ..., 2.4214e-07, 5.6718e-07, 1.1921e-06], [ 1.4063e-06, 9.1735e-07, 6.8732e-07, ..., 6.5193e-09, 8.3633e-07, 1.8440e-07], [ 1.7844e-06, 1.8533e-06, 1.1148e-06, ..., 1.1176e-08, 1.4100e-06, 1.3616e-06], ..., [-8.0168e-06, -2.7940e-06, 5.2992e-07, ..., 0.0000e+00, 7.2829e-07, -6.8098e-06], [-3.2429e-06, -7.9647e-06, -9.7677e-06, ..., 2.2352e-08, -1.1154e-05, 2.2817e-07], [ 9.1828e-07, 1.3802e-06, 1.1893e-06, ..., 4.6566e-09, 1.4324e-06, 6.6962e-07]], device='cuda:0') Epoch 150, bias, value: tensor([ 0.0294, -0.0203, -0.0106, -0.0017, 0.0095, 0.0038, -0.0182, 0.0035, 0.0207, 0.0159], device='cuda:0'), grad: tensor([ 6.8247e-06, 5.4128e-06, 9.7081e-06, 1.3880e-05, -9.4473e-05, 1.2025e-05, 5.3458e-06, -2.0817e-05, -1.8567e-05, 8.0585e-05], device='cuda:0') 100 0.0001 changing lr epoch 149, time 257.05, cls_loss 0.0017 cls_loss_mapping 0.0046 cls_loss_causal 0.5280 re_mapping 0.0065 re_causal 0.0204 /// teacc 99.04 lr 0.00010000 Epoch 151, weight, value: tensor([[ 0.0349, -0.0894, -0.1043, ..., -0.0376, -0.1619, -0.0152], [-0.0963, -0.1293, -0.1305, ..., -0.1134, -0.0692, -0.0002], [ 0.0741, 0.0526, -0.1324, ..., -0.1007, -0.1220, -0.0709], ..., [ 0.0135, 0.0397, -0.0605, ..., -0.0246, -0.1244, 0.0493], [ 0.0116, 0.1044, 0.0817, ..., -0.0837, 0.0670, -0.1245], [-0.0449, -0.0049, -0.0383, ..., -0.0308, -0.0606, 0.0501]], device='cuda:0'), grad: tensor([[-3.3766e-05, 4.6659e-07, -5.9195e-06, ..., 1.5460e-07, -4.0412e-05, 4.7497e-08], [ 1.9409e-06, 3.9153e-06, 2.8852e-06, ..., 1.0543e-06, 2.5816e-06, -2.2203e-06], [ 4.3251e-06, 1.3802e-06, 1.8859e-06, ..., 4.3679e-07, 6.1803e-06, 3.9488e-07], ..., [ 3.4366e-07, 3.6787e-07, 5.4110e-07, ..., 1.7416e-07, 1.1493e-06, 3.6974e-07], [ 9.3654e-06, 1.5646e-07, -5.1260e-06, ..., -2.4121e-06, 1.0036e-05, 1.5106e-06], [-1.6931e-06, -1.1839e-05, -2.9244e-07, ..., 5.4948e-08, -4.8913e-06, -1.0794e-06]], device='cuda:0') Epoch 151, bias, value: tensor([ 0.0293, -0.0201, -0.0109, -0.0014, 0.0098, 0.0036, -0.0176, 0.0033, 0.0204, 0.0158], device='cuda:0'), grad: tensor([-1.1760e-04, 1.0049e-06, 1.7658e-05, 5.9634e-05, 1.9502e-06, 7.4618e-06, 4.6529e-06, 3.7067e-06, 3.8594e-05, -1.7092e-05], device='cuda:0') 100 0.0001 changing lr epoch 150, time 256.29, cls_loss 0.0024 cls_loss_mapping 0.0054 cls_loss_causal 0.5286 re_mapping 0.0064 re_causal 0.0188 /// teacc 98.96 lr 0.00010000 Epoch 152, weight, value: tensor([[ 0.0352, -0.0895, -0.1050, ..., -0.0379, -0.1619, -0.0148], [-0.0971, -0.1295, -0.1312, ..., -0.1132, -0.0697, 0.0005], [ 0.0741, 0.0526, -0.1338, ..., -0.1017, -0.1225, -0.0713], ..., [ 0.0136, 0.0397, -0.0616, ..., -0.0245, -0.1245, 0.0488], [ 0.0118, 0.1043, 0.0815, ..., -0.0843, 0.0669, -0.1249], [-0.0452, -0.0046, -0.0389, ..., -0.0307, -0.0605, 0.0505]], device='cuda:0'), grad: tensor([[ 1.9614e-06, 7.9125e-06, 8.9332e-06, ..., 2.6934e-06, 6.7577e-06, 5.1595e-07], [ 7.3314e-06, 1.0356e-05, 4.0233e-07, ..., 1.7509e-07, 2.6077e-07, 3.5819e-06], [-1.3554e-04, -1.8990e-04, 3.0920e-06, ..., 1.9316e-06, 1.3839e-06, -7.7784e-05], ..., [ 1.1742e-04, 1.6725e-04, 1.1269e-07, ..., 1.3970e-08, 1.0338e-07, 6.7472e-05], [ 5.5470e-06, 9.5218e-06, 1.8582e-05, ..., 2.6543e-07, 1.9029e-05, 1.2806e-06], [ 5.3756e-06, 7.4878e-06, 4.8522e-07, ..., 4.1910e-08, 4.5728e-07, 2.7362e-06]], device='cuda:0') Epoch 152, bias, value: tensor([ 0.0295, -0.0198, -0.0112, -0.0014, 0.0101, 0.0036, -0.0176, 0.0030, 0.0203, 0.0158], device='cuda:0'), grad: tensor([ 2.6122e-05, 1.9804e-05, -3.4142e-04, -5.6803e-05, 9.1419e-06, 1.8418e-05, -4.1127e-05, 3.1137e-04, 3.9518e-05, 1.4499e-05], device='cuda:0') 100 0.0001 changing lr epoch 151, time 256.92, cls_loss 0.0018 cls_loss_mapping 0.0036 cls_loss_causal 0.5055 re_mapping 0.0064 re_causal 0.0193 /// teacc 98.93 lr 0.00010000 Epoch 153, weight, value: tensor([[ 0.0356, -0.0898, -0.1061, ..., -0.0379, -0.1627, -0.0155], [-0.0976, -0.1296, -0.1315, ..., -0.1133, -0.0699, 0.0009], [ 0.0748, 0.0530, -0.1346, ..., -0.1020, -0.1227, -0.0713], ..., [ 0.0134, 0.0398, -0.0624, ..., -0.0247, -0.1244, 0.0487], [ 0.0118, 0.1043, 0.0816, ..., -0.0843, 0.0670, -0.1255], [-0.0452, -0.0048, -0.0392, ..., -0.0308, -0.0608, 0.0506]], device='cuda:0'), grad: tensor([[ 1.6019e-07, 8.2422e-07, 2.8498e-07, ..., 4.2841e-07, 2.2165e-07, -3.8054e-06], [ 1.9930e-07, 1.0943e-06, 4.5449e-07, ..., 5.5041e-07, 3.7998e-07, 7.9907e-07], [-3.1944e-07, 5.1409e-07, 1.1735e-07, ..., 8.8662e-07, 1.4622e-07, 2.6077e-07], ..., [ 7.9721e-07, 1.8105e-06, 1.7043e-07, ..., 1.0617e-07, 1.7136e-07, 4.7032e-07], [ 3.1758e-07, 5.4799e-06, 2.1383e-05, ..., 1.5926e-07, 1.1884e-05, 3.5856e-07], [-1.7760e-06, -4.0308e-06, 8.8476e-07, ..., 4.7404e-07, 5.3924e-07, -9.7789e-07]], device='cuda:0') Epoch 153, bias, value: tensor([ 0.0294, -0.0196, -0.0108, -0.0012, 0.0116, 0.0034, -0.0176, 0.0027, 0.0202, 0.0146], device='cuda:0'), grad: tensor([-3.0503e-05, 1.1660e-05, 4.5970e-06, 6.4746e-06, -4.7088e-05, -2.9683e-05, 5.2452e-05, 5.5023e-06, 2.9013e-05, -2.3916e-06], device='cuda:0') 100 0.0001 changing lr epoch 152, time 256.64, cls_loss 0.0025 cls_loss_mapping 0.0052 cls_loss_causal 0.5423 re_mapping 0.0066 re_causal 0.0197 /// teacc 98.97 lr 0.00010000 Epoch 154, weight, value: tensor([[ 