Closed Bin-ze closed 4 months ago
Bin-ze
commented
5 months ago To verify that it was not the number of iterations, I trained for a whole day. This is the result at 255epoch：
Far worse than the official results in epoch 20，Do you know why this is? I used a 4-card configuration, while the official one is eight cards. I modified the learning rate to 1/10 of the original. Although it is very conservative, I don’t think this will lead to such a big difference.
Bin-ze
commented
5 months ago @gaotongxiao Young hero, it's time to take action！
pd162
commented
5 months ago Do you load the pretrained model checkpoints/abcnet_resnet50_fpn_pretrain.pth before finetuning on ICDAR2015? If not, the recognition head will not converge.
Bin-ze
commented
5 months ago Do you load the pretrained model
checkpoints/abcnet_resnet50_fpn_pretrain.pthbefore finetuning on ICDAR2015? If not, the recognition head will not converge.
Yes, I used a pretrained model. However, it can be seen that the detection head has not converged because the loss_bezier is very large, which is seriously different from the official log display.
However, I don't understand why the recognition head cannot converge. Is it because of the size of the data set or something else?
Do you have any experiences to share?
pd162
commented
5 months ago Do you load the pretrained model
checkpoints/abcnet_resnet50_fpn_pretrain.pthbefore finetuning on ICDAR2015? If not, the recognition head will not converge.Yes, I used a pretrained model. However, it can be seen that the detection head has not converged because the loss_bezier is very large, which is seriously different from the official log display.
However, I don't understand why the recognition head cannot converge. Is it because of the size of the data set or something else?
Do you have any experiences to share?
Could you provide your training log, especially before the first epoch? I doubt that the pretrained model didn't load successfully. So the loss is higher than the official version and your model cannot converge.
Bin-ze
commented
5 months ago Hi, I use 3 v100, single card batch=1, official batch 2*8=16, so I adjusted the learning rate to the official 1/10 (although it should be 1/5 according to the equivalent scaling), but using pretrain I Think it is reasonable to use a smaller lr
this is my log local：
System environment: sys.platform: linux Python: 3.8.17 | packaged by conda-forge | (default, Jun 16 2023, 07:06:00) [GCC 11.4.0] CUDA available: True numpy_random_seed: 882768434 GPU 0,1,2: NVIDIA Tesla V100-PCIE-16GB CUDA_HOME: /usr/local/cuda-11.3 NVCC: Cuda compilation tools, release 11.3, V11.3.109 GCC: gcc (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0 PyTorch: 1.12.1+cu113 PyTorch compiling details: PyTorch built with:
Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=11.3, CUDNN_VERSION=8.3.2, CXX_COMPILER=/opt/rh/devtoolset-9/root/usr/bin/c++, CXX_FLAGS= -fabi-version=11 -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_KINETO -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DSYMBOLICATE_MOBILE_DEBUG_HANDLE -DEDGE_PROFILER_USE_KINETO -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-unused-parameter -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Werror=cast-function-type -Wno-stringop-overflow, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_VERSION=1.12.1, USE_CUDA=ON, USE_CUDNN=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=OFF, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, USE_ROCM=OFF,
TorchVision: 0.13.1+cu113 OpenCV: 4.8.0 MMEngine: 0.8.4
2024/01/26 21:30:59 - mmengine - INFO - Config: bbox_coder = dict(type='mmdet.DistancePointBBoxCoder') custom_imports = dict( allow_failed_imports=False, imports=[ 'abcnet', ]) default_hooks = dict( checkpoint=dict(interval=20, type='CheckpointHook'), logger=dict(interval=5, type='LoggerHook'), param_scheduler=dict(type='ParamSchedulerHook'), sampler_seed=dict(type='DistSamplerSeedHook'), sync_buffer=dict(type='SyncBuffersHook'), timer=dict(type='IterTimerHook'), visualization=dict( draw_gt=False, draw_pred=False, enable=False, interval=1, show=False, type='VisualizationHook')) default_scope = 'mmocr' dictionary = dict( dict_file= '/home/guozebin/work_code/chenzhao/mmocr/projects/ABCNet/config/abcnet/../../dicts/abcnet.txt', same_start_end=False, type='Dictionary', with_end=False, with_padding=True, with_start=False, with_unknown=True) env_cfg = dict( cudnn_benchmark=False, dist_cfg=dict(backend='nccl'), mp_cfg=dict(mp_start_method='fork', opencv_num_threads=0)) find_unused_parameters = True icdar2015_textspotting_data_root = 'data/icdar2015' icdar2015_textspotting_test = dict( ann_file='textspotting_test.json', data_root='data/icdar2015', pipeline=[ dict(color_type='color_ignore_orientation', type='LoadImageFromFile'), dict( backend='pillow', keep_ratio=True, scale=( 2000, 4000, ), type='Resize'), dict( type='LoadOCRAnnotations', with_bbox=True, with_label=True, with_polygon=True, with_text=True), dict( meta_keys=( 'img_path', 'ori_shape', 'img_shape', 'scale_factor', ), type='PackTextDetInputs'), ], test_mode=True, type='OCRDataset') icdar2015_textspotting_train = dict( ann_file='textspotting_train.json', data_root='data/icdar2015', pipeline=[ dict(color_type='color_ignore_orientation', type='LoadImageFromFile'), dict( type='LoadOCRAnnotations', with_bbox=True, with_label=True, with_polygon=True, with_text=True), dict(type='RemoveIgnored'), dict(min_side_ratio=0.1, type='RandomCrop'), dict( max_angle=30, pad_with_fixed_color=True, type='RandomRotate', use_canvas=True), dict( keep_ratio=True, scales=[ ( 980, 2900, ), ( 1044, 2900, ), ( 1108, 2900, ), ( 1172, 2900, ), ( 1236, 2900, ), ( 1300, 2900, ), ( 1364, 2900, ), ( 1428, 2900, ), ( 1492, 2900, ), ], type='RandomChoiceResize'), dict( meta_keys=( 'img_path', 'ori_shape', 'img_shape', 'scale_factor', ), type='PackTextDetInputs'), ], type='OCRDataset') launcher = 'pytorch' load_from = '/home/guozebin/work_code/chenzhao/mmocr/ckpt/abcnet_resnet50_fpn_500e_icdar2015-326ac6f4.pth' log_level = 'INFO' log_processor = dict(by_epoch=True, type='LogProcessor', window_size=10) model = dict( backbone=dict( depth=50, frozen_stages=1, init_cfg=dict( checkpoint='open-mmlab://detectron2/resnet50_caffe', type='Pretrained'), norm_cfg=dict(requires_grad=False, type='BN'), norm_eval=True, num_stages=4, out_indices=( 0, 1, 2, 3, ), style='caffe', type='mmdet.ResNet'), data_preprocessor=dict( bgr_to_rgb=False, mean=[ 103.53, 116.28, 123.675, ], pad_size_divisor=32, std=[ 1, 1, 1, ], type='TextDetDataPreprocessor'), det_head=dict( centerness_on_reg=True, conv_bias=True, dcn_on_last_conv=False, feat_channels=256, in_channels=256, init_cfg=dict( layer='Conv2d', override=dict( bias=-4.59511985013459, name='conv_cls', std=0.01, type='Normal'), std=0.01, type='Normal'), module_loss=dict( bbox_coder=dict(type='mmdet.DistancePointBBoxCoder'), center_sample_radius=1.5, center_sampling=True, loss_bbox=dict(loss_weight=1.0, type='mmdet.GIoULoss'), loss_centerness=dict( loss_weight=1.0, type='mmdet.CrossEntropyLoss', use_sigmoid=True), loss_cls=dict( alpha=0.25, gamma=2.0, loss_weight=1.0, type='mmdet.FocalLoss', use_sigmoid=True), norm_on_bbox=True, num_classes=1, strides=[ 8, 16, 32, 64, 128, ], type='ABCNetDetModuleLoss'), norm_on_bbox=True, num_classes=1, postprocessor=dict( bbox_coder=dict(type='mmdet.DistancePointBBoxCoder'), strides=[ 8, 16, 32, 64, 128, ], test_cfg=dict( nms=dict(iou_threshold=0.5, type='nms'), nms_pre=1000, score_thr=0.3), type='ABCNetDetPostprocessor', use_sigmoid_cls=True, with_bezier=True), stacked_convs=4, strides=[ 8, 16, 32, 64, 128, ], type='ABCNetDetHead', use_scale=False, use_sigmoid_cls=True, with_bezier=True), neck=dict( add_extra_convs='on_output', in_channels=[ 256, 512, 1024, 2048, ], num_outs=6, out_channels=256, relu_before_extra_convs=True, start_level=0, type='mmdet.FPN'), postprocessor=dict( rescale_fields=[ 'polygons', 'bboxes', 'beziers', ], type='ABCNetPostprocessor'), roi_head=dict( rec_head=dict( backbone=dict(type='ABCNetRecBackbone'), decoder=dict( dictionary=dict( dict_file= '/home/guozebin/work_code/chenzhao/mmocr/projects/ABCNet/config/abcnet/../../dicts/abcnet.txt', same_start_end=False, type='Dictionary', with_end=False, with_padding=True, with_start=False, with_unknown=True), max_seq_len=25, module_loss=dict( ignore_char=-1, ignore_first_char=False, reduction='mean', type='CEModuleLoss'), postprocessor=dict( ignore_chars=[ 'padding', 'unknown', ], type='AttentionPostprocessor'), type='ABCNetRecDecoder'), encoder=dict(type='ABCNetRecEncoder'), type='ABCNetRec'), roi_extractor=dict( featmap_strides=[ 4, 8, 16, ], out_channels=256, roi_layer=dict( output_size=( 8, 32, ), sampling_ratio=1.0, type='BezierAlign'), type='BezierRoIExtractor'), type='RecRoIHead'), type='ABCNet') norm_on_bbox = True num_classes = 1 optim_wrapper = dict( clip_grad=dict(clip_value=1, type='value'), optimizer=dict(lr=0.0001, momentum=0.9, type='SGD', weight_decay=0.0001), type='OptimWrapper') param_scheduler = [ dict(by_epoch=False, end=1000, start_factor=0.001, type='LinearLR'), ] randomness = dict(seed=None) resume = False strides = [ 8, 16, 32, 64, 128, ] test_cfg = dict(type='TestLoop') test_dataloader = dict( batch_size=1, dataset=dict( ann_file='textspotting_test.json', data_root='data/icdar2015', pipeline=[ dict( color_type='color_ignore_orientation', type='LoadImageFromFile'), dict( backend='pillow', keep_ratio=True, scale=( 2000, 4000, ), type='Resize'), dict( type='LoadOCRAnnotations', with_bbox=True, with_label=True, with_polygon=True, with_text=True), dict( meta_keys=( 'img_path', 'ori_shape', 'img_shape', 'scale_factor', ), type='PackTextDetInputs'), ], test_mode=True, type='OCRDataset'), num_workers=4, persistent_workers=True, sampler=dict(shuffle=False, type='DefaultSampler')) test_evaluator = [ dict(type='E2EHmeanIOUMetric'), dict(type='HmeanIOUMetric'), ] test_pipeline = [ dict(color_type='color_ignore_orientation', type='LoadImageFromFile'), dict( backend='pillow', keep_ratio=True, scale=( 2000, 4000, ), type='Resize'), dict( type='LoadOCRAnnotations', with_bbox=True, with_label=True, with_polygon=True, with_text=True), dict( meta_keys=( 'img_path', 'ori_shape', 'img_shape', 'scale_factor', ), type='PackTextDetInputs'), ] train_cfg = dict(max_epochs=500, type='EpochBasedTrainLoop', val_interval=1) train_dataloader = dict( batch_size=1, dataset=dict( ann_file='textspotting_train.json', data_root='data/icdar2015', pipeline=[ dict( color_type='color_ignore_orientation', type='LoadImageFromFile'), dict( type='LoadOCRAnnotations', with_bbox=True, with_label=True, with_polygon=True, with_text=True), dict(type='RemoveIgnored'), dict(min_side_ratio=0.1, type='RandomCrop'), dict( max_angle=30, pad_with_fixed_color=True, type='RandomRotate', use_canvas=True), dict( keep_ratio=True, scales=[ ( 980, 2900, ), ( 1044, 2900, ), ( 1108, 2900, ), ( 1172, 2900, ), ( 1236, 2900, ), ( 1300, 2900, ), ( 1364, 2900, ), ( 1428, 2900, ), ( 1492, 2900, ), ], type='RandomChoiceResize'), dict( meta_keys=( 'img_path', 'ori_shape', 'img_shape', 'scale_factor', ), type='PackTextDetInputs'), ], type='OCRDataset'), num_workers=8, persistent_workers=True, sampler=dict(shuffle=True, type='DefaultSampler')) train_pipeline = [ dict(color_type='color_ignore_orientation', type='LoadImageFromFile'), dict( type='LoadOCRAnnotations', with_bbox=True, with_label=True, with_polygon=True, with_text=True), dict(type='RemoveIgnored'), dict(min_side_ratio=0.1, type='RandomCrop'), dict( max_angle=30, pad_with_fixed_color=True, type='RandomRotate', use_canvas=True), dict( keep_ratio=True, scales=[ ( 980, 2900, ), ( 1044, 2900, ), ( 1108, 2900, ), ( 1172, 2900, ), ( 1236, 2900, ), ( 1300, 2900, ), ( 1364, 2900, ), ( 1428, 2900, ), ( 1492, 2900, ), ], type='RandomChoiceResize'), dict( meta_keys=( 'img_path', 'ori_shape', 'img_shape', 'scale_factor', ), type='PackTextDetInputs'), ] use_sigmoid_cls = True val_cfg = dict(type='ValLoop') val_dataloader = dict( batch_size=1, dataset=dict( ann_file='textspotting_test.json', data_root='data/icdar2015', pipeline=[ dict( color_type='color_ignore_orientation', type='LoadImageFromFile'), dict( backend='pillow', keep_ratio=True, scale=( 2000, 4000, ), type='Resize'), dict( type='LoadOCRAnnotations', with_bbox=True, with_label=True, with_polygon=True, with_text=True), dict( meta_keys=( 'img_path', 'ori_shape', 'img_shape', 'scale_factor', ), type='PackTextDetInputs'), ], test_mode=True, type='OCRDataset'), num_workers=4, persistent_workers=True, sampler=dict(shuffle=False, type='DefaultSampler')) val_evaluator = [ dict(type='E2EHmeanIOUMetric'), dict(type='HmeanIOUMetric'), ] vis_backends = [ dict(type='LocalVisBackend'), ] visualizer = dict( name='visualizer', type='TextSpottingLocalVisualizer', vis_backends=[ dict(type='LocalVisBackend'), ]) with_bezier = True work_dir = 'work_dirs/'