0.0359, -0.0902, -0.1067, ..., -0.0395, -0.1634, -0.0155], [-0.0979, -0.1298, -0.1330, ..., -0.1136, -0.0715, 0.0009], [ 0.0753, 0.0531, -0.1353, ..., -0.1021, -0.1231, -0.0717], ..., [ 0.0132, 0.0398, -0.0634, ..., -0.0249, -0.1246, 0.0487], [ 0.0118, 0.1045, 0.0817, ..., -0.0843, 0.0673, -0.1260], [-0.0453, -0.0027, -0.0404, ..., -0.0314, -0.0612, 0.0519]], device='cuda:0'), grad: tensor([[ 1.5460e-07, 3.0082e-07, 2.0768e-07, ..., 6.5193e-08, 1.6857e-07, 5.2154e-08], [ 8.4378e-07, 1.7285e-06, 5.5321e-07, ..., -6.9849e-08, 6.6310e-07, -3.1292e-07], [-8.4192e-07, -4.6566e-09, 2.4494e-07, ..., 2.5146e-08, 3.5763e-07, 6.7055e-08], ..., [ 4.0568e-06, 9.2536e-06, 2.0154e-06, ..., 2.2352e-08, 2.6040e-06, 2.1718e-06], [-7.0743e-06, -1.6347e-05, -4.2282e-06, ..., 8.6613e-08, -5.9605e-06, 9.8720e-08], [ 1.7602e-07, -6.0722e-07, 8.6427e-07, ..., 2.7008e-08, 9.1922e-07, -2.4065e-06]], device='cuda:0') Epoch 154, bias, value: tensor([ 0.0292, -0.0196, -0.0109, -0.0020, 0.0098, 0.0050, -0.0182, 0.0025, 0.0202, 0.0160], device='cuda:0'), grad: tensor([ 5.7928e-07, 2.4643e-06, -9.6858e-07, 1.1295e-05, -1.7136e-07, -2.8521e-05, 2.5347e-05, 1.8656e-05, -2.6822e-05, -1.9185e-06], device='cuda:0') 100 0.0001 changing lr epoch 153, time 257.01, cls_loss 0.0031 cls_loss_mapping 0.0058 cls_loss_causal 0.5374 re_mapping 0.0065 re_causal 0.0189 /// teacc 99.02 lr 0.00010000 Epoch 155, weight, value: tensor([[ 3.6097e-02, -9.0778e-02, -1.0780e-01, ..., -4.0214e-02, -1.6465e-01, -1.5980e-02], [-9.8934e-02, -1.3101e-01, -1.3551e-01, ..., -1.1504e-01, -7.4972e-02, 6.9601e-05], [ 7.5909e-02, 5.3494e-02, -1.3598e-01, ..., -1.0218e-01, -1.2304e-01, -7.2211e-02], ..., [ 1.3141e-02, 3.9604e-02, -6.4776e-02, ..., -2.5000e-02, -1.2482e-01, 4.8568e-02], [ 1.2561e-02, 1.0596e-01, 8.3668e-02, ..., -8.4157e-02, 6.8812e-02, -1.2737e-01], [-4.5115e-02, -2.3898e-03, -4.0734e-02, ..., -3.1698e-02, -5.8896e-02, 5.4345e-02]], device='cuda:0'), grad: tensor([[-1.0338e-07, 4.4703e-08, 1.5739e-07, ..., 4.6566e-09, 1.9278e-07, 9.3132e-10], [ 1.3132e-07, 3.1199e-07, 1.0468e-06, ..., -1.6764e-08, 1.3895e-06, -1.8161e-07], [-1.8394e-06, -1.8794e-06, 5.6624e-07, ..., 2.7940e-09, 7.4226e-07, 5.4948e-08], ..., [ 1.4324e-06, 1.0375e-06, 1.1344e-06, ..., 6.5193e-09, 1.4799e-06, -3.0734e-08], [ 6.3330e-08, 2.3283e-08, 4.6566e-06, ..., 9.3132e-10, 5.4576e-06, 7.4506e-09], [ 1.5926e-07, 2.4680e-07, 1.1753e-06, ..., 5.5879e-09, 1.4743e-06, 1.1362e-07]], device='cuda:0') Epoch 155, bias, value: tensor([ 0.0292, -0.0221, -0.0094, -0.0028, 0.0097, 0.0038, -0.0180, 0.0018, 0.0221, 0.0177], device='cuda:0'), grad: tensor([ 1.6019e-07, 3.3975e-06, -3.2447e-06, -1.0502e-04, 1.0803e-07, 8.0764e-05, 2.1700e-07, 6.6459e-06, 1.2673e-05, 4.1872e-06], device='cuda:0') 100 0.0001 changing lr epoch 154, time 256.80, cls_loss 0.0024 cls_loss_mapping 0.0051 cls_loss_causal 0.5343 re_mapping 0.0067 re_causal 0.0198 /// teacc 98.88 lr 0.00010000 Epoch 156, weight, value: tensor([[ 0.0363, -0.0913, -0.1092, ..., -0.0413, -0.1655, -0.0162], [-0.1008, -0.1322, -0.1371, ..., -0.1161, -0.0768, 0.0003], [ 0.0761, 0.0533, -0.1371, ..., -0.1029, -0.1235, -0.0727], ..., [ 0.0137, 0.0401, -0.0659, ..., -0.0251, -0.1249, 0.0487], [ 0.0129, 0.1065, 0.0841, ..., -0.0840, 0.0694, -0.1278], [-0.0455, -0.0028, -0.0418, ..., -0.0320, -0.0592, 0.0540]], device='cuda:0'), grad: tensor([[ 6.9197e-07, 2.5779e-06, 1.5851e-06, ..., 1.1353e-06, 1.2834e-06, 1.2387e-07], [ 1.7192e-06, 2.0321e-06, 1.9465e-07, ..., 1.2666e-07, 1.5181e-07, -1.8626e-07], [ 4.6223e-05, 7.8604e-06, 5.8021e-07, ..., 2.6077e-07, 1.2154e-06, 2.6058e-06], ..., [-1.1347e-05, -3.3319e-05, 4.7218e-07, ..., 1.2107e-08, -1.3504e-07, -1.3784e-05], [ 1.6969e-06, 3.9004e-06, 2.1663e-06, ..., 1.6075e-06, 1.7593e-06, 2.5425e-07], [ 8.7097e-06, 4.7535e-06, 4.5169e-07, ..., 1.4249e-07, 1.7136e-07, 6.0238e-06]], device='cuda:0') Epoch 156, bias, value: tensor([ 0.0291, -0.0227, -0.0100, -0.0026, 0.0097, 0.0035, -0.0171, 0.0023, 0.0232, 0.0172], device='cuda:0'), grad: tensor([ 7.1265e-06, 5.0627e-06, 1.9097e-04, 4.6045e-05, -2.7943e-04, 1.6868e-05, -1.6823e-05, -3.3617e-05, 1.3895e-05, 4.9442e-05], device='cuda:0') 100 0.0001 changing lr epoch 155, time 257.26, cls_loss 0.0018 cls_loss_mapping 0.0047 cls_loss_causal 0.5349 re_mapping 0.0067 re_causal 0.0208 /// teacc 98.98 lr 0.00010000 Epoch 157, weight, value: tensor([[ 0.0361, -0.0916, -0.1099, ..., -0.0417, -0.1659, -0.0163], [-0.1011, -0.1323, -0.1375, ..., -0.1161, -0.0771, 0.0009], [ 0.0762, 0.0533, -0.1377, ..., -0.1030, -0.1238, -0.0734], ..., [ 0.0137, 0.0402, -0.0669, ..., -0.0252, -0.1249, 0.0484], [ 0.0131, 0.1067, 0.0840, ..., -0.0840, 0.0695, -0.1281], [-0.0450, -0.0028, -0.0415, ..., -0.0320, -0.0595, 0.0542]], device='cuda:0'), grad: tensor([[-4.9919e-06, -4.1723e-07, 1.3039e-07, ..., 1.3970e-08, 1.2852e-07, -4.4610e-07], [ 3.8557e-07, -1.5169e-05, 3.1944e-07, ..., 1.8626e-09, 3.4366e-07, -7.2345e-06], [-4.9081e-07, 4.7870e-06, 3.4086e-07, ..., 4.6566e-09, 5.0012e-07, 2.4922e-06], ..., [ 5.2154e-07, 3.1739e-06, 1.1548e-07, ..., 0.0000e+00, 1.5926e-07, 1.5628e-06], [ 1.2945e-07, 6.5751e-07, -2.8778e-06, ..., 9.3132e-09, -2.5742e-06, 1.1660e-06], [ 1.2964e-06, 9.4902e-07, 3.1758e-07, ..., 1.8626e-09, 3.7905e-07, 3.4552e-07]], device='cuda:0') Epoch 157, bias, value: tensor([ 0.0289, -0.0222, -0.0103, -0.0026, 0.0095, 0.0039, -0.0173, 0.0021, 0.0232, 0.0173], device='cuda:0'), grad: tensor([-1.5303e-05, -1.1456e-04, 4.0293e-05, 1.4221e-06, 1.8552e-06, 1.1958e-05, 2.2829e-05, 2.4483e-05, 1.8507e-05, 8.6799e-06], device='cuda:0') 100 0.0001 changing lr epoch 156, time 255.44, cls_loss 0.0017 cls_loss_mapping 0.0040 cls_loss_causal 0.5325 re_mapping 0.0060 re_causal 0.0194 /// teacc 98.98 lr 0.00010000 Epoch 158, weight, value: tensor([[ 0.0370, -0.0923, -0.1110, ..., -0.0422, -0.1668, -0.0160], [-0.1021, -0.1326, -0.1379, ..., -0.1164, -0.0774, 0.0020], [ 0.0766, 0.0536, -0.1381, ..., -0.1030, -0.1239, -0.0736], ..., [ 0.0150, 0.0412, -0.0675, ..., -0.0252, -0.1250, 0.0478], [ 0.0132, 0.1067, 0.0839, ..., -0.0842, 0.0694, -0.1285], [-0.0472, -0.0041, -0.0421, ..., -0.0321, -0.0599, 0.0536]], device='cuda:0'), grad: tensor([[-3.9861e-07, 2.2165e-07, 5.6811e-08, ..., 3.9116e-08, 6.4261e-08, 3.2596e-08], [ 9.0338e-07, 2.5425e-06, 5.2154e-08, ..., -7.4506e-09, 1.2759e-07, 9.6019e-07], [ 5.1223e-08, 2.7195e-07, 6.3330e-08, ..., 7.4506e-09, 1.0990e-07, 2.9430e-07], ..., [-2.2314e-06, -6.9849e-06, 5.9605e-08, ..., 2.7940e-09, -3.7253e-09, -3.7886e-06], [ 2.6915e-07, 3.0827e-07, 4.0047e-08, ..., 6.8918e-08, 1.2014e-07, 3.5297e-07], [ 7.0315e-07, 2.6915e-06, 1.7695e-08, ..., 2.7940e-09, 1.3877e-07, 1.2843e-06]], device='cuda:0') Epoch 158, bias, value: tensor([ 0.0290, -0.0215, -0.0100, -0.0027, 0.0095, 0.0039, -0.0169, 0.0018, 0.0231, 0.0167], device='cuda:0'), grad: tensor([ 1.2442e-06, 4.0345e-06, 1.6605e-06, 2.6189e-06, 3.3639e-06, 6.1579e-06, 4.7497e-07, -1.1720e-05, 3.6694e-06, -1.1519e-05], device='cuda:0') 100 0.0001 changing lr epoch 157, time 256.92, cls_loss 0.0020 cls_loss_mapping 0.0038 cls_loss_causal 0.5651 re_mapping 0.0062 re_causal 0.0193 /// teacc 98.97 lr 0.00010000 Epoch 159, weight, value: tensor([[ 0.0369, -0.0931, -0.1142, ..., -0.0423, -0.1698, -0.0173], [-0.1029, -0.1330, -0.1387, ..., -0.1174, -0.0780, 0.0026], [ 0.0769, 0.0539, -0.1387, ..., -0.1031, -0.1241, -0.0739], ..., [ 0.0150, 0.0413, -0.0683, ..., -0.0253, -0.1252, 0.0474], [ 0.0134, 0.1067, 0.0838, ..., -0.0844, 0.0695, -0.1292], [-0.0474, -0.0038, -0.0430, ..., -0.0322, -0.0602, 0.0539]], device='cuda:0'), grad: tensor([[-1.1520e-06, 3.8696e-07, 3.9069e-07, ..., 4.2003e-07, 4.9453e-07, 9.7789e-09], [ 1.2899e-07, 3.2317e-07, 3.5344e-07, ..., 3.9535e-07, 6.3051e-07, -8.2003e-07], [-6.6310e-07, -2.4214e-07, 1.8813e-07, ..., 1.3877e-07, 3.8324e-07, 5.7742e-08], ..., [ 3.3947e-07, 1.5832e-07, 1.0990e-07, ..., 6.9849e-09, 3.0128e-07, 9.0338e-08], [-5.4017e-08, -2.1514e-07, 8.6147e-08, ..., 3.9348e-07, 1.0710e-08, 7.5437e-08], [ 5.4203e-07, -1.6484e-07, 4.4145e-07, ..., 6.5193e-08, 5.9698e-07, 1.3737e-07]], device='cuda:0') Epoch 159, bias, value: tensor([ 0.0280, -0.0211, -0.0098, -0.0020, 0.0092, 0.0039, -0.0169, 0.0015, 0.0229, 0.0169], device='cuda:0'), grad: tensor([-2.6692e-06, -1.3458e-07, -1.4137e-06, -2.5015e-06, -2.6338e-06, 4.2934e-07, -1.2722e-06, 4.7460e-06, 1.3094e-06, 4.1313e-06], device='cuda:0') 100 0.0001 changing lr epoch 158, time 256.99, cls_loss 0.0024 cls_loss_mapping 0.0042 cls_loss_causal 0.5503 re_mapping 0.0062 re_causal 0.0184 /// teacc 99.02 lr 0.00010000 Epoch 160, weight, value: tensor([[ 0.0374, -0.0937, -0.1154, ..., -0.0433, -0.1706, -0.0170], [-0.1038, -0.1333, -0.1395, ..., -0.1177, -0.0785, 0.0027], [ 0.0773, 0.0542, -0.1394, ..., -0.1026, -0.1242, -0.0743], ..., [ 0.0154, 0.0417, -0.0694, ..., -0.0260, -0.1250, 0.0479], [ 0.0131, 0.1065, 0.0836, ..., -0.0845, 0.0694, -0.1306], [-0.0476, -0.0039, -0.0436, ..., -0.0325, -0.0604, 0.0534]], device='cuda:0'), grad: tensor([[ 4.5747e-06, 9.8422e-06, 6.5342e-06, ..., 2.3982e-07, 9.2164e-06, 6.4187e-06], [ 1.3888e-05, 1.8701e-05, 1.9789e-05, ..., 4.7032e-08, 1.7419e-05, -9.2566e-05], [ 2.0768e-06, -1.6987e-05, 1.0416e-05, ..., 2.6543e-07, 6.0014e-06, 5.2862e-06], ..., [ 1.4044e-05, 2.5868e-05, 3.4552e-06, ..., -5.6345e-08, 5.1074e-06, 9.5516e-06], [-5.5730e-05, -1.2088e-04, -7.0930e-05, ..., 5.8208e-08, -9.7632e-05, -4.9034e-07], [ 2.7612e-05, 6.8605e-05, 2.7582e-05, ..., 2.2352e-08, 4.4644e-05, 1.4238e-05]], device='cuda:0') Epoch 160, bias, value: tensor([ 0.0276, -0.0211, -0.0092, -0.0022, 0.0096, 0.0042, -0.0169, 0.0016, 0.0224, 0.0165], device='cuda:0'), grad: tensor([ 3.7342e-05, -1.3041e-04, 8.0645e-05, 2.2128e-05, -9.7930e-05, 6.7890e-05, 6.3121e-05, 7.8678e-05, -2.2733e-04, 1.0586e-04], device='cuda:0') 100 0.0001 changing lr epoch 159, time 257.34, cls_loss 0.0017 cls_loss_mapping 0.0044 cls_loss_causal 0.5315 re_mapping 0.0061 re_causal 0.0192 /// teacc 98.99 lr 0.00010000 Epoch 161, weight, value: tensor([[ 0.0378, -0.0937, -0.1157, ..., -0.0434, -0.1709, -0.0169], [-0.1041, -0.1336, -0.1400, ..., -0.1178, -0.0788, 0.0029], [ 0.0774, 0.0542, -0.1406, ..., -0.1029, -0.1245, -0.0744], ..., [ 0.0154, 0.0416, -0.0708, ..., -0.0261, -0.1252, 0.0478], [ 0.0135, 0.1069, 0.0841, ..., -0.0845, 0.0698, -0.1311], [-0.0479, -0.0037, -0.0448, ..., -0.0326, -0.0610, 0.0535]], device='cuda:0'), grad: tensor([[ 1.1146e-05, 1.1556e-05, 1.3739e-05, ..., 5.1223e-09, 1.8179e-05, 