2024/01/26 21:31:03 - mmengine - INFO - Hooks will be executed in the following order:
before_run:
(VERY_HIGH ) RuntimeInfoHook
(BELOW_NORMAL) LoggerHook
before_train:
(VERY_HIGH ) RuntimeInfoHook
(NORMAL ) IterTimerHook
(VERY_LOW ) CheckpointHook
before_train_epoch:
(VERY_HIGH ) RuntimeInfoHook
(NORMAL ) IterTimerHook
(NORMAL ) DistSamplerSeedHook
before_train_iter:
(VERY_HIGH ) RuntimeInfoHook
(NORMAL ) IterTimerHook
after_train_iter:
(VERY_HIGH ) RuntimeInfoHook
(NORMAL ) IterTimerHook
(BELOW_NORMAL) LoggerHook
(LOW ) ParamSchedulerHook
(VERY_LOW ) CheckpointHook
after_train_epoch:
(NORMAL ) IterTimerHook
(NORMAL ) SyncBuffersHook
(LOW ) ParamSchedulerHook
(VERY_LOW ) CheckpointHook
before_val: (VERY_HIGH ) RuntimeInfoHook
before_val_epoch:
(NORMAL ) IterTimerHook
(NORMAL ) SyncBuffersHook
before_val_iter: (NORMAL ) IterTimerHook
after_val_iter:
(NORMAL ) IterTimerHook
(NORMAL ) VisualizationHook
(BELOW_NORMAL) LoggerHook
after_val_epoch:
(VERY_HIGH ) RuntimeInfoHook
(NORMAL ) IterTimerHook
(BELOW_NORMAL) LoggerHook
(LOW ) ParamSchedulerHook
(VERY_LOW ) CheckpointHook
after_val: (VERY_HIGH ) RuntimeInfoHook
after_train:
(VERY_HIGH ) RuntimeInfoHook
(VERY_LOW ) CheckpointHook
before_test: (VERY_HIGH ) RuntimeInfoHook
before_test_epoch: (NORMAL ) IterTimerHook
before_test_iter: (NORMAL ) IterTimerHook
after_test_iter:
(NORMAL ) IterTimerHook
(NORMAL ) VisualizationHook
(BELOW_NORMAL) LoggerHook
after_test_epoch:
(VERY_HIGH ) RuntimeInfoHook
(NORMAL ) IterTimerHook
(BELOW_NORMAL) LoggerHook
after_test: (VERY_HIGH ) RuntimeInfoHook
after_run: (BELOW_NORMAL) LoggerHook
2024/01/26 21:31:07 - mmengine - INFO - load model from: open-mmlab://detectron2/resnet50_caffe 2024/01/26 21:31:07 - mmengine - INFO - Loads checkpoint by openmmlab backend from path: open-mmlab://detectron2/resnet50_caffe 2024/01/26 21:31:07 - mmengine - WARNING - The model and loaded state dict do not match exactly
unexpected key in source state_dict: conv1.bias
Name of parameter - Initialization information
backbone.conv1.weight - torch.Size([64, 3, 7, 7]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.bn1.weight - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.bn1.bias - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.0.conv1.weight - torch.Size([64, 64, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.0.bn1.weight - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.0.bn1.bias - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.0.conv2.weight - torch.Size([64, 64, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.0.bn2.weight - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.0.bn2.bias - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.0.conv3.weight - torch.Size([256, 64, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.0.bn3.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.0.bn3.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.0.downsample.0.weight - torch.Size([256, 64, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.0.downsample.1.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.0.downsample.1.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.1.conv1.weight - torch.Size([64, 256, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.1.bn1.weight - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.1.bn1.bias - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.1.conv2.weight - torch.Size([64, 64, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.1.bn2.weight - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.1.bn2.bias - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.1.conv3.weight - torch.Size([256, 64, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.1.bn3.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.1.bn3.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.2.conv1.weight - torch.Size([64, 256, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.2.bn1.weight - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.2.bn1.bias - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.2.conv2.weight - torch.Size([64, 64, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.2.bn2.weight - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.2.bn2.bias - torch.Size([64]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.2.conv3.weight - torch.Size([256, 64, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.2.bn3.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer1.2.bn3.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.0.conv1.weight - torch.Size([128, 256, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.0.bn1.weight - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.0.bn1.bias - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.0.conv2.weight - torch.Size([128, 128, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.0.bn2.weight - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.0.bn2.bias - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.0.conv3.weight - torch.Size([512, 128, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.0.bn3.weight - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.0.bn3.bias - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.0.downsample.0.weight - torch.Size([512, 256, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.0.downsample.1.weight - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.0.downsample.1.bias - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.1.conv1.weight - torch.Size([128, 512, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.1.bn1.weight - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.1.bn1.bias - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.1.conv2.weight - torch.Size([128, 128, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.1.bn2.weight - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.1.bn2.bias - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.1.conv3.weight - torch.Size([512, 128, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.1.bn3.weight - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.1.bn3.bias - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.2.conv1.weight - torch.Size([128, 512, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.2.bn1.weight - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.2.bn1.bias - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.2.conv2.weight - torch.Size([128, 128, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.2.bn2.weight - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.2.bn2.bias - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.2.conv3.weight - torch.Size([512, 128, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.2.bn3.weight - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.2.bn3.bias - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.3.conv1.weight - torch.Size([128, 512, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.3.bn1.weight - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.3.bn1.bias - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.3.conv2.weight - torch.Size([128, 128, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.3.bn2.weight - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.3.bn2.bias - torch.Size([128]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.3.conv3.weight - torch.Size([512, 128, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.3.bn3.weight - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer2.3.bn3.bias - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.0.conv1.weight - torch.Size([256, 512, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.0.bn1.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.0.bn1.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.0.conv2.weight - torch.Size([256, 256, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.0.bn2.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.0.bn2.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.0.conv3.weight - torch.Size([1024, 256, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.0.bn3.weight - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.0.bn3.bias - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.0.downsample.0.weight - torch.Size([1024, 512, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.0.downsample.1.weight - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.0.downsample.1.bias - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.1.conv1.weight - torch.Size([256, 1024, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.1.bn1.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.1.bn1.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.1.conv2.weight - torch.Size([256, 256, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.1.bn2.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.1.bn2.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.1.conv3.weight - torch.Size([1024, 256, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.1.bn3.weight - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.1.bn3.bias - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.2.conv1.weight - torch.Size([256, 1024, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.2.bn1.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.2.bn1.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.2.conv2.weight - torch.Size([256, 256, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.2.bn2.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.2.bn2.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.2.conv3.weight - torch.Size([1024, 256, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.2.bn3.weight - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.2.bn3.bias - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.3.conv1.weight - torch.Size([256, 1024, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.3.bn1.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.3.bn1.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.3.conv2.weight - torch.Size([256, 256, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.3.bn2.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.3.bn2.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.3.conv3.weight - torch.Size([1024, 256, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.3.bn3.weight - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.3.bn3.bias - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.4.conv1.weight - torch.Size([256, 1024, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.4.bn1.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.4.bn1.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.4.conv2.weight - torch.Size([256, 256, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.4.bn2.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.4.bn2.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.4.conv3.weight - torch.Size([1024, 256, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.4.bn3.weight - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.4.bn3.bias - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.5.conv1.weight - torch.Size([256, 1024, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.5.bn1.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.5.bn1.