1.4948e-07], [ 7.8883e-07, 1.7788e-06, 1.1008e-06, ..., 1.3970e-09, 1.4221e-06, -3.3192e-06], [ 1.6792e-06, 1.7192e-06, 2.4810e-06, ..., 2.7940e-09, 3.2391e-06, 4.7637e-07], ..., [ 3.9255e-07, -1.1306e-06, 6.3796e-07, ..., 0.0000e+00, 7.1200e-07, -8.0373e-07], [-2.6688e-05, -2.6658e-05, -2.9385e-05, ..., 2.8871e-08, -4.0770e-05, 1.6205e-07], [ 2.4848e-06, 2.1514e-06, 3.3118e-06, ..., 9.3132e-10, 4.2059e-06, 6.3190e-07]], device='cuda:0') Epoch 161, bias, value: tensor([ 0.0278, -0.0211, -0.0092, -0.0020, 0.0097, 0.0041, -0.0173, 0.0012, 0.0226, 0.0165], device='cuda:0'), grad: tensor([ 5.7191e-05, -1.6764e-08, 1.2390e-05, 3.2932e-05, -9.4175e-06, 7.2159e-06, 6.0275e-06, 8.1677e-07, -1.3137e-04, 2.4050e-05], device='cuda:0') 100 0.0001 changing lr epoch 160, time 257.03, cls_loss 0.0015 cls_loss_mapping 0.0044 cls_loss_causal 0.5269 re_mapping 0.0060 re_causal 0.0190 /// teacc 99.00 lr 0.00010000 Epoch 162, weight, value: tensor([[ 0.0382, -0.0940, -0.1162, ..., -0.0435, -0.1712, -0.0169], [-0.1045, -0.1337, -0.1402, ..., -0.1179, -0.0789, 0.0034], [ 0.0776, 0.0543, -0.1419, ..., -0.1032, -0.1249, -0.0748], ..., [ 0.0152, 0.0415, -0.0744, ..., -0.0263, -0.1256, 0.0475], [ 0.0138, 0.1071, 0.0840, ..., -0.0846, 0.0699, -0.1316], [-0.0480, -0.0038, -0.0453, ..., -0.0325, -0.0613, 0.0535]], device='cuda:0'), grad: tensor([[-4.6864e-06, -3.2000e-06, 1.7229e-07, ..., 7.3574e-08, 1.6065e-07, -9.3132e-09], [ 2.1514e-07, 2.4633e-07, 1.3225e-07, ..., 1.5367e-08, 1.5926e-07, -1.4761e-06], [ 7.0408e-07, 6.0722e-07, 4.2329e-07, ..., 2.2817e-08, 4.5542e-07, 2.0396e-07], ..., [-2.3236e-07, -6.4867e-07, 2.0163e-07, ..., -1.2107e-08, 2.1188e-07, -4.3306e-08], [ 4.2003e-07, 1.7649e-07, 1.1409e-06, ..., 1.3411e-07, 8.9174e-07, 4.2794e-07], [ 2.7400e-06, 2.3730e-06, 1.2806e-06, ..., 8.3819e-09, 1.1260e-06, 5.0012e-07]], device='cuda:0') Epoch 162, bias, value: tensor([ 0.0282, -0.0208, -0.0092, -0.0020, 0.0098, 0.0041, -0.0168, 0.0008, 0.0226, 0.0163], device='cuda:0'), grad: tensor([-1.5810e-05, -1.9539e-06, 3.1628e-06, 8.5831e-06, 2.5891e-07, -1.2860e-05, 1.3011e-06, 4.1537e-07, 3.9749e-06, 1.2897e-05], device='cuda:0') 100 0.0001 changing lr epoch 161, time 257.02, cls_loss 0.0024 cls_loss_mapping 0.0047 cls_loss_causal 0.5066 re_mapping 0.0065 re_causal 0.0185 /// teacc 98.90 lr 0.00010000 Epoch 163, weight, value: tensor([[ 0.0385, -0.0943, -0.1166, ..., -0.0437, -0.1716, -0.0171], [-0.1046, -0.1338, -0.1405, ..., -0.1179, -0.0791, 0.0038], [ 0.0760, 0.0521, -0.1478, ..., -0.1068, -0.1279, -0.0765], ..., [ 0.0152, 0.0416, -0.0758, ..., -0.0263, -0.1266, 0.0472], [ 0.0151, 0.1079, 0.0842, ..., -0.0839, 0.0702, -0.1333], [-0.0480, -0.0036, -0.0477, ..., -0.0326, -0.0621, 0.0531]], device='cuda:0'), grad: tensor([[ 6.8080e-07, 1.4929e-06, 1.6568e-06, ..., 6.5193e-08, 6.4215e-07, 7.6368e-08], [ 3.3528e-07, 1.5069e-06, 7.8324e-07, ..., 8.3819e-09, 5.0524e-07, 7.4971e-08], [-2.3339e-06, -4.8727e-06, 7.1153e-07, ..., 3.7253e-09, 2.8266e-07, 2.3888e-07], ..., [ 4.1444e-08, -2.2501e-06, 4.2282e-07, ..., 0.0000e+00, 2.2165e-07, -8.7498e-07], [ 8.8662e-07, 1.9595e-06, 6.3181e-05, ..., 3.0734e-08, 2.0206e-05, 1.6764e-07], [ 1.3918e-05, 2.4028e-07, -8.2552e-05, ..., 3.2596e-09, -2.5660e-05, -1.7323e-07]], device='cuda:0') Epoch 163, bias, value: tensor([ 0.0283, -0.0214, -0.0096, -0.0015, 0.0099, 0.0063, -0.0178, 0.0005, 0.0229, 0.0159], device='cuda:0'), grad: tensor([ 2.7511e-06, 5.4240e-06, -6.6087e-06, 2.5049e-05, -1.5950e-04, 2.2024e-05, 3.6694e-06, -3.0529e-06, 1.6773e-04, -5.7220e-05], device='cuda:0') 100 0.0001 changing lr epoch 162, time 257.02, cls_loss 0.0023 cls_loss_mapping 0.0052 cls_loss_causal 0.5453 re_mapping 0.0068 re_causal 0.0195 /// teacc 98.94 lr 0.00010000 Epoch 164, weight, value: tensor([[ 0.0384, -0.0950, -0.1177, ..., -0.0444, -0.1723, -0.0178], [-0.1050, -0.1340, -0.1409, ..., -0.1181, -0.0792, 0.0040], [ 0.0765, 0.0523, -0.1483, ..., -0.1072, -0.1283, -0.0769], ..., [ 0.0153, 0.0416, -0.0769, ..., -0.0260, -0.1268, 0.0460], [ 0.0152, 0.1083, 0.0851, ..., -0.0840, 0.0708, -0.1348], [-0.0480, -0.0036, -0.0484, ..., -0.0327, -0.0623, 0.0540]], device='cuda:0'), grad: tensor([[-3.1255e-06, -1.8347e-07, 1.4063e-07, ..., 1.0710e-08, 6.1560e-07, -2.9802e-08], [ 6.7055e-08, 2.3749e-08, 4.7870e-07, ..., -4.3306e-08, 2.5127e-06, -1.2591e-06], [ 1.2694e-06, 8.9407e-08, 3.0827e-06, ..., 1.3504e-08, 2.0057e-05, 2.9663e-07], ..., [ 4.9360e-08, 7.6368e-08, 1.8533e-07, ..., 1.5367e-08, 3.6741e-07, 4.9826e-07], [ 3.9209e-07, -3.4925e-08, 1.9427e-06, ..., 3.0268e-08, 1.7090e-06, 1.8720e-07], [ 1.0300e-06, -1.4482e-07, 2.4727e-07, ..., 2.3283e-09, 3.0361e-07, -3.2596e-08]], device='cuda:0') Epoch 164, bias, value: tensor([ 0.0276, -0.0214, -0.0094, -0.0010, 0.0097, 0.0062, -0.0175, -0.0005, 0.0232, 0.0164], device='cuda:0'), grad: tensor([-1.0930e-05, 4.8988e-06, 7.3969e-05, -8.6010e-05, 4.8149e-07, 2.5071e-06, 9.8906e-07, 2.6934e-06, 6.5230e-06, 4.8392e-06], device='cuda:0') 100 0.0001 changing lr epoch 163, time 257.32, cls_loss 0.0022 cls_loss_mapping 0.0042 cls_loss_causal 0.4982 re_mapping 0.0065 re_causal 0.0185 /// teacc 98.85 lr 0.00010000 Epoch 165, weight, value: tensor([[ 