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.5.conv2.weight - torch.Size([256, 256, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.5.bn2.weight - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.5.bn2.bias - torch.Size([256]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.5.conv3.weight - torch.Size([1024, 256, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.5.bn3.weight - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer3.5.bn3.bias - torch.Size([1024]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.0.conv1.weight - torch.Size([512, 1024, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.0.bn1.weight - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.0.bn1.bias - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.0.conv2.weight - torch.Size([512, 512, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.0.bn2.weight - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.0.bn2.bias - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.0.conv3.weight - torch.Size([2048, 512, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.0.bn3.weight - torch.Size([2048]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.0.bn3.bias - torch.Size([2048]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.0.downsample.0.weight - torch.Size([2048, 1024, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.0.downsample.1.weight - torch.Size([2048]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.0.downsample.1.bias - torch.Size([2048]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.1.conv1.weight - torch.Size([512, 2048, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.1.bn1.weight - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.1.bn1.bias - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.1.conv2.weight - torch.Size([512, 512, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.1.bn2.weight - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.1.bn2.bias - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.1.conv3.weight - torch.Size([2048, 512, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.1.bn3.weight - torch.Size([2048]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.1.bn3.bias - torch.Size([2048]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.2.conv1.weight - torch.Size([512, 2048, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.2.bn1.weight - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.2.bn1.bias - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.2.conv2.weight - torch.Size([512, 512, 3, 3]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.2.bn2.weight - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.2.bn2.bias - torch.Size([512]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.2.conv3.weight - torch.Size([2048, 512, 1, 1]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.2.bn3.weight - torch.Size([2048]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
backbone.layer4.2.bn3.bias - torch.Size([2048]): PretrainedInit: load from open-mmlab://detectron2/resnet50_caffe
neck.lateral_convs.0.conv.weight - torch.Size([256, 256, 1, 1]): XavierInit: gain=1, distribution=uniform, bias=0
neck.lateral_convs.0.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
neck.lateral_convs.1.conv.weight - torch.Size([256, 512, 1, 1]): XavierInit: gain=1, distribution=uniform, bias=0
neck.lateral_convs.1.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
neck.lateral_convs.2.conv.weight - torch.Size([256, 1024, 1, 1]): XavierInit: gain=1, distribution=uniform, bias=0
neck.lateral_convs.2.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
neck.lateral_convs.3.conv.weight - torch.Size([256, 2048, 1, 1]): XavierInit: gain=1, distribution=uniform, bias=0
neck.lateral_convs.3.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
neck.fpn_convs.0.conv.weight - torch.Size([256, 256, 3, 3]): XavierInit: gain=1, distribution=uniform, bias=0
neck.fpn_convs.0.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
neck.fpn_convs.1.conv.weight - torch.Size([256, 256, 3, 3]): XavierInit: gain=1, distribution=uniform, bias=0
neck.fpn_convs.1.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
neck.fpn_convs.2.conv.weight - torch.Size([256, 256, 3, 3]): XavierInit: gain=1, distribution=uniform, bias=0
neck.fpn_convs.2.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
neck.fpn_convs.3.conv.weight - torch.Size([256, 256, 3, 3]): XavierInit: gain=1, distribution=uniform, bias=0
neck.fpn_convs.3.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
neck.fpn_convs.4.conv.weight - torch.Size([256, 256, 3, 3]): XavierInit: gain=1, distribution=uniform, bias=0
neck.fpn_convs.4.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
neck.fpn_convs.5.conv.weight - torch.Size([256, 256, 3, 3]): XavierInit: gain=1, distribution=uniform, bias=0
neck.fpn_convs.5.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.cls_convs.0.conv.weight - torch.Size([256, 256, 3, 3]): NormalInit: mean=0, std=0.01, bias=0
det_head.cls_convs.0.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.cls_convs.0.gn.weight - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.cls_convs.0.gn.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.cls_convs.1.conv.weight - torch.Size([256, 256, 3, 3]): NormalInit: mean=0, std=0.01, bias=0
det_head.cls_convs.1.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.cls_convs.1.gn.weight - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.cls_convs.1.gn.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.cls_convs.2.conv.weight - torch.Size([256, 256, 3, 3]): NormalInit: mean=0, std=0.01, bias=0
det_head.cls_convs.2.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.cls_convs.2.gn.weight - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.cls_convs.2.gn.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.cls_convs.3.conv.weight - torch.Size([256, 256, 3, 3]): NormalInit: mean=0, std=0.01, bias=0
det_head.cls_convs.3.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.cls_convs.3.gn.weight - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.cls_convs.3.gn.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.reg_convs.0.conv.weight - torch.Size([256, 256, 3, 3]): NormalInit: mean=0, std=0.01, bias=0
det_head.reg_convs.0.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.reg_convs.0.gn.weight - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.reg_convs.0.gn.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.reg_convs.1.conv.weight - torch.Size([256, 256, 3, 3]): NormalInit: mean=0, std=0.01, bias=0
det_head.reg_convs.1.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.reg_convs.1.gn.weight - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.reg_convs.1.gn.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.reg_convs.2.conv.weight - torch.Size([256, 256, 3, 3]): NormalInit: mean=0, std=0.01, bias=0
det_head.reg_convs.2.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.reg_convs.2.gn.weight - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.reg_convs.2.gn.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.reg_convs.3.conv.weight - torch.Size([256, 256, 3, 3]): NormalInit: mean=0, std=0.01, bias=0
det_head.reg_convs.3.conv.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.reg_convs.3.gn.weight - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.reg_convs.3.gn.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
det_head.conv_cls.weight - torch.Size([1, 256, 3, 3]): NormalInit: mean=0, std=0.01, bias=-4.59511985013459
det_head.conv_cls.bias - torch.Size([1]): NormalInit: mean=0, std=0.01, bias=-4.59511985013459
det_head.conv_reg.weight - torch.Size([4, 256, 3, 3]): NormalInit: mean=0, std=0.01, bias=0
det_head.conv_reg.bias - torch.Size([4]): NormalInit: mean=0, std=0.01, bias=0
det_head.conv_bezier.weight - torch.Size([16, 256, 3, 3]): NormalInit: mean=0, std=0.01, bias=0
det_head.conv_bezier.bias - torch.Size([16]): NormalInit: mean=0, std=0.01, bias=0
det_head.conv_centerness.weight - torch.Size([1, 256, 3, 3]): NormalInit: mean=0, std=0.01, bias=0
det_head.conv_centerness.bias - torch.Size([1]): NormalInit: mean=0, std=0.01, bias=0
det_head.scales.0.scale - torch.Size([]):
The value is the same before and after calling init_weights of ABCNet
det_head.scales.1.scale - torch.Size([]):
The value is the same before and after calling init_weights of ABCNet
det_head.scales.2.scale - torch.Size([]):
The value is the same before and after calling init_weights of ABCNet
det_head.scales.3.scale - torch.Size([]):
The value is the same before and after calling init_weights of ABCNet
det_head.scales.4.scale - torch.Size([]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.backbone.convs.0.conv.weight - torch.Size([256, 256, 3, 3]):
Initialized by user-defined init_weights in ConvModule
roi_head.rec_head.backbone.convs.0.bn.weight - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.backbone.convs.0.bn.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.backbone.convs.1.conv.weight - torch.Size([256, 256, 3, 3]):
Initialized by user-defined init_weights in ConvModule
roi_head.rec_head.backbone.convs.1.bn.weight - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.backbone.convs.1.bn.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.backbone.convs.2.conv.weight - torch.Size([256, 256, 3, 3]):
Initialized by user-defined init_weights in ConvModule
roi_head.rec_head.backbone.convs.2.gn.weight - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.backbone.convs.2.gn.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.backbone.convs.3.conv.weight - torch.Size([256, 256, 3, 3]):
Initialized by user-defined init_weights in ConvModule
roi_head.rec_head.backbone.convs.3.gn.weight - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.backbone.convs.3.gn.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.encoder.layer.rnn.weight_ih_l0 - torch.Size([1024, 256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.encoder.layer.rnn.weight_hh_l0 - torch.Size([1024, 256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.encoder.layer.rnn.bias_ih_l0 - torch.Size([1024]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.encoder.layer.rnn.bias_hh_l0 - torch.Size([1024]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.encoder.layer.rnn.weight_ih_l0_reverse - torch.Size([1024, 256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.encoder.layer.rnn.weight_hh_l0_reverse - torch.Size([1024, 256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.encoder.layer.rnn.bias_ih_l0_reverse - torch.Size([1024]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.encoder.layer.rnn.bias_hh_l0_reverse - torch.Size([1024]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.encoder.layer.embedding.weight - torch.Size([256, 512]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.encoder.layer.embedding.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.decoder.embedding.weight - torch.Size([97, 256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.decoder.attn_combine.weight - torch.Size([256, 512]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.decoder.attn_combine.bias - torch.Size([256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.decoder.gru.weight_ih_l0 - torch.Size([768, 256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.decoder.gru.weight_hh_l0 - torch.Size([768, 256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.decoder.gru.bias_ih_l0 - torch.Size([768]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.decoder.gru.bias_hh_l0 - torch.Size([768]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.decoder.out.weight - torch.Size([97, 256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.decoder.out.bias - torch.Size([97]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.decoder.vat.weight - torch.Size([1, 256]):
The value is the same before and after calling init_weights of ABCNet
roi_head.rec_head.decoder.vat.bias - torch.Size([1]):
The value is the same before and after calling init_weights of ABCNet
2024/01/26 21:31:07 - mmengine - INFO - Load checkpoint from /home/guozebin/work_code/chenzhao/mmocr/ckpt/abcnet_resnet50_fpn_500e_icdar2015-326ac6f4.pth
2024/01/26 21:31:07 - mmengine - WARNING - "FileClient" will be deprecated in future. Please use io functions in https://mmengine.readthedocs.io/en/latest/api/fileio.html#file-io
2024/01/26 21:31:07 - mmengine - WARNING - "HardDiskBackend" is the alias of "LocalBackend" and the former will be deprecated in future.
2024/01/26 21:31:07 - mmengine - INFO - Checkpoints will be saved to /home/guozebin/work_code/chenzhao/mmocr/projects/ABCNet/work_dirs.
2024/01/26 21:31:20 - mmengine - INFO - Epoch(train) [1][ 5/334] lr: 5.0000e-07 eta: 4 days, 20:01:04 time: 2.5011 data_time: 1.1496 memory: 4739 loss: 80.0229 loss_cls: 0.1096 loss_bbox: 0.0759 loss_centerness: 0.5014 loss_bezier: 78.2609 loss_ce: 1.0751
2024/01/26 21:31:22 - mmengine - INFO - Epoch(train) [1][ 10/334] lr: 1.0000e-06 eta: 2 days, 21:20:22 time: 1.4948 data_time: 0.5781 memory: 4836 loss: 84.2790 loss_cls: 0.1101 loss_bbox: 0.0855 loss_centerness: 0.5475 loss_bezier: 82.6358 loss_ce: 0.9001
2024/01/26 21:31:24 - mmengine - INFO - Epoch(train) [1][ 15/334] lr: 1.5000e-06 eta: 2 days, 4:48:33 time: 0.4572 data_time: 0.0055 memory: 5016 loss: 74.7471 loss_cls: 0.0673 loss_bbox: 0.0788 loss_centerness: 0.5478 loss_bezier: 73.0744 loss_ce: 0.9788
2024/01/26 21:31:27 - mmengine - INFO - Epoch(train) [1][ 20/334] lr: 2.0000e-06 eta: 1 day, 21:14:26 time: 0.4559 data_time: 0.0052 memory: 4563 loss: 81.4511 loss_cls: 0.0765 loss_bbox: 0.0871 loss_centerness: 0.6037 loss_bezier: 79.5800 loss_ce: 1.1038
2024/01/26 21:31:29 - mmengine - INFO - Epoch(train) [1][ 25/334] lr: 2.5000e-06 eta: 1 day, 16:08:52 time: 0.4562 data_time: 0.0062 memory: 4580 loss: 119.4200 loss_cls: 0.1193 loss_bbox: 0.1129 loss_centerness: 0.6994 loss_bezier: 117.5438 loss_ce: 0.9447
2024/01/26 21:31:31 - mmengine - INFO - Epoch(train) [1][ 30/334] lr: 3.0000e-06 eta: 1 day, 13:09:58 time: 0.4533 data_time: 0.0063 memory: 4745 loss: 102.3057 loss_cls: 0.0933 loss_bbox: 0.0978 loss_centerness: 0.6362 loss_bezier: 100.4807 loss_ce: 0.9978
2024/01/26 21:31:34 - mmengine - INFO - Epoch(train) [1][ 35/334] lr: 3.5000e-06 eta: 1 day, 10:53:30 time: 0.4691 data_time: 0.0063 memory: 4943 loss: 76.5474 loss_cls: 0.0737 loss_bbox: 0.0869 loss_centerness: 0.5503 loss_bezier: 74.8164 loss_ce: 1.0201
2024/01/26 21:31:36 - mmengine - INFO - Epoch(train) [1][ 40/334] lr: 4.0000e-06 eta: 1 day, 9:09:04 time: 0.4552 data_time: 0.0066 memory: 4060 loss: 77.7166 loss_cls: 0.0607 loss_bbox: 0.0849 loss_centerness: 0.5446 loss_bezier: 76.1711 loss_ce: 0.8553
2024/01/26 21:31:38 - mmengine - INFO - Epoch(train) [1][ 45/334] lr: 4.5000e-06 eta: 1 day, 7:35:14 time: 0.4319 data_time: 0.0064 memory: 4689 loss: 82.8328 loss_cls: 0.0787 loss_bbox: 0.1086 loss_centerness: 0.6355 loss_bezier: 81.2414 loss_ce: 0.7686
2024/01/26 21:31:40 - mmengine - INFO - Epoch(train) [1][ 50/334] lr: 5.0000e-06 eta: 1 day, 6:23:04 time: 0.4167 data_time: 0.0060 memory: 4193 loss: 80.0396 loss_cls: 0.0762 loss_bbox: 0.0924 loss_centerness: 0.5722 loss_bezier: 78.4566 loss_ce: 0.8422
2024/01/26 21:31:42 - mmengine - INFO - Epoch(train) [1][ 55/334] lr: 5.5000e-06 eta: 1 day, 5:26:45 time: 0.4274 data_time: 0.0058 memory: 3738 loss: 65.4770 loss_cls: 0.0669 loss_bbox: 0.0621 loss_centerness: 0.4887 loss_bezier: 64.0226 loss_ce: 0.8367
2024/01/26 21:31:45 - mmengine - INFO - Epoch(train) [1][ 60/334] lr: 6.0000e-06 eta: 1 day, 4:49:05 time: 0.4527 data_time: 0.0061 memory: 5859 loss: 100.1543 loss_cls: 0.1282 loss_bbox: 0.1238 loss_centerness: 0.6831 loss_bezier: 98.3492 loss_ce: 0.8699
2024/01/26 21:31:47 - mmengine - INFO - Epoch(train) [1][ 65/334] lr: 6.5000e-06 eta: 1 day, 4:10:35 time: 0.4573 data_time: 0.0067 memory: 4377 loss: 104.2979 loss_cls: 0.1500 loss_bbox: 0.1296 loss_centerness: 0.6830 loss_bezier: 102.2921 loss_ce: 1.0432
2024/01/26 21:31:49 - mmengine - INFO - Epoch(train) [1][ 70/334] lr: 7.0000e-06 eta: 1 day, 3:33:16 time: 0.4310 data_time: 0.0076 memory: 4371 loss: 91.9972 loss_cls: 0.1127 loss_bbox: 0.1020 loss_centerness: 0.6043 loss_bezier: 90.2145 loss_ce: 0.9636
2024/01/26 21:31:51 - mmengine - INFO - Epoch(train) [1][ 75/334] lr: 7.5000e-06 eta: 1 day, 3:07:35 time: 0.4381 data_time: 0.0073 memory: 5935 loss: 76.3080 loss_cls: 0.0834 loss_bbox: 0.0833 loss_centerness: 0.5238 loss_bezier: 74.7150 loss_ce: 0.9026
2024/01/26 21:31:53 - mmengine - INFO - Epoch(train) [1][ 80/334] lr: 8.0000e-06 eta: 1 day, 2:37:15 time: 0.4335 data_time: 0.0067 memory: 4193 loss: 58.7434 loss_cls: 0.0729 loss_bbox: 0.0639 loss_centerness: 0.4451 loss_bezier: 57.1829 loss_ce: 0.9784
2024/01/26 21:31:55 - mmengine - INFO - Epoch(train) [1][ 85/334] lr: 8.5000e-06 eta: 1 day, 2:15:35 time: 0.4264 data_time: 0.0063 memory: 3673 loss: 58.4728 loss_cls: 0.0677 loss_bbox: 0.0745 loss_centerness: 0.4748 loss_bezier: 56.9404 loss_ce: 0.9154
2024/01/26 21:31:58 - mmengine - INFO - Epoch(train) [1][ 90/334] lr: 9.0000e-06 eta: 1 day, 2:00:22 time: 0.4551 data_time: 0.0058 memory: 5542 loss: 79.6476 loss_cls: 0.1198 loss_bbox: 0.1282 loss_centerness: 0.6383 loss_bezier: 77.8792 loss_ce: 0.8821
2024/01/26 21:32:00 - mmengine - INFO - Epoch(train) [1][ 95/334] lr: 9.5000e-06 eta: 1 day, 1:39:37 time: 0.4439 data_time: 0.0058 memory: 6427 loss: 72.2142 loss_cls: 0.1107 loss_bbox: 0.1196 loss_centerness: 0.5859 loss_bezier: 70.2818 loss_ce: 1.1162
2024/01/26 21:32:02 - mmengine - INFO - Epoch(train) [1][100/334] lr: 1.0000e-05 eta: 1 day, 1:26:24 time: 0.4392 data_time: 0.0062 memory: 5713 loss: 59.2292 loss_cls: 0.0712 loss_bbox: 0.0913 loss_centerness: 0.4926 loss_bezier: 57.4408 loss_ce: 1.1334
2024/01/26 21:32:04 - mmengine - INFO - Epoch(train) [1][105/334] lr: 1.0500e-05 eta: 1 day, 1:12:22 time: 0.4510 data_time: 0.0071 memory: 6089 loss: 80.8816 loss_cls: 0.1146 loss_bbox: 0.1357 loss_centerness: 0.6378 loss_bezier: 79.2265 loss_ce: 0.7671
2024/01/26 21:32:07 - mmengine - INFO - Epoch(train) [1][110/334] lr: 1.1000e-05 eta: 1 day, 1:00:27 time: 0.4465 data_time: 0.0072 memory: 4471 loss: 90.9681 loss_cls: 0.1105 loss_bbox: 0.1366 loss_centerness: 0.6766 loss_bezier: 89.2868 loss_ce: 0.7575
2024/01/26 21:32:09 - mmengine - INFO - Epoch(train) [1][115/334] lr: 1.1500e-05 eta: 1 day, 0:51:19 time: 0.4570 data_time: 0.0070 memory: 5474 loss: 88.1288 loss_cls: 0.0803 loss_bbox: 0.1307 loss_centerness: 0.6303 loss_bezier: 86.4037 loss_ce: 0.8839
2024/01/26 21:32:11 - mmengine - INFO - Epoch(train) [1][120/334] lr: 1.2000e-05 eta: 1 day, 0:40:31 time: 0.4538 data_time: 0.0071 memory: 4467 loss: 77.1714 loss_cls: 0.0687 loss_bbox: 0.1313 loss_centerness: 0.5666 loss_bezier: 75.5313 loss_ce: 0.8735
2024/01/26 21:32:14 - mmengine - INFO - Epoch(train) [1][125/334] lr: 1.2500e-05 eta: 1 day, 0:35:22 time: 0.4649 data_time: 0.0074 memory: 5768 loss: 65.3111 loss_cls: 0.0745 loss_bbox: 0.1095 loss_centerness: 0.5028 loss_bezier: 63.6560 loss_ce: 0.9683
2024/01/26 21:32:16 - mmengine - INFO - Epoch(train) [1][130/334] lr: 1.3000e-05 eta: 1 day, 0:32:54 time: 0.4971 data_time: 0.0078 memory: 6013 loss: 47.1523 loss_cls: 0.0884 loss_bbox: 0.1005 loss_centerness: 0.4677 loss_bezier: 45.5999 loss_ce: 0.8958
2024/01/26 21:32:18 - mmengine - INFO - Epoch(train) [1][135/334] lr: 1.3500e-05 eta: 1 day, 0:19:37 time: 0.4544 data_time: 0.0072 memory: 5009 loss: 83.1653 loss_cls: 0.1107 loss_bbox: 0.1436 loss_centerness: 0.6190 loss_bezier: 81.5045 loss_ce: 0.7875