0.0388, -0.0957, -0.1190, ..., -0.0459, -0.1733, -0.0187], [-0.1065, -0.1343, -0.1413, ..., -0.1184, -0.0792, 0.0047], [ 0.0767, 0.0522, -0.1490, ..., -0.1079, -0.1293, -0.0771], ..., [ 0.0158, 0.0422, -0.0774, ..., -0.0255, -0.1267, 0.0460], [ 0.0149, 0.1081, 0.0848, ..., -0.0841, 0.0705, -0.1366], [-0.0478, -0.0031, -0.0488, ..., -0.0329, -0.0620, 0.0542]], device='cuda:0'), grad: tensor([[ 1.4268e-06, 4.7013e-06, 1.0259e-05, ..., 2.3935e-06, 4.9174e-06, 2.0443e-07], [ 1.5944e-06, 7.7635e-06, 6.9588e-06, ..., 3.0966e-07, 8.4192e-06, 1.3225e-07], [ 4.6100e-07, 3.4031e-06, 5.0068e-06, ..., 7.3109e-07, 3.5875e-06, 1.2619e-07], ..., [ 1.0207e-06, 3.2857e-06, 2.0992e-06, ..., 2.0489e-08, 2.7828e-06, 7.9535e-07], [-4.0494e-06, -2.0146e-05, -1.1466e-05, ..., 1.0878e-06, -1.9640e-05, -3.5670e-07], [-5.6671e-07, -6.8033e-07, 1.4175e-06, ..., 9.5926e-08, 1.1828e-06, -1.3700e-06]], device='cuda:0') Epoch 165, bias, value: tensor([ 0.0276, -0.0205, -0.0097, -0.0023, 0.0090, 0.0070, -0.0169, -0.0006, 0.0225, 0.0169], device='cuda:0'), grad: tensor([ 1.7956e-05, 2.1100e-05, 9.3728e-06, 8.2105e-06, 1.3150e-05, 3.1680e-05, -6.4850e-05, 8.6203e-06, -4.6343e-05, 1.1390e-06], device='cuda:0') 100 0.0001 changing lr epoch 164, time 257.39, cls_loss 0.0023 cls_loss_mapping 0.0040 cls_loss_causal 0.5345 re_mapping 0.0063 re_causal 0.0186 /// teacc 99.05 lr 0.00010000 Epoch 166, weight, value: tensor([[ 0.0388, -0.0966, -0.1203, ..., -0.0467, -0.1741, -0.0188], [-0.1073, -0.1346, -0.1420, ..., -0.1189, -0.0795, 0.0051], [ 0.0768, 0.0519, -0.1498, ..., -0.1085, -0.1305, -0.0783], ..., [ 0.0159, 0.0425, -0.0795, ..., -0.0256, -0.1269, 0.0459], [ 0.0151, 0.1083, 0.0845, ..., -0.0845, 0.0704, -0.1371], [-0.0477, -0.0034, -0.0497, ..., -0.0331, -0.0623, 0.0544]], device='cuda:0'), grad: tensor([[ 1.4296e-07, 6.9626e-06, 1.0237e-05, ..., 9.3281e-06, 1.1094e-05, -5.9605e-08], [ 1.2638e-06, 3.5837e-06, 3.5204e-06, ..., 1.7760e-06, 4.6231e-06, -2.4354e-07], [ 7.3202e-07, 1.7555e-06, 1.4799e-06, ..., 6.0024e-07, 2.0768e-06, 3.0734e-08], ..., [ 1.7090e-07, 2.3376e-07, 2.9290e-07, ..., 2.5611e-08, 4.4936e-07, 3.3528e-08], [-2.4941e-06, 2.1428e-05, 3.6031e-05, ..., 3.6687e-05, 3.7193e-05, 8.5216e-08], [ 7.1246e-08, 4.6333e-07, 7.0827e-07, ..., 3.6089e-07, 7.5996e-07, 3.2596e-08]], device='cuda:0') Epoch 166, bias, value: tensor([ 0.0269, -0.0199, -0.0106, -0.0021, 0.0097, 0.0073, -0.0163, -0.0008, 0.0222, 0.0167], device='cuda:0'), grad: tensor([ 1.5914e-05, 8.7246e-06, 4.6492e-06, 8.8988e-07, 2.4922e-06, 8.8662e-06, -9.9719e-05, 1.1958e-06, 5.5313e-05, 1.7583e-06], device='cuda:0') 100 0.0001 changing lr epoch 165, time 257.50, cls_loss 0.0022 cls_loss_mapping 0.0037 cls_loss_causal 0.5468 re_mapping 0.0055 re_causal 0.0181 /// teacc 99.00 lr 0.00010000 Epoch 167, weight, value: tensor([[ 0.0384, -0.0971, -0.1215, ..., -0.0478, -0.1750, -0.0192], [-0.1082, -0.1352, -0.1426, ..., -0.1196, -0.0799, 0.0052], [ 0.0770, 0.0518, -0.1503, ..., -0.1089, -0.1310, -0.0787], ..., [ 0.0163, 0.0429, -0.0804, ..., -0.0261, -0.1269, 0.0461], [ 0.0154, 0.1089, 0.0850, ..., -0.0837, 0.0709, -0.1377], [-0.0472, -0.0035, -0.0501, ..., -0.0336, -0.0620, 0.0545]], device='cuda:0'), grad: tensor([[-1.1660e-06, 7.4506e-06, 3.2131e-08, ..., 1.2107e-08, 3.3062e-08, 5.4948e-08], [ 2.1188e-07, 1.3420e-06, 2.5611e-08, ..., -2.3283e-09, 7.3574e-08, 2.1560e-07], [ 1.0896e-07, 2.3525e-06, 4.8429e-08, ..., 1.1642e-08, 6.7521e-08, 1.5367e-06], ..., [-2.4727e-07, -2.9467e-06, 3.9116e-08, ..., 2.3283e-09, -9.9186e-08, -2.6822e-06], [ 2.1420e-07, 2.1840e-07, -1.6205e-07, ..., 2.4214e-08, -2.6450e-07, 6.3796e-08], [-3.0035e-07, -1.4350e-05, 2.2817e-07, ..., 2.3283e-09, 2.8405e-07, 4.2794e-07]], device='cuda:0') Epoch 167, bias, value: tensor([ 0.0252, -0.0199, -0.0108, -0.0022, 0.0100, 0.0061, -0.0154, -0.0004, 0.0223, 0.0175], device='cuda:0'), grad: tensor([ 2.8685e-05, 3.6545e-06, 7.6815e-06, 9.3356e-06, 2.0549e-05, -2.8033e-07, 2.8033e-06, -3.8482e-06, 3.4329e-06, -7.2062e-05], device='cuda:0') 100 0.0001 changing lr epoch 166, time 257.07, cls_loss 0.0020 cls_loss_mapping 0.0040 cls_loss_causal 0.5137 re_mapping 0.0059 re_causal 0.0181 /// teacc 98.88 lr 0.00010000 Epoch 168, weight, value: tensor([[ 0.0387, -0.0972, -0.1219, ..., -0.0479, -0.1755, -0.0193], [-0.1089, -0.1356, -0.1428, ..., -0.1217, -0.0801, 0.0053], [ 0.0772, 0.0521, -0.1505, ..., -0.1080, -0.1314, -0.0791], ..., [ 0.0166, 0.0430, -0.0810, ..., -0.0263, -0.1270, 0.0462], [ 0.0155, 0.1091, 0.0851, ..., -0.0841, 0.0710, -0.1382], [-0.0473, -0.0035, -0.0503, ..., -0.0335, -0.0623, 0.0545]], device='cuda:0'), grad: tensor([[ 1.1874e-07, 2.3469e-06, 1.7714e-06, ..., 1.9260e-06, 1.4901e-06, 4.4936e-07], [ 5.6811e-07, 1.3113e-06, 7.7020e-07, ..., 1.1083e-07, 8.8103e-07, 2.4587e-07], [ 2.5202e-06, 6.3032e-06, 2.2445e-06, ..., 8.4471e-07, 1.8580e-06, 4.6659e-07], ..., [-3.3844e-06, -6.9477e-06, 5.4715e-07, ..., 4.5169e-08, 5.4156e-07, -7.2969e-07], [-6.6450e-07, -2.1905e-06, 3.0044e-06, ..., -4.8196e-07, 2.9895e-06, 3.0035e-07], [ 3.6461e-07, 9.0059e-07, 8.8476e-07, ..., 3.3807e-07, 7.9209e-07, -2.5146e-08]], device='cuda:0') Epoch 168, bias, value: tensor([ 0.0254, -0.0195, -0.0109, -0.0021, 0.0105, 0.0060, -0.0158, -0.0002, 0.0221, 0.0170], device='cuda:0'), grad: tensor([ 4.7609e-06, 4.2804e-06, 1.7226e-05, 1.7405e-05, 3.9749e-06, -1.0304e-05, -2.7016e-05, -2.1294e-05, 6.3591e-06, 4.4927e-06], device='cuda:0') 100 0.0001 changing lr epoch 167, time 255.59, cls_loss 0.0021 cls_loss_mapping 0.0041 cls_loss_causal 0.5080 re_mapping 0.0060 re_causal 0.0181 /// teacc 99.01 lr 0.00010000 Epoch 169, weight, value: tensor([[ 0.0389, -0.0975, -0.1221, ..., -0.0481, -0.1757, -0.0194], [-0.1096, -0.1357, -0.1432, ..., -0.1215, -0.0802, 0.0056], [ 0.0772, 0.0509, -0.1509, ..., -0.1090, -0.1320, -0.0795], ..., [ 0.0168, 0.0439, -0.0819, ..., -0.0239, -0.1277, 0.0459], [ 0.0156, 0.1090, 0.0846, ..., -0.0845, 0.0707, -0.1386], [-0.0475, -0.0037, -0.0511, ..., -0.0337, -0.0627, 0.0544]], device='cuda:0'), grad: tensor([[-1.0477e-06, 4.8149e-07, 9.8255e-08, ..., 1.4435e-08, 6.5658e-08, 4.6473e-07], [ 6.4867e-07, 2.5239e-06, 1.3132e-07, ..., 5.1223e-09, 9.5926e-08, -1.7539e-05], [-4.7358e-07, 1.6354e-06, 7.6368e-08, ..., 4.6566e-09, 6.2864e-08, 1.2003e-05], ..., [-1.8030e-06, -1.3508e-05, 3.3760e-07, ..., 0.0000e+00, 1.9697e-07, -1.6773e-06], [ 6.4261e-07, 1.1688e-06, 5.9530e-06, ..., 1.0943e-07, 3.3472e-06, 5.8068e-07], [ 1.3830e-06, 5.3197e-06, 1.5926e-07, ..., 9.3132e-09, 2.0582e-07, 3.7681e-06]], device='cuda:0') Epoch 169, bias, value: tensor([ 0.0255, -0.0191, -0.0122, -0.0016, 0.0114, 0.0064, -0.0158, 0.0002, 0.0215, 0.0164], device='cuda:0'), grad: tensor([-1.4128e-06, -8.9943e-05, 5.4240e-05, 6.5714e-06, -1.0114e-06, -7.3463e-06, 2.9691e-06, 8.7619e-06, 1.1042e-05, 1.6049e-05], device='cuda:0') 100 0.0001 changing lr epoch 168, time 256.57, cls_loss 0.0015 cls_loss_mapping 0.0038 cls_loss_causal 0.5283 re_mapping 0.0060 re_causal 0.0184 /// teacc 98.94 lr 0.00010000 Epoch 170, weight, value: tensor([[ 0.0390, -0.0978, -0.1231, ..., -0.0494, -0.1762, -0.0196], [-0.1102, -0.1360, -0.1438, ..., -0.1229, -0.0805, 0.0058], [ 0.0777, 0.0511, -0.1511, ..., -0.1096, -0.1322, -0.0798], ..., [ 0.0167, 0.0442, -0.0830, ..., -0.0230, -0.1278, 0.0459], [ 0.0157, 0.1091, 0.0844, ..., -0.0846, 0.0706, -0.1390], [-0.0475, -0.0038, -0.0518, ..., -0.0339, -0.0630, 0.0546]], device='cuda:0'), grad: tensor([[-1.3039e-07, 1.7127e-06, 1.0170e-06, ..., 9.9931e-07, 9.8255e-07, 2.2352e-08], [ 1.5469e-06, 6.0052e-06, 2.6003e-06, ..., 6.7567e-07, 1.7611e-06, 7.5530e-07], [-2.4103e-06, 3.5446e-06, 7.9200e-06, ..., 2.2966e-06, 5.1819e-06, 2.2259e-07], ..., [ 3.2727e-06, 1.3085e-06, 1.2033e-06, ..., 2.3888e-07, 8.3447e-07, -2.1178e-06], [-1.8030e-05, -5.1975e-05, -3.9190e-05, ..., -9.0301e-06, -2.5153e-05, 1.7881e-07], [ 1.1120e-06, 3.0175e-06, 2.2352e-06, ..., 4.6100e-07, 1.5851e-06, 1.8161e-07]], device='cuda:0') Epoch 170, bias, value: tensor([ 0.0254, -0.0190, -0.0122, -0.0017, 0.0112, 0.0067, -0.0155, 0.0002, 0.0212, 0.0164], device='cuda:0'), grad: tensor([ 3.6098e-06, 1.2234e-05, -1.8030e-06, 2.5749e-05, 6.7987e-06, 1.0774e-05, 1.1250e-05, -2.5053e-07, -7.4804e-05, 6.2212e-06], device='cuda:0') 100 0.0001 changing lr epoch 169, time 256.87, cls_loss 0.0019 cls_loss_mapping 0.0040 cls_loss_causal 0.5350 re_mapping 0.0059 re_causal 0.0178 /// teacc 98.94 lr 0.00010000 Epoch 171, weight, value: tensor([[ 0.0397, -0.0979, -0.1234, ..., -0.0499, -0.1766, -0.0198], [-0.1110, -0.1369, -0.1444, ..., -0.1253, -0.0811, 0.0055], [ 0.0781, 0.0515, -0.1514, ..., -0.1083, -0.1322, -0.0800], ..., [ 0.0167, 0.0445, -0.0837, ..., -0.0230, -0.1279, 0.0462], [ 0.0158, 0.1095, 0.0848, ..., -0.0847, 0.0708, -0.1397], [-0.0479, -0.0043, -0.0530, ..., -0.0340, -0.0637, 0.0554]], device='cuda:0'), grad: tensor([[-2.1514e-07, 1.2154e-07, 1.6298e-07, ..., 7.7765e-08, 1.3877e-07, 3.8184e-08], [ 1.4063e-07, 2.9383e-07, 1.0896e-07, ..., 2.0023e-08, 1.5926e-07, -1.2601e-06], [-2.0675e-07, -1.1874e-07, 7.8697e-08, ..., 1.6298e-08, 1.0012e-07, 1.1083e-07], ..., [-9.4995e-08, -3.8324e-07, 1.0617e-07, ..., 9.3132e-10, 1.1548e-07, 3.7439e-07], [-8.2422e-08, -3.3760e-07, -6.2864e-08, ..., 1.3458e-07, -1.2852e-07, 1.5227e-07], [ 2.4866e-07, 1.8952e-07, 1.0384e-06, ..., 1.8161e-08, 8.7265e-07, 3.0873e-07]], device='cuda:0') Epoch 171, bias, value: tensor([ 0.0257, -0.0201, -0.0113, -0.0019, 0.0112, 0.0064, -0.0148, 0.0005, 0.0212, 0.0162], device='cuda:0'), grad: tensor([-4.5402e-07, -3.1274e-06, -3.3900e-07, 1.5572e-06, 7.5623e-07, -3.5241e-06, -2.8685e-07, 1.8869e-06, 3.7905e-07, 3.1609e-06], device='cuda:0') 100 0.0001 changing lr epoch 170, time 256.45, cls_loss 0.0025 cls_loss_mapping 0.0052 cls_loss_causal 0.5543 re_mapping 0.0059 re_causal 0.0174 /// teacc 98.94 lr 0.00010000 Epoch 172, weight, value: tensor([[ 0.0401, -0.0984, -0.1235, ..., -0.0502, -0.1771, -0.0199], [-0.1124, -0.1375, -0.1458, ..., -0.1267, -0.0820, 0.0057], [ 0.0782, 0.0514, -0.1516, ..., -0.1084, -0.1325, -0.0807], ..., [ 0.0170, 0.0450, -0.0845, ..., -0.0230, -0.1278, 0.0463], [ 0.0163, 0.1115, 0.0852, ..., -0.0850, 0.0722, -0.1392], [-0.0487, -0.0070, -0.0569, ..., -0.0344, -0.0675, 0.0552]], device='cuda:0'), grad: tensor([[-7.0035e-07, 3.7216e-06, 3.1721e-06, ..., 1.1399e-06, 3.9414e-06, -1.0785e-06], [ 1.1921e-07, 1.2010e-05, 1.2197e-05, ..., 1.9316e-06, 1.2815e-05, -4.6007e-06], [-2.4214e-08, 1.4678e-06, 1.7565e-06, ..., 6.8638e-07, 1.8701e-06, 