2024/01/26 21:32:20 - mmengine - INFO - Epoch(train) [1][140/334] lr: 1.4000e-05 eta: 1 day, 0:11:01 time: 0.4199 data_time: 0.0062 memory: 5420 loss: 107.2929 loss_cls: 0.1170 loss_bbox: 0.1522 loss_centerness: 0.6355 loss_bezier: 105.5978 loss_ce: 0.7905
2024/01/26 21:32:23 - mmengine - INFO - Epoch(train) [1][145/334] lr: 1.4500e-05 eta: 1 day, 0:07:17 time: 0.4610 data_time: 0.0057 memory: 5294 loss: 111.0490 loss_cls: 0.0893 loss_bbox: 0.1702 loss_centerness: 0.6066 loss_bezier: 109.3865 loss_ce: 0.7964
2024/01/26 21:32:25 - mmengine - INFO - Epoch(train) [1][150/334] lr: 1.5000e-05 eta: 1 day, 0:05:42 time: 0.4936 data_time: 0.0058 memory: 5443 loss: 86.5440 loss_cls: 0.0826 loss_bbox: 0.1665 loss_centerness: 0.5484 loss_bezier: 84.7832 loss_ce: 0.9633
2024/01/26 21:32:27 - mmengine - INFO - Epoch(train) [1][155/334] lr: 1.5500e-05 eta: 23:56:35 time: 0.4613 data_time: 0.0068 memory: 5189 loss: 62.8552 loss_cls: 0.0934 loss_bbox: 0.1478 loss_centerness: 0.5501 loss_bezier: 61.0990 loss_ce: 0.9649
2024/01/26 21:32:29 - mmengine - INFO - Epoch(train) [1][160/334] lr: 1.6000e-05 eta: 23:47:04 time: 0.4132 data_time: 0.0077 memory: 5393 loss: 61.0773 loss_cls: 0.0655 loss_bbox: 0.1471 loss_centerness: 0.5319 loss_bezier: 59.4582 loss_ce: 0.8746
2024/01/26 21:32:31 - mmengine - INFO - Epoch(train) [1][165/334] lr: 1.6500e-05 eta: 23:38:15 time: 0.4084 data_time: 0.0072 memory: 4628 loss: 58.5584 loss_cls: 0.0991 loss_bbox: 0.1539 loss_centerness: 0.5160 loss_bezier: 56.8251 loss_ce: 0.9642
2024/01/26 21:32:34 - mmengine - INFO - Epoch(train) [1][170/334] lr: 1.7000e-05 eta: 23:33:42 time: 0.4320 data_time: 0.0065 memory: 5546 loss: 52.6691 loss_cls: 0.1179 loss_bbox: 0.1633 loss_centerness: 0.5488 loss_bezier: 50.8770 loss_ce: 0.9620
2024/01/26 21:32:36 - mmengine - INFO - Epoch(train) [1][175/334] lr: 1.7500e-05 eta: 23:31:12 time: 0.4662 data_time: 0.0066 memory: 5460 loss: 51.5676 loss_cls: 0.0932 loss_bbox: 0.2016 loss_centerness: 0.5817 loss_bezier: 49.8051 loss_ce: 0.8860
2024/01/26 21:32:38 - mmengine - INFO - Epoch(train) [1][180/334] lr: 1.8000e-05 eta: 23:22:54 time: 0.4390 data_time: 0.0065 memory: 4049 loss: 70.9129 loss_cls: 0.1025 loss_bbox: 0.2569 loss_centerness: 0.6374 loss_bezier: 69.0977 loss_ce: 0.8184
2024/01/26 21:32:40 - mmengine - INFO - Epoch(train) [1][185/334] lr: 1.8500e-05 eta: 23:18:27 time: 0.4233 data_time: 0.0063 memory: 5381 loss: 72.7930 loss_cls: 0.2525 loss_bbox: 0.3509 loss_centerness: 0.7012 loss_bezier: 70.7547 loss_ce: 0.7337
2024/01/26 21:32:43 - mmengine - INFO - Epoch(train) [1][190/334] lr: 1.9000e-05 eta: 23:13:59 time: 0.4442 data_time: 0.0063 memory: 3379 loss: 80.5646 loss_cls: 0.3565 loss_bbox: 0.4810 loss_centerness: 0.8948 loss_bezier: 78.0133 loss_ce: 0.8189
2024/01/26 21:32:45 - mmengine - INFO - Epoch(train) [1][195/334] lr: 1.9500e-05 eta: 23:10:12 time: 0.4457 data_time: 0.0064 memory: 3754 loss: 61.7456 loss_cls: 0.1910 loss_bbox: 0.3464 loss_centerness: 0.6922 loss_bezier: 59.6708 loss_ce: 0.8451
2024/01/26 21:32:47 - mmengine - INFO - Epoch(train) [1][200/334] lr: 2.0000e-05 eta: 23:06:41 time: 0.4495 data_time: 0.0067 memory: 4325 loss: 45.2591 loss_cls: 0.1740 loss_bbox: 0.3140 loss_centerness: 0.6057 loss_bezier: 43.4050 loss_ce: 0.7604
2024/01/26 21:32:49 - mmengine - INFO - Epoch(train) [1][205/334] lr: 2.0500e-05 eta: 23:03:17 time: 0.4497 data_time: 0.0066 memory: 4762 loss: 54.1472 loss_cls: 0.1751 loss_bbox: 0.3158 loss_centerness: 0.6172 loss_bezier: 52.0880 loss_ce: 0.9511
2024/01/26 21:32:52 - mmengine - INFO - Epoch(train) [1][210/334] lr: 2.1000e-05 eta: 23:03:06 time: 0.4723 data_time: 0.0064 memory: 4864 loss: 40.8218 loss_cls: 0.1554 loss_bbox: 0.2707 loss_centerness: 0.5914 loss_bezier: 38.7319 loss_ce: 1.0725
2024/01/26 21:32:54 - mmengine - INFO - Epoch(train) [1][215/334] lr: 2.1500e-05 eta: 22:56:50 time: 0.4483 data_time: 0.0065 memory: 4804 loss: 33.7001 loss_cls: 0.2039 loss_bbox: 0.3154 loss_centerness: 0.6548 loss_bezier: 31.6031 loss_ce: 0.9229
2024/01/26 21:32:56 - mmengine - INFO - Epoch(train) [1][220/334] lr: 2.2000e-05 eta: 22:53:56 time: 0.4258 data_time: 0.0068 memory: 5783 loss: 22.5069 loss_cls: 0.1096 loss_bbox: 0.1927 loss_centerness: 0.3892 loss_bezier: 20.8786 loss_ce: 0.9368
2024/01/26 21:32:58 - mmengine - INFO - Epoch(train) [1][225/334] lr: 2.2500e-05 eta: 22:49:02 time: 0.4328 data_time: 0.0067 memory: 4311 loss: 45.1425 loss_cls: 0.1181 loss_bbox: 0.2430 loss_centerness: 0.4610 loss_bezier: 43.3964 loss_ce: 0.9239
2024/01/26 21:33:00 - mmengine - INFO - Epoch(train) [1][230/334] lr: 2.3000e-05 eta: 22:45:12 time: 0.4227 data_time: 0.0066 memory: 3918 loss: 76.6669 loss_cls: 0.1768 loss_bbox: 0.3636 loss_centerness: 0.6712 loss_bezier: 74.5914 loss_ce: 0.8639
2024/01/26 21:33:03 - mmengine - INFO - Epoch(train) [1][235/334] lr: 2.3500e-05 eta: 22:44:31 time: 0.4550 data_time: 0.0063 memory: 5371 loss: 57.4753 loss_cls: 0.1530 loss_bbox: 0.3649 loss_centerness: 0.6644 loss_bezier: 55.4106 loss_ce: 0.8823
2024/01/26 21:33:05 - mmengine - INFO - Epoch(train) [1][240/334] lr: 2.4000e-05 eta: 22:40:48 time: 0.4539 data_time: 0.0060 memory: 3783 loss: 34.0878 loss_cls: 0.1214 loss_bbox: 0.3369 loss_centerness: 0.5839 loss_bezier: 32.1566 loss_ce: 0.8890
2024/01/26 21:33:07 - mmengine - INFO - Epoch(train) [1][245/334] lr: 2.4500e-05 eta: 22:39:27 time: 0.4470 data_time: 0.0063 memory: 5814 loss: 23.9602 loss_cls: 0.1034 loss_bbox: 0.2744 loss_centerness: 0.4430 loss_bezier: 22.3092 loss_ce: 0.8301
2024/01/26 21:33:09 - mmengine - INFO - Epoch(train) [1][250/334] lr: 2.5000e-05 eta: 22:36:04 time: 0.4478 data_time: 0.0064 memory: 5355 loss: 23.2497 loss_cls: 0.1141 loss_bbox: 0.2547 loss_centerness: 0.4069 loss_bezier: 21.6908 loss_ce: 0.7831
2024/01/26 21:33:11 - mmengine - INFO - Epoch(train) [1][255/334] lr: 2.5500e-05 eta: 22:31:45 time: 0.4193 data_time: 0.0069 memory: 3856 loss: 44.2494 loss_cls: 0.2082 loss_bbox: 0.3757 loss_centerness: 0.6872 loss_bezier: 42.1251 loss_ce: 0.8533
2024/01/26 21:33:14 - mmengine - INFO - Epoch(train) [1][260/334] lr: 2.6000e-05 eta: 22:30:32 time: 0.4369 data_time: 0.0076 memory: 5130 loss: 49.1669 loss_cls: 0.2671 loss_bbox: 0.4137 loss_centerness: 0.7838 loss_bezier: 46.8786 loss_ce: 0.8236
2024/01/26 21:33:16 - mmengine - INFO - Epoch(train) [1][265/334] lr: 2.6500e-05 eta: 22:30:29 time: 0.4751 data_time: 0.0069 memory: 5184 loss: 30.3049 loss_cls: 0.2080 loss_bbox: 0.3534 loss_centerness: 0.6371 loss_bezier: 28.3275 loss_ce: 0.7789
2024/01/26 21:33:18 - mmengine - INFO - Epoch(train) [1][270/334] lr: 2.7000e-05 eta: 22:29:32 time: 0.4771 data_time: 0.0067 memory: 5307 loss: 19.7767 loss_cls: 0.1084 loss_bbox: 0.2727 loss_centerness: 0.5060 loss_bezier: 18.0340 loss_ce: 0.8555
2024/01/26 21:33:21 - mmengine - INFO - Epoch(train) [1][275/334] lr: 2.7500e-05 eta: 22:27:54 time: 0.4612 data_time: 0.0067 memory: 4939 loss: 13.6433 loss_cls: 0.0510 loss_bbox: 0.1664 loss_centerness: 0.3280 loss_bezier: 12.2820 loss_ce: 0.8160
2024/01/26 21:33:23 - mmengine - INFO - Epoch(train) [1][280/334] lr: 2.8000e-05 eta: 22:23:49 time: 0.4288 data_time: 0.0062 memory: 3783 loss: 16.0401 loss_cls: 0.0772 loss_bbox: 0.1605 loss_centerness: 0.3470 loss_bezier: 14.8560 loss_ce: 0.5993
2024/01/26 21:33:25 - mmengine - INFO - Epoch(train) [1][285/334] lr: 2.8500e-05 eta: 22:23:16 time: 0.4384 data_time: 0.0064 memory: 5999 loss: 32.8569 loss_cls: 0.0873 loss_bbox: 0.1906 loss_centerness: 0.4519 loss_bezier: 31.3577 loss_ce: 0.7694
2024/01/26 21:33:27 - mmengine - INFO - Epoch(train) [1][290/334] lr: 2.9000e-05 eta: 22:22:47 time: 0.4738 data_time: 0.0066 memory: 5487 loss: 37.6147 loss_cls: 0.0818 loss_bbox: 0.1979 loss_centerness: 0.5309 loss_bezier: 35.6744 loss_ce: 1.1297
2024/01/26 21:33:30 - mmengine - INFO - Epoch(train) [1][295/334] lr: 2.9500e-05 eta: 22:20:04 time: 0.4502 data_time: 0.0066 memory: 4780 loss: 30.1234 loss_cls: 0.1552 loss_bbox: 0.2331 loss_centerness: 0.6619 loss_bezier: 27.9297 loss_ce: 1.1435
2024/01/26 21:33:32 - mmengine - INFO - Epoch(train) [1][300/334] lr: 3.0000e-05 eta: 22:18:27 time: 0.4373 data_time: 0.0064 memory: 4334 loss: 37.2476 loss_cls: 0.1910 loss_bbox: 0.2894 loss_centerness: 0.7176 loss_bezier: 35.1753 loss_ce: 0.8742
2024/01/26 21:33:34 - mmengine - INFO - Epoch(train) [1][305/334] lr: 3.0500e-05 eta: 22:19:53 time: 0.4813 data_time: 0.0064 memory: 6204 loss: 41.1584 loss_cls: 0.1547 loss_bbox: 0.3165 loss_centerness: 0.7096 loss_bezier: 39.2043 loss_ce: 0.7732
2024/01/26 21:33:37 - mmengine - INFO - Epoch(train) [1][310/334] lr: 3.1000e-05 eta: 22:19:26 time: 0.4938 data_time: 0.0063 memory: 3673 loss: 41.0209 loss_cls: 0.1767 loss_bbox: 0.3328 loss_centerness: 0.7705 loss_bezier: 38.8929 loss_ce: 0.8480
2024/01/26 21:33:39 - mmengine - INFO - Epoch(train) [1][315/334] lr: 3.1500e-05 eta: 22:16:40 time: 0.4468 data_time: 0.0058 memory: 3902 loss: 34.0136 loss_cls: 0.1335 loss_bbox: 0.2599 loss_centerness: 0.6546 loss_bezier: 32.1330 loss_ce: 0.8325
2024/01/26 21:33:41 - mmengine - INFO - Epoch(train) [1][320/334] lr: 3.2000e-05 eta: 22:16:03 time: 0.4441 data_time: 0.0056 memory: 4187 loss: 23.8213 loss_cls: 0.0861 loss_bbox: 0.2173 loss_centerness: 0.5506 loss_bezier: 22.0932 loss_ce: 0.8742
2024/01/26 21:33:44 - mmengine - INFO - Epoch(train) [1][325/334] lr: 3.2500e-05 eta: 22:17:38 time: 0.4934 data_time: 0.0061 memory: 6338 loss: 25.9496 loss_cls: 0.0972 loss_bbox: 0.2045 loss_centerness: 0.5302 loss_bezier: 24.3024 loss_ce: 0.8154
2024/01/26 21:33:46 - mmengine - INFO - Epoch(train) [1][330/334] lr: 3.3000e-05 eta: 22:14:48 time: 0.4670 data_time: 0.0064 memory: 3820 loss: 34.3054 loss_cls: 0.1113 loss_bbox: 0.1850 loss_centerness: 0.5182 loss_bezier: 32.7535 loss_ce: 0.7375
2024/01/26 21:33:48 - mmengine - INFO - Exp name: abcnet_resnet50_fpn_500e_icdar2015_20240126_213051
2024/01/26 21:33:53 - mmengine - INFO - Epoch(val) [1][ 5/167] eta: 0:02:48 time: 1.0424 data_time: 0.7204 memory: 4886
2024/01/26 21:33:54 - mmengine - INFO - Epoch(val) [1][ 10/167] eta: 0:01:46 time: 0.6778 data_time: 0.3635 memory: 2951
2024/01/26 21:33:56 - mmengine - INFO - Epoch(val) [1][ 15/167] eta: 0:01:25 time: 0.3195 data_time: 0.0064 memory: 2951
2024/01/26 21:33:58 - mmengine - INFO - Epoch(val) [1][ 20/167] eta: 0:01:13 time: 0.3238 data_time: 0.0075 memory: 2951
2024/01/26 21:33:59 - mmengine - INFO - Epoch(val) [1][ 25/167] eta: 0:01:05 time: 0.3169 data_time: 0.0073 memory: 2951
2024/01/26 21:34:01 - mmengine - INFO - Epoch(val) [1][ 30/167] eta: 0:01:00 time: 0.3127 data_time: 0.0062 memory: 2951
2024/01/26 21:34:02 - mmengine - INFO - Epoch(val) [1][ 35/167] eta: 0:00:56 time: 0.3274 data_time: 0.0063 memory: 2951
2024/01/26 21:34:04 - mmengine - INFO - Epoch(val) [1][ 40/167] eta: 0:00:52 time: 0.3433 data_time: 0.0061 memory: 2951
2024/01/26 21:34:06 - mmengine - INFO - Epoch(val) [1][ 45/167] eta: 0:00:49 time: 0.3283 data_time: 0.0060 memory: 2951
2024/01/26 21:34:07 - mmengine - INFO - Epoch(val) [1][ 50/167] eta: 0:00:46 time: 0.3096 data_time: 0.0057 memory: 2951
2024/01/26 21:34:09 - mmengine - INFO - Epoch(val) [1][ 55/167] eta: 0:00:43 time: 0.3257 data_time: 0.0058 memory: 2951
2024/01/26 21:34:11 - mmengine - INFO - Epoch(val) [1][ 60/167] eta: 0:00:41 time: 0.3458 data_time: 0.0067 memory: 2951
2024/01/26 21:34:13 - mmengine - INFO - Epoch(val) [1][ 65/167] eta: 0:00:39 time: 0.3626 data_time: 0.0075 memory: 2951
2024/01/26 21:34:14 - mmengine - INFO - Epoch(val) [1][ 70/167] eta: 0:00:37 time: 0.3600 data_time: 0.0101 memory: 2951
2024/01/26 21:34:16 - mmengine - INFO - Epoch(val) [1][ 75/167] eta: 0:00:34 time: 0.3343 data_time: 0.0102 memory: 2951
2024/01/26 21:34:18 - mmengine - INFO - Epoch(val) [1][ 80/167] eta: 0:00:32 time: 0.3578 data_time: 0.0078 memory: 2951
2024/01/26 21:34:19 - mmengine - INFO - Epoch(val) [1][ 85/167] eta: 0:00:30 time: 0.3506 data_time: 0.0083 memory: 2951
2024/01/26 21:34:21 - mmengine - INFO - Epoch(val) [1][ 90/167] eta: 0:00:28 time: 0.3339 data_time: 0.0073 memory: 2951
2024/01/26 21:34:23 - mmengine - INFO - Epoch(val) [1][ 95/167] eta: 0:00:26 time: 0.3535 data_time: 0.0073 memory: 2951
2024/01/26 21:34:26 - mmengine - INFO - Epoch(val) [1][100/167] eta: 0:00:25 time: 0.4713 data_time: 0.0078 memory: 2951
2024/01/26 21:34:27 - mmengine - INFO - Epoch(val) [1][105/167] eta: 0:00:23 time: 0.4502 data_time: 0.0065 memory: 2951
2024/01/26 21:34:29 - mmengine - INFO - Epoch(val) [1][110/167] eta: 0:00:21 time: 0.3202 data_time: 0.0063 memory: 2951
2024/01/26 21:34:31 - mmengine - INFO - Epoch(val) [1][115/167] eta: 0:00:19 time: 0.3971 data_time: 0.0065 memory: 2951
2024/01/26 21:34:33 - mmengine - INFO - Epoch(val) [1][120/167] eta: 0:00:17 time: 0.3935 data_time: 0.0066 memory: 2951
2024/01/26 21:34:35 - mmengine - INFO - Epoch(val) [1][125/167] eta: 0:00:15 time: 0.3740 data_time: 0.0062 memory: 2951
2024/01/26 21:34:38 - mmengine - INFO - Epoch(val) [1][130/167] eta: 0:00:14 time: 0.5135 data_time: 0.0061 memory: 2951
2024/01/26 21:34:40 - mmengine - INFO - Epoch(val) [1][135/167] eta: 0:00:12 time: 0.4686 data_time: 0.0062 memory: 2951
2024/01/26 21:34:42 - mmengine - INFO - Epoch(val) [1][140/167] eta: 0:00:10 time: 0.3499 data_time: 0.0062 memory: 2951
2024/01/26 21:34:43 - mmengine - INFO - Epoch(val) [1][145/167] eta: 0:00:08 time: 0.3593 data_time: 0.0062 memory: 2951
2024/01/26 21:34:45 - mmengine - INFO - Epoch(val) [1][150/167] eta: 0:00:06 time: 0.3439 data_time: 0.0061 memory: 2951
2024/01/26 21:34:47 - mmengine - INFO - Epoch(val) [1][155/167] eta: 0:00:04 time: 0.3364 data_time: 0.0062 memory: 2951
2024/01/26 21:34:49 - mmengine - INFO - Epoch(val) [1][160/167] eta: 0:00:02 time: 0.3591 data_time: 0.0064 memory: 2951
2024/01/26 21:34:50 - mmengine - INFO - Epoch(val) [1][165/167] eta: 0:00:00 time: 0.3409 data_time: 0.0064 memory: 2951
2024/01/26 21:34:54 - mmengine - INFO - Evaluating hmean-iou...