7.2736e-07], ..., [-1.4156e-07, 4.1537e-07, 8.0839e-07, ..., 2.1327e-07, 1.3039e-06, 2.9057e-07], [ 4.0513e-08, -4.2629e-04, -4.3273e-04, ..., -1.2779e-04, -4.4346e-04, 2.1234e-07], [ 3.7113e-07, 1.2051e-06, 1.8263e-06, ..., 4.0326e-07, 1.9744e-06, 9.6206e-07]], device='cuda:0') Epoch 172, bias, value: tensor([ 0.0261, -0.0200, -0.0115, -0.0002, 0.0119, 0.0047, -0.0147, 0.0003, 0.0225, 0.0149], device='cuda:0'), grad: tensor([ 2.3283e-06, -2.2322e-05, 8.9258e-06, 3.4403e-06, 1.2442e-05, 5.0962e-06, 9.2173e-04, 1.8403e-05, -9.6178e-04, 1.1981e-05], device='cuda:0') 100 0.0001 changing lr epoch 171, time 256.79, cls_loss 0.0030 cls_loss_mapping 0.0055 cls_loss_causal 0.5372 re_mapping 0.0060 re_causal 0.0181 /// teacc 99.07 lr 0.00010000 Epoch 173, weight, value: tensor([[ 0.0409, -0.0981, -0.1238, ..., -0.0501, -0.1773, -0.0193], [-0.1131, -0.1378, -0.1461, ..., -0.1266, -0.0823, 0.0060], [ 0.0784, 0.0514, -0.1520, ..., -0.1083, -0.1329, -0.0821], ..., [ 0.0172, 0.0450, -0.0856, ..., -0.0225, -0.1280, 0.0464], [ 0.0165, 0.1122, 0.0863, ..., -0.0845, 0.0731, -0.1400], [-0.0500, -0.0070, -0.0576, ..., -0.0362, -0.0677, 0.0556]], device='cuda:0'), grad: tensor([[-2.6021e-06, 1.6578e-07, 6.2399e-08, ..., 4.6100e-08, 9.2201e-08, 6.2399e-08], [ 2.6356e-06, 8.7246e-06, 1.6037e-06, ..., 3.2596e-08, 3.3733e-06, 1.6317e-06], [ 1.7043e-06, 3.7588e-06, 1.9390e-06, ..., 3.1665e-08, 3.9563e-06, 4.2049e-07], ..., [-1.8934e-06, -9.5442e-06, 1.5413e-07, ..., 0.0000e+00, 1.3923e-07, -2.1309e-06], [-2.6654e-06, -3.9190e-06, -3.8091e-06, ..., 7.3109e-08, -7.7337e-06, 3.2131e-07], [ 2.2445e-06, -1.5553e-07, 3.7253e-08, ..., 3.2596e-09, -3.3202e-07, -7.9488e-07]], device='cuda:0') Epoch 173, bias, value: tensor([ 0.0234, -0.0196, -0.0119, -0.0004, 0.0131, 0.0050, -0.0161, 0.0002, 0.0231, 0.0173], device='cuda:0'), grad: tensor([-1.4469e-05, 1.8626e-05, 1.0550e-05, 1.0943e-06, 1.8850e-06, 5.1223e-09, 2.4354e-07, -1.5363e-05, -1.3724e-05, 1.1124e-05], device='cuda:0') 100 0.0001 changing lr epoch 172, time 256.68, cls_loss 0.0014 cls_loss_mapping 0.0036 cls_loss_causal 0.5386 re_mapping 0.0056 re_causal 0.0190 /// teacc 99.02 lr 0.00010000 Epoch 174, weight, value: tensor([[ 0.0413, -0.0980, -0.1240, ..., -0.0505, -0.1773, -0.0189], [-0.1137, -0.1400, -0.1466, ..., -0.1269, -0.0830, 0.0046], [ 0.0786, 0.0517, -0.1520, ..., -0.1083, -0.1325, -0.0823], ..., [ 0.0174, 0.0463, -0.0861, ..., -0.0225, -0.1280, 0.0482], [ 0.0167, 0.1126, 0.0870, ..., -0.0842, 0.0734, -0.1404], [-0.0504, -0.0073, -0.0579, ..., -0.0363, -0.0679, 0.0553]], device='cuda:0'), grad: tensor([[-1.3504e-07, 7.4953e-06, 1.9118e-05, ..., 1.3016e-05, 2.1663e-06, -9.4064e-08], [ 1.4622e-07, 1.9725e-06, 3.8696e-07, ..., 2.0256e-07, 1.2387e-07, 1.6550e-06], [-3.9209e-07, -2.7008e-07, 3.6089e-07, ..., 9.8720e-08, 2.3516e-07, 8.8476e-09], ..., [ 1.3597e-07, -1.6121e-06, 9.6858e-08, ..., 2.2352e-08, 6.2864e-08, -1.7332e-06], [-7.2792e-06, -5.7667e-05, -7.0691e-05, ..., -1.9357e-05, -4.3690e-05, 2.1420e-08], [ 1.2573e-07, 2.4773e-07, 8.5402e-07, ..., 3.2224e-07, 3.7486e-07, -4.6566e-10]], device='cuda:0') Epoch 174, bias, value: tensor([ 0.0236, -0.0212, -0.0114, -0.0005, 0.0129, 0.0049, -0.0165, 0.0019, 0.0229, 0.0172], device='cuda:0'), grad: tensor([ 4.7654e-05, 4.9323e-06, -4.7218e-07, -4.1634e-05, 8.6986e-07, 4.9949e-05, 1.3731e-05, -2.9281e-06, -7.5400e-05, 3.4235e-06], device='cuda:0') 100 0.0001 changing lr epoch 173, time 256.48, cls_loss 0.0018 cls_loss_mapping 0.0043 cls_loss_causal 0.5135 re_mapping 0.0057 re_causal 0.0180 /// teacc 98.94 lr 0.00010000 Epoch 175, weight, value: tensor([[ 0.0417, -0.0980, -0.1244, ..., -0.0507, -0.1775, -0.0191], [-0.1140, -0.1402, -0.1463, ..., -0.1266, -0.0831, 0.0047], [ 0.0789, 0.0518, -0.1523, ..., -0.1083, -0.1328, -0.0828], ..., [ 0.0176, 0.0466, -0.0878, ..., -0.0224, -0.1282, 0.0484], [ 0.0166, 0.1126, 0.0870, ..., -0.0844, 0.0733, -0.1417], [-0.0505, -0.0074, -0.0594, ..., -0.0368, -0.0684, 0.0552]], device='cuda:0'), grad: tensor([[-7.7719e-07, 7.1246e-08, 7.6368e-08, ..., 4.6566e-09, 1.2992e-07, 5.3085e-08], [ 1.8906e-07, 2.8405e-07, 1.5972e-07, ..., 1.3970e-09, 2.6915e-07, -2.4773e-07], [ 1.2480e-07, 1.4203e-07, 1.8720e-07, ..., 3.7253e-09, 3.1386e-07, 4.8894e-08], ..., [ 6.7521e-08, 1.2619e-07, 1.2852e-07, ..., 0.0000e+00, 2.2026e-07, 2.7986e-07], [-3.3714e-07, -5.5181e-07, -3.9162e-07, ..., 2.3749e-08, -2.9104e-07, 2.7940e-07], [-1.0710e-07, -1.2154e-07, 1.1073e-06, ..., 9.3132e-10, 2.0470e-06, -8.8476e-07]], device='cuda:0') Epoch 175, bias, value: tensor([ 0.0236, -0.0210, -0.0114, -0.0033, 0.0130, 0.0079, -0.0170, 0.0021, 0.0224, 0.0170], device='cuda:0'), grad: tensor([-3.8743e-06, 9.6485e-07, 1.2936e-06, -1.8895e-05, -2.7478e-05, 1.2361e-05, 1.9409e-06, 1.4696e-06, 1.8124e-06, 3.0428e-05], device='cuda:0') 100 0.0001 changing lr epoch 174, time 256.65, cls_loss 0.0023 cls_loss_mapping 0.0043 cls_loss_causal 0.5455 re_mapping 0.0061 re_causal 0.0176 /// teacc 98.94 lr 0.00010000 Epoch 176, weight, value: tensor([[ 0.0422, -0.0985, -0.1253, ..., -0.0518, -0.1781, -0.0195], [-0.1147, -0.1405, -0.1468, ..., -0.1269, -0.0835, 0.0049], [ 0.0789, 0.0518, -0.1528, ..., -0.1084, -0.1336, -0.0833], ..., [ 0.0178, 0.0465, -0.0895, ..., -0.0224, -0.1284, 