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.30, recall: 0.0067, precision: 0.0060, hmean: 0.0064
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.40, recall: 0.0067, precision: 0.0061, hmean: 0.0064
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.50, recall: 0.0067, precision: 0.0067, hmean: 0.0067
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.60, recall: 0.0067, precision: 0.0081, hmean: 0.0074
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.70, recall: 0.0063, precision: 0.0105, hmean: 0.0079
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.80, recall: 0.0019, precision: 0.0081, hmean: 0.0031
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.90, recall: 0.0000, precision: 0.0000, hmean: 0.0000
2024/01/26 21:34:54 - mmengine - INFO - Evaluating hmean-iou... 2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.30, recall: 0.5474, precision: 0.4911, hmean: 0.5178
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.40, recall: 0.5465, precision: 0.4969, hmean: 0.5205
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.50, recall: 0.5407, precision: 0.5348, hmean: 0.5377
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.60, recall: 0.5022, precision: 0.6036, hmean: 0.5482
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.70, recall: 0.4054, precision: 0.6818, hmean: 0.5085
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.80, recall: 0.1858, precision: 0.7830, hmean: 0.3004
2024/01/26 21:34:54 - mmengine - INFO - prediction score threshold: 0.90, recall: 0.0034, precision: 0.8750, hmean: 0.0067
2024/01/26 21:34:54 - mmengine - INFO - Epoch(val) [1][167/167] e2e_icdar/precision: 0.0105 e2e_icdar/recall: 0.0063 e2e_icdar/hmean: 0.0079 icdar/precision: 0.6036 icdar/recall: 0.5022 icdar/hmean: 0.5482 data_time: 0.0282 time: 0.3787 2024/01/26 21:34:56 - mmengine - INFO - Epoch(train) [2][ 5/334] lr: 3.3900e-05 eta: 22:14:31 time: 0.4678 data_time: 0.0654 memory: 4174 loss: 35.6368 loss_cls: 0.0931 loss_bbox: 0.2146 loss_centerness: 0.5189 loss_bezier: 33.7419 loss_ce: 1.0683 2024/01/26 21:34:59 - mmengine - INFO - Epoch(train) [2][ 10/334] lr: 3.4400e-05 eta: 22:13:00 time: 0.4855 data_time: 0.0659 memory: 5482 loss: 27.5721 loss_cls: 0.1140 loss_bbox: 0.2310 loss_centerness: 0.5587 loss_bezier: 25.7776 loss_ce: 0.8909 2024/01/26 21:35:01 - mmengine - INFO - Epoch(train) [2][ 15/334] lr: 3.4900e-05 eta: 22:10:46 time: 0.4344 data_time: 0.0058 memory: 5710 loss: 27.9195 loss_cls: 0.0990 loss_bbox: 0.1982 loss_centerness: 0.5401 loss_bezier: 26.4141 loss_ce: 0.6682 2024/01/26 21:35:03 - mmengine - INFO - Epoch(train) [2][ 20/334] lr: 3.5400e-05 eta: 22:09:58 time: 0.4422 data_time: 0.0049 memory: 4345 loss: 33.8623 loss_cls: 0.0922 loss_bbox: 0.2193 loss_centerness: 0.5632 loss_bezier: 32.1609 loss_ce: 0.8268 2024/01/26 21:35:06 - mmengine - INFO - Epoch(train) [2][ 25/334] lr: 3.5900e-05 eta: 22:10:49 time: 0.4808 data_time: 0.0061 memory: 5660 loss: 33.4696 loss_cls: 0.0928 loss_bbox: 0.2326 loss_centerness: 0.5116 loss_bezier: 31.6431 loss_ce: 0.9895 2024/01/26 21:35:08 - mmengine - INFO - Epoch(train) [2][ 30/334] lr: 3.6400e-05 eta: 22:09:57 time: 0.4798 data_time: 0.0066 memory: 6285 loss: 35.2096 loss_cls: 0.1010 loss_bbox: 0.2833 loss_centerness: 0.6660 loss_bezier: 33.2007 loss_ce: 0.9586 2024/01/26 21:35:10 - mmengine - INFO - Epoch(train) [2][ 35/334] lr: 3.6900e-05 eta: 22:08:48 time: 0.4535 data_time: 0.0066 memory: 5147 loss: 34.7616 loss_cls: 0.1284 loss_bbox: 0.2794 loss_centerness: 0.7090 loss_bezier: 32.8919 loss_ce: 0.7529 2024/01/26 21:35:13 - mmengine - INFO - Epoch(train) [2][ 40/334] lr: 3.7400e-05 eta: 22:08:49 time: 0.4647 data_time: 0.0067 memory: 5601 loss: 33.3299 loss_cls: 0.1407 loss_bbox: 0.2361 loss_centerness: 0.5852 loss_bezier: 31.6776 loss_ce: 0.6904 2024/01/26 21:35:15 - mmengine - INFO - Epoch(train) [2][ 45/334] lr: 3.7900e-05 eta: 22:06:41 time: 0.4506 data_time: 0.0067 memory: 5799 loss: 30.3966 loss_cls: 0.1292 loss_bbox: 0.2487 loss_centerness: 0.5827 loss_bezier: 28.6536 loss_ce: 0.7824 2024/01/26 21:35:17 - mmengine - INFO - Epoch(train) [2][ 50/334] lr: 3.8400e-05 eta: 22:05:47 time: 0.4376 data_time: 0.0070 memory: 5338 loss: 35.8284 loss_cls: 0.1651 loss_bbox: 0.2775 loss_centerness: 0.6990 loss_bezier: 33.8087 loss_ce: 0.8780 2024/01/26 21:35:19 - mmengine - INFO - Epoch(train) [2][ 55/334] lr: 3.8900e-05 eta: 22:05:49 time: 0.4666 data_time: 0.0074 memory: 4823 loss: 42.7366 loss_cls: 0.1759 loss_bbox: 0.2823 loss_centerness: 0.7456 loss_bezier: 40.7314 loss_ce: 0.8013 2024/01/26 21:35:22 - mmengine - INFO - Epoch(train) [2][ 60/334] lr: 3.9400e-05 eta: 22:05:05 time: 0.4686 data_time: 0.0073 memory: 5248 loss: 28.2224 loss_cls: 0.1166 loss_bbox: 0.2143 loss_centerness: 0.5628 loss_bezier: 26.5714 loss_ce: 0.7573 2024/01/26 21:35:24 - mmengine - INFO - Epoch(train) [2][ 65/334] lr: 3.9900e-05 eta: 22:04:22 time: 0.4577 data_time: 0.0075 memory: 4823 loss: 19.1592 loss_cls: 0.0679 loss_bbox: 0.1445 loss_centerness: 0.3875 loss_bezier: 17.8066 loss_ce: 0.7527 2024/01/26 21:35:26 - mmengine - INFO - Epoch(train) [2][ 70/334] lr: 4.0400e-05 eta: 22:02:39 time: 0.4430 data_time: 0.0077 memory: 4830 loss: 25.3838 loss_cls: 0.0726 loss_bbox: 0.1312 loss_centerness: 0.3274 loss_bezier: 24.1362 loss_ce: 0.7163 2024/01/26 21:35:28 - mmengine - INFO - Epoch(train) [2][ 75/334] lr: 4.0900e-05 eta: 22:01:38 time: 0.4380 data_time: 0.0070 memory: 4407 loss: 35.7047 loss_cls: 0.1318 loss_bbox: 0.2123 loss_centerness: 0.5868 loss_bezier: 34.0130 loss_ce: 0.7607 2024/01/26 21:35:31 - mmengine - INFO - Epoch(train) [2][ 80/334] lr: 4.1400e-05 eta: 22:00:43 time: 0.4488 data_time: 0.0063 memory: 4977 loss: 47.8859 loss_cls: 0.1534 loss_bbox: 0.3229 loss_centerness: 0.8296 loss_bezier: 45.7219 loss_ce: 0.8581 2024/01/26 21:35:33 - mmengine - INFO - Epoch(train) [2][ 85/334] lr: 4.1900e-05 eta: 21:57:58 time: 0.4217 data_time: 0.0064 memory: 4084 loss: 47.8201 loss_cls: 0.1219 loss_bbox: 0.2866 loss_centerness: 0.7140 loss_bezier: 45.8143 loss_ce: 0.8835 2024/01/26 21:35:35 - mmengine - INFO - Epoch(train) [2][ 90/334] lr: 4.2400e-05 eta: 21:56:30 time: 0.4124 data_time: 0.0064 memory: 3965 loss: 29.0756 loss_cls: 0.0906 loss_bbox: 0.1696 loss_centerness: 0.4885 loss_bezier: 27.5553 loss_ce: 0.7716 2024/01/26 21:35:37 - mmengine - INFO - Epoch(train) [2][ 95/334] lr: 4.2900e-05 eta: 21:56:03 time: 0.4464 data_time: 0.0065 memory: 4231 loss: 17.2184 loss_cls: 0.0841 loss_bbox: 0.1435 loss_centerness: 0.4109 loss_bezier: 15.9003 loss_ce: 0.6796 2024/01/26 21:35:39 - mmengine - INFO - Epoch(train) [2][100/334] lr: 4.3400e-05 eta: 21:54:59 time: 0.4518 data_time: 0.0067 memory: 5242 loss: 36.7983 loss_cls: 0.1439 loss_bbox: 0.2672 loss_centerness: 0.6804 loss_bezier: 35.0710 loss_ce: 0.6359 