0.0482], [ 0.0167, 0.1128, 0.0869, ..., -0.0846, 0.0735, -0.1421], [-0.0505, -0.0072, -0.0612, ..., -0.0355, -0.0691, 0.0555]], device='cuda:0'), grad: tensor([[-2.3050e-07, 3.2922e-07, -1.2338e-04, ..., -5.0664e-05, 2.5518e-07, 6.9989e-07], [ 1.5693e-07, 2.2305e-07, 2.9244e-07, ..., 8.0559e-08, 1.9604e-07, 1.0990e-07], [-1.4110e-07, 7.4040e-08, 1.1222e-07, ..., 1.6764e-08, 9.4995e-08, 3.0268e-08], ..., [-3.0734e-08, 1.8720e-07, 1.2713e-07, ..., 2.5611e-08, 1.0245e-07, 5.9465e-07], [-1.7090e-07, -5.6531e-07, 1.4249e-07, ..., 2.4308e-07, -3.9488e-07, 1.7695e-08], [ 2.6077e-07, -5.8301e-07, 1.7891e-06, ..., 6.1095e-07, 3.8743e-07, -1.6652e-06]], device='cuda:0') Epoch 176, bias, value: tensor([ 0.0236, -0.0206, -0.0116, -0.0031, 0.0153, 0.0077, -0.0168, 0.0017, 0.0222, 0.0164], device='cuda:0'), grad: tensor([-5.2452e-04, 2.8163e-06, -1.7667e-06, 2.0396e-06, 8.2096e-07, -1.5885e-05, 5.3024e-04, 3.5185e-06, 7.9302e-07, 1.8720e-06], device='cuda:0') 100 0.0001 changing lr epoch 175, time 256.99, cls_loss 0.0015 cls_loss_mapping 0.0031 cls_loss_causal 0.5168 re_mapping 0.0058 re_causal 0.0182 /// teacc 98.95 lr 0.00010000 Epoch 177, weight, value: tensor([[ 0.0426, -0.0986, -0.1248, ..., -0.0514, -0.1778, -0.0199], [-0.1152, -0.1408, -0.1473, ..., -0.1270, -0.0841, 0.0049], [ 0.0789, 0.0519, -0.1531, ..., -0.1085, -0.1342, -0.0838], ..., [ 0.0178, 0.0466, -0.0908, ..., -0.0224, -0.1285, 0.0482], [ 0.0167, 0.1130, 0.0871, ..., -0.0847, 0.0737, -0.1425], [-0.0505, -0.0070, -0.0613, ..., -0.0349, -0.0691, 0.0563]], device='cuda:0'), grad: tensor([[-1.0151e-07, 4.4703e-08, 1.8440e-07, ..., 7.4506e-09, 2.3376e-07, 2.2817e-08], [ 1.0384e-07, 1.8720e-07, 9.1270e-08, ..., 2.3283e-09, 1.3271e-07, -4.4238e-08], [ 7.6834e-08, 5.5972e-07, 7.5437e-08, ..., 1.3970e-09, 2.9709e-07, 3.3062e-08], ..., [-6.2212e-07, -1.9167e-06, 1.1083e-07, ..., 0.0000e+00, -5.1036e-07, -2.0489e-08], [ 2.7427e-07, 1.2247e-06, 5.1409e-06, ..., 1.3039e-08, 7.1935e-06, 4.2841e-08], [ 6.9849e-08, -9.4064e-07, -6.1318e-06, ..., 4.6566e-10, -8.8140e-06, -4.8988e-07]], device='cuda:0') Epoch 177, bias, value: tensor([ 0.0239, -0.0207, -0.0120, -0.0033, 0.0156, 0.0078, -0.0175, 0.0017, 0.0221, 0.0164], device='cuda:0'), grad: tensor([ 8.1956e-08, 6.2073e-07, 9.8348e-07, 2.4047e-06, 2.6636e-06, -1.1614e-06, 1.1101e-06, -3.3304e-06, 1.6227e-05, -1.9625e-05], device='cuda:0') 100 0.0001 changing lr epoch 176, time 256.56, cls_loss 0.0015 cls_loss_mapping 0.0029 cls_loss_causal 0.5241 re_mapping 0.0055 re_causal 0.0173 /// teacc 98.94 lr 0.00010000 Epoch 178, weight, value: tensor([[ 0.0427, -0.0989, -0.1250, ..., -0.0511, -0.1780, -0.0195], [-0.1158, -0.1411, -0.1479, ..., -0.1277, -0.0846, 0.0063], [ 0.0795, 0.0521, -0.1534, ..., -0.1084, -0.1347, -0.0839], ..., [ 0.0175, 0.0465, -0.0921, ..., -0.0224, -0.1288, 0.0467], [ 0.0168, 0.1134, 0.0874, ..., -0.0846, 0.0740, -0.1429], [-0.0505, -0.0065, -0.0615, ..., -0.0355, -0.0689, 0.0568]], device='cuda:0'), grad: tensor([[ 7.3109e-08, 5.4948e-07, 7.4040e-08, ..., 6.7055e-08, 4.5635e-08, 2.6496e-07], [ 3.1618e-07, 1.8738e-06, 4.7032e-08, ..., 3.7253e-08, 1.2806e-07, 5.4110e-07], [ 1.6624e-07, 1.3411e-06, 1.8626e-08, ..., -1.5367e-08, 5.3551e-08, 4.9453e-07], ..., [-2.6692e-06, -1.6913e-05, 1.0710e-08, ..., 3.2596e-09, -1.2536e-06, -5.6177e-06], [ 1.0654e-06, 1.1690e-05, 7.8827e-06, ..., 4.9993e-06, 5.1335e-06, 1.8254e-06], [ 7.7346e-07, 4.8913e-06, 8.7079e-08, ..., 4.7497e-08, 3.8184e-07, 1.7723e-06]], device='cuda:0') Epoch 178, bias, value: tensor([ 0.0241, -0.0201, -0.0115, -0.0033, 0.0153, 0.0077, -0.0178, 0.0004, 0.0222, 0.0165], device='cuda:0'), grad: tensor([ 1.6298e-06, 3.1814e-06, 3.1628e-06, 9.6764e-07, 1.2964e-06, 1.3812e-06, -1.0967e-05, -3.0875e-05, 2.0847e-05, 9.3430e-06], device='cuda:0') 100 0.0001 changing lr epoch 177, time 256.44, cls_loss 0.0019 cls_loss_mapping 0.0035 cls_loss_causal 0.5120 re_mapping 0.0056 re_causal 0.0164 /// teacc 98.83 lr 0.00010000 Epoch 179, weight, value: tensor([[ 0.0430, -0.0993, -0.1255, ..., -0.0515, -0.1786, -0.0199], [-0.1172, -0.1420, -0.1486, ..., -0.1280, -0.0851, 0.0061], [ 0.0800, 0.0525, -0.1538, ..., -0.1087, -0.1353, -0.0841], ..., [ 0.0181, 0.0478, -0.0929, ..., -0.0225, -0.1289, 0.0474], [ 0.0164, 0.1129, 0.0871, ..., -0.0850, 0.0736, -0.1443], [-0.0501, -0.0070, -0.0618, ..., -0.0343, -0.0686, 0.0564]], device='cuda:0'), grad: tensor([[-8.2050e-07, 2.2119e-08, 1.6997e-08, ..., 9.5461e-09, 1.5134e-08, 3.4459e-08], [ 6.3563e-08, 4.3167e-07, 5.1223e-09, ..., 1.3970e-09, 1.6298e-09, -7.3388e-07], [ 2.1467e-07, 1.2713e-07, 1.7928e-08, ..., 9.7789e-09, 2.4447e-08, 7.6368e-08], ..., [-3.1455e-07, -4.7348e-06, 3.9581e-09, ..., 0.0000e+00, 1.3970e-09, -2.2892e-06], [ 6.5425e-08, 7.0082e-08, 5.5647e-08, ..., 3.0734e-08, 4.3306e-08, 1.6438e-07], [ 5.3830e-07, 2.8890e-06, 8.8476e-09, ..., 6.9849e-10, 9.0804e-09, 1.8980e-06]], device='cuda:0') Epoch 179, bias, value: tensor([ 0.0241, -0.0207, -0.0113, -0.0030, 0.0146, 0.0074, -0.0167, 0.0014, 0.0212, 0.0166], device='cuda:0'), grad: tensor([-3.6471e-06, -2.2352e-06, 1.2340e-06, 2.4308e-06, 4.3376e-07, 7.1665e-07, 4.3050e-07, -8.7246e-06, 1.0040e-06, 8.3372e-06], device='cuda:0') 100 0.0001 changing lr