2024/01/26 21:35:41 - mmengine - INFO - Epoch(train) [2][105/334] lr: 4.3900e-05 eta: 21:52:47 time: 0.4237 data_time: 0.0067 memory: 5511 loss: 37.9046 loss_cls: 0.1701 loss_bbox: 0.2675 loss_centerness: 0.6910 loss_bezier: 36.1703 loss_ce: 0.6057 2024/01/26 21:35:44 - mmengine - INFO - Epoch(train) [2][110/334] lr: 4.4400e-05 eta: 21:52:22 time: 0.4330 data_time: 0.0066 memory: 5355 loss: 27.2889 loss_cls: 0.1184 loss_bbox: 0.1587 loss_centerness: 0.4666 loss_bezier: 25.7921 loss_ce: 0.7531 2024/01/26 21:35:46 - mmengine - INFO - Epoch(train) [2][115/334] lr: 4.4900e-05 eta: 21:52:06 time: 0.4630 data_time: 0.0066 memory: 5332 loss: 30.4800 loss_cls: 0.0987 loss_bbox: 0.1587 loss_centerness: 0.4475 loss_bezier: 29.0706 loss_ce: 0.7045 2024/01/26 21:35:48 - mmengine - INFO - Epoch(train) [2][120/334] lr: 4.5400e-05 eta: 21:50:20 time: 0.4408 data_time: 0.0067 memory: 4908 loss: 32.7542 loss_cls: 0.1292 loss_bbox: 0.2200 loss_centerness: 0.5994 loss_bezier: 31.1254 loss_ce: 0.6802 2024/01/26 21:35:50 - mmengine - INFO - Epoch(train) [2][125/334] lr: 4.5900e-05 eta: 21:48:02 time: 0.4067 data_time: 0.0063 memory: 4506 loss: 33.7011 loss_cls: 0.1472 loss_bbox: 0.2650 loss_centerness: 0.7328 loss_bezier: 31.7315 loss_ce: 0.8246 2024/01/26 21:35:52 - mmengine - INFO - Epoch(train) [2][130/334] lr: 4.6400e-05 eta: 21:46:45 time: 0.4134 data_time: 0.0057 memory: 4273 loss: 27.9848 loss_cls: 0.1280 loss_bbox: 0.2333 loss_centerness: 0.6392 loss_bezier: 26.2176 loss_ce: 0.7667 2024/01/26 21:35:54 - mmengine - INFO - Epoch(train) [2][135/334] lr: 4.6900e-05 eta: 21:46:50 time: 0.4524 data_time: 0.0058 memory: 4725 loss: 37.6867 loss_cls: 0.1194 loss_bbox: 0.2396 loss_centerness: 0.6958 loss_bezier: 35.7702 loss_ce: 0.8617 2024/01/26 21:35:56 - mmengine - INFO - Epoch(train) [2][140/334] lr: 4.7400e-05 eta: 21:44:45 time: 0.4379 data_time: 0.0065 memory: 4028 loss: 39.4029 loss_cls: 0.1266 loss_bbox: 0.2708 loss_centerness: 0.6999 loss_bezier: 37.2735 loss_ce: 1.0320 2024/01/26 21:35:59 - mmengine - INFO - Epoch(train) [2][145/334] lr: 4.7900e-05 eta: 21:44:55 time: 0.4391 data_time: 0.0064 memory: 4546 loss: 36.7799 loss_cls: 0.1267 loss_bbox: 0.2324 loss_centerness: 0.5585 loss_bezier: 34.8647 loss_ce: 0.9975 2024/01/26 21:36:01 - mmengine - INFO - Epoch(train) [2][150/334] lr: 4.8400e-05 eta: 21:44:12 time: 0.4620 data_time: 0.0064 memory: 4917 loss: 33.3326 loss_cls: 0.1060 loss_bbox: 0.1909 loss_centerness: 0.5776 loss_bezier: 31.6108 loss_ce: 0.8472 2024/01/26 21:36:04 - mmengine - INFO - Epoch(train) [2][155/334] lr: 4.8900e-05 eta: 21:44:59 time: 0.4726 data_time: 0.0075 memory: 5477 loss: 32.4475 loss_cls: 0.1121 loss_bbox: 0.2282 loss_centerness: 0.6992 loss_bezier: 30.6551 loss_ce: 0.7530 2024/01/26 21:36:06 - mmengine - INFO - Epoch(train) [2][160/334] lr: 4.9400e-05 eta: 21:44:16 time: 0.4727 data_time: 0.0078 memory: 4626 loss: 40.1348 loss_cls: 0.1363 loss_bbox: 0.2635 loss_centerness: 0.7966 loss_bezier: 38.2085 loss_ce: 0.7298 2024/01/26 21:36:08 - mmengine - INFO - Epoch(train) [2][165/334] lr: 4.9900e-05 eta: 21:44:37 time: 0.4651 data_time: 0.0072 memory: 4371 loss: 31.1086 loss_cls: 0.1214 loss_bbox: 0.2351 loss_centerness: 0.6848 loss_bezier: 29.2470 loss_ce: 0.8203 2024/01/26 21:36:10 - mmengine - INFO - Epoch(train) [2][170/334] lr: 5.0400e-05 eta: 21:43:47 time: 0.4626 data_time: 0.0067 memory: 4691 loss: 31.9128 loss_cls: 0.1310 loss_bbox: 0.2744 loss_centerness: 0.7046 loss_bezier: 29.9311 loss_ce: 0.8717 2024/01/26 21:36:12 - mmengine - INFO - Epoch(train) [2][175/334] lr: 5.0900e-05 eta: 21:42:03 time: 0.4247 data_time: 0.0065 memory: 3357 loss: 37.9630 loss_cls: 0.1370 loss_bbox: 0.3028 loss_centerness: 0.7711 loss_bezier: 35.9456 loss_ce: 0.8065 2024/01/26 21:36:15 - mmengine - INFO - Epoch(train) [2][180/334] lr: 5.1400e-05 eta: 21:42:13 time: 0.4422 data_time: 0.0066 memory: 5064 loss: 45.1929 loss_cls: 0.1673 loss_bbox: 0.2822 loss_centerness: 0.7925 loss_bezier: 43.1993 loss_ce: 0.7515 2024/01/26 21:36:17 - mmengine - INFO - Epoch(train) [2][185/334] lr: 5.1900e-05 eta: 21:43:22 time: 0.4954 data_time: 0.0068 memory: 5318 loss: 49.8830 loss_cls: 0.1768 loss_bbox: 0.3352 loss_centerness: 0.9359 loss_bezier: 47.7258 loss_ce: 0.7094 2024/01/26 21:36:20 - mmengine - INFO - Epoch(train) [2][190/334] lr: 5.2400e-05 eta: 21:43:42 time: 0.4989 data_time: 0.0067 memory: 4939 loss: 33.1001 loss_cls: 0.1125 loss_bbox: 0.2832 loss_centerness: 0.7342 loss_bezier: 31.2520 loss_ce: 0.7182
I don’t know why, this time the results are better, but loss_besize still dominates the gradient.
Bin-ze
commented
5 months ago loss_bezier will dominate the optimization direction, causing model training to fail.
pd162
commented
5 months ago I have noticed that you loaded the finetuned model to re-train on icdar2015. It seems strange because the result should keep with the official result. I hope you leave your contact information to talk about this problem in more detail.
Bin-ze
commented
5 months ago You only need to run the baseline yourself to find out. I think it is impossible for you to get consistent results with the official results.
You can contact me by email and add contact information. My email is on my github homepage
MikasaLee
commented
2 months ago Hi,I have encountered the same problem as you. May I ask how you ultimately handled it @Bin-ze @pd162
MikasaLee
commented
2 months ago Also, I noticed that the learning rate during training was initially 5e-6 and kept increasing. Why is this happening?
Prerequisite
Task
I'm using the official example scripts/configs for the officially supported tasks/models/datasets.
Branch
main branch https://github.com/open-mmlab/mmocr
Environment
There are no problems with the environment and the code runs normally.
Reproduces the problem - code sample
I used the icdar2015 data set, but I couldn't get convergence consistent with the official one：
Official training log： https://download.openmmlab.com/mmocr/textspotting/abcnet/abcnet_resnet50_fpn_500e_icdar2015/20221210_170401.log
Reproduces the problem - command or script
mim train mmocr config/abcnet/abcnet_resnet50_fpn_500e_icdar2015.py --work-dir work_dirs/ --launcher pytorch --gpus 4
Reproduces the problem - error message
No error message
Additional information
I want to know the documentation about E2E-None-Hmean, det-Hmean, but I didn't find the relevant information.
If E2E-None-Hmean is 0, what are the possible problems?
I don't understand:
dictionary = dict( type='Dictionary', dict_file='{{ fileDirname }}/../../dicts/abcnet.txt', with_start=True, with_end=True, same_start_end=True, with_padding=True, with_unknown=True)
The meaning of configuration, but when with_start=True, with_end=True, same_start_end=True, with_padding=True, When with_unknown=True, num_classes will be +1